COMPREHENSIVE ANALYTICAL CHEMISTRY
ELSEVIER B.V. Sara Burgerhartstraat 25 P.O. Box 211, 1000 AE Amsterdam, The Netherlands
ELSEVIER Inc. 525 B Street, Suite 1900 San Diego, CA 92101-4495 USA
ELSEVIER Ltd The Boulevard, Langford Lane Kidlington, Oxford OX5 1GB UK
ELSEVIER Ltd 84 Theobalds Road London WC1X 8RR UK
q 2004 Elsevier B.V. All rights reserved. This work is protected under copyright by Elsevier B.V., and the following terms and conditions apply to its use: Photocopying Single photocopies of single chapters may be made for personal use as allowed by national copyright laws. Permission of the Publisher and payment of a fee is required for all other photocopying, including multiple or systematic copying, copying for advertising or promotional purposes, resale, and all forms of document delivery. Special rates are available for educational institutions that wish to make photocopies for non-profit educational classroom use. Permissions may be sought directly from Elsevier’s Rights Department in Oxford, UK; phone: (+44) 1865 843830, fax: (+44) 1865 853333, e-mail:
[email protected]. Requests may also be completed on-line via the Elsevier homepage (http://www. elsevier.com/locate/permissions). In the USA, users may clear permissions and make payments through the Copyright Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923, USA; phone: (+1) (978) 7508400, fax: (+1) (978) 7504744, and in the UK through the Copyright Licensing Agency Rapid Clearance Service (CLARCS), 90 Tottenham Court Road, London W1P 0LP, UK; phone: (+44) 20 7631 5555; fax: (+44) 20 7631 5500. Other countries may have a local reprographic rights agency for payments. Derivative Works Tables of contents may be reproduced for internal circulation, but permission of the Publisher is required for external resale or distribution of such material. Permission of the Publisher is required for all other derivative works, including compilations and translations. Electronic Storage or Usage Permission of the Publisher is required to store or use electronically any material contained in this work, including any chapter or part of a chapter. Except as outlined above, no part of this work may be reproduced, stored in a retrieval system or transmitted in any form or by any means, electronic, mechanical, photocopying, recording or otherwise, without prior written permission of the Publisher. Address permissions requests to: Elsevier’s Rights Department, at the fax and e-mail addresses noted above. Notice No responsibility is assumed by the Publisher for any injury and/or damage to persons or property as a matter of products liability, negligence or otherwise, or from any use or operation of any methods, products, instructions or ideas contained in the material herein. Because of rapid advances in the medical sciences, in particular, independent verification of diagnoses and drug dosages should be made.
First edition 2004 Library of Congress Cataloging in Publication Data A catalog record is available from the Library of Congress. British Library Cataloguing in Publication Data A catalogue record is available from the British Library. ISBN: 0-444-50738-8 ISSN: 0166-526X 1 The paper used in this publication meets the requirements of ANSI/NISO Z39.48-1992 (Permanence of Paper). W Printed in The Netherlands.
COMPREHENSIVE ANALYTICAL CHEMISTRY ADVISORY BOARD
Professor A.M. Bond Monash University, Clayton, Victoria, Australia Dr T.W. Collette US Environmental Protection Agency, Athens, GA, U.S.A. Professor M. Grasserbauer Director of the Environment Institute, European Commission, Joint Research Centre, Ispra, Italy Professor M.-C. Hennion Ecole Supe´rieure de Physique et de Chimie Industrielles, Paris, France Professor G. M. Hieftje Indiana University, Bloomington, IN, U.S.A. Professor G. Marko-Varga AstraZeneca, Lund, Sweden Professor D.L. Massart Vrije Universiteit, Brussels, Belgium Professor M.E. Meyerhoff University of Michigan, Ann Arbor, MI, U.S.A.
Wilson & Wilson’s COMPREHENSIVE ANALYTICAL CHEMISTRY
Edited by ´ D. BARCELO Research Professor Department of Environmental Chemistry IIQAB-CSIC Jordi Girona 18-26 08034 Barcelona Spain
Wilson & Wilson’s COMPREHENSIVE ANALYTICAL CHEMISTRY
VOLUME XLII
NON-DESTRUCTIVE MICROANALYSIS OF CULTURAL HERITAGE MATERIALS Edited by K. JANSSENS R. VAN GRIEKEN University of Antwerp Department of Chemistry Universiteitsplein, 1 B-2610 Antwerp Belgium
2004
ELSEVIER AMSTERDAM – BOSTON – HEIDELBERG – LONDON – NEW YORK – OXFORD – PARIS SAN DIEGO – SAN FRANCISCO – SINGAPORE – SYDNEY – TOKYO
CONTRIBUTORS TO VOLUME XLII Annemie Adriaens Department of Chemistry, Ghent University, Krijgslaan 281 S12, B-9000 Gent, Belgium
[email protected] Michael Alram Coin Cabinet, Kunsthistorisches Museum Vienna, Burgring 5, A-1010 Vienna, Austria
[email protected] Marc Aucouturier C2RMF CNRS, Centre de recherche et restauration des muse´es de France, 6 rue de Pyramides, 75041 Paris cedex 01, France
[email protected] Dimitrios Bikiaris “ORMYLIA” Art Diagnosis Centre, Sacred Convent of the Annunciation, 63071 Ormylia, Greece
[email protected] Ewa Bulska Department of Chemistry, University of Warsaw, Pasteura 1, PL-02-083 Warsawa, Poland
[email protected] Thomas Calligaro C2RMF CNRS, Centre de recherche et restauration des muse´es de France, 6 rue de Pyramides, 75041 Paris cedex 01, France
[email protected] Yannis Chryssoulakis National Technical University of Athens, Athens, Greece and “ORMYLIA” Art Diagnosis Centre, Sacred Convent of the Annunciation, 63071 Ormylia, Greece
[email protected] Sister Daniilia “ORMYLIA” Art Diagnosis Centre, Sacred Convent of the Annunciation, 63071 Ormylia, Greece
[email protected] Evelyne Darque-Ceretti Ecole des Mines de Paris, CEMEF rue C. Daunesse, BP 207, F = 06904 Sophia-Antipolis cedex, France
[email protected]
vi
Contributors to volume XLII
Dalva L.A. de Faria Laboratoria de Espectroscopia Molecular, Instituto de Quimica da USP-University of Sao Paulo, Av. Prof. Lineu Prestes, 784 05508-900, Sao Paulo, Brazil
[email protected] Guy Demortier Laboratoire de Re´actions Nucleaires, Faculte´s Universitaires Notre-Dame de la Paix, 61 rue de Bruxelles, B-5000 Namur, Belgium
[email protected] Marc Dowsett Department of Physics, University of Warwick, Coventry CV4 7AL, United Kingdom
[email protected] Jean-Claude Dran C2RMF CNRS, Centre de recherche et restauration des muse´es de France, 6 rue de Pyramides, 75041 Paris cedex 01, France
[email protected] Howell G.M. Edwards Department of Chemical and Forensic Sciences, University of Bradford, Richmond Road, Bradford BD7 1DP, West Yorkshire, United Kingdom h.g.m.edwards@bradford:ac.uk Bernard Gratuze IRAMAT CNRS, Centre Ernest Babelon, 3D rue de la Fe´rollerie, F-45071 Orle´ans cedex, France
[email protected] Rene´ Van Grieken Centre for Micro- and Trace analysis, Department of Chemistry, University of Antwerp, B-2610 Antwerp, Belgium
[email protected] Annick Hubin Department of Metallurgy, Electrochemistry and Materials Science, Vrije Universiteit Brussel, Pleinlaan 2, B-1050 Brussel, Belgium
[email protected] Koen Janssens Centre for Micro- and Trace analysis, Department of Chemistry, University of Antwerp, B-2610 Antwerp, Belgium
[email protected]
vii
Contributors to volume XLII
Teresa E. Jeffries ICP-MS Facility, Department of Mineralogy, The Natural History Museum, Cromwell Road, London SW7 4BD, United Kingdom
[email protected] Georgios Karagiannis “ORMYLIA” Art Diagnosis Centre, Sacred Convent of the Annunciation, 63071 Ormylia, Greece
[email protected] Robert Linke Institute of Science and Technologies in Art, Academy of Fine Arts Vienna, Schillerplatz 3, A-1010 Vienna, Austria
[email protected] Franz Mairinger Academy of Fine Arts, Dapontegasse 12/16, A-1030 Vienna, Austria
[email protected] Luc Moens Laboratory of Analytical Chemistry, Ghent University, Proeftuinstraat 86, B-9000 Ghent, Belgium
[email protected] Joseph Salomon C2RMF CNRS, Centre de recherche et restauration des muse´es de France, 6 rue de Pyramides, 75041 Paris cedex 01, France
[email protected] Christos Salpistis “ORMYLIA” Art Diagnosis Centre, Sacred Convent of the Annunciation, 63071 Ormylia, Greece
[email protected] Manfred Schreiner Institute of Science and Technologies in Art, Academy of Fine Arts Vienna, Schillerplatz 3, A-1010 Vienna, Austria
[email protected] David Scott GCI Museum Research Laboratory, Getty Conservation Institute, 1200 Getty Center Drive Suite 700, Los Angeles CA 90049-1864, United States
[email protected]
viii
Contributors to volume XLII
Sophia Sotiropoulou “ORMYLIA” Art Diagnosis Centre, Sacred Convent of the Annunciation, 63071 Ormylia, Greece
[email protected] Herman Terryn Department of Metallurgy, Electrochemistry and Materials Science, Vrije Universiteit Brussel, Pleinlaan 2, B-1050 Brussel, Belgium
[email protected] Peter Vandenabeele Laboratory of Analytical Chemistry, Ghent University, Proeftuinstraat 86, B-9000 Ghent, Belgium
[email protected] Barbara Wagner Department of Chemistry, University of Warsaw, Pasteura 1, PL-02-083 Warsawa, Poland
[email protected] Heinz Winter Coin Cabinet, Kunsthistorisches Museum Vienna, Brugring 5, A-1010 Vienna, Austria
ix
This page is intentionally left blank
WILSON AND WILSON’S
COMPREHENSIVE ANALYTICAL CHEMISTRY VOLUMES IN THE SERIES Vol. IA
Vol. IB Vol. IC Vol. IIA
Vol. IIB
Vol. IIC
Vol. IID Vol. III
Vol. IV
Vol. V
Vol. VI Vol. VII Vol. VIII
Vol. IX
Vol. X
Analytical Processes Gas Analysis Inorganic Qualitative Analysis Organic Qualitative Analysis Inorganic Gravimetric Analysis Inorganic Titrimetric Analysis Organic Quantitative Analysis Analytical Chemistry of the Elements Electrochemical Analysis Electrodeposition Potentiometric Titrations Conductometric Titrations High-Frequency Titrations Liquid Chromatography in Columns Gas Chromatography Ion Exchangers Distillation Paper and Thin Layer Chromatography Radiochemical Methods Nuclear Magnetic Resonance and Electron Spin Resonance Methods X-Ray Spectrometry Coulometric Analysis Elemental Analysis with Minute Sample Standards and Standardization Separation by Liquid Amalgams Vacuum Fusion Analysis of Gases in Metals Electroanalysis in Molten Salts Instrumentation for Spectroscopy Atomic Absorption and Fluorescence Spectroscopy Diffuse Reflectance Spectroscopy Emission Spectroscopy Analytical Microwave Spectroscopy Analytical Applications of Electron Microscopy Analytical Infrared Spectroscopy Thermal Methods in Analytical Chemistry Substoichiometric Analytical Methods Enzyme Electrodes in Analytical Chemistry Molecular Fluorescence Spectroscopy Photometric Titrations Analytical Applications of Interferometry Ultraviolet Photoelectron and Photoion Spectroscopy Auger Electron Spectroscopy Plasma Excitation in Spectrochemical Analysis Organic Spot Tests Analysis The History of Analytical Chemistry
xi
Volumes in the series Vol. XI
Vol. XII
Vol. XIII
Vol. XIV Vol. XV Vol. XVI Vol. XVII Vol. XVIII Vol. Vol. Vol. Vol. Vol.
XIX XX XXI XXII XXIII
Vol. Vol. Vol. Vol. Vol. Vol. Vol. Vol. Vol. Vol. Vol.
XXIV XXV XXVI XXVII XXVIII XXIX XXX XXXI XXXII XXXIII XXXIV
Vol. Vol. Vol. Vol.
XXXV XXXVI XXXVII XXXVIII
Vol. XXXIX Vol. XL Vol. XLI
xii
The Application of Mathematical Statistics in Analytical Chemistry Mass Spectrometry Ion Selective Electrodes Thermal Analysis Part A. Simultaneous Thermoanalytical Examination by Means of the Derivatograph Part B. Biochemical and Clinical Applications of Thermometric and Thermal Analysis Part C. Emanation Thermal Analysis and other Radiometric Emanation Methods Part D. Thermophysical Properties of Solids Part E. Pulse Method of Measuring Thermophysical Parameters Analysis of Complex Hydrocarbons Part A. Separation Methods Part B. Group Analysis and Detailed Analysis Ion-Exchangers in Analytical Chemistry Methods of Organic Analysis Chemical Microscopy Thermomicroscopy of Organic Compounds Gas and Liquid Analysers Kinetic Methods in Chemical Analysis Application of Computers in Analytical Chemistry Analytical Visible and Ultraviolet Spectrometry Photometric Methods in Inorganic Trace Analysis New Developments in Conductometric and Oscillometric Analysis Titrimetric Analysis in Organic Solvents Analytical and Biomedical Applications of Ion-Selective Field-Effect Transistors Energy Dispersive X-Ray Fluorescence Analysis Preconcentration of Trace Elements Radionuclide X-Ray Fluorescence Analysis Voltammetry Analysis of Substances in the Gaseous Phase Chemiluminescence Immunoassay Spectrochemical Trace Analysis for Metals and Metalloids Surfactants in Analytical Chemistry Environmental Analytical Chemistry Elemental Speciation – New Approaches for Trace Element Analysis Discrete Sample Introduction Techniques for Inductively Coupled Plasma Mass Spectrometry Modern Fourier Transform Infrared Spectroscopy Chemical Test Methods of Analysis Sampling and Sample Preparation for Field and Laboratory Countercurrent Chromatography: The Support-Free Iiquid Stationary Phase Integrated Analytical Systems Analysis and Fate of Surfactants in the Aquatic Environment Sample Preparation for Trace Element Analysis
Contents Contributors to Volume Volumes in the Series . Series Editor’s Preface Preface . . . . . . . .
XLII . . . . . . . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
Chapter 1. Introduction and overview . . . . . K. Janssens and R. Van Grieken 1.1 Introduction . . . . . . . . . . . . 1.2 Overview of the analytical reference 1.3 Overview of the case studies section References . . . . . . . . . . . . . . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. vi . xi . xxv . xxvii
. . . . . . . . . . . . . . . . section . . . . . . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
1
. . . .
1 5 7 10
Chapter 2. UV-, IR- and X-ray imaging . . . . . . . . . . . . . . . . Franz Mairinger 2.1 Scientific investigations of works of arts and crafts . . . . 2.2 Application of electromagnetic radiation for the examination of cultural heritage objects . . . . . . . . . . 2.3 Instrumental basis . . . . . . . . . . . . . . . . . . . . . 2.3.1 Light and radiation sources . . . . . . . . . . . . 2.3.2 Imaging . . . . . . . . . . . . . . . . . . . . . . 2.3.3 Sensor systems . . . . . . . . . . . . . . . . . . . 2.3.4 Sensor subsystems . . . . . . . . . . . . . . . . . 2.4 Surface examinations . . . . . . . . . . . . . . . . . . . 2.4.1 Surface examinations with ultraviolet radiation . . 2.4.2 Instrumental techniques for UV-fluorescence photography . . . . . . . . . . . . . . . . . . . . 2.4.3 Instrumental techniques for reflected UV photography . . . . . . . . . . . . . . . . . . . . 2.4.4 Application of UV-fluorescence photography . . . . 2.4.5 Application of UV photography . . . . . . . . . . 2.5 Depth examinations . . . . . . . . . . . . . . . . . . . . 2.5.1 Depth examinations with infrared radiation . . . . 2.5.2 Depth examinations with X-rays and gamma-rays . References . . . . . . . . . . . . . . . . . . . . . . . . . . . .
15
PART I. ANALYTICAL REFERENCE SECTION
15 16 16 17 25 33 41 44 45 46 46 47 48 49 50 54 66
xiii
Contents
Chapter 3. Electron microscopy and its role in cultural heritage studies . . . . . . . . . . . . . . . . . . . . . A. Adriaens and M.G. Dowsett 3.1 Introduction . . . . . . . . . . . . . . . . . . . 3.1.1 Why use electron microscopy? . . . . . . 3.1.2 Imaging with electrons . . . . . . . . . . 3.1.3 Varieties of electron microscopy . . . . . 3.1.4 Recent developments in commercial SEM 3.2 The interaction of electrons with a solid—contrast mechanisms . . . . . . . . . . 3.2.1 Scattering . . . . . . . . . . . . . . . . 3.2.2 Secondary electron emission . . . . . . . 3.2.3 Backscattered electrons . . . . . . . . . 3.2.4 Cathodoluminescence . . . . . . . . . . 3.2.5 Core-level excitation and X-ray or Auger emission . . . . . . . . . . . . . . 3.2.6 Electron energy loss spectroscopy . . . . 3.2.7 Diffraction in TEM . . . . . . . . . . . . 3.2.8 Image contrast in TEM . . . . . . . . . 3.3 Components and optics of electron microscopes . . . . . . . . . . . . . . . . . . . . 3.3.1 Basic optics . . . . . . . . . . . . . . . . 3.3.2 The electron gun . . . . . . . . . . . . . 3.3.3 Focusing an electron beam . . . . . . . . 3.3.4 Newtonian lens model . . . . . . . . . . 3.3.5 The magnetic prism . . . . . . . . . . . 3.3.6 Detectors . . . . . . . . . . . . . . . . . 3.4 Sample preparation techniques . . . . . . . . . 3.5 Origin/ provenance studies . . . . . . . . . . . . 3.5.1 Ceramics . . . . . . . . . . . . . . . . . 3.5.2 Glass . . . . . . . . . . . . . . . . . . . 3.6 Technology and techniques of manufacture . . . 3.6.1 Seals . . . . . . . . . . . . . . . . . . . 3.6.2 Ceramics . . . . . . . . . . . . . . . . . 3.6.3 Glass . . . . . . . . . . . . . . . . . . . 3.6.4 Metals . . . . . . . . . . . . . . . . . . 3.7 Use . . . . . . . . . . . . . . . . . . . . . . . . 3.8 Degradation processes, corrosion and weathering . . . . . . . . . . . . . . . . . . . . 3.8.1 Metals . . . . . . . . . . . . . . . . . .
xiv
. . . . .
73
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
73 73 73 75 77
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
79 80 81 82 83
. . . .
. . . .
. . . .
. . . .
. . . .
84 86 88 91
. . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . .
92 92 92 93 95 97 98 102 104 104 105 106 106 107 113 114 117
. . . . . . . . . .
118 119
Contents
3.8.2 Glass . . . . . . . . . . 3.8.3 Ceramics . . . . . . . . 3.9 Authenticity and authentication 3.10 Conclusions . . . . . . . . . . Acknowledgements . . . . . . . . . . References . . . . . . . . . . . . . .
. . . . . .
. . . . . .
. . . . . .
. . . . . .
. . . . . .
. . . . . .
. . . . . .
. . . . . .
. . . . . .
. . . . . .
. . . . . .
. . . . . .
. . . . . .
. . . . . .
119 120 121 123 124 124
Chapter 4. X-ray based methods of analysis . . . . . . . . . . . . K. Janssens 4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . 4.2 Basic principles . . . . . . . . . . . . . . . . . . . . . . 4.2.1 X-ray wavelength and energy scales . . . . . . . 4.2.2 Interaction of X-rays with matter . . . . . . . . 4.2.3 The photoelectric effect; X-ray fluorescence . . . . . . . . . . . . . . . . . . . . 4.2.4 Scattering and diffraction . . . . . . . . . . . . 4.2.5 X-ray absorption fine structure and spectroscopy . . . . . . . . . . . . . . . . . . . 4.3 Instrumentation for X-ray investigations . . . . . . . . 4.3.1 X-ray sources . . . . . . . . . . . . . . . . . . . 4.3.2 X-ray detectors . . . . . . . . . . . . . . . . . . 4.3.3 X-ray fluorescence instrumentation . . . . . . . 4.3.4 XRD instrumentation . . . . . . . . . . . . . . 4.3.5 XAS instrumentation at SR beamlines . . . . . . 4.4 A survey of applications of X-ray methods in the cultural heritage sector . . . . . . . . . . . . . . . . . . . . . . 4.4.1 Compositional analysis of historic glass . . . . . 4.4.2 Pigments . . . . . . . . . . . . . . . . . . . . . 4.4.3 Lustre ware . . . . . . . . . . . . . . . . . . . 4.4.4 Metallic artefacts . . . . . . . . . . . . . . . . . 4.4.5 Analysis of graphic documents . . . . . . . . . . 4.4.6 Mn oxidation in odontolites . . . . . . . . . . . 4.4.7 Therapeutic and cosmetical chemicals of Ancient Egypt . . . . . . . . . . . . . . . . . 4.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . A4.1 Figures-of-merit for XRF spectrometers . . . . . . . . . A4.1.1 Analytical sensitivity . . . . . . . . . . . . . . . A4.1.2 Detection and determination limits . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . .
.
129
. . . .
129 131 131 131
. .
133 137
. . . . . . .
143 147 148 155 164 186 193
. . . . . . .
194 195 195 200 202 206 210
. . . . . . .
211 215 216 217 217 217 219
xv
Contents
Chapter 5. Ion beam microanalysis . . . . . . . . . . . . T. Calligaro, J.-C. Dran and J. Salomon 5.1 Historical background and motivation . . . . . 5.2 Fundamentals of ion beam analysis . . . . . . 5.2.1 Interaction of radiations with matter . 5.2.2 Particle-induced X-ray emission . . . . 5.2.3 Elastic scattering of particles . . . . . 5.2.4 Nuclear reaction analysis . . . . . . . 5.3 Specific arrangements for the study of art and archaeological objects . . . . . . . . . . . . . 5.3.1 External beams . . . . . . . . . . . . . 5.3.2 Nuclear microprobes . . . . . . . . . . 5.3.3 Micro and macro-imaging . . . . . . . 5.3.4 Portable systems . . . . . . . . . . . . 5.4 Applications in the field of art and archaeology 5.4.1 Materials’ identification . . . . . . . . 5.4.2 Provenance of the materials . . . . . . 5.4.3 Alteration phenomena . . . . . . . . . 5.4.4 Authentication and relative dating . . . 5.5 Survey of worldwide IBA activity in the field of cultural heritage . . . . . . . . . . . . . . . . 5.6 Conclusion and future prospects . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . .
. . . . . .
227
. . . . . .
. . . . . .
. . . . . .
. . . . . .
. . . . . .
. . . . . .
227 228 230 237 244 246
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
251 251 253 255 255 257 257 263 266 267
. . . . . . . . . . . . . . . . . .
268 268 271
Chapter 6. X-ray photoelectron and Auger electron spectroscopy Annick Hubin and Herman Terryn 6.1 Introduction . . . . . . . . . . . . . . . . . . . . . 6.2 The basic concepts of XPS and AES . . . . . . . . . 6.2.1 Principle of X-ray photoelectron spectroscopy 6.2.2 Principle of Auger electron spectroscopy . . . 6.3 XPS and AES instruments . . . . . . . . . . . . . . 6.3.1 General set-up . . . . . . . . . . . . . . . . 6.3.2 The vacuum system . . . . . . . . . . . . . 6.3.3 The X-ray source for XPS . . . . . . . . . . 6.3.4 The electron gun for AES . . . . . . . . . . 6.3.5 Detection of electron energy . . . . . . . . . 6.3.6 The ion gun . . . . . . . . . . . . . . . . . 6.3.7 The sample holder and stage . . . . . . . . . 6.4 Sample requirements . . . . . . . . . . . . . . . .
xvi
. . .
277
. . . . . . . . . . . . .
277 279 279 282 286 286 287 289 290 291 294 295 296
. . . . . . . . . . . . .
. . . . . . . . . . . . .
Contents
6.5
Information in XPS and AES spectra . . . 6.5.1 Surface analysis . . . . . . . . . . 6.5.2 Qualitative analysis . . . . . . . . 6.5.3 Quantitative analysis . . . . . . . 6.5.4 Chemical analysis . . . . . . . . . 6.5.5 In-depth analysis . . . . . . . . . 6.5.6 Data analysis . . . . . . . . . . . 6.5.7 Imaging . . . . . . . . . . . . . . 6.6 Comparison of XPS, AES and other surface analytical techniques . . . . . . . . . . . . 6.7 XPS and AES for chemical analysis of cultural heritage materials . . . . . . . . . Acknowledgements . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
297 297 297 298 301 303 306 306
. . . . . . . .
308
. . . . . . . . . . . . . . . . . . . . . . . .
308 311 311
Chapter 7. Laser ablation inductively coupled plasma mass spectrometry . . . . . . . . . . . . . . . . . . . . . . . . Teresa E. Jeffries 7.1 Introduction . . . . . . . . . . . . . . . . . . . . 7.2 The inductively coupled plasma mass spectrometer 7.2.1 Historical account . . . . . . . . . . . . . 7.2.2 Operational rationale . . . . . . . . . . . . 7.2.3 The inductively coupled plasma . . . . . . 7.2.4 The plasma sampling interface . . . . . . 7.2.5 Ion focusing . . . . . . . . . . . . . . . . 7.2.6 Quadrupole mass analyser . . . . . . . . . 7.2.7 The vacuum system . . . . . . . . . . . . 7.2.8 Ion detection and signal handling . . . . . 7.3 Laser ablation: essential components . . . . . . . 7.3.1 Development of the laser . . . . . . . . . . 7.3.2 The association of lasers with ICP-MS . . . 7.3.3 Stimulated emission . . . . . . . . . . . . 7.3.4 Nd:YAG laser (resonator) cavity . . . . . . 7.3.5 Harmonic generation . . . . . . . . . . . . 7.3.6 Harmonic separation . . . . . . . . . . . . 7.3.7 Energy attenuation and control. . . . . . . 7.3.8 Beam delivery and viewing optics . . . . . 7.3.9 Ablation cell and sample transport . . . . 7.4 Analytical concepts and factors affecting analysis . 7.4.1 Why use the technique? . . . . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . .
313
. . . . . . . . . . . . . . . . . . . . . .
313 314 314 315 317 321 323 323 327 329 329 329 331 333 334 337 338 340 341 342 343 343
. . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . .
xvii
Contents
7.4.2 7.4.3 7.4.4 7.4.5 7.4.6
Sample preparation and mounting . . . . . Analysis of transient signals . . . . . . . . Factors affecting analysis . . . . . . . . . Optimization and calibration . . . . . . . . Figures of merit and analytical performance targets . . . . . . . . . . . . . . . . . . . 7.5 Continuing developments and final remarks . . . . 7.5.1 Continuing developments . . . . . . . . . 7.5.2 Final remarks . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . Chapter 8. Infrared, Raman microscopy and fibre-optic Raman spectroscopy (FORS) . . . . . . . . . . . . . . Howell G.M. Edwards and Dalva L.A. de Faria 8.1 Introduction . . . . . . . . . . . . . . . . . . . 8.2 Comparison of the potential use of IR and Raman spectroscopies for the non-destructive analysis of art works . . . . . . . . . . . . . . . . . . . 8.3 Some theoretical aspects of IR and Raman spectroscopies . . . . . . . . . . . . . . . . . . 8.4 Instrumentation . . . . . . . . . . . . . . . . . 8.5 Sampling . . . . . . . . . . . . . . . . . . . . . 8.6 Resonance Raman . . . . . . . . . . . . . . . . 8.7 SERS . . . . . . . . . . . . . . . . . . . . . . . 8.8 Intensity measurements in Raman scattering . . 8.9 Raman spectroscopy with fibre optics . . . . . . 8.9.1 Sampling considerations . . . . . . . . . 8.9.2 Probe design . . . . . . . . . . . . . . . 8.9.3 Probe background . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . .
. . . .
. . . .
. . . .
. . . .
344 345 347 350
. . . . .
. . . . .
. . . . .
. . . . .
352 355 355 356 357
. . . . .
359
. . . . .
359
. . . . .
360
. . . . . . . . . . .
. . . . . . . . . . .
367 373 381 387 390 390 392 392 392 393 393
. .
397
. . . . . .
397 399 399 410 420 420
. . . . . . . . . . .
. . . . . . . . . . .
Chapter 9. Secondary ion mass spectrometry. Application to archaeology and art objects . . . . . . . . . . . . . . . . . . Evelyne Darque-Ceretti and Marc Aucouturier 9.1 Introduction . . . . . . . . . . . . . . . . . . . . . . 9.2 Principles and equipment . . . . . . . . . . . . . . . 9.2.1 Principles . . . . . . . . . . . . . . . . . . . . 9.2.2 Equipment and choice of analytical parameters 9.3 Analysis procedures . . . . . . . . . . . . . . . . . . 9.3.1 Elemental identification, sensitivity . . . . . .
xviii
. . . . . . . . . . .
. . . . . .
Contents
9.3.2 Quantitative analysis . . . . . . . . . . . . 9.3.3 In-depth analysis and depth resolution . . . 9.3.4 Surface analysis . . . . . . . . . . . . . . . 9.3.5 Imaging, lateral resolution . . . . . . . . . . 9.3.6 Chemical compound analysis and distribution 9.4 Examples of applications for cultural heritage . . . . 9.4.1 Dating and/or provenance studies based on isotopic analysis . . . . . . . . . . . . . 9.4.2 Dating (not based on isotopic analysis) . . . 9.4.3 Provenance studies not based on isotopic analysis . . . . . . . . . . . . . . . . . . . 9.4.4 Surface layer analysis on artefacts . . . . . 9.4.5 Interface studies on coated layers . . . . . . 9.4.6 ToF-SIMS applications . . . . . . . . . . . . 9.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . .
. . . . . .
. . . . . .
425 429 435 438 439 440
. . . . . .
440 444
. . . . . .
. . . . . .
447 448 453 455 458 459
. . . . . . .
465
. . . . . . . . . .
. . . . . . . . . .
465 467 470 474 477 480 485 488 490 490
. . .
493
. . . . . .
493 493
. . .
493
. . . . . .
PART II. CASE STUDIES SECTION
Chapter 10. The non-destructive investigation of copper alloy patinas . . . . . . . . . . . . . . . . . . . . David A. Scott 10.1 A brief historical account . . . . . . . . . . 10.2 Optical examination . . . . . . . . . . . . . 10.3 Environmental scanning electron microscopy 10.4 X-ray fluorescence analysis . . . . . . . . . 10.5 Scanning X-ray fluorescence microanalysis . 10.6 XRD analysis . . . . . . . . . . . . . . . . . 10.7 FTIR spectroscopy . . . . . . . . . . . . . . 10.8 Conclusions . . . . . . . . . . . . . . . . . Acknowledgements . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
Chapter 11. Precious metals artefacts . . . . . . . . . . . . . . G. Demortier 11.1 Introduction . . . . . . . . . . . . . . . . . . . . . 11.2 Non-destructive analysis of gold jewellery items . . . 11.2.1 Contribution of atomic and nuclear (but non-radioactive) methods to the analysis of ancient gold jewellery items . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
xix
Contents
11.2.2 Illustration of the analytical performances of non-vacuum PIXE for gold artefacts . . . . . . 11.3 The soldering of gold . . . . . . . . . . . . . . . . . . 11.3.1 Ancient recipes for gold soldering . . . . . . . 11.3.2 Iranian goldsmithery from the 4th century BC . 11.3.3 Tartesic gold artefacts . . . . . . . . . . . . . 11.3.4 Later Iranian goldsmithery . . . . . . . . . . . 11.3.5 Preparations of low-melting brazing alloys . . . 11.3.6 A new reading of Elder Pliny’s Natural History 11.3.7 Italian jewellery . . . . . . . . . . . . . . . . 11.3.8 Gold artefacts from Slovenia . . . . . . . . . . 11.3.9 The Guarrazar treasure . . . . . . . . . . . . 11.3.10 Merovingian and late Byzantine jewellery . . . 11.4 Pre-Hispanic gold artefacts of Mesoamerica . . . . . . 11.4.1 Archaeological context . . . . . . . . . . . . . 11.4.2 A selection of typical artefacts . . . . . . . . . 11.4.3 Differential PIXE . . . . . . . . . . . . . . . . 11.4.4 Application to the measurement of the gold enhancement at the surface of tumbaga . . . . 11.5 Characterization of complex items . . . . . . . . . . . 11.5.1 XRF induced by a g-ray source . . . . . . . . . 11.5.2 Gamma-ray transmission measurements . . . 11.5.3 Study of a composite gold jewellery artefact . . 11.6 Gold coins . . . . . . . . . . . . . . . . . . . . . . . 11.6.1 Fineness measurements of gold coins . . . . . 11.6.2 Gold coins from the ancient world . . . . . . . 11.6.3 Gold coins from the new world . . . . . . . . . 11.7 Conclusions . . . . . . . . . . . . . . . . . . . . . . . Acknowledgements . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . .
496 498 498 499 502 506 509 517 520 521 525 528 530 530 534 536
. . . . . . . . . . . .
. . . . . . . . . . . .
538 544 544 544 545 548 548 548 556 558 559 560
Chapter 12. Diagnostic methodology for the examination of Byzantine frescoes and icons. Non-destructive investigation and pigment identification. . . . . . . . . . . . . . . . . . . . . Sister Daniilia, Sophia Sotiropoulou, Dimitrios Bikiaris, Christos Salpistis, Georgios Karagiannis and Yannis Chryssoulakis 12.1 Introduction. . . . . . . . . . . . . . . . . . . . . . . . . 12.2 The Entry of the Mother of God into the Temple . . . . . . 12.2.1 Description . . . . . . . . . . . . . . . . . . . . .
xx
565
565 566 566
Contents
12.2.2 The preparation of the plaster: materials and technique . . . . . . . . . . . . . . 12.2.3 The drawing . . . . . . . . . . . . . . . 12.2.4 Materials and painting techniques . . . . 12.2.5 Study of the colour palette . . . . . . . . 12.2.6 Conclusions . . . . . . . . . . . . . . . 12.3 Mother of God Hodegetria . . . . . . . . . . . . 12.3.1 Description . . . . . . . . . . . . . . . . 12.3.2 Construction and state of preservation of the support . . . . . . . . . . . . . . 12.3.3 State of preservation of the surface . . . 12.3.4 The ground . . . . . . . . . . . . . . . . 12.3.5 The drawing . . . . . . . . . . . . . . . 12.3.6 Materials and technique of the painting . 12.3.7 Conclusions . . . . . . . . . . . . . . . A12.1 Experimental details . . . . . . . . . . . . . . . A12.1.1 Non-destructive analysis . . . . . . . . A12.1.2 Micro-sampling analysis . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
567 568 570 581 584 585 585
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
587 588 592 592 593 601 602 602 603 604
.
605
. . . . . . . . . . .
605 606 606 608 613 622 622 624 629 630 630
Chapter 14. Pigment identification in illuminated manuscripts . . . Peter Vandenabeele and Luc Moens 14.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . .
635
Chapter 13. The provenance of medieval silver coins: analysis with EDXRF, SEM/EDX and PIXE . . . . . . . . . . . . . . . . . . Robert Linke, Manfred Schreiner, Guy Demortier, Michael Alram and Heinz Winter 13.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . 13.2 The Friesacher Pfennig . . . . . . . . . . . . . . . . . 13.2.1 Introduction . . . . . . . . . . . . . . . . . . . 13.2.2 Experimental . . . . . . . . . . . . . . . . . . 13.2.3 Results . . . . . . . . . . . . . . . . . . . . . . 13.3 The Tiroler Kreuzer . . . . . . . . . . . . . . . . . . . 13.3.1 Introduction . . . . . . . . . . . . . . . . . . . 13.3.2 Experimental . . . . . . . . . . . . . . . . . . 13.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . .
635
xxi
Contents
14.2 Combined method approach . . . . . . . . . . . . 14.2.1 Analysis of manuscripts . . . . . . . . . . 14.2.2 Sources of impurities . . . . . . . . . . . . 14.3 Analysis of the manuscripts from the collection of Raphael De Mercatellis . . . . . . . . . . . . . 14.3.1 Introduction . . . . . . . . . . . . . . . . 14.3.2 Pigment identification with TXRF and MRS 14.3.3 Intra-manuscript comparison of Expositio problematum Aristotelis . . . . . . . . . . 14.3.4 Analysis of Decretum Gratiani . . . . . . . 14.4 Conclusion . . . . . . . . . . . . . . . . . . . . . Acknowledgements . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . .
636 639 652
. . . . . . . . . . . .
654 654 655
. . . . .
. . . . .
. . . . .
657 658 659 660 660
Chapter 15. Provenance analysis of glass artefacts . . . . . . . . . Bernard Gratuze and Koen Janssens 15.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . 15.2 Obsidian, a natural glass used since the Paleolithic era . 15.3 Bronze and Iron age glasses . . . . . . . . . . . . . . . 15.3.1 Neolithic first artificial glassy materials and the discovery of glass during Bronze Age . . . . 15.3.2 When trade beads reached Europe . . . . . . . . 15.3.3 Middle Bronze Age plant ash soda-lime glasses . 15.3.4 Late Bronze Age mixed soda – potash glasses . . 15.3.5 Iron Age and Antiquity natron soda-lime glasses 15.3.6 Protohistoric glass trade routes . . . . . . . . . 15.3.7 Glass chrono-typo-chemical models: a dating tool? . . . . . . . . . . . . . . . . . . . 15.4 Glass trade towards and from Central Asia and the Indian world during Antiquity . . . . . . . . . . . . 15.5 Carolingian glass production: some unusual lead glass composition smoothers . . . . . . . . . . . . . . . 15.6 Late Middle Age recycled glass . . . . . . . . . . . . . 15.7 Glass technology transfer during the 16th –17th century to and from Antwerp . . . . . . . . . . . . . . . . . . . 15.8 Trade beads: the glass trade internationalization, during the Post-Medieval period . . . . . . . . . . . . . 15.9 Conclusions . . . . . . . . . . . . . . . . . . . . . . . Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . .
.
663
. . .
663 665 670
. . . . . .
670 671 672 674 675 678
.
678
.
680
. .
687 691
.
699
. . . .
702 705 706 707
xxii
. . . . .
Chapter 16. Corrosion of historic glass and enamels . . . . . . . Manfred Schreiner 16.1 Introduction . . . . . . . . . . . . . . . . . . . . . . 16.2 The weathering of medieval stained glass . . . . . . . 16.2.1 SEM investigations of the corrosion phenomena on naturally weathered medieval glass . . . . 16.2.2 The determination of hydrogen in the leached surface layer by SIMS and NRA . . . . . . . . 16.2.3 Leaching studies of glass with medieval composition . . . . . . . . . . . . . . . . . . 16.2.4 IRRAS investigations on leached glass with medieval composition . . . . . . . . . . . . . 16.2.5 Weathering phenomena on glass with medieval composition studied with TM-AFM . . . . . . 16.3 The degradation of medieval enamels . . . . . . . . . 16.3.1 SEM investigations of the enamel of the medieval goblets . . . . . . . . . . . . . . . . 16.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . Acknowledgements . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . .
. .
713
. . . .
713 716
. .
719
. .
725
. .
731
. .
735
. . . .
738 742
. . . .
. . . .
745 750 752 753
.
755
. . . . . .
755 756 758 760 763 766
. .
766 768
.
770
.
773
. .
774 775
Chapter 17. A study of ancient manuscripts exposed to iron – gall ink corrosion . . . . . . . . . . . . . . . . . . . . . . . . . . Ewa Bulska and Barbara Wagner 17.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . 17.1.1 Iron –gall ink . . . . . . . . . . . . . . . . . . . 17.1.2 Iron –gall ink corrosion . . . . . . . . . . . . . 17.1.3 Investigated artefacts . . . . . . . . . . . . . . 17.2 Analytical methods . . . . . . . . . . . . . . . . . . . 17.2.1 Inspection by scanning electron microscopy . . . 17.2.2 Compositional analysis by X-ray fluorescence spectrometry . . . . . . . . . . . . . . . . . . . 17.2.3 Electron probe micro-analysis . . . . . . . . . . 17.2.4 Laser ablation inductively coupled plasma mass spectrometry . . . . . . . . . . . . . . . . 17.2.5 Elemental analysis by inductively coupled plasma mass spectrometry . . . . . . . . . . . . . . . . 17.2.6 Graphite furnace atomic absorption spectrometry . . . . . . . . . . . . . . . . . . . 17.2.7 Mo¨ssbauer spectrometry . . . . . . . . . . . . .
xxiii
Contents
17.2.8 Investigation of Fe(II)/Fe(III) by X-ray absorption near edge spectroscopy . . . . . . . . . . . . . . 17.3 Searching for the conservation treatment . . . . . . . . 17.3.1 Reconstitution of manuscript by model samples . 17.3.2 Requirement for conservation treatment . . . . 17.3.3 Investigation of the model samples . . . . . . . 17.4 Concluding comments . . . . . . . . . . . . . . . . . . Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . .
778 780 780 782 784 784 785 785
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
789
xxiv
Series Editor’s Preface The investigation of cultural heritage materials is a field that, contrary to other fashion or curiosity-driven research topics, will always need to be funded. It becomes obvious to all of us that to achieve knowledge and progress in our society, we need to know more about our past, represented by artefacts of previous civilisations such as ancient silver and gold coins, frescoes or old manuscripts. In this respect Janssens and Van Grieken have produced a timely book on the non-destructive analysis of cultural heritage materials and it is a useful addition to the Comprehensive Analytical Chemistry Series. Although we have recently published books on molecular and spectroscopic techniques, the artefacts that comprise our cultural heritage require specific techniques and instrumentation which should preferably be non-destructive or micro-destructive and the measurements often need to be made in situ. By using non-destructive or micro-destructive techniques, the excellent ancient pieces of work that need to be examined do not suffer any damage or no visible damage, respectively. To get a good overview of the contents of this book, I strongly recommend that you first read chapter one, where the editors describe in detail the content and structure of the book. The book contains 17 chapters. In the first part the emphasis is on the techniques used in this field. There are 9 chapters, which include a concise description of non-destructive microanalytical techniques, such as X-ray based methods, electron microscopy and Raman microscopy, among others. The second part is intended to show to the reader how the different analytical techniques are being applied in the real world, and comprises 8 chapters, with applications concerning the diagnostics of different important pieces of work such as Byzantine frescoes and icons, medieval silver coins or glass artefacts among other cultural heritage materials. In summary, the book gives a state-of-the art report on the major techniques used for the non-destructive analysis of cultural heritage artefacts or materials. It is with a great deal of pleasure that I take this opportunity to present this exceptional and unique volume and I would like to thank the editors and the authors for their excellent work.
xxv
Series Editor’s Preface
With this book we have now published 42 volumes and we appear to be on course to reach volume 50 by the year 2006. I would like to thank not only the editors and authors of past, present and future volumes but also the scientific community for the broad acceptance of the Comprehensive Analytical Chemistry series. Damia Barcelo´ Barcelona, June 2004
xxvi
Preface
During the last decade, considerable progress has been made in the instrumental and methodological aspects of microscopic analysis. These improvements are apparent in rather diverse fields, ranging from infrared reflectography over Raman microscopy and X-ray microanalysis by means of electrons, protons and photons to mass spectrometric methods that employ laser sampling. Considerable improvements in detector technology, instrument-computer interfacing, focusing optics, and in the performance of radiation sources suitable for use in various parts of the electromagnetic spectrum lie at the basis of the advances. In the present context of investigations of cultural heritage materials, which often must be performed in situ, an important recent development has been the ongoing miniaturization of components, permitting the design of compact, portable and sometimes handheld analytical instruments that are able to provide analytical data of comparable quality to those produced by ‘regular’ laboratory instruments. Since many of these technological improvements are predominantly application-driven, the above-mentioned progress has led to an effective increase in the applicability of the various methods that are described in this book. The augmented applicability is, on the one hand, realized through an increase in the sensitivity, the detection power and the lateral, spatial and/or spectral resolution of the methods, thus allowing information on more subtle variations in specific material properties to be obtained in a more accurate and/or more precise manner. In some cases, an increasing number of independent properties can simultaneously be recorded from the same location in a material. In some of the application-oriented chapters in the second part of the book, the advantages of the combined use of various atomic and molecular spectroscopies for gathering different types of information on the same material, objects or problem is already apparent. The applicability of (micro)analytical methods for the investigation of materials from the cultural heritage field is also realized through the relaxation of the boundary conditions or circumstances that have traditionally limited the application range of a number of the conventional methods of
xxvii
Preface
investigation. In earlier decades, conventional spectroscopic methods could exclusively be used for the investigation of ‘ideal’ samples (e.g., dilute solutions of previously dissolved materials, polished disks of metals or glass), usually requiring considerable sample preparation. Through the development of microscopic and/or solid-sampling variants of the original methods, complex objects, composed of different materials and having non-flat, irregular shapes can now also be analyzed with some reliability, either with minimal or no damage whatsoever to the objects. All chapters have been written by experts in their own fields, whether they be predominantly methodological and instrumental in nature or more specifically directed towards the study of a particular kind of cultural heritage problem or artifact type. The multi-author approach, although inevitably leading to a certain variability in style of presentation, in our opinion, outweighs any advantages of uniformity and homogeneity that characterize a single-author book. Since many different methods and problems are covered in this volume, it is in practice impossible for a single individual to cover a significant fraction of the relevant methodological, technical and applied aspects of non-destructive micro-analysis. From the beginning, our intention was to target two types of practitioners with this book. On the one hand, we hope that MSc and PhD level students, and spectroscopists and analysts with a background in the natural sciences, seeking to broaden their knowledge on methods of non-destructive microanalysis and how (one or combination of several of) these methods may be applied in the cultural heritage area, will profit from this book. On the other hand, by presenting in the same volume a collection of applied studies, where several methods of analysis are employed together and in a problem-oriented fashion, we also hope to have provided a resource for conservators and museum curators who are faced with questions about which combination of analytical methods or services would be optimal to shed light on a particular curatorial or conservation problem. K. Janssens R. Van Grieken
xxviii
Chapter 1
Introduction and overview K. Janssens and R. Van Grieken
1.1
INTRODUCTION
For the study, conservation and restoration of materials and artefacts of culturo-historical value, there is a well-defined need for analytical methods that are able to provide information on (see Fig. 1.1): † the chemical nature/composition of selected parts of cultural heritage artefacts and materials in order to elucidate their provenance; † the state of alteration (on the surface and /or internally) of objects as a result of short-, medium- and long-term exposure to particular environmental conditions; † the effect/effectiveness of conservation/restoration strategies during and after application. According to Lahanier et al. [1], the ideal method for analysing objects of artistic, historic or archaeological nature should be: (a) non-destructive, i.e., respecting the physical integrity of the material/ object. Often valuable objects can only be investigated when the analysis does not result in any (visible) damage to the object. Usually this completely eliminates sampling or limits it to very small amounts; (b) fast, so that large numbers of similar objects may be analysed or a single object investigated at various positions on its surface; this property is very valuable since this is the only way of being able to discern between general trends in the data and outlying objects or data points; (c) universal, so that by means of a single instrument, many materials and objects of various shapes and dimensions may be analysed with minimal sample pre-treatment; (d) versatile, allowing with the same technique to obtain average compositional information as well as local information of small areas (e.g., millimetre to micron-sized) from heterogeneous materials; Comprehensive Analytical Chemistry XLII Janssens and Van Grieken (Eds.) q 2004 Elsevier B.V. All rights reserved
1
K. Janssens and R. Van Grieken
Fig. 1.1. Interaction between cultural heritage materials, the use of analytical techniques and environmental factors.
(e) sensitive, so that object grouping and other types of provenance analysis can be done by means of not only major elements but also trace-element fingerprints; and (f) multi-elemental, so that in a single measurement, information on many elements is obtained simultaneously and, more importantly, information is also obtained on elements which were not initially thought to be relevant to the investigation. While most methods that are described in the chapters of this book fulfil several, but usually not all, requirements described above, it is obvious that in the cultural heritage field, the analytical techniques should preferably be non-destructive or micro-destructive. Non-destructive techniques allow analytical information to be obtained with no damage to the sample or (in some cases) to the artefacts in question. When micro-destructive methods are used, all visible damage is avoided and the objects under examination remain aesthetically unimpaired [2]. The possibility of using these types of methods is of enormous advantage when sampling is not feasible or when fragments used for analysis need to be put back in their original location at the end of the investigation. Among the truly non-destructive methods are the spectroscopies based on ultraviolet, visual and infrared (IR) radiations, as well as the X-ray-based methods.
2
Introduction and overview
Objects and monuments of culturo-historical significance comprise a wide variety of materials (metals, ceramics, glass, various igneous and sedimentary rocks, textile, leather, wood, horn, parchment, paper, etc.) and usually exhibit a fairly complex three-dimensional structure and a heterogeneous chemical composition. Especially in the case of artefacts of precious nature (e.g., jewellery, weaponry, religious objects), cultural heritage objects often: † are composed of various materials (e.g., Au/Ag alloy artefacts adorned with gemstones), † consist of a base material covered with one or more layers of pigmentation (e.g., polychrome wooden statues, easel paintings, illuminated manuscripts) or † show significant (surface) alteration due to burial or atmospheric exposure (e.g., bronze statues, silver coins, etc.) [3]. In addition to the requirement that the method(s) of analysis that are employed to assess the material state of such objects/materials are as non-destructive as realistically possible, in most cases, preference is then given to methods that are able to yield information on well-defined areas of the artefacts in question [4]. Sometimes, these areas are microscopically small; in other cases, a lateral resolution of 1 mm2 suffices. Techniques having a high spatial resolution are considered by some to take an intermediate place between destructive and non-destructive methods. In recent years, partially driven by the increasing importance of hightech materials of a complex micro-structural nature and mainly as a result of miniaturization of components (radiation sources, radiation guide tubes, detectors), a number of portable and /or microscopic versions of established analytical methods have come into existence. Such methods are well suited for inspection and/or analysis of objects of great cultural value as measurements can be made on site (e.g., pigment identification in frescoes) thus eliminating the need to sample or even move the objects out of their normal surroundings (e.g., a museum or an archaeological site). Thus, most of today’s major museums have at their disposal one or more (trans)portable instruments for high spatial resolution examination of the artefacts in their collection. As well as to being useful for analysing the cultural heritage artefacts themselves, such techniques are also employed in support studies, where under controlled laboratory circumstances, e.g., the deterioration of building materials (mortar, sandstone, etc.) under the influence of rain, exposure to solar radiation, pollution gases, etc., is examined.
3
K. Janssens and R. Van Grieken
Although such methods have been described in the specialized literature of the field or sub-field to which they belong (e.g., portable and microscopic X-ray fluorescence analysis (XRF) in the atomic spectrometry literature, fibre-optic IR reflectography in the molecular spectroscopy literature), the flow of information towards the larger field of potential users has been limited because: † the technical literature that pertains to each of the sub-fields from which the techniques have emerged (physics, analytical chemistry, atomic/ molecular spectroscopy) is quite fragmented; † the potential users (archaeologists, art-historians, conservators, architects, museum conservators) are not sufficiently trained in the technical/ scientific aspects to allow them to absorb/monitor the above-mentioned state-of-the-art innovations. In view of the above, we have gathered together a group of authors who combine (a) a number of concise descriptions of non-destructive microanalytical techniques that have shown their effectiveness in the cultural heritage field with (b) a number of case studies where one or a series of artefacts of particular, generic nature (e.g., bronze statues, illuminated manuscripts, glass artefacts, etc.) are studied from the provenance or conservation point of view, preferably using an interdisciplinary, multitechnique approach. Among the chapters in the first section of this volume, the methods based on the use of energetic radiation belonging to the X-ray or Ro¨ntgen part of the electromagnetic spectrum are well represented. X-ray-based methods of non-destructive analysis are very frequently employed in the cultural heritage area for various purposes. Starting in earnest in the 1950s by judicious use of the available means then, this is still very much the case today where many technological advances increase the applicability of the methods, resulting in an extensive literature. This was also apparent during the seventh edition of the international conference on “Non-destructive Investigations and Micro-analysis for the Diagnostics and Conservation of the Cultural and Environmental Heritage” (Art 2002) that was organized in Antwerp (Belgium) by the editors of this book [5]. Of the ca. 200 papers presented at this conference, 115 included the use of analytical techniques for characterization of cultural heritage artefacts or materials. Among these contributions, 63 employed one or more X-ray-based or related technique (such as XRF analysis, scanning electron microscopy (SEM), transmission electron microscopy (TEM), protoninduced X-ray emission (PIXE), X-ray diffraction (XRD), X-ray absorption
4
Introduction and overview
spectroscopy (XAS), total-reflection X-ray fluorescence (TXRF) and X-ray microtomography) while 28 made use of IR spectroscopy, 14 used Raman microscopy, 12 used visible light spectrometry or reflectometry and six employed gas chromatography coupled to mass spectrometry (GC – MS). Thus, the first part of this volume can be considered to be the analytical reference section of the book, introducing and bringing together all technical information, library spectra (if any) and literature references of various methods. Each description consists of an explanation of the basic principles of the technique (keeping a cultural heritage user profile in mind), information on the availability of equipment, skills required, etc., and is illustrated with a few concrete examples of applications in the cultural heritage field. The second part of the book is intended to show the reader how these methods can be employed, either separately or in combination with each other, to solve concrete, real-world problems and provide overviews of the literature in specific application areas. Among these case studies, both investigations that aim to extract provenance information from cultural artefacts as well as studies that seek to evaluate and monitor preservation/ restoration treatments have been included. 1.2
OVERVIEW OF THE ANALYTICAL REFERENCE SECTION
The first section of the book starts with a overview (Chapter 2) of imaging and photographic techniques in the IR, visual, ultraviolet and X-ray part of the electromagnetic spectrum. Strictly speaking, the techniques discussed here are not microscopic in nature, but in practice will be very frequently employed prior to or in combination with some of the micro-beam methods that are discussed in later chapters. Chapter 3 is intended to introduce the reader to various forms of electron microscopy that are currently used to study cultural heritage materials. It outlines the principles underlying the technique and continues to discuss the role of electron microscopy in the field under study. The instrumentation, analytical possibilities and limitations of both SEM and TEM are discussed. An overview of recently published work involving the application of SEM and /or TEM analyses in the cultural heritage field concludes this chapter. In addition to X-ray radiography and tomography, discussed in Chapter 2, X-ray emission techniques are very frequently employed for non-destructive analysis of cultural heritage materials and artifacts. Most X-ray emission techniques involve irradiation of a material with a beam of X-ray photons
5
K. Janssens and R. Van Grieken
(or other particles), followed by detection of element-specific fluorescent radiation. In Chapter 4, besides a brief outline of the theoretical background of interactions between X-rays, and matter, the systematics of X-ray line spectra, instrumentation of X-ray spectrometry, the principles of quantitative analysis by means of XRF, and various specialized forms of this method that are of use in the cultural heritage sector are described, including totalreflection XRF, microscopic XRF and portable XRF for in situ investigations. Illustrative examples of their use are given. In addition, the use of more exotic X-ray techniques based on synchrotron radiation such as (microscopic) X-ray absorption near edge structure spectroscopy (m-XANES) and (microscopic) X-ray diffraction analysis (m-XRD) in combination with (m-)XRF are also mentioned. In Chapter 5, ion-beam methods of analysis (IBA) are discussed in detail. After a description of the fundamentals of the interaction of heavy charged particles with matter, attention is focused on the PIXE method and its companion techniques, particle-induced gamma emission (PIGE), Rutherford backscattering spectrometry (RBS) and elastic recoil detection analysis (ERDA). Specific attention is given to instrumentation devoted to the analysis of artistic and archaeological artifacts in so-called “external beam” facilities. Examples of the application of IBA techniques for surface and /or for bulk analysis of different materials are included in this chapter. In Chapter 6, the topic of surface analysis of materials by means of electron spectroscopy is discussed with a description of the theoretical background and the instrumentation for X-ray photo-electron spectroscopy (XPS) and (scanning) Auger spectrometry (SAM). The current state-of-theart of the method is described, with special attention to the lateral resolution currently achievable with both methods. This chapter also discusses sample preparation procedures and provides a review of SAM and XPS applications in the cultural heritage domain. Chapter 7 looks at the relatively new technique of laser ablation inductively coupled plasma mass spectrometry (LA-ICP-MS). The chapter begins with a short background showing the major steps leading to the development of both lasers and ICP mass spectrometers. This is followed by a more detailed technical discussion of a “generic instrument,” dealing separately with the ICP-MS and the laser. A section concerning the analytical technique describes how the two components work together. The reader is guided through appropriate sample preparation techniques towards taking the right choice in setting instrument parameters and dealing with the date produced. This chapter also concludes with a brief review of the relevant literature.
6
Introduction and overview
In the next chapter (Chapter 8) on IR and Raman microscopy (including fiber-optics Raman spectroscopy, FORS), first a resume´ of the requirements of IR, Raman spectroscopy and fiber-optic coupling in dealing with archaeological and art specimens is presented; this is followed by some case-type studies to illustrate the advantages and problems encountered in the experimental use of these techniques. Studies involving the combined use of XRD, SEM and GC – MS along with Raman microscopy are now becoming more frequent and some of these are discussed as examples. A description of secondary ion mass spectrometry or -microscopy (SIMS) can be found at the end of the reference section in Chapter 9. After some introductory considerations on the non-destructive nature of SIMS, and a description of the basic principles, the instrumentation for SIMS measurements is discussed together with an outline of the analytical procedures employed here. The quantitative, imaging and in-depth modes of operation are described and the chapter ends with a review of current applications in the cultural heritage sector, including the measurement of lead isotopes, determining the origin of gems and metal ingots and interface studies of coatings. 1.3
OVERVIEW OF THE CASE STUDIES SECTION
The second part of the book starts with Chapter 10, which is devoted to the analysis of corroded Cu-alloy materials, a very frequently encountered type of artistic or archaeological artefact. The combined use of methods of nondestructive analysis such as SEM, in situ XRD (with imaging system) and optical examination by bench microscopic methods for characterization of the alteration layers found at the surface of Cu-alloy objects is described. In situ examination of such artefacts by means of XRF in point-analysis and /or scanning mode is also discussed. The application of these types of methods is discussed for the examination of a gilded bronze Osiris from ancient Egypt, an important Greek-inscribed copper plaque of the 7th century BC , and small fragments of corroded copper alloy Phiale from the Southern Italian site of Francavilla Marittima, dated to the 5 – 6th century BC ; the contribution of analytical methods to the understanding of the development of patinas and their authenticity is evaluated. The use of bulk in situ XRD to study the corrosion of test coupons in assessing the museum environment is exemplified by examples of copper alloy coupons exposed to fixed amounts of formic and acetic acid pollutants. Chapter 11 extensively discusses the characterization of gold artifacts of various nature and the materials employed in various historical periods to
7
K. Janssens and R. Van Grieken
manufacture these items. The role of nuclear and atomic methods of nondestructive analysis of noble metals is described with emphasis on the ability of these methods to provide information on the bulk composition of the materials. The complementary use of RBS– PIXE and nuclear reaction analysis (NRA) for the compositional study of silver and gold artifacts is outlined, including the investigation of thick objects by means of gamma-ray transmission. Next, the study of the procedures of soldering on gold jewelry objects of different origin (Achemenide, Roman, Greek, Artesic and Etruscan artifacts) is treated, including the reproduction in laboratory circumstances of ancient soldering and brazing procedures. The discussion includes a comparison of modern analytical results with the description of ancient recipes listed in the Natural History of Pliny (1st century AD ). In this chapter, a study on the gilding of Mesoamerican tumbaga (a copper– gold alloy) artifacts is also presented. Through the application of diagnostic analytical methods for studying Byzantine iconography, the current state of preservation of works and the associated creative process of both wall-painting ensembles and portable icons may be revealed. In Chapter 12, the usefulness of this approach in the detailed evaluation of the aesthetic and historic value of the artworks is discussed. The use of stratigraphic data concerning structural painting materials and techniques as applied to two major works of art is described in detail: (a) the frescoes of the Protaton Church (end of the 13th century), attributed to Manuel Panselinos, who is considered to be the chief exponent and a legendary icon painter of the Macedonian School and (b) the icon of the Mother of God Hodegetria, Church of St Modestos, Kalamitsi, Chalkidiki, Greece, a representative portable icon of the 16th century Cretan style of Byzantine art. The artworks were examined through the application of both non-destructive and micro-analytical methods including, on the one hand, digital photography and macrophotography, stereomicroscopic photography, ultraviolet fluorescent photography, IR reflectography, X-radiography, image processing and colorimetry (measurement and representation) and, on the other hand, sampling of small fragments followed by optical microscopic observation and digital capture of cross-sections, under polarized white light and UV light, selective staining and micro-Raman and microscopic Fouriertransform IR (m-FTIR) spectroscopies of polished sections. Chapter 13 focuses on the advantages and disadvantages of analysing corroded silver coins by means of energy-dispersive (ED) XRF compared to SEM / EDX and PIXE, for the identification of the coin’s mint by its chemical composition. The objects of investigation are Austrian silver coins of the 12th and 15th centuries. As most of the coins were found in soil, where they have
8
Introduction and overview
been buried for hundreds of years, corrosion effects influence the qualitative as well as the quantitative results that can be obtained. The different information depths of all three techniques mentioned above are a main point of discussion. Comparison of the Ag-Ka/Ag-L count ratios of the EDXRF intensities with SEM / EDX measurements on cross-sections outlines the inefficiency of quantitative non-destructive analysis for the investigation of corroded objects. In Chapter 14, the identification and trace analysis of pigments encountered in illuminated manuscripts is described by means of a combination of micro-Raman spectroscopy and total-reflection XRF. First, the analysis of the various components of an illuminated manuscript, such as the inorganic pigments, the dyes and inks, the binding material and the parchment substrate, is discussed in general. In the second part, the analysis of manuscripts from the collection of Raphael de Mercatellis and of the “Brevarium Mayer van de Bergh,” an illuminated 16th century prayer book, is described. In the latter description, attention is given to a procedure that allows us to differentiate between the different workshops that were involved in the manufacture of this manuscript. In Chapter 15, different examples chosen from various historical contexts (from Protohistory to the Post-Medieval period) and from various geographical areas (mainly Europe and the Indian world) are used to illustrate the information provided by the chemical analysis of historical glasses for trade and provenance studies on one hand, and for understanding the development and the history of sciences and techniques on the other hand. A long time before artificial glass was invented, pre-historic populations were using obsidian, a natural glass, to make tools. The importance of trace analysis to understand and reconstruct the trade and exchange patterns of this material during Neolithic times is briefly illustrated. During the Bronze Age, artificial glass, as glass beads, was the object of long-distance trade. Relationships between the chemical composition of these objects, their chronology and the production area of the raw material can be used to build a distribution model of glass. In a similar manner, the Indian glass trade and manufacture at the beginning of our era is studied, with emphasis on the trade between India and the Mediterranean world. In order to investigate the European recipes used for glass making in Carolingian times, lead isotopes ratio analysis has been used to identify the birthplace of these recipes and follow the distribution of glass products through Europe. In the Post-Medieval period, the Venetian and fac¸on-de-Venice glass became very popular in various parts of Europe. By means of information on the major to trace composition, transfer of technology and recipes in 15 – 17th century Europe in various
9
K. Janssens and R. Van Grieken
stages can be followed, and the influence of the economic and political changes in the Low Countries on the local glass technology and workshop traditions is described. During the exposure of medieval stained glass used in window panes of cathedrals, churches and other historic buildings, weathering crusts consisting of gypsum (Ca2SO4·2H2O) and syngenite (K2SO4·CaSO4·H2O) are formed as crystalline corrosion products and mainly hydrated silica as amorphous material on the external surface of such glass objects. Chapter 16 discusses the manner in which these alternation layers can be characterized by means of several methods. After a discussion of the chemical exchange reactions that cause the weathering phenomena, analytical results obtained from specimens of medieval glass objects by SEM as well as SIMS and NRA are presented. Similar investigations were carried out on specimens from medieval enamel of the Burgundian Treasure of the Vienna Kunsthistorisches Museum. Additionally, the monitoring of in situ weathering tests, carried out on sample glass similar in chemical composition to medieval glass and enamel by means of atomic force microscopy (AFM), is described. In the final chapter (Chapter 17), the use of several analytical methods (together with their advantages and limitations) for the investigation and conservation of 16th century manuscripts endangered by iron-gall ink corrosion will be described. After a discussion of the fundamental chemical interaction that causes the mechanical strength of cellulosic materials to decrease dramatically due to the presence of Fe2þ in the ink, the results obtained by means of various analytical methods are described. This includes destructive and non-destructive investigations performed by SEM and electron probe micro-analysis, XRF spectrometry, inductively coupled plasma mass spectrometry, atomic absorption spectrometry, Mo¨ssbauer spectrometry and m-XANES. Also, the use of these techniques for optimization of a suitable conservation treatment for documents suffering from damage induced by iron-gall ink is described.
REFERENCES 1
2
10
Ch. Lahanier, G. Amsel, Ch. Heitz, M. Menu and H.H. Andersen, Proceedings of the International Workshop on Ion-Beam Analysis in the Arts and Archaeology, Pont-A-Mousson, Abbaye des Premontre´s, France, February 18 –20, 1985—Editorial, Nucl. Instr. Meth. Phys. Res., B14 (1986) 1. E. Ciliberto and G. Spoto (Eds.), Modern Analytical Methods in Art and Archaeology, Chemical Analysis Series, Vol. 155. Wiley, Chichester, 2000.
Introduction and overview 3 4 5
A.M. Pollard and C. Heron, Archaeological Chemistry. Royal Society of Chemistry, London, UK, 1996. D.C. Creagh and D.A. Bradley (Eds.), Radiation in Art and Archaeometry. Elsevier, Amsterdam, The Netherlands, 2000. R. Van Grieken, K. Janssens, L. Van ’t Dack and G. Meersman (Eds.), Proceedings of Art 2002: Seventh International Conference on Non-destructive Testing and Microanalysis for the Diagnostics and Conservation of the Cultural and Environmental Heritage, 2–6 June 2002, Congress Centre Elzenveld, Antwerp, Belgium, University of Antwerp (UIA), Antwerp, Belgium, 2002, 780 pp.
11
This page is intentionally left blank
Part I: Analytical Reference Section
This page is intentionally left blank
Chapter 2
UV-, IR- and X-ray imaging Franz Mairinger
2.1
SCIENTIFIC INVESTIGATIONS OF WORKS OF ARTS AND CRAFTS
Works of art are the result of mental processes in material shape. The realization of an artistic idea in fine arts is bound to a proper choice of an artistic technique and appropriate materials. Appearance, technique and material are closely connected. For example, the soft “sfumato” transition between different hues in baroque paintings can only be achieved with (drying) oil paints [1], whereas the precise rendering of finest details of an object (like furs, hair, jewellery) in gothic panel paintings is reserved for special distemper techniques [2]. Similar considerations hold for all other branches of arts and crafts. These considerations prove that any work of (traditional) fine arts needs a material base, so the tasks and aims of scientific examinations are easily defined: questions of the material nature in context with material history and the state of preservation of an object can be answered. This includes: – The analysis of utilized materials. – The investigation of artistic techniques. – The investigations of ageing processes of ancient materials and modern industrial products (used also in conservation). To answer such questions two types of methods are applied: area and point examinations. These terms are more or less self-explanatory. Area examinations cover the states of the total or greater parts of surface of an object (including macro- and microscopic investigations). This fact meets the holistic approach of art historians. Aims of these methods are to make invisible or imperceptible surface states or the inner structure of opaque objects visible to the naked eye. Infrared, UV and X-ray radiation can serve for this purpose. The methods are non-destructive, in a strict sense no Comprehensive Analytical Chemistry XLII Janssens and Van Grieken (Eds.) q 2004 Elsevier B.V. All rights reserved
15
Franz Mairinger
samples are taken. Reliable statements about the material composition of the object are possible only in special cases (e.g., in radiography). Questions of the material composition or layer structures are the domain of point examinations. Among the variety of modern instrumental techniques of analysis that are currently available [3,4] non-destructive ones are preferred. But to answer questions concerning the structure of complex paint layers or the chemical nature of binding media in paints, sampling is unavoidable, a procedure that is always destructive. 2.2
APPLICATION OF ELECTROMAGNETIC RADIATION FOR THE EXAMINATION OF CULTURAL HERITAGE OBJECTS
For area examinations of works of art, five regions of the electromagnetic spectrum are of special interest: – radiation in the visible range (400 – 780 nm) for colour and black and white documentation by photographic emulsion and digital photography. – near or long-wave ultraviolet radiation (320 –400 nm) for UV-fluorescence and reflected UV examinations. The use of middle (280– 320 nm) and short-wave (200 – 280 nm) ultraviolet radiation is only permitted in special cases (minerals, gems), since it is harmful for many art objects. – near infrared radiation (780– 3000 nm) for depth examination of paintings, objects of graphic art or textiles. – radiation in the intermediate (3 –6 mm) and far (6– 15 mm) IR are used in IR-thermography (a topic that will not be discussed here) which is useful for conservation of historic buildings and multispectral aerial surveying. – X-rays for the radiography of opaque (metallic and non-metallic) objects. They are produced by X-ray tubes. A wide range of tube voltages (5– 400 kV) is used, depending on the nature of the objects. – g-rays emitted by radioactive isotopes. They are distinguished from X-rays by their source, rather than by their nature and are mainly used for metallic objects. A disadvantage is that their intensity is not easily controlled. 2.3
INSTRUMENTAL BASIS
The generation of images of objects by visible and non-visible radiation and their registration is a complex process that involves many factors including – lighting (resp. irradiation); – imaging by optical systems;
16
UV-, IR- and X-ray imaging
– sensing; and – registration. In this chapter the instrumental basis of these four steps is discussed. 2.3.1
Light and radiation sources
The choice of the proper illumination resp. radiation source is of paramount importance for the imaging of objects by photographic and electronic means. The spectral output and physical intensity distribution of the illumination sources are important. Poor lighting design may obscure important details of the object. For the choice of a proper radiation source not only the spectral content of the source (broadband, narrowband, monochromatic) is significant; other questions are also involved: availability, costs, useful life time and also questions concerning the optical components: spectral sensitivity of the sensor, blooming, lag, drift or noise of the sensor at high or low lighting levels, limits of lighting power level for objects sensitive to light or heat and in case of UV-examinations: auto-fluorescence of any optical component. In the following sections, a short discussion of the various sources will be given. 2.3.1.1 Ultraviolet radiation sources There are many light sources that emit UV radiation, but not all of them can be used for UV examinations. Sunlight contains a fairly small portion of long and middle wave UV, as compared to the amount of visible radiation. The UV intensity depends on atmospheric and seasonal conditions and is therefore not reliable. Incandescent lamps (tungsten, tungsten – halogen) need not to be considered at all since they only produce a small quantity of ultraviolet radiation. Only wire-filled flash lamps are suitable for reflected ultraviolet photography. A continuous arc discharge between carbon or metallic (Fe) electrodes provide strong UV-emissions, but it is difficult to maintain a constant output. These sources require an enclosure in a light tight box and produce harmful vapours and heat. Gas discharge lamps are at the present state of development the most preferable sources of UV. Mercury high- and low-pressure lamps, fluorescent tubes, xenon arcs, electronic flash lamps and metal halide lamps belong to this group.
17
Franz Mairinger
All gas discharge lamps have a so-called falling characteristic, i.e., their internal electrical resistance decreases sharply with rising temperature, so that the current flow increases rapidly; without protection the lamp is destroyed within fractions of a second. Therefore, they can be operated only with a ballast; chokes are used for this purpose (and starting gear). Nearly all of these sources also emit visible radiation with high intensity, so that the use of appropriate lamp and/or camera filters is obligatory. If these UV sources are used with reflectors, it should be kept in mind that the average reflectivity of silver coatings rapidly drops off at wavelength shorter than 400 nm (9% reflectivity at 320 nm), whereas bare aluminium (and also rhodium) retains its high reflectance down to 200 nm.
Mercury vapour lamps Mercury vapour emits in an electric discharge or arc a large number of extremely bright spectral lines in the visible and ultraviolet region (Fig. 2.1). The intensity distribution is strongly dependent on vapour pressure and temperature. Two types of mercury vapour discharge lamps are commercially available: high-pressure (long-wave) and low-pressure (short-wave) lamps. Such lamps are produced in power ratings of a few watts to many kilowatts with pressure loadings from 1 mm Hg to 200 bar. Most manufacturers produce both types. In the following sections the production palette of Philips is used as representative example.
Fig. 2.1. Spectral lines of mercury vapour.
18
UV-, IR- and X-ray imaging
High-pressure lamps High-pressure lamps consist of a small quartz tube in which mercury vapour is generated under a high pressure. An outer envelope made of dark Wood glass, which absorbs the visible radiation, surrounds the burner. The main output in the UV is at 366.3 nm, some middle (334 nm) and short wave (313 nm) ultraviolet is also emitted; 20% of the total emission consist of a continuous spectrum. As an example for this lamp-type the PHILIPS HPW (125 W) is chosen. Its spectral power distribution is shown in Fig. 2.2b. The HPW lamp has the common E 27 base (Edison thread) similar to domestic lamps. A ballast gear (choke) is required for operation; there are no restrictions in burning position; a warm-up period of several minutes is required for full intensity. When the lamp is turned off, re-ignition is possible only after a longer cooling period. The average useful lifetime is 6000 h. This lamp type is advantageous in illuminating smaller objects with high ultraviolet brightness. Low-pressure lamps and fluorescent tubes The peak output of low-pressure mercury vapour lamps is in the short-wave UV region at 253.7 nm. This radiation excites a strong fluorescence in some minerals and gems, has a strong germicidal action and a very painful effect on the eyes and the human skin (use of goggles and gloves is recommended).
Fig. 2.2. Spectral power distribution of (a) a Philips HPW lamp and (b) a Philips TL(D)/08 40W “black light” fluorescent tube.
19
Franz Mairinger
This radiation should not be used for the examination of light sensitive art objects. It is not transmitted by glass camera lenses. Philips produces TUV germicidal lamps with power ratings from 4 to 30 W. The UV emission of such sources can be modified by special phosphors. This is accomplished in fluorescent tubes. The wall of the tube is internally coated with an inorganic fluorescent powder, which converts the short-wave radiation into a long-wave UV radiation in the range of 320 –400 nm with a peak emission at 350 nm. Their spectral power distribution is shown in Fig. 2.2a. These “black light” tubes can be obtained in various lengths from 14 to 122 cm. They can be operated in standard fluorescent light fixtures with standard starter and choke. They are used to illuminate large areas with UV radiation. Their useful lifetime is over 2000 h. No eye protection is necessary. An example of this type is the Philips TL(D)/08 (40 W) “black light” fluorescent tube; it has a dark-coloured envelope which transmits the UV-A radiation and a small amount of visible light in the violet region. Electronic flash lamps and xenon arc lamps Electronic flash lamps are primarily designed for photographic work within the visible range, but they provide also a fairly intense emission in the near UV region. They contain a mixture of noble gases like xenon, krypton and argon and the actual UV output depends on the composition of this mixture. Some tubes are coated with a yellowish lacquer layer to give a better colour rendition by absorbing the UV part. These types are less suitable for UV photography. For UV examinations, the visible part of the emission must be excluded by using an appropriate short-pass filter (e.g., Wratten #18, or Schott UG 2) in front of the flash lamp. A problem in using electronic flash lamps is that since fluorescence cannot be observed during the very brief flash-time, a preliminary inspection with a continuous source is necessary. Nevertheless, this source is quite convenient for routine UV and fluorescence work, after working conditions have been established by test exposures. For feeble fluorescence the “multiflash-technique” is a convenient method: the room is darkened, the shutter of the camera is opened and a series of flashes is fired. High-pressure, continuous xenon arc lamps produce a sun-like emission with a nearly continuous spectrum encompassing the ultraviolet, visible and near infrared regions, as the diagram of the spectral power distribution (Fig. 2.3) indicates.
20
UV-, IR- and X-ray imaging
Fig. 2.3. Spectral power distribution of xenon arc lamps.
They are used as a standard light source for colour measurements. Disadvantages are the high cost of the lamps and ballast gears and the lower UV efficiency compared with mercury vapour lamps. 2.3.1.2 Infrared radiation sources All hot bodies emit infrared radiation in a continuous spectrum. So all types of incandescent lamps (tungsten, tungsten –halogen) are excellent infrared radiation sources. At 3000 K, the filament temperature of a common 100 W lamp, the peak emission is at 950 nm. For infrared examinations it is not necessary to operate the bulbs at full mains voltage; a (small) voltage reduction will protect thermal sensitive objects from unnecessary radiant heat. It has been already mentioned that electronic flash tubes are also excellent sources for infrared photography. The above-mentioned “multiflash-technique” can be applied and prevents excessive heat strains for sensitive objects. Narrow-band, but feeble, incoherent sources are IR transmitting diodes, which are used in remote controls for electronic gears and for optical communications (in optical fibres). Their wavelength range lies between 800 and 1550 nm. It is fairly easy to build a homemade diode array for the illumination of medium-sized objects. Sony makes use of such diodes for the built-in IR illumination source for the “night-shot” option in their digital still cameras and camcorders. Since the bandwidth is small, about 10 MHz, IR transmitting diodes can be used in imaging spectroscopy of small object areas as a cheap replacement of expensive narrow-band interference filters.
21
Franz Mairinger
2.3.1.3 X-ray and gamma sources X-ray sources The mechanism of production of X-rays is well known, so it will be discussed only briefly: they are generated in a tube by accelerating electrons onto a target material (see Chapter 4 for more details). The character and intensity of the generated polychromatic radiation (bremsstrahlung) depend on the applied tube voltage. The total X-ray intensity of the continuous radiation Icont is given by: Icont ¼ AiZV m
ð2:1Þ
where A is a proportionality constant, i, the tube current, Z, the atomic number of the tube anode, V, the tube voltage and m, a constant with a value between 2 and 5, depending on the voltage and the filter type used. Since for maximum yield, Z should be as high as possible, for radiography the high melting tungsten (W) is the best choice. There are many X-ray generators commercially available but the choice of an appropriate machine affords consideration, it depends on the type of work to be done [5,6]. An X-ray tube needs for operation two power supplies: a (low voltage) filament supply and a high-voltage anode supply. The tube current is controlled by temperature changes of the filament. This is usually accomplished by a variable-voltage transformer which energizes the primary of the filament transformer. The high-voltage supply consists of a step-up transformer, and an autotransformer for the adjustment of the primary voltage of the former and for fixed machines a rectifier. Some industrial X-ray tubes are designed for the direct application of an AC high-voltage (50 or 60 Hz); in this case, the X-ray tube acts as its own rectifier since the tube current only flows during the positive half-period. Modern mobile X-ray machines generate their own AC voltages with higher frequencies than 50 (or 60) Hz; their X-ray output is much higher than that of 50 Hz half-wave machines. Seifert (Agfa) produces such middle-frequency generators up to 260 kV. The various machines can be roughly classified according to their maximum voltages. But it should be noted, that it is impossible to cover a voltage range from 5 to 300 kV with one machine. There is not only an upper voltage limit, but also lower one. The efficiency of a 300 kV generator at a tube voltage of 20 kV is near zero. This is caused by the penetration factor (transparency): at low voltages a dense cloud of electrons (space charge) is formed around the cathode, which cannot be drawn away by the anode, so that no additional electrons can leave the cathode.
22
UV-, IR- and X-ray imaging
For that reason two different machines must be acquired if radiographs of objects of low absorption (e.g. organic materials such as paper or parchment) and high absorption (metal items of greater thickness) should be made. For the best definition in a radiograph the size of the focal spot of the X-ray tube is an important factor. To minimize the geometric unsharpness, it should be as small as possible. Fine-focus tubes usually have focal spots of 0.5 £ 0.5 – 1 £ 1 mm2. These small areas are only possible at voltages up to 100 kV. There are also micro-focus tubes with spot sizes down to 8 £ 8 mm2 available. Since nearly 98% of the electric energy applied to the tube is converted into heat, the anode resp. the target must be cooled; oil, water or air are used as coolants. The continuous operation time of oil-cooled tubes is normally limited to a few minutes, than a cooling period of twice the operation time is necessary. An other essential point is the material of the tube window; it should have a low absorbance for soft X-rays and a high mechanical rigidity. Thin (0.5 mm) beryllium sheets are the best choice. This is important for a high contrast in radiographs of object with low absorption. Tubes with a voltage range up to 300 kV have to withstand large loads, have much larger focus spots (2.5 £ 2.5 – 6 £ 6 mm2). In this case long source-film distances will aid in showing finer details (with increased exposure times). Table 2.1 is meant TABLE 2.1 Typical X-ray machines Maximum voltage (kV)
Type of objects and thickness limits
Screens
Focal point size
5–50; 8–100
Paper, parchment, textiles paintings on canvas and wood, wooden sculptures, thin metal sheets (iron, brass, bronze) Iron (up to 40 mm), bronze (30–35 mm), gold (,0.5 mm) objects, rock (marble up to 150–200 mm) Iron (50 –60 mm), bronze (30–40 mm), gold (0.6 mm) objects Iron (up to 80 mm), bronze (50–60 mm)
None
0.1 £ 0.1 –1 £ 1 mm2
Lead foil fluorescent
2.5 £ 2.5 mm2
(25–) 200
260 –300
400
Lead foil
Lead foil
23
Franz Mairinger
to be a rough guide for X-ray machines; it lists voltage ranges and some possible applications. Gamma-ray sources Radiography with gamma rays has several advantages. The radioactive sources are simple and compact and have no need for external power. The latter is important when objects such as monumental outdoor bronze sculptures are to be examined. In contrast to X-ray machines that emit a broad continuous band of bremsstrahlung, gamma ray sources emit one or a few gamma lines with discrete wavelengths, i.e., they are equivalent to monochromatic radiation sources. Their energy is given in kilo- or megaelectron volts (keV, MeV). There is no defined correspondence between monochromatic and polychromatic radiation. As a rule of thumb: the penetration of monochromatic radiation expressed in keV corresponds approximately to that of polychromatic radiation expressed in kV multiplied by 2. So the monochromatic radiation of 60Co at 1200 keV will have similar penetration properties as that of polychromatic radiation emitted by an X-ray tube operated at 2400 kV. Compared to X-ray tubes the intensity of gamma-ray sources is rather small, so that long exposure times are required. As can be seen in Table 2.2, quite a few radioactive isotopes with lines of very different energies can be used in gamma radiography. Gamma-ray sources lose activity according to their half-lives; this necessitates more or less frequent adaptation of the exposure time. Another disadvantage is the considerable hardness of their radiation which causes a rather low subject and film contrast. TABLE 2.2 Radioactive isotopes used in gamma radiography Radioactive element
Symbol
Half-life
Specific g-radiation constant (R m2/h Ci)
Energy in MeV (number of lines)
Half-valuelayer in lead (mm)
Caesium 137 Cobalt 60 Iridium 192 Selen 75 Thulium 170 Ytterbium 169
137
30 a 5.3 a 74 d 118.5 d 128 d 32 d
0.35 1.30 0.48 0.203 0.0025 0.125
0.66(1) 1.17–1.33(2) 0.3 –0.6(,10) 0.066–0.4(9) 0.052–0.084 0.063–0.308
8.4 13 2.8
24
Cs Co 192 Ir 75 Se 170 Tm 169 Yb 60
0.88
UV-, IR- and X-ray imaging
2.3.2
Imaging
In an abstract sense optical images are non-uniform 2D patterns of brightness. In the ultraviolet, visible and infrared regions of the spectrum, these patterns are generated by lens systems, bearing in mind that common camera lenses are corrected for the visible spectrum. For registration, the images are projected on the surface of a sensor. The latter can be a photographic emulsion or an electronic detector with sensitivity to radiation of the appropriate wavelength range. 2.3.2.1 UV optics UV illumination is used for surface investigations of works of art in two different ways: for reflected UV-photography and for exciting UV-fluorescence. Photographic documentation of the latter presents no problems or restrictions since the fluorescence is in the visible range. Difficulties are associated with the recording of UV radiation reflected by an object. Optical glasses, depending on their composition, feature a fairly low transmission for wavelengths shorter then 335 – 360 nm. Thus, most custom cameras can be utilized for reflected UV photography with the 365 nm Hg line. Only modern high-refracting rare-earth glasses (e.g. lanthanum flint) have a transmission that drops sharply already at a wavelength a little shorter than 400 nm. Other difficulties may arise from the anti-reflection coating and the cement of modern lenses. Some of them exhibit quite a strong fluorescence when irradiated by UV. They produce glare light that deteriorates the contrast of the image. There is a rule of thumb for checking the antireflection coating: in daylight the lens surface appears always slightly coloured. Lenses with bluish or purple reflections exhibit normally less fluorescence than tinges of green, brown or amber. Dust particles also exhibit fluorescence, so that the lens surface should be cleaned thoroughly. Multi-layer anti-reflection coatings suppress reflections only in a relatively narrow band of wavelengths. With camera lenses the coating is usually optimized for green light; for UV radiation the (internal) reflectance may increase considerably. A further point is that the chromatic aberrations of camera lenses are corrected for the visible range; in the UV region their focal length is somewhat shorter. Since the focusing is done in the visible region, however, the UV image may be slightly blurred. Decreasing the lens aperture can compensate for this.
25
Franz Mairinger
In general, before a commercial camera is used for UV-examinations, test exposures are necessary. For short-wave UV documentations down to 200 nm, true UV lenses must be employed. They are made of fused quartz or synthetic fluorite, are corrected in the range between 220 and 1000 nm and are fairly expensive. Nikon produces a UV-Nikkor 4.4/105 mm lens and Zeiss a UV-Sonar 4.3/105 mm lens. 2.3.2.2 IR optics In contrast to the UV region most optical glasses and optical cements transmit near IR up to 2600 nm freely. Only for the long-wave IR radiation, materials such as fused silica (185 nm – 4 mm with an absorption peak at 2.7 mm), germanium (1.8 – 25 mm) or sapphire (150 nm – 6 mm) must be utilized. IR thermography cameras have silicon lenses that are opaque to visible light. Similar to the long-wave UV, any camera lenses that work in the visible will work satisfactorily in the 1000 – 2500 nm band if some precautions are observed that were in part discussed in Section 2.3.2.1. Since the focal length f of a lens increases with increasing wavelength by a factor of f/200 – f/300, an image focused in the visible is slightly out of focus in the IR range. In infrared photography this can be reduced by focusing through a (deep) red filter or by using a smaller aperture, but it should be remembered, that diffraction effects increase with wavelength and plays a significant role in the IR region. Some 35 mm camera lenses have special infrared markings (red dot or line), but these are only helpful up to 1000 nm. In general the resolving power of camera lenses decreases in the infrared, since the spherical and chromatic aberrations increase. Lenses with apochromatic correction have a better IR performance than achromatic ones. Very often coated lenses feature a reduced transmission in the near IR; e.g., the Nikon Micro-Nikkor 105 mm lens transmit in region between 750 and 1000 nm only 50% of the intensity at 600 nm [7]. All these effects and the penetration of IR radiation into the surface of many objects can contribute to the often observed blurring of the IR images. 2.3.2.3 Filters In this section the various types of filtering devices for the infrared, ultraviolet and X-ray region are discussed. For the photographic and electronic recording of invisible object states, filters must be applied between object and recording device since most of the latter are sensitive to unwanted spectral regions, in
26
UV-, IR- and X-ray imaging
order to avoid a superposition of the various images. The use of filters in X-ray radiography is a special field, which is also discussed. Filters are passive devices; they can attenuate light resp. radiation wavelength-invariant (neutral-density filters) or wavelength-selective by absorption, interference, or selective scattering. They can be utilized in front of camera lenses as part of the imaging system or in front of light (radiation) sources. Different materials are employed: – Dyed gelatine filters can be used for image-forming work as well as for lighting (at low power levels) in the UV, visible and infrared region. The attenuation occurs by selective absorption. They are the least expensive filter type, but are easily scratched, sensitive to moisture, fingerprints and temperatures above 50 8C. Kodak (Wratten) produces this filter type for a wide spectral range. Coloured plastic filters have similar properties but have more defects. – Ion-coloured glass filters consist of a solid solution of inorganic salts or elements in a glass matrix. They have a much higher scratch and temperature resistance; therefore, they can be utilized in front of radiation sources at high power levels. Such filters are manufactured, e.g., by Schott [8] and Corning. Also available are glass filters that attenuate radiation by colloidal scatter; they are produced by controlled annealing. During this process microcrystalline nodes are formed, which scatter and absorb certain spectral regions. Ruby glass is an example of this type of filter. – Interference filters are produced by carefully controlled vacuum deposition of non-conducting transparent materials with different indices of refraction or metallic layers on top of a glass substrate (up to 20 layers). They are used as narrow-band or edge filters. – Infrared mirrors are a special type of interference devices. They have a wavelength-sensitive surface. There are IR-reflecting or “hot” mirrors: they reflect infrared (heat) and transmit most of the visible spectrum. Their counterpart are IR-transmitting or “cold” mirrors: they transmit IR (800 nm – 2,5 mm) and reflect the visible part of the radiation. They are used effectively for lighting purposes. As far as the spectral behaviour is concerned, a distinction between the following filter types can be made: – Band-pass filters have a single transmittance band flanked by two rejection bands.
27
Franz Mairinger
– Edge filters have a more or less abrupt border between a region of high transmission and an area of rejection. Filters based on absorption and on interference are used for this purpose. When the transmission occurs in the region of shorter wavelengths, the device is designated as a short-pass filter. The counterpart, who transmits longer wavelengths and excludes shorter wavelength is called a long-pass filter. – Narrow-band interference and line filters isolate very narrow bands (down to a few nm) of radiation. They are used for critical applications such as imaging spectroscopy. – Dichroic filters are used when the transmittance band corresponds to a spectral band within the visible spectrum. The same filter has a different colour whether the light is transmitted or reflected. Such a filter can pass, e.g. the blue band and reflects the yellow band (additive mixing of red and green light). Sets of such filters are used for colour separation (RGB or CMYK) or image segmentation work. Filters for UV-examinations In Section 2.4.1 it will be discussed in detail that there are two different methods of using UV-radiation for examinations: the reflected-UV-method, analogous to ordinary photography, and the fluorescent-light-method where objects are irradiated by UV and emit visible fluorescent light. Two different types of filtration are to be applied. 1. For the reflected-UV-method, short-pass filters resp. band-pass-filters are utilized. They transmit (long-wave) UV radiation and reject visible light. They are also called exciter filter and are applied either in front of UV sources or in front of the camera lenses. In the latter case the filter must be well polished. For the long-wave UV these filters are made of a special barium-sodium-silicate glass tinted with 9% nickel oxide (called Wood glass). A filter of this type is incorporated in the HPW-lamps of Philips. For short-wave applications special tinted quartz filters are available (e.g. Schott UG 5). Data of some UV exciter-filters are shown in Table 2.3. Figure 2.4 shows the transmittance curve of the Schott UG1 filter that is especially suitable for the selection of the 365 nm Hg-line; this filter transmits also in the infrared region, so it could be used as a long-pass filter for IR examinations. This IR transmission may be suppressed by adding an appropriate band-pass filter such as the Schott BG 39 heat reflection filter. 2. For the fluorescent-light-method, a long-pass filter, also called barrier filter, is placed in front of the camera lenses. It absorbs UV that is
28
TABLE 2.3
Manufacturer
Designation
Filter-type
Kodak Corning Glass #5840 Corning Glass 9863 Schott
18 A (2 mm) CS7-60 (2 mm) CS7-54 (5 mm) UG 1
Band-pass, Band-pass, Band-pass, Band-pass,
Schott Schott
Transmission band (nm)
Remarks
glass glass glass glass
310 –400 310 –400 250 –380 310 –400
UG 5
Band-pass, ion-coloured glass
240 –480
UG 11
Band-pass, ion-coloured glass
260 –390
Transmits IR Transmits IR Transmits IR Transmits IR (710–850 nm and 2.4 – 4.4 mm) Transmits IR (660 nm –2.7 mm), for 254 nm Hg-line Transmits IR (690–750 nm)
ion-coloured ion-coloured ion-coloured ion-coloured
UV-, IR- and X-ray imaging
Exciter filters for UV-fluorescence
29
Franz Mairinger
Fig. 2.4. Transmittance of Schott UG 1 long-wave UV exciter filter (1 mm). T is transmission; R, transmission without reflection losses.
reflected or scattered by the object. The selection of the appropriate barrier filter depends on the colour of the excited fluorescent light and on the spectral purity of the UV-source. Gelatine filters with an edge near 410 – 420 nm are normally utilized. For critical applications glass – plastic compound filters (e.g., Schott KV 418) are the better choice. Table 2.4 lists characteristics of a selection of long-pass barrier filters. Some of the ioncoloured glass filters exhibit intrinsic fluorescence, when irradiated directly by UV. This can cause blurred images. If there is no violet or bluish fluorescence, any cheap light yellow gelatine filter (e.g., Wratten #9 or 11) can serve for this purpose. It should also be kept in mind that thick glass filters are a special type of optical flat. Since they displace a light ray laterally (image shift) without changing its direction (dependent on the filter thickness and incidence angle), a degradation of the image quality can occur.
Filters for infrared examinations For infrared examinations and documentation, the visible part of the spectrum must be excluded, since some infrared imaging devices (including infrared sensitive films) are sensitive to light. Without a filter the visual and the infrared image would be superimposed and the result would resemble a normal panchromatic recording. A long-pass filter that absorbs the visible radiation must be used in front of the camera lenses. These filters have a
30
TABLE 2.4 Barrier (long-pass) filters for photographic recording of UV-fluorescence Designation
Filter-type, material
Cut-on wavelength (nm)
Remarks
Kodak Kodak Kodak Kodak Kodak Schott
Wratten Wratten Wratten Wratten Wratten GG 420
395 410 420 480 510 420
Pale yellow Pale yellow Pale yellow Yellow Deep yellow Pale yellow slight intrinsic fluorescence
Schott
GG 495
495
Yellow, slight intrinsic fluorescence
Schott Schott Schott
LP 400 LP 430 KV 408
400 430 408
Interference filter Interference filter Free of fluorescence
Schott
KV 418
Long-pass, gelatine Long-pass, gelatine Long-pass, gelatine Long-pass, gelatine Long-pass, gelatine Long-pass, ion-coloured glass, annealed Long-pass, ion-coloured glass, annealed Long-pass Long-pass Long-pass, glass – plastic compound Long-pass, glass – plastic compound
418
Free of fluorescence
2B 2A 2E 9 12
UV-, IR- and X-ray imaging
Manufacturer
31
Franz Mairinger
deep red or black appearance. In an IR photography the cut-on wavelength should correspond with the sensitization of the emulsion. The Kodak Wratten filters #87 and 87 C are perfectly suited for this purpose. But even red filters (Wratten 25) can be utilized. Table 2.5 shows the data of some suitable filters. For infrared imaging spectroscopy narrow-band interference filters with a bandwidth of 10 – 15 nm are used. These (expensive) filters are commercially available in the range between 700 and 1550 nm in 10 nm steps by Schott and Electrophysics. For heat sensitive objects such as panel paintings, this technique has a severe disadvantage; high-power sources are needed for lighting, since only a small amount of the total radiant energy passes these filters. X-ray filters In radiography with X-rays, filters are used to reduce excessive subject contrast (and hence radiographic contrast) by hardening the radiation. Although in most cases the highest possible contrast is desired, there are certain instances where too much contrast is a definite disadvantage. This holds for metallic specimens having a wide variation in thickness or for paintings on canvas with a priming consisting of substrate layers containing lead white onto which thin paint layers were applied; with unfiltered radiation the radiograph would show in this case only the texture of the weave. Figure 2.5 graphically illustrates the process of filtering. The longer wavelengths (softer radiation) do not penetrate the filter so that the beam emerging from the filter has a higher portion of the more penetrating wavelengths (harder radiation). Overall, the total intensity of radiation is TABLE 2.5 Infrared long-pass filters Manufacturer
Designation
Cut-on wavelength
Remarks
Kodak Wratten Kodak Wratten Kodak Wratten Electrophysics Electrophysics Schott Schott Schneider (B þ W) Schneider (B þ W)
#87 #87 C #88 A LPF 750, LPF 800 LPF 1000, LPF 1500 RG 780 (3 mm) RG 1000 (3 mm) #092 (¼ RG830) #093 (¼ RG1000)
740 nm 800 nm 730 nm 750 nm, 800 nm 1000 nm, 1500 nm 780 nm 1000 nm (1 mm) 830 nm 1000 nm
Gelatine Gelatine Gelatine Glass Glass Glass, annealed Glass Glass, annealed Glass
32
UV-, IR- and X-ray imaging
Fig. 2.5. Effect of a filter in front of an X-ray source. The longer wavelengths are removed; the overall intensity of the beam is reduced.
reduced. Since such filters remove most of the wavelengths that would not be able penetrate the thicker portions of the specimen, overexposure of the thinner parts is avoided and scattering is reduced. Filtering is in fact analogous to an increase of the tube voltage, which causes a decrease in contrast. The net effect is determined by the nature of the individual specimen. The choice of the filter material depends on the tube voltage. For voltages up to 50 kV, a range that is used for radiography of non-metallic specimens (wood sculptures, paintings), aluminium filters of varying thickness can be employed; copper, brass and lead foils are utilized at higher voltages. A lead foil, mounted in close contact to the film will reduce not only the effect of scattered radiation from all sources on the film; at voltages above 100 kV, it also acts as an intensifying screen that increases the photographic action on the film by the emission of photo-electrons. 2.3.3
Sensor systems
In the previous sections the basic principles of lighting and imaging by lenses in the UV, visible and IR bands were discussed. This section is concerned with the registration of these visible and non-visible images projected onto the surface of a photon detector with finite area. The detection is a threshold
33
Franz Mairinger
process: any photon of energy greater than the threshold value will give rise to a detectable signal. There are two main groups of photodetectors: – photographic emulsions and – electro-optic image sensors. In photographic emulsions absorbed photons cause silver halide grains to become developable to metallic silver. Films can be utilized in the X-ray, UV, visible and near-infrared bands. Electro-optic devices rely on the photoelectric effect. The photon images, which have been focused onto the photosensitive surface, are converted by a vacuum tube or by a solid-state sensing array to electron images. These electronic images are reconverted to visible images. 2.3.3.1 Photographic materials The basis of common photography is the decomposition of the silver halides AgCl, AgBr and AgI by photons of appropriate energy to metallic silver. For photographic films the use of silver bromide suspended in gelatine as protective colloid greatly predominates. The spectral sensitivity of silver bromide ranges from the shortest X-ray wavelengths to the blue region of the visible spectrum. With the addition of some silver iodide the sensitivity is further extended into the blue-green. Yellow or red light is not sufficiently energetic to activate AgBr, although the heat of formation of silver bromide is DHf ¼ 2100:44 kJ; which would correspond to a luminosity factor of photons of a wavelength at 1390 nm. The yellowish colour of silver bromide indicates an absorption in the blue part of the visible region only. Vogel discovered in 1873 that certain dyestuffs, when added to photographic emulsion, extended the sensitivity to longer wavelengths. Research for sensitizing dyes stimulated this discovery and in the 1930s the first colour films and infrared-sensitive emulsions became commercially available. The theoretical sensitivity limit of 1350 nm was achieved with special dyestuffs of the cyanine-group. Photographic emulsions are still the standard of comparison for many other detector devices in terms of sensitivity (1– 10 photons), dynamic range (106) and resolution (pixel size 10 mm). In the following sections the properties of photographic materials for the different regions of the spectrum are discussed. X-ray films The use of X-ray films in radiography is still a very appropriate way to obtain and store a maximum of information. Modern X-ray films consist of a flexible
34
UV-, IR- and X-ray imaging
blue-tinted cellulose triacetate base, coated on both sides with thin (,20 – 25 mm) gelatin layers containing fine-grained silver halides in high concentration. This measure is due to the fact that X-rays are absorbed in such thin layers only to a very small amount; this holds especially for hard radiation. Thus the application of two layers increases film speed. When highest detail visibility is essential (such as in microradiography), films coated only on one side of the base are used. There are two types of X-ray films: medical and industrial. Medical films have a high speed (thick emulsion layers) to keep the necessary radiation dose for exposure low, but have much lower resolution than the low-speed industrial films. There are three different forms of packaging for X-ray sheet films: – sheet films: the sheets are interlaced or non-interlaced and must be loaded in cassettes or film holders. The blank films are used (in a darkened room) for radiography at very low tube voltages and for electron radiography; – envelope (day – light) packing: each sheet is enclosed in a light tight packing, having the advantage that no cassettes or film holders are necessary. – envelope (vacuum) packing between two fluoro-metallic foils, mostly containing lead (of 27 mm thickness, when manufactured by Agfa) or lead oxide foils (when manufactured by Kodak) for tube voltages .120 kV. The foils act as intensifier screens and as protection against scattered radiation. These films are available at different speeds; a lower speed corresponds to a higher resolution. Roll films (with widths up to 1 m and length up to 10 m) and large format X-ray sensitive paper are also available.
Films for reflected-UV-photography All photographic materials are sensitive to UV radiation down to as far as 230 nm, but the sensitivity of photographic emulsions is in most cases much lower than the rated speed, because the gelatin binder and the overcoat absorb UV, so that test exposures are necessary. Below 230 nm the speed is further reduced drastically [9]. For long-wavelength UV photography with common glass lenses, any black and white ortho- or panchromatic sensitized film can be utilized; the use of colour film has no advantage.
35
Franz Mairinger
Black and white and colour films for the visible wavelength range For the visible range, photographic materials of different sensitization are available. Unsensitized, orthochromatic (red-blind), panchromatic colournegative and colour-reversal films in all formats and sizes are commercially available. For the registration of UV-fluorescence in the visible wavelength range, black-and-white or colour films can be utilized, but colour films are the first choice, because they allow for the visual differentiation between colours of equal brightness but different hue that would be rendered by means of identical grey tones.
Infrared emulsions Two decades ago a great variety of infrared films with a sensitivity range between 850 and 1350 nm were commercially available. With the advent of electronic IR-sensors the assortment has been reduced to a few emulsions listed in Table 2.6. A material worthy of note is the Kodak Ektachrome Infrared Film (Type 2236). It is a 35 mm false colour daylight type reversal film, with one of the three layers sensitized to infrared up to about 900 nm. It was mainly used for aerial cameras to detect camouflage of military objects, in various crop surveys for mapping infected areas and in archaeology for photoarchaeological records. The film is very contrast-rich. It has three layers sensitized to green, red and infrared. Since all of them are sensitive to blue, use of a yellow filter (e.g., Wratten #12) is obligatory. Upon processing of the green sensitive layer, a positive yellow image is formed. The red sensitive layer yields a positive magenta and the infrared a positive cyan image. This film adds an infrared component to the visible record; thus, visually equal or similar colours are rendered in different hues, depending on their infrared behaviour. Table 2.7 lists the colour rendition modified by IR reflection of the object. It is also possible to sensitize common black-and white films to IR by immersing them in a highly diluted solution of IR-sensitizers such as xenocyanine followed by a rapid drying. The procedure is described in detail by Bru¨gel [10].
2.3.3.2 Electronic imaging detectors The input image for electronic imaging detectors may be formed in any band of the electromagnetic spectrum from the UV to the far-IR region.
36
TABLE 2.6 Commercial infrared films Kodak infrared high-speed type 4143
Kodak Ektachrome infrared type 2236
Konica infrared 750 black and white film
Film format
35 mm, Sheets 4 £ 500 ,þ138C or deep freeze With Wratten # 87C 200/248 (Incandescent l.3400 K) 300 –920 nm (,820 nm)
EU: 135, USA: 135, 4 £ 500 ,þ138C or deep freeze With Wratten #12: 100/218 (electronic flash lamps) 360–900 nm (,750 nm)
35 mm, rollfilm
Daylight, electronic flash, tungsten Wratten #25, 29, 70 (red filters) IR: 87, 88A, 87C 1 s (1/2 stop)
Daylight, electronic flash, tungsten Flash: Wratten #12 Tungsten: Wratten #12 þ CC20C þ Schott BG22 1/10 s (1 stop)
1 s (?)
D 76, D19, HC 110
Kodak Laboratories
D 76, DK 20, ID19
Storage Speed (ISO) Spectral sensitivity (maximum IR) Recommended light source Recommended filters
Reciprocity failure starts at (correction) Development
,þ138 or deep freeze With Wratten #25 32/168 (daylight) 400 –820 nm (750 nm) Daylight, electronic flash, tungsten Wratten #25, 29, 70
UV-, IR- and X-ray imaging
Film type
37
Franz Mairinger TABLE 2.7 Object-colour rendition of Ektachrome infrared film Object colour
Rendition (no IR reflection)
Rendition (with IR reflection)
Red Magenta Green Yellow Blue White Black (Grey) Infrared
Green Green Blue Cyan Black Cyan Black –
Yellow Yellow Magenta White Red Red Red Red
Their electric response varies with the level of irradiance and is proportional to the radiant input. There are four types of photoelectric sensors : – Photoemissive surfaces, used in vacuum tube cameras and image intensifiers. Photons cause the emission of secondary electrons, which can be accelerated, collected and measured or used to form an image on a screen. – Photoconductors, are light-dependent resistors. Their resistivity varies with irradiance. For example, the photosensitive surface of a vidicon camera tube is photoconductive. Photoconductors are passive devices that cannot produce an output signal directly; they only influence the electrical current of an external bias source. – Photovoltaic cells (photodiodes) generate (small) voltages on incidence of radiation. They do not need an external voltage source for their operation. – p– n junction devices are the most recent type of photodetectors and are based on photoconductive and photovoltaic phenomena. An electronic read-out device is integrated. UV detectors The basic principles of ultraviolet detectors are very similar to those for the visible range. They are based on the principles of photoconductivity or of photoelectric emission. Imaging detectors need additional components to record the spatial intensity distribution information. The response of silicon detectors decreases quite rapidly towards shorter wavelengths, as can be seen in Fig. 2.6. This is caused by a quick recombination of electron– hole pairs near the surface. Applying a thin fluorescent film of polycyclic organic phosphors (coronene, lumogen) to the surface of a CCD, similar to fluorescent tubes, can increase the sensitivity in
38
UV-, IR- and X-ray imaging
Fig. 2.6. Spectral response (Sl) and quantum efficiency (QE) of a silicon-CCD sensor.
the UV region [11]. Special UV-CCD arrays (thinned backside illuminated) as large as 4096 £ 4096 pixels are available, next to UV sensitive targets down to 200 nm for vidicon tubes. A wide variety of imaging photoemissive UV detectors has been developed. They consist of a photo-cathode (mostly alkali metals and their compounds) onto which the image is projected, and a phosphor screen that emit light when the accelerated photoelectrons impact on it. The resolution of such devices is rather poor. IR detectors Imaging (quantum) IR photodetectors operate similar to visible spectrum photo-detectors on the basis of the photoelectric effect. But in contrast to the visible range (390 –780 nm) where photodetectors are almost exclusively based on silicon, a great variety of semiconductor materials are used, since the IR comprises a range from 0.78 to 500 mm (with decreasing photon energies). Two classes of materials employed for IR quantum detectors: up to 8 mm intrinsic and for longer wavelengths (up to 150 mm) extrinsic semiconductor materials are used [11]. The spectral behaviour of intrinsic IR detectors is determined by their inherent bandgaps; outside this limit their sensitivity drops drastically. Two primary types are in use: bulk photoconductive sensors such as PbS, PbSe, HgCdTe (also called MCT—mercury cadmium telluride) and junction sensors made from Ge, GaAs, InAs, InSb or MCT (Table 2.8).
39
Franz Mairinger TABLE 2.8 Intrinsic IR semiconductor detectors Material
Typical operating temperature (K)
Peak wave-length (mm)
Usable range
Si GaInAs InGaAsP Ge PbS InAs Hg0.7Cd0.3Te PtSi
300 300 300 300 300 77 195 90
0.9 1.6 1.3 1.5 2.4 3.1 4.5 0.9
0.6 –1.1 0.9 –1.7 1.0 –1.6 0.9 –1.6 1.1 –3.5 1.8 –3.8 1.0 –5.5 0.8 –5.0
The performance of extrinsic quantum IR detectors is not so much governed by band gaps of the materials used, by rather by the type of the added doping compounds for the Ge or Si bulk substrate. Si:Ga, Si:As, Ge:Cu, Ge:Zn combinations are used. The extrinsic quantum detectors have their peak wavelength in the far IR region; they are used mainly in IRthermography and will not be discussed further here. Silicon detectors can be used in imaging applications in the spectral range 0.6 –1.3 mm with a sensitivity peak at 0.9 mm, corresponding approximately to that of IR-sensitized photographic emulsions. Lead sulfide (PbS) is a reliable photoconductor material. It is sensitive from 1.0 to 4.0 mm with a peak at 2.2 mm (Fig. 2.7). It becomes more sensitive when cooled, but also allows for stable operation at room temperature; it is employed for targets in IR-vidicons. Gallium –indium –arsenide (GaInAs) is used for imaging detectors in IR video cameras with a good sensitivity up to 1.6 mm. Sensors in this material can be operated at room-temperature. An important class of IR detectors are Schottky-barrier photodiodes. These devices are formed by depositing metals such as Pt or Pd onto the surface of a p-type silicon substrate, which (after a heat treatment) is mounted on a CCD. These platinum silicide (PtSi) focal-plane arrays are useful between 1.1 and 5.0 mm; however, they must be operated at low temperature (77 – 90 K). They have a linear response, but their quantum efficiency is quite low. X-ray detectors Real-time electronic X-ray imaging presented a difficult problem because most video cameras cannot be exposed to intense X-ray radiation without a
40
UV-, IR- and X-ray imaging
Fig. 2.7. Spectral response of IR vidicons with PbS/PbO targets, manufactured by Hamamatsu.
severe degradation or even destruction of the image sensors. To avoid this, an X-ray conversion process with an X-ray sensitive phosphor that fluoresces in the visible and fibre-optics must be used. For X-rays with an energy above 20 keV, scintillation devices on the basis of NaI or CsI doped with thallium are used (in connection with video cameras). 2.3.4
Sensor subsystems
In this section the selection and the key parameters of solid-state and vacuum tube cameras for UV and IR documentation are discussed. For the choice of a particular camera not only the imaging performance is important, but also its physical and electrical characteristics. A key parameter is the video signal format; it provides a standardized transfer to equipment outside the camera by appropriate interfaces. There are two mutually non-compatible (colour) TV standards: PAL (CCIR) in Europe and NTSC (RS-170) in the US. They have different frame rates, scan lines, bandwidths and scanning frequencies. This is important to know for the use of monitors and A/D converters. Nowadays solid-state devices are more popular than tube cameras. There are many reasons for this: they are more or less maintenance-free, are free of geometric distortions and image drift or lag, do not suffer from image burn-in (in normal circumstances), have a better temporal stability and are
41
Franz Mairinger
lightweight and compact. However, vacuum tube cameras sometimes offer a degree of versatility that digital solid-state devices cannot match. Since camera tubes such as vidicons are analogue devices, their photo-surfaces and targets are nearly continuous in nature; the sampling is done only in one direction by the line raster, so the raster line number can be freely chosen for better vertical resolution, whereas solid-state imagers have well-defined pixels, which are sampled in both directions. Exposure and sensitivity control of tubes are also easier to manage. 2.3.4.1 UV video cameras Solid-state video cameras based on standard silicon CCD sensors have quite a low sensitivity in the UV region even in the long-wave band (Fig. 2.6); the quantum efficiency at 400 nm is only about 2%. As has been already mentioned, applying a thin fluorescent layer to the surface of the imaging CCD chip can increase the UV sensitivity; the film acts as a “converter” of UV into visible light. However, all these methods are rather cumbersome and costly. Vidicons of tube cameras for the visible with a normal Sb2S3 target are the better choice for work in the long-wave UV band. For the middle and short-wave ultraviolet special UV-vidicons with quartz faceplates are available (e.g., Hamamatsu N 983); their range extends from 200 to 700 nm. The horizontal resolution at centre is up to 900 lines. 2.3.4.2 IR video cameras Imaging silicon CCD chips in commercial digital still cameras and camcorders have their maximum sensitivity at 900 nm (Fig. 2.3) that would correspond approximately to IR photographic emulsion. This responsivity causes problems in the rendition of colours and grey tones in common photography. Hence most digital cameras are equipped with a builtin IR (cut-off) filter (e.g., Schott BG-38). Therefore, it is not possible to use these cameras for IR documentation. Some digital cameras (e.g., SONY DCS 828 with an 8 Mpixel sensor, and older models like DSC 707 and 717) and camcorders (SONY Mini DV and Digital 8) are provided with a special feature called night-shot. The IR-cut-off filter in front of the sensor can be mechanically retracted, so that near infrared recording (up to 1.1 mm) is possible. The image quality of the digital still cameras is quite good, that of the camcorders fair, so that this technique can be used as an alternative to infrared emulsions. An example of such an application is shown in Fig. 2.8. Hamamatsu, Oriel and Sony manufacture monochrome CCD cameras with increased IR sensitivity up to 1.3 mm.
42
UV-, IR- and X-ray imaging
Fig. 2.8. Friederichsmeister, Wiener Neusta¨dter Altar, detail of the Predella. (a) Normal recording, (b) IR recording with Sony DRV 9, “night shot”, IR filter B þ W 092, captured with Sony DVBK-2000E Board.
For IR work up to 2.2 mm video cameras with special IR imaging sensors must be used. Following Van Asperen de Boer [12 – 15] this technique is called infrared reflectography. Both vacuum tube and solid-state cameras can be employed for this purpose. Infrared vidicons with a lead oxisulfide (PbS/PbO) target (photosensitive front plate) are still quite popular for the examination of paintings in musea. Their (IR) sensitivity maximum is situated around 1.9 mm and ends at 2.2 mm. A typical example for this type is the N2606-06 vidicon made by Hamamatsu. Its spectral response together with other IR vidicons is shown in Fig. 2.7. The horizontal resolution of vidicons depends on several factors such as object contrast and lighting level; it varies between 250 and 500 lines, i.e., rather low. These imaging tubes can be utilized after some adjustments in regular control units, but semi-professional ones with auxiliary electronic circuits (automatic gain control, video booster, contrast enhancement) give a better image quality [16]. Several manufacturers such as Hamamatsu, Ikegami, Quantex or Sony offer such devices. Vidicon cameras have several disadvantages. Features such as image lag, blooming or even burn-in at higher lighting levels, geometric and radiometric distortions, or thermal instability with a loss of image contrast are well known and frequently encountered phenomena. In the early 1970s IR sensitive solid-state imaging devices for military applications were developed [17,18]. On the basis platinum silicide (PtSi) focal plane photodiode arrays (FPA) with silicon CMOS read-out devices became available. When the sensor is cooled to liquid nitrogen temperature
43
Franz Mairinger
(77 K), radiation in the spectral range between 1.1 and 5 mm can be detected. When such cameras are equipped with a 1.1 – 2.2 band-pass filter, they are an excellent tool for the study of underdrawings in paintings. Sterling and thermo-electric cooling of the sensors is used (see Section 2.3.4.3). These camera’s are manufactured by Mitsubishi (Japan), Infratec, Thermosensorik (both Germany) and Inframetrics (USA) and are quite expensive. Cameras with InGaAs sensors are also quite attractive. They can be operated at room temperature (18 8 C) and are sensitive in the 0.9 – 1.7 mm band and are offered by Sensors Unlimited (USA). Rockwell (US) manufactures high-priced cameras with a MCT (HgCdTe/Al2O3) sensor with a cut-off wavelength near 2.5 mm. 2.3.4.3 Sensor cooling Several parameters of IR-sensors such as random noise, responsivity or spectral sensitivity are associated with temperature. Thus, cooling provides a better performance and is even obligatory for focal plane arrays based on PtSi-sensors. This can be accomplished by means of several methods. The simplest and cheapest way is the use of a heat sink in connection with forced air-cooling. But this method is not very effective and can be a source of noise. An other possibility is the use of dewar vessels filled with liquefied gases (He 4.2 K, H2 20 K, N2 77 K, dry ice 195 K). The primary disadvantage is that the dewar must be refilled quite frequently and that the system is bulky. A Joule – Thompson cryostat makes use of the sudden expansion of a highpressure gas through an expansion valve. Similar in function to an ordinary refrigerator is the Sterling cycle refrigerator. This device makes use of the compressing and expanding action of two coupled opposed pistons (moved by linear motors). This device is very efficient; it can cool sensors down to near liquid nitrogen temperatures rapidly. It is frequently used for cooling portable thermographic cameras. Thermoelectric (TE) cooling makes use of the Peltier effect. Modern Peltier modules utilize semiconductor materials (such as BiTe); the elements are formed into large arrays (TE-modules) that can be mounted thermally in series. Such multi-stage devices can cool down sensors to about 2 1008C and allow a precise electronic temperature control. They are very reliable and are also used for cooling CPUs in computers. 2.4
SURFACE EXAMINATIONS
This group of examinations comprises methods, which reveal non-destructively surface states of an object invisible to the naked eye. Natural and
44
UV-, IR- and X-ray imaging
Fig. 2.9. Methods of surface examination.
monochromatic lights as well as UV radiation are used [3]. Figure 2.9 provides an overview of these methods. The registration of the results is achieved by photographic or electronic means. Since the methods and possibilities of macro- and micro-examinations are well known they will not be discussed any further, apart from a few remarks about the application of raking light. By choosing a large angle of incidence, the rendition of the surface relief can be exaggerated. This is a simple method to increase the legibility of engraved or embossed patterns or inscriptions. Surface imperfections of paintings such as cracks fissures or lacunae show up quite clearly, but also brush marks, impastos, primary cusps (scalloped weave deformations of canvas) and blocked areas can become (more) visible. 2.4.1
Surface examinations with ultraviolet radiation
The application of UV radiation for the examination of works of art and cultural heritage materials is an important and well-known tool [3,4,9, 19 – 21]. Since this radiation is in most cases reflected or absorbed already in the top layers of an object, mainly surface states invisible to the naked eye can be observed. This is due to two properties of UV: the excitation of visible fluorescence and the different reflectivity resp. absorption behaviour of UV compared to that of visible radiation. As already stated, here are two different methods of using UV: the photographic or electronic recording of the radiation reflected by the object and the photographic recording of the (feeble) visible fluorescence. Both methods are easy to perform with common photographic equipment; only some special filters are required. Since UV photography and fluorescence are complementary, both should be carried out in course of an investigation to facilitate the interpretation.
45
Franz Mairinger
2.4.2
Instrumental techniques for UV-fluorescence photography
Many inorganic and organic materials subjected to UV irradiation will give emit fluorescent light (of low intensity) in the visible range; it can be registered by common photographic emulsions. Its colour may comprise the entire visible spectrum; violet, blue and green hues are frequently observed. The object reflects also a considerable part of the incident UV radiation. For registering the visible fluorescence only, barrier (long-pass) filters such as Wratten 2B, 2E or Schott KV17 that exclude the reflected UV are placed in front of the lens. The registration is done in a darkened room. Since many common materials such as cotton textiles, human skin, teeth, and even eyes fluoresce quite strongly, the operator should leave the room during the long exposure. Colour films are the first choice for registration of UV-fluorescence. A differentiation between colours of equal brightness (e.g., red-green) is possible, which a black and white emulsion would render in equal grey hues. The intensity of the fluorescent light is very low, so that excessively long exposures (up to 5 or 10 min) result. The long exposure times cause a reciprocity failure: the film speed goes down. For black and white films, this causes no difficulties since for professional films correction factors are listed. For colour (reversal) films, this does present a problem. The spectral composition of the fluorescence is similar to daylight; therefore, a daylighttype film should be used. Colour films have three emulsion layers with different reciprocity failure characteristics. At exposure times longer than 0.1 s an erroneous colour rendition results. At longer exposures, no correction is possible. There are two possibilities to solve this problem: either by employing a high-speed daylight film or by using a tungsten light balanced colour film with a conversion filter (e.g., Wratten 85B, the barrier filter; e.g., Wratten 2E in front of it). These films provide a nearly correct colour rendition even at longer exposures. 2.4.3
Instrumental techniques for reflected UV photography
The reflected-ultraviolet-method works analogous to ordinary photography: the UV radiation reflected from the specimen is registered. To achieve this, the camera lens is covered with an exciter (short-pass) filter that transmits UV only and rejects visible light; Wratten 18A or Schott UG 2 (2 mm) filter are recommended. The object is evenly illuminated with mercury vapour lamps or fluorescent tubes. The correct exposure time is determined by a series of test exposures. With these results, it is possible to
46
UV-, IR- and X-ray imaging
calibrate a conventional exposure meter, if the filter that is used for the recording is placed in front of the meter-window. Electronic flash lamps can also be used. If one flash is not enough, the multi-flash-technique is applied; the shutter is opened and a series of flashes is fired. 2.4.4
Application of UV-fluorescence photography
Although many inorganic and organic materials exhibit a specifically coloured fluorescence, the application of this method for identification is not reliable. Traces of active impurities (even dirt) may cause strong fluorescence. An example of such behaviour is the colourless mineral calcite. Its fluorescence can vary from red or orange to blue or violet depending on the deposit. The same holds for natural chalk. In paint layers the characteristic fluorescence of pigments [19,21,22] is masked by the binding media and varnish layers There are also substances that can quench the intrinsic fluorescence of materials. Among pigments, the material verdigris, a basic copper acetate, which was used in gothic paintings as a green glaze, quenches the fluorescence of the natural resins mastic and dammar [22]. The same holds for earth pigments such as ochres, siena earths or umber. Thus, the identification of a material on the base of fluorescence colours is not reliable. UV fluorescence is frequently used to examine the state of preservation of paintings and polychromed sculptures. Very often later additions such as retouches or over-paints appear as dark spots or areas on the greenish fluorescence of old varnish layers. Since with increasing age also these additions gradually develop primary fluorescence, after 100 years they are difficult to detect. Similar considerations are valid for the examination of signatures and datings on paintings. If they are under an aged varnish layer, they will not show up in fluorescence examinations. If they do show up, they are on top of the varnish layer and are later additions or gone over by a restorer. Even local cleaning of such regions with organic solvents is easily detected [23]. Generally speaking, the fluorescence of paint layers, which are always complex mixtures, is mainly determined by the binding media that were employed (drying oils, resins, egg) or by the (aged) varnish layers and not by pigments. This is due to the poor penetration ability of UV radiation, which is already strongly absorbed in the top layer. Only in watercolours, lean tempera binders or wall paintings the primary fluorescence of pigments can be observed.
47
Franz Mairinger
There are many other applications of UV fluorescence. Mechanical erasure, chemical bleaching, fading of inks or paints or bio-deterioration on graphic documents can be detected quite clearly. Microscopic cracks and other defects in bronze objects can be detected by spraying a fluorescent dye to the surface. The excess is wiped off carefully; the defects show up in longwave UV quite clearly. Examinations of porcelain [24], gems [25], minerals, dyed textiles, resin coatings or pigmented glazes on gilded grounds of paintings or on metal sculptures or the state of preservation of amber objects are other examples of applications of UV fluorescence. 2.4.5
Application of UV photography
When UV radiation strikes the surface of an object, three kinds of interaction are possible: – the incident radiation is reflected or scattered on the surface while the wavelength is unchanged. UV photography renders such areas, depending on their reflectivity, in light shades. In fluorescence recordings they appear black (or dark blue), since UV is absorbed by the barrier filter. – UV radiation is absorbed and transformed into heat. Such regions reproduce in UV and fluorescence photography as black or greyish, depending on the grade of absorption. – UV radiation is absorbed by the object and excites the emission of visible (coloured) fluorescence light. Such areas appear in UV photography as black. In this manner, UV photography allows for an additional discrimination between areas that appear black in fluorescence, i.e., between UV absorbing and reflecting regions. However, the interpretation of such recordings is very often difficult. There are fewer applications [26 – 28] for this method compared to UV fluorescence photography, but it is a useful technique for the examination of graphic documents. Iron-gall ink absorbs long-wave UV strongly without generation of fluorescence, so the legibility of faded, bleached or erased parts of handwriting can be improved considerably. Regions on paper or parchment where biodeterioration, caused by bacteria or fungi, has taken place, which seem to the naked eye untouched, show up as more or less greyish spots. Figure 2.10 illustrates such applications for a fragment of a manuscript [29].
48
UV-, IR- and X-ray imaging
Fig. 2.10. Fragment of a Syrian manuscript Syr. 5 (Marc 6, 20–26, 27–35, Austrian National Library). (a) Normal recording, (b) UV-fluorescence, (c) reflected UV.
2.5
DEPTH EXAMINATIONS
Depth imaging investigations reveal in a non-destructive manner the internal structure of opaque objects or of complex strata under an opaque surface layer. They can also be used to test the structural integrity of components and assemblies. Two techniques are commonly applied: examinations with infrared radiation and with X- or g-rays (Fig. 2.11). The physical bases of these methods are quite different: the penetration ability of infrared radiation is based on its reduced scatter in turbid media, whereas the great penetrating power of X-rays is due to the high energy of the photons. Both procedures generate images that are invisible; they must be transformed by photographic emulsions, electronic detectors or fluoroscopic devices into visible pictures.
Fig. 2.11. Methods of depth examinations.
49
Franz Mairinger
Again, similar to UV and UV fluorescence investigations, both procedures are complementary: infrared examinations reveal details of the object that cannot be seen by radiography and this also holds true for radiographic investigations. 2.5.1
Depth examinations with infrared radiation
In contrast to UV radiation that is absorbed or scattered in the surface layers of an object, IR radiation quite frequently can penetrate layers or materials that are opaque in the visible region of the spectrum. This property makes such IR examination a valuable tool in archaeology, art history and conservation [4,19,30]. There are two possibilities of use: – The absorption of infrared radiation of (coloured) substances and materials differs very often from their behaviour in the visible, so that two pigments in a paint layer of the same hue are rendered differently by means of infrared registration. The same holds for optically alike additions applied to an object during restoration. – Quite a few opaque materials such as paper, parchment, wood, human skin and turbid media such as fog, haze, paint layers and printing inks transmit in the near-IR region and become transparent. This ability permits, e.g. in certain cases the visualization of underdrawings executed on the (white) ground of paintings. The mechanism of these phenomena will be discussed briefly in the next section. The obtained invisible IR images are registered by photographic or electronic means, according to the wavelength region. 2.5.1.1 Interaction of near-IR radiation with turbid media A turbid medium consists of a homogeneous, transparent matrix and small particles dispersed within. Fog, smoke, paints, and colloidal liquid or solid solutions belong to such systems. The interactions of near-infrared radiation with these turbid media are basically the same as those of visible light. Reflection, refraction, dispersion, scattering and specific absorption are the determining factors for the transmission resp. the transparency of these media [10,12]. In case of paint layers, the opacity strongly depends on the ability for scattering (and the specific absorption of coloured pigments). An example of a matrix where scattering is the dominating factor is white paint layers. Their hiding power in the visible is exclusively due to scattering (and the thickness of the layer).
50
UV-, IR- and X-ray imaging
Scattering in paint layers is described quantitatively by the Kubelka – Munk theory. For white pigments the dominating factors are the scattering coefficients and the geometric shape of the particles. Apart from the geometric factors, the scattering coefficients depend in an intricate way on the wavelength of the incident radiation and on the difference of the refraction indices of pigment and medium [31]. Generally speaking, the transparency increases with the wavelength of the incident radiation and the decreasing difference of the refraction indices. Van Asperen de Boer [12] showed that for most pigments the maximum of transmittance occurs in the spectral region between 1.8 and 2.2 mm, which is inaccessible with photographic emulsions. Images within this range can only be made visible by video cameras equipped with imaging IR detectors. 2.5.1.2 Infrared luminescence The phenomenon of fluorescence occurs not only by interaction of UV radiation with matter; blue-green light can stimulate in many inorganic and organic materials, such as natural resins, parchment, leather, minerals, pigments and wood, a invisible near-infrared luminescence. The intensity is very low but it can be registered by IR photography. Bridgeman and Gibson [32] used this method for examinations of pigments, paintings, stamps and dyed textiles. 2.5.1.3 Instrumental techniques of infrared photography Infrared photography works similarly to ordinary photography. The IR radiation reflected (or transmitted) by the object is registered on a film sensitized up to a wavelength of 900 nm (e.g. Kodak High Speed Infrared). The recording is achieved by covering the camera lens with a long-pass filter that only transmits IR radiation and rejects visible light. Gelatin filters such as Wratten 87 and 87 C or glass filters such as Schott RG 780 are recommended. The focusing should be done with a red filter, since the focal length f increases in the IR approximately by 1/200f. The best radiation sources are incandescent lamps or electronic flashes. The lighting of the objects should be as uniform as possible in order to avoid hard, pitch-dark shadows, bare of contours. The reduced scatter of IR radiation is the cause of this phenomenon. For 3D objects, diffuse lighting must be applied. For small objects a light tent is recommended [33]. The correct exposure can be determined by conventional exposure meters after test exposures. It has been stated already that layers or thin sections of opaque materials such paper, parchment, paintings (on canvas), wood (up to 5 mm), textiles,
51
Franz Mairinger
minerals (rocks) are often translucent in the near-IR region; transmission lighting is possible. The set-up for this technique is fairly complicated. Since only the transmitted IR is allowed to enter the camera, no stray light is permitted. A light-tight box must be used in a darkened room. The exposure times are extremely long (30– 40 min) [34 –36]. For the Kodak Ektachrome Infrared film, an electronic flash must be used for lighting, since the reciprocity failure starts at exposure times longer than 1/25 s. A yellow filter (e.g. Wratten No. 12) must be placed over the camera lens to absorb the UV, violet and blue parts of the spectrum. The film has a very small latitude ð 12 stop). So at least three exposures of one object (varied by a 12 stop) should be made. For excitation of infrared luminescence, a source of blue-green light is needed. A tungsten or an electronic flash lamp are used, equipped with a blue-green (glass) band-pass filter that has no transmittance in the infrared region (e.g. Schott BG 39 or DMZ 20). To register the infrared emission of an object the camera is loaded with a high-speed infrared film and an IR longpass Filter (Wratten No. 87) is placed in front of the lens. The exposures are about 6 min at f/5.6 when the object is illuminated by two 500 W light sources at a distance of 1 m. For small objects an (filtered) electronic flash lamp at a distance of 10 cm is the better choice. A single flash at a lens opening of f/2.8 is enough for most objects. 2.5.1.4 Instrumental techniques for infrared reflectography It has been stated that photographic emulsions are incapable of registering the important spectral region between 1.8 and 2.2 mm; instead, video cameras equipped with IR imaging detectors are employed. Both types— vacuum tube cameras and solid-state cameras—are used [11,12,37 – 41]. The advantages and disadvantages of vidicon cameras were already discussed. The main problem of all video cameras is their fairly low resolution. In order to detect fine structures such as drawn lines, only small, overlapping (100 – 150 pixels) sections of the object should be registered. For large paintings their number may go up to 100 (and more). Thermal instability and image lag and burn-in of vidicons present quite a problem for this process. In this respect solid-state cameras have the much better performance. At the start the optimum camera settings should be adjusted on places of highest contrast of the object and should not be changed during the entire registration process. The recording of so many overlapping frames affords a precise positioning system for the camera [42 – 44]. The single frames are digitized by a frame grabber and stored in a computer. After geometric and radiometric corrections
52
UV-, IR- and X-ray imaging
a mosaic program assembles them. Such programs were published by Wecksung et al. [45], Billinge et al. [42] and Mairinger and Papst [46]. 2.5.1.5 Applications of infrared examinations Infrared examinations are employed in many fields of research and conservation of cultural heritage objects; only a few applications in graphic arts, paintings and related fields will be mentioned here [47 – 49]. 2.5.1.6 Graphic arts The non-destructive infrared examination of works of graphic arts as drawings, prints and illuminated manuscripts on paper, papyrus and parchment answer many questions of art historians, historians and conservators. The legibility of manuscripts, documents, palimpsests and papyri can be improved considerably. Very useful tools for such tasks are the already mentioned digital still or video cameras with the night-shot feature. Inks and pigments that appear identical in the visible are rendered differently by IR (false colour) photography. Thus, brownish writing and drawing fluids such as sepia, bistre, (aged) iron-gall and bark inks, of which the differentiation is cumbersome, can be discriminated [48]. In this manner, texts rendered illegible due to mechanical erasure, chemical bleaching, charring, obliteration or other deteriorating procedures can be made at least partially visible. Especially of soot inks, even remains of it, are easily detected [50,51]. As in other cases, complementary examinations by reflected UV and UV-fluorescence must be carried out. The application of infrared reflectography for such tasks has severe restrictions. The sensitivity of the usual solid state sensors (GaInAs, PtSi) in the spectral range between 0.78 and 1.2 mm is very low (for PtSi , 0) and in addition all common inks, except soot and pigmented inks or paint layers, vanish completely at wavelengths greater than 1.3 mm. Nevertheless, a discrimination between the various writing fluids is possible by using a tube camera (equipped with an IR vidicon) and appropriate narrow-band filters. 2.5.1.7 Paintings Nowadays infrared examinations of paintings have gained nearly the same importance as radiography for the stylistic approach of paintings and for conservation problems. Two fields of applications can be mentioned: – the detection and differentiation of compositional alterations by the author and of later addition, as over-paintings, retouches and reconstructions, by another hand. This can be accomplished by infrared (digital) photography. IR reflectography is less suitable, since a great number of
53
Franz Mairinger
paint layers become transparent near 2 mm. The base for it is the different infrared absorption or reflectivity of materials, which look alike in the visible. This information is quite interesting for art historians and restorers. Infrared photography provides also a possibility to locate old retouches and over-paintings under old, discoloured varnish layers, which are hard to detect by UV fluorescence examinations, because they are masked by the strong luminescence of the varnish. – the revealing of underdrawings and the use of stencils in paintings [12,41, 52 – 56]. Most artists made a preparatory drawing of the intended composition with brush, pen, charcoal or black natural chalk on the grounded support of a painting. The character of such underdrawings varies between rough sketches and detailed drawings. Opaque paint layers cover the underdrawings in later phases of the work. Sometimes parts of the underdrawing are even visible for the (trained) naked eye due to saponification of the basic white lead, which was the only important white pigment till 1835. For white grounds (chalk, gypsum) and black colouring matters, these drawings can be made visible again by infrared reflectography and with some restrictions (blue and green areas) by IR photography. These drawings, untouched by later hands give vital information about the author, his workshop and the optional use of stencils for contemporary or later copies. The infrared visibility of underdrawing techniques has been discussed by Jennings [57]. 2.5.1.8 Other cultural heritage materials Infrared radiation penetrates quite often saline incrustations, resinous crusts and other forms of patina on pottery or glass and allows to see the buried patterns. This holds for other materials too. Dyestuffs used on textiles exhibit quite often a characteristic infrared reflectance or luminescence; e.g. restorations on tapestries are easily detected by means of infrared colour film. 2.5.2
Depth examinations with X-rays and gamma-rays
The ability of X-rays to penetrate opaque objects and to blacken photographic emulsions was used quite early after their discovery by Ro¨ntgen in 1895 [58,59] in various fields. In 1913 the physician Faber [60 – 62] made the first radiograph of a painting and obtained a patent for this method, which proved to be a hindrance for the general application in examinations of works of art.
54
UV-, IR- and X-ray imaging
A further obstacle was a heated discussion about the alleged destructive actions of X-rays on paintings in the twenties of the last century. But soon after this debate radiography became an important, indispensable tool for art historians and archaeologists. In 1938 Wolters [63] published a book about the possibilities and importance of X-ray examinations in art history, where the methodology of reading and interpreting radiographs was developed and refined. So today X-ray examinations are among the most important methods for depth examination of art objects [6,19,63 – 67]. In the following sections the interaction of X-rays with matter is briefly described (see Chapter 4 and 5 for more detailed information), followed by practical and technical considerations and finally the application of radiographic investigations of paintings are discussed. It is of course impossible to present a detailed instruction for the production of radiographs and their interpretation; only a few general rules will be outlined. 2.5.2.1 Interaction of X-rays with matter—the attenuation laws When X-ray or gamma ray photons interact with any form of matter, some are transmitted, some are absorbed and some are scattered from their path of incidence; as a result of these processes, the incident beam is attenuated [68]. Behind the object a shadow image is generated that can be registered by means of fluorescent screens or photographic emulsions. The attenuation depends on the quality of the applied radiation, the thickness and the material composition of the object. The understanding of laws of attenuation is of paramount importance for the interpretation of radiographs. Quantitatively the attenuation is described by the equation: I ¼ I0 e2mL x where I0 is the primary intensity of the incident beam, I, the intensity of the transmitted beam, mL ; the linear attenuation coefficient (given in cm21) and x, the thickness of the material. mL is very often replaced by the more fundamental mass attenuation coefficient m, which is obtained by dividing mL by the density r: m ¼ mL =r Attenuation coefficients vary not only by the nature of the materials, but also by the wavelength of radiation (i.e. photon energy). This dependence can be expressed by the approximate equation: m ¼ cl3 Z4
55
Franz Mairinger
where c is a universal constant, l; the wavelength (of the monochromatic radiation) and Z, the atomic number of the absorbing material. For complex mixtures such as paint layers, m is calculated by adding the contributions according to the percentage of the present atomic species. The equation implies also that organic pigments and the common binding media (oil, distemper, glue, gums) in paints make only small contributions to the overall attenuation, whereas pigments containing lead, mercury or tin attenuate strongly already in thin layers. The absorption ability of various materials can be depicted by the size of their half-value layer d 12 : This is the layer thickness of materials that diminishes the intensity of the incident radiation by 50%. Figure 2.12 shows this value for several pigments and metal foils used in paintings and for ˚ gilding up to the 19th century. Values are given for three wavelengths: 0.2 A ˚ (for silver) and 0.71 A ˚ (for molybdenum). (Ka radiation of tungsten), 0.56 A An important factor for the quality of a radiograph is the radiation contrast behind the specimen, which governs the film contrast and thus the readability of the recording. Figure 2.13 illustrates the correlation between utilized wavelength, linear attenuation coefficient and contrast. The above considerations are strictly valid only for monochromatic radiation, but X-ray tubes produce polychromatic radiation that complicates the attenuation processes. A defined correspondence between these two types of radiation does not exist. This is partly due to the fact that for thicker specimens a hardening of the incident radiation occurs. After the soft
Fig. 2.12. Half-value layers d 12 (in mm) of metal foils and pigments. 56
UV-, IR- and X-ray imaging
Fig. 2.13. Attenuation of X-rays by a softwood step-wedge.
components are absorbed in the upper layers of the object, the radiation becomes more and more homogenized and gains a greater average penetration ability. For hard radiation (.100 kV) and thick specimens, scattering (incoherent and coherent) becomes the dominant factor for attenuation and is a problem in radiography with both X-rays and gamma rays. The intensity of the scattered radiation increases with thickness of the specimens, but is nearly independent of wavelength. The greater portion of scattered radiation that is blackening the film originates from the specimen under examination, but any part of the surroundings (e.g., walls, the floor, even parts of the film cassette) that is in the path of the direct radiation from the X-ray tube becomes a source of scattered radiation, which can affect the radiography and its contrast. It is, therefore, recommended that the diameter of the beam is adjusted to the dimensions of the object by the use of cut-out (lead) diaphragms and that the film is covered by a thin lead foil (see filters). In view of the above, it is possible to draw the following conclusions for practical work: – The attenuation depends on: (a) thickness of the object, (b) the quality of radiation (third power of l), and (c) the atomic number (third power of Z) of the absorbing material; – The attenuation is caused by absorption and scattering. Both depend on the atomic number, but the latter is nearly independent of wavelength. – The quality of radiation, the chemical composition, the thickness and the density of the object determine the radiation contrast.
57
Franz Mairinger
– Since the readability of a radiograph depends strongly on contrast and scatter, the softest radiation is preferred, which must, of course, penetrate the specimen to a certain amount. For hard radiation diaphragms and masks are used to minimize the scattered radiation. – For best definition and sharpness of radiographs, tubes with small focal spots are preferred; for soft radiation the best window material is beryllium. The film must be in close contact with the object. If this is not possible the tube-object distance should be increased. 2.5.2.2 Special radiographic techniques There are art objects where standard radiographic techniques give unsatisfactory results or fail. Examples are paintings where the support (mostly wood panels) bears on both sides paint layers so that the two radiographs are superimposed, or paintings on metallic supports (copper, iron). Other examples are radiographs of 3D objects such as sculptures with complicated internal structures and differences of materials, which are very difficult to interpret, the detection of watermarks in illuminated manuscripts or radiography of very small archaeological objects (jewellery). For such objects special radiographic methods must be applied.
Stereoscopic techniques in radiography Visual 3D depth sensation is perceived because the two human eyeballs are laterally separated by about 67 mm. So two slightly different images are produced, which are fused into a 3D impression by the visual process. A similar impression is obtained when a binocular viewer presents a pair of stereoscopic photographs to the eyes. The same technique can be used in radiography of 3D objects. The technical layout for it is presented in Fig. 2.14. Two slightly different images are taken either by moving the source or the object relative to each other. Three conditions must be fulfilled for a 3D perception of the object: – Both film sheets are in the same plane. – The central ray is always orthogonal to the film plane. – The film-focus distance must be the same for both exposures. The two radiographs are posed on two separate illuminated screens and viewed with a mirror stereoscope.
58
UV-, IR- and X-ray imaging
Fig. 2.14. Technical layout for stereoscopic radiographs.
Tomographic techniques Tomography is a special technique that provides a (relatively) distinct image of a selected plane in a 3D object. Originally, it was developed for medical radiography, often termed “body-section radiography”. Since about 1980 it was replaced by computerised tomography (CT). Hounsfield [69] developed in the 1970s a new tomographic technique, which he called computerised transversed axial tomography (CAT scanning). In the first generation of these instruments a rotating X-ray pencil beam, which was collinearly connected with a detector, was scanned in small steps over the object to be examined. In this arrangement, source and detector are mounted on a circular frame, called the gantry. The attenuation data of each step are stored in a computer and special algorithms generate an image of the chosen section.
59
Franz Mairinger
CT examinations permit to obtain radiographs of any selected plane of 3D objects without any interference from other layers. A basic requirement is that the beam must penetrate the specimen with sufficient intensity, since otherwise artefacts result. Medical CT instruments use tube voltages between 120 and 130 kV. They can be utilized for non-metallic art objects [70 –72]. For technical examinations CT instruments with tube voltages up to 420 kV are available, which can even be used for study of bronze sculptures, e.g. to determine their wall thickness and casting defects [73]. Since the range of grey values in the image can be chosen, even a core, if still present, can be made visible [74,75].
Microfocal radiography In standard radiography the film should be as close as possible to the object (scale 1:1) and the source-film distance comparatively long in order to minimize geometric unsharpness caused by the (effective) size of the focal spot of the X-ray source. This is not a very satisfying technique for small archaeological objects such as earrings, necklaces or needles [76]. Since about two decades, commercial microfocal tubes are available. Their focal spot has an effective size in the range down to 4 – 8 mm, which makes these tubes the equivalent of point sources. This allows a direct enlargement of the radiographic shadow picture by placing the object near the focal spot and making the distance of the film (or the detector system) as large as possible. Additionally, the intensity of scattered radiation is reduced in this manner. Tube voltages up to 200 kV are possible in some microfocal tubes, so that metallic objects can be examined [77].
2.5.2.3 Electrons as imaging medium Electrons also are able to blacken photographic emulsions and can be used for radiographic examinations. There are two groups of methods, which can be combined: – Electron radiography and beta-radiography. Both methods apply an external source of electrons, which can be transmitted through thin objects and are registered by photographic emulsions. – Auto-electron-radiography. Here, the object itself is the source of the electrons; they are generated by irradiating an object with hard radiation. This method works only when elements with a high atomic number are present.
60
UV-, IR- and X-ray imaging
Electron- and beta-radiography For electron radiography the object is brought into close contact between a thin lead foil (20 mm) and an X-ray film having an emulsion only on one side of the base. This “sandwich” is wrapped in a light tight envelope [78] (Fig. 2.15a). The radiation is applied to the foil side where photo- and Comptonelectrons are produced. The kinetic energy of the electrons that determines their penetration power, depend on the applied tube voltage. Typically voltages in the 100 – 400 kV range are used. The electrons are able to penetrate thin objects such as paper or parchment and blacken the film. This is an excellent method to register watermarks of paper on tightly printed pages where transillumination gives unsatisfactory results. For beta-radiography a foil coated with a beta-emitter (mostly 14C) serves as (low intensity) electron source. A limitation in this respect is that the energy of the emitted electrons cannot be varied. The foil is brought in close contact with the object and the film, after which exposures taking many hours or days commences. Mundry et al. [79] used this technique for the recording of watermarks. Auto-electron-radiography It has been already discussed that the absorption of X-ray photons is accompanied by an ejection of electrons from the inner shells of the atoms in the irradiated material. The probability of ejection of photoelectrons increases with the atomic number of the elements, corresponds to the increasing mass absorption coefficients of the heavy nuclei such as lead, mercury or tin in contrast to calcium, aluminium or silicon. All these elements are components
Fig. 2.15. The use of electrons as imaging radiation. Technical layout for (a) electron radiography and (b) auto-electron radiography.
61
Franz Mairinger
of inorganic pigments, have different emission properties and therefore result in different blackenings on photographic emulsions. The technical layout for such examinations of paintings is rather simple (Fig. 2.15b). Hard radiation is produced by an X-ray generator (150– 400 kV). The soft parts of the radiation, which would strongly blacken the film, are absorbed by thick copper or lead filters. The bare film is pressed tightly onto the surface of the painting, so that examination must be done in a darkened room or under red safelight. This method permits the radiography of paintings on copper or stone slabs. For very thin or weakly absorbing objects such as stamps, documents, illuminated manuscripts or engravings a combination of electron- and autoelectron-radiography is possible. The technical layout is the same as shown in Fig. 2.15a (lead foil/object/film). The registration (i.e., the film blackening) of paint layers may look quite differently, depending on the layer thickness and tube voltage, since the effects of absorption and emission processes are superposed. Paint layers may also absorb photoelectrons generated by the lead foil. Such areas are rendered transparent on the film [80]. These methods are very interesting for the non-destructive examinations of valuable illuminated manuscripts [79]. They can reveal important information on painting techniques and provide hints about pigment usage and later additions. 2.5.2.7 Compton X-ray backscatter techniques Conventional X-ray testing techniques require access to both sides of an object. For thick or strongly absorbing specimens this can be disadvantageous, since high tube voltages must be applied and the image contrast is low. The Compton X-ray backscatter technique requires only access to one side of the object. It is based on the detection of the radiation that is scattered by the specimen [80]. An X-ray pencil beam scans the chosen area in a raster fashion while solid-state detectors, located on both sides of the scanning beam record the generated Compton backscatter radiation. A series of 22 slices is generated, each made at a constant depth below the surface of the object. The depth resolution of each slice is 0.4 mm, almost free of the influence of the other slices while the maximum depth is 50 mm. After amplification and digitisation of the signals, a computer generates the images. Such an instrument is manufactured by Philips named ComScan. This technique has several advantages: – It is possible to examine wall paintings, since only access to one side of the object is needed.
62
UV-, IR- and X-ray imaging
– Radiographs of paintings on metal or stone slabs are possible. – 3D sets of data are obtained without complicated reconstructions as in CT. – The image contrast is the result of local changes of radiographic density in the object, whereas in conventional radiography the mean value of the attenuation coefficient is determining for the contrast. Niemann and Roy [83] have shown the application of this method for the examination of sculptures. 2.5.2.4 Application and interpretation of radiographic examinations The trivial fact that any variation in image grey values (i.e., blackening on a film) is caused by a variation in layer-thickness or a change in chemical composition is a physically correct statement, but it is not very helpful for the interpretation of the structure of complex art objects, since a radiograph is a summation image of all (absorbing) layers. Without a thorough knowledge of material history, art technology, conservation techniques and the permanent comparison with the examined object, the possibilities of interpretation are limited. So the following considerations should not be understood as a foolproof guide for reading radiographs. Examples are presented for paintings and graphic arts, sculptures (wood, ivory, metal) and archaeological objects. A sophisticated systematic approach has been developed for the interpretation of radiographs of paintings [63]; the latter are relatively easy to produce so that there is a vast stock of recordings. Additionally, there are many reliable historical sources on technical aspects of painting available which assist the interpretation. For radiographs of sculptures the situation is worse, due to a more difficult technical layout and less knowledge about historical carving and casting techniques. On the other hand, for archaeological objects the situation is much better. The refined technical methods of non-destructive testing of materials are quite helpful for the interpretation of radiographs of historical arms, tools and other objects of daily life. Paintings Radiographs of paintings supply information about the materials used (e.g., support, pigments), the techniques employed, including peculiarities of specific artists and their workshops, compositional and dimensional changes, temporal changes and damages such as the effects of aging processes, cracks, paint losses, later additions by restorers, etc. Information on support materials. The structure and construction of wooden supports, like growth rings, number of boards, textures
63
Franz Mairinger
(puttied knots, cracks), tool marks, joining techniques, worm tunnelling and later accretions can be seen quite clearly even on panel paintings that bear paint layers on both sides. In gilded panels the gold foil is invisible in a radiograph because the thickness of the gold leaf is around 1 mm. The same holds for silver foils. The structure of textile supports can only be depicted when a ground layer was applied on it, since the X-ray absorption of vegetable fibres (such as flax, hemp, cotton) is too low. In most cases, a very clear (negative) impression of the ground layer will be visible. This is important for relined paintings, where the original support is covered. The structure and diameter of the yarn (hand- or machine twisted) and the weave and seams become visible. These properties can be characteristic for an artistic period or landscape. Other features that may show up are primary cusps, i.e., scalloped weave deformations near the borders of the canvas. Before application of the ground layer, the canvas is fastened by nails on a strainer. Thereby the weave is slightly distorted and the dried ground fixes these distortions. If the cusps are still visible in the radiograph, the dimensions of the painting were unaltered. Information on painting technique and layer structure. The interpretation of these aspects of a radiograph is a complex task that cannot be covered by brief considerations. Very often the radiographs are quite different from the surface image with respect to contours, distribution of lights and contrasts and even details of the composition (pentimenti). There are many reasons for this: the superposition of all layers, the difference of X-ray and optical absorption of pigments or the presence of later additions. Thus, the dark blue veil of a gothic Madonna where a blue pigment with low X-ray absorption was applied, may be rendered very light in a radiograph, when the artist has used a greyish underpaint consisting of a mixture of white lead and bone black, due to the strong absorption by the former pigment [84,85]. Optically very light coloured flesh parts (carnation) are rendered dark grey when gesso was used as a source of light and the colouring is done with thin organic glazes. The detection of damages and (puttied) losses in paint layers can be fairly easily performed. In contrast to UV fluorescence examinations the true size of the damage can be recognized. Cracks in the paint layer are normally rendered black; if they are white the area was over painted. On the other hand, the loss of glazes in the surface layers is very difficult to detect; for this purpose, a microscopic examination of the structure of cracks is better suited.
64
UV-, IR- and X-ray imaging
Graphic arts The examination of painting techniques in illuminated manuscripts is quite similar to that of paintings. Soft radiation or electrons must be used for radiographs [29,86]. Sculptures Whereas radiographs of paintings and their art-historical interpretation are nearly ubiquitous, radiographs of sculptures are less numerous [87 –91]. There are several reasons for this: – Many and very different inorganic and organic materials have been used for sculptures: different kinds of wood (polychromed and un-polychromed), ivory, amber, wax, papier-maˆche´ (carta pesta), rocks (limestone, marble, sandstone, granite, alabaster, jade, artificial stone and others), burned clay, porcelain, metals and their alloys, composite materials (metals – enamel, Gold – ivory and others) and since the last century organic polymers casts. For all these materials quite different techniques (carving, modelling, casting, embossing) and various tools are applied so that only a limited number of historical sources are available compared with reports on painting techniques. – For radiographs of these materials, quite different types of X-ray sources are needed and the accompanying measures for radiological protection can be costly. – Sculptures are 3D objects with complex inner structures; the 2D projections and the inverse perspective make the interpretation complicated. Traces of tooling are difficult to detect in radiographs. This situation is now gradually changed by the application of modern industrial CAT-scanners. With medical scanners only sculptures made of organic materials (tube voltages usually are restricted to about 130 kV) and limited size (determined by the diameter of the gantry) can be examined. In the absence of systematic studies on the subject, the complex of answerable questions is limited. For wooden sculptures, information about the method of construction, the presence of cavities, the joining techniques employed for the various parts, the preservation of the polychromy and the presences of restorations is fairly simple to obtain. Even the depth of penetration of solidifying polymer solutions for damaged wood can be detected by adding (radiopaque) contrast media that are used for medical radiography to the solutions. Goebbels et al. [91] successfully made precise measurements of wall thickness of antique
65
Franz Mairinger
great-bronzes by CAT scans. Casting defects, residual cores, welding processes, repairs cracks and other flaws are also detectable. Archaeological objects The technical layout and the problems of radiography of archaeological objects are very similar to those described in the section above. The radiological examinations of archaeological specimens are more difficult than those of modern (metallic) materials. Radiographs of very small objects are difficult, microfocal techniques must be used [76]. Good radiographs of ceramics provide also difficulties, because voids and other faults are very small. But the main problem is always the interpretation of the obtained radiographs. Usually long experience and the examination of larger series of similar objects are necessary to gain reliable results and statements. For example, there exist very few published results on braze welding of nonferrous metals; the book by Driehaus [67] provides detailed instructions for such examinations. REFERENCES 1 2 3 4
5 6
7 8 9 10
66
A.P. Laurie, The Painters Methods & Materials, Chapter XI: The Optical Properties of Oil. Seeley Service & Co., London, reprinted 1960. P.B. Coremans, La technique de Primitifs Flamands, Stud. Conserv., 1 (1952) 1–2. see also pp. 8–29, 145 –161. D.C. Creagh and D.A. Bradley (Eds.), Radiation in Art and Archeometry. Elsevier, Amsterdam, 2000. M. Matteini and A. Moles, Scienca e restauro, metodi di indagine. Firenze 1986. German translation (A. Burmester), Naturwissenschaftliche Untersuchungsmethoden in der Restaurierung. Mu¨nchen, 1990. R.A. Quinn and C.C. Sigl (Eds.), Radiography in Modern Industry. Eastman Kodak Company, Rochester, NY, 1980. A. Gilardoni, A.O. Orsini and S. Taccani, X-rays in Art. Gilardoni SpA, Mandello Lario (Lecco), Italy, 2nd ed., 1994, Realizzazione editorale Grafica & Arte Bergamo. E. Walmsley, C. Fletcher and J. Delaney, Evaluation of system performance of near-infrared imaging devices, Stud. Conserv., 37 (1992) 120 –131. Schott, Optische Filter—Glasfilter, Interferenzfilter und Spezialfilter (No. 3555d). Also available on floppy disk. Kodak Publication No. M-27, Ultraviolet & Fluorescence Photography. Eastman Kodak Company, Rochester, NY, 1968. W. Bru¨gel, Physik und Technik der Ultrarotstrahlung. Curt R. Vincentz Verlag, Hannover, 1961.
UV-, IR- and X-ray imaging 11 12
13 14 15 16 17 18 19 20 21
22
23 24 25
26
27 28 29
M.W. Burke, Image Acquisition. Handbook of Machine Vision Engineering, Vol. 1. 1996, 861 pp. J.R.J. Van Asperen de Boer, Infrared Reflectography. A Contribution to the Examination of Earlier European Paintings, Thesis, Univ. of Amsterdam, 1970, Amsterdam. J.R.J. Van Asperen de Boer, Reflectography of paintings using an infra-red vidicon television system, Stud. Conserv., 14 (1969) 96–118. J.R.J. Van Asperen de Boer, Infrared reflectography: a method for the examination of paintings, Appl. Optics, 7 (1968) 1711–1714. J.R.J. Van Asperen de Boer, Infrared reflectograms of panel paintings, Stud. Conserv., 11 (1966) 45–46. Nowadays these expensive additions can be replaced by using image processing software. F.D. Shepherd, Silicide infrared staring sensors, Proc. SPIE Infrared Detectors Arrays, 930 (1988) 2–10. J. Silvermann, J.M. Moony and F.D. Shepherd, Infrared video cameras, Sci. Am., 3 (1992) 58–63. F. Mairinger, Strahlenuntersuchung an Kunstwerken. E.A. Seemann, Leipzig, 2003. E.R. De la Rie, Ultraviolet Radiation Fluorescence of Paint and Varnish Layers, PACT 13, 1986, pp. 91 –108 (Scientific Examination of Easel Paintings). H.P. Autenrieth, A. Aldrovandi, P. Turek, Die Praxis der UV-Fluoreszenzfotografie Z. Kunsttechnol. Konserv., 4(2) (1990) 215 –234 and Nachtra¨ge zur Praxis der UV-Fluoreszenzfotografie. 6(1) (1992) 195 –196. Both papers are concerned with the UV-fluorescence of wall paintings. R.L. Feller, Artist’s Pigments—A Handbook of Their History and Characteristics, Vol. 1. National Gallery of Art Washington, Cambridge University Press, Cambridge, 1986. E.R. De la Rie, Fluorescence of paint and varnish layers. Part 1: Stud. Conserv., 27 (1982) 1–7 (1), Part 2: 27 (1982), 65–69 (2), Part 3: 27 (1982), 102 –108 (3). J. King, The examination of porcelain etc. by ultraviolet light, Apollo, 58 (1953) 74. J. De Ment, Handbook of Fluorescent Gems and Minerals: An Exposition and Catalogue of the Fluorescent and Phosphorescent Gems and Minerals Including the Use of Ultraviolet Light in the Earth Sciences. Mineralogist Pub. Co., Portland 15, OR, 1949. I.N. Gilgendorf, Study and restoration of lost ancient inscriptions on the dry plaster by the method of infrared and ultraviolet photography, ICOM Comm. f. Conserv., Fourth Triennial Meeting Venice (1975), Preprints 1, 75/4/8, pp. 1–9. C.S. Holliday, The application of ultra-violet light to prehistoric rock art, SAMAB, 7 (1961) 179 –184. B. Hallstro¨m, The use of UV-reflectograms for the examination of paintings, ICOM Comm. f. Conserv., Fourth Triennial Meeting Venice (1975) 75/VI/10 F. Mairinger, Physikalische Methoden zur Sichtbarmachung verblasster oder getilgter Tinten, Restaurator, 5(1–2) (1981/82) 45–56.
67
Franz Mairinger 30
31 32 33 34 35
36
37
38
39 40
41
42
43
44 45
46
68
F. Mairinger, The infrared examination of paintings. In: D.C. Creagh and D.A. Bradley (Eds.), Radiation in Art and Archaeometry. Elsevier, Amsterdam, 2000, pp. 56–75. G. Kortu¨m, Reflexionsspektroskopie. Verlag Chemie, Weinheim, 1968. C.F. Bridgeman and H.L. Gibson, Infrared luminescence in the photographic examinations of paintings and other art objects, Stud. Conserv., 8 (1963) 77 –83. C.E. Engel, Ultraviolet and fluorescence recording, Chapter 8, in Photography for the Scientist, 1968, pp. 363–382. Academic Press, London. D.A. Kushel, Application of transmitted IR radiation to the examination of artefacts, Stud. Conserv., 30 (1985) 301– 310. L. Lazzarini, Studio tecnico-scientifico di un dipinto di Guiseppe Porta detto II Salviati. Notizie da Palazzo Albani (Univ. Urbino) Anno III, no. 2/3 (1974) 38 –47. C.F. Bridgeman, in: S.A. Pollack and H.C. Pamphlet (Eds.), Uses of Radiation in Philately and Examination of Paintings. Eastman Kodak Co., Rochester NY, 1965, 28pp, Plates (Reprint: “The Science of Ionising Radiation” ed. L.E. Etter, 1965, Chapter 27). D.L. Clark, et al., Design and performance of a 486 £ 640 pixel platinum silicide IR imaging system, SPIE—The International Society for Optical Engineering, 1590 (1991) 303 –311. M. Cohen and G. Olsen, Infrared detectors. Tuned in the near infrared. InGaAs detector arrays offer silicon performance in infrared applications from blood testing to industrial inspections, Photonics Spectra, 28(7) (1994) 132 –134. D.P. Leech and I. Gutmanis, The U.S. infrared detector industry. Prospects for commercial diversification, SPIE—Int. Soc. Opt. Engng, 1683 (1992) 2– 12. E. Walmsley, C. Metzger and J.K. Delaney, Evaluation of platinum silicide cameras for use in infrared reflectography. ICOM Com. Conserv. 10th Triennial Meeting Washington, DC, 1993, pre-prints 57– 62. E. Walmsley, C. Metzger, J.K. Delaney and C. Fletcher, Improved visualisation of underdrawings with solid-state detectors operating in the infrared, Stud. Conserv., 39 (1994) 217– 231. R. Billinge, J. Cupitt, N. Dessipris and D. Saunders, A note on an improved procedure for the rapid assembly of infrared reflectogram mosaics, Stud. Conserv., 38 (1992) 92–98. A. Burmester, J. Cupitt, H. Derrien, N. Dessipris, A. Hamber, K. Martinez, M. Mu¨ller and D. Saunders, The Examination of Paintings by Digital Image Analysis, Conference Volume, Third International Conference on Non-destructive Testing of Works of Art. Italian Society for Non-Destructive Testing, Siena, 1992, pp. 201– 214. I. Sandner, FH Ko¨ln (1998). Private Communication. G. Wecksung, R. Evans, J. Walker, M. Ainsworth, J. Brealy and G. Carriveau, Assembly of infra-red reflectograms by digital processing using a portable data collecting system, ICOM Comm. Conserv. Eighth Triennial Meeting, (1987) pp. 107–109. F. Mairinger and A. Papst, Die Erstellung von Infrarot-Reflektogrammen mittels des Programmpakets IREI¨KON, Fourth International Conference on
UV-, IR- and X-ray imaging
47 48
49 50
51 52
53
54
55 56
57
58 59
60 61 62 63
Non-destructive Testing of Works of Art. Deatsche Gesellschaft fu¨r Zersto¨rungsfreie Psu¨fung, Berlin, Vol. 1 (1994) 175 –182. J.M. Cabrera and M.C. Garrido, Estudio te´cnico del Guernica, Boletin del Museo del Prado, 2(6) (1981) 147 –156. A. Burmester and K. Renger, Neue Ansa¨tze zur technischen Erforschung von Handzeichungen: Untersuchungen der ‘Mu¨nchner Rembrandt-Fa¨lschungen’ im nahen Infrarot, Maltechnik-Restauro, 92(3) (1986) 9–34. 21 ill., 55 ref. N. Goethghebeur and L. Kockaert, Le Grand Calvaire d’Albert Bouts au Muse´e des Beaux-Arts de Bruxelles, Bull. IRPA XVIII, Vol. 18 (1980/81) 5 –20. S. Fletcher, A Preliminary study of the use of infrared reflectography in the examination of works of art on paper ICOM Comm. f. Con.; Preprints 7th Triennial Meeting Copenhagen 1984, 84.14.24-28. H. Ku¨hn, Die Zeichenmaterialien des Niklaus Manuel, Maltechnik-Restauro, (1982) 156–157. D. Hollanders-Favert, M. Lietaerd-Parmentier, R. Van Schoute and H. Verougstraete-Marcq, Le dessin sous-jacent chez Albert Bouts, Acta Lovanensis, 4 (1975) 41–135. E.D. Bosshard, Fortschritt in der naturwissenschaftlichen Gema¨lde-untersuchung. Die Erforschung der Unterzeichnung mit dem Infrarot-Fernsehgera¨t, Z. fu¨r Schweizerische Archa¨ol. Kunstgeschichte, 39 (1982) 76–80. H. Irgang, Introduction to the study of underdrawing in early Netherlandish paintings. In: B. Cory (Ed.), Student Papers Presented at the Art Conservation Training Programs, 11th Annual Conference, May 2 & 3, 1985; Art Conservation Program, Winterthur Museum/Univ. of Delaware, New York, 1986, pp. 85 –101. K.H. Weber, Die sixtinische Madonna, Maltechnik Restauro, 90(4) (1984) 9–28. Bibliographie de l’Infrarouge et du de dessin sous-jacent dans, Le Dessin SousJacent dans la Peinture, Van Schoute, R. et Verougstrate-Marcq, H., Louvain-la Neuve Colloques, 1975–2003. J. Jennings, Infrared visibility of underdrawings techniques and media. In: R. Van Schoute and H. Verougstrate-Marcq (Eds.), Le Dessin Sous-Jacent dans la Peinture, Colloques IX 12 –14 Septembre 1993. Colle`ge Erasme—Place Blaise Pascal, Louvain-la-Neuve, 1993, pp. 241 –252. W.C. Roentgen, Eine neue Art von Strahlen. Sitzungsbericht der physikal.medizinischen Gesellschaft, Wu¨rzburg 137 (Dez 1895). ¨ ber eine neue Art von Strahlen W.C. Roentgen, in: W. Gerlach u. Krafft (Eds.), U Essay. Kindler Verlag, Mu¨nchen, 1972, Naturwissenschaftliche Texte bei Kindler. A. Faber, Eine neue Anwendung der Ro¨ntgenstrahlen. Die Umschau, 18 12 (21. 3. 1914). ¨ lbildern. Umschau 15, A. Faber, Die Ergebnisse der Ro¨ntgenuntersuchung von O 24 (11.6.1921), 325– 327. C.F. Bridgeman, The amazing patent on the radiography of paintings, Stud. Conserv., 9 (1964) 134 –139. Ch. Wolters, Die Bedeutung der Gema¨ldedurchleuchtung mit Ro¨ntgenstrahlen fu¨r die Kunstgeschichte dargestellt an Beispielen aus der niederla¨ndischen und deutschen Malerei des 15. und 16. Jahrhunderts. Vero¨ffentlichungen zur Kunstgeschichte 3, Frankfurt am Main 1938.
69
Franz Mairinger 64 65
66 67 68 69 70 71
72 73
74
75 76
77 78 79
80 81
70
M. Hours, A la De´couverte de la Peinture par les Me´thodes Physiques. Analyse Scientifique et Conservation des Peintures, Office du Livres, Fribourg, 1976. R. Van Schoute and H. Verougstraete-Marcq (Eds.), Radiography, Chapter VII, in Scientific Examination of Easel Paintings, PACT 13 (1986), Conseil de L’Europe, Strasbourg, pp. 131 –154. K. Nicolaus, Gema¨lde-Untersucht-Entdeckt-Erforscht. Klinkhardt & Biermann, Braunschweig, 1979. J. Driehaus, Archaeologische Radiographie, 1968, Archa¨o-Physica Bd. 4. Du¨sseldorf, Rheinlandverlag, 112 pp. D. Robakowski, Ro¨ntgenfotographie zur Pru¨fung kulturhistorischer Objekte— Wege zum optimalen ro¨ntgenbild, Maltechnik-Restauro, 93(3) (1987) 32–48. G.N. Hounsfield, Computed medical imaging, J. Comput. Ass. Tomogr., 4 (1980) 665 –674. P. Reimers and J. Goebbels, New possibilities of non-destructive evaluation by X-ray computed tomography, Mat. Eval., 41 (1983) 732 –737. B. Illerhaus, J. Goebbels, P. Reimers and H. Riesemeierr, The principle of computerized tomography and its application in the reconstruction of hidden surfaces in objects of art, Proceedings of the Fourth International Conference on Non-destructive Testing of Works of Art, Berlin, Vol. 1. 1994, pp. 41 –49. M.M. Ter-Pogossian, M.E. Raichle and B.E. Sobel, Positron emission tomography, Sci. Am., 243 (1980) 140 –155. W.-D. Heilmeier, Technical investigation of Roman large bronzes, Fourth International Conference on Non-destructive Testing of Works of Art, Berlin, Vol. 1. 1994, pp. 11–20. W.B. Gilboy and J. Foster, Industrial application of computerized tomography with X- and gamma radiation. In: R.S. Sharp (Ed.), Research Techniques in Nondestructive Testing, Vol. 6. Academic Press, New York, 1981, pp. 255 –287. P. Keil, Fortschritte auf dem Gebiete der Ro¨ntgen-Computer Tomographie, Phys. Bl., 39 (1983) 2–8. H. Reiter, H. Moesta and W. Reinhard, X-ray methods as means of assessment of archeological findings as well as for the revealing of their production process (in German), Proceedings of the Fourth International Conference on NonDestructive Testing of Works of Art, Berlin, Vol. 1. Deutsche Gesellschaft zur Zersto¨rungsfreien Pru¨fung e.v., Berlin, 1994, pp. 75 –84. R.S. Sharpe and R.W. Parish, Engineering applications of microfocal radiography, Microfocal Radiography, Ely, R.V., London, 1980, pp. 43 –81. H.S. Tasker and S.W. Towers, Electron radiography using secondary betaradiation from lead intensifying screens, Nature, 156 (1945) 50–51. E. Mundry, D. Schnitger, J. Riederer, C. Schro¨der, Radiographie und Autoradiographie mit Elektronen. Fourth International Conference on NonDestructive Testing of Works of Art, Berichtsband 45 Teil 2. Deutsche Gesellschaft zur Zersto¨rungsfreien Pru¨fung e.v., Berlin, 1994, pp. 775–786. Ch.F. Bridgeman, S. Keck and H.F. Sherwood, The radiography of paintings by electron emission, Stud. Conserv., 3 (1958) 175 –182. E. Ziesche and D. Schnitger, Elektronenradiographische Untersuchungen der Wasserzeichen des Mainzer Catholicon von 1460, Archiv f. Gesch. d. Buchwesens (AGB)XXI, 5– 6 (1980) 1303– 1359.
UV-, IR- and X-ray imaging 82 83
84 85 86 87
88 89 90 91
R.S. Holt, et al., Gamma-ray scattering techniques for non-destructive testing and imaging, Nucl. Inst. Meth., 221 (1985) 98. W. Niemann and W. Roy Modern X-ray imaging techniques: radioscopy and backscatter imaging. Fourth International Conference on Non-Destructive Testing of Works of Art, Berlin Berichtsband 45 Teil 1. Deutsche Gesellschaft zur Zersto¨rungsfreien Pru¨fung e.v., Berlin, 1994, pp. 21 –30. F. Mairinger, Strahlenuntersuchung an Kunstwerken, E.A. Seemann, Leipzig, 2003, pp. 186–187. K.H. Weber, Die sixtinische Madonna, Maltechnik-Restauro, 90(4) (1984) 9–28. C.F. Bridgeman, Radiography of paper, Stud. Conserv., 10 (1965) 8–17. F. Drilhon, L’Examen radiographique de sculptures en cire du XVIe au XIXe sie`cle. ICOM Comm. Conserv. Seventh Triennial Meeting Copenhagen 1984, 84.1.58 –84.1.61 L. Hatziandreov and G. Ladopoulos, Radiographic examination of the marble statue of Hermes at Olympia, Stud. Conserv., 26 (1981) 24 –28. G.F. Alfrey and K. James, The gamma-ray radiography of decorative plasterwork, Stud. Conserv., 31(2) (1986) 70 –76. V. Vitali, J. Darcovich and W. Williams, Construction of a fudo-myoo sculpture: an X-radio-graphic study, Stud. Conserv., 31(4) (1986) 185 –189. J. Goebbels, J. Haid, D. Hainisch, B. Illerhaus, H.-J. Maltte and D. Meinel, Ancient bronze—a challenge to radiographic technique, Fourth International Conference on NDT Testing of Works of Art, Berlin, Vol. 2. 1994, pp. 733 –742.
71
This page is intentionally left blank
Chapter 3
Electron microscopy and its role in cultural heritage studies A. Adriaens and M.G. Dowsett
3.1 3.1.1
INTRODUCTION Why use electron microscopy?
In the sense of straightforward imaging, the electron microscope allows a sample to be examined at far higher magnification and lateral resolution than is possible with light microscopy. A further advantage is a very large depth of field which, for example, allows rough samples to be studied with the whole surface remaining in focus at one time. However, where cultural heritage materials are concerned, perhaps it is in the huge range of contrast mechanisms, and the possibility of simultaneous imaging and localized chemical analysis where the main advantage of the electron microscope lies. Cultural heritage related studies present a very heterogeneous challenge to the microscopist, and involve materials and problems similar to those encountered in many other areas of materials science—e.g., ceramics, metal alloys, biological materials, manufacturing assessment and corrosion studies. Because of this we have included a fairly complete overview of electron microscope techniques applied to such materials. However, a chapter such as this can only give a brief description of electron microscopy and its application in this area. For more basic information on the technique itself the reader is strongly urged to consult Goodhew et al. [1] and Watt [2]. Applications have been limited to the study of inorganic materials. Itemization is done from the perspective of museum-specific problems. 3.1.2
Imaging with electrons
The human eye provides the ultimate limitation in microscopy. It has a resolution reye of ,0.1 mm—about the diameter of a human hair. This means Comprehensive Analytical Chemistry XLII Janssens and Van Grieken (Eds.) q 2004 Elsevier B.V. All rights reserved
73
A. Adriaens and M.G. Dowsett
that two small objects placed about 20 cm from the eye can just be seen as distinct when they are ,0.1 mm apart. The limitation arises from the intrinsic magnification of the eye, and the separation of the sensing elements on the retina. An optical microscope has a resolution rmic and improves that of the unaided eye in a way which is limited by the wavelength of the light used to illuminate the object. For visible light, this corresponds to rmic ,0.2 mm. There is no point in building a microscope with a significantly higher magnification than reye /rmic—the result would be widely separated but fuzzy images with no improvement in detail. If one allows that 0.1 mm is straining the eye somewhat, a magnification ,1000 is the useful limit for conventional light microscopy [3]. Louis de Broglie [4] established that a wave is associated with any moving particle, and that the corresponding wavelength l is given by l¼
h p
ð3:1Þ
where h is Planck’s constant, 6.626 £ 10234 J s, and p is the particle’s (relativistic) momentum. For an electron with energy E (eV) [1], the associated wavelength is 1:225 l ðnm21 Þ ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi E þ 1026 E2
ð3:2Þ
For the energy range 103 – 106 eV used in modern electron microscopes, l is comfortably smaller than the atomic diameter. The resolution of the microscope is then determined by other factors, such as imperfections in the imaging (aberrations) due to defects in the optical components. Magnifications of (depending on the microscope) 105 – 106 can be achieved allowing single atoms (or even orbitals within them) to be imaged directly at the highest end of the range. In the first place, therefore, electron microscopy allows far smaller detail in an object to be examined than does light microscopy. This benefit comes with some inherent features of the electron microscope, which may be disadvantageous in some circumstances, especially where delicate materials are concerned. Electrons will not travel far through air, and electron microscopes are (usually) vacuum-based instruments—therefore, the specimen must be vacuum tolerant, and not a significant gas source in its own right (but see section 3.1.4). In addition, intense irradiation by electrons can damage or destroy a sample through heating, or other effects. Even in quite good vacua, samples can become coated in the microscope obscuring true surface detail and distorting
74
Electron microscopy and its role in cultural heritage studies
chemical analyses (at 1026 mbar, approximately one mono-atomic layer of gas hits the sample surface per second). Many samples, not least those relating to cultural heritage, are insulating, and charge up under the electron beam. This degrades the resolution, and usually requires some countermeasure such as coating the sample with a conducting layer. However, because of the way that electrons interact with matter, the electron microscope has the ability to do far more than just produce magnified images of the sample. Through secondary electron emission it allows the imaging of topography and major compositional changes (changes of density or atomic number). Via electron diffraction it can provide information on the crystallography and other features of a sample microvolume (e.g., strain). Because electrons can lose characteristic amounts of energy to individual atoms, causing internal excitation and X-ray emission, for example, the chemistry of the microvolume can be determined by using techniques such as energy dispersive spectroscopy (EDS) to measure the emitted X-ray spectrum or electron energy loss spectroscopy (EELS) to determine the energy lost by the primary electron in the interaction. What is more, the chemistry and crystal structure can be correlated with the other electron-imageable properties, depending on the type of microscope used. 3.1.3
Varieties of electron microscopy
There are two basic types of electron microscope (Fig. 3.1). If the detectors are mounted on the same side of the sample as the impinging beam, particles emitted from the front of the sample are detected, and one has a means of characterizing the surface and near-surface regions of a sample of any thickness. This is the basis of the scanning electron microscope (SEM) (Fig. 3.1(a)) and its close cousin the electron probe microanalyser (EPMA). Whereas the design of the SEM usually emphasizes resolution and multi-technique imaging, which results in (typically) a fine electron probe with a small current, EPMA is optimized for the analysis of X-rays characteristic of the elemental composition of the material. The effect on the instrument design is to allow for the use of a higher current broader probe, and more accurate wavelength dispersive spectroscopy (WDS) for the X-ray analysis, in addition to EDS. If the detectors are mounted behind the sample, then electrons and other particles transmitted through a thin section of material can be detected, and the internal details of such a sample can be examined. This is the basis of
75
A. Adriaens and M.G. Dowsett
Fig. 3.1. General types of electron microscope. (a) Scanning electron microscope (SEM); (b) transmission electron microscope (TEM).
transmission electron microscopy (TEM) (Fig. 3.1(b)), and scanning transmission electron microscopy (STEM). A further division in the microscopy can be discerned in the way the magnification is achieved. In the scanning microscopies (SEM, STEM), the optics of the microscope are used to form a sharply focussed electron beam which may be from ,1 mm – ,0.1 nm in diameter. This is scanned across a small square area of the sample, dwelling on a regular array of points known as pixels. More than 106 pixels may cover the scanned area. Typically, the most efficient spacing between pixels is determined by the area around the point of impact of the beam from which secondary particles are emitted (the diameter of the interaction volume – see section 3.2). The signals from each pixel are recorded and stored individually. The magnification is
76
Electron microscopy and its role in cultural heritage studies
determined by the ratio of the size of the displayed image to the scanned area. The resolution is determined ultimately by the probe diameter, but more typically, by the size of the interaction volume. In the TEM, magnification is achieved by using lenses underneath the sample to project the image formed by the transmitted electrons onto a recording device. In this case, the magnification is determined by the optical system and the resolution by the aberrations (imperfections) in the lens performance. The interaction volume plays a diminishing role as the sample becomes thinner, and as the energies of the detected electrons approach that of the primary beam. Most electron microscopy of cultural heritage materials uses SEM, usually with a magnification at the low end of the possible range. The microscopist usually requires complementary information to that available from optical microscopy, perhaps combined with chemical analysis on the micron scale. However, an outline of other modes of operation is included here because of their potential use in specialized applications. 3.1.4
Recent developments in commercial SEM
The increasingly widespread application of the Schottky field emission gun (FEG), originally developed ca. 1969, (e.g., Swanson and Crouser [5]) combined with greatly improved pumping technology has given rise to rapid development of the commercial SEM. Two related areas, highly relevant to cultural heritage materials are described here. As we explain later, the Schottky emitter allows SEMs (and TEMs) to achieve far smaller, brighter foci at the sample than with the older tungsten filament technology, and at much lower beam energies (SEM). Beam diameters quoted at 1.5 nm at 15 kV are not uncommon. First, electron beam energies as low as 1 keV can be used at high resolution in the low-voltage SEM (LVSEM). Even with a conventional electron gun, low-voltage SEM may have some attractions for the examination of cultural heritage materials. The resolution will still be adequate for many purposes, and selection of a (sample-dependent) electron energy somewhere in the 1 –15 keV range such that the secondary electron emission coefficient is $1 can minimize charging problems, thus simplifying sample preparation. Aspects of electron beam damage, especially those due to localized heating, can also be reduced because of the lower power density input. Note, however, that low-energy electrons cannot excite some X-ray transitions which may be required for chemical analysis.
77
A. Adriaens and M.G. Dowsett
The fact that the electron microscope is a vacuum-based instrument has already been mentioned. This restricts the types of sample which can be examined to those which are both vacuum tolerant, and which will preserve the vacuum (which do not, for example contain significant amounts of water, other volatile species, or gas as part of their structure, or through porosity). Conversely, it implies that samples can rarely be examined in their natural state, and that, at best, the chemical composition of the surface may have changed through loss of adsorbed gas. The second development, the environmental SEM (ESEM) is a solution to this problem. First introduced in around 1981 by Danilatos and co-workers [6,7], and commercially available from 1988, ESEM microscopy is a rapid growth area. It has long been known that the primary electron beam is not seriously defocused in SEM by transit through a millimetre or so of gas at pressures approaching 1 atm. Instead, electrons which interact with gas molecules are strongly scattered and form a more or less uniform halo around the beam. The resolution of the microscope is therefore preserved, although there may be an increase in background noise from the samplerelated signals excited by the halo. In the ESEM, therefore, the sample is placed in a separate chamber with its own pumping system. This is linked to a second independently pumped chamber containing the electron column by a pressure-limiting aperture. Detectors may be placed in either chamber, depending on type and the signal detected. The sample is placed close to the aperture (typically 0.5 mm from an aperture diameter 0.5 mm) through which the primary and secondary electrons pass. The microscope column, and the electron gun provide further stages of pumping, which is important because the Schottky field emission source, requires an operating vacuum ,1029 mbar. High vacuum conditions can then be maintained in the microscope column and, to a lesser extent around the detectors, and the electrons need only to travel a few mm in poor vacuum over the sample. A resolution down to 10 nm can be maintained, and analytical techniques such as EDS can be carried out on, for example, wet or hydrated specimens in their natural state [8,9]. Phenomena such as crystallization and dissolution can be observed directly as they happen. Moreover, samples may be analysed in controlled atmospheres. A key advantage is that ionization of the gas over the sample minimizes charging effects, and insulators can be analysed without coating. However, as always, there exists the possibility of electron beam damage which may be increased in some experiments by the presence of species such as water which generate highly reactive radicals under electron irradiation [10,11].
78
Electron microscopy and its role in cultural heritage studies
3.2
THE INTERACTION OF ELECTRONS WITH A SOLID—CONTRAST MECHANISMS
The power and versatility of electron microscopy derives from the variety of ways in which the primary electron beam interacts with the sample, and the fact that the strength of the various interactions is very dependent on the sample’s physical structure, topography, crystallography and chemistry, thus giving rise to contrast between different regions in an image. Many types of secondary or modified primary particle are emitted, and most of them find a use in topographic, structural or chemical characterization. Some of the emission is summarized in Fig. 3.2(a,b). Electrons are very strongly scattered by matter, even at the level of a single atom. The signals that contribute to contrast in an image are, therefore, themselves strong, and this is another reason why the electron microscope can image such small structures. One important effect of the interaction is that the achievable resolution is dependent on the emitted species examined, and is usually worse (often much worse) than the probe diameter or the limitations imposed by aberrations. This is because most of the incident electrons scatter strongly and repeatedly, loosing their energy in a tear-drop shaped volume (Fig. 3.2(a)), typically 1 mm across, exciting secondary electrons and photons in the process. If the sample is thinned (as it must be in TEM), more electrons pass through relatively unscattered, and the effect is reduced (Fig. 3.2(b)). The lateral resolution of EDS and SEM can be strongly dependent on the size of the interaction volume, but techniques such as electron diffraction, and EELS which employ electrons which have elastic, or near elastic scattering
Fig. 3.2. Interaction of electrons with a solid showing effects of interaction volume. (a) SEM sample; (b) sample thinned for TEM.
79
A. Adriaens and M.G. Dowsett
have resolution similar to the probe size (or the aberration limited performance in TEM). Some of the possible interactions and associated contrast mechanisms are summarized below, together with their microscopical or analytical application. 3.2.1
Scattering
An electron is scattered when its direction of travel is changed through interaction with atoms or electrons in the sample (Fig. 3.3). Most individual scattering events are peaked in the forward direction—i.e., the probability of the electron being scattered through a large angle is small. Thus if the sample’s thickness is similar to the mean distance between scattering events (<100 nm typically), most of the electrons injected into the front of the sample will emerge from the back either unscattered, or scattered once or twice through a relatively small angle. This is the emission used in TEM and STEM (Fig. 3.3(a,b)). If the sample is sufficiently thick, multiple scattering will occur, and most of the primary electrons will stop in the sample. However, their interactions will cause significant internal excitation and result in the emission of secondary electrons, and photons from the front of the sample. These are the source of signals in SEM (Fig. 3.3(c)). Some primary electrons may pass sufficiently close to the atomic nucleus to be scattered through a large angle, and out through the front of the sample. Similarly, an electron may suffer a few scattering events with little energy loss, and again emerge through the sample surface (Fig. 3.3(a,c)). These backscattered electrons are also an important source of information in SEM.
Fig. 3.3. Electron scattering and emission. (a) Elastic scattering examples; (b) inelastic scattering; (c) secondary electron emission.
80
Electron microscopy and its role in cultural heritage studies
If the kinetic energy of the scattered electron remains the same before and after the event, the scattering is said to be elastic. Otherwise, it is inelastic. Elastic scattering occurs when an electron is deviated by the Coulomb force due to the combined effect of the charge on an atomic nucleus and the electrons around it, without transferring any energy to the atom itself. Most of the intensity in TEM diffraction patterns comes from this process. Inelastic scattering is responsible for most of the imaging and analytical modes used in SEM and TEM. For example, a primary electron may excite a core level in the atom, losing some keVs in energy, and being scattered through a small angle. This electron is detected and its energy loss (characteristic of the excitation) measured using EELS in the TEM. When the core level de-excites, an X-ray photon may be emitted and its energy or wavelength measured using EDS or WDS. Thus the same basic process provides two complimentary ways of measuring the sample chemistry. 3.2.2
Secondary electron emission
Most of the electron flux emitted from the sample in SEM has energies below 50 eV, and is known as secondary electron emission. These electrons are either primaries which have suffered multiple inelastic scattering or electrons excited from valence levels. Because of their low energies, their range within the sample is small, and their escape depth a fraction of a micron. Secondary electron emission is concentrated around the primary spot, and constitutes a high-resolution source of near-surface information. It has a high depth of field combined with a high lateral resolution (e.g., a surface with up to 4 mm peak-to-peak roughness can be imaged at a resolution of 10 nm) and forms the principal SEM imaging tool through topographic contrast. The secondary electron emission coefficient d increases at first as the angle of incidence u of the primary beam to normal increases: d < d0 =cos u
ð3:3Þ
Surfaces which present different angles to the incident beam, therefore, emit different fluxes of secondary electrons, so that topography is highlighted immediately. The low-energy electrons are easy to collect, and the detector will receive flux from non-line of sight sources on the surface. Details within channels and holes in the sample may, therefore, be resolved.
81
A. Adriaens and M.G. Dowsett
3.2.3
Backscattered electrons
The detection of the higher energy backscattered electrons gives rise to a number of sources of contrast in SEM. This includes topographic contrast, atomic number contrast, and mechanisms which give crystallographic information on very small volumes of material in situ—channelling contrast and electron backscatter diffraction (EBSD). The lateral resolution, typically ,100 nm, is not quite as good as for secondary electrons, because highenergy electrons can escape from greater depths where the lateral extent of the interaction volume is large. However, this is adequate for most cultural heritage purposes. Backscattered electrons are collected where the detector has a line of sight to the point of emission. This fact, combined with the dependence of backscattering on angle of incidence tends to result in topographic contrast which is more absolute (either black or white) than for secondary electrons. However, the line-of-sight detection, combined with the use of segmented detectors enables this source of topographic contrast to be greatly enhanced by subtracting the signals from two segments from one another (see sections 3.3.6 and 3.6.1), or virtually eliminated by adding them. The latter strategy is used to isolate other sources of contrast, such as that due to change in atomic number in the backscattered flux (see, e.g., section 3.7). Atomic number contrast arises because the backscattering coefficient rises monotonically with atomic number of the scattering atoms. The effect is sufficiently large to discriminate between adjacent elements in the Periodic Table and high- and low-density phases in a ceramic body may easily be distinguished, for example. Channelling contrast (Fig. 3.4) is significantly weaker than atomic number contrast, but the spatial resolution may be as high as 10 nm depending on the electron gun used on the microscope and the mean atomic number of the sample. In polycrystalline materials, such as metal alloys, the orientation of the polycrystals to the incident beam will vary depending on the microstructure. Where an open crystallographic direction (a channel) is presented to the beam, it will penetrate deeper into the material making the escape of backscattered electrons more difficult. There are several ways of obtaining diffraction patterns in the SEM of which EBSD is probably the most versatile, although it requires a specially positioned detector. In a FEG SEM the lateral resolution may exceed 15 nm. The EBSD pattern contains a huge amount of information on the local crystal structure of a bulk sample and is an attractive alternative to TEM in many cases. The X-rays produced by the interaction of the electron beam
82
Electron microscopy and its role in cultural heritage studies
Fig. 3.4. The origin of channelling contrast in SEM.
with the sample will also diffract whilst escaping from a crystalline region. The collected X-ray diffraction or Kossel pattern [2] may also have some cultural heritage application. 3.2.4
Cathodoluminescence
Many materials, including minerals and biological tissues emit light when struck by a beam of keV electrons. The colour of the light is characteristic of the local composition. For example, in an insulator, the electron beam excites electron hole pairs into the conduction and valence bands. After a while, these recombine giving rise to photon emission characteristic of the band gap of the matrix, and modified by any impurities present. If the light is collected and analysed, even quite coarsely using red, green and blue filters, chemically related images of the sample can be generated. If the sample is cooled to approaching liquid helium temperature, the rather broad-band emission condenses into a line spectrum which can be used to detect impurities down to 0.01 ppm—far lower than EDS, for example.
83
A. Adriaens and M.G. Dowsett
3.2.5
Core-level excitation and X-ray or Auger emission
In a rare, but useful, interaction a core level in an atom may absorb energy in the range 100 eV– 10 keV depending on its electronic structure, releasing an electron and creating a core level hole. The atom can de-excite (a higher level electron can fall into the hole) releasing an X-ray photon (Figs. 3.3(b) and 3.5), or an Auger electron to carry away the excess energy. Auger electron spectroscopy, the quantitative analysis of a sample using the latter effect is dealt with elsewhere in this book, so we will confine the discussion here to the X-ray emission. The X-ray energy is characteristic of the electronic structure of the atom concerned, and an X-ray spectrum taken from the sample gives a qualitative, or quantitative, compositional analysis of elements present at concentrations .0.1 –1%. This is the basis of EDS and WDS. Elements with atomic number Z $ 4 (beryllium) emit X-rays in this way. The number of possible electronic transitions increases with Z, as does the energy of the deepest core levels. The element-specific X-ray fingerprint spectrum becomes complex as one moves up the Periodic Table. However, the primary electron energy must be greater than that of the core level to excite it, and for efficient X-ray production, the electron energy must be three times that of the level concerned. Which X-rays are observed, therefore, depends on the beam energy. The shortest wavelengths corresponding to deep core levels
Fig. 3.5. Transitions giving rise to X-ray emission for molybdenum (after Goodhew, Humphreys and Beanland (2001).
84
Electron microscopy and its role in cultural heritage studies
in heavy elements are only seen in TEM where the beam energy is 100 keV or more. However, heavy elements can still be detected in EDS in SEM up to 30 keV through lower energy transitions. The basic technique has a spatial resolution around 1 mm because of the effect of interaction volume, but this can be significantly improved by thinning the sample as in TEM (see, e.g., section 3.9). An EDS detector is usually protected from condensation from the vacuum by a window made of low Z material. Unfortunately, this will absorb the softer X-rays making it impossible to detect elements below sodium in the Periodic Table. On the other hand, improved vacua have made windowless detectors more common, and these can go as low in atomic number as beryllium. The WDS dispersion system has no inlet window, and is more suitable for light element detection, although ultimately limited by the window on the detector. EELS is an alternative technique for the assay of light elements. WDS has a superior signal-to-noise ratio, and higher energy resolution than EDS, and is especially useful where X-ray peaks are closely spaced (see, e.g., section 3.6.2.2). On the other hand, its collection efficiency and the parallel collection of all energies (hence speed of analysis) make EDS the most widely used method in electron microscopy. The X-ray fingerprint spectrum provides a rapid method of determining which elements are present in the sample. However, because of the size of the interaction volume and primary electron scattering effects, care must be taken that the data come from the actual region of interest (see, e.g., Chapter 6 of Goodhew et al. [1]) and that spurious peaks do not originate from other parts of the microscope [12]. Quantification of the data is problematic, because the interaction of the emergent X-rays with the sample is dependent on the sample composition itself, so the intensity of the X-ray flux from a particular concentration of an element, depends on the other elements present in that sample. In general, the quantification is handled by the software supplied with the microscope, but it is useful to understand the nature of the corrections involved, and the possible errors which can arise. The quantification method varies depending on whether the sample is thin (as in TEM) or bulk (as in most SEM). In the thin sample, interaction is minimized, and use can be made of the ratio method due to Cliff and Lorrimer [13] where CA N ¼ kAB A CB NB
ð3:4Þ
For any two elements A and B in the sample, CA and CB are the corresponding weight fractions, NA and NB the detected X-ray intensities, 85
A. Adriaens and M.G. Dowsett
and kAB a scaling factor depending on the operating conditions, the detector, and the elements concerned. This is usually determined from measurements of reference materials. In a bulk sample there are three effects which require correction because they modify the X-ray intensity in a matrix-dependent way. (i)
Atomic number correction factor GZ : The primary electron slows down in the sample until it has too little energy to excite X-rays. Its range is Z dependent, and so is the probability that it will scatter without exciting an X-ray. The emergent intensity is, therefore, dependent on the average atomic number in the sampled volume. (ii) Absorption correction factor GA : X-ray intensity is lost through composition-dependent absorption over the escape depth. The correction factor becomes larger for the softer X-rays, and for a higher average Z along the escape path. (iii) Fluorescence correction factor GF : One way in which X-rays are absorbed is through the excitation of core levels in atoms along the path and the subsequent emission of lower energy X-rays. The emission from low Z elements in the presence of high Z elements will be increased, in general, by this process.
It is important that the sample and any reference materials are polished flat and that the angle to the detector is known (see section 3.3.6), because of the path length dependence of the correction factors. If Nsample is the X-ray intensity for a particular element from the sample and Nref its intensity under the same conditions from the reference material then Csample ¼
Nsample GZ GA GF Nref
ð3:5Þ
This so-called ZAF correction system is applied iteratively by the software, until G values consistent with the resulting composition are found. Typical estimates of the quantitative accuracy vary from ^1.5% for WDS to ^6% for EDS, but errors may be much larger for low-intensity peaks, for example. 3.2.6
Electron energy loss spectroscopy
Electrons which suffer some inelastic scattering lose an amount of energy characteristic of the scattering process. In TEM/STEM such electrons are transmitted through the sample and can be dispersed by a magnetic prism
86
Electron microscopy and its role in cultural heritage studies
(see section 3.3.5) to generate a spectrum, whose general appearance is shown in Fig. 3.6. In order of increasing energy loss, the main scattering processes are: a few eV—phonon scattering where the electron excites atomic vibrations, essentially sound waves, in the material, or scatters from such a wave. Up to 50 eV—energy loss to excitation of molecular or low-energy atomic orbitals and plasmon excitation (a plasmon is a wave in the valence or conduction band electrons). Above 50 eV—the excitation of core levels, perhaps with local structure due to lower energy processes. For example, an electron which has ionized an atom which subsequently emits an X-ray photon detected in EDS, will lose an energy close to that of the X-ray. From the point of view of cultural heritage materials, that part of the spectrum which appears above an energy loss of 50 eV is probably of most interest. Contrary to EDS, it provides a way of analysing for low Z elements including lithium at high lateral resolution (1– 10 nm is possible). The peak shape is dependent on the excitation causing the energy loss (K loss peaks have a sharp edge at low-energy loss, with a power law decay whilst L23 peaks for low Z elements are more rounded, e.g.). The peaks also appear on a local background which needs to be removed before quantitative analysis is possible. An additional feature of the EEL spectrum is the fine structure which appears on the low and high-energy loss sides of the peak. On the low-loss
Fig. 3.6. General features of an EELS spectrum for diamond, showing the three regions of energy loss. Note that the EXELFS structure varies significantly from diamond, graphite and C60.
87
A. Adriaens and M.G. Dowsett
side is the energy-loss near edge structure (ELNES). This varies depending on the chemical environment of the element, and can therefore be used to identify, or at least compare different local environments. On the high-loss peak decay region the extended energy loss fine structure (EXELFS) is found. This structure is due to an electromagnetic interference phenomenon, and varies according to the bonding and coordination of the atom. The energy spectrum may be used in one of two basic ways: in the first place it provides a means of chemical analysis which is both complementary and supplementary to EDS. Secondly, the prism may be used in such a way as to allow energy-filtered images to be collected in a STEM or in a TEM (by means of the Omega or Gatan filters). Thus high-resolution elemental mapping becomes possible.
3.2.7
Diffraction in TEM
Electrons which have been elastically scattered through small angles are a rich source of crystallographic information in TEM through diffraction effects, and give rise to several powerful techniques, including immediate cross correlation with EDS (see section 3.6.3). These are seldom applied to cultural heritage materials, despite the fact that very small volumes of material can be studied in comparison with X-ray diffraction (see, e.g., section 3.9). In section 3.1.1, we mentioned that an electron is associated with an energy-dependent wavelength of sub-atomic dimensions at the energies used in electron microscopy. The proof of this lies in the diffraction and interference phenomena which are exhibited by electrons, just as they are by light. In particular, when an electron beam is scattered from the electronic equivalent of a diffraction grating, the scattered intensity is observed at angles u given by Bragg’s law l u ¼ arcsin ð3:6Þ 2d where d is the spacing of the grating. Because l is so small, the planes of atoms in a crystal make a good grating for electrons, and then d is the spacing between the planes. The back focal plane of the first lens in an optical system contains an image in angle (see section 3.3.4) and rays leaving the object (the thin sample in TEM) at particular angles pass through the same point in the back focal
88
Electron microscopy and its role in cultural heritage studies
plane of the objective (Fig. 3.1(b)). If the intermediate lens is set so as to focus this plane onto the detector, a spot diffraction pattern is observed for crystalline samples larger than the beam diameter. (If the sample is polycrystalline on a scale length smaller than the beam diameter, superimposed spot patterns or rings are observed.) Each spot corresponds to a particular plane of atoms (Fig. 3.7(a)). Again, since l is so small, u is also small (typically ,0.58), and so u¼
l 2d
ð3:7Þ
Only those planes of atoms which are nearly parallel to the incident beam will contribute significantly to the diffraction pattern. Unless the final image is energy filtered, the diffraction pattern will be superimposed on a background intensity due to inelastically scattered electrons. This becomes more intense as the sample becomes thicker, but contains valuable information itself. To a first approximation, the spot diffraction pattern can be regarded as a scale map of the reciprocal lattice, and the actual distance between spots is inversely proportional to the actual distance between the planes producing them. Since the effective magnification of the system (known as the camera constant) can be measured, the diffraction pattern identifies the crystal structure and the interatomic spacing. The background of inelastically scattered electrons also contains intensity variations known as Kikuchi lines. These are caused by the diffraction of electrons initially inelastically scattered through particular angles so as to meet the Bragg condition for crystal planes with a particular orientation. The effect is to transfer some high-intensity background to a low-intensity region producing a pair of parallel lines, one dark and one bright. Their separation is the same as the spots corresponding to the same planes (Fig. 3.7(b)). Kikuchi lines allow the exact direction of the incident beam with respect to the lattice to be determined, or particular directions to be set. Diffraction patterns from regions down to 1 mm or so across can be obtained in TEM using a field aperture in an intermediate image between the sample and the detector to determine the area to be viewed. However, much smaller areas may be studied using convergent beam electron diffraction CBED where the primary beam is focussed into a spot #100 nm diameter over a thin, flat region of the sample (Fig. 3.7(c)). More information on this technique can be found in Goodhew et al. [1] and Loretto [14].
89
90 Primary beam
O P
Sample
1
Objective lens
Obj b ective lens
Undiffracted Diffracted from plane 1
Lens
1
(a)
0
Intensity
Diffracted from plane 2
B
Inelastic background
Diffraction pattern
A Scattering angle
2
(b)
(c)
Fig. 3.7. Diffraction in TEM. (a) Spot diffraction pattern; (b) formation of Kikuchi lines; (c) convergent beam electron diffraction.
A. Adriaens and M.G. Dowsett
2
Convergent beam
P'
Electron microscopy and its role in cultural heritage studies
3.2.8
Image contrast in TEM
If the sample plane is imaged onto the detector(s) by shortening the focal length of the intermediate lens(es), an image with a large variety of possible contrast mechanisms is obtained. For example, as the scattering increases with both atomic number and thickness, thicker, or higher density regions of the sample will appear darker on the image (Fig. 3.8). Many diffraction effects are also present in the image, and these can be enhanced by imaging with beams corresponding to particular diffraction conditions, by selecting the angle of incidence of the primary beam, and placing an aperture in the back focal plane of the intermediate lens to control the range of angles contributing to the image (see section 3.4). Although TEM methods have not been widely used for cultural heritage materials so far, it is certainly worth looking at the huge range of information available, especially if high lateral resolution information is required (see, e.g., section 3.6.2.2).
Fig. 3.8. Origin of mass–thickness contrast in TEM.
91
A. Adriaens and M.G. Dowsett
3.3 3.3.1
COMPONENTS AND OPTICS OF ELECTRON MICROSCOPES Basic optics
Charged particles such as electrons can be manipulated using electric and magnetic fields. By creating such fields with a particular symmetry and strength, components analogous to those familiar in light optics can be created, viz. lenses (focusing), prisms (dispersion) and mirrors (deflection) [3,15,16]. In the electron microscope, the beam gains its kinetic energy through acceleration from the source by an electrostatic field and its focus from rotationally symmetric magnetic fields. Scanning and steering are achieved using magnetic fields transverse to the beam path. There are, in addition, many other optical components used in electron microscope columns such as stigmators and filters of various types which are beyond the scope of this chapter. 3.3.2
The electron gun
There are two mechanisms used to source electrons for the microscope: thermionic emission and field emission. The thermionic emission electron gun [17,18] was once the most common, especially in SEM, although FEGs, especially Shottky FEGs [19] are used increasingly in modern microscopes. In the thermionic gun (Fig. 3.9(a)), a sharply pointed tungsten filament surrounded by a control grid or Wehnelt is heated to a temperature of 2700 – 2900 K. Electrons in the metal acquire enough energy to escape through the surface potential barrier into the vacuum and are extracted through a small hole in the Wehnelt by the anode. This type of gun produces a sharp cross-over between the Wehnelt and the anode (or sometimes just beyond the anode) and it is this which is imaged by subsequent lenses to form the final spot. The brightness (current emitted per unit solid angle per unit area) and energy spread of the source are important parameters which limit the current density in, and diameter of, the beam at the sample. The simple thermionic gun has a lower brightness and higher energy spread (several eV) than the FEG which is why the latter is used for STEM, ESEM and LVSEM. However, the filament is cheap, and the gun will operate in relatively poor vacua (,1026 mbar). Improved brightness and lower energy spread (because both the operating temperature and emitting area are lower) can be obtained using an LaB6 emitter in place of the tungsten filament [20], but better vacuum conditions are required in the gun, and the tool is less tolerant of user error.
92
Electron microscopy and its role in cultural heritage studies
Fig. 3.9. Schema of electron guns. (a) Thermionic emission using a tungsten hairpin; (b) the Schottky field emission gun.
Modern microscopes often use either a cold FEG (STEM) or a Schottky FEG (all types) gun (Fig. 3.9(b)). The advantages of these sources are 100 – 1000 times the brightness of the thermionic type, combined with an energy spread of a fraction of an eV. The cold FEG has the ultimate brightness and is used in very high-resolution columns, but is more noisy than the Schottky type which is now replacing the thermionic gun for general purpose applications. FEGs require better vacua in the gun than thermionic emitters, but these are fairly easy to achieve with modern pumping technology. The Schottky field emitter (Fig. 3.9(b)) consists of a sharply pointed tip (radius 0.4 –1 mm) made of zirconium-coated tungsten, surrounded by a grid disc whose function is to suppress thermionic emission from the shoulders of the tip. The tip is heated to ,1800 K and biased at several kV with respect to an extraction electrode. The small tip radius gives rise to a local electric field which is sufficiently strong to pull electrons through the work function of the metal. This process gives rise to an extremely small virtual source, just inside the tip.
3.3.3
Focusing an electron beam
An electron moving through a magnetic field (B-field) experiences the Lorentz force F which is at right angles to the electron’s velocity vector v and
93
A. Adriaens and M.G. Dowsett
the magnetic field B F ¼ 2eðv ^ BÞ
ð3:8Þ
where 2e is the charge on the electron and ^ is the vector cross-product. The effect, as shown in Fig. 3.10(a), is that an electron with an initial velocity at an angle to a uniform B-field moves in a helical (corkscrew-shaped) trajectory. An electron lens in a microscope has a non-uniform field generated by soft iron pole pieces within a current-carrying coil (Fig. 3.10(b)). The strength of the lens depends on its geometry (fixed) and the current (variable). An electron travelling exactly along the axis of the lens experiences no force because its velocity is always parallel to the field. Offaxis, however, the helical motion described above is established, giving rise to a tangential velocity vt. This interacts with the axial component of the field Bz to produce a radial force reducing and reversing any radially outward velocity component, and giving a focusing effect (Fig. 3.11). If one considers the overall action of the lens on a beam rather than a single electron, it is clear that there is a net rotation as the beam passes through the lens, and so images will be rotated as well. This has important consequences in TEM if images taken with lenses run at different strengths are to be compared (e.g., diffraction and field images) because the orientation of the images may be different. Note that the focal length of the lens depends on B2 so that lenses are always converging, irrespective of the direction of the field. Only the
Fig. 3.10. (a) Electron trajectory in a uniform magnetic field; (b) in the non-uniform rotationally symmetric field of a lens.
94
Electron microscopy and its role in cultural heritage studies
Fig. 3.11. Focusing and rotation effects in the lens field.
sense of the rotation is altered. In modern microscopes, therefore, combinations of lenses can be used so as to cancel out the rotation. 3.3.4
Newtonian lens model
Ignoring the rotation, the action of the magnetic lens may be understood using the Newtonian optics applied to the paraxial operation of glass lenses, provided one limits oneself to objects and images outside the field of the lens (lenses where the object or an intermediate image are within the lens (immersion lenses) are common in electron microscopy, and the reader is referred to Chapter 2 of Joy et al. [12] for a more detailed treatment). So, cardinal planes may be defined for a magnetic lens, in an analogous manner to a glass lens, especially (with reference to Fig. 3.12) focal planes F1 and F2 and principal planes H1 and H2, where subscripts 1 and 2 relate to image and object space, respectively. The focal lengths of the lens, f1 and f2 are given by P1F1 and P2F2, respectively. Since we are considering the object and image to be outside the field of the lens, f1 ¼ f2 ; for completeness, we indicate that the principal planes are crossed (i.e., unlike the norm in light optics, the object principal plane is on the image side and vice-versa, but the
95
A. Adriaens and M.G. Dowsett
Fig. 3.12. Geometrical optics of a simple lens also showing the effect of apertures in different planes. The aperture at A in the back focal plane and any subsequent images of this will limit the range of angles (and, therefore, that part of the diffraction pattern) which contributes to the field image. The aperture at B and any subsequent images of it will determine the field of view.
main point of the geometric construction is to show that there are two types of image formed by (any) ideal lens. At the second focal plane (or back focal plane) of the first lens in an optical system, and any subsequent images of this, all rays emitted from the object at a particular angle pass through the same point. In electron microscopy, especially TEM, this has particular significance because it is where the diffraction image is located. For an object located a distance x1 outside the first focal plane, a real image is created a distance x2 beyond the second focal plane, with a magnification M where x f M¼2 2 ¼2 x1 f
ð3:9Þ
Thus, in TEM, e.g., by changing the focal length of the lens system below the sample, one may select for either a diffraction image, or a field image to be in focus at the detector. Real lenses depart from the ideal behaviour described above, first because they are used at finite aperture, and secondly because they may have mechanical imperfections, and finally because the beam itself will have a finite energy spread. Any non-ideal behaviour is generally referred to as an aberration. For example, rays entering a lens at different angles from a point source on axis are not all focused at the same point (Fig. 3.13(a)), a phenomenon known as spherical aberration. Rays entering the lens along the same trajectory, but with different energies, will also be focused at a
96
Electron microscopy and its role in cultural heritage studies
Fig. 3.13. Examples of lens aberrations. (a) Spherical aberration; (b) chromatic aberration.
different point (Fig. 3.13(b)). This is known as chromatic aberration. For a complete elementary treatment of aberrations, the reader is referred to Chapter 7 of Grivet et al. [15]. The apertures in an electron microscope play a crucial role in determining the beam and image quality. Simple examples of aperturing are shown in Fig. 3.12. For example, an aperture located in the second focal plane of the first lens in an optical system will control the maximum angle of ray admitted to the image (and hence, the spherical aberration) (or which part of the diffraction pattern contributes to the image in TEM) whereas an aperture located in an image plane will determine which part of the object is imaged by subsequent lenses. In practice, apertures are found in other places and used for many other purposes as well, especially to prevent scattered electrons reaching the sample, or the detectors. 3.3.5
The magnetic prism
A glass prism disperses light according to its wavelength to generate a spectrum, with short wavelengths (higher energy photons) being dispersed the most. Thus, the contribution of different wavelengths to an incident beam can be measured. A magnetic prism performs the same function for an injected electron beam, except that the dispersion is opposite (high-energy electrons have their trajectories bent the least). This is how the energy spectrum of electrons which have lost small characteristic amounts of energy to atoms in a solid may be measured as in EELS, for example. The uniform field is the simplest example of a magnetic prism (Fig. 3.14(a)), but in practice real prisms are more sophisticated with inclined entry and exit faces, and shaped pole faces in some cases, to control the focusing and aberrations (Fig. 3.14(b)). In this way, an energy resolution ,1 eV can be
97
A. Adriaens and M.G. Dowsett
Fig. 3.14. The magnetic prism. (a) Uniform field neglecting edge effects; (b) schematic of prism design for EELS showing angular focussing.
achieved. The spectrum is collected in older instruments by scanning the magnetic field to disperse different energies through a slit onto the detector. A more efficient method uses a parallel detector to collect a significant part of the required bandwidth simultaneously (parallel EELS or PEELS). 3.3.6
Detectors
The processes by which electrons or X-rays emitted from the sample are converted to viewable images or electrical signals are similar to those occurring in the sample itself. Some of the main types of detector used in SEM and TEM are described here. Most SEM images are collected using an Everhart– Thornley (E – T) detector (Fig. 3.15). Under the primary electron bombardment, the sample emits low-energy secondary electrons, and higher energy backscattered electrons (see sections 3.2.2 and 3.2.3). The E –T detector can collect the former by means of a weak electric field between the detector’s outer grid and the sample, and also detects any line of sight backscattered electrons. If the external field is reversed, only the backscattered electrons are detected. Internally, the detector consists of a scintillator biased at about þ10 kV with respect to the grid. This is mounted on the end of a light conducting pipe, coupling the scintillator to a photomultiplier tube outside the microscope vacuum. Secondary electrons passing through the grid are accelerated to
98
Electron microscopy and its role in cultural heritage studies
Fig. 3.15. The Everhart –Thornley detector principally used for secondary electron detection.
,10 keV and strike the scintillator, causing it to emit visible photons. These are conducted to the photomultiplier and converted to pulses or a current. The output from the photomultiplier will respond very fast to changes in input flux, so the E – T is an ideal imaging detector for SEM. Output from the detector can be digitised and placed in computer memory, at a location synchronized with the beam scan, or used to modulate the intensity of the beam of a cathode ray tube, which is scanned in the same was as the primary electron beam. In the TEM, images can be produced directly by projecting the image of transmitted electrons using the lenses between the sample and a phosphor screen, very similar to that used in a black and white television. This can be viewed from above, and photographed (in older microscopes) from below using fine grain film. In more modern instruments, the image can be captured by a CCD camera coupled to the screen by a fibre-optic block. Large area solid state detectors which operate in a similar way to the lithium-drifted detector described below are used to detect backscattered electrons in SEM and forward scattering in STEM. They consist of thin semiconductor wafers mounted on the pole-piece of the SEM objective (Fig. 3.16(a)), or below the sample in STEM. In SEM the detector may be in the form of an annulus split into four quadrants (Fig. 3.16(b)). The signals from the quadrants are collected separately and may be recombined arithmetically to enhance different contrast mechanisms. For example, subtracting the signals from opposite quadrants can enhance topographic contrast, whilst adding them can enhance chemical contrast.
99
A. Adriaens and M.G. Dowsett
Fig. 3.16. (a) Solid state diode backscattered electron detector mounted on the pole piece of the SEM objective. (b) The detector may take the form of an annulus split into four quadrants. Then A þ B þ C þ D gives chemical contrast, and A –C and B –D give topographic contrast.
X-ray detection in electron microscopy is done with detectors of two different types [21]. In EDS, a solid-state device—a lithium drifted silicon diode—subtending as large an angle as possible at the sample is used (Fig. 3.17(a)). This detects all X-ray energies in parallel and outputs electrical pulses whose height is proportional to X-ray energy to first order. The collection efficiency of this device is high because of the large area, but it has both higher background noise and poorer energy resolution than the WDS spectrometer. However, modern digital signal processing and deconvolution techniques, combined with the speed of parallel detection make the EDS very attractive.
Fig. 3.17. Lithium-drifted silicon diode used for EDS. (a) Possible detector geometry. Angles a and b must be known accurately for quantification; (b) detector schematic.
100
Electron microscopy and its role in cultural heritage studies
The device is made from a thin piece of silicon (or in some cases germanium) (Fig. 3.17(b)). The front and back surfaces are coated with gold as a contact. The front layer is very thin ,20 nm so that its X-ray absorption is minimized. In the top of the silicon is a 2 – 3-mm thick region doped with lithium (n-type) and boron ( p-type) to the point where there are approximately the same numbers of donors and acceptors (so-called intrinsic material). Towards the back, the n-doping is allowed to dominate, producing an internal junction. A potential difference is applied between the front (cathode) and back (anode) contacts so that the junction is reverse biased and an internal electric field exists across the intrinsic region. When an X-ray enters such a material it is absorbed by an atom, exciting a photoelectron from an inner shell. This energetic particle slows down in the material, creating electron –hole pairs. These drift apart under the influence of the field and are collected as a charge pulse on the contacts. The number N of electron– hole pairs created by an incident X-ray of energy E (per eV) and, therefore, the pulse height, is given by N<
E 3:8
ð3:10Þ
where 3.8 eV is the average energy for pair formation in silicon. The detector must be operated at 77 K, both to reduce noise, and to protect the lithium structure. Adsorption of contamination onto the front surface seriously degrades the performance, so the device is normally operated in a good vacuum behind a thin (7– 12 mm) beryllium window. Unfortunately, this absorbs low-energy X-rays and is essentially opaque below energies of 1 keV. This limits the detector to elements from sodium onwards in the Periodic Table. Windowless, or Mylar windowed detectors allow detection down to beryllium, but the combination of the gold layer, and a “dead” layer of silicon beneath (,100 nm) seriously reduce the efficiency (boron K a is detected at only 10% of the efficiency of sodium K a) and prevent the detection of lower energy emission. In WDS, a small solid angle of the X-ray emission from the sample is admitted to a crystal spectrometer through a narrow slit (Fig. 3.18). The surface of the crystal acts as a diffraction grating for the X-rays, dispersing them according to wavelength l via Bragg’s law l¼
2d sin u n
ð3:11Þ
where d is the lattice spacing in the crystal, and n ¼ 1; 2; 3; … is the order of diffraction.
101
A. Adriaens and M.G. Dowsett
Fig. 3.18. Crystal spectrometer for WDS. The sample surface, the crystal surface and the detector slit lie on the dash dot locus known as the Rowland circle.
A gas-filled ionization counter is used to detect the diffracted intensity. To measure different wavelengths, the detector and the crystal must be moved differentially, whilst maintaining a very accurate relationship between their respective positions and that of the sample. The sample height also needs to be set accurately, and samples must be polished flat to better than 1 mm. The device then collects data sequentially, rather than in parallel as in EDS. The spectrometry is free from the background signals generated in the EDS detector, and the resolution is better. The window on the gas-filled detector still cuts the detection efficiency for low-energy X-rays, unless an organic film supported by a grid is used, in which case the beryllium Ka line (11.4 nm) can be detected. 3.4
SAMPLE PREPARATION TECHNIQUES
Sample preparation techniques depend critically on the type of information that is sought. SEM applications, for instance, can involve the analysis of asreceived surfaces, as is usually the case for the study of weathered or corroded surfaces. For these studies minimal or sometimes no sample preparation will be necessary. An important element is to take care that the samples are electrically conducting. Objects made, e.g., of ceramic, glass,
102
Electron microscopy and its role in cultural heritage studies
stone or flint, will need special attention in this sense. Charge compensation in conventional SEM instruments is usually done by sputter coating the sample with a very thin layer of carbon. For low-voltage instruments, or in the ESEM, however, it may be possible to analyse these non-conducting samples without any surface preparation (see section 3.1.4). When analyses need to be performed within the bulk of a sample, the sample is often embedded in epoxy, ground and polished. The latter is done by mechanical polishing in sequential steps of decreasing grain size. Typically one starts out with SiC paper which covers a grain size from 180 to 4000 mesh /inch. In a second stage polishing cloth is used with diamond spray. In particular for quantitative EDS analyses, the mechanical polishing is an essential step in the sample preparation. The maximum size and shape of samples is limited by the size and geometry of the sample chamber. Typically this is of the order of 2.5 cm in diameter. The depth can be adjusted. ESEM instruments sometimes have a relatively big environmental stage so that an entire object, e.g., a statuette, can be fitted as a whole but needs to be oriented so that the area of interest can be analysed. On the other hand, in many cases it is advisable to work with very small samples. The minimum sample size will depend on the question that needs to be addressed but can be of the order of 1 mm3. Bronk and Freestone [22], for instance, report on the development of a microsampling technique for glass objects. The method, which can easily be carried out in a museum, makes use of diamond-coated grinding files to make a scratch in the glass object. Abraded particles from the object can be collected and consequently selected using an optical microscope. Soft tissues will not be discussed here, but may undergo degradation through radiation damage. For TEM analysis, the sample preparation is much more elaborate as one aims at obtaining samples with a thickness of the order of tens of nanometres. In addition the sample must be representative of the entire sample in structure and composition. Depending on the type of material, the technique used to make ultra-thin samples will vary. Basically it includes the removal of unwanted material by chemical or mechanical means and the cutting of the sample to the desired thickness. Goodhew et al. [1] give a comprehensive overview of the most commonly used techniques: electropolishing, mechanical polishing, ion and atom milling, focused ion beams, cleaving, ultramicrotomy and replication. A specific method for the preparation of stratigraphic sections of paint fragments is described by San Andres et al. [23]. Their methodology involves three main stages and is rather similar to the preparation of biological material: dehydration of the paint fragment (1 £ 1 £ 0.5 mm3) in a stove, impregnation and inclusion in a
103
A. Adriaens and M.G. Dowsett
stable epoxy resin, and cutting of the embedded sample using a microtome. The sections are placed on a mesh grid for analysis. 3.5
ORIGIN/ PROVENANCE STUDIES
Provenance studies involve characterizing and locating the natural sources of the raw materials used to make artefacts, therefore contributing to establishing patterns of raw material procurement, trade or exchange. Besides using stylistic arguments, provenancing of artefacts is usually done in two ways. One way is by comparing the composition of the artefact to resources with a known origin. Alternatively the composition of the artefact can be compared to that of a control group of artefacts of which the origin is already established. 3.5.1
Ceramics
In the case of ceramics, either the chemical or the mineralogical composition can be compared with the geological source, but it is especially in the latter case that the potentials of SEM – EDS or EPMA surface. Chemical characterization mainly implies determining trace elements, as their presence and quantity, in particular, are usually very specific for a given geological origin of the raw materials. The contribution of EPMA and SEM – EDS are limited in this case and commonly used techniques for trace element analysis in ceramics include neutron activation analysis (NAA), inductively coupled plasma mass spectrometry (ICP-MS) and X-ray fluorescence spectrometry (XRF). When the mineralogical composition is to be compared, petrological data need to be obtained that relate a ceramic fabric to its geological source. Here, in general, the standard petrological approach remains the preferred technique to be used: thin sectioning provides information on the mineralogy of the pottery’s non-plastic inclusions, which can be very characteristic for a certain ware and its place of origin. The application of electron microscopy provides useful supplementary information, such as the identification of unknown minerals, very small minerals or minerals with ambiguous optical properties [24]. In the case of North Mesopotamian metallic ware, for instance, one variety of this pottery group, possibly representing a specific production centre, is characterized by the presence of small potassium feldspar inclusions [25]. In thin section these inclusions can be identified as feldspars, but because of their small size no further specification is possible. Using electron microscopy, the small inclusions become clearly visible and
104
Electron microscopy and its role in cultural heritage studies
Fig. 3.19. BSE image of “metallic ware” from Tell Beydar (Syria), showing potassium feldspar inclusions (light grey) and quartz inclusions (dark grey).
are easily identified as potassium feldspar when their elemental composition is revealed by EDS (Fig. 3.19). Freestone and co-workers [24,26] give a comprehensive discussion about the potential of SEM – EDS for mineralogical inferences from non-plastic inclusions. In addition to aiding mineral identification, SEM can also extend the application and power of textural analysis due to its automated image analysis facilities [24]. Contrary to optical microscopy, these analyses can include the characterization of high-temperature mineral phases next to non-plastic inclusions. 3.5.2
Glass
Similarly for glass objects, the chemical composition will allow the origin to be identified, provided a compositional classification has been previously established with a significant number of well-authenticated glasses made in certain manufacturing centres or during certain historical periods [27]. The analytical procedure to determine the composition of ancient glasses is complicated in the sense that a large number of elements must be analysed, again most of which are trace elements. Also here SEM –EDS and EPMA contribute to only part of the information and it is necessary to complement the analyses with techniques that can attain lower detection limits. In work by De Raedt et al. [28,29], the authors report on a study of fac¸on-de-Venise
105
A. Adriaens and M.G. Dowsett
glass excavated in Antwerp and aim to resolve the question whether the vessels had been manufactured locally or had been imported from Venice. The results of the EPMA analyses in combination with multivariate statistics allow the classification of objects into various groups that clearly link composition with typology but cannot give a definite answer concerning their origin. Additional analyses using synchrotron radiation XRF (SR-XRF) and laser ablation ICPMS (LA-ICPMS) show that only differing levels of trace elements such as Zr, Hf and some rare earths can highlight a difference in silica source, hence differentiating between local and Venetian production. Another main limitation of SEM – EDS and EPMA is their low sensitivity to light elements, which may be present in ancient glasses. Here a windowless EDS spectrometer would be useful. Advantages of the technique lie in the fact that composite glasses can be analysed as well as thin layers that are applied on other materials. 3.6
TECHNOLOGY AND TECHNIQUES OF MANUFACTURE
A large part of SEM-aided examination of artefacts is taken up by technological studies, which involve the identification of the materials and production techniques used. The information sought in answering questions with regard to technology very much depends on the type of material that is being studied. 3.6.1
Seals
Sax and Meeks [30] have studied the methods, which have been used for engraving Mesopotamian cylinder seals. While the seals themselves were examined by optical microscopy, SEM was used to observe experimental seal impressions of the artefacts under investigation. The advantage of using impressions is that they facilitate the observation of the depth of the engraved features and of details within these features. Even though no elemental information is required in this case, backscattered electron images were used for best contrast and visual display of depth (Fig. 3.20). Four basic techniques of lapidary engraving were recognized, all having some characteristic features, some observable using optical microscopy, others using SEM. A temporal progression was noticed from the use of relatively simple microchipping techniques through the use of sophisticated wheelcutting techniques. The criteria for recognition of these techniques were recently confirmed by experiments [31].
106
Electron microscopy and its role in cultural heritage studies
Fig. 3.20. BSE images of features drilled on rock crystal and chalcedony with flint and copper tools and comparative seals [31]. (a) Feature drilled on rock crystal with hand-held flint tool and olive oil /quartz paste (40 mesh). (b) Head of figure on rock crystal seal, Jemdet Nasr, c. 3000 BC . (c) Feature drilled on Chalcedony, using drill press at c. 1200 rpm with copper tool and olive oil/emery paste (220 mesh); the flattipped drill was not reshaped. (d) Feature on agate seal, Mitannian, c. 1300 BC . Reproduced with permission from RLAHA, Oxford.
3.6.2
Ceramics
Comprehensive overviews that describe the role of electron microscopy in technological studies of ceramics include papers by Tite et al. [32] and Freestone [24]. SEM – EDS has proven to be an invaluable tool in contributing answers to various questions involving technological aspects, even though not all steps in the pottery production sequence are as appropriate for investigation. Three aspects of pottery production are particularly well suited: (i) identification of the mineralogy of the body, which is the clay part of the ceramic (as opposed to slips and glazes), (ii) the resources and techniques used for surface coatings (slips, paints, glazes…) and (iii) the firing technology. However, SEM – EDS, like any other analysis method, often needs to be complemented by additional
107
A. Adriaens and M.G. Dowsett
techniques in order to develop a comprehensive image of the production technology. Both fresh fracture surfaces as well as resin-impregnated polished sections can be analysed when using SEM. The advantage of using fresh fractured surfaces is, in the first instance, the minimal sample preparation but also the fact that ceramics tend to fracture along the glass phase when it is present, thereby making this phase easily visible [32]. The disadvantage of fresh fractured surfaces is that they do not provide a representative picture of the distribution of inclusions within the clay/glass matrix; so in some cases it is worth taking the time to prepare polished sections. Moreover, if quantitative elemental analysis is planned using an attached X-ray detector, the sample requires a flat surface for precise measurements. 3.6.2.1 Determination of the composition of body’s raw materials The characterization of used resources has already been discussed with respect to provenance studies, but it also aids in understanding the working behaviour of the raw materials and, therefore, it also plays a role in technology studies. The preferred technique to identify and describe non-plastic inclusions within the clay matrix is polarizing microscopy. The clay minerals themselves, however, are too small to be observed by optical microscopy, but can by identified by X-ray diffraction (XRD). It must be noted here that once a certain temperature is reached during the initial firing of the pottery, these clay minerals will decompose. The edges of the grains will bond together by ion diffusion (sintering) and eventually vitrification will occur and new high-temperature minerals will form during cooling. This means that fairly unaltered clay minerals will only be present in low-fired ceramics or in ceramics made using high refractory clays. In these cases SEM can be used to identify the clay minerals by their shape and size [33,34]. However, a technique more often used for the identification of clay minerals in low-fired ceramics is the establishment of the elemental composition of the clay fraction using EPMA [32]. Small areas of the clay matrix where non-plastic inclusions are absent are analysed. The resulting composition can then be compared to that of standard clay minerals such as those given by Deer et al. [35], or the Si:Al ratio, an indicator of the clay type, can be compared to those given by Newman [36]. Since many ceramics are fired above the temperatures at which clay minerals start breaking down, the composition of the original raw materials can no longer be directly identified. However, the newly formed hightemperature silicates, present within the glass phase, can sometimes be
108
Electron microscopy and its role in cultural heritage studies
Fig. 3.21. BSE image of a ceramic waster from Tell Beydar (Syria), showing diopside crystals in unreacted glass phase.
identified using SEM. Their identification can already reveal information about the original raw material. For instance, the presence of diopside, a high-temperature calcium silicate, indicates that calcareous resources were used (Fig. 3.21). For a more precise determination of the raw material’s composition, SEM –EDS can be used to analyse the unreacted glass phase, excluding the newly formed high-temperature silicates. Tite et al. [32] have noted that especially in highly fired non-calcareous clays an apparently continuous glass phase can reveal multiple phases after etching of the samples. Finally, when the composition of the unreacted glass phase is established, it can again be compared to that of standard clay minerals. 3.6.2.2 Investigation of surface coatings EPMA and SEM – EDS are also used in the study of ceramic surface coatings such as slips, washes, paints and glazes. By analysing the applied surface layer in cross-section, information can be obtained about its structure. This can be done using fresh fractures, or thin sections, in which case the surface layer/ body interface is more clearly revealed [32]. Figure 3.22(a) shows a secondary electron image of a cross-section of Samarra fine ware (Tell Baghouz, Syria) [37] which is characterized by a dark paint on a light coloured background (Fig. 3.22(b)). EDS analyses show that the pigments used for painting are forms of iron oxide. The dark colour of the decorations in combination with the pale surface is an indication of firing under neutral
109
A. Adriaens and M.G. Dowsett
Fig. 3.22. (a) SE image of a cross-section of Samarra fine ware. EDS spectra of body and slip are superimposed; (b) example of Samarra fine ware [37]. Reproduced with permission from CNRS Editions.
or slightly reducing circumstances at sufficiently high temperatures to cause the pigment to sinter or melt. As a result the painted parts will vitrify more than the underlying clay body, which can be easily seen in Fig. 3.22(a) [37]. In an EPMA-based study, Jacobs has tried to explain why Old Babylonian and Kassite ceramics from Iraq are characterized by a pale colour, even
110
Electron microscopy and its role in cultural heritage studies
though they contain a considerable amount of iron [38,39]. Although these ceramics have no specifically applied surface layer, EPMA allowed studying the differences between body and surface. Based on the analysis of both artefacts and test bars with known composition, fired at different temperatures, the author concluded that several factors play a role in the colouring of the ceramics, namely the presence of calcium in the clay, the salt content of the clay and the firing conditions. A study of paint layers on Greek and especially Mesopotamian ceramics has been carried out by Noll and co-workers [40– 43]. They describe in detail the principles of several ceramic painting techniques such as hematite red, black by iron reduction, manganese black, carbon black and several dichrome and polychrome painting techniques. Based on the study of samples from a broad region and a variety of periods, the authors try to trace back these techniques to where they might have been “invented.” They showed that the three-stage firing cycle used to produce Attic red- and blackFigure vases was already used in the 5th millennium BC in the production of polychrome Halaf pottery. Many studies of ceramic glazes and other vitreous materials are SEM – EDS and EPMA based [44 –51]. In a recent study based on the analysis of ceramics from the Near East and Egypt, Mason and Tite [52] discuss the origin and development of tin-opacified glazes. Polished sections of both the interior and exterior of the vessels were examined. BSE images were generated for distinguishing the phases, while the bulk chemical composition of the glazes and the composition of crystalline phases within the glaze was determined using EDS. However, for measurement of tin WDS was used, since adjacent spectral peaks of potassium may result in spurious tin results. For the same reason, lead was also measured by WDS. The analyses showed that tin was first used in glazes during the first half of the 8th century AD in Iraq. This development is situated in the tradition of pre-Islamic opaqueglaze technology, since the “traditional” opacifiers such as air bubbles, quartz, feldspar, wollastonite and diopside are still found in the first phase of development of tin-opacified glazes. Tin oxide was first applied as a surface coating to the body. Later in the 8th century it was dispersed throughout the thickness of the glaze. Only in the 9th century was the concentration of tin oxide disposed through the glaze was sufficiently high (5– 8% SnO2) for no other opacifying agent to be required. In a study of 10 high-quality luxury wares, Kingery and Vandiver [46] use TEM for the analysis of very fine particles and emulsions on Song Dynasty jun ware and Kashan lusterware. The basis of translucence in jun glazes is the formation of an emulsion of two different liquids in the glaze.
111
A. Adriaens and M.G. Dowsett
˚ across, Since droplets in the emulsion of a jun glaze are in the order of 800 A only a powerful microscope such as the TEM can resolve them. The investigators observed that emulsion formation requires a temperature at or below 12008C, while above this temperature the jun compositions form a clear glaze. If the glaze is cooled too rapidly, the typical blue cloudy quality will not be achieved. 3.6.2.3 Study of firing technology As we mentioned before, the internal mineralogy of a ceramic is usually altered during its firing and subsequent cooling. The newly formed hightemperature phases are not only a derivative of the kind of clay that was used, they are also an indication of the applied firing temperature. These phases can only form within a certain range of temperatures. Therefore, the presence or absence of certain mineral phases can be used to establish minimum and maximum limits for the initial ceramic firing. However, these newly formed minerals are usually only clearly visible in high-fired ceramics (Fig. 3.21). In pottery where the temperature at which these phases form is exceeded only to a minimum extent, or during only a short period of time, the new crystals will be too small to be observed. In these cases, XRD remains a better-suited tool for mineral phase determination. Or alternatively, TEM would be able to contribute here by electron diffraction and EDS on ,10 nm particles. The extent of the glass phase increases with increasing firing temperature and provides a useful parameter for estimating the firing temperature. This approach has been explored, in particular, by Tite and Maniatis [53], Maniatis and Tite [54] and Tite et al. [32]. In practice, this is done by comparing the ceramic’s microstructure in SEM with samples of similar composition refired at known temperatures. Figure 3.23 shows progressive changes in the degree and texture of vitrification. The precision of this method is of the order of 50 – 1008C, which is comparable to other methods that are commonly used for estimating the firing temperature. These involve the study of macroscopic properties such as porosity, hardness and thermal expansion and are, in fact, also dependent on the extent of vitrification. SEM measurements, however, are less precise in regions where the vitrification is stable (850 – 10508C for ceramics made from calcareous clay) or changing only slowly [32]. It is important to note that the way a clay body reacts when it is fired not only depends on temperature and the presence of fusing agents (composition), but also on factors like heating rate, firing atmosphere, grain size, material expansion rate and others [55]. Therefore, it may be desirable to check the estimated temperature with a single refiring, in order to ensure
112
Electron microscopy and its role in cultural heritage studies
Fig. 3.23. SE images of a clay matrix showing the progressive changes in the degree and texture of vitrification as a function of temperature. (a) 6508C; (b) 8508C; (c) 10508C. Magnification 2000.
that factors other than chemical composition have not affected the vitrification of the ceramic significantly [24]. Alternatively, refiring of the pottery under investigation can be done over a range of temperatures. Only when the initial firing temperature is exceeded, will a change in vitrification occur. Again, this technique is restricted when the initial firing temperature was within a temperature range where the vitrification is stable. 3.6.3
Glass
Barber and Freestone [56] report on the TEM investigation of the Lycurgus cup, a decorated late Roman cut glass cage cup. Their aim was to determine the origin of the colour, which appears a deep wine-red in transmitted light and an opaque green in reflected light. Earlier studies on the same cup had already shown the presence of finely dispersed particles. Using electron
113
A. Adriaens and M.G. Dowsett
Fig. 3.24. (a) TEM image of a silver–gold alloy crystal. (b,c) electron diffraction patterns from the crystal recorded by tilting to obtain strong Bragg spots [56]. Reproduced with permission from RLAHA, Oxford.
diffraction and EDS spectra, the new results showed the presence of both non-metallic (NaCl and CaPO4) as well as metallic particles (Ag/Au), the latter causing the dichroic effect (Fig. 3.24). 3.6.4
Metals
Electron microscopy and more specifically SEM – EDS or EPMA are routinely used for providing insight into metal-working technology. In most cases it is used as a complementary technique to conventional metallography using reflected light microscopy. Electron microscopy, in the first instance, contributes in providing information on the nature of the metal or alloy
114
Electron microscopy and its role in cultural heritage studies
employed to make the object, which is reflected in its chemical composition. For example, many debased silver objects are made from silver-copper alloys. Both constituents can be seen in polished and etched sections using light microscopy, but often it may be very difficult to obtain an idea of the composition this way [57]. Here additional evidence must be obtained using a suitable analytical method. This can in principle be done using various analytical techniques. The most cited are atomic absorption spectroscopy (AAS), ICPMS, XRF and also SEM –EDS . The advantage of SEM – EDS and XRF in comparison to the other techniques is that the sample does not need to be dissolved which is usually a time-consuming procedure and which at the same time involves the risk of contamination. SEM –EDS furthermore has the additional benefit over XRF that the microstructure also can be studied in more detail. As in conventional metallography this provides information on the manufacturing process used to produce the object. The shape and size of the grains assist in determining, for instance, whether an object was cast into a mould or worked to shape by hammering and annealing. One of the main difficulties with the study of the microstructure of ancient metals lies in problems associated with the selection of the sample and the actual sampling of the object. In this case basically the same criteria apply as for metallographic sampling, in the sense that the samples need to be representative of the object as a whole. Also the microstructure should not be altered in the process of removing the sample and the orientation of the sample in relationship to the object should be carefully recorded [57]. SEM applications for metals analysis are numerous. One of the principal areas of metal research has been in elucidating the earliest forms of metallurgy. The main attention here is focused on knowing which alloys were used. For instance, for the copper-based alloys in the Near-East and Europe, it has been possible to trace the chronological development of alloy types from copper to arsenical copper to tin bronze, leaded bronze and brass during later periods [58,59]. SEM – EDS or EPMA proves to be useful here in providing information on the chemical composition and phase analysis of the metallic remains that are excavated [60 – 63]. In addition SEM is aimed at obtaining a better insight into the smelting and mining techniques used. This is usually achieved by trying to obtain an overall picture of the metallurgical site. Chemical analyses can contribute in determining the chemical composition of the metallurgical debris that is found on the archaeological site, such as slags (metallurgical waste material) and crucible accretions [64,65]. The analysis of the metallurgical debris using EPMA allows addressing questions such as the composition of the
115
A. Adriaens and M.G. Dowsett
entrapped metal, and also the fact whether metals were smelted from ore or just remelted from existing metal objects or refined. In addition simulation experiments are often performed, of which the products are again analysed with SEM – EDS (Fig. 3.25), to test potential hypotheses [66 – 68]. Other studies concentrate on revealing the technology behind the decoration process of metallic objects, in other words how objects were gilded, silvered or tinned. In a study by Meeks [69], for instance, the author makes a thorough examination of the various technologies behind tin-rich surfaces on bronze objects. The paper attempts to identify the different microstructures associated with different process of tinning. SEM imaging and X-ray analyses in conjunction with metallography are used to compare the results of a set of simulation experiments with a library of microstructures. SEM analyses also often contribute to technical studies of unique museum objects. In this case the questions that need to be answered very much depend under the object under study. In a report by Lascalea et al. [70] the authors perform an overall examination of a Santamarian ceremonial axe using a range of analytical techniques. Again SEM is used to determine
Fig. 3.25. BSE image of a copper–arsenic alloy. The alpha phase is composed of regular grains, containing 3% As. Along the grain boundaries Cu3As intermetallics, with an As concentration up to 28% (bright phase), are present next to copper sulphide inclusions (dark phase). Magnification 400 [67]. Reproduced with permission from TMS Publications.
116
Electron microscopy and its role in cultural heritage studies
the elemental composition at various locations to get an idea of the homogeneity of the object. 3.7
USE
Use-wear studies of stone tools aim at providing information on how the tools were used and what materials were worked. Different human activities, such as cutting plant stalks or hide cleaning, leave different markings on the tools used in these tasks. The investigations of the tools involve the study of the morphology of the tool, plus the edge fractures and the alteration of the stone surface, which have been caused by use and wear. In particular, the microwear polish is considered to be the most significant diagnostic trace [71]. Usually optical microscopy is used to perform these investigations, but visual examination can be enhanced considerably by SEM. Alvarez et al. [72], for instance, report on a study of microwear traces on experimental lithic tools used in the replication of rock art carvings. In their study the use of SEM not only confirms the results obtained by optical microscopy, but also allows a contextual view and characterization of the wear traces by allowing broader portions of the tool edges to be in focus at high magnification without losing resolution and depth of field. Use studies are also quite customary in the case of pottery. The focus here is primarily devoted to the study of organic residues, in which case electron microscopy does not have a major contribution. Nevertheless crucible accretions also, for instance, are investigated to help determine the ore or metal that was being processed. Adriaens et al. [73] report on a SEM-WDS study for the analysis of ceramic fragments from Kestel/Go¨ltepe (Turkey). In this study it was strongly hypothesized that the excavated vessels were crucibles in which tin ore from the nearby Kestel mine was smelted. Using SEM-WDS and SIMS the intention was to examine whether any remains of tin smelting activities could be found inside the crucibles to resolve the question of whether tin was indeed smelted at the site in Kestel /Go¨ltepe. Cross-sections were embedded in epoxy resin in such a manner that the inside of the material could be analysed. Figure 3.26 shows a backscattered electron image of such a crucible cross-section, with next to it an X-ray map of the same area for tin. SEM-WDS results have shown that the tin-bearing layer actually consists of a silicate phase with an average of 40% of tin, which corresponds in composition to a typical metallic tin slag, an indication that the crucibles were very likely to have been used for the processing of tin-bearing materials.
117
A. Adriaens and M.G. Dowsett
Fig. 3.26. (a) Backscattered electron image of a crucible cross-section; (b) tin X-ray map of the same area [73]. Reproduced with permission from Academic Press.
3.8
DEGRADATION PROCESSES, CORROSION AND WEATHERING
The study of deterioration processes, corrosion and weathering products, is besides fundamental interest, in particular, useful for conservation and preservation purposes. The information one wants to acquire at this stage is
118
Electron microscopy and its role in cultural heritage studies
the nature and the type of the attack and in addition the environmental conditions that contributed to the corrosion or weathering. The combined knowledge will enable the conservator to choose the best possible cleaning treatments, conservation and storage parameters. 3.8.1
Metals
For various inorganic materials, electron microscopy equipped with EDS has proven to useful in providing information on the composition of their deterioration products. In particular for metals analysis, the technique has obtained a remarkably favourable acceptance and studies in this field are again numerous [74 –80]. Corrosion of metals is understood to be the degradation of the metallic structure at its surface through chemical reactions with species of the environment, which results in the transition of metal atoms from the metallic state to the non-metallic state. As a result of this, SEM –EDS data are often complemented with methods such as XRD or XPS, IR or Raman that can provide speciation information. Samples are often analysed as cross-sections as this gives the additional advantage of elucidating the layered structure of the corrosion. Chapters 10 and 13 of this book give a comprehensive overview of copper alloy patinas and corroded silver coins, respectively. 3.8.2
Glass
Next to metals analysis, electron microscopy also contributes to the study of deterioration processes of glass [81 – 86]. Deterioration of glass includes both chemical and structural changes. The initial stage of attack is a process that involves ion exchange between the alkali ions, which are present in the silicate structure of the glass, such as Na, K, and hydrogen from the environment. This leads to the formation of a leached layer or so-called “gel layer” in which alkaline elements are depleted (Fig. 3.27). Provided the glass is in continuous contact with water, the attack will eventually lead to a complete breakdown of the silicate structure. In case of atmospheric attack, the leached ions will interact with components from the ambient atmosphere such as carbon dioxide and sulphur dioxide which will lead to a crust formation including products such a calcite (CaCO3) and gypsum (CaSO4·2H2O) [87]. In practice SEM provides information about the state of deterioration in that the structure and the thickness of the corrosion layer can be assessed, together with the extent to which leaching has occurred; and the formation of certain minerals can be observed. For more details on glass
119
A. Adriaens and M.G. Dowsett
Fig. 3.27. SE image of the so-called “gel layer” in corroded glass. Reproduced from Ref. [81].
corrosion and the use of SEM in this field the reader is referred to Chapter 16 of this book.
3.8.3
Ceramics
Although at first sight perhaps less likely than glass or metal corrosion, the surfaces of ceramics may also undergo changes whilst buried. Substantial changes in major element or mineralogical composition may affect provenance studies as well as technological investigations such as those involving refiring experiments, porosimetry or phase analysis [88]. Conclusions in the current literature on the topic of post-depositional alterations of ceramics are at this point still controversial and not univocal [88,89]. Franklin and Vitali [90] have conducted a series of survey-type corrosion experiments in order to study the environmental stability of ceramics. They subjected samples of a ceramic of known composition, structure and manufacture to a variety of aqueous solutions, designed to simulate a broad range of soil environments. Major emphasis was placed on microscopic observations of the surfaces and cross-sections of the test briquettes over a wide range of magnifications in order to investigate the corrosion mechanisms. SEM –EDS was used to study the morphology and composition of the deposits and the surfaces of the treated and untreated briquettes. The authors concluded that the formation of the deposit on the surface, having a
120
Electron microscopy and its role in cultural heritage studies
thickness in the order of 100 mm, acted as a protective layer and thus prevented any further alteration of the inner composition. Freestone et al. [88] have used polarizing microscopy, SEM – EDS, EPMA and XRD to attack the same issue. Unlike Franklin and Vitali [90], they used archaeological ceramics for their study. Elemental distributions across the thin-sections of the shards were obtained by defocusing the beam of EPMA to give a spot of ca. 80 mm diameter and thus analysing the fine-grained matrix between the coarser temper inclusions. The analyses showed changes in the concentrations of P2O5, CaO, FeOt (the total Fe oxide content expressed as FeO) and Al2O3 across the shards. Results showed that calcium phosphates readily form during the weathering of ceramics, but did not support the view that these act as effective diffusion barriers on an archaeological time scale, at least not in all potshards, since precipitation of phosphate had occurred through to the cores of the shards. Limiting conditions on any alteration that may occur include environmental conditions, internal structure of the ceramics, the composition of the ceramic and exposed phases and the tendency for the precipitation for weathering products to seal off diffusion pathways into the ceramic, as shown by Franklin and Vitali [90]. 3.9
AUTHENTICITY AND AUTHENTICATION
Authenticity is one of the major concerns in museums. Judgements on the genuineness or authenticity of a work of art or an archaeological object are based on thorough examinations including the characterization of the materials, the methods of manufacture and age determination. It has to be noted here that the scientific data cannot provide definitive attribution, but can provide information that shows whether a particular cultural property is consistent with other cultural property from a certain time period, from a specific place of origin or from a certain maker. For paintings, frescoes and manuscripts, pigment identification is used to trace imitations and forgeries. Chapters 12 and 14 of this book describe this issue in detail and demonstrate the use of various analytical techniques in this field. Within the electron microscopy methods, SEM again takes up the majority of the work, though applications using TEM and especially ESEM are also gaining ground. Burnstock and Jones [91] give a comprehensive overview of the use of SEM imaging for the study of paint materials. Next to pigments, the authors also describe the analysis of coatings and binding media as well as deterioration studies. An extensive literature list is provided. Imaging SEM is, in particular, useful for the characterization of the pigment particle
121
A. Adriaens and M.G. Dowsett
morphology. Inorganic pigments tend to show unique characteristic features, which enable characterization. For some 19th and 20th century pigments, however, elemental analysis is needed as supplementary technique for differentiation [91]. TEM-EDS measurements for the study of pigments date back to as early as 1967. Kleber et al. [92] report on the examination and investigation of “Maya Blue,” a blue pigment characteristic to the Mayan civilization in ancient America. The capabilities of TEM-EDS for the study of pigments are described in an overview by Bulcock [93]. The author points out several advantages of the technique in comparison to SEM, the main one being the simultaneous acquisition of elemental and crystal structure information, which is obtained by electron diffraction patterns. Secondly the samples are much thinner and the instrument usually operates at much higher accelerating voltages, which results in less scattering and broadening of the electron beam and hence a higher spatial resolution (e.g., Fig. 3.2(b)). This is, in particular, advantageous for the analysis of multilayered samples. By means of an example, Bulcock [93] discusses the investigation of a 15th century Venetian panel painting in view of an authenticity study and subsequent restoration. EDS spectra and electron diffraction patterns were obtained from extremely small paint fragments (200 £ 20 mm2 maximum). The acquired electron diffraction patterns were compared with an available database to identify the mineral under investigation. Results were readily confirmed with the elemental analysis obtained by EDS. Doehne and Bower [94] and Bower et al. [95] make a comparison of ESEM with conventional SEM for the analysis of pigments from paintings. The main advantage in their work is the possibility of analysing non-conducting samples without the need for coating as the instrument operates under moderate pressure (0.75 – 150 kPa). Disadvantages, on the other hand, lie in the higher noise level of the images and the smaller field of view. Tracing imitations of glass objects is usually done by classification and provenance studies. Both are often based on stylistic evolutions, but here also the chemical composition proves to be very helpful. In section 3.5.2, we already discussed an example of provenance/authenticity studies, in which the authors aimed at distinguishing imitation from real Venetian glass objects [28,29]. Another type of glass that has been of great interest to collections is the Art Nouveau iridescent glass, the most famous manufacturers being Tiffany (USA) and Loetz (Austria). In the mean time many imitations have been made. Jembrih et al. [96] report on a study in which they aim at developing a recognition procedure based on non-destructive analytical techniques. Using XRF and SEM – EDS differences can be
122
Electron microscopy and its role in cultural heritage studies
Fig. 3.28. Elemental distribution of Si, Pb, K, P and Ag of a cross-sectioned Loetz sample. Magnification 30 [96]. Reproduced with permission from Springer.
identified on the basis of the structure and chemical composition of the surface layer. A characteristic of Tiffany glass fragments is leaded bulk glass, whereas in the case of Loetz K – Ca – Si bulk glass could be determined. Modern glass fragments show a high amount of Na and Sr. In addition, the cross-sections of Tiffany show a layered structure of the bulk without a specific surface layer whereas the cross-sections of Loetz glass reveal a homogeneous bulk material with one or two homogeneous surface layers as can be seen in the X-ray mappings (Fig. 3.28). 3.10
CONCLUSIONS
The application of electron microscopy to the study of cultural heritage materials has a relatively long history. Elemental analysis using EPMA, for instance, was demonstrated as early as the 1960s [97,98]. Later also SEM – EDS came into use. Both techniques contribute to the understanding of the heterogeneity of the composition of materials by providing morphological data as well as elemental analysis. EPMA and SEM – EDS are used for a large variety of applications, such as microscopic studies of metallographic sections, fragments of ceramic, stone, glass and paintings. In addition, organic materials, such as plant materials, animal and hominid remains are also being examined, although these are not addressed here. A comprehensive overview of possible applications in the latter fields is given by Olsen [99].
123
A. Adriaens and M.G. Dowsett
TEM, on the other hand, is used surprisingly little in this research area, in spite of the fact that it combines elemental and structural information. It is the method of choice whenever sub-surface properties are to be studied at nanometre spatial resolution. A major disadvantage of the technique is that ultra-thin microtome prepared sections are needed. Environmental SEM (ESEM) is another instrument which is gaining an increasing interest. It provides the added capability of working at near atmospheric pressure and, therefore, provides several significant advantages over conventional SEM analysis. Firstly, vacuum sensitive materials, such as moist, liquid, or outgassing samples may be analysed. Secondly, analysis and imaging of uncoated, non-conductive materials can be done without local surface charging. And thirdly, the atmosphere within the analysis chamber is now a variable that can be modified to facilitate the examination of dynamic processes, such as wetting and drying, at high magnifications. The electron microscope is, without doubt, the most sophisticated and versatile imaging and analytical tool that the human race has devised to date. It will continue to make a huge contribution to the study of the cultural heritage (of which it is an integral part like other tools and artefacts). In addition to developments such as LVSEM and ESEM, one should remember that modern electron microscopy owes much to the rapid evolution of the microcomputer—another ubiquitous aspect of our culture. Thus this complex instrument has become easier to use. Its intrinsic analytical and imaging power is made generally available through rapid spectrum and image processing software and the increasing use of parallel detection and the multispectral imaging techniques which the mass storage and speed of the modern computer makes feasible. The ability of electron microscopy to make the correlation between microchemistry and microstructure which is vital in cultural heritage studies will continue to be enhanced by developments in this area.
Acknowledgements The authors would like to thank Tom Broekmans for kindly sharing his broad knowledge on ceramics. REFERENCES 1
124
P.J. Goodhew, J. Humphreys and R. Beanland, Electron Microscopy and Analysis. Taylor & Francis, London, 2001.
Electron microscopy and its role in cultural heritage studies 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25
26 27 28 29 30 31 32
I.M. Watt, The Principles and Practice of Electron Microscopy. Cambridge University Press, Cambridge, 1997. P. Grivet, P.W. Hawkes and A. Septier, Electron Optics Part 2. Pergamon Press, Oxford, 1965. L. de Broglie, Optique Electronique et Corpusculaire. Herman, Paris, 1950. L.W. Swanson and L.C. Crouser, J. Appl. Phys., 40 (1969) 4741. G.D. Danilatos, Scanning, 4 (1981) 9. G.D. Danilatos, J. Microsc., 162 (1991) 391. H.A. Videla, P.S. Guiamet and S. Gomez de Saravia, Int. Biodeteriorat. Biodegred., 46 (2000) 335. M.N. Timofeeff, T.K. Lowenstein and W.H. Blackburn, Chem. Geol., 164 (2000) 171. C.P. Royall, B.L. Thiel and A.M. Donald, J. Microsc., 204 (2001) 185. S. Kitching and A.M. Donald, J. Microsc., 190 (1998) 357. D.C. Joy, A.D. Romig Jr. and J.I. Goldstein, Principles of Analytical Electron Microscopy. Plenum Press, New York, 1986. G. Cliff and G.W. Lorrimer, J. Microsc., 103 (1975) 203. M.H. Loretto, Electron Beam Analysis of Materials. Chapman & Hall, London, 1988. P. Grivet, P.W. Hawkes and A. Septier, Electron Optics Part 1. Pergamon Press, Oxford, 1965. P. Dahl, Introduction to Electron and Ion Optics. Academic Press, London, 1973. M.E. Haine and P.A. Einstein, Br. J. Appl. Phys., 3 (1952) 40. C.W. Oatley, J. Phys. E Sci. Instrum., 8 (1975) 1037. D. Tuggle, L.W. Swanson and J. Orloff, J. Vac. Sci. Technol., 16 (1979) 1699. A.N. Broers, J. Phys. E, 2 (1969) 273. J. Ruste, in: F. Maurice, L. Meny and R. Tixier (Eds.), Microanalysis and Scanning Electron Microscopy. Les Editions de Physique, Orsay, 1979. H. Bronk and I.C. Freestone, Archaeometry, 43(4) (2001) 517. San Andres, M.I. Baez, J.L. Baldonedo and C. Barba, J. Microsc., 188(1) (1997) 42. I.C. Freestone, Archaeometry, 24(2) (1982) 99. T. Broekmans, A. Adriaens and K. Van Lerberghe, Modern trends in scientific studies on ancient ceramics. In: T. Kiligoglou, A. Hein and Y. Maniatis (Eds.), British Archaeological Reports (BAR) series 1011, 1011. Archaeopress, Oxford, 2002, 343 pp. I.C. Freestone and A.P. Middleton, Mineralog. Mag., 51 (1987) 21. S. Hreglich and M. Verita, Scanning Electron Microsc., 2 (1986) 485. I. De Raedt, K. Janssens and J. Veeckman, J. Anal. At. Spectrom., 14 (1999) 493. I. De Raedt, K. Janssens, J. Veeckman, L. Vincze, B. Vekemans and T.E. Jeffries, J. Anal. At. Spectrom., 16 (2001) 1012. M. Sax and N.D. Meeks, Archaeometry, 37(1) (1995) 25. M. Sax, J. McNabb and N.D. Meeks, Archaeometry, 40(1) (1998) 1. M.S. Tite, I.C. Freestone, N.D. Meeks and M. Bimson, in: J.S. Olin and A.O. Franklin (Eds.), Archaeological Ceramics. Smithonian Institution Press, Washington DC, 1982, 109 pp.
125
A. Adriaens and M.G. Dowsett 33 34 35 36 37 38 39
40 41 42 43 44 45 46 47
48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63
126
R.E. Grim, Clay Mineralogy. McGraw-Hill, New York, 1968. R.W. Grimshaw, The Chemistry and Physics of Clays and Allied Ceramic Materials. Wiley, New York, 1971. W.A. Deer, R.A. Howie and J. Zussman, Rock Forming Minerals, Sheet silicates, Vol. 3. Longmans, London, 1962. A.C.D. Newman, Chemistry of clays and clay minerals. Longman Scientific and Technical, London, 1987. O. Nieuwenhuyse, L. Jacobs, B. van As, T. Broekmans and A. Adriaens, Pale´orient, 27/1 (2002) 147. L. Jacobs, Newsletter of the Department of Pottery Technology (Leiden University), 9/10 (1991/1992) 7. L. Jacobs, in: S. Me´ry (Ed.), Earth Sciences and Archaeological Ceramics. Experimentations, Applications. Institut Ge´ologique Albert de Lapparent, Cergy, 1992, 121 pp. W. Noll, Neues Jahrbuch fu¨r Mineralogie Abhandlungen, 127(3) (1976) 261. W. Noll, Acta Praehist. Archaeol.,(7/8) (1976/1977) 15. W. Noll, Antike Welt, 8(2) (1977) 21. W. Noll, R. Holm and L. Born, Angewandte Chem. Int. Ed., 14(9) (1975) 602. R.E.M. Hedges, Archaeometry, 18(2) (1976) 209. R.E.M. Hedges and P.R.S. Moorey, Archaeometry, 17(1) (1975) 25. W.D. Kingery and P. Vandiver, Ceramic Masterpieces. Art, Structure, Technology. Free Press, New York, 1986. B. Kleinmann, in: J.S. Olin and M.J. Blackman (Eds.), Proceedings of the 24th International Archaeometry Symposium. Smithsonian Institution Press, Washington, DC, 1986, 73 pp. M.S. Tite, Archaeometry, 29(1) (1987) 21. M.S. Tite, Archaeometry, 31(2) (1989) 115. M.S. Tite and M. Bimson, World Archaeol., 21 (1989) 87. M.S. Tite, I. Freestone, R. Mason, J. Molera, M. Vendrell-Saz and N. Wood, Archaeometry, 40(2) (1998) 241. R.B. Mason and M. Tite, Archaeometry, 39(1) (1997) 41. M.S. Tite and Y. Maniatis, Nature, 257(11) (1975) 122. Y. Maniatis and M.S. Tite, J. Archaeol. Sci., 8 (1981) 59. B. Velde and I.C. Druc, Archaeological Ceramic Materials. Origin and Utilization. Springer, Heidelberg, 1999. D.J. Barber and I. Freestone, Archaeometry, 32(1) (1990) 33. D.A. Scott, Metallography and Microstructure of Ancient and Historic Metals. The Getty Conservation Institute, Los Angeles, 1991. R. Tylecote, A History of Metallurgy. Institute of Materials, London, 1992. P. Craddock, Early Metal Mining and Production. Edinburgh University Press, Edinburgh, 1995. A. Isakov, P.L. Kohl, C.C. Lamberg-Karlovsky and R. Maddin, Archaeometry, 29(1) (1987) 90. N. Ryndina, G. Indenbaum and V. Kolosova, J. Archaeol. Sci., 26 (1999) 1059. S. Shalev and P. Northover, Archaeometry, 35(1) (1993) 35. S. Klein and A. Hauptman, J. Archaeol. Sci., 26 (1999) 1075.
Electron microscopy and its role in cultural heritage studies 64
65 66 67 68 69 70 71 72 73 74 75 76 77 78
79 80 81 82
83 84
85 86 87 88 89 90 91 92 93
J. Merkel and B. Rothenberg, in: A. Hauptmann, E. Pernicka, T. Rehren and U. Yalc¸in (Eds.), The Beginnings of Metallurgy. Der Anschnitt, Bochum, 1999, 149 pp. ¨ zbal, A. Adriaens and B. Earl, Pale´orient, 25(1) (2000) 57. H. O H. Lechtman and S. Klein, J. Archaeol. Sci., 26 (1999) 497. B. Earl and A. Adriaens, JOM, 52(3) (2000) 14. ¨ zbal and K.A. Yener, Int. J. Mining Miner., A. Adriaens, B. Earl, H. O 2 (2002) 35. N.D. Meeks, Archaeometry, 28(2) (1986) 133. G.E. Lascalea, A. Pifferetti, M. Fernandez De Rapp, N.E. Walsoe de Reca and J.P. Northover, Archaeometry, 44(1) (2002) 83. S. Yamada and M. Shimura, in: A.D. Romig Jr. and J.I. Goldstein (Eds.), Microbeam Analysis. San Francisco Press, San Francisco, 1984, 223 pp. M. Alvarez, D. Fiore, E. Favret and R.C. Guerra, J. Archaeol. Sci., 28 (2001) 457. A. Adriaens, K.A. Yener and F. Adams, J. Archaeol. Sci., 26 (1999) 1069. D.A. Scott, Stud. Conserv., 30 (1985) 49. D.A. Scott, J. Am. Inst. Conserv., 33(1) (1994) 1. H. Strandberg, L.G. Johansson and O. Lindqvist, Mater. Corrosion, 48 (1997) 721. L. Robbiola, J.M. Blengino and C. Fiaud, Corrosion Sci., 40(12) (1998) 2083. M. Wadsak, I. Constantinides, G. Vittiglio, A. Adriaens, K. Janssens, M. Schreiner, F. Adams, Ph. Brunella and M. Wuttmann, Mikrochim. Acta, 133(1 –4) (2000) 159. R. Linke and M. Schreiner, Mikrochim. Acta, 133 (2000) 165. I. Constantinides, A. Adriaens and F. Adams, Appl. Surf. Sci., 189(1 –2) (2002) 90. A. Aerts, Microscopic analysis of Roman vessel glass, Dissertation. University of Antwerp, Antwerp, Belgium, 1998. K. Janssens, A. Aerts, L. Vincze, F. Adams, C. Yang, R. Utui, K. Malmqvist, K.W. Jones, M. Radke, S. Garbe, F. Lechtenberg, A. Knochel and H. Wouters, Nucl. Instr. Meth. Phys. B, 109/110 (1996) 690. M. Schreiner, Mikrochim. Acta II, (1991) 255. O. Schalm, Characterization of Paint Layers in Stained-glass Windows: Main Causes of the Degradation of Nineteenth Century Grisaille Paint Layers, Dissertation. University of Antwerp, Antwerp, Belgium, 2000. A. Orlando, F. Olmini, G. Vaggelli and M. Bacci, Analyst, 121 (1996) 553. G. Woisetschla¨ger, M. Dutz, S. Paul and M. Schreiner, Mikrochim. Acta, 135 (2000) 121. R.H. Doremus, Glass Science. Wiley, New York, 1973. I. Freestone, N.D. Meeks and A.P. Middleton, Archaeometry, 27(2) (1985) 161. P.M. Rice, Pottery Analysis. A Sourcebook. University of Chicago Press, Chicago, 1987. U.M. Franklin and V. Vitali, Archaeometry, 27(1) (1985) 3. A. Burnstock and C. Jones, in: D.C. Creagh and D.A. Bradley (Eds.), Radiation in Art and Archaeometry. Elsevier, Amsterdam, 2000, 202 pp. R. Kleber, L. Masschelein-Kleiner and J. Thissen, Stud. Conserv., 12/2 (1967) 41. S. Bulcock, in: D.C. Creagh and D.A. Bradley (Eds.), Radiation in Art and Archaeometry. Elsevier, Amsterdam, 2000, 232 pp.
127
A. Adriaens and M.G. Dowsett 94 95 96 97 98 99
128
E. Doehne and N.W. Bower, Microbeam Anal., 2 (1993) s39. N.W. Bower, D.C. Stulik and E. Doechne, Fresenius, J. Anal. Chem., 348 (1994) 402. D. Jembrih, M. Schreiner, M. Peev, P. Kresja and C. Clausen, Mikrochim. Acta, 133 (2000) 151. A.P. Hornblower, Archaeometry, 5 (1962) 37. W.J. Young, in: G. Thomson (Ed.), Recent Advances in Conservation. Butterworths, London, 1963. S.L. Olsen, Scanning Electron Microscopy in Archaeology, BAR International Series 452, Oxford, 1996.
Chapter 4
X-ray based methods of analysis K. Janssens
4.1
INTRODUCTION
X-rays were discovered by Ro¨ntgen in 1878 and were first considered to be a mysterious type of radiation with properties very different from electromagnetic radiation in the ultraviolet, visual and infrared range of the spectrum: Ro¨ntgen’s first experiments suggested that X-rays could not be reflected, refracted or polarized but instead penetrated fairly deeply into materials, apparently without causing harm or other (permanent) changes. Although it is now well documented that X-rays exhibit analogous characteristics to the radiation of longer wavelengths and that high doses of this type of radiation on (biological) tissues can effectively induce discolourations and other damages, irradiation of inorganic materials with X-rays is generally considered not to cause them any harm. Thus, next to being used for radiographic purposes (see Chapter 2), X-rays are also frequently employed for compositional and structural analysis of cultural heritage (and many other) materials. A well-established method of quantitative element analysis is X-ray fluorescence (XRF) analysis, which is based on the ionization of the atoms of the material being investigated by an energetic beam of primary X-rays. The characteristic radiation that is emitted by the ionized atoms upon relaxation contains information on the nature and the abundance of the elemental constituents present. Since XRF meets a number of the requirements of the “ideal method” for non-destructive analysis of cultural heritage materials (see Chapter 1) [1], analysis of objects of artistic and/or archaeological value with conventional XRF is fairly common; it is in fact one of the most often applied methods for obtaining qualitative and semi-quantitative information on the materials these objects are made of [2]. Several textbooks cover the fundamental and methodological aspects of the method and its many variants [3]. In the Comprehensive Analytical Chemistry XLII Janssens and Van Grieken (Eds.) q 2004 Elsevier B.V. All rights reserved
129
K. Janssens
specific context of non-destructive analysis of (often) irregular and/or heterogeneous materials, the use of conventional XRF for reliable quantitative analysis is severely hampered by the fact that the irradiated area usually is large. This prevents the separate analysis of details of decorations, distinct features, etc. Also, the irradiation geometry and sample surface are usually non-ideal and/or not well defined, possibly introducing systematic errors in the quantification. Using smaller X-ray beams can circumvent some of these limitations. In such a case, motorized sample movement allows to extend the local analysis capability towards two-dimensional imaging of certain elements on the surface of artefacts. Since many objects of artistic and/or archaeological nature are fairly large and bulky (e.g., statues, oil paintings, vases, treasury objects), recently available instrumentation that can accommodate objects of various shapes and is able to operate in air atmosphere is very useful. X-ray diffraction (XRD) makes use of the coherent scattering of X-rays by atomic electron clouds and the constructive interference that takes place between rays scattered by regularly spaced series of atoms. The orientation and relative intensity of the reflexes contain information on the crystallographic structure of the materials being studied. Via the use of appropriate data bases, XRD allows to qualitatively identify the compounds present and quantitatively determine their relative abundance in case a mixture of the latter is under scrutiny. It is beyond the scope of this chapter to provide a detailed description on the fundamentals of XRD; more detailed information may be found in many excellent reference texts [4 – 8]. With the advent of more powerful X-ray sources called synchrotrons, other forms of X-ray spectroscopy, requiring highly monochromatic primary radiation of which the energy or wavelength may be continuously varied, have also come into use. Notably some of the forms of X-ray absorption spectroscopy (XAS) are proving themselves to be useful for non-destructive characterization of cultural heritage materials while these sources have also significantly stimulated the development of the microscopic equivalents of XRF and XRD. By means of XAS it is possible to obtain information on the chemical state of specific (inorganic) constituents of a material, such as the oxidation state of Mn and Fe in weathered glass (often responsible for giving this material its greenish/brownish colour). In what follows, the fundamentals of the interaction of X-rays with matter are described, after which an overview of the available instrumentation for XRF, XRD and XAS is presented. The chapter concludes with a literature survey on applications of X-ray methods in the cultural heritage sector.
130
X-ray based methods of analysis
4.2 4.2.1
BASIC PRINCIPLES X-ray wavelength and energy scales
The X-ray or Ro¨ntgen region of the electromagnetic spectrum starts at ca. 10 nm and extends towards the shorter wavelengths. The energies of X-ray photons are of the same order of magnitude as the binding levels of inner-shell electrons (K, L, M,… levels) and therefore can be used to excite and/or probe these atomic levels. The wavelength l of an X-ray photon is inversely related to its energy E according to: ¼ 12:4=EðkeVÞ lðAÞ
ð4:1Þ
where 1 eV is the kinetic energy of an electron that has been accelerated over a voltage difference of 1 V (1 eV ¼ 1.602 £ 10219 J). Accordingly, the X-ray energy range starts at 100 eV and continues towards higher energies. X-ray analysis methods most commonly employ radiation in the 1 – 50 keV ˚ ) range. The established unit of measure for wavelengths in the (10 –0.2 A ˚ is equal to 10210 m. X-ray region is the angstrom; 1 A 4.2.2
Interaction of X-rays with matter
When an X-ray beam passes through matter, some photons will be absorbed inside the material or scattered away from the original path, as illustrated in Fig. 4.1. The intensity I0 of an X-ray beam passing through a layer of thickness d and density r is reduced to an intensity I according to the well-known law of Lambert –Beer: I ¼ I0 e2mrd
ð4:2Þ
Fig. 4.1. X-ray—matter interactions.
131
K. Janssens
The number of photons (the intensity) is reduced but their energy is generally unchanged. The term m is called the mass attenuation coefficient and has the dimension cm2/g. The product mL ¼ mr is called the linear absorption coefficient and is expressed in cm21. m(E) is sometimes also called the total cross-section for X-ray absorption at energy E. Fig. 4.2 shows a log – log plot of the energy dependence of the mass attenuation coefficient of several chemical elements in the X-ray energy range between 1 and 100 keV. The absorption edge discontinuities (due to photoelectric absorption—see below) are clearly visible. Low Z materials attenuate X-rays of a given energy less than high-Z materials. A given material will attenuate high energy (i.e., hard) X-rays less than low energy (soft) X-rays. The mass absorption coefficient m(M) of a complex matrix M consisting of a mixture of several chemical elements (e.g., an alloy such as brass), can be calculated from the mass attenuation coefficient of the n constituting elements: Xn mðMÞ ¼ i¼1 wi mi ð4:3Þ where mi is the mass attenuation coefficient of the ith pure element and wi its mass fraction in the sample considered. This is called the mixture rule. The mass absorption coefficient m plays a very important role in quantitative XRF analysis. Both the exciting primary radiation and the fluorescence radiation are attenuated in the sample. To relate the observed fluorescence intensity to the concentration, this attenuation must be taken into account. As illustrated in Fig. 4.1, the absorption of radiation in matter is the cumulative effect of several types of photon – matter interaction processes
Fig. 4.2. Energy dependence of the mass absorption coefficient m of several elements.
132
X-ray based methods of analysis
that take place in parallel. Accordingly, in the X-ray range the mass attenuation coefficient mi of element i can be expressed as: mi ¼ ti þ si
ð4:4Þ
where ti is the cross-section for photoelectric ionization and si the crosssection for scattering interactions. All the above-mentioned cross-sections are energy (or wavelength) dependent. Except at absorption edges (see below), m is more or less proportional to Z 4l3—see also Chapter 5. 4.2.3
The photoelectric effect; X-ray fluorescence
In the photoelectric absorption process (see Fig. 4.3), a photon is completely absorbed by the atom and an (inner shell) electron is ejected. Part of the photon is used to overcome the binding energy f of the electron and the rest is transferred to the electron in the form of kinetic energy. After the interaction, the atom (actually an ion now) is left in a highly excited state since a vacancy has been created in one of the inner shells. The atom will almost immediately return to a more stable electron configuration by emitting an Auger electron or a characteristic X-ray photon. The latter process is called X-ray fluorescence (XRF). The ratio of the number of emitted characteristic X-rays to the total number of inner shell vacancies in a particular atomic shell that gave rise to it, is called the fluorescence yield of that shell (e.g., vK). For light elements ðZ , 20Þ; predominantly Auger electrons are produced during the relaxation upon K-shell ionization (vK , 0.2) while the medium to heavy elements are preferentially relaxing in a radiative manner (0.2 , vK , 1.0). See Fig. 6.5 for a plot on the dependence of vK, vL3 and vM5 on the atomic number Z. Photoelectric absorption can only occur if the energy of the photon E is equal or higher than the binding energy f of the electron. For example, an X-ray photon with an energy of 15 keV can eject a K-electron (fK ¼ 7.112 keV) or an L3-electron (fL3 ¼ 0.706 keV) out of an Fe atom. However, a 5 keV photon can only eject L-shell electrons from such an atom. Since photoelectric absorption can occur at each of the (excitable) energy levels of the atom, the total photoelectric cross-section ti is the sum of (sub)shell-specific contributions: ti ¼ ti;K þ ti;L þ ti;M þ … ¼ ti;K þ ðti;L1 þ ti;L2 þ ti;L3 Þ þ ðti;M1 þ … þ ti;M5 Þ þ …
ð4:5Þ
In Fig. 4.4, the variation of tMo with energy is plotted. At high energy, e.g., above 50 keV, the probability for ejecting a K-electron is rather low and
133
K. Janssens
Fig. 4.3. Photoelectric ionization can be followed by either radiative relaxation, causing the emission of characteristic fluorescent X-rays or non-radiative relaxation, involving the emission of Auger electrons.
that for ejecting an L3-electron is even lower. As the energy of the X-ray photon decreases, the cross-section increases, i.e., more vacancies are created. At the binding energy fK ¼ 19.99 keV, there is an abrupt decrease in the cross-section because X-rays with lower energy can no longer eject
134
X-ray based methods of analysis
Fig. 4.4. Variation of tMo as a function of X-ray photon energy. The K, L1, L2 and L3 absorption edges are visible.
electrons from the K-shell. However, these photons continue to interact with the (more weakly bound) electrons in the L and M-shells. The discontinuities in the photoelectric cross-section are called absorption edges. The ratio of the cross-section just above and just below the absorption edge is called the jump ratio, r. As XRF is the result of selective absorption of radiation, followed by spontaneous emission, an efficient absorption process is required. An element can therefore be determined with high sensitivity by means of XRF when the exciting radiation has its maximum intensity at an energy just above the K-edge of that element.
4.2.3.1 Selection rules, characteristic lines and X-ray spectra Characteristic X-ray photons are produced following the ejection of an inner orbital electron from an excited atom, and subsequent transition of atomic orbital electrons from states of high to low energy. Each element present in the specimen will produce a series of characteristic lines making up a polychromatic beam of characteristic and scattered radiation coming from the specimen. The systematic (IUPAC) name of the X-ray line arising from a vacancy in the K-shell of an atom, which is filled by an electron originally belonging to the L3-shell of that atom is the K –L3 transition. However, this transition is more commonly referred to as the Ka1-line (non-systematic or Siegbahn nomenclature); similarly, fluorescent X-rays resulting from L3-M5 transitions are better known as La1-photons. Table 4.1 lists a number of observed X-ray lines and their corresponding IUPAC and Siegbahn names. Moseley first established the relationship between the wavelength l of a characteristic X-ray photon and the atomic number Z of the excited element
135
K. Janssens TABLE 4.1 Principal X-ray lines (IUPAC and Siegbahn notations) and their approximate intensities relative to the major line in each subshell Series
IUPAC name
Siegbahn name
Relative intensity
K-lines
K–L3 K–L2 K–M3 K–M2 L3 –M5 L3 –M4 L3 –N5,4 L3 –M1 L3 –N1 L2 –M4 L2 –N4 L2 –M1 L2 –O1 L1 –M3 L1 –M2 L1 –N3 L1 –N2 M5 –N7 M5 –N6 M5 –N6
Ka1 Ka2 Kb1 Kb3 La1 La2 Lb2,15 Ll Lb6 Lb1 Lg1 Lh Lg6 Lb3 Lb4 Lg3 Lg2 Ma1 Ma2 Mb
100 ,50 ,17 ,8 100 ,10 ,25 ,5 ,1 100 ,20 ,3 ,3 100 ,70 ,30 ,30
L3-lines
L2-lines
L1-lines
M-lines
(see Fig. 4.5). Moseley’s law is written as: 1=l ¼ KðZ 2 sÞ2
ð4:6Þ
where Z is the atomic number and K and s are constants. s is the shielding constant and takes a value close to one. K has a different value for each of the line series considered (e.g., the Ka-lines, the La-lines—see Table 4.1). Each unique atom has a number of available electrons that can take part in the transfer and, since millions of atoms are typically involved in the excitation of a given specimen, all possible de-excitation routes are taken. These de-excitation routes can be defined by a simple set of selection rules that account for the majority of the observed wavelengths. Application of the selection rules indicates that in, e.g., the K series, only L2 ! K and L3 ! K transitions are allowed. There are equivalent pairs of transitions for the L, M, N, etc., shells. Figure 4.6 shows the lines that are observed in the K series.
136
X-ray based methods of analysis
Fig. 4.5. Variation of characteristic line wavelengths with atomic number.
4.2.4
Scattering and diffraction
Scattering is the interaction between radiation and matter which causes a photon to change direction. If the energy of the photon is the same before and after scattering, the process is called elastic or Rayleigh scattering. Elastic scattering takes place between photons and bound electrons and forms the basis of XRD. If the photon loses some of its energy, the process is called inelastic or Compton scattering. Accordingly, the total cross-section for scattering si can be written as the sum of two components: si ¼ sR;i þ sC;i
ð4:7Þ
Fig. 4.6. Characteristic X-ray lines observed in the K series.
137
K. Janssens
where sR,i and sC,i, respectively, denote the cross-sections for Rayleigh and Compton scatter of element i. Compton scattering occurs when X-ray photons interact with weakly bound electrons. After inelastic scattering over an angle f, a photon (see Fig. 4.7), with initial energy E, will have a lower energy E 0 given by the Compton equation: E0 ¼
E E 1þ ð1 2 cos fÞ m0 c2
ð4:8Þ
where m0 denotes the electron rest mass. 4.2.4.1 Interference and diffraction When a wave front of X-rays strikes an atom, the electrons in this atom will scatter the X-rays (see above: Rayleigh scattering). The elastically scattered wave is immediately re-emitted in all directions and can be imagined as a spherical wave front (see Fig. 4.8). When a line of identical atoms, a distance a apart, is considered, and a ~ where lkl ~ ¼ 1=l) approaches this wave (with wavelength l and wave vector k; line under an angle u, the wave front arriving at the second atom will have a path different d1 ¼ OP ¼ a cos(u) relative to that arriving at the first atom
Fig. 4.7. Geometry for Compton scattering of X-ray photons.
138
X-ray based methods of analysis
Fig. 4.8. Incident plane wave and scattered spherical wave during XRD experiments.
(see Fig. 4.9). When two rays, scattered under an angle w relative to the row of atoms are considered, a second path difference d2 ¼ QR ¼ a cos(w) is observed that will partially compensate this offset, so that finally, the path difference between the two rays scattered at both atoms is given by: D ¼ d1 2 d2 ¼ ðQR 2 OPÞ ¼ a½cosðuÞ 2 cosðwÞ Only when this path difference is equal to an integral multiple n of the wavelength l, constructive interference will take place between the scattered rays, giving rise to an increased intensity in specific directions w relative to the line of atoms. Depending on the value of the order n1 (n1 ¼ 0, 1, 2,…), this set of directions form cones of decreasing opening angle, as shown in Fig. 4.10 for the cases of u ¼ 908 and u – 908. The condition D ¼ n1l can also be expressed by making use of the scalar product between the wave ~ Þ and scattered waves ðk ~ Þ and the vector a ~ between vectors of the incident ðk 0 1 139
K. Janssens
Fig. 4.9. Path difference between rays scattered at two adjacent atoms spaced a distance a apart.
Fig. 4.10. Diffraction cones (left) produced by interaction of a planar X-ray wave front with a linear array of atoms for (a) u ¼ 908 and (b) u – 908.
140
X-ray based methods of analysis
adjacent atoms along the line: ~ ¼n ~ ·k~ 0 2 a ~ ·k a 1 1
ð4:9Þ
When along some angle relative to the first, a second line of regularly spaced ~ (thus forming a twoatoms is considered, characterized by a spacing vector b dimensional net), this second line will independently generate a series of ~ The intersection of both sets cones at the same origin, but centred around b: of cones will be a series of lines (see Fig. 4.11) and only along directions pointing to these lines, strong constructive interference will be observed. In case of irradiation of a three-dimensional lattice (characterized by a ~ and ~c), the incident X-ray beam k ~ will be diffracted ~; b unit cell with vectors a 0 ~ along those directions k1 that simultaneously fulfil the following equations, termed the Laue conditions: 8 ~ 2k ~ Þ¼n > ~ ·ðk a > 0 1 1 > < ~ Þ¼n ~b·ðk ~ 2k ð4:10Þ 0 1 2 > > > : ~ ~ Þ¼n ~c·ðk0 2 k 1 3
Fig. 4.11. Diffraction induced by a two-dimensional net of atoms is only strong along intersecting lines between two cones.
141
K. Janssens
~ ~; b corresponding to the intersection of three sets of cones cantered around a and ~c: The result is that, for a three-dimensional lattice of atoms irradiated by X-radiation, strong constructive interference will only occur in specific directions and for specific conditions of incidence. In 1913 Bragg realized that the conditions for constructive interference of X-rays are equivalent to that of a simple plane reflecting the X-rays like a optical mirror. This is shown in Fig. 4.12 where the incident radiation strikes the planes of atoms at an angle u. For constructive interference, the path difference GY þ YH between rays A and B (reflected, respectively, off the first and second plane, separated a distance d from each other) must be an integral multiple of the wavelength l. Since GY ¼ YH ¼ d sin(u), it follows that: 2d sin u ¼ nl
ð4:11Þ
This equation is known as Bragg’s law and is very important in XRD and crystallography. When u is maximum, i.e., at 908, the diffracted rays are back-reflected by the crystal and sin u ¼ 1 so that 2d ¼ nl; thus the smallest d-spacing than can be observed with X-rays in any crystal will be equal to one-half of the wavelength of the incident radiation. From Bragg’s law it follows that when a crystal is irradiated with a beam of X-rays of known wavelength l and this beam is diffracted over an angle 2u,
Fig. 4.12. Geometry for Bragg reflection of an X-ray wave front by sets of crystallographic planes spaced a distance d apart.
142
X-ray based methods of analysis
it is possible to determine the interplanar spacing d of the set of atomic planes inside the crystal that caused the diffraction. By either varying the wavelength or the orientation of the impinging beam on the crystal, many principal planes of a crystal can be brought into diffracting mode. The set of recorded d-spacings can then be employed to identify the crystallographic phase that was irradiated and/or to determine the dimensions of the unit cell. 4.2.5
X-ray absorption fine structure and spectroscopy
As shown in Fig. 4.2, the mass absorption coefficient vs. energy curves of the elements feature sharp discontinuities, called absorption edges, corresponding to the energies required for electronic transitions from deep core levels to the vacuum level. However, when one considers in more detail the shape of the m(E) vs. E curves in the immediate vicinity of these edges, additional features become apparent. XAS is a method that employs this fine structure to extract information on the chemical environment of the absorbing atom. In an X-ray absorption experiment, a monochromatic X-ray beam is directed at the sample. The photon energy of the X-rays is gradually increased such that it traverses one of the absorption edges of the elements contained within the sample. Below the absorption edge, the photons cannot excite the electrons of the relevant atomic level and thus absorption is low. Transitions at energies E smaller than the binding energy E0 occur only when the absorbing atoms possesses localized unoccupied (or only partially occupied) states, e.g., d-orbitals in the case of transition metals (see Fig. 4.13a). When the photon energy is just sufficient to excite the electrons (E ¼ E0), then a large increase in absorption occurs known as the absorption edge. At X-ray energies higher than E0, transitions to continuum states take place, i.e., to electronic states not localized on the absorbing atom. The resulting photoelectrons have a low kinetic energy and can be backscattered by the electron shells of the atoms surrounding the emitting atom (see Fig. 4.13c – e). The photoelectrons (with kinetic energy E 2 E0) leaving the atoms can also be regarded as an expanding spherical wave with wavenumber k given by: sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2m0 k¼ ðE 2 E0 Þ "2
ð4:12Þ
An X-ray absorption spectrum is typically divided into two energy regions: the X-ray absorption near-edge structure or XANES region, which extends
143
K. Janssens
Fig. 4.13. Cr-K edge X-ray absorption spectrum of K2CrO4; the physical processes leading to the different features of an X-ray absorption spectrum are indicated, as well as the XANES and EXAFS regions. (a) Pre-edge peak(s) caused by transitions to unoccupied orbitals; (b) multiple scattering peaks; (c) single scattering region showing oscillations on the absorption curve caused by (d) destructive interferences and (e) constructive interference between the outgoing and back-reflected photoelectron waves.
from a few eV below an element’s absorption edge to about 50 eV above the edge, and the extended X-ray absorption fine structure or EXAFS region, which extends from about 50 eV to as much as 1000 eV above the edge (Fig. 4.13 b – e). The XANES region is also sometimes referred to as the near-edge X-ray absorption fine structure or NEXAFS region. The range just before the edge is called the pre-edge region (see Fig. 4.13a). The physics of the processes responsible for XANES and EXAFS spectral features is different; thus, these spectral regions provide different types of information about an element and its local environment. XANES can provide information about vacant orbitals, the electronic configuration and site symmetry of the absorbing atom. The absolute position of the edge contains information on the oxidation state of the absorbing atom. In general, with increasing oxidation state of the central atom [e.g., Mn0, Mnþ2, Mnþ3,…, Mnþ7], the absorption profiles shift towards higher energies by a few eV per oxidation state. In the near-edge region, multiple scattering events dominate (see Fig. 4.13b). Theoretical multiple
144
X-ray based methods of analysis
scattering calculations can be compared with experimental XANES spectra in order to determine the geometrical arrangement of the atoms surrounding the absorbing atom. Hence the technique provides complementary information to EXAFS. The basis of physical process leading to the oscillations of the absorption in the EXAFS region is the interference between the photoelectron wave leaving the emitting atom with its backscattered equivalents, as pictorially represented in Fig. 4.13d and e. Depending on the wavelength (i.e., on the energy of the ejected photoelectron) of the interfering waves and the distance between the central atom and the backscattering neighbours, the interference can be constructive, leading to a higher value for the absorption coefficient m (see Fig. 4.13e) or destructive, giving rise to a lower value for m (Fig. 4.13d) relative to the case when the central atom would not have any backscattering neighbours. The net result is a series of oscillations on the high photon energy side of the absorption edge, called the X-ray absorption fine structure (XAFS). After transforming the energy scale of the XAS spectrum to wave numbers, the EXAFS oscillations x(k) are isolated from the total absorption coefficient curve by fitting a “background absorption” curve m0(k) to it and using the expression: xðkÞ ¼
mðkÞ 2 m0 ðkÞ m0 ðkÞ
The different coordination shells around the absorbing atom each contribute to the total EXAFS signal. By means of a Fourier transformation of kn xðkÞ (where n is an integer, e.g., 2 or 3) the contributions can be sorted, yielding a radial distribution function (RDF) around the absorbing atom. This function shows peaks at a distance where the neighbouring atoms that cause the EXAFS oscillations are situated. The RDFs contain information on the ˚ . By isolating and environment of the absorbing element out to ca. 5 A individually fitting the contributions to x(k), the true interatomic distances and number of neighbouring atoms per shell can be determined. More detailed information on the theory, techniques, and applications of XAS can be found in the book by Koningsberger and Prins [9]. Stern [10] and Durham [11] describe in a relatively brief form the theory of EXAFS and XANES spectroscopies. Applications of XAS in materials and earth sciences are summarized by Wong [12] and Brown et al. [13]. Since its development in the early 1970s, synchrotron-based XAS has proven itself to be a versatile structural probe for studying the local environments of cations in a variety of materials ranging from crystalline
145
K. Janssens
solids, glasses, and high-temperature liquids to aqueous sorption systems which involve metal complexes associated with (or sorbed at) solid/water interfaces. XAS is an atom-specific, local structure probe. Both physical and electronic structure are probed, but the probe range is generally only the first two shells of atoms around the absorber atom, which is generally less than ˚ . In some special cases (e.g., when the sample is highly crystalline and 5–6 A the second and more distant shells around the absorbing atom contain atoms that backscatter photoelectrons well), information about more distant shells can be obtained, usually with greater uncertainty than in the near-shell cases. XAS generally requires a synchrotron X-ray source for several important reasons: – A high X-ray flux is required in an XAS experiment in order to obtain high signal-to-noise data in a reasonable time frame (of the order of 30 – 40 min per spectrum). This requirement is particularly critical if the element of interest is at low concentration in a sample. Synchrotron sources provide X-rays of five or more orders of magnitude greater flux than conventional laboratory X-ray sources. – A broad spectral range at uniformly high flux is required because a typical X-ray absorption spectrum covers about 1000 eV. Tunable monochromators with appropriate d-spacings can be used to scan through a broad range of energy; thus one can choose the most appropriate energy range for an experiment. – High stability in flux, energy, and beam position is required in an XAS experiment and can be achieved with a synchrotron X-ray source. The main advantages of XAS over other structural methods are its element specificity and the fact that it can be used with practically any atom in any state of organization (solid, liquid, or gas). In addition, the sensitivity of XAS can be at the tens to hundreds of ppm level of an element, so that it can be used to study the structural environment of an element at trace levels in a chemically complex matrix. These attributes coupled with its ability to provide quantitative information on interatomic distances and the number and identities of atoms in the first and, in favourable cases, the second shell around an absorber, make XAS an extremely versatile and often unique probe of an atom’s local environment. The local nature of the XAS probe is both an advantage (e.g., XAS is particularly useful in studying the environment of atoms in amorphous materials where there is no longrange order and where X-ray scattering methods suffer from lack of element specificity) and a disadvantage (e.g., XAS provides essentially no information
146
X-ray based methods of analysis
on long-range order in a solid, including cation order –disorder among nonequivalent sites in a crystal structure). Disadvantages of XAS relative to XRD includes a limited resolution for sets of similar bond lengths and a high sensitivity to disorder. 4.3
INSTRUMENTATION FOR X-RAY INVESTIGATIONS
XRF spectrometry typically uses a polychromatic beam of short wavelength / high-energy photons to induce the emission of longer wavelength/lower energy characteristic lines in the sample to be analysed. Modern X-ray spectrometers may use either the diffracting power of a single crystal to isolate narrow wavelength bands (wavelength-dispersive XRF (WDXRF)) or an energy-selective detector may be employed to isolate narrow energy bands (energy-dispersive XRF (EDXRF)) from the polychromatic radiation (including characteristic radiation) that is produced in the sample. Because the relationship between emission wavelength and atomic number is known, isolation of individual characteristic lines allows the unique identification of an element to be made and elemental concentrations can be estimated from characteristic line intensities. Thus this technique is a means of material characterization in terms of chemical composition. WDXRF instrumentation is almost exclusively used for (highly reliable and routine) bulk analysis of materials, e.g., in industrial quality control laboratories. In the field of EDXRF instrumentation, next to the equipment suitable for bulk analysis, several important variants have evolved in the last 20 years. Both total reflection XRF (TXRF) and microscopic XRF (m-XRF) are based on the spatial confinement of the primary X-ray beam so that only a limited part of the sample (þsupport) is irradiated. This is realized in practice by the use of dedicated X-ray sources, X-ray optics, and irradiation geometries. In XRD instruments, on the other hand, a (quasi)monochromatic beam is normally employed to irradiate the samples, unless a form of energydispersive diffraction is being employed. In instruments employing laboratory sources, the wavelength of the radiation is fixed and equal to the most intense emission line of the X-ray tube anode being used. A typical X-ray tube is schematically shown in Fig. 4.14. Frequently encountered anode materials and the corresponding primary wavelengths and X-ray tube operating conditions are listed in Table 4.2. XAS makes use of synchrotron radiation sources and double crystal (or more sophisticated) monochromators in order to reach a level of
147
K. Janssens
Fig. 4.14. Cross-section of a sealed X-ray tube.
monochromaticity (DE/E or Dl/l) of around 1024 (see below) in experimental setups dedicated to this type of measurements. 4.3.1
X-ray sources
When an energetic electron beam impinges upon a (high-Z) material, X-rays in a broad wavelength band are emitted. This radiation is called bremsstrahlung as it is released during the sudden deceleration of the primary electrons, as a result of their interaction with the electrons of the lattice atoms in the target. At each collision, the electrons are decelerated and part of the kinetic energy lost is emitted as X-ray photons. TABLE 4.2 X-ray tube anode materials and associated characteristics Anode material
Cr
Fe
Cu
Mo
Atomic number ˚) l(Ka1) (A ˚) l(Ka2) (A ˚) l(Ka) (A ˚) l(Kb1) (A Critical potential (kV) Operating voltage range (kV) Operating current (mA)
24 2.2896 2.2935 2.2909 2.0848 5.99 30–40 10
26 1.9360 1.9399 1.9373 1.7565 7.11 35–45 10
29 1.5405 1.5443 1.5418 1.3922 8.98 35 –45 20
42 0.70926 0.71354 0.71069 0.63225 20.0 50 –55 20
148
X-ray based methods of analysis
(In addition, characteristic X-ray lines (see below) of the target materials are also produced.) Since during one collision, an electron of energy E can lose any amount between zero and E, the resulting bremsstrahlung continuum features photons with energies in the same range. On a wavelength scale, the ˚ ) ¼ 12.4/Emax continuum is characterized by a minimal wavelength lmin (A (keV) ¼ 12.4/V (kV) where Emax is the maximum energy of the impinging electrons and V the potential used to accelerate them. The continuum distribution reaches a maximum at 1.5 – 2lmin so that an increase in the accelerating potential V causes a shift of the continuum towards shorter wavelengths. In Fig. 4.15 bremsstrahlung spectra emitted by X-ray tubes operated at different accelerating potentials are shown. Four different types of X-ray sources are being employed in X-ray analysis: (a) sealed X-ray tubes and (b) radioactive sources are the most commonly employed, while to a lesser extent primary X-rays produced in (c) rotating anode tubes and (d) synchrotron radiation facilities are also utilized for analytical purposes. Most commercially available X-ray spectrometers utilize a sealed X-ray tube as an excitation source, and these tubes typically employ a heated tungsten filament to induce the emission of thermionic electrons in a vacuum chamber. Figure 4.14 shows a schematic cross-section of a sealed X-ray tube. After acceleration by means of a high voltage V, the electrons are directed towards a layer of high-purity metal (e.g., Cr, Rh, W, Mo, Rh, Pd,…) that serves as anode. In the metal layer, the bremsstrahlung continuum is produced upon which the characteristic lines of the anode material are superimposed (see Fig. 4.15). The shape of the emission spectrum can be modified by changing the electron acceleration voltage. The broad band radiation is well suited for the excitation of the characteristic lines of a wide range of atomic numbers. The higher the atomic number of the anode material, the more intense the beam of radiation produced in the tube. In typical X-ray tubes employed in XRF spectrometers, accelerating voltages of 25 – 50 kV are used, while electron currents in the range of 20 – 50 mA are employed. For WDXRF, frequently, 3 kW X-ray tubes are used; in EDXRF spectrometers, depending on the manner of sample excitation, tubes in the 50 – 1000 W range are employed. The efficiency of an X-ray tube is relatively low: only about 1% of the electric power is converted into X-rays, the rest is dissipated as heat. Accordingly, the anode of high-power tubes (.100 W) usually is water-cooled to avoid melt down of the metal block. A key factor in the design of a X-ray tube is the maximum powder loading (expressed in W/mm2) it can stand. The high-voltage power supplies used together with X-ray tubes are highly stable so that a wide conical X-ray beam
149
K. Janssens
Fig. 4.15. Polychromatic excitation spectra emitted by an Rh X-ray tube operated at various accelerating voltages. The excitation spectrum consists of a bremsstrahlung continuum upon which the characteristic lines of the anode material are superimposed.
of nearly constant intensity (to within a few percent relative) is being emitted. For applications requiring higher power levels as 3 kW, rotating rather than fixed anode tubes are employed. In these devices, the anode is a fast-spinning water-cooled metal cylinder covered with the desired anode material. During each revolution of the anode, only a small area on the surface is bombarded by the electrons during a short fraction of the time, so that the rest of the period can be used for heat removal. Rotating anode tubes that can be operated up to a total power of 18 kW are commercially available. In most XRD experiments only one well-defined wavelength of radiation is needed. Quasi-monochromatic X-rays of fixed wavelength can be obtained by filtering the white spectrum emerging from an X-ray tube. Filtering is probably the cheapest and simplest way of obtaining approximately monochromatic X-rays. One can achieve far “cleaner” radiation by using a monochromator, however, the cost of a monochromator could be 10,000 times the cost of a thin sliver of metal of the correct thickness. Table 4.3 lists suitable filters for various X-ray generators. In Fig. 4.16, next to an X-ray
150
X-ray based methods of analysis TABLE 4.3 Composition and thickness of Kb filters for commonly used X-ray generators to reduce the integrated Kb1/Ka1 intensity ratio to 1/500 Anode material
b-filter
Filter thickness (mm)
Areal density (g/cm2)
Loss (%) at Ka1
Ag
Pd Rh Zr Ni Co Fe Mn Mn2O3 MnO2 V V2O5
92 92 120 23 20 19 18 42 42 17 56
0.110 0.114 0.078 0.020 0.017 0.015 0.013 0.019 0.021 0.010 0.019
74 73 71 60 57 54 53 59 61 51 64
Mo Cu Ni Co Fe
Cr
emission curve of a Mo-anode X-ray tube, the absorption curve of a Zr foil is also shown. The absorption edge of Zr falls just in between the Ka and Kb lines of Mo. In other words, if we pass Mo radiation through a sheet of Zr metal, the Zr metal will absorb the Mo-Kb radiation far more strongly than the Ka photons. Figure 4.16 also shows schematically what the resulting distribution of radiation intensities will look like after filtering. Radioactive a-, b-, and g-sources may also be employed for EDXRF analysis. Generally, these sources are very compact compared to X-ray tubes and can, e.g., be used in portable analysis systems. a-sources are suited for the analysis of low atomic number elements. Frequently used sources are 244 Cm, with a half-life (t1/2) of 17.8 y that emits 5.76 and 5.81 MeV a-particles, and 210Po, having a half-life of 138 days and emitting 5.3 MeV as. b-sources can also be employed, either for direct EDXRF excitation of a sample or for producing bremsstrahlung radiation in a target to be used for subsequent sample excitation. 22Na (t1/2 ¼ 2.6 y), 85Kr (t1/2 ¼ 10.7 y) and 63Ni (t1/2 ¼ 100 y) are b-emitters that can be used for the former purpose, emitting, respectively, b2-particles of ca. 550, 670 and 66 keV. For bremsstrahlung production, 147Pm (t1/2 ¼ 2.6 y, 225 keV) in combination with a Zr target and 3H (t1/2 ¼ 12.4 y, 19 keV, Ti target) are useful. In Table 4.4, some characteristics of radioactive sources emitting X-ray or g-ray lines are listed. The X-ray emitting sources usually contain nuclides that decay by means of the electron-capture mechanism. During the decay,
151
K. Janssens
Fig. 4.16. Original and filtered emission curves of a Mo-anode X-ray tube. The absorption curve of Zr is also indicated.
an inner shell electron is captured by the neutron-deficient nucleus, transforming a proton in a neutron. This results in a daughter nuclide that has a vacancy in one of its inner shells, which results in the emission of corresponding characteristic radiation. For example, when a 55Fe nucleus (26 protons and 29 neutrons) captures a K-electron and becomes a 55Mn nucleus, an Mn K – L3,2 (Mn-Ka) or K – M3,2 (Mn-Kb) photon will be emitted. Other sources (such as 241Am or 57Co) emit g-rays of suitable energy as a result of different nuclear transformations. In Fig. 4.17, the range of elements that can be usefully analysed by means of various radioactive and X-ray tubes sources is summarized. In a number of specialized cases, X-ray analysis experiments also make use of synchrotron sources. Synchrotron radiation (SR) is produced by high-energy (GeV) relativistic electrons or positrons circulating in a storage ring. This is a very large, quasi-circular vacuum chamber where
152
X-ray based methods of analysis TABLE 4.4 Radioactive sources used for XRF analysis (flux in photons/s/sr) Radioisotope
Half-life (years)
X-ray or g-ray energy (keV)
Flux
55
2.7 88 1.3 0.16 433 0.66 0.74
5.9– 6.5 (Mn-K X-rays) 14.6–22 (U-L X-rays) 22–25 (Ag-K X-rays) 27–32 (Te-K X-rays) 59.6 (g-ray) 41.48 (Eu-K X-rays) 122.136 (g-ray)
7 £ 106
Fe Cm 109 Cd 125 I 241 Am 153 Gd 57 Co 244
8 £ 106 6 £ 107 4 £ 108 4 £ 106
strong magnets force the particles on closed trajectories. X-radiation is produced during the continuous acceleration (change in velocity vector in this case) of the particles. SR sources are several (6 – 12) orders of magnitude more bright than X-ray tubes, have a natural collimation in the vertical plane and are linearly polarized in the plane of the orbit. The spectral distribution is continuous when the emission of radiation is induced by bending magnets (see Fig. 4.18a), and the most simple way of employing SR is to use the full white beam to irradiate the sample (see below—m-XRF). When more sophisticated magnetic arrays called undulators are used to produce the radiation, even more intense X-ray beams can be generated. The photons are emitted in specific energy bands called harmonics (see Fig. 4.18b). Undulators are complex multi-pole magnetic structures of which a number of parameters such as the distance between the magnet poles and the magnetic field strength can be modified during experiments to suit the users needs. Figure 4.18b shows the energy spectrum of the undulator device at ESRF ID22. By changing undulator parameters such as the gap width and the magnetic field strength, the energy of the harmonics can be adjusted so that the output flux of an undulator in a specific energyrange can be optimized. In view of their quasi-monochromatic nature, undulator sources therefore are more suitable for performing m-XRF experiments involving monochromatic primary micro-beams, of which the energy can optionally be tuned. In this manner, it is possible to employ selective excitation of a series of elements in the sample, yielding optimal detection conditions (see below—TXRF). An additional advantage is the high degree of polarization of synchrotron radiation, causing spectral backgrounds due to scatter to be greatly reduced when the detector is placed at 908 to the primary beam and in the storage ring plane.
153
K. Janssens
Fig. 4.17. Range of elements that can be analysed using (top) radioactive sources, (bottom) X-ray tubes with different anodes, showing excitation of K- and L-lines.
The combination of a high primary beam intensity and low spectral background causes detection limit values of synchrotron XRF to go down to the ppb level. A disadvantage of the use of synchrotron radiation is that the source intensity decreases with time (due to a gradual loss of orbiting particles in between ring refills) so that measurement of unknown samples must be bracketed between standards and/or by continuously monitoring the primary beam intensity. The fact that synchrotron radiation can be efficiently focused into (sub)microscopic beams also permits very small samples to be employed during XAS and XRD experiments.
154
X-ray based methods of analysis
Fig. 4.18. (a) Energy distribution in the white beam used for sample excitation at Beamline L, HASYLAB (DORIS storage ring), Germany; for comparison, the white spectrum produced by the National Synchrotron Light Source (NSLS, Upton, NY, USA) is also shown; (b) Typical energy spectrum emitted by the undulator source at the ID22 Beamline, Grenoble, France. bw: bandwidth.
4.3.2
X-ray detectors
As any radiation detector, an X-ray detector is a transducer for converting X-ray photon energy into easily measurable and countable voltage pulses. All detector types work through a process of photoionization in which the interaction between an entering X-ray photon and the active detector material produces a number of electrons. By means of a capacitor and a resistor, the current produced by the electrons is converted to a voltage pulse,
155
K. Janssens
in such a way that one digital voltage pulse is produced for each X-ray photon that enters the detector. Next to being sensitive to photons of the appropriate energy range, there are two other important properties that the ideal detector should possess: proportionality and linearity. A detector is said to be proportional when the height of the voltage pulse that is produced upon entry of a photon is proportional to the energy of the photon. Proportionality is needed when, through pulse-height selection, only pulses of a particular height, i.e., corresponding to X-ray photons within a specific energy band, are to be measured. When the rate with which voltage pulses are being recorded is the same as the rate with which X-ray photons enter the detector, the latter is said to have a linear response. This property is important when the recorded count rates of various X-ray lines are to be used as measures of the photon intensities of these lines produced in a sample. The detector resolution is the precision/repeatability with which the energy of a specific type of X-ray photons (e.g., the Mn-Ka line at 5.9 keV) can be determined and is therefore a measure of the capability of the detector to distinguish between X-rays of very similar energy but different origin (e.g., the As-Ka1 line at 10.543 keV and the Pb-La1 line at 10.549 keV). In WDXRF spectrometers (see section 4.3.3.1), gas flow proportional ˚ ) and scintillation counters (for counters (for long wavelengths, l . 2 A ˚ wavelengths shorter than 2 A) are used to count X-rays. Both types of detectors usually are combined in a tandem detector that covers the entire wavelength range used in WDXRF spectrometry. Since neither of these detectors has a sufficient resolution to separate multiple wavelengths/ energies on its own, they are employed together with an analysing crystal. In case of energy-dispersive spectrometry, solid-state detectors of higher resolution are used. A gas flow proportional counter (see Fig. 4.19) consists of a cylindrical tube about 2 cm in diameter, carrying a thin (25 – 50 mm) wire along its radial axis. The tube is filled with a mixture of an inert gas and a quench gas—typically 90% argon/10% methane (P-10). The cylindrical tube is grounded and a voltage of ca. 1400 – 1800 V is applied to the central wire. The wire is connected to a resistor shunted by a capacitor. An X-ray photon entering the detector produces a number of ion pairs (n), each comprising one electron and one Arþ ion. The first ionization potential for argon is about 16 eV, but competing processes during the conversion of photon energy to ionization cause the average energy required to produce an ion pair to be greater than this amount. The fraction relating the average energy to
156
X-ray based methods of analysis
Fig. 4.19. Schematics of a gas-filled proportional counter.
produce one ion pair, to the first ionization potential, is called the Fano factor F. For argon, F is between 0.5 and 0.3 and the average energy 1 required to produce one primary ion pair is equal to 26.4 eV. The number of ion pairs produced by a photon of energy E will equal: n ¼ E=1
ð4:13Þ
Following ionization, the charges separate with the electrons moving towards the (anode) wire and the argon ions to the grounded cylinder. As the electrons approach the high field region close to the anode wire they are accelerated sufficiently to produce further ionization of argon atoms. Thus a much larger number N of electrons will actually reach the anode wire. This effect is called gas gain, or gas multiplication, and its magnitude is given by M ¼ N/n. For gas flow proportional counters used in X-ray spectrometry M typically has a value of around 105. Provided that the gas gain is constant the size of the voltage pulse V produced is directly proportional to the energy E of the incident X-ray photon. In practice not all photons arising from
157
K. Janssens
photon energy E will be exactly equal to V. There is a random process associated with the production of the voltage pulses and the resolution of a counter is related to the variance in the average number of ion pairs produced per incident X-ray photon. While the gas flow proportional counter is ideal for measurement of longer wavelengths, it is rather insensitive to wavelengths shorter than ˚ . For this shorter wavelength region it is common to use a about 1.5 A scintillation counter (see Fig. 4.20). The scintillation counter consists of two parts, the phosphor (scintillator) and the photomultiplier. The phosphor is typically a large single crystal of sodium iodide that has been doped with thallium, denoted as a NaI(Tl) crystal. When X-ray photons fall onto the phosphor, blue light photons are produced (with a wavelength of 410 nm), where the number of blue light photons is related to the energy of the incident X-ray photon. These visual light photons produce electrons by interaction with the surface of the photocathode in the photomultiplier, and the number of electrons is linearly increased by a series of secondary surfaces, called dynodes, inside the photomultiplier. The current produced by the photomultiplier is then converted to a voltage pulse, as in the case of the gas flow proportional counter. Since the number of electrons is proportional to the energy of the incident X-ray photon, the scintillation counter also has a proportional response. Because of inefficiencies in the X-ray/visual-light/ electron conversion processes, the average energy to produce a single event within a scintillation counter is more than a magnitude higher than the equivalent process in a flow counter. For this reason, the resolution of scintillation counters is much worse than that of flow counters. The output pulses produced by both the above-mentioned detectors are further processed by a linear amplifier and a discriminator circuit. Usually the number of pulses is counted during a preset amount of time and the accumulated counts stored in computer memory for display and further processing. The processing of an X-ray event by the detector and its associated electronics takes a finite amount of time. After the arrival of one X-ray the detection system is said to be “dead” during this length of time, because X-rays arriving within this dead period will not be counted. The dead time is of the order of 200 – 300 ns after the arrival of each photon; this implies that count rates of up to 106 photons/s can be handled. The detectors used in the various forms of EDXRF are semiconductor detectors. Conventionally, two types, i.e., lithium drifted silicon [Si(Li) and hyperpure germanium (HP-Ge)] detectors are used. Their main advantages are their compact size, the non-moving system components, and relatively good energy resolution, which optimally is of the order of 120 eV at 5.9 keV.
158
X-ray based methods of analysis
Fig. 4.20. Schematics of a scintillator detector.
159
K. Janssens
Because of their operating principles, these detectors have an inherent simultaneous multi-element capacity, which leads to a short measuring time for all elements as the detectors select the energy and collect counts at the same time. Disadvantages include the need for liquid nitrogen (LN2) cooling during operation, the necessity of having a relatively thin (8 – 25 mm) Be window and the fact that the maximum processable number of counts is limited to about 40,000 cps. This figure can be increased to 100,000 cps, but with loss of optimal performance characteristics. The detector crystal itself typically is a disk of very pure Si or Ge with dimensions of 4 – 10 mm diameter and 3 – 5 mm thickness. Even after careful production of the Si ingots from which the disks are cut, some trace impurities in the Si lattice will be left. To compensate and bind all free electrons, lithium ions are drifted (allowed to diffuse at elevated temperature) into the silicon crystal to neutralize the Si crystal defects in a particular zone, the so-called intrinsic zone. Afterwards, Au contacts are evaporated onto the crystal and a reverse voltage applied. In the crystal, the energy difference 1 (band gap) between the valence and conduction band is 3.8 eV. At room temperature, the conduction band is partially populated so that the crystal is a (semi)conductor. To keep the leakage current as low as possible, the crystal is cooled with LN2 by placing it in a vacuum cryostat. At 2 1968C almost all electrons remain in the valence band. The radiation to be measured needs to enter the cryostat through a thin entrance window, usually made of Be. By applying a reverse voltage to the charge carrier free intrinsic zone, an absorbed X-ray photon is converted into charge by ionization. Electrons are promoted from the valence to the conduction band, leaving “positive holes” in the valence band; thus the crystal temporarily becomes conducting. n ¼ E/1 of electron – hole pairs are created. The electrons and holes are quickly swept to the contact layers by the electric field created by the applied reverse bias on the crystal. Figure 4.21 shows the operation principle schematically. The charge induces a signal at the gate of a cooled field effect transistor (FET) that is the input stage of a charge sensitive preamplifier. The output signal is fed to a pulse processor that shapes the pulse and amplifiers it further. This signal is in the range up to 10 V and is proportional to the energy of the absorbed photon. The pulse height is digitized by means of an analog-to-digital converter (ADC) and the resulting digital value stored in a multi-channel analyser (MCA). This is an array of memory cells, called channels; by using the digital value associated with a single event as address offset into the memory array, the content of the appropriate channel is incremented with one count. Thus, all detector events having the same pulse height are stored
160
X-ray based methods of analysis
Fig. 4.21. Scheme of the working principle of a Si(Li) detector.
in the same channel. For example, upon entry in the detector of a Cu-Ka1 photon (E ¼ 8.05 keV), 2117 electron – hole pairs may be generated, which may lead to the formation of a preamplifier voltage pulse of, e.g., 42.0 mV. After further amplification and shaping, this is converted into a bell-shaped pulse of 4.20 V; the pulse-height is then digitized by an ADC, resulting for instance in a digital number of 420. Ultimately, this causes the content of channel 420 to be incremented with one count. After readout, the MCA memory (typically 1024 or 2048 channels in size, each corresponding to a 10 – 20 eV wide energy range) yields a pulse-height distribution of the detected events or an energy-dispersive X-ray spectrum, as shown in Fig. 4.22. In the spectra, always a broadening of the X-ray lines can be observed, i.e., the counts associated with photons of a specific energy, which normally should end up in a single channel, are distributed in a quasi-Gaussian fashion over several adjacent channels. This gives rise to the bell-shaped X-ray peaks in the spectrum. This line-broadening is caused by statistical fluctuations in the number of electron– hole pairs created when an X-ray photon of a given energy enters the detector; electronic noise in the amplifiers causes the uncertainty on the pulseheight to increase further. Even under conditions in which all noise contributions in the electronics are minimized, the line broadening remains a significant phenomenon, causing frequent peak overlap to occur in X-ray spectra, e.g., between lines of adjacent elements such as the Mn-Kb and Fe-Ka peaks. The resolution of energy-dispersive detectors 161
162 K. Janssens
Fig. 4.22. Energy-dispersive XRF spectrum of a multi-element standard, obtained in a TXRF spectrometer.
X-ray based methods of analysis
conventionally is expressed as the full-width-at-half-maximum of the Mn-Ka (Mn K –L2,3) peak (at 5.98 keV) and typically is around 150 eV. In the most optimal case, this value can be also low as 120 eV. The time to process and X-ray event (dead time) is of the order of 10 – 30 ms; conventional EDXRF spectrometers can therefore only operate at count rates up to 40,000 cps. In view of the presence of a Be window in the detector cryostat, X-ray photons below 2 keV are hard to detect in a conventional Si(Li) detector, although thin-window models are commercially available. Roughly since 1995, several types of compact and thermoelectrically cooled ED detectors have become available. The most significant advantage of these detectors is that they do not require liquid nitrogen cooling, allowing the instrument they are incorporated in to be much smaller. These type of detectors is suitable for employment in portable equipment. Thermoelectrically cooled Si-PIN, Cd12xZnxTe (CZT) and HgI2 detectors are fairly inexpensive devices. The currently available Si-PIN diode detectors mostly have a thickness of about 300 mm which makes the detector useful up to X-ray energies of 20 keV and an energy resolution in the range 180 – 200 eV at Mn-Ka, i.e., slightly worse than that of Si(Li) or HPGe detectors. CZT detectors are targeted towards the higher energy range with a thickness of up to 2 mm, allow efficient detection of X-rays up to 150 keV with a resolution of ca. 250 eV at Mn-Ka (5.9 keV) and 1 keV at 60 keV. Similarly, HgI2 detectors (with thicknesses of a few millimetres) can also be used in this range with a resolution of ca. 200 eV at Mn-Ka. A very useful and more recent type of solid-state detector is the solid-state drift chamber (SSD) detector, featuring excellent energy resolution at high count rates. A FWHM below 140 eV at 5.9 keV can be achieved with thermoelectrical cooling (Peltier effect). SSDs exist in a large variety of sizes up to 2 cm2 diameter. They still show excellent spectroscopic behaviour at count rates as high as 2 £ 106 counts/cm2/s. The compact design, the relatively low price, the absence of the need for liquid nitrogen cooling, the high count rate capability and the insensitivity to noise pick-up make these systems attractive alternatives to conventional semiconductor detectors. The resolution of a number of different of X-ray detectors in the range 1 – 100 keV (wavelengths of ca. 1 –0.01 nm) are compared in Fig. 4.23. It is clear that scintillators and proportional counters are not even able to separate the Ka-lines of adjacent elements whereas this is the case for most 163
K. Janssens
Fig. 4.23. Energy resolution (expressed as FWHM of the Ka line of a given energy) of different X-ray detectors in the 1–100 keV range. The difference in Ka line energy between adjacent elements is also shown (symbols).
of the solid-state detectors. A more extensive overview of recently developed X-ray detectors can be found elsewhere [14]. In XRD instruments, typically employing radiation in the 8 – 18 keV range, usually collimated scintillator detectors are mounted on the 2u arm of the goniometer. In XAS setups, in many cases, ionization chambers similar to the abovementioned proportional counters are employed to measure the incoming and transmitted X-ray beam intensities. 4.3.3
X-ray fluorescence instrumentation
4.3.3.1 Wavelength-dispersive XRF A typical WDXRF system consists of an X-ray tube, a specimen support holder, a primary collimator, an analysing crystal and a tandem detector. The typical WDXRF irradiation/detection geometry is shown in Fig. 4.24. Wavelength-dispersive spectrometers employ diffraction by a single crystal to separate characteristic wavelengths emitted by the sample. A single crystal of known interplanar spacing d is used to disperse the collimated polychromatic beam of characteristic wavelengths that is coming from the sample, such that each wavelength l will diffract at a specific angle u, given by Bragg’s law.
164
X-ray based methods of analysis
Fig. 4.24. Schematic drawing of a wavelength-dispersive XRF spectrometer.
A goniometer is used to maintain the required u/2u relationship between sample and crystal/detector. Prior to impinging on the analyser crystal, by means of a collimator or slit, the spread in initial directions of the sample-to-crystal beam is limited. Since the maximum achievable angle on a typical WDXRF spectrometer is around 738, the maximum wavelength that can be diffracted by a crystal of spacing d is equal to ca. 1.9d. The angular dispersion du/dl of a crystal with spacing 2d is given by: du n ¼ dl 2d cos u and is therefore inversely proportional to its d-spacing. Thus, high dispersion can only be obtained at the expense of reducing the wavelength range covered by a particular crystal. Several crystals therefore are likely to be employed for covering a number of analyte elements. Typically, 4 – 6 different analyser crystals (with different d-spacings) and two different collimators are provided in this type of instrument, allowing for a wide choice in dispersion conditions. The smaller the d-spacing of the crystal, the better the separation of the lines, but the smaller the wavelength range that can be covered. The separating power of the crystal spectrometer is dependent upon
165
K. Janssens
the divergence allowed by the collimators (which mainly determine the width of the diffracted lines) in the 2u spectrum, but the angular dispersion of the analysing crystal itself and the intrinsic width of the diffraction lines also play a role. In Table 4.5, some characteristics of a few commonly employed analyser crystals are listed. Classically, large single crystals have been used as ˚ ), only a dispersive elements. For dispersion of long wavelengths (.8 A limited number of natural materials are available; the most commonly ˚ ), allowing measureemployed is thallium acid phthalate (TAP, 2d ¼ 26.3 A ment of the Mg, Na, F and O-K lines. As alternative, several other materials with large 2d-spacings have been used and since the 1980s layered synthetic multilayers (LSMs) are in use. These consist of stacks of alternate electronrich (e.g., W) and electron-poor (e.g., graphite) layers of atoms or molecules, deposited on a sufficiently smooth substrate. Since the composition and interplanar distance of the LSM to a certain extent can be optimized for particular applications, a factor 4 – 6 improvement in peak intensities compared to TAP crystals can be achieved. Among wavelength-dispersive spectrometers, a distinction can be made between single-channel instruments and multi-channel spectrometers. In the former type of instrument, a single-dispersive crystal/detector combination is used to sequentially measure the X-ray intensity emitted by a sample at a series of wavelengths when this sample is irradiated with a beam from a high power (2 –4 kW) X-ray tube. In a multi-channel spectrometer, many crystal/detector sets are used to measure many X-ray lines/elements simultaneously. Single channel instruments are also referred to as scanning spectrometers; this type is the most common. During an angular scan, the angle u between sample and analyser crystal is continuously varied; in order to maintain an identical angle between analyser crystal and detector, the
TABLE 4.5 Analysing crystals used in wavelength-dispersive X-ray spectrometry Crystal
Planes
˚) 2d (A
K-line range
L-line range
Lithium fluoride (LiF) Lithium fluoride (LiF) Pentaerythritol (PET) Thallium acid pthalate (TAP) LSMs
220 200 002 001 –
2.848 4.028 8.742 26.4 50– 120
.Ti .K Al –K F–Na Be –F
.La .Cd – – –
166
X-ray based methods of analysis
latter moves at double the angular speed as the crystal. In this manner, X-ray intensity vs. 2u diagrams are obtained. By means of tables, the recorded peaks can be assigned to the characteristic lines of one or more elements. In Fig. 4.25, a typical 2u-spectrum obtained from a brass sample is shown. Simultaneous wavelength-dispersive spectrometers were introduced in the early 1950s, and sequential systems about a decade later. At this time, about 30,000 or so wavelength-dispersive instruments have been supplied commercially. The two major categories of wavelength-dispersive X-ray spectrometers differ mainly in the type of source used for excitation, the number of elements that they are able to measure at one time, the speed at which they collect data and their price range. For high specimen throughput quantitative analysis where speed is of the essence, and where a high initial cost can be justified, simultaneous wavelength-dispersive spectrometers are optimal. For more flexibility, where speed is important but not critical and where moderately high initial cost can be justified, sequential wavelengthdispersive spectrometers are probably more suited. Both the instruments are, in principle at least, capable of measuring all elements in the periodic classification from Z ¼ 9 (F) and upwards, and most modern wavelength-dispersive spectrometers can do some useful measurements down to Z ¼ 6 (C). Both can be fitted with multi-sample handling facilities and automated. Both are capable of precision of the order of a few tenths of a percent and both have sensitivities down to the ppm level. Single-channel wavelength-dispersive spectrometers are typically employed for both routine and non-routine analysis of a wide range of products, including ferrous and non-ferrous alloys, oils, slags and sinters, ores and minerals and thin films. These systems are very flexible but, relative to multi-channel spectrometers, are somewhat slow. The multi-channel wavelength-dispersive instruments are used almost exclusively for routine, high-throughput analyses where there is need of fast accurate analysis, but where flexibility is of no importance. 4.3.3.2 Energy-dispersive XRF for bulk analysis Energy-dispersive spectrometers became commercially available in the early 1970s with the advent of high-resolution solid-state detectors; today there are of the order of 20,000 units in use. In principle, EDXRF instruments have a much simpler mechanical design than WDXRF instruments, as the detection system does not include any moving parts and the solid-state detector (most commonly an Si(Li) detector) itself acts as a dispersion agent. The high geometrical efficiency of the semiconductor detector permits a great variety in excitation conditions. The manner in which the radiation that
167
168 K. Janssens Fig. 4.25. Wavelength-dispersive X-ray spectrum of a brass sample, showing the characteristic lines of the major elements Cu and Zn, and the minor constituents Cr, Fe, Ni and Pb, superimposed on a continuous background.
X-ray based methods of analysis
originally exits from the X-ray tube is “pretreated” before it reaches the sample varies according to the type of EDXRF instrument. The final analytical capabilities and in particular the detection limits that can be attained by the instrument strongly depend on the sophistication with which this is done. In Fig. 4.26a, the most simple of EDXRF instrumental configurations is shown. A low power X-ray tube (e.g., 50 W) and a Si(Li) detector are both placed at an angle of 458 with respect to the sample. Collimators are used to confine the excited and detected beams to a sample area between 0.5 and 2 cm2. In such a “direct-excitation” configuration, the distance between the components can be fairly small (typically a few cm) and since both the tube anode lines and the bremsstrahlung component of tube output spectrum are used to irradiate the sample, only a limited tube power is required. Since the bremsstrahlung continuum not only ensures a uniform excitation of many elements, but also causes a significant scatter background to be present in the recorded EDXRF spectra, most direct-excitation systems are equipped with a set of primary beam filters to alter the tube spectrum. By selection of an appropriate filter, the excitation conditions for a particular range of elements can be optimized. In order to facilitate the determination of low-Z elements, commercial systems can be either evacuated or flushed with He, thus reducing the absorption of low energy radiation and scatter.
Fig. 4.26. Schematic drawings of (a) a direct-excitation XRF instrument, (b) a secondary target XRF instrument, (c) a polarized XRF instrument employing a cartesian (X YZ) irradiation geometry.
169
K. Janssens
The schematic of a “secondary target” EDXRF system is shown in Fig. 4.26b. In such a configuration, a high-power (1 kW) X-ray tube irradiates a metal disk (the secondary target, e.g., made of Mo), causing it to emit its own characteristic radiation lines (Mo-Ka and Mo-Kb). This “bichromatic” fluorescent radiation is then used to excite the sample to be examined. The advantage of the secondary target scheme is that, as a result of the bichromatic excitation, the background in the resulting EDXRF spectra is significantly lower as in the direct excitation case. This leads to better detection limits. By using a filter that preferentially absorbs the Kb component of the secondary target radiation (e.g., a Zr foil in case of a Mo target; see Fig. 4.16), a quasi-monochromatic form of sample excitation can be realized. By interchanging the target (and matching filter), different element ranges can be excited optimally. For example, to obtain the best conditions for determination of trace concentrations of the elements Rb – Nb in geological samples, an Rh secondary target may be selected while for optimal detection of Cr in the same material, a Cu target would be more beneficial. The stationary arrangement of components used in energy-dispersive X-ray fluorescence (EDXRF) is ideally suited for geometrical configurations that exploit polarization phenomena to reduce background and thereby improve signal-to-noise ratios. Figure 4.26c shows a configuration employed to achieve a reduction in the background level of EDXRF spectra obtained in direct excitation conditions. In this case, one or more energy bands of the tube emission spectrum are scattered and/or diffracted under (nearly) 908 by means of a suitable scatterer material and/or diffraction crystal. Because scattering rather than fluorescence is used to “reflect” the primary tube spectrum onto the sample, the X-ray beam that impinges on the sample is linearly polarized in the plane perpendicular to the tube – scatterer– sample plane. When the Si(Li) detector is also positioned in the former plane at 908 relative to the scatterer-sample axis, the lowest background level will be recorded. The reason for this background reduction is that the polarized photons will preferentially be scattered out of the plane of polarization and therefore will not reach the detector. The optimal geometrical configuration is therefore that tube, scatterer, sample and detector are arranged in an XYZ (also called “Cartesian”) geometry, as shown in Fig. 4.26c. For polarization of medium-to-hard radiation (E . 10 keV) by Barkla-scattering, fairly thick slabs of low-Z materials such as Al2O3, B4C and B3N are suitable materials. For polarization of softer radiation, the above-mentioned materials are not suitable since for
170
X-ray based methods of analysis
E , 10 keV, photoelectric absorption dominates over scattering. In the region 1 – 10 keV, radiation can be polarized through Bragg diffraction over 2u < 908 by using a suitable crystal. For example, highly oriented pyrolithic graphite (HOPG) is an excellent Bragg polarizer for the (002) reflection of the Rh-La radiation (u ¼ 43.28). Multiple layer scatterers, for example consisting of a thin layer of HOPG glued on top of an Al2O3 substrate, in combination with an Rh tube are useful to determine a wide range of elements simultaneously with good detection limits and sensitivities.
4.3.3.3 Portable EDXRF instruments Next to EDXRF spectrometers that are intended for use in the laboratory, a number of portable EDXRF instruments are also commercially available. These devices are used in various fields for on-site analysis of works of art, environmental samples, forensic medicine, industrial products, waste materials, etc. In its simplest form, the instruments consist of one or more radioisotope sources combined with a scintillation or gas proportional counter. However, also combinations of radiosources with thermoelectronically cooled solid-state detectors are available in compact and light-weight packages (i.e., below 1 kg). In Fig. 4.27, schematics of various types of radiosource-based EDXRF spectrometers are shown. The X-ray source can be present in the form of a ring; radiation from the ring irradiates the sample from below while the fluorescent radiation is efficiently detected by a solid-state detector positioned at the central axis. Shielding prevents radiation from the source to enter the detector. Figure 4.27 also shows X-ray sources of other shapes, requiring a different type of shielding. Next to equipment using radioisotopes as X-ray source, portable equipment that includes miniature
Fig. 4.27. Radioisotope-excited X-ray fluorescence analysis by means of (left) an annular source, (middle) a central source and (right) a side-looking source.
171
K. Janssens
Fig. 4.28. Photograph of a portable XRF analyser.
low power X-ray tubes is also available; in such devices, almost exclusively the direct excitation form of EDXRF is employed. As shown in Fig. 4.28, handheld devices allowing for rapid on-site identification of alloys and other inorganic materials are available from various companies, such as TN technologies, Oxford Instruments plc., Niton Inc., Metorex International Oy and EDAX Portable Products Division [15]. The most recent versions of these devices incorporate several radioisotope sources (e.g., a combination of 55Fe, 109Cd and 241Am is frequently encountered) in combination with a compact solid-state detector such as a drift chamber detector or a PIN diode. Usually such analysers are equipped with “standardless” data-processing software that is intended for use in a specific type of application, such as the sorting of waste metal or the detection of hazardous waste, by means of which the quantitative results can be calculated and displayed to the user immediately after the data collection. An overview of currently available instruments can be found in Ref. [16].
172
X-ray based methods of analysis
4.3.3.4 Total reflection XRF When X-rays impinge upon an (optically) flat material under a very small angle (typically a few mrad), i.e., nearly grazing the surface, total external reflection occurs. This means that instead of penetrating the material, the X-ray photons will only interact with the top few nm of the material and then be reflected. The material that is present on top of the reflecting surface will be irradiated in the normal manner, and will interact with both the primary and the reflected X-rays. The major difference between conventional EDXRF and TXRF therefore is the excitation geometry. In the standard case of EDXRF the angle between the primary incident radiation and the sample is 45deg; while the detector is placed normal to the incident beam so that the angle between sample and detector is also 458. The principle setup of TXRF is shown in Fig. 4.29. The largest angle at which total external reflection still takes place is called the critical angle of total reflection fcrit. A narrow, almost parallel beam impinges at angles below the critical angle on the surface of the reflector that carries the sample as randomly distributed micro-crystals in the centre part of its surface. Since the X-rays scarcely penetrate the reflector, the contribution from scattered primary radiation from the substrate is minimized. As a result of the double excitation of the sample by both the primary and the reflected beams, the fluorescent signal is practically twice as intense as in the standard EDXRF excitation mode. The critical angles are in the range of a few milliradians for typical reflector materials such as quartz or Si and primary radiation of 9.4 keV (from a W anode X-ray tube) or 17.5 keV (from a Mo anode tube). With higher energies
Fig. 4.29. Schematic layout of a TXRF spectrometer.
173
K. Janssens
in the exciting spectrum, adjustments must be made for the proper incident angle below the critical angle, which (for glass reflectors) is given by: fcrit ðmradÞ ¼ 20:7=EðkeVÞr1=2 ðg=cm3 Þ
ð4:14Þ
The main advantages of TXRF are: (a) The background caused by scattering of the primary radiation on the substrate is reduced. (b) The fluorescence intensity is doubled as the primary and reflected beams pass through the sample giving efficient excitation. (c) The distance between the sample on the reflector surface and the detector can be made small, thus the solid angle for detection is large. (d) Depending on the X-ray source and the spectral modification devices, the absolute limits of detection are in the pg range for instruments employing 2 –3 kW X-ray tubes and in the fg range when synchrotron radiation excitation is used (see below). Figure 4.22 shows a typical TXRF spectrum. TXRF permits to simultaneously determine minor to trace elements in samples of small volume. Additional advantages are insensitivity to matrix effects, easy calibration, fast analysis times and low costs. 4.3.3.5 Microscopic XRF The basic measuring strategy of microscopic X-ray fluorescence (m-XRF) analysis is illustrated in Fig. 4.30. This micro-analytical variant of bulk EDXRF is based on the localized excitation and analysis of a microscopically small area on the surface of a larger sample, providing information on the lateral distribution of major, minor and trace elements in the material under study. Essentially, a beam of primary X-rays with (microscopically) small cross-section irradiates the sample and induces the emission of fluorescent X-rays from a micro-spot. A suitable detector system collects the fluorescent radiation that carries information on the local composition of the sample. When the sample is moved either manually or under computer control in the X-ray beam path, either spot analyses, line-analysis or image collection is possible. The difficulties in the exploitation of this method reside with the production of sufficiently intense X-ray beams to allow for sensitive microanalysis. Techniques to do this have only recently appeared; in the past, X-rays were considered to be notably difficult to focus into beam of small dimensions. Variants on the basic mode of operation either reside with the method employed for X-ray beam concentration/focussing or with the source
174
X-ray based methods of analysis
Fig. 4.30. Principle of m-XRF.
type employed: conventional X-ray tubes or synchrotron radiation sources. Especially the increased performance of compact and relatively inexpensive X-ray focusing devices and in particular the development of (poly)capillary X-ray focusing optics, permitting X-ray beams to be focused to below 10 mm diameter spots, has made the development of m-XRF possible. When used in combination with X-ray tubes, absolute detection limits in the pg range are obtained for thin samples. In massive samples, relative detection limits of around 10 ppm have been reported for transition elements. At synchrotron facilities, the capabilities of the m-XRF method (both regarding spot sizes and detection limits) are significantly better: fg to ag level absolute detection limits are obtained with beams as small as 0.5 – 2 mm in diameter. By the use of monochromatic beams of polarized radiation, optimal peak-to-background ratios in the resulting EDXRF spectra can be obtained, resulting in relative detection limit values in the 10 – 100 ppb range in biological materials. The application of m-XRF to a great variety of problems and materials has been described, including geochemistry, archaeology, industrial problems and environmental studies [17]. Especially the fact that quantitative data on (trace) constituents can be obtained at the microscopic level without sample damage is of use in many different circumstances. Microfocussing X-ray optics Figure 4.31 illustrates the functioning principle of capillary optics. Through repeated total external reflection on the inner walls of the tube, X-ray photons are transported along the length of the tubes; thus, straight capillaries act as effective apertures over a wide energy range (0– 40 keV and beyond). Since metal pin holes or cross-slit systems are not effective to collimate the high-energy portions of the white beam of a
175
K. Janssens
Fig. 4.31. Functioning principle of capillary optics: (a) total external reflection off an optical flat surface; (b) multiple total reflection on the inner surface of a straight capillary; (c) X-ray photon propagation in a bent capillary (degree of bending is strongly exaggerated); (d) point-to-point focusing of X-rays with a polycapillary lens.
synchrotron source to a micro-spot with well-defined shape, straight capillaries are employed for this purpose. However, these devices do not “concentrate” the X-ray beam. In conical capillaries, the X-ray beam that enters the wide end of the taper is gradually “squeezed” to the taper dimensions at the fine end during the many reflections that take place along the way. In case ellipsoidal lead-glass capillaries are employed [18], the capillary is positioned in such a way that its tip is about 0.5 – 1 mm away from the sample surface. Inside ellipsoidal capillaries, the X-ray photons are subject to only one or a few reflections and part of the radiation is focused in this manner. A polycapillary lens consists of a large number of hollow glass fibres. The fibres are bent towards a common focal point, situated outside the polycapillary. While passing through these tubes, photons are redirected towards the focal point, resulting in focussing of the X-ray beam [19]. Polycapillary lenses can operate in the energy range from 2 to 30 keV and have a number of other properties that render them advantageous for use in m-XRF and m-XANES experiments. They have a focal point that is situated outside the lens; the focal distance is typically a few millimetres to a few centimetres, allowing for a safe distance between optic and sample. Also, the compact size of these devices contributes to their practical applicability. Mono- and polycapillary X-ray optics are very frequently used in m-XRF instruments since they are compact, robust and relatively inexpensive optical elements. In almost all laboratory m-XRF apparatus either capillary optics or pin-hole apertures are employed for focusing or collimating the polychromatic spectrum emerging from a micro-focus X-ray tube.
176
X-ray based methods of analysis
Capillary optics have the additional advantage of being “non-imaging optics”, indicating that the diameter of the X-ray micro-beams they generate is largely independent of the size of the X-ray source employed. Generally, polycapillary optics produce beam sizes in the range 10 – 50 mm while operating efficiently in the 5 – 25 keV range. A more detailed treatment of their characteristics can be found elsewhere [20]. At synchrotron m-XRF setups, also more sophisticated types of optics are also in use: compound refractive lenses (CRL) [21], Fresnel zone plates (FZP) [22] and curved mirror systems are examples of the latter [23]. These imaging optics are capable of producing X-ray beams with submicron dimensions but can only be used at synchrotron sources such as the European Synchrotron Radiation Facility (Grenoble, France). Laboratory m-XRF equipment Several commercial companies offer laboratory m-XRF instruments, including EDAX and Horiba. In Fig. 4.32, photographs of the exterior and interior of the EDAX Eagle instrument are shown. The heart of this spectrometer is a 40 W Mo or Rh anode micro-focus X-ray tube that can be interfaced to a monocapillary collimator or polycapillary lens, yielding X-ray beams in the range 20 mm to 3 mm. X-rays are detected with a 80 mm2 Si(Li) detector. This instrument is available with a large vacuum chamber in which medium-sized artefacts (up to 25 £ 20 £ 9 cm3 in size) can be analysed without encountering practical problems.
Fig. 4.32. Photograph of (a) the exterior and (b) the inside of the Eagle vacuum chamber, showing the positioning of an artefact on the motorized stage (shown retracted). By means of a camera, the interior can be observed when the vacuum chamber is closed.
177
K. Janssens
In general, the sensitivity of laboratory m-XRF setups is limited when compared to their synchrotron-based counterparts, as a result of the very large difference in primary intensity between regular X-ray tubes and synchrotron sources (Fig. 4.33). Provided appropriate spectrum collection times are employed (typically in the range 100 – 1000 s per point), relative detection limits in the range 20 – 50 ppm can be achieved (for transition elements in a glass matrix). This makes laboratory m-XRF instruments
Fig. 4.33. The ArTAX m-XRF spectrometer, mounted onto a flexible support, allowing easy positioning of the device in front or above large artefacts.
178
X-ray based methods of analysis
significantly more suitable for the determination of low levels of the elements Ti-Mo in various materials than scanning electron microprobes. A recent direction in m-XRF instrument development is the construction of compact and/or portable small-beam instruments, consisting, e.g., of an air-cooled mini-focus X-ray tube, a compact optical element for beam focussing/collimation and a Peltier-cooled energy-dispersive detector. Such instruments, offering beams of 50 – 200 mm cross-section, are very useful for in situ investigations of archaeological and artistic materials, i.e., in the museum, gallery or archaeological site they are normally present. In some cases, the spectrometer can be as simple as a compact X-ray tube equipped with a collimator tube, defining the X-ray beam to ca. 1– 2 mm diameter, an optional beam filter and a small pin-diode detector attached to the side of the tube. Such a system can be very useful for in situ analysis of bronze statues [24,25]. Other systems make use of drift chamber detectors [26,27]. Bronk et al. [28] have described the design of a mobile spectrometer for m-XRF with the requirements of archaeometry in mind. The ArtTAX instrument offers non-destructive and sensitive multi-elemental analysis, a sub-mm resolution with the possibility of working outside the laboratory. The spectrometer consists of an air-cooled, low-power molybdenum tube, a polycapillary X-ray lens that focuses the beam to ca. 100 mm, a silicon drift detector, a CCD camera, and three light diodes for sample positioning. The motor-driven measurement head is fixed on an x; y; z-flexible tripod support which can be assembled and dismantled within minutes. The detection limits of this device are in the range of 10 mg/g for transition elements in glass. Open helium purging in the excitation and detection paths enables the determination of elements down to sodium, thus avoiding vacuum conditions or a size-limiting sample chamber. To demonstrate the potential of ArtTAX in the field of art and archaeology, a selection of qualitative and quantitative results on pigment, metal, glass, and enamel analyses was presented [28]. Bichlmeier et al. [29,30] and Vittiglio et al. [31] have described another compact micro-XRF instrument, consisting of a similar set of compact components and applied it to the analysis of noble metals objects (such as coins, burial artefacts and statuettes in bronze and brass), multi-coloured beads and 20th century decorate glass objects and pigmented materials of various nature (18th century illuminated parchments various manuscripts from the 13th to the 19th centuries prepared with ferro-gallic ink.)
179
K. Janssens
Synchrotron m-XRF instrumentation In Fig. 4.34a the schematic layout of number of different m-XRF spectrometers that make use of synchrotron radiation are shown. The capabilities of some of these spectrometers is limited to elemental microanalysis while by means of other facilities, elemental analysis can be combined with other types of X-ray spectroscopy such as m-XANES and m-XRD. Polychromatic synchrotron m-XRF. The experimental setup used for polychromatic synchrotron m-XRF measurements installed at Beamline L of the HASYLAB (Hamburg, Germany) synchrotron laboratory is schematically depicted in Fig. 4.34a. Figure 4.18a shows the energy distribution in the white beam, produced in a bending magnet source, as seen through a 10 £ 10 mm2 pin-hole, positioned at 20 m from the DORIS synchrotron ring. The bending magnets of this storage ring produce a white spectrum that contains appreciable
Fig. 4.34. Schematics of (a) a polychromatic m-SRXRF spectrometer, (b) a combined m-SRXRF and m-XANES facility employing monochromatic excitation and (c) a confocal m-XRF excitation geometry, making use of pink-beam excitation. The setup shown in (d) is a monochromatic m-XRF/XRD facility at Beamlines ID18F and ID22, ESRF, Grenoble, France. Setups (a –c) are in operation at Beamline L, HASYLAB (DORIS III storage ring), Hamburg, Germany.
180
X-ray based methods of analysis
Fig. 4.34 (continued).
amounts of more energetic photons (above 60 keV), something which is not the case with many other synchrotron facilities, such as e.g., the National Synchrotron Light Source (NSLS, Upton, NY, USA). The beam that originates from the storage ring is first collimated down to ca. 100 £ 100 mm2 by motorized cross-slits before entering a straight glass capillary. The latter is mounted onto a motorized XYuw-stage for alignment to the beam. Usually, straight borosilicate glass capillaries of inner diameter 10 – 50 mm are employed for micro-beam formation [32]; the capillary – sample distance is typically of the order of a few mm. The sample itself is mounted onto a motorized XYZu stage, allowing it to be moved in increments
181
K. Janssens
of 1 mm and 0.18, respectively. The sample surface plane is vertical and oriented at 458 to the incoming micro-beam and is also in the focal plane of the long-distance optical microscope, which is placed horizontally. Fluorescent signals are detected by a HPGe (high-purity germanium) solid-state detector; the latter is located at a distance of 5 – 7 cm from the sample and is shielded by a Ta/Pb enclosure. The detector collects the fluorescent and scattered radiations in a solid angle of ca. 0.001 steradians. The vertical position of the detector is such that it is exactly in the (horizontal) plane formed by the storage ring and the micro-beam. Since the HPGe detector is oriented at 908 to the micro-beam in the plane of maximum linear polarization, the XRF spectra which are collected in this way feature optimal peak-to-background ratios as the scatter-induced continuum background is reduced in intensity [33]. Behind the sample, an ionization chamber is placed to monitor the transmitted beam intensity; a similar monitor can be placed in the beam path between slits and capillary (not shown in Fig. 4.34a). During conventional m-XRF measurements, involving point analyses, line scans or two-dimensional (2D) mapping, the sample is moved through the beam by means of the motorized stage (in the XYZ directions) so that the appropriate locations on the surface are irradiated; the rotation stage (u stage with vertically oriented rotation axis) normally is only useful during tomographic measurements [34]. Correlated stage movement and spectrum acquisition allows the collection of nx £ ny individual XRF spectra (nx, ny ¼ number of pixels in horizontal and vertical directions of the image) which during or after the acquisition can be processed to yield (net) elemental maps, line profiles or area/phase-specific sum spectra of the irradiated material. By means of appropriate calibration models, the latter can be converted into quantitative images or local composition values [35]. The analytical characteristics of this facility are described more in detail elsewhere [36]. A special feature of this experimental facility is the fact that K-line XRF measurements can be performed on the elements ranging from K ðZ ¼ 19Þ to Pb ðZ ¼ 82Þ because of the high-energy components of the white beam. In the transition metal range, fg-level absolute detection limits in thin samples and ppm-level relative detection limits in thick organic or silicate-based samples are obtained while for the lanthanides, equivalent values are situated around 10 fg and 5 – 10 ppm, respectively, within 1000 s of irradiation time [37]. Next to the high intensity of the synchrotron beam, the high sensitivity over an extended element range is also the consequence of the fact that in SR –XRF spectra, a reduced background level is observed. This is the result of the polarized character of the radiation and the fact that the XRF is detected in the plane of the synchrotron ring (see Ref. [38] for
182
X-ray based methods of analysis
details). A disadvantage of the use of the strongly penetrating primary and fluorescent radiation is that the analytical signals in thick samples can originate from extensive depths (e.g., for REE, this can be up to several mm in glass samples), thereby reducing the effective lateral resolution of the technique in one direction. This blurring effect can be avoided by employing thin samples. m-XRF and m-XANES using tunable monochromatic excitation. In the setup shown in Fig. 4.34b, a narrow energy-band is selected from the continuum by means of a fixed-exit Sik111l monochromator [39]. After passing cross-slits, the monochromatic beam is demagnified by a polycapillary lens. What is shown in Fig. 4.34b and c are so-called polycapillary halflenses, suitable for focussing of (quasi-parallel) synchrotron radiation. An important feature of this type of lens is the large beam acceptance area, typically several square millimetres in size. This permits to capture a large fraction of the photons in the impinging beam, resulting in a high throughput and gain factor. When the orientation of the monochromator crystals is kept constant, m-XRF measurements with monochromatic excitation can be performed. This excitation form results in XRF spectra with very low background levels, leading to relative detection limits situated in the 0.1 – 10 ppm range and at the 0.1 –1 fg level in thin samples. The increase in flux density is of such a magnitude that it becomes possible to perform m-XANES measurements at a second generation synchrotron bending magnet source where a very significant part of the initial beam is lost in the monochromatization process [40,41]. Confocal m-XRF and m-XANES in monochromatic and pink-beam mode. A striking difference between m-XRF and micro-analytical methods such as electron probe micro-analysis (EPMA), secondary ion microscopy (SIMS) and laser-ablation inductively coupled mass spectrometry (LA-ICP-MS) is that X-ray photons of sufficient energy can penetrate very deeply into the material being analysed. Since this penetration depth can be one to several orders of magnitude higher than the diameter of the focused X-ray beam, often a degradation of the lateral resolution is observed [42,43] when samples thicker than the beam size are investigated. In fact, rather than measuring the lateral distribution of the elemental constituents at the surface of the materials, a two-dimensional projection of the three-dimensional distribution of these constituents in the sample is obtained. This penetrative characteristic of X-ray beams is employed to advantage in XRF tomography experiments, where through the measurement of a series of the aforementioned projected distributions under various angles and the use of appropriate mathematical back projection algorithms [44,45], it is
183
K. Janssens
possible to perform three-dimensional elemental analysis in a non-destructive manner. Since this method involves rotation of the sample over 1808 or 3608 relative to the primary beam, it is limited to the investigation of relatively small objects, typically of dimensions of a few mm. Figure 4.34c shows the same arrangement as that of Fig. 4.34a and b, but in this case, instead of one optical element (PC1, to focus the primary beam), a second optic (PC2) is placed between sample and detector. The polycapillary half-lens PC1 focuses (quasi-)parallel radiation that enters the wide end of the device into a focal spot that is situated ca. 5 mm outside the narrow end of the lens. Between the lens end and the focal spot, the resulting X-ray beam is strongly convergent, while after focus, it is strongly divergent. Around the focal spot, the primary X-ray beam has a narrow “waist”, as schematically shown in Fig. 4.35. At a primary energy E0 of 21 keV, the diameter of the waist is ca. 10 mm. The narrow part of the waist extends in both directions for several hundreds of mm before significantly getting broader. Similarly, secondary radiation emitted by the sample atoms in the direction of the detector will pass through lens PC2 only when their point of origin is situated in the waist associated with PC2. Thus, when the two lenses are situated perpendicular to each other in one plane, in such a way that the intersection of the two waists form a cube-like volume, only fluorescent signals generated in this volume will reach the detector. Within the limits imposed by (a) the absorption of the radiation in the sample and (b) the free distance to the end of the lenses, it is possible to arbitrarily position this micro-volume within the material being investigated. This method is henceforth denoted confocal m-XRF. When combined with energy scanning, confocal m-XANES measurements can also be performed in this arrangement. The spatial resolution obtained with such an arrangement
Fig. 4.35. Detail of the intersecting waists of two polycapillary lenses aligned perpendicularly relative to each other in the same horizontal plane.
184
X-ray based methods of analysis
is dependent on the energy of the primary and the fluorescent radiations being employed and is situated in the 10 –40 mm range [46– 48]. Since the introduction of the second lens before the detector causes the detected net count rate to decrease by a factor 10 or more, instead of using the highly monochromatic form of excitation required for XAS studies, “pinkbeam” mode can also be employed. In this case, the Sik111l crystals of the monochromator are replaced by a pair of multilayer-coated reflectors. These reflectors select a relatively wide energy band ðDE=E < 1022 Þ out of the excitation spectrum (Fig. 4.18a), leading to a primary beam intensity that is ca. 30 – 50 times higher than in the case of monochromatic excitation. Combined m-SRXRF and m-SRXRD at the ESRF ID22 and ID18F beamlines. ID18F and ID22 are undulator beamlines of the European Synchrotron Radiation Facility (ESRF), Grenoble, France. Instead of a continuous energy distribution, only photons within specific energy bands (undulator harmonics) are produced (see Fig. 4.18a). In Fig. 4.34d, a schematic of the experimental arrangement at ESRF ID18F is shown [49]. After the monochromator, the beam is focussed onto the sample by means of a CRL; an additional pin-hole is used to reduce the scatter background in the sample area and to define the beam in the horizontal direction. In this manner, beam sizes as small as 0.5 £ 0.9 mm2 can be obtained. Also with other optics, such as pairs of curved mirrors, beam sizes of these submicroscopic dimensions can be produced. Next to energy-dispersive X-ray detector, at right angles to the primary beam, behind the sample, a wide angle XRD camera can be placed. This camera allows to record the diffracted beams generated in the sample up to a scattering angle 2u of ca. 408. By scanning a sample through the beam, simultaneously local composition information (from the generated XRF spectrum) as well as structural information (from the two-dimensional diffraction pattern) can be recorded. 4.3.3.6 Quantitative analysis With the exception of TXRF, all forms of XRF analysis suffer from matrix effects, i.e., the observed net X-ray intensity of a specific element is not only a function of the concentration of the analyte element itself but also depends on that of the other constituents. This effect is caused by the absorption of X-rays in the sample and by fluorescence enhancements when the radiation is emerging from the sample. In samples with nonideal homogeneity, the texture and size of the particles that constitute a porous sample may significantly influence the observed X-ray signals. Thus, in XRF very frequently curved rather than straight calibration
185
K. Janssens
curves are obtained; only in case of the analyses of trace constituents in a constant matrix, in which the absorption and enhancement effects are constant or of very thin foil samples, in which no such non-linear phenomena occur, is the quantification of XRF signals relatively straightforward. The same applies to the quantification of TXRF data, where the equivalent of “thin” samples is being analysed. Accordingly, a variety of correction and calibration models and procedures is in use in X-ray spectrometry based on (a) standard addition and dilution, (b) thin-film samples, (c) matrix dilution, (d) comparison to type standards, (e) internal standard elements and (f) mathematical corrections. A number of these procedures are destructive in nature (e.g., because they require dissolution of the material to be analysed) and therefore are not always applicable to analyses of cultural heritage materials. In these situations, the so-called “fundamental” approaches to the quantification problem in XRF are frequently employed. Here, knowledge on X-ray physics is employed to model all X-ray matter interactions in the sample in an attempt to correct for the matrix effects. Among the software packages that are available, a distinction can be made between those making use of (sometimes sophisticated) interpolation strategies between X-ray intensities derived from a large number of standard samples strongly resembling the unknown materials (e.g., a series of Cu-containing alloys when bronzes are analysed, a series of glass standards in the case of ceramics are studied) [50] or the software models that are based on a detailed knowledge of the XRF spectrometer response to different materials [51]. A more detailed treatment of quantification procedures is beyond the scope of this chapter. 4.3.4
XRD instrumentation
4.3.4.1 The Laue method In the Laue method, polychromatic radiation is used to irradiate a single crystal that is stationary. The patterns that are recorded by means of photographic film, image plates or a CCD camera can be regarded as a stereographic projection of the planes of the crystal. Either a transmission or a back-reflection geometry is employed. Since the crystal is fixed in space, the interplanar distances d and the Bragg angle u is fixed for every set of planes in the crystal. The only possible variations are the diffraction order n and the wavelength l. Thus, the diffracted beams form arrays of spots that lie on curves on the film. Each set of planes picks out and diffracts the particular wavelength from the white radiation that
186
X-ray based methods of analysis
satisfies the Bragg law for the values of d and u involved. Each curve therefore corresponds to a different wavelength. The spots lying on any one curve are reflections from planes belonging to one zone. Laue reflections from planes of the same zone all lie on the surface of an imaginary cone whose axis is the zone axis. In the back-reflection Laue arrangement (see Fig. 4.36a), the film is placed between the X-ray source and the crystal. The beams that are diffracted in a backward direction are recorded. One side of the cone of Laue reflections is defined by the transmitted beam. The film intersects the cone, with the diffraction spots generally lying on a hyperbola. In the transmission Laue method (Fig. 4.36b), the film is placed behind the crystal to record beams that are transmitted through the crystal. As before, one side of the cone of Laue reflections is defined by the transmitted beam. The film intersects the cone, with the diffraction spots generally lying on an ellipse.
Fig. 4.36. The Laue method in (a) back-reflection geometry and (b) transmission geometry.
187
K. Janssens
Crystal orientation is determined from the position of the spots. Each spot can be indexed, i.e., attributed to a particular plane, using special charts. Greninger charts are used for back-reflection patterns and Leonhardt charts for transmission patterns. The Laue technique can also be used to assess crystal perfection from the size and shape of the spots. If the crystal has been bent or twisted, the spots become distorted and smeared out. 4.3.4.2 Powder diffraction When a monochromatic beam of wavelength l is directed at a single crystal, then only one or two diffracted beams may result (Fig. 4.37a). When the crystal is now rotated around the impinging beam so that the
Fig. 4.37. The powder method. Irradiation of (a) a single crystal results in one or a few reflections and (b) a powdered sample results in cones of diffracted radiation which can be recorded on film (c).
188
X-ray based methods of analysis
angle between the diffracting planes and the primary beam remains the same, then the reflected beam will describe a cone with the crystal at its apex. The same effect can be obtained when instead of rotating a single crystal, the irradiated material consists of hundreds to thousands of very small crystals, all having a random orientation of their own. The diffracted beams are seen to lie on the surface of several cones. The cones may emerge in all directions, forwards and backwards, as illustrated in Fig. 4.37b. A circle of film can be used to record the diffraction pattern as shown in Fig. 4.37c. Each cone intersects the film, giving rise to arcs on the film. For every set of crystal planes, by chance, one or more crystals will be in the correct orientation to give the correct Bragg angle to satisfy Bragg’s equation. Every crystal plane is thus capable of diffraction. Each diffraction line is made up of a large number of small spots, each from a separate crystal. Each spot is so small as to give the appearance of a continuous line. If the crystal is not ground finely enough, the diffraction lines appear speckled. This is the basis of the Debye-Scherrer or powder method of XRD; it is probably the most commonly applied method in X-ray crystallography. The powder method is used to accurately determine the value of the ~ and ~c) of a ~; b lattice parameters (the magnitude of the unit cell vectors a crystal. In practice, the powder camera consists of a metal cylinder at the centre of which is placed the sample (see Fig. 4.38). Powdered material is typically glued to a glass rod with an amorphous type of adhesive. A strip of X-ray film is placed inside the cylinder with two holes punched into it: one for the
Fig. 4.38. Apparatus for Debye-Scherrer recordings: (left) side view, (right) front view.
189
K. Janssens
primary beam to enter the camera and a second for the beam to stop, diametrically opposed to the first. The camera enclosure is light-tight and placed in front of the X-ray beam. A typical Debye-Scherrer pattern obtained with this type of camera is shown in Fig. 4.37c. 4.3.4.3 The rotating crystal method In the rotating crystal method, a single crystal is mounted on an axis normal to a monochromatic X-ray beam. A cylindrical film is placed around it and the crystal is rotated about the chosen axis. As the crystal rotates, sets of lattice planes will at some point make the correct Bragg angle for the monochromatic incident beam, and at that point a diffracted beam will be formed. Since the reflected beams are located on the surface of imaginary cones, their intersections with the film (after flattening) are horizontal lines, as shown in Fig. 4.39a. The main use of the rotating crystal method is in the determination of unknown crystal structures. The precession, oscillation, and Weissenberg methods use variations of this theme. The Gandolfi camera. When confronted with phase identification problems in the field of art and archaeometry, an additional difficulty may be the fact that only very small samples (if any) are available for analysis. Sample may be tiny single crystals or agglomerates of a few single crystals. The Gandolfi XRD camera allows non-destructive analysis of single crystals smaller than 100 mm in diameter via the recording of a “powder pattern”. In this device, a single grain or crystal is doubly rotated in the X-ray beam in order to sequentially expose all available atomic planes to diffracting positions. A schematic of the Gandolfi camera is shown in Fig. 4.40.
Fig. 4.39. The crystal rotation method: (a) general arrangement, (b) detail of the crystal mounting head.
190
X-ray based methods of analysis
Fig. 4.40. Photograph (left) and schematic (right) of the Gandolfi camera. The diffracted beam hits the film at R. The angle x (which can be freely chosen in a 4-circle diffractometer) is fixed at 458 while the specimen rotates about the V and F axes.
It closely resembles a Debye-Scherrer camera. The sample is mounted on the tip of a needle that is mounted at 458 to the camera axis V. The needle rotates around its own axis (F-rotation) as well as around the camera axis. Requiring a sample barely visible to the naked eye, this technique is capable of identifying the mineral constituents of minute samples of paint used, e.g., by prehistoric artists or applied in very thin layers by easel painters. Recently, a Gandolfi camera equipped with an image plate detector has been developed [52]. See Fig. 12.2 for an example of an XRD pattern collected with a Gandolfi camera. Diffractometers. Of many of the XRD instruments described above that make use of X-ray film, more modern equivalents are available that either use imaging plates (i.e., laser readable and erasable X-ray-sensitive surfaces), X-ray-sensitive CCD cameras or Geiger/scintillator counters as X-ray detectors. The latter are normally mounted on computer-controlled and commercially available 4-circle diffractometers (see Fig. 4.41) that only require the crystal to be aligned in the X-ray beam; the instruments’ software will autonomously measure and refine unit-cell dimensions, record all reflections out to a maximum u-angle and store their intensities. Microscopic XRD. Chapter 6 discusses a number of case studies where a powder XRD spectrometer equipped with Goebel mirrors for microscopic beam focussing is used to characterize corrosion products of metallic artefacts. 4.3.4.4 Quantitative XRD analysis: Rietveld refinement A drawback of the conventional method of powder diffraction is that the XRD patterns of the constituent phases of a mixture strongly overlap, thereby preventing proper determination of the structure and /or its the quantitative analysis of the mixture. In the 1960s, Rietveld [53,54] developed a method to effectively separate the individual constituent of an unknown XRD pattern
191
K. Janssens
Fig. 4.41. Schematic layout of a 4-circle diffractometer.
from each other in a quantitatively reliable manner, thereby allowing an accurate determination of the structure. By fitting the experimental data to a (relatively simple) model, sample properties (such as the particle size distribution, the preferred orientation of grains, lattice constants, site occupancies, etc.) and a number of instrumental effects that influence the recorded diffraction patterns (such as the variation of the peak widths with diffraction angle, the type of X-ray tube, the slit settings, the goniometer radius, the polarization of the radiation, etc.) can be accounted for. In such circumstances, quantitative XRD analysis becomes equivalent to multiple linear regression of unknown XRD patterns (of mixtures of phases) against a library of patterns of single constituents. For reliable quantitative analysis of powders, however, it is essential that two conditions be fulfilled: (a) samples must be carefully prepared to comply to the definition of a powder: they must be homogeneous and a sufficient number of particles with random orientation must be present and (b) the structure factors of the constituents must be correctly calculated. In complex materials, as encountered in the cultural heritage area, especially condition (a) is not always fulfilled, which may limit the reliability of the analysis results. In general, however, the Rietveld refinement approach has been so successful that nowadays the structure of materials, in the form of powders, is routinely being determined with a comparable accuracy as obtained by means of single crystal diffraction techniques [55]. Rietveld refinement is widely used for quantitative phase analysis, resulting yearly in more than a thousand scientific papers applying it [56].
192
X-ray based methods of analysis
4.3.5
XAS instrumentation at SR beamlines
As already explained above, for most XAS experiments, highly monochromatic synchrotron radiation is required. The degree of monochromaticity is usually expressed as a relative energy width DE/E, which typically is of the order of 1024 (i.e., DE ¼ 1 eV at E ¼ 10 keV) in case the Si(111) reflection is used for monochromatization. It is the most basic form, a double crystal monochromator consisting of two parallel crystals mounted on a common rotation table so that the diffraction angle u can be continuously varied (see Fig. 4.34b and c). The first crystal selects the wavelength (and its higher order overtones) while the diffraction at the second crystal ensures that the monochromatic beam regains a horizontal direction. In between the sample and the monochromator, X-ray focusing optics of various types can be placed in order to concentrate the X-ray beam in the horizontal and/or vertical direction. If the beam is focused to dimensions below ca. 100 mm, the use of a fixed-exit monochromatic is recommended, to ensure that the focused beam spot does not change in size and position during energy scans. The experimental setup shown in Fig. 4.34b allows to simultaneously record XAS data in two manners: Absorption XAS. The most straightforward mode of operation is to directly record the intensity of a monochromatic beam before and after (a homogeneous) sample as a function of energy E by means of two ionization chambers (see IONI1 and IONI2 in Fig. 4.34b and c) and calculate the m(E) vs. E curve: IðEÞ ln ¼ 2mðEÞ·rd ð4:15Þ I0 ðEÞ where the sample areal density rd is a scaling factor that usually is normalized out. An ionization chamber essentially consists of two isolated parallel metal plates between which a voltage (several 100 V) is applied. The electrical current flowing between the two condensator plates is proportional to the flux I of the X-ray beam that passes in between. XAS data with sufficiently signal-to-noise ratio can be obtained when (a) the absorption in the sample is not too large so that I(E) and I0(E) do not differ too much and (b) the concentration of the absorbing element is high enough (.0.1% w/w) so that a significant difference between I(E) and I0(E) can be recorded. Fluorescent XAS. When the sample under investigation is so strongly absorbing (or so thick) that the transmitted intensity is no longer measurable or when XAS data from a minor/trace constituent are to be recorded, instead
193
K. Janssens
of directly measuring the absorption, one of the signals produced as a result of photoelectric absorption can be recorded as a function of the primary energy. By means of the energy-dispersive detector, positioned at 908 to the primary beam (see Fig. 4.34b and c), the fluorescence intensity profile Ifl(E) vs. E can be recorded. When the absorbing element is not too strongly concentrated (,0.1 –0.5%), the normalized fluorescent intensity Ifl(E)/I0(E) is proportional to m(E), provided appropriate dead-time corrections, etc., are taken into account. An alternative to detection of the fluorescent intensity is to record the total electron yield (TEY), i.e., to measure the number of (photo)electrons emitted by the sample during the irradiation as a function of the primary beam energy E. The TEY mode of detection is especially useful at low primary energies and when low atomic number elements are studied. 4.4
A SURVEY OF APPLICATIONS OF X-RAY METHODS IN THE CULTURAL HERITAGE SECTOR
Synchrotron (m-)XRF, offering (sub)ppm level detectability for many elements [57] on the basis of the irradiation of minute samples, can be used for detailed and quantitative finger-print analysis of materials in order to gain a better understanding on their provenance. The use of a microscopic X-ray beam of monochromatic energy also permits the detailed investigation of the processes that have altered the surface composition of a material, through the use of XRF, XRD or XAS. Though suitable for trace-level micro-analysis of organic materials (e.g., paper, pigments dispersed in a organic binder) or of silicate-rich materials such as pottery or glass, the high primary intensity of synchrotron microbeams usually is not compatible with the XRF analysis of metallic materials (e.g., artefacts made in bronze, iron, silver, gold or alloys of these metals). Laboratory m-XRF can be conveniently used for this purpose. Another strong point of laboratory m-XRF is the possibility to perform local (quantitative) analysis on objects whose size, shape or nature is incompatible with the vacuum and the small sample enclosures employed by most conventional micro-analytical techniques such as EPXMA and m-PIXE. In this respect, m-XRF offers similar possibilities to external-beam PIXE [2] (see also Chapter 5), but in some cases with a better lateral resolution. Even for smaller objects (such as, e.g., coins), which might be analysed as a whole in conventional XRF apparatus, the use of a small beam instrument offers advantages. A small X-ray beam permits to analyse an object at various locations, e.g., to verify that all parts of a statue are made of the same material, or to investigate the homogeneity of the material used. XRF on
194
X-ray based methods of analysis
curved or otherwise non-flat surfaces can lead to errors in the quantification (especially for metallic materials) [58]; with a small beam, it is in general easier to select locations on an object that more closely resemble the ideal, flat and polished surface normally required for reproducible quantitative measurements. In case corroded objects are under investigation, only a small area of the (altered) surface needs to be removed in order to expose the underlying original material. In what follows, after briefly describing the instrumentation employed for the investigations, a limited number of case studies highlighting the aboveoutlined types of investigations are outlined. Depending on the nature of the investigation and the type of information required, one of the various forms of X-ray analysis described above may be employed. In several books [59 – 61] and special journal issues [62], several other collections of similar case studies can be found. 4.4.1
Compositional analysis of historic Glass
The composition of glass, produced from prehistoric to the modern ages, has been subject to a number of distinct changes, depending on the raw materials available and the state of glass-making technology and know-how. Thus, chemical analysis of excavated glass fragments can be very useful for gaining information on glass-making technology and/or glass trade in different periods. X-ray-based methods have been used extensively for analysing this type of material [63 – 68], next to other (trace-level) methods of (micro)analysis [69 – 71], as detailed in Chapter 15. 4.4.2
Pigments
Essentially two X-ray-based approaches to pigment analysis in painted works of art (oil paintings, frescoes, miniatures and illuminated manuscripts) can be distinguished: (a) the in situ approach where portable equipment is employed to irradiate a small part of a large artefact and (b) the ex situ approach where the pigmented layer(s) to be examined are sampled with minimal damage to the work of art and the resulting minute fragments or particles are examined by means of one or more techniques in specialized laboratories. For the detailed characterization of paint layer stratigraphies, traditionally small fragments are removed from oil paintings by means of a hollow needle or a fine scalpel. However, as outlined below, such a destructive procedure is not always necessary.
195
K. Janssens
The pallet of inorganic pigments contains a wide range of different types with well-known chemical compositions (see Table 4.6). Inorganic pigments have been used for all colours of the artists’ pallet because of their high colouring power and their stability against changes in temperature, climate and light. Each inorganic pigment is characterized by its colour and by one or more constituting elements; when these key-elements are observed in the correct proportions, the pigment can be identified. Furthermore impurities can give information on the provenance and/or manner of manufacture of specific colouring materials. Since the chronological use of most pigments is known, it is possible in some cases to determine an approximate date for the genesis of painted historical objects. Due to the circumstance that some pigment appeared on artists’ pallets after a certain time (post quem) or another pigment disappeared before that time (ante quem), it is possible to distinguish between original materials, restored parts or fakes [20,72]. However, a number of problems render pigment identification by means of XRF or X-ray emission analysis alone difficult. A first difficulty derives from the fact that, as is apparent from Table 4.6, many pigments share key elements (e.g., the presence of the element Sb may point to the presence of antimony white or to that of Naples Yellow, that of Cu to many green or blue pigments) so that the observation of a specific element in the XRF spectrum cannot always be used as an unambiguous indication of the presence of a specific pigment. Especially in the case of Cu-containing greens and Fe-containing yellow-browns, the data provided by XRF or emission analysis are rather unspecific. A second problem has to do with the fact that during X-ray irradiation of artefacts that have been covered with more than one pigmented layer (e.g., varnish, ground and support layers next to one or more actual paint layers), the constituents of many (if not all) of these layers will contribute to the observed XRF spectrum. Strong X-ray lines of, e.g., a lead-white base layer may then obscure much weaker X-ray lines of the top-most layers while generally the interpretation becomes more cumbersome. Similarly, when pigments are mixed to obtain a particular colour, the identification of the individual components can be rendered more difficult. In quite a few cases, however, the identification is unproblematic and the XRF method can even be used to determine the mixing ratio(s) of the original pigments. In other cases, specific measures need to be taken to either (a) increase the specificity of the analysis or (b) limit the depth in the paint layer stack from which the signals in the XRF spectrum originate. The use of compact portable XRF equipment for pigment identification has been described predominantly for the analysis of frescoes and ceramics [25,73]. With these works of art, usually a single pigmented layer covers a
196
X-ray based methods of analysis TABLE 4.6 Inorganic pigments and the key elements that can be used for their identification based on X-ray techniques (adapted from Ref. [76]) Colour
Pigment name
Chemical composition
Key element(s)
White
Antimony white Lithophone Permanent white Titanium white White lead Zinc white Zirconium oxide Chalk Gypsum Auripigmentum Cadmium yellow Chrome yellow Cobalt yellow Lead-tin yellow Massicot Naples yellow Strontium yellow Titanium yellow Yellow ochre Zinc yellow Cadmium red Cadmium vermilion Chrome red Molybdate red Realgar Red lead Red ochre Vermilion Basic copper sulphate Chromium oxide Chrysocolla Cobalt green Emerald green Guignent green Malachite Verdigris
Sb2O3 ZnO þ BaSO4 BaSO4 TiO2 2PbCO3·Pb(OH)2 ZnO ZrO2 CaCO3 CaSO4·2H2O As2S3 CdS 2PbSO4·PbCrO4 K3[Co(NO2)6]·1.5H2O Pb2SnO4/PbSn12xSixO7 PbO Pb(SbO3)2/Pb3(SbO4)2 SrCrO4 NiO·Sb2O3·20TiO2 Fe2O3·n H2O (20–70%) K2O·4ZnO·4CrO3·3H2O CdS þ CdSe CdS þ HgS PbO·PbCrO4 7PbCrO4·2PbSO4·PbMoO4 As2S3 Pb3O4 Fe2O3 (up to 90%) HgS Cux(SO4)y(OH)z Cr2O3 CuSiO3·n H2O CoO·5ZnO Cu(CH3COO)2·3Cu(AsO2)2 Cr2O3·n H2O þ H3BO3 CuCO3·Cu(OH)2 Cu(CH3COO)2·nCu(OH)2
Sb Zn, Ba Ba Ti Pb Zn Zr Ca Ca As Cd Cr K, Co Sn Pb Pb, Sb Sr, Cr Ni, Sb, Ti Fe Zn, Cr Cd, Se Cd, Hg Pb, Cr Pb, Cr, Mo As Pb Fe S, Hg Cu Cr Cu Co, Zn Cu, As Cr Cu Cu
Yellow
Red
Green
continued
197
K. Janssens
TABLE 4.6 (continuation)
Colour
Pigment name
Chemical composition
Key element(s)
Blue
Azurite Cerulean blue Cobalt blue Cobalt violet Egyptian blue Manganese blue Prussian blue Smalt Ultramarine Antimony black Black iron oxide Carbon/charcoal black Cobalt black Ivory black Manganese oxide
2CuCO3·Cu(OH)2 CoO·SnO2 CoO·Al2O3 Co3(PO4)2 CaO·CuO·4SiO2 BaSO4·Ba3(MnO4)2 Fe4[Fe(CN)6]3 Co-glass (K2O,SiO2,CoO) Na8 – 10Al6Si6O24S2 – 4 Sb2O3 FeO·Fe2O3 C (95%) CoO C þ Ca3(PO4)2 MnO þ Mn2O3
Cu Co, Sn Co, Al Co Ca, Cu, Si Ba, Mn Fe Si, K, Co Si, Al, Na, S Sb Fe – (K) Co P, Ca Mn
Black
substrate material so that most of the observed XRF signals can straightforwardly be attributed to constituents of the coloured layer. In some cases, however, also with frescoes, complicated stratigraphies can be encountered. For example, during the study of the frescoes of Giotto in the Chapel of the Scrovegni in Padova, Italy [74], it could be established by means of portable XRF and without any sampling that the golden haloes in these frescoes are composed of different layers: (a) a superficial layer of calcium sulphate, due to pollution, (b) a (on average) 1.6 mm thick layer of gold leaf, (c) a Cu-containing glue layer, having an equivalent Cu thickness of ca. 1 mm, (d) a layer of white lead, having an equivalent Pb thickness of around 5 mm, (e) possibly a layer of azurite (Cu) and finally (f) the Fe- and Srcontaining plaster substrate. Bronk et al. [28] have used the ARTax spectrometer for studying the evolution of the use of pigments in enamelled objects of the 16th– 19th century where, e.g., the occurrence of Ni, As and Bi impurities in Co-blue enamels is indicative of a 16 – 17th dating of the objects while the absence of the former elements is observed in Co-blue enamel layers of objects (re)produced in the 19th century. In Chapter 14, the use of TXRF for analysis of minuscule amounts of pigments, removed from illuminated manuscripts by a virtually nondestructive sampling method, is discussed in greater detail. By means of
198
X-ray based methods of analysis
this method, pigments from the top-most pigmented layer of (oil) paintings can be analysed, provided the varnish layer is removed (e.g., during restoration activities). Analysis of paint multilayers by means of the confocal variant of m-XRF is a relatively new possibility which shows a lot of promise for application in the cultural heritage sector [48]; preliminary experiments on car paint multilayers, consisting of up to 9 different coatings showed that each layer could be non-destructively analysed separately (see Fig. 4.42). This method was also
Fig. 4.42. Left panel: car paint multilayer stack, consisting of a series of nine (possibly 10) differently pigmented or transparent layers with thicknesses in the range 20 – 100 mm; right panel: X-ray intensity depth profiles of various elements. The sample was analysed with the blue layer on top. Layer 1 (blue) can be observed to contain Ti, some Co, Cu and Pb; layer 2 (red) contains Cr, Fe, Cu, Zn and some Sr; layer 3 (black) is not associated with any strong XRF intensity; layer 4 (orange) contains Ti, Fe, Co and Zn, but less Fe and more Zn than layer 2; layers 5–7 (black) also do not show strong XRF signals, except layer 6 that is associated with Sr and some Co; finally layer 8 (orange) shows a similar pattern as layer 4. Some of the signals originating from this deep layer (ca. 350 mm deeper than the surface) are strongly absorbed by the layers on top of it. Layer 9 is a transparent varnish layer. ( For a colored version of this figure, see Plate 4.I.)
199
K. Janssens
demonstrated to be applicable to quantitative analysis of complex paint layers stratigraphies encountered in 16th century oil paintings [75]. 4.4.3
Lustre ware
Pottery from the Middle Ages and Renaissance Era often is decorated with glazes showing gold and copper-coloured metallic reflections and iridescence. These effects are called lustre and consists of a metal deposition on a tinopacified lead glaze that produces the brilliant metallic reflections of different colour and iridescence [77]. The technique of lustre was developed by the Islamic culture in Mesopotamia, during the 9th century AD . Since muslims were not allowed to use gold in artistic representations, a way was found to create the same effect without using real gold. It arrived in Spain during medieval times, following the expansion of Arabian culture. From there it was introduced into the centre of Italy, where it was exploited to produce polychrome lustre Renaissance pottery. Thus, lustre became one of the most important decorative techniques of the Medieval and Renaissance pottery of the Mediterranean basin. Centuries later, the lustre is still visible thanks to the high quality of the film and its resistance to atmospheric oxidation and burial weathering. The specific optical properties of the films are the consequence of the presence of silver and copper nanoparticles, dispersed homogeneously in the glassy matrix of the ceramic glaze. To create these nanoparticles, artisans applied a mixture of copper and silver salts and oxides, together with vinegar, ochre and clay, on the surface of previously glazed pottery. The object was then placed in a kiln for heating to about 6008C in a reducing atmosphere. The high temperature caused the glaze to soften so that copper and silver ions could migrate into the outer layers of the glaze. In the reducing atmosphere, the ions reduced to metals, which upon aggregation formed nanoparticles that gave rise to the desired colour and optical properties. Different recipes were used to obtain different lustre colours, ranging from goldlike to copperlike. A red originates from the migration of copper ions and their subsequent reduction. In the case of gold shades, in principle only silver is needed, but craftsmen used both copper and silver. Researchers are still trying to find out why they used both materials and what procedure was employed, since copper needs higher temperatures than silver in order to stimulate the formation of nanoclusters. The lustre technique shows that craftsmen had an advanced technological and empirical knowledge of material science.
200
X-ray based methods of analysis
Several techniques have been used to characterize the chemical and physical properties of these films. Detailed compositional information can be obtained by Rutherford Backscattering Spectrometry (RBS, see Chapter 5), while the optical properties can be investigated by optical absorption in the visible – ultraviolet region. By electron microscopy (TEM and SEM) it was possible to visualize the metallic particles present in the glaze; silver and copper nanocrystals can be well separated in TEM pictures. They are quasi-spherical, with a diameter between 5 and 100 nm. Silver nanocrystals, larger than copper ones, appear grouped together among copper crystals and close to the glaze surface. From XRD patterns, the crystalline phase of the particles could be derived. In order to determine the valence state of Ag and Cu and to describe the local atomic environment around the metallic species, XAS measurements were performed. These permitted to study the amorphous oxide phase in which the metals are present. Lustre samples of glazed Renaissance pottery (end of XV– XVI century) from Deruta and Gubbio, Italy were analysed [78]. Examples of lustre pottery from these localities are on display in many important international museums, such as the Louvre Museum (Paris, France) and the Metropolitan Museum of Art, New York, USA. The lustre on the examined artefacts consists of a heterogeneous metal – glass composite film, some hundreds of nanometres thick, analogous to that present in the modern metal – glass nanostructured composites synthesized for high-technology applications [79,80]. It was found that in the case of red lustre, the colour was mainly due to the presence of copper nanoclusters, while in the case of gold lustre the colour was caused by nanoclusters of silver. In the gold lustre, Agþ, Cuþ and Cu2þ ions are present, while in the red lustre, according to the historical recipes [81,82], the dominant ion is Cuþ. The nanoclusters are confined to the more external glaze while the oxidized forms are present at larger depths. Figure 4.43 summarizes some of the XANES results obtained at the Cu-K edge (8979 eV). When XANES data are recorded in the fluorescence mode of detection, where the lustre layer is probed down to a depth of 50 mm, Cu appears to be predominantly present under an oxidized form (i.e., as Cuþ in Cu2O, cuprite). However, if the TEY mode of detection is used, where only the top 100 nm of the same lustre layer is analysed, a different profile is obtained, showing the presence of Cu in the metallic state. The analysis of the EXAFS spectra obtained for three red lustre samples, performed in fluorescence mode, suggests the presence of copper coordinated ˚ , which is a distance very close to the by oxygen atoms at a distance of 1.86 A
201
K. Janssens
Fig. 4.43. XANES spectra obtained in fluorescence (fluo) and total electron yield (TEY) mode of the red lustre, compared to the XANES profiles of Cu2O and metallic Cu.
˚ ). The Cu – O first shell coordination distance for the first shell of Cu2O (1.85 A also shows a number of neighbours not very far from 2, the value shown by the Cu2O standard. In general, one can state that the chemical state and local environment of the Cu ions are similar to those found in copper– alkali ion-exchanged silicate glass samples. This finding strongly supports the view that lustre formation is mediated by a copper- and silver–alkali ion exchange as a first step, followed by nucleation and growth of metal nanocrystals as a spontaneous process due to the supersaturation of the elemental concentrations in the region where clustering occurs. 4.4.4
Metallic artefacts
COPRA, short for “Compact Ro¨ntgen Analyser”, is a fully equipped microXRF instrument intended for easy transportation to musea and galleries. Bichlmeier et al. [83] have used this instrument for the analysis of five golden medallions and two Roman coins (dated 550 – 575 AD ) that were all fitted with small attachment rings (Fig. 4.44). The questions regarding these artefacts were the elemental composition of the rings and medallions and whether different or identical gold alloys could be identified. The average compositions of the artefacts’ rear sides and rings were obtained from three measurements on different spots (Table 4.7). The standard deviations of the Cu, Ag and Au concentrations were 2 –8%. This relatively small uncertainty
202
X-ray based methods of analysis
Fig. 4.44. Series of seven late-Roman gold medallions, fitted with attachment rings.
allowed the grouping of identical artefact and ring materials according to their compositions. For example, the prominent high silver content of medallions M4 and M5 leads to the conclusion that both belong to one group; also, their composition of gold and copper match well. The composition of the bulk materials can be combined easily with the shape of the objects: Both Roman coins (Fig. 4.44, Nos. 3 and 6), the two smaller medallions with a convex projection (Fig. 4.44, Nos. 4 and 5) and the three richly ornamented, large medallions (Fig. 4.44, Nos. 1, 2 and 7) consist of the same material. Also, the attached rings of the larger medallions and the Roman coins show identical compositions. The standard deviations of the copper, silver and gold TABLE 4.7 Results of the micro-XRF analyses (wt%), average data for three measurements (40 kV, 0.8 mA) % w/w
M1 M2 M3 M4 M5 M6 M7
Back side
Ring
Fe
Cu
Ag
Au
Fe
Cu
Ag
Au
1.0 0.3 1.5 0.3 0.3 0.2 1.4
0.5 0.6 0.4 1.8 1.7 0.5 0.6
8.6 8.7 1.2 18.7 18.4 1.4 8.1
90.1 90.3 96.9 79.4 79.6 98.1 90.3
1.5 1.0 0.9 0.3 0.2 1.4 8.5
1.3 1.4 2.4 1.8 2.0 1.5 3.0
8.3 6.7 8.6 15.2 15.0 8.3 8.8
88.8 91.1 88.0 82.9 83.0 88.2 79.9
203
K. Janssens
concentrations were fairly small (2– 8%), while the standard deviation of the iron concentration was about 40%. A possible explanation is that iron particles may have been introduced mechanically into the artefacts (e.g. by hammering) and therefore are distributed inhomogeneously. For this reason, the iron concentration was not considered further. For the gold artefacts it was assumed that the rings of medallions 1 and 2 were attached in the manufacturing process, whereas medallion 7, which shows distinct signs of mending, was repaired subsequently. The mounting of rings on the Roman coins took place in a second step of production, since the material of the bulk and of the eye do not have the same composition. This implies that the coins originally used as currency were transformed into pendants and not manufactured in one piece. There were no apparent distinctions between the rings and medallions M4 and M5, so that it was not possible to determine if the rings were attached originally or subsequently. It was possible to distinguish different ring materials: those of medals M4 and M5 show the same composition, and also M1, M2, M3 and M6. Owing to this grouping it can be assumed that medallions M1 and M2 were made from the same material and therefore in the same workshop where coins M3 and M6 were modified. In Chapter 13, the analysis of coins for elucidating their provenance is discussed in greater detail. The above-described studies concerned investigations of materials that could either be sampled or the complete object could be easily brought into the laboratory. Sometimes, however, the analytical instrument must be brought to the objects. During an exhibition of ancient gold artefacts from the Iranian National Museum in the Kunsthistorisches Museum (Vienna, Austria), a number of these valuable items were analysed on site by means of the COPRA instrument. The complete setup of the instrument took ca. 45 min. Five objects were analysed in various places with the aim of determining the type of gold alloy they were manufactured from and in order to document possible difference in origin of these objects. The spectral data ¨ gussa, Austria). were calibrated against a series of Au/Ag/Cu standards (O The average compositions are summarized in Table 4.8. Quasi-straight line calibration curves between observed X-ray intensities and the concentration of the above-mentioned elements were established. Since it concerns a metallic matrix, the obtained concentration data was normalized to 100%. Three of the objects are made in fine gold (pure Au with some Cu present) while two others are made in different Au/Ag/Cu alloys. The variability of the composition within most objects was small. In Fig. 4.45, a photograph taken during the irradiation of one of the artefacts (a griffin) is shown. In this case, the object is positioned before the X-ray beam by means of a manual
204
X-ray based methods of analysis TABLE 4.8 Average composition derived from in situ irradiation of five Iranian Au artefacts Object
Number of analysis points
Au (%)
Ag (%)
Cu (%)
Other lines (not quant.)
Lion Griffin Bracelet Beaker Seal
18 8 9 3 5
98.8 99.0 99.2 64.8 83.6
0.0 0.0 0.0 31.7 12.4
1.2 1.0 0.8 3.4 4.0
Fe Fe K, Ca, Fe Fe Fe
laboratory stage. When combined with a motorized stage, the control software of the instrument allows for coordinated sample stage movement and spectrum collection, so that, e.g., element distributions along a line on the specimen surface can be recorded. The COPRA instrument is currently installed at the Academy of Fine Arts, Vienna, Austria. Other types of cultural heritage artefacts that were investigated with it include Medieval brass and bronze burial artefacts (Cu, Zn, Sn), early 20th century decorated glass objects (Mn, Fe, Co, Ni, Zn, Sn), daguerreotypes (Cu, Au, Ag), differently pigmented areas of 18th century illuminated parchments (Ca, Mn, Fe, Cu, Au, Pb) and various manuscripts from the 13th to the 19th centuries of the Austrian State Archives prepared with ferro-gallic ink, next to modern car paints [31], multi-coloured beads and industrial materials [31].
Fig. 4.45. Use of the COPRA instrument for in situ (i.e., in the museum gallery) analysis of a gold artefact.
205
K. Janssens
For the chemical characterization of corrosion compounds analysis on ancient bronze objects, De Ryck et al. [84] have compared different microbeam techniques. They include optical microscopy, SEM – EDX, secondary ion microscopy, synchrotron-based FTIR, synchrotron-based XRD and XANES. The objective was to investigate which combination of analysis methods is most suitable for this type of investigation, taking into account aspects such as limited sampling and the ability of obtaining spatial information. These authors state that XRD in combination with optical microscopy and SEM-EDX is able to provide a complete description of the layered structure both on the elemental and the molecular levels. Chapter 10 discusses in some detail the use of laboratory microscopic XRD for characterization of corroded Ci-alloy base artefacts. 4.4.5
Analysis of graphic documents
Historical objects such as documents, illuminated manuscripts and coloured prints are not only valuable objects of cultural heritage but also documents of human history. The knowledge about historical materials is very important to answer archaeometrical questions and to develop and/or refine restoration and conservation concepts. Historic inks contain, next to the constituents that are responsible for its colour (e.g., carbon/soot in bistre-type inks, Fe in ferro-gallic inks,…), impurities of various nature. Together, these non-essential components of the inks can form a characteristic fingerprint that can provide information on the manufacturing technology of the ink or that can be used to investigate additions, alternations and or falsifications made in different inks to historical documents. Haller and Kno¨chel [36] and Mommsen [85] compared the concentrations of Cu and Pb in ink of the Gutenberg Bible to that of other early single leaf copies and books. The pioneering work with PIXE by Cahill et al. [86,87] identified the (X-ray based) analysis of ink as a new research tool for the history of early printing (from the second half of the 15th century) [88,89]. It is assumed that each printer/printing-office can be recognized by a specific ink-preparation as reflected in the trace element signature of the dried ink. The recipes for the ink were kept as a secret; in the 15th century, not a single specification of composition is documented [90]. Mommsen et al. [85] employed a 0.5 £ 1 mm2 polychromatic X-ray beam derived from the synchrotron storage ring ELSA (Bonn, Germany) for irradiating single leaves of early 15th century printed paper. Ink and paper of 22 different works from different locations in Germany, Italy and
206
X-ray based methods of analysis
Switzerland were analysed. The energy deposited in the paper during the measurements (300 s) was estimated to be ca. 15 m W/cm2, i.e., about a factor 70 lower than bright sunlight. Comparison of the spectra from paper and paper/ink combinations from page II,316 of the single leaf 42-line Gutenberg bible revealed that Ni, Cu and Pb are present in the ink at concentration levels of a few 10 mg/cm2 while K, Ca, Ti, Mn, Fe and Zn originate from the paper base. In six of the 22 leaves, printed areas could not be distinguished from blank areas, suggesting that the corresponding inks were prepared from mainly C-bearing material such as lamp black or soot [90]. In some of the other inks, in addition to Ni, Cu and Pb, also K, Ca and Fe were present. On all analysed pages except the Gutenberg leaf, the same ink composition on the recto and verso side was found. On the recto side of the Gutenberg B-42 leaf, the ink thickness was ca. three times larger than on the verso side, although both sides show the same Cu/Pb and Ni/Pb ratios (respectively, 1.0 ^ 0.5 and 0.007 ^ 0.003). For the same page (II, 316) of the Harvard Gutenberg bible, by means of PIXE recto and verso Cu/Pb ratio’s of 1.15 ^ 0.05 and 1.44 ^ 0.07 were reported [85– 88]. For three specific printers, the constancy of their ink composition over large time intervals was tested by analysing leaves of several books produced by them. They appear to have changed the composition of their inks fairly frequently. It therefore appears difficult to establish a definite trace element pattern specific for one printer. When only the trace elements which occur in the ink alone (and not in the paper) are considered, only Pb and Cu concentrations can be used for distinguishing between early printers. As a result of this study, Mommsen et al. [85] concluded that a systematic investigation of the ink composition in works printed 10 – 15 years after Gutenberg’s first bible edition is needed to learn more about the early recipes for ink preparation and find specific reasons for the presence of metallic impurities in the ink. The synchrotron-based XRF method was found to be suitable for non-destructive measurements on this fragile type of material and appropriate for performing the large number of measurements required to reveal systematic trends in the composition of paper and inking. Next to synchrotron-based XRF, laboratory m-XRF is also an excellent technique for the analysis of valuable documents, e.g., to determine their authenticity. Typically, different inks, while having the same visible appearance, will have different chemical compositions. Larsson [91] and Stocklassa and Nillson [92] described analysis by m-XRF of a 500-year-old Swedish possession letter (dated April l, 1499). The document, a sales contract for an estate, showed signs of alterations. It was suspected that the alteration
207
K. Janssens
was made in the 1530s, when the Swedish king restituted land back to certain nobles, who previously had been stripped of their estates by an earlier ruler. Although alteration was suspected by visual inspection (the original name of the owner had been removed by scraping), the original text was unreadable. By employing m-XRF generated Ca and Zn maps (a trace constituent of the original ink), the original name could be established, however. In the Ca map, the (falsified) visible text was visible, featuring the family name “Ga¨smestad i Bo¨re”; in the Zn map of the same area, however, a completely different text reading “Bøtinge i Asbo” could be read. Obviously, the forger used different inks for the alteration, accounting for the change in chemical makeup. In Fig. 4.46, the elemental ratios Mn/Fe, Cu/Fe and Zn/Fe obtained by means of m-XRF investigations of Goethes manuscripts of Faust I and Faust II are intercompared. Folios GSA25/XVII,2,12; GSA25,XVII,1,2; (Faust I, original and corrections) and GSA25/XVIII,5,7 (Faust II) were examined by means of the ARTax spectrometer at the Goethe und Schiller-Archiv (Weimar, Germany), yielding fingerprint spectra of the inks. It can be observed that the two inks employed by Goethe to write Faust I have a similar composition. The ink used for the corrections to Faust I are similar to the ink employed for writing Faust II, indicating that Goethe did not modify the first part until working on the second part (Fig. 4.47). These and other valuable manuscripts (e.g., music scores by Bach and Mozart) also have
Fig. 4.46. Analysis results of J.W. von Goethes manuscripts of Faust I and II by means of the ARTax m-XRF spectrometer. The two inks used in Faust I (squares) show a similar composition; the ink used for corrections to Faust I (triangles) is similar to that used for Faust II (circles).
208
X-ray based methods of analysis
Fig. 4.47. Mn K-edge XANES spectra of: (a) fossilized ivory; (b) 6008C heated fossilized ivory; (c, d) turquoise-blue collection odontolites and (e) synthetic apatite Ba5(PO4)2.5(MnO4)0.5Cl taken as reference mineral with Mn5þ in tetrahedral coordination. Inset: Pre-edge structure of tetrahedral Mn5þ in (b) heated fossilized ivory; (c, d) odontolites and (e) synthetic Ba5(PO4)2.5(MnO4)0.5Cl are observed at 0.9 and 1.9 eV below those of tetrahedral Mn6þ in BaMnO4 and Mn7þ in KMnO4, respectively.
been examined in the context of oxidative corrosion inflicted to the paper by ferro-gallic inks [93] (see also Chapter 17). The ARTax spectrometer was also used to investigate the pigments used for colouration of two similar copper engravings of Albrecht Du¨rer (Kupferstichkabinett, Berlin, Germany) entitled “Petrus und Johannes heilen einem Lahmen” (Peter and John healing a cripple). Both 16th century prints were coloured at a later time. m-XRF analysis revealed that one of the prints (KK Berlin Inv.-Nr. A137) was coloured with pigments such as azurite (blue), malachite (green), white lead, cinnabar (red), ochre (yellow-brown), red lead, gold water colour and crayon, consistent with a colouration made in
209
K. Janssens
the 16th – 17th century. In the second print (KK Berlin Inv.-Nr. A551), however, zinc oxide red and chromium green pigments were encountered, pointing to a colouration in the 19th century [94]. Silver point drawings belong to the most precious and rarest treasures of print collections. They were essentially created during the Renaissance and are characterized by extremely thin grey-brownish strokes on paper that had been coated before use. So far, only little chemical information on these drawing materials could be obtained because the drawings are very delicate, and, therefore, analyses are very difficult to perform without any damage. A chemical fingerprint can be obtained by the determination of ratios of selected elements, reflecting not only the origin of the used materials but also the processing and the storage conditions. In order to precisely characterize the chemical composition of drawings by Albrecht Du¨rer (1471 –1528) and Jan van Eyck (ca. 1395 – 1441), Reiche et al. [95,96] employed synchrotron radiation-induced m-XRF. 4.4.6
Mn oxidation in odontolites
In the Middle Ages, Cistercian monks created odontolite, a turquoise-blue gemstone, by heating fossilized mastodon ivory found in 13 – 16 million-yearold Miocene geological layers next to the Pyrenees. They thought they had produced the mineral turquoise because of the resemblance of odontolite with this semi-precious stone. Odontolite was used for the decoration of medieval art objects such as the reliquary bronze crosses. Fossilized ivory and its mysterious colour change upon heating have been investigated by several naturalists and gemmologists, among them Re´aumur (1683 – 1757). Reiche et al. [97] investigated odontolite decorations on a 13th century cross, made in a Limoges workshop. Although vivianite, a blue-coloured iron phosphate or copper salts were proposed to be the colouring phases, none of these minerals were found in the odontolite material. Rather, it consisted of fluorapatite, Ca5(PO4)3F, containing trace amounts of iron, manganese, barium, lead, rare earth elements and uranium and presenting crystallites of 100 – 500 nm in size. The crystal size was about 10 times larger than that of unheated fossilized ivory and suggested that odontolite was heated at about 6008C [98]. As potential colouring agents manganese and iron were then studied. Luminescence and optical spectroscopy permitted the exclusion of iron as a colouring ion and suggested that manganese ions could be responsible for the colouration of odontolite. XAS was employed to follow changes in the local environment and the valence state of manganese on heating. XANES and EXAFS spectra were recorded at the K-edge of
210
X-ray based methods of analysis
manganese impurities (200 – 650 ppm) in fluorapatite, which is a strongly absorbing matrix at this energy (6.5 keV). An energy resolution of 0.4 eV, realized by using the a Si(220) monochromator of ESRF Beamline ID26, was used to measure the position and the intensity of the pre-edge structure of manganese (see Fig. 4.47). They indicate most clearly the changes in the structural environment and oxidation state of manganese. Unheated (white) fossil ivory showed Mn2þ ions in octahedral coordination while in the (turquoise-blue) fossilized ivory heated at 6008C and in the two odontolite samples, the major part of the manganese was found to be in the 5þ oxidation state as indicated by the pre-edge structure observed at 6541.3 eV. In addition, a comparison of the XANES spectra of heated fossilized ivory, odontolites and a reference synthetic Mn-chlorapatite indicated that the Mn5þ substitutes for P5þ at the tetrahedral sites; in such a coordination Mn5þ ions give rise to an intense turquoise-blue colour. On the basis of the XANES data, it could be concluded that the transformation of white mastodon ivory into turquoise-blue odontolite involves two phenomena: (1) a fossilization accompanied by an uptake of metal ions, specifically Mn ions (Mn2þ), possibly by sorption on apatite crystallites; and (2) a deliberate heating process in air above 6008C that oxidizes Mn2þ into Mn5þ, which substitutes for P5þ in the tetrahedral site of the apatite structure. This substitution occurs during the heat-induced crystal growth of apatite. Thus, the origin of the colour change in fossilized ivory during heating could clearly be demonstrated using XAS and, in contrast to former hypotheses, it could be shown that odontolite owes its turquoise-blue colour to traces of Mn5þ ions in a distorted tetrahedral environment of four O22 ions. 4.4.7
Therapeutic and cosmetical chemicals of Ancient Egypt
The funerary furniture discovered in Egyptian tombs, dating from between 2000 BC and 1200 BC , provides lot of information about the customs of everyday life in Ancient Egypt [99]. Among these objects there was an abundance of toilet accessories: mirrors, hairpins, eyeliner applicators, combs or spatulas, and make-up receptacles, some of which are now preserved in the Egyptian Department of the Louvre Museum (Fig. 4.48). Inside these three 4000-year-old containers made of marble, alabaster, wood or reed, cosmetic powders in an exceptionally good state of conservation were found. In order to obtain information on their composition and on the methods used in their elaboration, the organic fractions of the cosmetics were analysed by chromatographic techniques and the mineral content by
211
K. Janssens
Fig. 4.48. X-ray radiography of different makeup receptacles from the Egyptian collections of the Louvre Museum. The white areas show the distribution of the X-ray absorbing lead powders present in the make-up. (a) Reed case, still full of makeup; (b) alabaster recipient with a fabric lid; (c) alabaster recipient and cover. It contains a small amount of makeup attached on the inner wall.
scanning electron microscopy, FTIR spectrometry and powder XRD [100]. Conventional quantitative laboratory XRD was impeded by several factors: (a) owing to the high archaeological value of the powders, only small quantities could be extracted and analysed; (b) the as-found cosmetics are highly absorbing mixtures of lead-based compounds; (c) most mixtures contained as many as 10 phases, i.e., the resulting diffractograms display a complex series of overlapping Bragg lines. Measurements carried out at two synchrotron beamlines (ESRF BM16 and LURE DW22) were able to take advantage of the high flux, the high energy and the high resolution of the exciting radiation. The Rietveld refinement method was applied to determine the respective crystalline phase mass fractions. Taking into account the
212
X-ray based methods of analysis
anisotropic line profile of some phases, it was possible to significantly improve the fit agreement factors (to less than 10%) and detect quantities of minerals down to 0.5% (see Fig. 4.49). Two natural compounds bound with some animal grease were identified: crushed ore of black galena (PbS) and cerussite (PbCO3). Galena is still the basic constituent of many khols traditionally used in North Africa, Asia and the Middle-East nowadays. White cerussite was added to the composition in order to obtain a grey-to-white make-up. Rietveld analysis of the XRD patterns of the mixture revealed the presence of two more white constituents: laurionite (PbOHCl) and phosgenite (Pb 2Cl 2CO3). These products are very rare in nature and could not have been extracted from the mines in sufficient quantities for the preparation of the cosmetics. These products could have been formed by chemical alteration and ageing, assuming the original content of the make-up receptacles had been in contact with carbonated and chlorinated waters. However, no clear trace and evidence of such alteration processes could be found in any of the 49 recipients. Therefore one major conclusion of the work was that laurionite and phosgenite were intentionally manufactured by the Egyptians. The texts of Pliny the Elder and Dioscorides (1st century AD ) report on a number of medical recipes. In particular some of them refer to the use of lead oxide that was ground and diluted into salted and sometimes carbonated
Fig. 4.49. XRD pattern of a sample of Egyptian cosmetic powder (upper curve) and Rietveld refinement. Open circles: calculated pattern on the basis of four components: black galena (PbS), cerussite (PbCO3), phosgenite (Pb2Cl2CO3) and laurionite (PbOHCl) (all white); lower curve: residual.
213
K. Janssens
(natron) water. This wet process was mimicked in the laboratory. By maintaining the solution at a neutral pH, a slow reaction yields white precipitates of either laurionite or phosgenite. This is the first indication that wet chemistry has been practiced since 2000 BC . The reason for adding the white lead derivatives PbOHCl and Pb2Cl2CO3 to black PbS, instead of white cerussite (PbCO3) is that since the earliest periods of Egyptian history, cosmetics were not only employed for aesthetic purposes, but also for their therapeutic and magic/religious properties. Greco-Roman texts mention, e.g., that white precipitates synthesized from PbO are appropriate for eye and skin care; these lead compounds could be used as a bactericide and as a protection for the eye against exposure to the suns rays. The width of the diffraction peak profiles also permitted to compare the strain and crystallite size in archaeological, synthetic and natural powders [101,102]. XRD peak breadth analysis combined with SEM observations showed that the PbS ore present in the cosmetics was ground and sorted according to grain size. The resulting granulometry of galena provided the make-up with the expected texture and its metallic brightness. By contrast the Bragg line broadening of PbOHCl and Pb2Cl2CO3 is free from any strain: this suggests that they have been directly synthesized as fine powders and have not been prepared by crushing. In this manner, the XRD line broadening related to the crystallographic microstructure contributed to a better understanding of the origin and the process of elaboration of the archaeological powders. The same authors also studied the effect of Pb-containing cosmetics on the structure and composition of ancient hair samples, obtained from different Egyptian mummies, by means of X-ray micro-beam techniques [104]. In native hair, the lipids are present as calcium soaps; m-XRF and m-XRD investigations performed at ESRF ID22 and ID13 showed that appropriate lead treatment considerably enhances the organized lipid features in hair [104]. The elemental micro-analysis of Egyptian mummy hair cross-sections, compared to native modern hair, showed a significant increase in content of specific trace elements. The observed elemental distribution across the hair sections differed noticeably between hairs originating from different mummies. XRF measurements of the hair of one of the mummies revealed a notable increase in calcium (146 mmol/g on average), zinc (37 mmol/g), iron (14 mmol/g) and lead (1.1 mmol/g) contents, compared to the maximum contents of these elements observed in native samples. Other trace constituents of native hair such as manganese (1.8 mmol/g), bromine (1.1 mmol/g), titanium (1.1 mmol/g) and strontium (0.62 mmol/g) also showed
214
X-ray based methods of analysis
a significant increase. The elemental distribution across hair sections from a second mummy was found to be very heterogeneous (see Fig. 4.49). The localization of sulphur, mainly originating from hair proteins, enables a clear visualization ion of the hair section contour. Three types of elemental distribution could be distinguished: (a) elements such as Ca that are similarly distributed in the fibber section as in current hair; (b) elements concentrated specifically in some of the histological zones: Mn, Zn, Pb, Sr in the medullar canal and Fe, Cu and Pb in the peripheral areas of the hair strand; (c) elements presents as impurities, deposited on the surface of the hair strand (Ti, Zn). Considering that natron, the dehydration and purification agent used during mummification, is a complex mixture of sodium chlorides, sulphates, carbonates and bicarbonates, as well as containing small quantities of calcium carbonate, part of the trace element load in the hair must be assumed to originate from this treatment. Significant amounts of calcium and chlorine were observed in all samples while all trace elements associated with natron of the Greco-Roman period (Ca, Sr, Br, Cl, Fe, Mg) were found to be in excess in one of the examined hair samples. Apart from these elements, a very high content of Mn and Pb in the hair strands from both mummies was observed. These elements are particularly concentrated at the hair periphery and in the medulla of one of the samples. The presence of these elements was attributed to the use of Pb-based make-up and of Mn-based hair dyes in Ancient Egypt. The specific concentration of lead, calcium and other cations within the medullar canal, where the lipids were concentrated before mummification, could be related to the saponification of hair lipids. The absorption of exogenous metal cations within the fibber may have locally increased the electronic density of the material. The diffusion of metal ions can moreover have a structuring effect, by regularly organizing the keratins around the metal sites. This can be an explanation for the increase in contrast observed in the hair diffraction pattern (Fig. 4.50). 4.5
CONCLUSION
It can be concluded that X-ray techniques offer a variety of possibilities for non-destructively compositional and/or structural analysis of cultural heritage materials. Apart from determinations of the (local)
215
K. Janssens
Fig. 4.50. X-ray fluorescence maps of a Mummy hair section. The hair diameter is ca. 150 mm. The elemental content (in mmol/g w/w) is indicated, followed by the ratio to average native hair trace element content.
composition, which may be performed via XRF analysis, structural information on crystalline and amorphous phases can also be obtained via XRD and XAS. Next to standard laboratory equipment, which is not always suitable for analysis of large artefacts, open beam instrumentation is available, either under the form of compact transportable equipment for in situ measurements or as part of larger installations such as synchrotron facilities. Provenance analysis investigations as well as conservation studies can be executed in many cases by means of X-ray techniques.
Acknowledgements The author would like to thank many colleagues and collaborators at the University of Antwerp and at other institutions for the interesting collaborations, joint project activities and fruitful discussions. These include I. Deraedt, A. Aerts, O. Schalm, K. Proost, L. Vincze, B. Vekemans, P. Van Espen, R. Van Grieken and F. Adams (at the University of Antwerp, Belgium), A. Simionovici and A. Somogyi (at the European Synchrotron Radiation Facility, Grenoble, France), G. Falkenberg (at the Hamburger Synchrotron Labor, Hamburg, Germany) and A. Rindby (Gotebo¨rg University, Sweden).
216
X-ray based methods of analysis
APPENDIX A4.1
FIGURES-OF-MERIT FOR XRF SPECTROMETERS
A4.1.1 Analytical sensitivity When XRF analysis of thin film samples is performed (i.e., in samples where the product rd of sample thickness d and sample density r is so small that absorption of the incoming exciting and the outgoing fluorescent radiation in the material can be neglected, there is a linear relation between the collected net X-ray intensity Ni of a given characteristic line of element i and the irradiated mass mi ; which usually is also proportional to the concentration ci of that element in the sample: Ni ¼ Spi mi t ¼ Si ci t
ðA4:1Þ
The proportionality constants Si for the various elements are called the sensitivity coefficients of the XRF spectrometer for determination of these elements (typically expressed in counts/s/(g/cm3) or in counts/s/mg) and are important figures-of-merit of the instrument. By a selection of the excitation conditions (tube anode material, excitation voltage), the shape and location of the maximum in the sensitivity curve can be influenced to suit the needs of the application at hand. Instead of using the X-ray intensity collected during a specific time t, it is often more convenient to use the net X-ray count rate Ri: Ri ¼ Ni =t ¼ Si ci
ðA4:2Þ
A4.1.2 Detection and determination limits In reality, it is not possible to directly measure the net peak intensity Ni ; rather, a total intensity Ti ¼ Ni þ Bi is measured. The background intensity Bi can be written as the sum of various contributions: X Bi ¼ Bscatter þ Bdetector þ j–1 Boverlap þ Bblank ðA4:3Þ i i i i;j denotes the contribution to the spectral background below the where Bscatter i analytical line of element i due to scattering of the primary radiation in the sample itself, in the sample environment gas (air or Helium, if any) and (in some cases) on the sample holder materials. These phenomena cause a continuous background upon which the characteristic peaks are superimposed. Bdetector denotes the background contribution in the i same energy/wavelength region due to detector artefacts, Boverlap is the i;j 217
K. Janssens
contribution to the peak intensity resulting from unresolved overlap between lines of an element j – i and the analytical line of element i and Bblank i denotes the contributions to the peak intensity of element i not originating from the sample, i.e., a blank value. When the magnitude of Bi is experimentally determined and this measurement is repeated n times, the results will be distributed around a mean value kBil with a standard deviation sB. In modern instruments, most sources of systematic and random errors (e.g., due to mechanical or electrical instabilities) are small compared to the inherent uncertainty on the intensity measurements resulting from counting statistics. When Bi is obtained by means of a counting procedure (which usually is the case), Poisson (or counting) statistics govern the measurements so that s2B ¼ kBi l: The Union of Pure and Applied Chemistry (IUPAC) defines the limit of detection as “the lowest concentration level than can be determined to be statistically significant from an analytical blanc.” The lowest net X-ray intensity Ni,LD that can still be distinguished in a statistically significant manner from the average background level can be written as: Ni;LD ¼ kBi l þ ksB
ðA4:4Þ
where k is an integer constant depending on the significance level considered. The limit of detection concentration ci,LD corresponding to Ni,LD can be written as: Ci;LD
pffiffiffiffi Ni;LD 2 kBi l ksB k RB pffi ¼ ¼ ¼ Si t Si t Si t
where RB ¼ kBi l=t is the background count rate. When the irradiation of a standard sample (with known concentration ci ) during a time t results in net and background intensities Nistd ; and Bstd i ; so that the sensitivity Si can be std std approximated by the ratio Ni /ci /t, it follows that the lowest detectable concentration (or relative detection limit) ci,LD can be estimated from this measurement by using the relation: qffiffiffiffiffiffi k Bstd i std
ci;LD ø ci
Nistd
ðA4:5Þ
When during such an experiment, a known mass mstd was irradiated, the i lowest detectable mass (or absolute detection limit) mi,LD can be calculated 218
X-ray based methods of analysis
by means of: qffiffiffiffiffiffi k Bstd i std
mi;LD ø mi
Nistd
ðA4:6Þ
Relative detection limits are useful figures-of-merit for bulk XRF equipment, where it usually is relevant to know the lowest concentration level at which the spectrometer can be used for qualitative or quantitative determinations. In instruments where very small sample masses are being irradiated (e.g., in the pg range for m-XRF and TXRF), the absolute detection limit is another useful figure-of-merit since that provides information on the minimal sample mass than can be analysed in a given setup. In the literature, usually detection limit values for k ¼ 3 (corresponding to a statistical confidence level of 99%) are reported. A related figure-of-merit is the determination limit, which is defined as the lowest concentration (or mass) at which a quantitative determination with a relative uncertainty of at least 10% is possible. This quantity can be calculated by setting k ¼ 10 in the above expressions.
REFERENCES 1
2 3 4
5 6 7
8
Ch. Lahanier, G. Amsel, Ch. Heitz, M. Menu and H.H. Andersen, Proceedings of the International Workshop on Ion-Beam Analysis in the Arts and Archaeology, Pont-A-Mousson, Abbaye des Premontre´s, France, February 18 –20, 1985—editorial, Nucl. Instr. Meth. Phys. Res., B14 (1986) 1. S.A. Johansson and J.L. Campbell, PIXE: A Novel Technique for Elemental Analysis. Wiley, Chichester, 1988. R. Van Grieken and A. Markowicz (Eds.), Handbook of X-ray Spectrometry, 2nd ed., Marcel Dekker, New York, 2002. Ch. Hammond, The Basics of Crystallography and Diffraction, IUCr Texts on Crystallography, No. 3. International Union of Crystallography/Oxford University Press, Oxford, 1997, 249 pp, ISBN 0-19-855966-6. D.E. Sands, Introduction to Crystallography. Dover Publications, Dover, UK, 1994, ISBN: 0486678393. A. Guinier, X-Ray Diffraction in Crystals, Imperfect Crystals, and Amorphous Bodies. Dover Publications, Dover, UK, 1994, ISBN: 0486680118, 378 pp. D.L. Bish and J.E. Post (Eds.), Modern Powder Diffraction, Reviews in Mineralogy, Vol. 2. The Mineralogical Society of America, Washington, DC, 1990, 369 pp, ISBN 0-939950-24-3. R.A. Young (Ed.), The Rietveld Method, IUCr Monograph on Crystallography, No. 5, International Union of Crystallography/Oxford University Press, Oxford, 1993, 298 pp, ISBN 0-19-855577-6.
219
K. Janssens 9
10
11
12 13
14
15
16
17 18
19 20
21
22
220
D.C. Koningsberger and R. Prins (Eds.), X-ray Absorption: Principles, Applications, Techniques of EXAFS, SEXAFS and XANES. Wiley, New York, 1988, pp. 3–51. E.A. Stern, Theory of EXAFS. In: D.C. Koningsberger and R. Prins (Eds.), X-ray Absorption: Principles, Applications, Techniques of EXAFS, SEXAFS and XANES. Wiley, New York, 1988, pp. 3–51. P.J. Durham, Theory of XANES. In: D.C. Koningsberger and R. Prins (Eds.), X-ray Absorption: Principles, Applications, Techniques of EXAFS, SEXAFS and XANES. Wiley, New York, 1988, pp. 53– 84. J. Wong, Extended X-ray absorption fine structure: a modern structural tool in materials science, Mater. Sci. Engng, 80 (1986) 107 –128. G.E. Brown Jr., G. Calas, G.A. Waychunas and J. Petiau, X-ray absorption spectroscopy and its applications in mineralogy and geochemistry. In: F. Hawthorne (Ed.), Spectroscopic Methods in Mineralogy and Geology, Reviews in Mineralogy, Vol. 18. Mineralogical Society of America, 1988, pp. 431–512. L. Stru¨der, G. Lutz, P. Lechner, H. Soltau and P. Holl, Semiconductor detectors for (imaging) X-ray spectroscopy. In: K. Tsuji, J. Injuk and R. Van Grieken (Eds.), X-Ray Spectrometry: Recent Technological Advances. Wiley, Chichester, 2004, pp. 134– 193, Chapter 4.1. S. Piorek, Radio-isotope exited X-ray analysis. In: R. Van Grieken and A. Markowicz (Eds.), Handbook of X-ray Spectrometry, 2nd ed., Marcel Dekker, New York, 2002, Chapter 7. R. Cesareo, A. Brunetti, A. Castellano and M.A. Morales Medina, Portable equipment for X-ray fluorescence analysis. In: K. Tsuji, J. Injuk and R. Van Grieken (Eds.), X-Ray Spectrometry: Recent Technological Advances. Wiley, Chichester, 2004, pp. 307– 341, Chapter 5.3. K. Janssens, F. Adams and A. Rindby, Microscopic X-Ray Fluorescence Analysis. Wiley, Chicester, 2000. K. Janssens, L. Vincze, B. Vekemans, F. Adams, M. Haller and A. Kno¨chel, The use of ellipsoidal lead-glass capillaries for micro-focussing of highly energetic (0 – 60 keV) synchrotron radiation, J. Anal. Atom. Spectrom., 13 (1998) 339–350. S.B. Dabagov, S.V. Nikitina and M.A. Kumakhov, Nucl. Inst. Meth. Phys. Res. B, 103(1) (1995) 99 –105. N. Gao and K. Janssens, Polycapillary X-ray optics. In: K. Tsuji, J. Injuk and R. Van Grieken (Eds.), X-Ray Spectrometry: Recent Technological Advances. Wiley, Chichester, 2004, pp. 89–110, Chapter 3.3. A. Simionovici, C. Schroer and B. Lengeler, Parabolic compound refractive X-ray lenses. In: K. Tsuji, J. Injuk and R. Van Grieken (Eds.), X-Ray Spectrometry: Recent Technological Advances. Wiley, Chichester, 2004, pp. 111–131, Chapter 3.4. B. Lai, B. Yun, D. Legnini, Y. Xiao, J. Chrzas, P.J. Vicaro, V. White, S. Bajikar, D. Denton, F. Cerrina, E. Di Fabrizio, M. Gentili, L. Grella and M. Baciocchi, Hard X-ray phase zone plate fabricated by lithographic techniques, Appl. Phys. Lett., 61 (1992) 1877–1879.
X-ray based methods of analysis 23
24
25 26
27
28
29
30
31
32
33
34
35 36 37
M. Yanagihara and K. Yamashita, Multilayers for soft and hard X-rays. In: K. Tsuji, J. Injuk and R. Van Grieken (Eds.), X-Ray Spectrometry: Recent Technological Advances. Wiley, Chichester, 2004, pp. 63–87, Chapter 3.1. G. Vittiglio, K. Janssens, B. Vekemans, F. Adams and A. Oost, A compact smallbeam XRF-instrument for in-situ analysis of objects of historical and/or artistic value, Spectrochim. Acta B, 54 (1999) 1697–1710. P. Moioli and C. Seccaroni, Analysis of art objects using a portable X-ray fluorescence spectrometer, X-ray Spectrom., 29 (2000) 48–52. C. Fiorini, J. Kemmer, P. Lechner, K. Kromer, M. Rohde and T. Schulein, A new detection system for X-ray microanalysis based on a silicon drift detector with Peltier cooling, Rev. Sci. Instr., 68 (1997) 2461 – 2465. C. Fiorini and A. Longoni, Application of a new noncryogenic X-ray detector in portable instruments for archaeometric analyses, Rev. Sci. Instr., 69 (1998) 1523 –1528. H. Bronk, S. Ro¨hrs, A. Bjeoumikhov, N. Langhoff, J. Schmalz, R. Wedell, H.E. Gorny, A. Herold and U. Waldschla¨ger, ArTAX—a new mobile spectrometer for energy-dispersive micro X-ray fluorescence spectrometry on art and archaeological objects, Fresenius J. Anal. Chem., 371 (2001) 307 –316. S. Bichlmeier, K. Janssens, J. Heckel, D. Gibson, P. Hoffmann and H.M. Ortner, Component selection for a compact micro-XRF spectrometer, X-ray Spectrom., 30 (2001) 8–14. S. Bichlmeier, K. Janssens, J. Heckel, P. Hoffmann and H.M. Ortner, Comparative material characterization of historical and industrial samples by using a compact micro-XRF spectrometer, X-ray Spectrom., 31(1) (2002) 87–91. G. Vittiglio, S. Bichlmeier, P. Klinger, J. Heckel, W. Fuzhong, L. Vincze, K. Janssens, P. Engstro¨m, A. Rindby, K. Dietrich, D. Jembrih-Simbu¨rger, M. Schreiner, D. Denis, A. Lakdar and A. Lamotte, A compact m-XRF spectrometer for (in-situ) analyses of cultural heritage and forensic materials, Nucl. Instr. Meth. Phys. Res. B, 213 (2004) 693–698. K. Janssens, B. Vekemans, L. Vincze, F. Adams and A. Rindby, A micro XRF spectrometer based on rotating anode generator and capillary optics. Spectrochimica Acta: Part B: Atomic Spectrometry, Spectrochim. Acta, B51 (1996) 1661 –1678. L. Vincze, K. Janssens, F. Adams and K.W. Jones, A general Monte-Carlo simulation of ED-XRF spectrometers: part III, polarized polychromatic radiation, homogeneous samples, Spectrochim. Acta B, 50 (1995) 1481 –1500. L. Vincze, K. Janssens, B. Vekemans and F. Adams, Modeling of photon scattering at high X-ray energies: experiment vs. simulation, J. Anal. Atom. Spectrom., 14 (1999) 529–533. K. Janssens, L. Vincze, A. Aerts, F. Adams and J. Hertogen, Automated segmentation of m-XRF image sets, X-ray Spectrom., 26 (1997) 333–346. A. Kno¨chel and M. Haller, X-ray fluorescence analysis using synchrotron radiation (SYXRF), J. Trace Microprobe Techn., 14(3) (1996) 461–488. K. Janssens, L. Vincze, B. Vekemans, C.T. Williams, M. Radtke, M. Haller and A. Kno¨chel, The non-destructive determination of REE in fossilized bone using
221
K. Janssens
38
39
40
41
42
43 44
45
46
47
48
49
50
51
222
synchrotron radiation induced K-line X-ray microfluorescence analysis, Fres. J. Anal. Chem., 363 (1998) 413–420. L. Vincze, K. Janssens, B. Vekemans and F. Adams, Monte Carlo simulation of X-ray fluorescence spectra: part 4: photon scattering at high X-ray energies, Spectrochim. Acta B, 54 (1999) 1711–1722. G. Falkenberg, O. Clauss, A. Swiderski and Th. Tschentscher, Upgrade of the X-ray fluorescence beamline at HASYLAB/DESY, X-Ray Spectrom., 30 (2001) 170–173. L. Vincze, F. Wei, K. Proost, B. Vekemans, K. Janssens, Y. He, Y. Yan and G. Falkenberg, Suitability of polycapillary optics for focussing of monochromatic synchrotron radiation as used in trace level micro-XANES measurements, J. Anal. Atom. Spectrom., 17 (2002) 177– 182. K. Proost, L. Vincze, K. Janssens, N. Gao, E. Bulska, M. Schreiner and G. Falkenberg, Characterization of a polycapillary lens for use in microXANES experiments, X-ray Spectrom., 32 (2003) 215 –222. K. Janssens, Comparison to other techniques. In: K. Janssens, F. Adams and A. Rindby (Eds.), Microscopic X-Ray Fluorescence Analysis. Wiley, Chicester, 2000, Chapter 7. K. Janssens, L. Vincze, F. Adams and K.W. Jones, Synchrotron X-ray microanalysis, Anal. Chim. Acta, 283 (1993) 89 –119. L. Vincze, B. Vekemans, I. Szalo´ki, K. Janssens, R. Van Grieken, H. Feng, K.W. Jones and F. Adams, High-resolution X-ray fluorescence micro-tomography on single sediment particles. In: U. Bonse Ulrich (Ed.), Proceedings SPIE Conference on Developments in X-ray Tomography: 3, Bellingham, Washington, 2002, pp. 240 –245. C.M. Hansel, M.J. Laforce, S. Fendorf and S. Sutton, Spatial and temporal association of As and Fe species on aquatic plant roots, J. Environ. Sci. Technol., 36 (2002) 1988–1994. A.A. Bjeoumikhov, N. Langhoff, J. Schmalz, R. Wedell, V.I. Beloglazov and N.F. Lebeder, Polycapillary conic collimator for micro-XRF, Proc. SPIE, 3444 (1998) 430–435. K. Janssens, K. Proost and G. Falkenberg, Confocal microscopic X-ray fluorescence at the HASYLAB microfocus beamline: characteristics and possibilities, Spectrochim. Acta B, (2004) in press. B. Kanngießer, W. Malzer and I. Reiche, A new 3D micro X-ray fluorescence analysis set-up—first archaeometric applications, Nucl. Instr. Meth. Phys. Res. B, 211 (2003) 259 –264. A. Somogyi, M. Drakopoulos, L. Vincze, B. Vekemans, C. Camerani, K. Janssens, A. Snigirev and F. Adams, ID18F: a new micro-X-ray fluorescence radiation facility at the European synchrotron radiation facility (ESRF): preliminary results, X-ray Spectrom., 30 (2001) 242– 252. J.L. De Vries and B.A.R. Vrebos, Quantification of infinitely thick specimen by XRF analysis. In: R. Van Grieken and A. Markowicz (Eds.), Handbook of X-ray Spectrometry, 2nd ed., Marcel Dekker, New York, 2002, Chapter 5. F. He and P. Van Espen, General approach for quantitative energy dispersive X-ray fluorescence analysis based on fundamental parameters, Anal. Chem., 63 (1991) 2237–2244.
X-ray based methods of analysis 52
53 54 55 56 57
58
59 60 61 62 63
64
65 66 67
68
69
70
H.H. Otto, W. Hofmann and K. Schro¨der, A double-radius Gandolfi X-ray camera for the generation of powder-like diffractograms of small single crystals, using an imaging plate detector, J. Appl. Cryst., 35 (2002) 13–16. H.M. Rietveld, Line profiles of neutron powder-diffraction peaks for structure refinement, Acta Crystallogr., 22 (1967) 151–152. H.M. Rietveld, A profile refinement method for nuclear and magnetic structures, J. Appl. Crystallogr., 2 (1969) 65–71. A. Albinati and B.T.M. Willis, J. Appl. Crystallogr., 15 (1982) 361 –374. L.B. McCusker, R.B. Von Dreele, D.E. Cox, D. Louer and P. Scardi, Rietveld refinement guidelines, J. Appl. Cryst., 32 (1999) 36–50. F. Adams, K. Janssens and A. Snigirev, Microscopical X-ray fluorescence analysis and related methods with laboratory and synchrotron radiation sources, J. Anal. Atom. Spectrom., 13 (1998) 319 –331. M. Milazo and C. Cicardi, Simple methods for quantitative X-ray fluorescence analysis of ancient metal objects of archaeological interest, X-ray Spectrom., 26 (1997) 211–216. E. Ciliberto and G. Spoto (Eds.), Modern analytical methods in art and archaeology, Chem. Anal., Vol. 155. Wiley, New York, 2000. D.C. Creagh and D.A. Bradley (Eds.), Radiation in Art and Archaeometry. Elsevier, Amsterdam, 2000. K. Tsuji, J. Injuk and R. Van Grieken (Eds.), X-ray Spectrometry: Recent Technological Advances. Wiley, Chichester, 2004, ISBN 0-471-48640-X. E. Selin (Ed.), Special Millenium Issue on Cultural Heritage, X-ray Spectrometry, Vol. 29. 2000, pp. 1–129. P. Wobrauschek, G. Halmetschlager, S. Zamini, C. Jokubonis, G. Trnka and M. Karwowski, Energy-dispersive X-ray fluorescence analysis of Celtic glasses, X-ray Spectrom., 29(1) (2000) 25–33. C. Jokubonis, P. Wobrauschek, S. Zamini, M. Karwowski, G. Trnka and P. Stadler, Results of quantitative analysis of Celtic glass artifacts by energy dispersive X-ray fluorescence spectrometry, Spectrochim. Acta B, 58(4) (2003) 627–633. A. Aerts, K. Janssens and F. Adams, Trace level microanalysis of Roman glass from Khirbet Qumran, Israel, J. Archaeol. Sci., 26 (1999) 883 –891. A. Aerts, B. Velde and K. Janssens, Change in silica sources in Roman and post Roman glass, Spectrochim. Acta B, 58 (2003) 659 –667. P. Hoffmann, S. Bichlmeier, M. Heck, C. Theune and J. Callmer, Chemical composition of glass beads of the Merovingian period from graveyards in the Black Forest, Germany, X-Ray Spectrom., 29(1) (2000) 92 –100. P. Hoffman, Analytical determinaton of coloring elements and their compounds in glass beads from graveyards of the Merowings time, Fres. J. Anal. Chem., 349 (1994) 320 –333. B. Velde, Alumina and calcium oxide content of glasses found in western and northern Europe, First to Ninth centuries, Oxford J. Archaeol., 9(1990) (1990) 105–117. D.C.W. Sanderson, J.R. Hunter and S.E. Warren, Energy dispersive X-ray fluorescence analysis of 1st millennium AD glass from Britain, J. Archaeol. Sci., 11 (1984) 53–70.
223
K. Janssens 71 72
73
74
75 76
77 78
79
80
81 82 83
84
85 86
224
O. Dussart and B. Velde, La composition du verre Hellenistique en Jordanie et Syrie du Nord, Syria, 67 (1990) 687– 693. A. Longoni, C. Fiorini, P. Leutenegger, S. Sciuti, G. Fronterotta, L. Stru¨der and P. Lechner, A portable XRF spectrometer for non-destructive analysis in archaeometry, Nucl. Instr. Meth. Phys. Res. A, 409 (1998) 395 –400. E. Aloupi, A. Karydas and T. Paradellis, Pigment analysis of wall paintings and ceramics from Greece and Cyprus. The optimum use of X-ray spectrometry on specific archaeological issues, X-ray Spectrom., 29 (2000) 18 –24. R. Cesarao, A. Castellano, G. Buccolieri, S. Quarta, M. Marabelli, P. Santopadre, M. Ieole and A. Brunetti, Portable equipment for energydispersive X-ray fluorescence analysis of Giotto’s frescoes in the Chapel of the Scrovegni, Nucl. Instr. Meth. Phys. Res. B, 213 (2004) 703– 706. Z. Smit, K. Janssens, K. Proost and I. Langus, Confocal m-XRF depth analysis of paint layers, Nucl. Instr. Meth. Phys. Res. B, 219 –220 (2004) 35 –40. R. Klockenka¨mper, A. von Bohlen and L. Moens, Analysis of pigments and inks in oil paintings and historical manuscripts using total reflection X-ray fluoresence spectrometry, X-ray Spectrom., 29 (2000) 119 –129. A. Caiger-Smith, Luster Pottery. Technique, Tradition and Innovation in Islam and the Western World. Faber & Faber, London, UK, 1985. S. Padovani, C. Sada, P. Mazzoldi, B. Brunetti, I. Borgia, A. Sgamellotti, A. Giulivi, F. D’Acapito and G. Battaglin, Copper in glazes of Renaissance luster pottery: nanoparticles, ions, and local environment, J. Appl. Phys., 93(12) (2003) 10058–10063. J. Pe´rez-Arantegui, J. Molera, A. Larrea, T. Pradell, M. Vendrell-Saz, I. Borgia, B.G. Brunetti, F. Cariati, P. Fermo, M. Mellini, A. Sgamellotti and C. Viti, Luster pottery from the thirteenth century to the sixteenth century: a nanostructured thin metallic film, J. Am. Ceram. Soc., 84 (2001) 442– 446. I. Borgia, B. Brunetti, I. Mariani, A. Sgamelotti, F. Cariati, P. Fermo, M. Mellini, C. Viti and G. Padeletti, Heterogeneous distribution of metal nanocrystals in glazes of historical pottery, Appl. Surf. Sci., 185 (2002) 206–316. W.D. Kingery and P.B. Vandiver, Ceramic Masterpieces. Art, Structure and Technology. The Free Press, New York, 1986, pp. 111– 121. C. Piccolpasso, Li Tre Libri dell’Arte del Vasaio, 1557. Edizioni all’Insegna del Giglio, Firenze, 1976. S. Bichlmeier, K. Janssens, J. Heckel, P. Hoffmann and H.M. Ortner, Comparative material characterization of historical and industrial samples by using a compact micro-XRF spectrometer, X-Ray Spectrom., 31(1) (2002) 87–91. I. De Ryck, A. Adriaens, F. Adams and E. Pantos, A comparison of microbeam techniques for the analysis of corroded ancient bronze objects, Analyst, 128 (2003) 1104– 1109. H. Mommsen, Th. Beier, H. Dittmann, D. Heimermann, A. Hein, A. Rosenberg, M. Boghardt, E.-M. Manebutt-Benz and H. Halbey, Archeometry, 38 (1996) 347. T. Cahill, B. Kusko and R.N. Schwab, Analyses of inks and papers in historical documents through external beam pixe techniques, Nucl. Instr. Meth. Phys. Res., B181 (1981) 205.
X-ray based methods of analysis 87 88 89
90 91
92 93
94
95
96
97
98
99 100
101
102
T. Cahill, B. Kusko, R.A. Eldred and R.N. Schwab, Archaeometry, 26 (1984) 3. B.T. Kusko and R.N. Schwab, Historical analyses by PIXE, Nucl. Instr. Meth. Phys. Res. B, 22 (1987) 401 –406. R.N. Schwab, T.A. Cahill, B.H. Kusko, R.A. Eldred, G. Moller and D. Dutschke, Cyclotron analysis of the ink in the 42-line Bible, Papers Biblio. Soc. Am., 77 (1983) 285–315. C.H. Bloy, A History of Printing Ink, Balls and Rollers 1440 – 1850. Butterworths, London, 1967. S. Larsson, X-Ray microbeam spectroscopy, Thesis, Department of Physics, Chalmers University of Technology, S-41296 Goteborg, Sweden, ISBN 91-7032579-0, Bibliotekets ReproService, Goteborg, Sweden, 1991. B. Stocklassa, G. Nillson and N. Paulson, paper presented at the European Conference on EDXRS, Myconos, Greece, May 30–June 6, 1992. R. Fuchs, O. Hahn and D. Oltrogge, Geist und Seele sind verwirret (die tintenfraßproblematik der autographen Johann Sebastian Bachs, Restauro, 106(2) (2000) 116–121. A. Denker, O. Hahn, B. Kanngießer, W. Malzer, S. Merchel, M. Radtke, S. Ro¨hrs, I. Reiche and H. Stege, Chemie der Kunst—Zersto¨rungsfreie Analyse von Kunst—und Kulturgu¨tern, MP Materialpru¨fung, 45(11 – 12) (2003) 485–503. I. Reiche, A. Berger, W. Go¨rner, S. Merchel, M. Radtke, J. Riederer, H. Riesemeier and M. Roth, Following the traces of Albrecht Du¨rer: analysis of silverpoint drawings by spatially resolved synchrotron-induced X-ray fluorescence analysis, Nucl. Instr. Meth. Phys. Res. B, (2004) in press. I. Reiche, M. Radtke, A. Berger, T. Ketelsen, S. Merchel, W. Go¨rner, H. Riesemeier, J. Riederer and M. Roth, Spatially resolved synchrotroninduced X-ray fluorescence analyses of metal point drawings and their mysterious inscriptions, Spectrochim. Acta B, (2004) in press. I. Reiche, C. Vignaud, B. Champagnon, G. Panczer, C. Brouder, G. Morin, V.A. Sole´, L. Charlet and M. Menu, From mastodon ivory to gemstone: the origin of turquoise color in odontolite, Am. Mineral., 86 (2001) 1519 –1524. I. Reiche, C. Vignaud, T. Calligaro, J. Salomon and M. Menu, Comparative analysis of odontolite, heated fossil ivory and blue fluorapatite by PIXE/PIGE and TEM, Nucl. Instr. Meth. B, 161 (2000) 737 –742. A. Lucas and J.R. Harris, Ancient Egyptian Materials and Industries. Edwards Arnold Ltd, London, 1963. Ph. Walter, P. Martinetto, G. Tsoucaris, R. Bre´niaux, M.A. Lefebvre, G. Richard, J. Talabot and E. Dooryhee, Manufacturing cosmetics in ancient Egypt, Nature, 397 (1999) 483 –484. P. Martinetto, M. Anne, E. Dooryhee and Ph. Walter, Analysis of X-ray diffraction line profile of galena powders: a clue to some practices of mineral crushing in ancient Egypt. Proceedings of the Sixth European Conference on Powder Diffraction (EPDIC6), Materials Science Forum, 1998. T. Ungar, P. Martinetto, G. Ribarik, E. Dooryhee, P. Walter and M. Anne, Revealing the powdering methods of black makeup in Ancient Egypt by fitting microstructure based Fourier coefficients to the whole X-ray diffraction profiles of galena, J. Appl. Phys., 91 (2003) 2455–2465.
225
K. Janssens 103
104
226
L. Bertrand, P. Chevallier, J. Doucet, A. Simionovici, G. Tsoucaris and P. Walter, Apport du rayonnement synchrotron a l’etude de cheveux archeologiques, J. Phys. IV, 12 (2002) 237 –243. L. Bernard, J. Doucet, A. Simionovici, G. Tsoucaris and P. Walter, Lead revealed lipid organisation in human hair, Biochem. Biophys. Acta, 1620 (2003) 218– 224.
Chapter 5
Ion beam microanalysis T. Calligaro, J.-C. Dran and J. Salomon
5.1
HISTORICAL BACKGROUND AND MOTIVATION
Electrostatic accelerators were the first type of charged particle accelerators to have been designed back in the early 1930s. They were originally dedicated to Nuclear Physics, but the constant need for higher and higher energy by nuclear physicists left them progressively unused. Their availability soon attracted the interest of solid-state physicists and material scientists as potential tools for both materials’ processing and analysis. A new set of analytical techniques labelled ion beam analysis (IBA) quickly grew and was extensively applied in various scientific fields. The birth of two IBA techniques, Rutherford backscattering spectrometry (RBS) and nuclear reaction analysis (NRA) occurred in 1957. Then the field of IBA rapidly expanded in the early 1960s, mostly due to the development of semiconductor detectors, and accelerators entirely devoted to IBA were built all around the world. Another analytical method, namely particle-induced X-ray emission (PIXE) was developed in 1970. International conferences on this new field were soon held: on IBA in 1973 and on PIXE in 1977. Applications to art history and archaeology were initiated in 1972 and rapidly grew because of the intrinsic capabilities of these new tools, the most important of which is the non-destructiveness. In 1985, an international workshop was specially held on that topic in Pont-a`-Mousson, France (see proceedings in Ref. [58]). Its conclusions led to the design and building, at the end of 1987, of an IBA facility located in the Louvre museum dedicated to the study of cultural heritage. More details on this historical background can be found in several review articles [1]. In the following sections we briefly describe the main features of IBA techniques and their applications to issues relevant to cultural heritage. We will emphasize the major impact of external beam set-ups, which have rendered these techniques totally non-destructive and thus fully adapted to Comprehensive Analytical Chemistry XLII Janssens and Van Grieken (Eds.) q 2004 Elsevier B.V. All rights reserved
227
T. Calligaro et al.
the study of precious items. This will be highlighted by selected examples from recent literature. A more extensive bibliography covering the research activities classified according to the types of materials is provided in appendix. 5.2
FUNDAMENTALS OF ION BEAM ANALYSIS
The aim of the present chapter is to provide to the reader unfamiliar with this field the basic knowledge on IBA. IBA techniques rely on the interaction of light ions of energy in the MeV range with constituent atoms of materials and the detection of secondary products, which can be either photons or ions and have an energy characteristic of the target atom (Fig. 5.1). We assume that the emission happens during irradiation (prompt) and hence analytical techniques based on delayed emission such as particle activation are not considered. The main IBA techniques and
Fig. 5.1. Physical principles of IBA techniques. Particle-induced X-ray emission (PIXE) is a two-step process: an inner-shell electron of the target atom is expelled by the impinging ion, then follows an electronic rearrangement accompanied by X-ray emission. Rutherford backscattering spectrometry (RBS) relies on a purely elastic process based upon the electrostatic repulsion between positively charged projectile and target nuclei. NRA occurs when the projectile and the target nuclei come close enough to undergo a nuclear reaction with emission of characteristic photons or charged particles.
228
Ion beam microanalysis
their acronym are the following: † † † †
particle-induced X-ray emission (PIXE) Rutherford backscattering spectrometry (RBS) elastic recoil detection analysis (ERDA) nuclear reaction analysis (NRA), comprising particle-induced gammaray emission (PIGE) and charged particle detection
The physical principles on which these techniques are based are briefly presented below. Let us recall the main qualities of IBA techniques: 1. non-destructive for most materials, except organic compounds which can be sensitive to heat or radiation damage 2. highly quantitative with an accuracy generally better than 5% 3. multielemental, including light elements 4. one of them (PIXE) is very sensitive 5. complementary and can be implemented simultaneously 6. can provide spatial information via depth profiling and micro-mapping 7. can be operated directly on artefacts with external beams Their limitations are the following: 1. only probe the near-surface of materials (10 – 20 mm) and thus the bulk composition can be smeared by surface alteration 2. do not give any information on the chemical state of elements 3. the analysis of insulating materials can be problematic because of electrical charging It is out of the scope of this chapter to describe these techniques in detail. We will only introduce some basic background on the interaction of radiation and matter, as well as the principle of the various methods. Many textbooks give a thorough description of the IBA methods. Some of them describe several techniques [2,3], while some others are dedicated to a specific one: PIXE [4,5], RBS [6], ERDA [7], NRA [8]. A detailed description of the nuclear microprobe and its applications can also be found [9,10]. We intend to give the reader some useful and practical information, and answer the following questions: How do IBA methods work? What are the benefits and disadvantages of IBA? How has the technique been adapted to the study of cultural heritage? Which other techniques can complement IBA? Which questions in art and archaeology can IBA methods answer? What are some recent examples of applications?
229
T. Calligaro et al.
5.2.1
Interaction of radiations with matter
With IBA methods the identification of the target constituents always relies on the measurement of the energy of a radiation emitted by the sample subjected to particle bombardment, which can be either a charged particle or a photon. It is important to note that charged particles and photons exhibit a distinct behaviour in the interaction with the sample material. As they penetrate the sample, the number of photons decreases, while their energy is left unchanged. On the contrary, the number of charged particles is kept constant, while their energy is gradually decreased until complete stopping down. This has a dramatic consequence on the capabilities of photon and charged particle-based IBA techniques. As the photons emitted by atoms bear no information about the depth where they were produced, photonbased methods like PIXE or PIGE have a poor profiling capability and therefore are preferred for bulk analysis. On the other hand, it is possible to measure simultaneously a large number of elements. On the contrary, the energy of an outgoing charged particle depends both on the nature of the emitting atom and on the depth from where it was emitted. Provided that these parameters can be unfolded, methods based on the use of charged particles like RBS and ERDA are excellent for the determination of layer thickness or depth profiles. The drawback is that the characterization of samples containing more than a few elements can be a very difficult task. 5.2.1.1 Ion – matter interaction Average energy loss The understanding of the way in which MeV light ions lose energy through materials is essential for the mastering of IBA methods. Indeed, this energy loss process enables the conversion of the energy axis of ion spectra in depth units in the target. Ions lose their energy during the passage through matter by colliding with the clouds of electrons or the nuclei of the target atoms. The rate of energy loss per length unit, dE=dx; is conveniently described in two regimes. At high energy (above 25 keV for H and 100 keV for He) dE=dx is approximately proportional to Z21 Z2 =E; where Z1 and Z2 are, respectively, the atomic numbers of the incident ion and target element. Therefore dE=dx decreases when E increases. Because of the Z21 dependence, the energy loss of helium ions is four times higher than for protons. In the low-energy regime the situation is reversed as dE=dx decreases with the energy of the ion. But this regime is not so important for the IBA techniques, because useful signals usually originate from the beginning of the trajectory where the ions still have a large energy. The dE=dx values can be expressed in keV/mm, but data
230
Ion beam microanalysis
in keV/g/cm2 or eV/1015 atoms are often preferred since they do not imply the knowledge of the sample density. Range MeV ions continually lose their kinetic energy in collisions until they come to rest at some depth within the material. The rate of energy loss increases until the ion penetrates further, and reaches a maximum close to the end of range of the trajectory. The average range R which ions travel through matter before rest is defined as: R¼
ðE0 0
dE dx
!21 dE
ð5:1Þ
Straggling The passage of an initially monokinetic beam through matter induces a spread for the energies of the ions. The energy straggling sE is defined as the width of the energy distribution of the ions. This straggling is an important parameter as it limits the depth resolution for the determination of elemental profiles by RBS and NRA. Moreover, the ions randomly deviate from the linear trajectory due to scattering with nuclei, and this is known as the angular straggling su. It is interesting to note that straggling also sets a limit for the spatial resolution achievable with external beams. Practically, the calculation of energy loss, range and straggling are performed using several computer codes, the most famous being the TRIM Monte Carlo simulation program [11]. Figure 5.2 shows the trajectories of 3 MeV protons and He ions simulated by TRIM (SRIM 2003 version) in a copper target, as well as the ion distribution at the end of the path. Table 5.1 gives, for 3 MeV protons and helium ions, the dE=dx; R and sx values for some materials commonly used for works of art and archaeological artefacts. In solids, the range of ions varies between 30 and 150 mm for protons, and only a few microns for helium particles. Protons can therefore pass through a sheet of paper but for most bulk samples only the very close surface is probed. The reduced lateral spread (3– 7 mm) shows that, contrary to electrons, ion paths keep almost parallel. Consequently ion beams are suitable for depth profiling and as a high spatial resolution probe. Beam damage Even though ion beam techniques are generally considered as nondestructive, some damage induced by the impinging beam has been observed on delicate and sensitive samples. The change of visual
231
T. Calligaro et al.
Fig. 5.2. Simulation of trajectories and ranges of 3 MeV protons (upper graphs) and helium ions (lower graphs) in a bulk copper target. These graphs were obtained by running the SRIM 2003 program (based on the original work by Biersack and Haggmark (1980), it can be downloaded from www.srim.org). Note the much longer range and the wider range distribution for protons than for helium ions.
appearance has been evaluated by spectrophotometric methods and the mechanical weakening by resistance tests. Most of the mineral and metallic samples do not suffer from the analyzing beam, while organic compounds are more sensitive. The damage results from the destruction of chemical bonds (radiolysis) leading to breaking of polymeric molecules or gaseous decomposition, loss of water (dehydration), and possibly migration of mobile elements. Zeng et al. have investigated the damage induced in paper-like samples, leading to a discoloration and reduction of tensile strength [12]. Alterations in the form of dark or brownish stains have been reported in glass and glazes [13]. In paintings, various types of degradation of organic constituents such as binders or varnishes have been observed: darkening of pigments such as white lead (damage to organic binding oils) and blistering of the varnish surface. The damage is correlated to the deposited dose per unit of irradiated area (beam fluence expressed in mC/mm2), rather than to the heating of the sample. Alterations may appear with integrated doses in the micro-Coulomb range. Surprisingly in several cases the damage turned out to be
232
TABLE 5.1 Energy loss dE=dx; range R and lateral spread sx of ions of the beam at the end of the trajectory for 3 MeV hydrogen and helium ions, for various materials of cultural heritage relevance Material
Composition
r (g/cm3) Protons 3 MeV
Helium 3 MeV
Paper Plexiglas Quartz Soda-glass Lead-glass Lead Copper Bronze Silver Gold Air (gas) Helium (gas)
C6H10O5 C4H6O2 SiO2 Na10MgAlSi25O60 Na7Si24K4Pb5O59 Pb Cu Cu89Zn9Pb2 Ag Au N3O1 He
1 1.2 2.32 2.3 4.8 11.3 8.9 8.8 10.5 19.3 0.0012 0.0002
12 14 21 22 35 42 55 53 53 72 0.012 0.0022
150 126 82 82 51 45 35 35 36 27 1.4 £ 104 7.7 £ 105
sx (mm)
dE=dx (keV/mm) R (mm)
4.6 3.7 3.5 3.5 3.6 6.7 2.7 3.0 3.7 3.9 4.9 £ 103 1.3 £ 104
124 150 220 219 344 330 490 470 456 560 0.13 0.026
sx (mm)
18 0.5 15 0.4 11 0.5 11 0.5 7.0 0.4 7.3 0.9 5.5 0.5 5.6 0.5 5.4 0.5 4.7 0.6 1.7 £ 103 530 89 £ 103 1.4 £ 103
Ion beam microanalysis
dE=dx (keV/mm) R (mm)
233
T. Calligaro et al.
reversible and has disappeared after a few days. In some exceptional cases sample damage has been observed even with the lowest beam intensity suitable for analysis [14]. Consequently, the most sensitive targets (manuscripts, papyrus, paintings) are analyzed with a very low beam intensity (0.1 nA) applied during a short time (100 s). The beam can be spread or scanned on a larger area to lower the beam fluence. More generally, as it is very difficult to predict the behaviour of the sample under the beam greatest care should be taken prior and during analysis. Preliminary checks of irradiation have to be carried on a less important part of the object and constant visual monitoring of the impact point during the experiment is mandatory. The beam dose has to be kept as small as possible by improving the detection efficiency (tight geometry, large detector diameter). An alternative to PIXE when the material under study is too fragile to be directly exposed to the proton beam (case of paintings, textiles, woods or biological samples like skin or hairs) is the PIXE-induced XRF technique. This approach will be briefly described in section 5.4.2.1. 5.2.1.2 Photon –matter interaction Interaction of photons with matter intervene at various levels of photonbased IBA experiments like PIXE or PIGE: self-absorption in the sample, selection of appropriate absorbers, performance of photon detectors, etc. Its knowledge is therefore necessary to optimize the experimental set-up and adjust the parameters for processing the collected spectra. The incidence of photon interaction on detector features (efficiency, spectrum artefacts, etc.) are common to other photon-based analytical techniques like XRF or SEM – EDX. This chapter deals with the aspects relevant to ion beam techniques, and for an extensive description of X-ray detection we recommend the reader to refer to Chapter 4. Absorption of X-rays and g-rays When N0 photons of energy E pass through a sample of thickness x in cm or t in g/cm2, the number N of photons transmitted is expressed by the Beer– Lambert law N ¼ N0 expð2mp xÞ or N ¼ N0 expð2mtÞ
ð5:2Þ
where mp is the linear attenuation coefficient in units of cm21 and m ¼ mp/r is the mass attenuation coefficient in units of cm2/g. The latter has the advantage of being independent of the specific gravity r of the medium. The N0 2 N photons which are “lost” within the sample are absorbed or 234
Ion beam microanalysis
scattered by four phenomena: † † † †
photoelectric absorption Rayleigh or elastic scattering Compton or inelastic scattering pair production
We should mention an additional inelastic scattering process named Raman scattering, but its low intensity makes it seldom parasitic apart in very specific set-up like PIXE – XRF. The dependence of the scattering of photons as a function of their energy for a lead target is depicted in Fig. 5.3. Each process dominates in a particular energy domain. In the low-energy range (0 – 100 keV), the photoelectric effect is dominant. In the intermediate regime (100 – 1000 keV) Compton scattering is prevailing. For very high-energy photons (above 1 MeV) the pair production has the main contribution. The Rayleigh or elastic scattering has a constant contribution up to 10 keV where it starts
Fig. 5.3. The different modes of photon matter interaction for a lead target as a function of the photon energy. Each process dominates in a particular energy domain. In the low-energy range (0 –100 keV), the photoelectric effect is dominant. In the intermediate regime (100 –1000 keV) Compton scattering is prevailing. For very high-energy photons (above 1 MeV) the pair production is the main contributor.
235
T. Calligaro et al.
to decrease. Since it produces a photon of the same energy as the initial one its effect is usually negligible. As IBA techniques produce photons in a wide range of energy (soft X-rays from to 1 keV in PIXE mode up to g-rays of several MeV in PIGE mode), all four phenomena might be implied. The common artefacts found in g-rays spectra are exemplified in Fig. 5.9 showing the PIGE spectrum of an emerald (Be3Al2Si6O18). The Compton scattering gives rise to an asymmetric broad structure located on the low energy side of a main peak of energy E0. This structure starts at an energy E called the Compton edge which depends on the scattering angle u and extends towards zero: E¼
E0 ; E0 in keV E0 ð1 2 cos uÞ 1þ 511
ð5:3Þ
The sum of Compton contributions of all peaks generates a global background in g-ray spectra which hampers the detection limit of the PIGE method. The Compton effect can also be observed for high-energy X-ray lines in PIXE spectra recorded in high-Z matrices. The eþe2 pair production process occurring above 1 MeV induces two escape peaks accompanying each photopeak. Indeed the annihilation of the positron with an electron of the detector crystal generates two 511 keV photons. One or both of them might escape from the detector, and the total energy deposited in the detector is reduced by the corresponding amount. For very high-energy photons (e.g., 6 – 7 MeV g-rays produced by 19F(p,ag)16O nuclear reaction) the escape lines might have a higher intensity that the photopeak. In the photoelectric regime which prevails in a PIXE experiment, the attenuation coefficient is uniformly decreasing proportionally to E 23, and exhibits discontinuities for certain energies called absorption edges, which are due to the binding energies of electrons. These sharp step-shaped discontinuities correspond roughly to a tenfold increase of the mass attenuation coefficient just above the absorption edge. Selective absorbers take advantage of this phenomenon to attenuate strong X-ray lines occurring just above a searched line (such as copper filter for analyzing traces in gold, Fig. 5.15). The following formulae (5.4) can be used to approximate the m values in three photon energy ranges bounded by the absorption edges of the material: m ¼ 30:03 Z3:94 =AE3 ; E . EK ;
m ¼ 0:978 Z4:30 =AE3 ; EL1 , E , EK ;
m ¼ 0:78 Z3:94 =AE3 ; EM1 , E , EL3
236
ð5:4Þ
Ion beam microanalysis
As can be seen, m is approximately proportional to Z 4, and therefore low-Z materials have the best transmission for soft X-rays. This explains why materials used for the entrance windows of Si(Li) detectors are often made of low-Z compounds (e.g., Be, BN, C – H polymers). In PIXE experiments, an additional absorber is required to block particles backscattered by the sample towards the Si(Li) crystal. It is possible to take advantage of this absorber to strongly attenuate the X-rays produced by the main constituents of the sample, in order to be able to count X-rays of minor and trace elements. Table 5.2 shows the transmission of absorbers of various materials having the thickness required to stop 3 MeV protons. The Be absorber has the best transmission at low energy (extending down to Na). The Al foil is a “silicon-killer” absorber useful for analyzing trace elements in Si-containing mineral samples, but at the expense of a reduced transmission for heavier elements such as Fe. The Cu foil is a selective filter: it strongly attenuates the Au lines by a factor 4000 while preserving a 3% transmission for Fe lines. The poor transmission of 140 mm of air shows the necessity to replace it with a gas absorbing much less X-rays such as He. 5.2.2
Particle-induced X-ray emission
PIXE is by far the most used IBA technique in the field of cultural heritage. For this reason it will receive a more detailed description than the other ones. Many of its features are common to other X-ray emission techniques (Chapter 4). 5.2.2.1 Physical principle PIXE is a two-step atomic process involving (1) an inner shell ionization of the target atom by the incoming ion; (2) the filling of the subsequent electronic vacancy by an outer shell electron and the release of excess energy by emission of a characteristic X-ray. The energy E of the emitted X-ray is like in other methods such as XRF or SEM – EDX related to the Z of the material through Moseley’s law (5.5) E ¼ CðZ 2 sÞ2
ð5:5Þ
where C is a different constant for each spectral series and s a shielding constant close to 1. The production rate of X-rays depends on both the energy of the impinging particle and the atomic number of the target atom. As can be seen in Fig. 5.5, the X-ray yield for Na, Si, Cu and Ba increases with the energy of the beam. But for a fixed beam energy, the X-ray yield decreases steeply with the Z of the target element. Apart from very general and rough
237
238 TABLE 5.2 Transmission of absorbers made of various materials for X-rays of Na, Al, Si, Au and Ag
Material
Thickness
Be Kaptonw Al Cu Air
100 mm 120 mm 80 mm 40 mm 140 mm
Na 1.041 keV
Al 1.487 keV
Si 1.740 keV
Fe 6.407 keV
AuL 9.77 keV
Ag 22.161 keV
5 £ 1025 – – – –
0.033 – 1.4 £ 1024 – –
0.12 4.5 £ 1025 – – –
0.96 0.82 0.13 0.03 0.71
0.99 0.94 0.54 2.5 £1024 0.91
0.996 0.995 0.946 0.402 0.991
The absorber thickness is chosen as to block 3 MeV backscattered protons.
T. Calligaro et al.
Absorber
Ion beam microanalysis
scaling laws, there is no accurate formula fitting the X-ray production for all elements at all energies. Instead, tabulated data either theoretically calculated or experimentally measured are available. In a typical PIXE experiment the incident particles used are protons having an energy of 2 – 3 MeV. The induced X-rays are collected by a solidstate detector made of lithium-drifted silicon. The lowest measurable energy by such a detector is about 1 keV and consequently all elements with Z . 11 can be detected simultaneously via either their K or L lines. Because of the very efficient X-ray production, the technique is very sensitive and the measurements very rapid (a few minutes). In addition the lower background in comparison with the electron microprobe, due to the much smaller Bremsstrahlung radiation induced by protons than by electrons, enhances the sensitivity and leads to a detection limit in the 1 – 10 ppm range (Fig. 5.4). These characteristics make PIXE suitable for trace element analysis. However, quantitative analysis of thick samples has not been an easy task because of complex matrix effects (slowing down of incident protons, absorption of X-rays, fluorescence). Several software packages have been developed to tackle this problem, among which the GUPIX program is the most widespread [15]. They provide quantification of sample composition with an accuracy better than 5%.
Fig. 5.4. Comparison between PIXE and SEM-EDX spectra of a copper target obtained, respectively, with 3 MeV protons and 20 keV electrons. Note the markedly lower background of the PIXE spectra due to the much reduced Bremsstrahlung radiation of protons.
239
T. Calligaro et al.
5.2.2.2 Optimal conditions for PIXE analysis The best sensitivity for PIXE analysis is usually obtained with a 3 MeV proton beam. Below this energy, the detection of heavy elements is hampered. As a rule of thumb, a 1 MeV decrease of the beam energy induces a 10-fold reduction in the X-ray yield for a medium-Z element (Fig. 5.5). Conversely, the use of a beam energy exceeding 3 MeV is also detrimental. The high g-ray yield from light elements induces, by Compton effect, a flat background in the X-ray spectrum that may obscure small peaks and consequently reduces the sensitivity in PIXE mode. Moreover, above 3 MeV, the background due to secondary electron Bremsstrahlung extends above 6 keV, hampering the sensitivity for elements heavier than iron.
Fig. 5.5. Variations of PIXE yield with energy for various target atoms. The yield continuously but smoothly increases with beam energy. For a fixed beam energy, the yield steeply decreases when increasing the atomic number of the target.
240
Ion beam microanalysis
The beam current used is generally in the range of 0.1 – 10 nA, depending on the sample composition and detector arrangement. Under these conditions, the power deposited on the sample is low (0.3 – 30 mW), and this is basically the origin of the non-destructiveness of the PIXE technique. Early PIXE experiments used to be conducted in two steps. First the matrix elements, supposedly composed of low-Z elements, were measured with 1 MeV protons without absorber in front of the detector window. Then the high-Z elements present at trace level were analyzed with a 3 MeV proton beam with a suitable filter whose role is to absorb X-rays from matrix elements and prevent backscattered particles from reaching the detector crystal (e.g., 50 mm aluminium). To reduce the irradiation dose on the sample and speed up the experiment, it is interesting to record in a single experiment the X-rays emitted by both the major components and trace elements. One possibility is to use a filter drilled with a pinhole ( funny filter), the hole providing a reduced solid angle for the X-rays of major constituents, while the X-rays from trace elements are collected with the full efficiency. The second solution, now widespread, relies on the use of two X-ray detectors to record simultaneously and separately the spectra of major and trace elements in a single experiment at 3 MeV (Fig. 5.6). The detector dedicated to
Fig. 5.6. Layout of double X-ray detector external beam set-up. On top, a detector flushed with helium is dedicated to the measurement of low-energy X-rays (usually corresponding to light elements of the matrix). Bottom, a detector equipped with a filter dedicated to the measurement of minor and trace elements. Note the beam dose monitoring by the signal emitted by the exit window.
241
T. Calligaro et al.
light elements, without absorber, has to be equipped with a magnet to deflect the trajectories of backscattered ions. This set-up can be optimized for most materials by choosing a combination of adequate filters and sample to detector distances [16,17]. 5.2.2.3 Quantification of sample composition Quantitative analysis requires to establish a reliable correlation between the number of counts NZ in the X-ray line of element Z under given experimental conditions and the weight concentration CZ, of this element. The simplest situation occurs for a thin target for which both the energy loss of the incident protons (and the subsequent variation of the ionization probability) and the absorption of the X-rays of interest are negligible, corresponding to a thickness of less than 1 mm in light matrices. Thin targets The relationship between NZ and CZ is straightforward NZ ¼ kZ QrxCZ
ð5:6Þ
where kZ is a characteristic parameter depending on the experimental arrangement, Q the number of incident protons, r the specific gravity and x the thickness of the material. The kZ parameters can be calculated theoretically and therefore this method should be absolute and should not require the use of any calibration standard. With this approach, the accuracy can be better than 10% when using K lines due to the accurate knowledge of K shell ionization probability, but poorer with L lines. Conversely, this method can be used to get the kZ coefficients versus Z (yield calibration curve of the analytical system) by running a series of thin samples, monoelemental or compounds. Once this calibration curve has been established, it is valid as long as long as the analytical system is not modified, and can be subsequently applied to a sample of unknown composition. Upon a new experiment, one or two standards are sufficient to check the constancy of the system. If a small shift is discovered, then the new results can be used for the normalization of the entire curve. Thick targets Two phenomena intervene to complicate the counts/concentration relationship for a thick target. First, the incident protons progressively slow down within the sample from their initial energy E0 to zero at the end of their range. Consequently the X-ray production rate regularly decreases as the
242
Ion beam microanalysis
particle penetrates the sample. In addition X-rays produced in depth are attenuated by the over-layer of material. The yield YZ is the conversion factor between counts and concentration for element Z for a bulk target. It is expressed in counts per unit concentration and permits to retrieve the concentration CZ of element Z in the sample: CZ ¼
NZ YZ
ð5:7Þ
The determination of YZ is more complex than the determination of the kZ parameters for a thin target, since it depends on the bulk composition of the sample, which is a priori unknown. The problem is solved by iteration after a first guess of the matrix composition. The YZ values are calculated by dividing the sample into successive thin layers and adding their contribution to the global X-ray production. These complex matrix effects are taken into account by the computer programs used for spectrum processing, such as the GUPIX code. Three methods have been developed to minimize the influence of uncertainty in the estimation of the YZ values on the final sample composition. Internal standard If one element of atomic number Z0 present in the spectrum, has a known concentration CZ0 ; the concentrations of all other elements can be deduced from the relation CZ ¼ CZ0
N Z Y Z0 N Z0 Y Z
ð5:8Þ
where NZ and NZ0 are the peak areas for elements Z and Z0 : In the ratio of the yields Y the beam dose Q and all geometrical factors vanish, and the matrix effects are those of the same matrix for different elements. However, it remains the ratio of the values of the detector efficiency for different elements. This procedure can be applied to samples which have been previously analyzed with a complementary technique (e.g., XRF, SEM – EDX, ICP-MS), which is seldom the case. External standard This approach consists in using an external standard (NIST, USGS, or a sample previously analyzed) and comparing its PIXE spectrum to that of the sample. The concentration of element Z in the unknown sample is related to
243
T. Calligaro et al.
that in the standard by the following formula: NZ YZstd CZ ¼ Cstd ð5:9Þ Z NZstd YZ This time, all the instrumental parameters vanish from the yield ratio. The main drawback of this method is the need to accurately know the proton dose Q received by both the standard and the sample. In order to apply this method to Art and Archaeology, it has been necessary to develop a reliable technique of beam monitoring in external beam mode. Summing up the components to 100% The concentrations of all components in the sample satisfy the relation: ! NZ X YZ CZ ¼ CZ ¼ 100% ð5:10Þ ! which implies X N Z0 Z Y Z0 Z0 Such a constraint substitutes to the measurement of Q; which is quite advantageous with an external beam where charge integration is unfeasible. It is only applicable to samples for which all constituent elements appear in the spectrum. This is, e.g., the case of metallic objects, made of bronze, brass or gold. It can be extended to samples based on oxides (stones, ceramics, etc.) by assuming that the sum of oxides is 100%. 5.2.3
Elastic scattering of particles
5.2.3.1 Rutherford backscattering spectrometry RBS, the second type of IBA technique, relies on the Coulomb (electrostatic) interaction between the incident ion and the nucleus of the target atom. At a given scattering angle (typically in the range 1508 –1708 with respect to the beam direction), the energy of the elastically scattered ions is related to the mass of the target nucleus via a parameter called the kinematic factor K 92 8 1=2 > 2 2 > > þM1 cos u > < M2 2 M1 sin u = E
K¼ ¼ ð5:11Þ > > E0 M1 þ M2 > > : ; where E0 and E are the energies of the incident particle before and after scattering, M1 and M2 the mass of the incident particle and target atom and u the scattering angle. Ideally the scattering probability follows the Rutherford law, i.e., is proportional to Z2 E22 ½sinðu=2Þ24 ; where Z is the
244
Ion beam microanalysis
atomic number of the target atom. The Z 2 dependence makes this technique appropriate to the analysis of intermediate or heavy elements in a light matrix. But for very heavy elements, the K factors are so close that it is difficult to distinguish atoms with neighbouring mass. The preferred incident particle is a 3 MeV He ion, since it allows to separate target atom masses M2 up to around 50, and the backscattering rate follows closely the Rutherford formula. Backscattered ions are collected using a solid-state Si diode with a thickness of a few hundred of micrometers usually placed at backward angle of 1658. As shown in Fig. 5.7, the spectrum yielded by a thick target exhibits a particular shape with successive steps having their edge at characteristic energies and their height proportional to the atomic concentration of the corresponding element. So RBS technique allows to identify the atoms constituting the sample through their mass M2. The spectrum contains intrinsically an information on the depth distribution of the constituent elements because of the energy loss of the incident ion on the inward path and of the scattered ion on the outward path. This feature permits to reconstruct the corresponding elemental
Fig. 5.7. RBS spectrum taken on a lead glass showing the successive steps corresponding to the major constituents. The experimental data have been fitted using the RUMP program.
245
T. Calligaro et al.
depth profile, a task performed with computer programs among which RUMP is the most popular [16,17]. 5.2.3.2 Non-Rutherford elastic scattering The scattering of protons on light target nuclei and to a lesser extent that of alpha particles might depart from the Rutherford law, as the interaction is not purely electrostatic but involves nuclear forces also. The backscattering yield can exhibit strong resonances at given energy values where it can gain an order of magnitude compared to Rutherford law. This enhanced diffusion yield can be used to measure the content of the corresponding light element in a target containing heavy elements in spite of the strong background due to the latter. The most common application is the measurement of oxygen content by means of the resonant scattering of alpha particles on 16O occurring at 3.045 MeV. 5.2.3.3 Elastic recoil detection analysis ERDA is an alternative technique for the measurement of the depth profiles of light elements in solids. It relies on the same basic interaction as RBS, namely elastic scattering, the only difference being that the incident ion must have a greater mass than that of the target atom. The light atom recoils in a forward direction and, if the sample is tilted in a grazing geometry, can escape and be detected. The energy of the scattered ion provides information of its original depth. It is usually necessary to cover the solid-state detector with a thin absorber to prevent the detection of the scattered part of the beam. In a conventional ERDA experiment, a 3 MeV He beam impinges on the sample with a 158 angle to the surface and the detector is placed symmetrically at 158 of the sample surface, and a 10 mm Mylar absorber is placed in front of the charged particle detector (Fig. 5.8). This technique is currently used for depth profiling hydrogen in various materials. However, due to the need of a planar surface for the target, it has not yet been, to our knowledge, applied to items relevant to art and archaeology. 5.2.4
Nuclear reaction analysis
5.2.4.1 Principle of nuclear reaction analysis NRA is based on nuclear reactions induced by light ions with target atoms. For instance a nuclear reaction occurs when an impinging proton penetrates the nucleus of a fluorine atom, resulting in a nucleus of 20Ne in an excited state. This compound nucleus very quickly decays into an
246
Ion beam microanalysis
Fig. 5.8. Layout of a typical ERDA arrangement. The incident beam impinges on the sample with a 158 angle relative to the sample surface and the detector is placed symmetrically at 158. An appropriate absorber (10 mm Mylar for 3 MeV He) is placed in front of the detector to stop ions of the beam elastically scattered by the target.
16
O nucleus, an a-particle and a g-ray of 7 MeV. So the nuclear reaction can be written as p þ 19 F ! 20 Nep ! 16 O þ a þ g
ð5:12Þ
and is commonly abbreviated as 19F(p,ag)16O. In the energy range provided by small accelerators, such reactions are restricted to p, d, 3He or 4He impinging on light target nuclei, because the electrostatic repulsion (Coulomb barrier) of heavier nuclei prevents the ion to come close enough to the nucleus in order to induce the reaction. The detected reaction product, either photon or secondary ion, is characteristic of the target nucleus, thus providing a high selectivity (even at the isotopic level) to the technique. Two varieties can be distinguished according to the type of reaction product used for analysis. The techniques based on ion– gamma reactions are known as PIGE and are often complementary to PIXE. Those relying on ion – ion reactions can provide information on the depth distribution of light elements via energy loss depth profiling and are thus complementary to RBS for the measurement of light elements. Of special interest are the resonant
247
T. Calligaro et al.
reactions, which exhibit a sharp increase in their yield for a given energy and can be used for depth profiling by stepwise increasing of the incident energy. It should be mentioned that the use of narrow resonances for elemental depth profiling is hardly practicable in air because of the unavoidable energy straggling which considerably deteriorates the depth resolution. 5.2.4.2 Ion – gamma reactions Non-resonant reactions: PIGE With the same 3 MeV proton beam used for PIXE, the PIGE technique can simultaneously provide the lithium, beryllium, boron and fluorine content (Fig. 5.9). This is due to the high g-ray production rate for these specific elements (see Table 5.3). The number of g-rays produced is still lower than the number of X-rays, but this is partly balanced by an improved detection efficiency due to a much higher volume and solid angle of the g-ray detectors. The g-rays can be collected with either a high-purity germanium detector when a good energy resolution is needed or with sodium iodide or bismuth germanate crystals when a good efficiency for high-energy g-rays is required.
Fig. 5.9. PIGE spectrum taken on an emerald (Be3Al2Si6O18). Note the Compton broad peak on the left of strong lines and the escape peaks for Be line due to the pair creation process.
248
Ion beam microanalysis TABLE 5.3 Selected non-resonant reactions used in PIGE for the bulk analysis of light elements with proton and deuteron beams Element
Reaction
Eg (keV)
3.5 MeV protons 7 Lithium Li(p,p0 g)7Li Beryllium Fluorine
9
6
19
F(p,p g) F
Sodium
23
Na(p,p0 g)23Na
Be(p,ag) Li 0
19
Yield/mC/Sr
Applications
478
9 £ 106
Emeralds
3562 197
6
2.5 £ 10 3 £ 106
Emeralds Bone
439
9.6 £ 106
Glass
1779
1.2 £ 10
6
Glass
6
Bronze, gold
Silicon
28
Copper
28
0
28
Cu(p,p g) Cu
152
2.3 £ 10
Silver
28
0
28
Ag(p,p g) Ag
309
1.1 £ 105
Gold
Gold
28
0
28
279
6 £ 10
4
Gold
1.8 MeV deuterons 12 Carbon C(d,pg)13C 14 Nitrogen N(d,pg)15N 16 Oxygen O(d,pg)17O 32 Sulphur S(d,pg)33S
3089 7301 871 841
1.5 £ 107 1.2 £ 107 1.1 £ 107 2.4 £ 105
Si(p,p0 g)28Si
Au(p,p g) Au
Copper Copper Copper Copper
alloys alloys alloys and gold alloys
Data taken from Ref. [3].
Under ideal conditions the count rates in the PIXE and PIGE channels can be brought to the same level and the two spectra can be collected during the same experiment. Some elements like sodium, aluminium and silicon can be measured by both PIXE and PIGE. Because g-rays are not attenuated in the sample, the PIGE signal comes from a much greater depth than for PIXE. Due to the low production yields, oxygen, carbon and nitrogen are hardly determined using a 3 MeV proton beam, but this can be overcome with the use of a deuteron beam, as it produces a strong PIGE emission for these light elements among others [18]. Unfortunately sample can be left activated after irradiation with deuterons and the neutron flux produced by (d,n) reactions imposes some radiation shields for the facility and may damage the electronics in the vicinity of the beam impact (detectors, video cameras, etc.).
Resonant reactions Some ion– gamma reactions can also exhibit sharp resonances at specific energy value, a feature that can be usefully applied to determine elemental depth profiles by a stepwise increase of the beam energy (Table 5.4).
249
250 TABLE 5.4 Resonant ion –gamma nuclear reactions used for the determination of depth profiles for some elements Reaction
Er (MeV)
Eg (MeV)
DR (mm)
R (mm)
Applications
H H F Na S
1
6.385 16.2 0.872 1.01 3.094
4.44 6– 7 6– 7 1.632 2.23
0.004 0.08 0.1 0.1
2–3
Quartz [75], obsidian dating [74] Glass weathering [70] Flint dating [76] Glass weathering [68] Bronze patina
H(15N,ag)12C H(19F,ag)16O 19 F(p,ag)16O 23 Na(p,ag)20Ne 1
32
S(p,p0 g)32S
1.4 0.5
Er is the energy of the resonance; Eg the energy of the emitted g-ray; DR the depth resolution at the sample surface; R the maximum probing depth in a silica matrix.
T. Calligaro et al.
Element
Ion beam microanalysis
5.2.4.3 Ion – ion reactions These reactions can be considered as the counterpart of RBS for light target nuclei. Indeed, the secondary product being a charged particle, the corresponding spectrum contains some depth information. The appropriate nuclear reactions are exo-energetic and therefore communicate to the outgoing product, an important energy. Consequently the elastically scattered part of the analyzing beam can be stopped by a suitable absorber without altering significantly the interesting signal. Most of these reactions use deuteron beams, because they provide the highest yield. The corresponding spectra are fitted with specific programs such as SIMNRA [19]. Table 5.5 gives the specifications of some resonant ion –ion nuclear reactions used for the study of cultural heritage. 5.3
5.3.1
SPECIFIC ARRANGEMENTS FOR THE STUDY OF ART AND ARCHAEOLOGICAL OBJECTS External beams
The PIXE yields vary smoothly with the energy of the incident particle (Fig. 5.5). Accordingly, this technique is not too demanding on the energy definition of the beam and can be operated at atmospheric pressure. The passage of the beam through the exit foil necessary to keep a good vacuum in the beam lines and the few millimetre gas path towards the target induce an acceptable energy loss (a few tens of keV), energy straggling of a few tens of keV, and lateral spread of the beam of a few tens of microns. Usually the air is replaced with helium to reduce the beam degradation, either in a closed arrangement or more conveniently as a gas flow. An additional advantage is that the very low X-ray mass attenuation coefficient of helium allows the detection of elements down to sodium. TABLE 5.5 Ion–ion nuclear reactions used for the determination of depth profiles of C, N, O and S Element
Reaction
E
DR (mm)
R (mm)
Applications
C N O S Na
12
2.0 2.5 1.6 2.0 0.592
0.5 0.7 0.15
4.0 15 7
0.015
0.7
Bronze, gold surfaces Bronze patina Bronze patina Gold soldering Glass weathering [68]
13
C(d,p) C N(d,p)15N 16 O(d,p)17O 32 S(d,p)33S 23 Na(p,a)20Ne 14
The last three reactions are of resonant type. E is the energy of incident ion; DR the depth resolution; R the maximum probing depth in the relevant material.
251
T. Calligaro et al.
The direct operation in air with the external beam provides several benefits. First, it is a non-invasive technique, which means that the analysis of objects is carried out directly on the artefact without any sampling. Items with large size, complex shapes or whose placement in vacuum are precluded are easily handled. Secondly, it is a non-destructive technique, as the sample is left unchanged after analysis. The surrounding atmosphere favours the cooling of the sample and the evacuation of electric charges, avoiding the coating of the sample with a conductive material. All these advantages explain the increase in numerous external beam set-ups in the field of cultural heritage [20]. The only serious drawback of the external beams is the difficult measurement of the beam dose, which is a critical point for an accurate quantitative analysis. Figure 5.10 shows a typical external beam set-up.
Fig. 5.10. View of the AGLAE external beam set-up.
252
Ion beam microanalysis
Recently the external beam technology has been extended to other ion beam techniques: RBS, NRA and even ERDA. With an optimized layout, RBS can be performed with specifications reaching what is obtained in vacuum. For deuteron-based NRA the presence of air or helium is not a problem as usually a thick absorber is anyhow needed to stop backscattered particles of the beam. Table 5.6 gives some exit foil materials and thicknesses used in various ion beam experiments, ranging from a 0.1 mm thick Si3N4 foil to a 10 mm thick aluminium one. 5.3.2
Nuclear microprobes
The beam produced by an accelerator has usually a diameter of a few millimetres. In order to be able to analyze small details from the sample, a smaller probe size is needed. The first attempt to reduce the beam diameter was achieved by letting the beam through a pinhole diaphragm. Such collimated early nuclear microprobes were used for the study of inks and papers of Gutenberg’s Bible [21] or Galileo’s manuscripts [22]. But these systems suffered from a limited beam resolution (,100 mm) and a low beam intensity (,0.1 nA), enabling only the PIXE technique. In the 1980s nuclear microprobes based on the focusing of the ion beam became widely available. Such a microprobe line is composed of a series of magnetic lenses (doublet, triplet or quadruplet of magnetic quadrupoles) de-magnifying the image of a rectangular diaphragm. The sample is placed in vacuum, and the beam can routinely be focused to a micron-size level. The beam position is rasterscanned on the sample surface with magnetic deflectors, enabling acquisition of elemental maps. Its principle is rather similar to a scanning electron microprobe, but it offers a better sensitivity with PIXE and the ability to detect light elements such as lithium or beryllium with PIGE. Despite a clear success in many fields such as geology or biology, the nuclear microprobes have been scarcely used in art and archaeology. This probably stems from the TABLE 5.6 Specifications of various foils used to extract the beam in air Material
t (mm) DE (keV) sE (keV) su (mm/mm) B (mm) Use
Aluminium 10 Kaptonw 8 Zirconium 2 Si3N4 0.1
233 129 73 8
15.9 9.0 7.3 5
14 6 15 0
150 500 200 10
PIXE-RBS PIXE PIGE RBS d-NRA All
t indicates the foil thickness; DE the energy loss; sE the energy spread (straggling); su the angular spread for 3 MeV protons; B the beam diameter on the target.
253
T. Calligaro et al.
need of sampling and sample preparation. Once a micro-sample is picked up, many other competing techniques can be used, like the analytical scanning electron microscope (SEM – EDX), which is a routine tool for the study of works of art. Moreover, with the high beam density at the beam impact (nA/ mm2), the risk of sample damage is increased. Nevertheless, in vacuo microprobes were used for the study of small artefacts or prepared samples (ceramic, glaze and painting cross-sections). The recent advent of microprobe operated at atmospheric pressure has boosted applications to cultural heritage. Such systems rely on the use of extremely thin foils able to stand the pressure difference between the atmosphere and the vacuum of the beam line and resistant to beam damage (0.1 mm Si3N4). With such foils, the beam is extracted in air with a 10 mm diameter, allowing direct observation of very small details on a whole object [23]. This somewhat worsening of the probe size compared to the resolution obtained in vacuum systems is not too detrimental: it matches the particle range in thick targets, yielding the analysis of a 10 £ 10 £ 10 mm3 volume. As pointed out by Swann [24,25], milli-probes with a diameter of 20 –50 mm are appropriate for most of the studies in the cultural heritage field. Figure 5.11 gives the general layout of an external microbeam line.
Fig. 5.11. Layout of the AGLAE external microprobe. (1) exit nozzle; (2) Si(Li) detector for 1– 15 keV range; (3) Si(Li) detector for 5– 40 keV range; (4) charged particle detector; (5) HPGe detector for gamma-rays; (6) Peltier-cooled X-ray detector for monitoring the beam dose using the PIXE signal emitted by the Si3N4 exit foil.
254
Ion beam microanalysis
5.3.3
Micro and macro-imaging
The capability to build elemental maps is an important point for the study of works of art and archaeology. The comparison between the shapes and colours from the visible image and elemental maps of the same region is usually very informative. The elemental images are obtained in a similar way as in SEM – EDX, by a raster scan of the beam on the sample surface. Two separate scanning modes are available (Fig. 5.12). The first one, based on the use of the scanning coils provided with the original nuclear microprobe, permits to draw elemental micro-maps over an area of 0.5 £ 0.5 mm2 at maximum, limited by the dimensions of the beam exit window. The second mode is a mechanical scan using high-precision stepping motors fixed on the sample holder. With such a system, quite large areas (up to several tens of centimetres in width and height) can be scanned, a feature that obviously cannot be attained by conventional microprobes and could be a good argument for the in-air microprobe. Although the nuclear microprobe has a poorer lateral resolution than the SEM – EDX, it combines two noticeable advantages: a possible direct operation in air and a much improved detection limit. Early applications involved conventional operation of the microprobe under vacuum, but the trend is now to use external microbeams for elemental mapping. As examples of high-resolution micromapping performed under vacuum, we can mention the study of glaze/fabric interface of ceramics [26] and the investigation of usewear on prehistoric tools [27]. Imaging of soldering areas of jewels has been obtained in air to understand the goldsmith techniques used in ancient jewellery [28]. Concerning graphic documents, many works have been carried out with large scans in air such as the reconstruction of worn out characters in ancient papyrus [29], the study of the alteration of ancient manuscripts written with iron –gall ink [30] or the identification of the palette used for an Egyptian polychrome papyrus [31].
5.3.4
Portable systems
The availability of X-ray detectors operating without liquid nitrogen (Peltier cooled or room temperature such as silicon-drift, silicon-PIN or HgI2) has opened new perspectives for the development of portable IBA systems. With suitable radioactive sources replacing the accelerator on the field IBA or at least PIXE is conceivable. A remarkable example of a highly portable system is the alpha-proton-X-ray spectrometer (APXS) constructed for the Mars
255
256 T. Calligaro et al. Fig. 5.12. Elemental maps at different scales showing the capabilities of the external nuclear microprobe. The left four images are elemental maps and visual aspects under daylight and UV light of an altered manuscript showing the migration of Ca and Fe around the characters written using an iron –gall ink. The scanned area is 10 £ 10 mm2. On the left are elemental maps of mineral inclusions in a natural ruby that was used to fingerprint the provenance of ancient rubies. The 100 £ 100 mm2 scanned area shows a combination of ilmenite FeTiO3, zircon ZrSiO4 and calcite CaCO3 inclusions.
Ion beam microanalysis
Pathfinder [32]. In 1997, this system provided a complete and detailed chemical elemental analysis of Martian soil and rocks near the landing site. The lightweight spectrometer (0.5 kg) used a set of nine 244Cm sources (5.8 MeV a-particles). For comparison, the flux of a-particles emitted at contact by the 50 mCi sources is equivalent to a 0.15 nA beam produced by an accelerator. Elements from Na onwards were measured by PIXE using a Peltier-cooled Si-PIN diode, carbon and oxygen by RBS and selected elements (F, Na, Mg, Al, Si) using (a,p) nuclear reactions using a telescope Si-detector. In the field of art and archaeology an similar system has been developed by Pappalardo et al. [33] for the in situ study of pigments in pottery. It was based on an annular 210Po source of 1 mCi, which is a pure a-emitter (sources emitting simultaneously g-rays are expected to hamper the PIXE sensitivity), which is equivalent to 1.25 MeV protons for PIXE. The difficulty lies in the ion source, being much weaker than a conventional beam and having a short lifetime (138 days for 210Po). The challenge is therefore to design very efficient irradiation and detection geometries and to produce a reliable pinpoint radioactive source of high intensity complying with the radiation safety regulations. 5.4
APPLICATIONS IN THE FIELD OF ART AND ARCHAEOLOGY
PIXE is by far the most used IBA technique in the field of art and archaeology, sometimes coupled with PIGE. Its popularity stems from its easy implementation in air and its high sensitivity, which permits the measurements of trace elements, highly significant in archaeology. In contrast, IBA techniques based on the detection of charged particles (RBS, ERDA or NRA) have been very rarely applied because until recently they could only be implemented under vacuum to avoid detrimental energy straggling. In the following sections, we will give some examples of applications taken from recent literature and classified according to archaeological issues of increasing complexity, namely materials’ identification, provenance, alteration phenomena and indirect dating. 5.4.1
Materials’ identification
Identification of the material constituting an artefact is a key issue for the study of cultural heritage. In the field of archaeology, it reveals ancient material usage, technology and skills, such as glass recipes or metal alloys. In the field of art history, this identification permits the characterization of
257
T. Calligaro et al.
artistic techniques, such as the pigments used in a painter’s palette. It also allows the identification of restored portions of work of arts and sometimes can provide an indirect dating and the detection of forgeries by the presence of anachronistic materials. With the ion beam techniques, identification is achieved by means of the chemical composition. For instance, the yellow orpiment pigment (As2S3) or an emerald (Be3Al2Si6O18) is recognized by the specific chemical formula. Among the various IBA methods, PIXE and PIGE are the most efficient techniques for the determination of the bulk target material constituents. The elements ranging from sodium to the end of the periodic table are quantified by PIXE, while lighter elements are given by PIGE. 5.4.1.1 Homogeneous samples: single layers, bulk materials The simplest situation is that of a thin layer on top of a substrate, such as manuscripts, drawings or watercolours. The inked or painted areas are considered as thin targets, so no matrix correction has to be applied and the quantification of layer composition is straightforward. The support (paper, parchment) usually composed of low-Z elements only produces a low signal and is often thin enough to let the beam go through. This reduces the risk of damage and allows the measurement of the beam current in a Faraday cup placed downstream. This makes in-air PIXE an almost unrivalled technique for the analysis of these types of samples. A vast number of applications have been performed, such as the identification of papers and inks of the Gutenberg bible [34] or the pigments used on miniatures from medieval manuscripts [35] and more recently the metal point used in Du¨rer’s drawings (Figs. 5.13 and 5.14) [36]. A slightly more complex case involves homogeneous bulk samples. One has to take into account matrix effects such as slowing down of incident beam, self-absorption of X-rays in the sample and possibly secondary fluorescence. The homogeneous criterion stating that the sample has a constant composition both laterally and in-depth is very important. Indeed, the signals for different elements are coming from different depths in the sample. One reason is that while the projectile is slowing down in the target, the PIXE production rate decreases, favouring the emission during the beginning of the path. A second reason is the self-absorption by the sample: the low-energy X-rays emitted by light elements, are only escaping from a shallow depth. As a result, in the orpiment pigment As2S3, e.g., 99% of the sulphur and arsenic X-rays are produced within a depth of 4 and 60 mm, respectively. In PIXE, the sample roughness also affects the measurements of light elements (Na –Si) as the size of the surface accidents (a few mm) is not
258
Ion beam microanalysis
Fig. 5.13. View of a drawing by Du¨rer placed in the external beam set-up.
negligible when compared to the production depth of the low-energy X-rays of those elements. Fortunately many items of cultural heritage relevance meet this condition of homogeneity, such as minerals [37,38], ceramics [39, 40] jewellery [41,42] mentioned in Chapter 11 and more generally materials that have not suffered alteration.
259
260 T. Calligaro et al. Fig. 5.14. PIXE spectrum of the metal point marks of Du¨rer’s drawing. The low-energy X-ray detector permits to identify the preparation layer made of bone white. The high-energy X-ray detector gives the main constituents of the metal point: silver, copper, zinc and lead. Presence of mercury is likely due to pollution.
Ion beam microanalysis
5.4.1.2 Samples with depth-dependent composition Many works of art or archaeological artefacts exhibit a depth-variable composition, such as paintings, glazed ceramics or objects with metallic coatings. Despite its poor profiling capability, the PIXE method has been applied either by varying the energy of the impinging beam or its incident angle. The variation of the proton energy can be obtained by either changing the acceleration voltage or keeping the latter constant and placing aluminium foils of different thicknesses in front of the beam to reduce the incident energy. The benefits of this approach is to keep the beam on the same spot for all energies and facilitate the beam monitoring based on the X-ray signal emitted by the exit window. A computer code has been developed to reconstruct the paint layer sequence from the recorded PIXE spectra [43]. With a single energy, two solutions provide some information on the elemental depth distribution. First, the relative intensity of different X-ray lines of the same element at a single energy (e.g., K/L ratio or within the L series) [44] can provide a limited information. A second solution for probing at increasing depths in the material is to change stepwise the incidence angle [45]. Similarly one can make use of the comparative results delivered by PIXE and PIGE in the same run. However, the best solution is to rely on RBS and NRA techniques that have an intrinsic profiling ability. When a sample is composed of layers of heavy elements lying on a substrate made of light elements, the most powerful method is RBS. For example, it enables to determine the nature and thickness of the gilding of items (Fig. 5.15). Using RBS it is even possible to distinguish between an alloy from a multilayer gilding and assign the layer ordering and thickness. A proton beam can be used to probe the thickest layers (up to 10 mm), while a helium beam has a shorter range but better separates heavy elements. RBS can be advantageously combined to PIXE for the determination of paint layer arrangement in paintings [46,47]. First PIXE is used to identify the pigments present in the sample. Then, by varying the energy of the proton beam, an increasing depth of the material is involved in the X-ray emission. The variation of the X-ray yield when the energy of the beam is increased from 0.5 to 3 MeV bears the information about the layer thickness and ordering. RBS is used to evaluate the possible varnish thickness and confirms the layer thickness and ordering. A similar RBS/PIXE/NRA combination has been used for the characterization of the gilding on Mesoamerican jewellery items [48] and PIXE, RBS, XPS for the study of the iridescent coating of glass [49].
261
T. Calligaro et al.
Fig. 5.15. RBS spectra of gilded objects obtained with an external beam of 3 MeV protons. The dotted spectrum corresponds to the scabbard of an historical sabre kept at the Fontainebleau palace, France. The solid line spectrum corresponds to a Khmer statuette kept at the Guimet museum for Asian art, Paris, France. Note the distinct shapes of the spectra, the first one being characteristic of massive gold –copper alloy while the second one is typical of a gilded item. Note the helium peak that can be used as a monitor.
5.4.1.3 Complementary techniques for the identification of materials Since IBA techniques are insensitive to the chemical state, chemical structure and inefficient for the determination of organic compounds the help of complementary techniques is sometimes needed to fully specify the material. Moreover, IBA has to be complemented by techniques able to probe deeper in the sample. High-energy XRF or g-ray transmission (GRT) has been employed to correct the data obtained with PIXE on altered surfaces of coins [50]. More generally, portable XRF has appeared as a useful tool for the in museum pre-selection of paintings for further studies with PIXE [51]. Raman spectrometry was used in conjunction with PIXE for the determination of pigments of Botticelli paintings [52].
262
Ion beam microanalysis
Indeed, artistic pigments such as malachite (CuCO3·Cu(OH)2) and verdigris (Cu(CH3COO)2·nCu(OH)2), anatase and rutile (both TiO2), or massicot (PbO) and red lead (Pb3O4) cannot be distinguished using PIXE, while they are easily identified using Raman spectrometry. Conversely, Raman gives a similar spectra for most of garnets used in jewellery (pyrope– almandine – spessartite family) while these gemstones are easily distinguished by PIXE on the basis of their chemical composition. Raman spectrometry was also used for the identification of mineral inclusions in garnets mounted on early Middle Age jewels, since they are located too deep in the crystal to be reached by the ion beam [53]. 5.4.2
Provenance of the materials
After the identification of the materials composing an artefact, the determination of their origin is the second important question, as it allows to trace commercial routes and cultural exchanges between ancient communities. For instance, most of the studies of pottery fabric deal with provenancing, i.e., finding the source of the clays used by the craftsman. The provenance is usually assessed by the trace element content of the sample material. For this purpose, PIXE is the best-suited technique. Its high sensitivity reaches the mg/g level for elements in the range 20 , Z , 40 with the K lines, and elements in the range 60 , Z , 90 with L lines. PIGE also exhibits a sensitivity better than 100 mg/g for some specific elements such as sodium, fluorine or lithium. Often major and trace element concentrations are combined using multivariate methods. Potsherds being a widespread archaeological material, numerous studies have been performed, ranging from Hellenistic and Byzantine ceramics [54] to Chinese pottery imported to the Islamic world [39]. The same remark holds for ancient glass. Provenance study has also been applied to precious items kept in museums. 5.4.2.1 Specific sensitivity improvements When analyzing by PIXE metallic items such as bronze or gold without special caution, the intense X-ray emission due to the main constituents hampers the detection of trace elements. This problem, related to the use of energy-dispersive detectors, is not specific to PIXE analysis and also affects XRF or SEM – EDX. The high count rates of X-ray lines from major constituents induces pile-up peaks in the region of the double of the energy and, at low energy the tailing of these strong lines masks small peaks of lighter elements. A first way to overcome this drawback is to use selective
263
T. Calligaro et al.
absorbers in front of the detector to attenuate or even completely filter out the unwanted X-rays [24]. For the analysis of gold, a 75 mm copper filter attenuates the Au La lines by a 1027 factor and permits to determine trace elements around silver (Ru, Rh, Pd, Cd, In, Sb, Sb) with a 30% transmission and a detection limit of 20 mg/g. In the case of copper-based samples, a complex filter composed of 10 mm Ni, 20 mm Co, 10 mm Fe and 10 mm Cr allowed to determine Pd and Ag with a sensitivity of 5 mg/g [55]. Finally one can consider the use of a wavelength detection system, at the expense of lower efficiency and higher cost. Another way to solve this problem is to induce the emission of X-rays from trace elements in the sample without that from the major constituents. Here a variant of the XRF is used, called PIXE – XRF [56,57]. The quasi-monochromatic emission of a secondary target bombarded with protons can selectively induce the X-ray fluorescence of elements otherwise non-measurable by PIXE, their lines being immediately inferior to the strong lines of the matrix constituents. This technique was applied to the measurement of traces in gold (Fig. 5.16): Zn was measured in gold down to 50 ppm with a Ge intermediate target [58] and Pt down to 100 ppm with an As target [59]. It should be stressed that the use of a microbeam may provide a more reliable determination of trace elements. For instance, the provenancing of gemstones relies on the measurement of trace elements homogeneously incorporated in the crystal lattice, but the presence of foreign mineral inclusions can be misleading. Using a broad beam having 1 mm diameter on a pure sapphire crystal (Al2O3) containing a 20 mm ilmenite inclusion (TiFeO3) would wrongly lead to a 500 ppm iron and titanium average content. The use of a microbeam permits to avoid such an inclusion and select a more representative area of the sample. By this way it was possible to locate the provenance of rubies inlaid in a Mesopotamian statuette [37] and emeralds set on Visigothic royal crowns [38]. 5.4.2.2 Complementary techniques for provenancing materials Often several techniques are combined with PIXE and PIGE to infer the full fingerprint of the sample material. For instance, PIXE, PIGE, XRF, SEM – EDX and LA-ICP-MS have been associated with the study of provenance and manufacture of Medieval central European glass [60]. An overview of the usefulness of chemical and isotopic characterization for the determination of provenance, fabrication process and relative dating of ancient glasses is given in Chapter 15. For high-Z materials such as metals, activation using high-energy ion beams has also proved to be very useful. For instance, trace elements in ancient coins were measured by activation with a 12 MeV proton
264
Ion beam microanalysis
Fig. 5.16. Sensitivity improvements provided by two specific set-ups for the measurement of trace elements in a gold matrix. Top is the PIXE spectrum recorded without absorber showing various artefacts: strong gold lines, sum peaks. Bottom is the spectrum recorded with a 75 mm thick copper absorber filtering out the Au lines. It allows the measurement of elements in the vicinity of the Ag lines. In the middle are the PIXE –XRF spectra recorded on pure gold and a gold sample containing 400 ppm platinum using an As intermediate target. The peak rising at 9.4 keV corresponds to the overlap of the Pt signal and the Raman scattering of the As Kb. Using this PIXE – XRF set-up, the detection limit for platinum was lowered to 50 ppm, compared to 1000 ppm at best for PIXE.
beam produced by a cyclotron [61]. Guerra et al. [62] has compared the results obtained with XRF, PIXE and charged particle activation for the characterization of ancient coinage. Sometimes the trace element fingerprint alone is not sufficient, and a complementary technique has to be used. The combination of the age of the obsidian deposits obtained by the fission track dating method with the chemical data given by PIXE allowed to distinguish various deposits in Mesoamerica [63]. For gemstones, mineral inclusions are classically used to asses the provenance. Raman micro-spectrometry conveniently allows to identify these inclusions, usefully complementing trace element measurements with PIXE for the location of the sources of garnets used on Middle Age jewels [53].
265
T. Calligaro et al.
5.4.3
Alteration phenomena
As stated earlier, the surface of materials relevant to cultural heritage has frequently a distinct composition from the bulk. This feature originates from either a deliberate chemical treatment or weathering during long-term conservation or burial. The combined use of several IBA techniques permits in many cases to evaluate the extent of alteration and via artificial aging tests to decipher the involved mechanism. In this context, it has been demonstrated using m-RBS and m-PIXE depth profiling that the superficial gold enrichment in the “Tumbaga” Mesoamerican gold alloy stems from an acidic surface treatment permitting the release of most copper [64]. Similarly, the artificial patina on various archaeological objects made of copper alloys (see Chapter 10) has been extensively investigated by means of RBS and NRA techniques. The patina appears to be constituted by a stratified structure of different copper compounds (mostly oxides, carbonates and organic salts) [65,66]. The NRA technique, yielding the C, N, O depth profiles is particularly well suited for the study of patina (see Fig. 5.17). Weathering of glass is a major issue in conservation science (Chapter 16). Many analytical techniques have been employed to determine the underlying mechanism including those based on synchrotron radiation [67]. The unique ability of IBA techniques to obtain hydrogen depth profiles has been
Fig. 5.17. Combined RBS –NRA spectra of patina on historical Japanese sabre hilts kept in the Guimet museum, Paris. The hilts are made of a traditional Cu –Au alloy labelled shakudo. The RBS spectra were obtained with 3 MeV protons and the NRA spectra with 2 MeV deuterons. The RBS spectrum of shakudo 1 exhibits a superficial deficit in copper relative to shakudo 2. This is correlated with a higher oxygen content as shown by the NRA spectrum.
266
Ion beam microanalysis
exploited for many years [68]. Among the numerous works in this field, Ma¨der et al. [69] have inferred the extent of corrosion through the Na depletion evidenced by PIXE – PIGE combination. On the other hand, Kosionides et al. [70] have investigated the H, O, C and Si profiles by NRA. Graphic documents are also prone to deterioration, the main causes being the acidic hydrolysis of cellulose or the interaction of the latter with some of the graphic materials like metallo-gallic inks or pigments (Chapter 17). In the study by Remazeilles et al. [30], the large-scale mapping capability of the external nuclear microprobe has been exploited to estimate the extent of lateral diffusion of iron – gall ink on manuscripts (see Fig. 5.12). 5.4.4
Authentication and relative dating
The chronology of the introduction of most pigments being relatively well established, it is in principle possible to determine a relative date of the painted work or the restored part of it. This would also allow the detection of forgeries by the presence of anachronistic pigments. Among the few examples that can be found in the literature one can mention the case of a painting wrongly attributed to Lucas Cranach which contains Naples’ yellow, a lead antimony pigment only available after the 17th century [71]. The detection of such anachronistic components is sometimes the source of controversy such as in the case of the “Vinland Map”, the alleged earliest representation of the coast of the New World. PIXE analysis has indeed revealed the presence of titanium oxide in the contours of the map but in minute amount, which makes unlikely an intentional modern intervention [72]. The chronology of Galileo’s writings has been reconstructed on the basis of ink composition [73]. Another possibility of indirect dating is offered by the modification of the surface composition of ancient objects. One of the first attempt was performed by measuring on obsidian the hydrogen (water) penetration at the surface using 1H(15N,ag)12C resonant nuclear reaction at 6.385 MeV [74]. The high-energy g-rays (4.44 MeV) were measured using a high-efficiency NaI detector and the profiles were obtained by increasing the energy of the beam by steps of 15 keV (corresponding to 10 nm in depth). But the diffusion process of water in glass is still debated, and seems to be very dependent on humidity and temperature. A more recent attempt to measure the penetration of hydrogen in quartz was investigated using the same reaction in view to develop an authentication method for ancient artefacts such as pre-Colombian skulls [75]. Another application deals with the fluorine penetration at the surface of flint used for dating chipped tools found in an archaeological site [76]. Indeed, the fluorine present in natural
267
T. Calligaro et al.
groundwater slowly diffuses into the flint material. The shallow fluorination depth (of the order of a micrometer for a burial period of a few thousands of years) was measured using the resonant nuclear reaction 19F(p,ag)16O at 872 keV. The energy of the beam was scanned between 800 and 900 keV and the detection of the high-energy g-rays (6– 7 MeV) was performed with a bismuth germanate g-ray detector. A good correlation was found between the estimated archaeological age (4000– 900 BC ) and the penetration depth of fluorine (Fig. 5.18). 5.5
SURVEY OF WORLDWIDE IBA ACTIVITY IN THE FIELD OF CULTURAL HERITAGE
Applications of IBA to art and archaeology are constantly growing, as illustrated by the now quite regular sessions devoted to this research field in IBA-related international conferences. A large number of laboratories equipped with particle accelerators occasionally perform studies on issues connected to cultural heritage. However, from the current literature, it can be seen that only a few of them have a regular activity in this field. It is worth mentioning that until now, the facility located in the Louvre is the only one exclusively dedicated to works of art and archaeology. Most of the laboratories listed in Table 5.7 are Western European, a fact likely linked to the wealth of cultural heritage in the old continent. A network of European laboratories has been constituted in the framework of two successive COST actions (scientific and technical cooperation) of the European Commission. The aim of the first one, labelled G1 “Ion beam analysis techniques applied to art and archaeology,” which covered the period 1994 –2000, was to favour the exchange of information and arouse cooperation between participants [77]. The program involved 19 laboratories from 12 countries. In 2001, a second action G8 extended the scope to other types of non-destructive techniques of examination and analysis applied to museum objects. 5.6
CONCLUSION AND FUTURE PROSPECTS
Thanks to their outstanding analytical capability and their non-destructive character, accelerator-based analytical techniques have proved their usefulness in the study of works of art and archaeology. The development of external beam lines has been a decisive step, since this type of set-up permits in-air analysis, thus avoiding taking off even minute samples from precious items or putting them inside a vacuum chamber, a potentially detrimental operation. The PIXE technique, often combined with PIGE, is still the most
268
Ion beam microanalysis
Fig. 5.18. Fluorine depth profile obtained on flint archaeological artefacts with the resonant nuclear reaction 19F(p,ag)16O. The fluorine penetration depth appears to be correlated with the burial time of the object.
widely spread technique in this field, due to its high sensitivity and ease of implementation in air. It can be used either for the quick identification of materials via the detection of major elements, or for provenance determination using the trace element content. However, as PIXE hardly provides
269
270
TABLE 5.7 List of laboratories currently active in ion beam analysis application to cultural heritage Accelerator type
Experimental set-ups
Main research fields
Belgium Lie`ge Belgium Namur Finland Helsinki France Paris C2RMF
Cyclotron 2 MV tandem 5 MV tandem 2 MV tandem
External External External External
Germany Berlin Germany Rossendorf Greece Athens Hungary Budapest Hungary Debrecen Italy Florence Mexico UNAM Slovenia Ljubljana Spain Madrid Spain Sevilla Sweden Lund UK Oxford USA Bartol Delaware
Cyclotron 5 MV tandem 5 MV tandem 2 MV Van de Graaff 5 MV Van de Graaff 3 MV Van de Graaff 2 MV tandem 2 MV tandem 5 MV tandem 2 MV tandem 3 MV tandem 1.7 MV tandem 2 MV Van de Graaff
External beam External beam External beam External beam Microprobe External beam External beam External beam External beam External beam External beam External beam External beam
USA Tempe China Shanghai Japan Tokyo Singapore Taiwan Taipei Australia Lucas Heights
2 MV tandem 3 MV tandem 2 MV tandem 2.5 MV Van de Graaff 3 MV tandem 3 MV Van de Graaff 10 MV tandem 5 MV Van de Graaff
External beam External beam External beam Microprobe External beam
Paintings glasses Gold metallurgy Gold metallurgy paintings Ceramics stones gems metals glasses enamels glazes manuscripts paintings Metals Bones paintings drawings glasses Metals ceramics Manuscripts bronzes Paintings gemstones glasses Manuscripts ceramics Stones, pigments, jewellery Coins stones Ceramics Jewellery ceramics Manuscripts glasses Bones glazes manuscripts Gold alloys ceramics bronzes enamels glasses Ceramics Ceramics manuscripts metals Ceramics Bones gemstones Coins Ceramics obsidians
South Africa Faure
beam beam microprobe beam beam microprobe
External beam
microprobe
microprobe microprobe microprobe microprobe microprobe microprobe
Ceramics
T. Calligaro et al.
Laboratory
Ion beam microanalysis
depth information, much effort has been made to associate it with RBS and/or NRA, two methods that can yield elemental depth profiles particularly useful for inferring the conservation state of artefacts and alteration mechanisms. Recently designed external beam set-ups are capable of performing simultaneous PIXE – RBS runs on the same spot. In addition, the use of ultra-thin exit windows permits to extract microbeams with a size as low as 10 mm and to obtain elemental micro-maps with lateral resolution of about the same value, but with the possibility to scan areas much larger than the conventional nuclear microprobe operating under vacuum. In spite of these qualities, IBA techniques are not a panacea, especially because they do not provide any information on either the structure of materials or the chemical state of constituent elements. It is frequently needed to rely on complementary techniques such as X-ray diffraction or Raman spectrometry. The latter is gaining a growing popularity in the field of cultural heritage, due to the development of portable instruments and the wealth of information it yields. Moreover IBA techniques necessitate the access to medium scale facilities ordinarily not situated close to museums. In the mean time, XRF portable systems have been developed which to some extent can be a satisfactory alternative to IBA. On the other hand, analytical methods based on large-scale facilities like synchrotron rings are increasingly applied (see Chapter 4). Although ion beam techniques are now facing competing analytical tools of high performance, there remain some domains where they have absolutely no counterpart, namely the non-destructive analysis of light elements and elemental depth profiling, a specific contribution of, respectively, PIGE and RBS – NRA techniques.
REFERENCES 1 2 3 4 5
6
J.R. Bird, P. Duerden and D.J. Wilson, Ion beam techniques in archaeology and the arts, Nucl. Sci. Appl., 1 (1983) 357. L.C. Feldman and J.W. Mayer, Fundamentals of Surface and Thin Film Analysis. North Holland/Elsevier, New York, 1986. J.R. Bird, R.A, Brown, D.D. Cohen and J.S. Williams, in: J.R. Bird and J.S. Williams (Eds.), Ion Beams for Material Analysis. Academic Press, Sidney, 1989. S.A.E. Johansson, J.L. Campbell and K.G. Malmqvist, Particle Induced X-Ray Emission Spectrometry (PIXE). Wiley, New York, 1995. J.-C. Dran, T. Calligaro and J. Salomon, Particle induced X-ray Emission. In: Ciliberto and Spoto (Eds.), Modern Analytical Methods in Art and Achaeology, Chemical Analysis, Vol. 155. John Wiley and sons, New York, 2000. W.-K. Chu, J.W. Mayer and M.A. Nicolet, Backscattering Spectrometry. Academic Press, Boston, 1978.
271
T. Calligaro et al. 7 8 9 10 11
12 13
14
15 16
17 18
19 20
21 22
23
24 25
272
J. Tirira, Y. Serruys and P. Trocellier, Forward Recoil Spectrometry: Application to Hydrogen Determination in Solids. Plenum Press, New York, 1996. J.R. Tesmer and M. Nastasi, Handbook of Modern Ion Beam Analysis. Material Research Society, Pittsburg, 1995. M.B.H. Breese, D.N. Jamieson and P.J.C. King, Materials Analysis Using a Nuclear Microprobe. Wiley, New York, 1996. F. Watt and G.W. Grime, Principles and Applications of High-Energy Ion Microbeams. Adam Hilger, Bristol, 1987. J.P. Biersack and L.G. Haggmark, A Monte Carlo computer program for the transport of energetic ions in amorphous targets, Nucl. Instr. Meth., 174 (1980) 257. X. Zeng, X. Wu, Q. Shao, J. Tang and F. Yang, Radiation damage in PIXE analysis of museum paper-like objects, Nucl. Instr. Meth., B47 (1990) 143. M. Chiari, A. Migliori and P.A. Mando`, Investigation of beam-induced damage to ancient ceramics in external-PIXE measurements, Nucl. Instr. Meth., B188 (2002) 151. S. Matsuyama, H. Endo, K. Ishii, H. Yamazaki, Y. Tokai, A. Sugimoto, K. Yamamoto and T. Satoh, Radiation damage of paper samples in in-air PIXE analysis, Int. J. PIXE, 9 (1999) 47. J.A. Maxwell, J.L. Campbell and W.J. Teesdale, Nucl. Instr. Meth., B43 (1988) 21. T. Calligaro, J.D. MacArthur and J. Salomon, An improved experimental setup for the simultaneous PIXE analysis of heavy and light elements with a 3-MeV proton external beam, Nucl. Instr. Meth., B109/110 (1996) 125. L.R. Doolittle, Algorithms for the rapid simulation of Rutherford backscattering spectra, Nucl. Instr. Meth., B9 (1985) 344. A.Z. Kiss, I. Biron, T. Calligaro and J. Salomon, Thick target yields of deuteron induced gamma-ray emission from light elements, Nucl. Instr. Meth., B85 (1994) 118. M. Mayer, SIMNRA User’s Guide, Technical Report I, Max Planck Institut fu¨r Plasmaphysik, Garching, Germany, 1997, 9/113 pp. P.A. Mando`, Advantages and limitations of external beams in applications to arts & archaeology, geology and environmental problems, Nucl. Instr. Meth., B85 (1994) 815. B.H. Kusko, T.A. Cahill, R.A. Eldred and R.N. Schwab, Proton milliprobe analysis of the Gutenberg Bible, Nucl. Instr. Meth., B3 (1984) 689. P. Del Carmine, F. Lucarelli, P.A. Mando` and A. Pecchioli, The external PIXE setup for the analysis of manuscripts at the Florence University, Nucl. Instr. Meth., B75 (1993) 480. T. Calligaro, J.-C. Dran, E. Ioannidou, B. Moignard, L. Pichon and J. Salomon, Development of an external beam nuclear microprobe on the Aglae facility of the Louvre museum, Nucl. Instr. Meth., B161/163 (2000) 322. C.P. Swann and S.J. Fleming, Selective filtering in PIXE spectrometry, Nucl. Instr. Meth., B49 (1990) 65. C.P. Swann, Recent applications of nuclear microprobes to the study of art objects and archaeological artifacts, Nucl. Instr. Meth., B130 (1997) 289.
Ion beam microanalysis 26 27 28
29 30
31
32
33
34
35
36
37
38
39
40
R.A. Jarjis, Backscattering spectroscopy development for the University of Oxford Scanning External Milliprobe (SEMP), Nucl. Instr. Meth., B118 (1996) 62. Z. Smit, G.W. Grime, S. Petru and I. Ratja, Microdistribution and composition of usewear polish on prehistoric stone tools, Nucl. Instr. Meth., B150 (1999) 565. G. Demortier, D. Decroupet and S. Mathot, Performance of PIXE on nuclear microprobes for the study of gold soldering process, Nucl. Instr. Meth., B54 (1999) 346. N.E.G. Lo¨vestam and E. Swietlicki, PIXE analysis and imaging of papyrus documents, Nucl. Instr. Meth., B145 (1990) 307. C. Remazeilles, V. Quillet, T. Calligaro, J.-C. Dran, L. Pichon and J. Salomon, PIXE elemental mapping on original manuscripts with an external microbeam. Application to manuscripts damaged by iron –gall ink corrosion, Nucl. Instr. Meth., B181 (2001) 681. A.-M.B. Olsson, T. Calligaro, S. Colinart, J.-C. Dran, N.E. Lo¨vestam, B. Moignard and J. Salomon, Micro-PIXE analysis of an ancient Egyptian papyrus: identification of pigments used for the “Book of the Dead”, Nucl. Instr. Meth., B181 (2001) 707. R. Rieder, H. Wa¨nke, T. Economou and A. Turkevich, Determination of the chemical composition of the Martian soil and rocks: the alpha proton X-ray spectrometer, J. Geophys. Res., 102(E2) (1997) 4027. L. Pappalardo, A portable PIXE system for the in situ characterisation of black and red pigments in neolithic, copper age and bronze age pottery, Nucl. Instr. Meth., B150 (1999) 576. T.A. Cahill, B. Kusko, R.A. Eldred and R.N. Schwab, Gutenberg’s inks and papers: non-destructive composition analyses by proton milliprobe, Archaeometry, 26 (1984) 3. J.D. MacArthur, P. Del Carmine, F. Lucarelli and P.A. Mando`, Identification of pigments in some colours on miniatures from the medieval age and early Renaissance, Nucl. Instr. Meth., B45 (1990) 315. I. Reiche, A. Berger, A. Duval, W. Go¨rner, H. Guicharnaud, S. Merchel, M. Radtke, J. Riederer and H. Riesemeier, Non-destructive investigations of Du¨rer’s silver point drawings by PIXE and SY-XRF. Proceedings of the 7th International Conference on Non-Destructive Testing and Microanalysis for the Diagnostics and Conservation of Cultural and Environmental Heritage, ART 2002, R. Van Grieken, K. Janssens, L. Van’t Dack and G. Meersman, Eds., University of Anwerp (UIA), Antwerp-Wilrijk, Belgium, 2002. T. Calligaro, A. Mossman, J.-P. Poirot and G. Querre´, Provenance study of rubies from a Parthian statuette by PIXE analysis, Nucl. Instr. Meth., B136/138 (1998) 846. T. Calligaro, J.-C. Dran, J.-P. Poirot, G. Querre´, J. Salomon and J.-C. Zwaan, PIXE/PIGE characterisation of emeralds using an external micro-beam, Nucl. Instr. Meth., B161/163 (2000) 769. S.J. Fleming and C.P. Swann, Recent application of PIXE spectrometry in archaeology. II Characterisation of Chinese pottery imported to the Islamic world, Nucl. Instr. Meth., B64 (1992) 528. S.J. Fleming and C.P. Swann, Roman mosaic glass: a study of production processes, using PIXE spectrometry, Nucl. Instr. Meth., B150 (1999) 622.
273
T. Calligaro et al. 41 42 43 44
45
46 47 48 49
50 51
52
53
54 55 56 57
58
59
274
G. Demortier, Ion beam analysis of gold jewelry, Nucl. Instr. Meth., B64 (1992) 481. G. Demortier, Ion beam studies of archaeological gold jewellery items, Nucl. Instr. Meth., B113 (1996) 347. I. Brissaud, G. Lagarde and P. Midy, Study of multilayers by PIXE technique. Application to paintings, Nucl. Instr. Meth., B117 (1996) 179. C. Neelmeijer, I. Brissaud, T. Calligaro, G. Demortier, A. Hautoja¨rvi, M. Ma¨der, L. Martinot, M. Schreiner, T. Tuurnala and G. Weber, Paintings—a challenge for XRF and PIXE analysis, X-ray Spectrom., 29 (2000) 101. G. Weber, J.M. Delbrouck, D. Strivay, F. Kerff and L. Martinot, Use of a variable incidence angle PIXE arrangement for studying pigment multilayers, Nucl. Instr. Meth., B139 (1998) 196. C. Neelmeijer and W. Wagner, Non-destructive pigment analysis in paint layers using ion-beam methods on air, Naturwissenschaften, 81 (1994) 553. C. Neelmeijer and M. Ma¨der, The merits of particle induced X-ray emission in revealing painting techniques, Nucl. Instr. Meth., B189 (2002) 293. J.-L. Ruvalcaba-Sil and G. Demortier, Elemental concentration profile in ancient gold artifact by ion beam scattering, Nucl. Instr. Meth., B113 (1996) 275. D. Jembrih, C. Neelmeijer, M. Schreiner, M. Ma¨der, M. Ebel, R. Svagera and M. Peev, Iridescent Art Nouveau glass—IBA and XPS for the characterisation of thin iridescent layers, Nucl. Instr. Meth., B181 (2001) 698. G. Demortier, Y. Morciaux and D. Dozot, PIXE, XRF and GRT for the global investigation of ancient gold artefacts, Nucl. Instr. Meth., B150 (1999) 640. C. Neelmeijer, I. Brissaud, T. Calligaro, G. Demortier, A. Hautoja¨rvi, M. Ma¨der, L. Martinot, M. Schreiner, T. Tuurnala and G. Weber, Paintings—a challenge for XRF and PIXE analysis, X-Ray Spectrom., 29 (2000) 101. C. Andalo`, M. Bicchieri, P. Bocchini, G. Casu, G.C. Galletti, P.A. Mando`, M. Nardone, A. Sodo and M. Plossi Zappala`, The beautiful Trionfo d’Amore attributed to Botticelli: a chemical characterisation by proton-induced X-ray emission and micro-Raman spectroscopy, Anal. Chim. Acta, 429 (2001) 279. T. Calligaro, S. Colinart, J.-P. Poirot and C. Sudres, Combined external-beam PIXE and m-Raman characterisation of garnets used in Merovingian jewellery, Nucl. Instr. Meth., B189 (2002) 320. S.Y. Waksman, A. Pape and Ch. Heitz, PIXE analysis of Byzantine ceramics, Nucl. Instr. Meth., B85 (1994) 824. G. Demortier and Y. Morciaux, PIXE gadgets, Nucl. Instr. Meth., B85 (1994) 112. X. Zeng, X. Wu, H. Yao and F. Yang, PIXE-induced XRF with transmission geometry, Nucl. Instr. Meth., B75 (1993) 99. Ts. Lin, Ch. Luo and J. Chou, The (pX,X) experiment—a new approach for partial sensitivity enhancement in trace element analysis, Nucl. Instr. Meth., 151 (1978) 439. M. Peisach, Ion beam analytical methods for attempts to establish the provenance of archaeological objects in the absence of reference artefacts, Nucl. Instr. Meth., B14 (1986) 99. M.F. Guerra and T. Calligaro, Gold traces to trace gold, Archaeol. Sci., (2003) to be published.
Ion beam microanalysis 60
61 62
63
64
65
66
67
68 69
70
71
72
73 74
Z. Smit, P. Pelicon, G. Vidmar, B. Zorko, M. Budnar, G. Demortier, B. Gratuze, S. Turm, M. Neemer, P. Kump and M. Kos, Analysis of medieval glass by X-ray spectrometric methods, Nucl. Instr. Meth., B161/163 (2000) 718. P. Meyers, Non destructive activation analysis of ancient coins using charged particles and fast neutrons, Archaeometry, 11 (1969) 67. M.F. Guerra, The study of the characterization and provenance of coins and other metalwork using XRF, PIXE and activation analysis. In: D.C. Creagh and D.A. Bradley (Eds.), Radiation in Art and Archaeometry. Elsevier, Amsterdam, 2000, pp. 378 –416. L. Bellot-Gurlet, T. Calligaro, O. Dorighel, J.-C. Dran, G. Poupeau and J. Salomon, PIXE analysis and fission track dating of obsidian from South American prehispanic cultures (Colombia, Ecuador), Nucl. Instr. Meth., B150 (1999) 616. J.L. Ruvalcaba-Sil and G. Demortier, Scanning RBS –PIXE study of ancient artefacts from south America using a microbeam, Nucl. Instr. Meth., B130 (1997) 297. E. Ioannidou, D. Bourgarit, T. Calligaro, J.-C. Dran, M. Dubus, J. Salomon and P. Walter, RBS and NRA with external beams for archaeometric applications, Nucl. Instr. Meth., B161/163 (2000) 730. G. Kalliabakos, S. Kossionides, P. Misialides, C.T. Papadopoulos and R. Vlastou, Determination of sulfur and copper depth distribution in patina layers using nuclear reaction techniques, Nucl. Instr. Meth., B170 (2000) 467. K. Janssens, A. Aerts, L. Vincze, F. Adams, C. Yang, R. Utui, K. Malmqvist, K.W. Jones, M. Radke, S. Garbe, F. Lechtenberg, A. Kno¨chel and H. Wouters, Corrosion phenomena in electron, proton and synchrotron X-ray microprobe analysis of a Roman glass from Qumran, Jordan, Nucl. Instr. Meth., B109/110 (1996) 690. W.A. Lanford, Ion beam Analysis of glass surfaces: dating, authentication and conservation, Nucl. Instr. Meth., B14 (1986) 123. M. Ma¨der, D. Grambole, F. Herrmann, C. Neelmeijer, M. Schreiner and G. Woisetschla¨ger, Non-destructive evaluation of glass corrosion states, Nucl. Instr. Meth., B136/138 (1998) 863. S. Kossionides, M. Kokkoris, A.G. Karydas, T. Paradellis, G. Kordas and G. Moraitou, Analysis of ancient glass using ion beams and related techniques, Nucl. Instr. Meth., B195 (2002) 408. T. Tuurnala and A. Hautoja¨rvi, Original or forgery—pigment analysis of paintings using ion beams and ionising radiation. In: G. Demortier and A. Adriaens (Eds.), Ion Beam Study of Art and Archaeological Objects, a contribution by members of the COST G1 Action, European Commission, EUR 19218, 2000. T.A. Cahill, R.N. Schwab, B.H. Kusko, R.A. Eldred, G. Mo¨ller, D. Dutschke and D.L. Wick, The Vinland Map, revisited: new compositional evidence on inks and parchment, Anal. Chem., 59 (1987) 829. L. Giuntini, F. Lucarelli, P.A. Mando, W. Hooper and P.H. Barker, Galileo writings: chronology by PIXE, Nucl. Instr. Meth., B95 (1995) 389. T. Laursen and W.A. Lanford, Hydration of obsidian, Nature, 276 (1978) 153.
275
T. Calligaro et al. 75 76
77
276
O. Dersch and F. Rauch, Water uptake of quartz investigated by means of ionbeam analysis, Fresenius J. Anal. Chem., 365 (1999) 114. P. Walter, M. Menu and J.-C. Dran, Dating of archaeological flints by fluorine depth profiling: new insights in the mechanism of fluorine uptake, Nucl. Instr. Meth., B64 (1992) 494. G. Demortier, and A. Adriaens (Eds.), Ion Beam Study of Art and Archaeological Objects, a contribution by members of the COST G1 Action, European Commission, EUR 19218, 2000.
Chapter 6
X-ray photoelectron and Auger electron spectroscopy Annick Hubin and Herman Terryn
6.1
INTRODUCTION
In X-ray photoelectron spectroscopy (XPS) and Auger electron spectroscopy (AES), electrons emitted after the interaction between primary X-rays or electrons and a sample are detected. The interaction is illustrated in Fig. 6.1 for AES. The amount of electrons having escaped from the sample without energy loss is typically measured in the range of 20 – 2000 eV. The data is represented as a graph of intensity versus electron energy. Due to the impact of the primary beam, the atoms in the sample are ionized and electrons are liberated from the surface, either as a result of the photoemission process (XPS) or of the radiationless de-excitation in the Auger electron emission process (AES). Although XPS and AES are comparable methods in the sense that both are based on the use of a spectrometer to measure electrons of relatively low energy, the main difference between the two methods consists in the source of the primary radiation, which is necessary to provoke ionization of the atoms. AES makes use of an electron gun while XPS relies on soft X-rays. As a consequence of that, one of the main differences is the lateral resolution of the two methods. Since there is a continuous evolution in the design of the equipment and the performance of the techniques, it is difficult to express an absolute value for it, but for AES the lateral resolution typically is situated in the 10 – 100 nm range, while by means of XPS only a lateral resolution of a few to 100 mm can be reached. In both methods low energy electrons are measured, giving rise to comparable depth and sensitivity values, which are, respectively, in the order of nanometres (see Fig. 6.1) and of about 0.1% atomic concentration. This type of measurement is necessarily performed under high vacuum conditions, and only samples restricted in size can be analysed. Comprehensive Analytical Chemistry XLII Janssens and Van Grieken (Eds.) q 2004 Elsevier B.V. All rights reserved
277
A. Hubin and H. Terryn
Fig. 6.1. Schematic representation of the primary beam– sample interaction in the case of an AES analysis.
From this point of view, XPS and AES cannot be considered as nondestructive techniques, although the analyses themselves are not destructive in nature. On the other hand, thanks to their spatial resolution, a small amount of material suffices for the analysis. Moreover, samples can be analysed with XPS and AES in the as-received condition, and can in many cases be put back in their original position afterwards. In this chapter, we try to give practical information on XPS and AES, so that the reader can understand the basic principles, the characteristics and the potential of the methods and their instrumentation. More detailed information may be found in dedicated books. Strongly recommended are, e.g., the books of Briggs and Seah [1], Carlson [2], Watts [3], Agius et al. [4], Thompson et al. [5] and Nefedov [6]. More practical information, e.g., spectra and data, is given in Refs. [7 –9]. We can also refer to the proceedings of important international conferences such as ECASIA “European Conference for Application of Surface and Interface Analysis” and QSA “Quantitative Surface Analysis.” Many papers on XPS and AES can be found in journals. “Surface Science” and “Applied Surface Science,” e.g., mostly communicate about the fundamental physical background of the XPS and AES analyses while, e.g., “Surface and Interface Analysis” and “Journal of the Vacuum Science and Technology” report more on the applications.
278
X-ray photoelectron and Auger electron spectroscopy
At the end of this chapter, a paragraph is dedicated to a literature overview of the use of XPS and AES for the analysis of cultural heritage materials. 6.2 6.2.1
THE BASIC CONCEPTS OF XPS AND AES Principle of X-ray photoelectron spectroscopy
In the case of XPS, electrons are liberated from the specimen as a result of a photoemission process. An electron is ejected from an atomic energy level by an X-ray photon, mostly from an Al-Ka or Mg-Ka primary source, and its energy is analysed by the spectrometer. The XPS process is schematically represented in Fig. 6.2 for the emission of an electron from the 1s shell of an atom. The experimental quantity that is measured is the kinetic energy of the electron, which depends on the energy hn of the primary X-ray source. The characteristic parameter for the electron is its binding energy. The relation between these parameters is given by Eq. (6.1) EB ¼ hn 2 EK 2 W
ð6:1Þ
where EB and EK are, respectively, the binding and the kinetic energies of the emitted photoelectron, hn the photon energy, and W the spectrometer work function. In a first approximation, the work function is the difference between the energy of the Fermi level EF and the energy of the vacuum level
Fig. 6.2. Schematic representation of the XPS process.
279
A. Hubin and H. Terryn
EV , which is the zero point of the electron energy scale: W ¼ EF 2 EV
ð6:2Þ
This quantity is to be determined by calibration for the spectrometer used. From Eq. (6.1) it is clear that only binding energies lower than the exciting radiation (1486.6 eV for Al-Ka and 1253.6 eV for Mg-Ka) are probed. Each element has a characteristic electronic structure and thus a characteristic XPS spectrum. Figure 6.3 shows the XPS spectrum of Ag. In the spectrum, a number of peaks appears on a background. The background originates from photoelectrons, which undergo energy changes between photoemission from the atom and detection in the spectrometer. The observed peaks can be grouped into three types: peaks originating from photoemission (i) from core levels, (ii) from valence levels at low binding energies (0– 20 eV) and (iii) from X-ray excited Auger emission (between 1100 and 1200 eV).
Fig. 6.3. XPS spectrum of Ag. Conditions: Al-Ka at 350 W, pass energy ¼ 58.7 eV.
280
X-ray photoelectron and Auger electron spectroscopy
Although valence level spectra have their analytical value (e.g., in the study of the electronic structure of materials), they contribute little to the analysis of cultural heritage materials, and will not be discussed here. The main information comes from the core level peaks and the Auger peaks. As can be seen in Fig. 6.3, the notation for XPS and AES peaks is different. The nomenclature employed to describe XPS and AES features is based on the momenta associated with the orbiting paths of electrons around atomic nuclei, indicated by the quantum numbers n; l; j and yet the translation into the notation is different for both techniques. XPS uses the spectroscopic notation: first the principal quantum number ðn ¼ 1; 2; 3; …Þ; then l ¼ 0; 1; 2; … indicated as s, p, d,…, respectively, and finally the j value given as a suffix ð1=2; 3=2; 5=2; …Þ: The AES nomenclature on the contrary usually follows the X-ray notation: the states with n ¼ 1; 2; 3; … are designated K, L, M,…, while the combinations of l and j are given the suffixes 1; 2; 3; …: Both notations are listed in Table 6.1. In the case of Ag, the Al-Ka radiation is only energetic enough to probe the core levels up to the 3s shell. The non-s level peaks are doublets. This is related to the fact that when l . 0; two possible states characterized by j arise (see Table 6.1). More details on the difference in energy between the two states and on their relative intensities can be found in Ref. [1]. It is also noted that the core level peaks have different intensities and widths. The relative intensities are governed by the ionization efficiencies of the different core shells, designated by ionization cross-section. The line width, defined as the full width at half-maximum (FWHM) intensity, is a convolution of several TABLE 6.1 XPS and X-ray notation, respectively, used for XPS and AES peaks Quantum number n
l
1 2 2 2 3 3 3 3 3
0 0 1 1 0 1 1 2 2 etc.
Notation j
XPS
X-ray
1/2 1/2 1/2 3/2 1/2 1/2 3/2 3/2 5/2
1s1/2 2s1/2 2p1/2 2p3/2 3s1/2 3p1/2 3p3/2 3d3/2 3d5/2 etc.
K L1 L2 L3 M1 M2 M3 M4 M5 etc.
281
A. Hubin and H. Terryn
contributions: the natural width of the core level, the width of the X-ray line and the resolution of the analyser. The Auger peaks in the spectrum, MNN peaks in the case of Ag, originate from the decay of the ionized atoms to their ground state. The principle of the Auger emission is discussed in section 6.2.2. 6.2.2
Principle of Auger electron spectroscopy
AES uses, in contrast to XPS, electrons as primary radiation. The analysed electrons are not the emitted core electrons, but the Auger electrons that are ejected as a consequence of the return of the ionized atom to its ground state. Figure 6.4 shows a schematic representation of the processes involved in the emission of an Auger electron. For the example shown here, a hole is created on the K level in the initial ionization step. This requires a primary energy greater than the binding energy of the electron in that shell. For the ionization to be efficient, a primary energy of about five times the binding energy is taken. In practice, typical primary energies are 5 and 10 keV. The hole can be produced by either the primary beam or the backscattered secondary electrons. The atom relaxes by filling the hole with an electron coming from an outer level, in the
Fig. 6.4. Schematic representation of the X-ray fluorescence (a) and Auger electron emission (b) processes.
282
X-ray photoelectron and Auger electron spectroscopy
example shown as L1. As a result, the energy difference EK 2 EL1 becomes available as excess energy, which can be used in two ways. The emission of an X-ray at that energy may occur or the energy may be given to another electron, either in the same level or in a more shallow one, as is the case in the example, to be ejected. The first of the two competing processes is X-ray fluorescence, the second Auger emission. Fortunately, the probability for Auger emission is much higher for core levels with binding energies below about 2 keV, as illustrated in Fig. 6.5. An AES transition is written as ABC, where A indicates the level of ionization, B the level where the second electron involved in the transition comes from, and C the level from which the Auger electron is emitted. For A, B and C the X-ray notation (see Table 6.1) is used. The Auger transition represented in Fig. 6.4 is called KL1L2,3. Electrons originating from the valence band are often denoted by V. For example in the case of Ag, the M4,5N4,5N4,5 doublet (see Fig. 6.6) is an MVV transition. More details on the notations and the corresponding electronic configurations can be found in Ref. [1]. The kinetic energy EABC of an Auger transition ABC in an atom of atomic number Z is given by Eq. (6.3) EABC ðZÞ ¼ EA ðZÞ 2 EB ðZÞ 2 EC ðZÞp 2 W
ð6:3Þ
where EI are the binding energies on the Ith atomic level. The star in EpC indicates that it is the binding energy on C in the presence of a hole. In good approximation Eq. (6.4) applies [10]: EABC ðZÞ ¼ EA ðZÞ 2
1 2
½EB ðZÞ þ EB ðZ þ 1Þ 2
1 2
½EC ðZÞ þ EC ðZ þ 1Þ 2 W
ð6:4Þ
Fig. 6.5. Relative probabilities of X-ray fluorescence and Auger emission.
283
A. Hubin and H. Terryn
Fig. 6.6. AES spectrum of Ag, (a) direct spectrum, (b) differentiated spectrum. Conditions: Ep ¼ 5 keV, DE=E ¼ 0:25%:
284
X-ray photoelectron and Auger electron spectroscopy
Equation (6.4) shows that the kinetic energy of an Auger electron is independent of the type of primary beam (i.e., electrons or X-rays) and its energy. For this reason, AES spectra are always plotted on a kinetic energy scale. Since the kinetic energy is only a function of the atomic energy levels, all elements of the periodic table have a unique spectrum. Figure 6.6 shows, as an example, the Auger spectrum of Ag. The AES peaks are superimposed on an important background of different types of secondary electrons. This is the reason why, in many cases, AES spectra are represented in the differentiated form. Note that the MNN peaks of Ag in the AES and the XPS spectrum are at different energies. This is because both techniques do not use the same energy scale: kinetic energies for AES spectra against binding energies for XPS spectra. The energy positions of X-ray and electron-induced Auger transitions differ, as shown by Eq. (6.1), by the value hn; the energy of the X-ray radiation. If in XPS the X-ray source is changed from Mg-Ka to Al-Ka, the position of the XPS peaks remains, while the AES peaks move by 233 eV. This allows the two types of peaks to be easily distinguished. As is the case for XPS peaks, the relative intensities of AES transitions are governed by their respective core shell ionization efficiencies due to the primary electron beam. Yet, for AES the situation is more complex, since there is additional ionization due to backscattered electrons (see Fig. 6.1). The backscattering factors depend on both the energy and the angle of incidence of the primary beam, and they influence the intensities as well as the spatial distribution of the detected Auger electrons as illustrated in Fig. 6.7.
Fig. 6.7. Influence of backscattering on intensity and spatial distribution [11]. Ip is peak intensity due to primary ionization, Ib to backscattered-induced ionization. Primary beam at (a) 5 keV, (b) 20 keV and 408 angle of incidence.
285
A. Hubin and H. Terryn
The sample consists of a 40 nm thick Al layer on Au. The Al KLL peak is shown. In general, AES peaks are broader than XPS peaks. This is related to the complex multiplet splitting due to the number of final states after the transition that are permitted (see Table 6.1). The KLL series, e.g., consists of five components, i.e., KL1L1 (one transition), KL1L2,3 (two transitions) and KL2,3L2,3 (three transitions, of which one is forbidden). Two elements contribute to the peak broadening in solid species: peak overlap in the multiplet structure and solid-state peak broadening. In what follows, the analytical possibilities and the strengths and weaknesses of XPS and AES are discussed. This is, however, impossible without considering the instrumentation first. 6.3 6.3.1
XPS AND AES INSTRUMENTS General set-up
The techniques are comparable in configuration and contain mainly the following parts: (i) a primary beam source: for AES, an electron gun and for XPS, an X-ray source; (ii) an electron energy analyser, combined with a detection system; and (iii) a sample stage, all contained within a vacuum chamber. As for most techniques, the system is operated and controlled by a computer, usually provided with software allowing mathematical treatment of the data. An example of a full set-up is given in Fig. 6.8, while Fig. 6.9 shows a schematic diagram of an XPS configuration with an X-ray source, a
Fig. 6.8. Picture of a full set-up: 5800 Multitechnique system of Physical Electronics combining XPS and AES.
286
X-ray photoelectron and Auger electron spectroscopy
Fig. 6.9. Schematic representation of an XPS set-up with a hemispherical sector analyser.
monochromator and an hemispherical sector analyser (HSA). In Fig. 6.10 a different configuration, used in AES, containing a cylindrical mirror analyser (CMA) with a central electron gun is depicted. XPS and AES apparatus is in a continuous evolution, but in our opinion, with both techniques the most radical progress has been the improvement of the lateral resolution. By the introduction of FEAES (field emission AES), the level of 10 nm can now be reached. The lateral resolution of XPS currently reaches the level of a few micrometres, and XPS mapping facilities are becoming more frequently available than in the past. In the following sections, the most essential parts of the instrumentation will be discussed in more detail. 6.3.2
The vacuum system
The electron spectrometer and sample room must be operating under ultra high vacuum (UHV), typically in the range of 1028 – 10210 torr. The reason for this is 2-fold. The low energy electrons are elastically and non-elastically scattered by residual gas molecules leading to a loss of intensity and energy so that not only the intensity of the peaks is affected but also the noise in the spectrum increases. The second reason is that lowering the vacuum level to, e.g., 1026 torr, would immediately result in the formation of a monolayer of
287
A. Hubin and H. Terryn
Fig. 6.10. Schematic representation of a cylindrical mirror analyser used in AES.
residual gas absorbed on the sample surface in less than a second. A vacuum of 10210 torr allows measurements to be carried out for about an hour before a monofilm is formed. The fact that AES and XPS are methods achieving a depth resolution of a few nanometre, with a detection limit lower than 1% of a monolayer, clearly establishes the high requirements on the vacuum. Even in the case of a vacuum of 10210 torr, typically carbon peaks are found as a result of surface contamination. Therefore, the sample is often cleaned by a slight sputtering, prior to the analysis. The sputtering technique will be discussed in section 6.5.5, but at this stage we already want to stress that sputtering may induce variations in the surface composition and should be handled with care, especially for a nondestructive analysis of cultural heritage material. The way in which such a vacuum is obtained depends on the specific design, but is generally based on diffusion, sputter ion and turbo molecular pumps, combined with auxiliary tools such as Ti sublimation pumps. The high vacuum level imposes high requirements on the used materials. Stainless steel is often selected for the fabrication of the analysis chamber and the joints are usually made of Cu rings. The trajectory of the electrons is strongly influenced by the earth’s magnetic and electric fields and consequently a screening is placed around the system. All UHV systems need occasionally to be baked out to remove contaminants from the chamber walls, the stage and other contact points.
288
X-ray photoelectron and Auger electron spectroscopy
6.3.3
The X-ray source for XPS
In contrast to the electron source, the X-ray source energy depends on the choice of the anode material, resulting in the availability of a number of discrete energies rather than a continuous variation of the energy, as exists for electron and ion guns. The photon energy must be sufficiently high to excite intense photoelectron peaks from all elements of the periodic table (see section 6.2.1). For XPS analysis, it is very important to consider the energy resolution of the primary X-rays. As explained in section 6.2.1, the width of the XPS line depends, among other factors, on the line width of the primary X-ray line, and this strongly affects the energy resolution obtained in the spectra. The most commonly applied configuration consists of a twin anode, providing Al-Ka and Mg-Ka photons with line energies/line widths of, respectively, 1486.6 eV/0.85 eV and 1253.6 eV/0.70 eV. In rare cases, higher energy anodes such Si-Ka (1739.5 eV/1 eV), Zr-La (2042.4 eV/ 1.7 eV) and Ti-Ka (4510.0 eV/2 eV) are used for the excitation of higher energy electron levels such as the 1s level of heavy elements. The X-ray line width can be reduced from 0.85 to 0.4 eV by using a monochromator. Its action is based on Bragg reflection of the X-rays on a single crystal, e.g., natural quartz, as shown in Fig. 6.9. The quartz crystal is placed on the surface of a Rowland or focussing sphere, together with the anode and the sample. The X-rays are dispersed by diffraction on the crystal and refocussed on the sample surface. Other benefits are the removal of contributions to the XPS spectrum of satellite peaks and of the Bremsstrahlung continuum coming from the X-ray spectrum of the anode. The drawback of the use of a monochromator is a severe loss in intensity of the primary Xrays, e.g., up to 40% for Al-Ka radiation, and thus of the resulting XPS peaks. Acquisition times are drastically increased, and detecting low elemental concentrations becomes impossible. As a result, in practice a monochromator is only used in cases where a high energy resolution, typically 0.4 eV, is required (see section 6.5.4). Focusing of X-rays is much more complicated compared to electron beams, and it is only recently that focussed X-ray systems enabling to reach a lateral resolution of a few micrometres are introduced in XPS. Depending on the constructor of the equipment, two trends may be noted. The first possibility is to apply an aperture system in order to select only a fraction of the emitted photoelectrons for detection. The newest development is the socalled microfocus monochromator, where focusing of the X-rays is realized by the intermediate of an electron gun.
289
A. Hubin and H. Terryn
6.3.4
The electron gun for AES
An AES system employs an electron gun, either based on a thermionic or a field emitter source. A thermionic source uses thermal energy to give electrons sufficient energy to escape over the work function of the source, while in a field emission source the work function barrier is reduced. The primary beam source in an AES system is comparable to that in scanning electron microscopes (SEM). For both techniques, there is an increasing trend of using a field emitter type electron source. In the case of a field emission (FE) source, the techniques of AES and SEM are, respectively, denoted FEAES and FESEM. In conventional AES, the only function of the incident beam is to produce ionization in the core levels of the atoms in order to initiate the Auger transitions (see section 6.2.2). From that point of view, the energy level and energy dispersion of the primary electrons are less important than for XPS. The main characteristic of the electron gun is its brightness. It is defined as the number of electrons emitted per unit area and unit solid angle. It determines the number of primary electrons impinging on the surface of the analysed material and therefore directly determines the data acquisition time and sensitivity of the AES system. The simplest form of a thermionic system is a W wire in the shape of a hairpin. A small electric current provokes heating in the wire, so that the electrons achieve energies higher than the work function of about 4.5 eV and are able to escape from the wire. Yet, most AES systems with a thermionic source use a LaB6 filament because its brightness is higher compared to that of W filament sources. A real breakthrough on the level of brightness was the introduction of field emission sources [12]. Field emission is achieved by applying a strong electrostatic field between a filament, in the shape of a needle with a tip radius of about 50 nm, and an extraction electrode. The filament is usually a wire of a W single crystal fashioned into a sharp point. The significance of the small tip radius is that the electric field can be concentrated to extreme levels, up to 109 V/m or more. Applying a high field results in narrowing the barrier in width as well as in height, allowing the electrons to tunnel directly through the barrier and omitting the requirement of thermal energy. The small area of emission from the tip into a small solid angle provides a high brightness compared to thermionic sources. Moreover, a more chromatic (i.e., mono-energetic) beam is obtained by reducing, e.g., chromatic aberrations in the electron lenses, so that smaller spot sizes can be achieved. Cleanliness of the W single crystal is extremely important, since adsorption
290
X-ray photoelectron and Auger electron spectroscopy
Fig. 6.11. Relation between spot size, beam current and energy for an FE gun.
of impurities will increase the work function. Therefore the gun compartment of the instrument is differentially pumped. The final spot size on the sample surface with a particular FE electron gun is a function of the lens system, the beam current and the energy, as illustrated in Fig. 6.11. Figure 6.12 shows an example of an FEAES analysis, resolving Cu particles with a dimension smaller than 100 nm proving that it is one of the best methods to combine high lateral resolution (10 nm range) with high depth resolution (nm range). Note, however, that the obtained lateral resolution depends also on the characteristics of the sample under investigation. 6.3.5
Detection of electron energy
Mainly two types of detectors are used in AES and XPS systems: the CMA and the HSA. In the past, the CMA was preferred for AES systems
Fig. 6.12. FEAES analysis of small Cu particles. (a) SEM image, (b) AES image based on the Cu mapping.
291
A. Hubin and H. Terryn
merely for geometrical reasons and for its high analyser transmission function, and the HSA analyser for XPS for its superior energy resolution. More and more, however, depending on the constructor and the requirements of the customer, the HSA detector is also used in AES systems. A multifunctional AES – XPS system is evidently operated by only one electron detector, and mostly a HSA type is chosen. Yet, it is important to mention that not all the requirements imposed by the type of analysis coincide for XPS and AES. 6.3.5.1 The cylindrical mirror analyser The CMA, shown in Fig. 6.10, consists of two concentric cylinders, the inner cylinder at ground potential while the potential of the outer cylinder is ramped negatively. A proportion of the emitted Auger electrons will pass through the defining aperture in the inner cylinder. Depending on the potential applied on the outer cylinder, electrons of the desired energy pass through the detector aperture and are refocused on the electron detector and measured by a channel electron multiplier. By scanning the potential, a signal proportional to the number of emitted electrons is obtained as a function of the kinetic energy. Unfortunately, the measured spectrum not only contains Auger electrons, but also low energy (typically between 0 and 50 eV) secondary electrons and elastically and inelastically backscattered electrons, with their energies depending on the primary energy that is used (usually between 1 and 10 keV). The AES peaks are superimposed as weak features on a relatively intense background. Therefore, very often the spectrum is recorded while applying a small ac potential modulation to the analyser, so that an analogue derivative spectrum is obtained. Nowadays, with the use of more sensitive multiple electron detection systems and powerful computer systems, most of the spectra are recorded in the direct mode. The signal-to-noise ratio is improved by scanning the energy over the cylinders several times so that counts are accumulated. After recording the spectrum in the direct mode, it can be differentiated numerically as shown in Fig. 6.6. The sensitivity of the CMA is related to the analyser transmission function, giving the effective number of electrons that are measured by the analyser for a particular energy, and is superior compared to the HSA. The energy resolution of a CMA is relative to the energy of the peak. Its best relative energy resolution (about 0.25% DE=E) is clearly inferior compared to that of the HSA, as will be discussed below. Therefore, the CMA is not often used for XPS analysis, except maybe in the past where a
292
X-ray photoelectron and Auger electron spectroscopy
kind of double CMA was introduced. Yet this configuration is no longer considered in new systems. The major advantage of the CMA is that there are less shadowing effects when analysing rough surfaces and small particles, which is often the case for cultural heritage materials (see also section 6.7). This is due to the coaxial position of the electron gun and the detector, as noted in Fig. 6.10, ensuring that the collection of the electrons emitted from the surface is done coaxially around the electron gun. This effect is shown in Fig. 6.13, where the SEM picture and the AES images (see section 6.5.7) of Ni spheres on an In substrate are shown for a coaxial geometry. The SEM picture is taken by the secondary electron detector incorporated in the AES system. The AES image is obtained by scanning the primary lens system over the surface. At each position, the intensity of a number of AES peaks is measured, providing information on the lateral distribution of elements (see also section 6.5.7). Operated in this mode, the technique is also called scanning Auger microscopy (SAM). A drawback of the use of an electron gun and an analyser operating under the same angle is that angle-resolved depth profiling (see section 6.5.5) is not possible. 6.3.5.2 The hemispherical sector analyser All constructors use HSA detectors for electron dispersion analysis in XPS. A typical configuration is shown in Fig. 6.9. The HSA is designed to have a constant and as high as possible energy resolution for the detection of photoelectrons. The best energy resolution in XPS is 0.4 eV, corresponding to the line width of the monochromator. In order to reduce the size of the analyser, it is standard practice to retard the kinetic
Fig. 6.13. Images of Ni spheres on an In substrate, (a) SEM and (b) coaxial AES imaging based on the Ni mapping.
293
A. Hubin and H. Terryn
energies of the photoelectrons either to a user-selected analyser energy, called pass energy, or by a user-selected ratio. The first mode is called fixed analyser transmission mode (FAT), also known as the constant analyser energy mode (CAE). In this mode of operation, which is applied for the detection of photoelectrons, a constant voltage is applied across the hemispheres allowing electrons of a particular energy to pass between them. The most important characteristic in this case is a constant energy resolution in the spectrum as a function of the energy, in contrast to the CMA analyser where a relative energy resolution is obtained. The electrons are emitted from the specimen and transferred to the focal point of the analyser by the lens assembly. At this point they are retarded electrostatically before entering the analyser itself. Those electrons with energies matching the pass energy of the analyser are transmitted, detected and counted by the electron detector. The retarding field potential is then ramped, and so the electrons are counted as function of energy. Since Auger electrons have a higher kinetic energy than photoelectrons, they need to be more retarded. When recording X-ray-induced AES (XAES) or AES spectra, the second mode is applied. In the so-called constant retard ratio (CRR), also known as fixed retard ratio (FRR), the voltages of the hemispheres of the analyser are changed with the energy of the spectrum so that the ratio of the electron kinetic energy to the pass energy is constant. In this mode, the transmission function is optimized at the expense of the energy resolution. To improve the sensitivity of the HSA, the electron detection is done by a multichannel detector system. Depending on the type of system, the number of electron multipliers may go up to 16. This parallel electron detection is especially useful when a monochromator is used due to the loss of intensity of the primary X-rays. 6.3.6
The ion gun
In AES and XPS instruments, sample sputtering can be performed by means of beams of energetic primary ions. Sputtering is useful for two reasons. One is to clean the sample prior to the analysis, because the surface is often contaminated with dirt or residual gas from the atmosphere. The second reason to sputter a sample is to record depth profiles, where the composition is probed in depth by the collection of AES or XPS spectra as a function of the sputtering time. In that case, the ion bombardment is carried out in a sequential manner with the ion gun switched off when the spectrum is recorded. The depth profile is built up by the measurement of the intensity of
294
X-ray photoelectron and Auger electron spectroscopy
the recorded peaks versus etching or sputtering time. The main parameters for sputtering are the energy and the current of the ion beam, the distribution of the current in the sputtered zone and finally the spot size. Three different types of ion guns are in use in XPS and AES systems: the cold static spot gun, the electron impact source and the duoplasmatron type. It is important to note that their characteristics are very distinct, resulting in large differences in obtained sputter profiles. The cold static spot gun has usually a large beam size of 5 –10 mm and is therefore only used for large spot XPS depth profiling or precleaning of the sample. The latter operation is often done in a configuration where the ion gun is mounted in a separate chamber, called preparation vacuum chamber. The gun is back-filled with an inert gas such as Ar having a pressure of about 1026 torr, with a variable potential between 1 and 10 kV. A discharge to form Arþ is realized by an external magnet and the positive ions are accelerated and extracted by a simple focus electrode. In the electron impact source, the ions are created by electrons emitted by a heated filament, accelerated into a cylindrical grid, where they collide with Ar gas atoms. The ion energy is controlled by the potential applied to the grid and ion energies up to 5 kV can be achieved. This ion source produces an ion beam with a narrow energy spread and spot sizes from 2 to 50 mm. It can be operated in static mode or it can be rastered over the surface to produce a larger and more uniform crater. For the rapid removal of material and etching of large areas, the duoplasmatron design of ion source is sometimes preferred. A magnetically constricted arc is used to produce a dense plasma from which the ion beam is extracted, focused and rastered across the specimen by a set of deflector plates. This type of gun provides intense ion beams with a narrow energy spread, very suitable for small spot focusing. However, due to its high price, it is not so much used in AES or XPS systems in contrast to secondary ion microscopy (SIMS) instruments. 6.3.7
The sample holder and stage
The mounting of the samples on the sample holder should be done in such a way that electrical conduction is guaranteed. This is achieved by using metallic clips or bolt-down assemblies. Alternatively a metal loaded tape may also be used. In the case of powders, the particles can be pressed into an indium foil. The sample holder is mounted on a sample stage which allows for high resolution positioning in the x; y; z and u directions. To an increasing extent,
295
A. Hubin and H. Terryn
especially in new systems (and highly appreciated in industrial laboratories) remote control stages are introduced. In this manner, in an automated fashion, different sample areas may be analysed, and if a parking stage is available, an automatic exchange of samples is possible. Triggered by the growing interest of the semiconductor industry, the vacuum systems are made accessible for large samples, allowing to analyse semiconductor wafers without preceding sample preparation. In most research instruments in use in university laboratories, the sample size is limited in area to a few cm2 and in thickness to a few mm. In some systems additional sample manipulation facilities are present such as, e.g., cooling or heating of the sample holder in situ during the analysis. A valuable tool in materials characterization is a separate preparation chamber connected to the analysis system, enabling sample treatments such as heating or tensile testing. After pretreatment, the sample is introduced into the vacuum chamber for analysis through a lock gate without coming in contact with the air. The angle between the ion beam, the primary beam and the detected electron beam may induce shadowing and redeposition effects when sputtering rough surfaces. To optimize the sputtering conditions, a sample stage with tilt and rotation capabilities is useful. In Ref. [13] the Zalar rotation is applied to this purpose. 6.4
SAMPLE REQUIREMENTS
Sample requirements are different for AES and XPS. Although it is also possible to detect and analyse Auger peaks in an XPS facility, the AES method is based on the use of a primary beam of electrons. This means that in principle the sample has to be conductive. In general, a sample can be analysed in AES if one can obtain good SEM pictures from it without coating. For AES investigations, isolating samples cannot be coated with a conductive C or Au layer as is commonly done in SEM –EDX analysis since the escape depth of the Auger electrons is only a few nm. Thus, AES is especially appreciated for the analysis of metals and passive films on metals, but there is a growing tendency to analyse semiconducting and ceramic materials with it. Our own experience shows that, e.g., Al covered with a non-conducting oxide film of about 0.1 mm is measurable [13]. In other cases, special attributes, such as a conductive grid to be placed on the sample, are necessary to limit charging effects. Another possibility is to adapt the working conditions. This can be done by varying the angle of incidence of the primary electrons or, more complicated, by performing the analysis under
296
X-ray photoelectron and Auger electron spectroscopy
conditions, i.e., low primary electron energy and low current, where the impinging primary current is balanced by escaping electrons. In principle the requirement of a conductive sample is less stringent with a primary X-ray beam. However, practice shows that in XPS also charging effects may occur when analysing non-conducting samples, especially when using a monochromator. Therefore the constructors of XPS equipment add in their newest configurations a neutralization system which is able to compensate the charge automatically. Both for XPS and AES, the samples should be stable in a UHV system of 10210 torr. Polymers with low vapour pressure, wet samples and porous materials need to be handled with care. Many cultural heritage materials belong to this category, as will be shown in section 6.7. Most operators prefer flat samples, first because the interpretation of the data is more straightforward, and secondly because roughness deteriorates the depth resolution in sputtering mode (see section 6.5.5). 6.5 6.5.1
INFORMATION IN XPS AND AES SPECTRA Surface analysis
For XPS as well as for AES, the primary beam has a penetration depth of a few micrometres. The photoelectron (XPS) or Auger electron (AES) can only travel a limited distance, called attenuation length l, before being inelastically scattered. The characteristic depth d from which photoelectrons and Auger electrons are emitted, called the escape depth, is given by d ¼ lðEÞcos u
ð6:5Þ
where u is the angle of emission from the surface normal. The attenuation length varies according to the element, which is emitting the electron and to the matrix, and depends on the energy of the emission. Typical values of l are, as shown in Fig. 6.14, in the range of 1 –10 atom layers [14]. This is the basis for the surface sensitivity of XPS and AES spectroscopy. 6.5.2
Qualitative analysis
The qualitative analysis of a specimen consists in identifying the elements that are present. For this purpose, a survey or wide energy scan spectrum is recorded. As outlined in section 6.2, each element has a characteristic XPS and AES spectrum. In Refs. [7,8], the spectra of all elements can be found. Nowadays instrumentation, both for XPS and AES, is equipped with data
297
A. Hubin and H. Terryn
Fig. 6.14. Dependence of attenuation length l on the emitted electron energy [14].
treatment systems running automatic peak identification tools. The identification of the composing elements of a sample under investigation is in most cases straightforward, except if peaks are overlapping. Which of the techniques to choose for a qualitative analysis depends mainly on two factors. First, the nature of the specimen is important: conductive specimens can be analysed by both techniques, while for non-conductive specimens XPS is more appropriate, although some remedies against charging effects exist (see section 6.4). Secondly, the lateral resolution required for the analysis should be considered. If surface distributions on the micrometre scale are expected, AES is the proper method. The combined use of both techniques can be very useful for the analysis of complex spectra. This is illustrated in Fig. 6.15 for a Cr sample with an oxide layer at the surface. In the AES spectrum, the Cr and O peaks overlap, while in the XPS spectrum they are clearly separated. 6.5.3
Quantitative analysis
The determination of the surface composition of a specimen is more complicated. To this purpose, data is collected in the multiplex mode. The relevant energy windows are selected, and the peak intensities are measured with high energy resolution and signal-to-noise ratio.
298
X-ray photoelectron and Auger electron spectroscopy
Fig. 6.15. Complementary information in the AES (a) and XPS (b) spectra of an oxidized Cr sample. Conditions: Ep ¼ 5 keV for AES, Al-K a at 350 W, pass energy ¼ 58.7 eV for XPS.
299
A. Hubin and H. Terryn
The calculation of the composition based on first principles is in practice hardly ever done. The proportion between the peak intensity and the concentration of the emitting element depends in a complex manner on the intensity of the primary beam, the ionization cross-section, the probability of Auger transition in the case of AES, the attenuation length, the spectrometer transmission efficiency and the detector efficiency. A more practical approach consists in incorporating those parameters into sensitivity factors. The intensity of a signal from an element A, IA ; in a solid is proportional to its molar fraction xA xA ¼ IA =IA0
ð6:6Þ
where IA0 is the intensity from a pure A sample and may be considered as a sensitivity factor. In general what is used in the commercial data treatment software is a set of relative sensitivity factors, normalized to a reference element. For this reference, in most cases Cu or Ag, the sensitivity factor is set to unity. The molar fraction xA in a homogeneous specimen composed of i elements is then given by .X .X Ni ¼ ðIA =SA Þ xA ¼ N A ðIi =Si Þ ð6:7Þ where Ii is the area of the peak generated by constituent i; Ni its number of moles and Si its relative sensitivity factor. Since the sensitivity factors include instrumental parameters, they need to be determined on each spectrometer and for each primary beam intensity. Moreover, since the escape depth is matrix and surface roughness dependent, using the values of sensitivity factors included in the data treatment software introduces errors. A more correct quantification can be obtained by using sensitivity factors that are determined on an appropriate set of reference samples with representative matrix effects, as shown in Ref. [15]. This may, however, be very difficult for cultural heritage materials. Anyway, the technique cannot be applied rigorously on heterogeneous samples, since in that case the assumption of constant sensitivity factors is not valid. The problem is more severe in AES analysis, due to the contribution of backscattered electroninduced ionization. An additional difficulty encountered in quantitative analysis is the determination of the peak areas Ii (see Eq. (6.7)). XPS peaks, and even more pronounced AES peaks, appear on a background. In some cases, the background correction is not straightforward. In most commercial data handling systems different possibilities exist. The simplest one is the straight line between two suitably chosen points. More complex methods are the Shirley and the Tougaard background correction procedures [1].
300
X-ray photoelectron and Auger electron spectroscopy
6.5.4
Chemical analysis
XPS and AES not only allow to identify and quantify the constituting elements of a sample, but also make it possible to obtain information on their chemical state. In this respect, XPS is favoured to AES, and the reason for it is to be found in the nature of the respective transitions that are used in both methods. The core level peaks in XPS show a clear shift in binding energy, the socalled chemical shift, related with differences in the chemical environment of the emitting element. This is shown in Fig. 6.16 for the C peak of a polyethylene terephthalate (PET) sample. The capability to distinguish between different chemical states is the main characteristic of XPS. Due to this, another acronym is in use for this technique: ESCA, which stands for “electron spectroscopy for chemical analysis.” The shifts are typically a few to 10 eV or more, and therefore detectors with a high energy resolution are used (see section 6.3.5.2). To subtract chemical information, it is imperative to determine peak positions as accurately as possible. The line of interest is preferentially evoked by means of a monochromatic X-ray source, and recorded with the highest possible energy resolution. When dealing with small chemical shifts, overlapping peaks may anyway occur in the spectra. Peak deconvolution and peak fitting
Fig. 6.16. Chemical shift of the C 1s peak as a function of its bonding with O.
301
A. Hubin and H. Terryn
tools are available in the commercial data handling systems. An illustration is seen in Fig. 6.16, where the C 1s peak is deconvoluted in its components. In AES, the situation is less promising. First, most AES lines are by nature broader than XPS lines. Secondly, the Auger transition involves three electrons and the overall chemical shift is influenced by the three energy levels concerned. In general, similar shifts of a few eV as for XPS are to be expected, especially when core electrons are involved in the transition. Since in AES-specific equipment, detectors with lower energy resolution are used (see section 6.3.5.1) it is clear that chemical shifts are not often measured accurately. Moreover, peaks corresponding to transitions with valence electrons are broad and poorly defined, which makes the assignment of chemical shifts nearly impossible. In that case, the identification of the chemical state may be done by a peak shape analysis. In fact the shape of a CVV or CCV peak, C being a core level, is related to the density of states (DOS) in the valence band. Since the DOS varies from one chemical environment to another, a variation in peak shape may be observed. This effect is commonly noticed in Auger spectra of non-metallic elements such as C, S, O, N. Figure 6.17 shows a few examples of chemical effect in differentiated AES spectra.
Fig. 6.17. Chemical shifts and line shape effects in AES spectra of Al, Si and C.
302
X-ray photoelectron and Auger electron spectroscopy
An additional difficulty that occurs when determining chemical shifts is the electrostatic charging of the specimen. In case of metallic specimens, this problem can be easily avoided by using a proper mounting procedure, but for poorly conducting materials a “charging” shift can be observed, especially in AES where an electron beam is used as primary radiation. However, also in XPS the phenomenon is encountered. Although an internal standard, e.g., the C 1s for XPS or the C KVV line for AES, may be used to estimate the value of the charging shift, it is not very accurate. In order to solve this problem, the Auger parameter was introduced [16]. The Auger parameter a is the difference between the kinetic energies of a photoelectron line and an Auger line in the XPS spectrum. Taking Eq. (6.1) into account, the expression for a becomes a ¼ EK þ EB 2 hn
ð6:8Þ
where EK is the kinetic energy of the sharpest Auger line and EB the binding energy of the most intense photoelectron line. The modified Auger parameter a0 is introduced to ensure a positive value: a0 ¼ a þ hn ¼ EK þ EB
ð6:9Þ
Since a and a0 are related to a difference in energy measured on the same sample in the same spectrometer, any static charge effects cancel out. The Auger parameter has a unique value for each chemical state and can be used as a fingerprint. Tabulations of the Auger parameter can be found in handbooks (see, e.g., Ref. [7]). 6.5.5
In-depth analysis
XPS and AES can be used to provide compositional information as a function of depth. It can be obtained by non-destructive and destructive techniques. The two most commonly applied methods, i.e., angular-resolved measurements and ion sputtering, are discussed below. 6.5.5.1 Angular-resolved measurements This method is almost exclusively used in XPS and is non-destructive since no material is removed. It is called angle-resolved XPS (ARXPS). The principle of this method is represented in Fig. 6.18. The intensity I of electrons emitted from a depth d is given by the Beer– Lambert relationship I ¼ I0 expð2d=l sin aÞ
ð6:10Þ
303
A. Hubin and H. Terryn
Fig. 6.18. Principle of ARXPS. d ¼ sampling depth, l ¼ attenuation length.
where I0 is the intensity from an infinitely thick clean substrate, a the electron take-off angle relative to the sample surface and l and d are as defined above. At 908, 95% of the signal intensity emerges from a distance 3l, while at 158, this is reduced to a distance of 0.8l. In the newest equipment, the measurements are performed in such a way that the analysed area remains constant throughout the tilt range. For example, in the case of a metal surface M covered by a thin organic overlayer containing C, the ratio of the peak intensities IM and IC ; when measured as a function of a, will vary. Valuable information on the thickness of the overlayer is gained, but the technique is limited to thin layers (of a few nm). 6.5.5.2 Sputter profiles Sputtering is a destructive method. The sample is bombarded with highly energetic ions (mostly Arþ ions with an energy of 1– 5 keV), the surface atoms are sputtered away and the residual surface is analysed. By this technique, layers up to 1 – 2 mm are accessible. AES or XPS spectra are recorded, either discontinuously after the subsequent sputter steps or simultaneously with the sputtering. The original data consist of signal intensities of the detected elements, mostly peak-topeak heights in AES and peak areas in XPS, as a function of sputtering time. A typical example is shown in Fig. 6.19. To obtain the original concentration distribution of the elements and/or their chemical states two transformation steps are required.
304
X-ray photoelectron and Auger electron spectroscopy
Fig. 6.19. XPS sputter profile of a Ti oxide layer on top of a TiNb alloy. Sputter conditions: 4 kV Arþ on a 2 £ 2 mm2 area. Analysis conditions: Al-Ka at 350 W.
Depth scale calibration is the first step. The sputter rate, which relates sputter time and depth, depends on instrumental parameters, but is also affected by the specimen. The instrumental effects can be determined by a calibration measurement on a layer of known thickness. In practice, very often a standard reference sample of Ta2O5/Ta with 30 and 100 nm thickness is used for calibration. In general, however, the sputter rate varies with the composition, and is by consequence sample dependent. Even for a constant composition, sputtering-induced effects, e.g., where a component is preferentially sputtered, may cause non-linearities in the sputter time and depth relationship. In the second step, peak intensities are translated into concentrations of elements taking into account their respective sensitivity coefficients. The obtained concentration profile deviates from the real one because differences in escape depth of Auger or photoelectrons originating from different elements are neglected. A number of effects disturbs the sputtering profiles. This is expressed by the depth resolution. The most common definition of the depth resolution is the difference in depth coordinate between 84 and 16% of the intensity change at an interface. Besides the differences in escape depth, other parameters contribute to the broadening of profiles. The most important are instrumental parameters, sample characteristics and radiation-induced effects. The influence of the instrumentation is on the level of the quality
305
A. Hubin and H. Terryn
of the vacuum (risk of reaction of the sputtered surface with residual gases), the purity of the ion beam (importance of a differentially pumped ion gun) and of the etch crater. The etch crater must coincide with the analysis area, and analysis of the crater walls must be avoided. In this perspective, XPS depth profiling is unfavourable compared to AES due to its larger spot size. The main sample characteristics limiting depth resolution are (a) surface roughness, possibly causing shadowing effects between the ion beam and the primary radiation and non-uniform sputter yields, and (b) preferential sputtering of one component in the study of alloys or compounds. Finally radiation-induced effects comprise among others implantation of primary ions, ion-induced reactions, e.g., reduction of Cu2þ to Cuþ, and atomic mixing through knock-on effects (displacement of atoms to deeper layers). 6.5.6
Data analysis
Data analysis is a technique that may be helpful in interpreting multidimensional data sets with overlapping spectral features. They are generated either in chemical analysis problems where the same spectral region is studied on different samples (e.g., metal oxide, hydroxide, sulphide,…) or in depth profiling studies where the same sample is measured under different conditions (e.g., a paint layer on top of a substrate). The two most commonly applied multivariate techniques are LLSF (linear least squares fitting) and FA (factor analysis). Both assume that each spectrum can be represented by a linear combination of component spectra ½D ¼ ½R½C
ð6:11Þ
where ½D is the data matrix of recorded XPS or AES spectra, ½R the matrix of spectra of independent components and ½C a matrix of weighting factors or concentrations. In LLSF, ½C is determined by a linear least squares minimalization. Data analysis by means of FA is more complex, and several techniques exist, but the basic idea is to first identify the number of independent components and then to determine the true component spectra together with the concentrations. Illustrations of data analysis can, e.g., be found in Refs. [17,18]. 6.5.7
Imaging
Distributions of elements and chemical states over the surface are measurable using rastering techniques. Obviously, for this application AES is in favour compared to XPS in view of its smaller analysed area (see section 6.3).
306
X-ray photoelectron and Auger electron spectroscopy
Imaging in AES is also named SAM. A well-focused incident electron beam is rastered over the surface, and the relevant spectra are collected. Peak intensities are translated into gray scale values. Examples are shown in Figs. 6.12 and 6.13. The smaller the spot size, the smaller the influence of the backscattering [1]. For spot sizes larger than 100 nm, the backscattering phenomena may modify the images in a complex manner. With the developments in small spot XPS, imaging in XPS is nowadays also feasible, as illustrated in Fig. 6.20.
Fig. 6.20. (a) Cross-section image and (b) C, O, Cl and Si XPS images of a painted substrate.
307
A. Hubin and H. Terryn
6.6
COMPARISON OF XPS, AES AND OTHER SURFACE ANALYTICAL TECHNIQUES
XPS and AES are often combined in one experimental set-up because they are complementary. Sometimes, the system is also equipped with other sources and detectors, so that other techniques can be applied simultaneously. A few examples are: energy-dispersive X-ray analysis (EDX), secondary ion mass spectroscopy (SIMS), scanning tunneling microscopy (STM) or ellipsometry. Making a list of all the available surface analytical methods is, even for the most experienced surface scientist, an extremely difficult job as the number of techniques continuously increases, especially when newer techniques such as scanning probe or light reflection methods are considered. Each of the many techniques has its advantages, and it is beyond the scope of this chapter to compare them all. The main common characteristic of XPS and AES is their high surface sensitivity. Another technique often used for the same reason is SIMS. In Table 6.2 these techniques are compared. The characteristic values that are listed here are only indicative since all equipment are in continuous evolution. Moreover, some differences in performance exist between the commercially available systems. 6.7
XPS AND AES FOR CHEMICAL ANALYSIS OF CULTURAL HERITAGE MATERIALS
From the principles explained above, it is clear that XPS and AES can yield valuable information on the surface characteristics of archaeological and artistic artefacts, often essential for the understanding of their manufacturing techniques or ageing, and for the selection of the conservation methods. A literature search shows that Lambert et al. [19,20] were pioneers in exploring the applicability of XPS for the analysis of artefacts. More recently they provided an introductory overview of the use of XPS in archaeology [21]. Ciliberto and Spoto [22] edited a monograph “Modern Analytical Methods in Art and Archaeology” including the use of XPS and AES [23]. They provide a comprehensive overview of the variety of cultural heritage materials analysed by the techniques outlined above. Pottery plays an important role in art historical analysis. Among other techniques, XPS has contributed to its chemical analysis [23]. As pottery is a ceramic material, AES cannot be used for its characterization. Lambert et al. [20], e.g., studied Mycenaeen pottery, while Bruno et al. [24,25] investigated medieval pottery. XPS is used for elemental identification,
308
Comparison of the main characteristics of XPS, AES and SIMS Characteristics
XPS
AES
SIMS
Primary beam Analysed beam Type of sample Area of analysis Surface selectivity Elemental identification Sensitivity
X-rays Electrons (energy) All, charging possible 10 mm 1– 5 nm All except H, He 0.1%
Electrons Electrons (energy) Conductive 10 nm 1–5 nm All except H, He 0.1%
Quantification Molecular identification Nature of chemical bonding Depth profiling Destructive nature
Requiring standards Not possible Shift, straightforward
Requiring standards Not possible Shift and shape requiring data analysis Elemental, chemical None if not sputtered
Ions Ions (mass) All, charging possible 100 nm 0.1 –1 nm All , 1 ppm (dynamic), 100 ppm (static) Requiring close standards Mostly possible Not possible
Elemental, chemical None if not sputtered
Elemental Always
X-ray photoelectron and Auger electron spectroscopy
TABLE 6.2
309
A. Hubin and H. Terryn
quantification and especially for elemental state identification. Bruno et al. [24] have demonstrated in their investigations the importance of multivariate data treatment methods. Also XPS and AES studies on glasses are reported [23], although for AES results the possible effects of the electron beam on the surface composition are clearly pointed out. By means of XPS, it could be demonstrated that, e.g., the oxidation state of Cu influences the colour of the glass [19,21]. Metallic objects of archaeological or artistic interest have frequently been studied by XPS and AES, owing to their conductive nature [23]. Many efforts have been devoted to the analysis of ancient bronzes by Polak et al. [26], Paparazzo et al. [27– 30] and Spoto et al. [31]. Points of interest are metallurgical aspects, corrosion damage and formation of patinas. XPS is mostly used to distinguish between the different oxidation states of Cu in bronze (Cu0, Cuþ, Cu2þ), while SAM reveals the distribution of the elements—Cu and Sn as the major alloying elements, but also C, O, S, Cl for their role in patina formation—and their chemical states on the surface. Also Pb objects made during the Roman Empire have been intensively studied [23]. Paparazzo et al. [27,32,33] applied XPS and AES for the investigation of Roman lead pipes (fistulae), with special interest in soldering techniques, again validating the strong points of both spectroscopies. Ingo et al. [34 –36] investigate early Fe metallurgy slags and Fe ores with small spot XPS and XAES. Determination of the oxidation state of Fe, and chemical analysis of Si, Al, Ti based on the Auger parameter provides data allowing to elucidate some aspects of the early Fe making process and on the geological origin of the used ores. For the study of dyes, pigments and paintings [23], XPS is much in favour compared to AES: first because chemical information is sought and secondly because organic materials, subject to degradation processes, are involved. For example, Wilson-Yang et al. [37] used XPS data to gain insight in the colour degradation of mural paintings. Ciliberto and Spoto [23] illustrate the difficulties encountered when analysing paper, i.e., a non-conductive, porous, rough and non-homogeneous material. Heating effects may modify the chemical bonding of C, the major constituent. The use of a cold stage is advisable in this case. AES has not found significant applications in the study of paper, but SAM appeared useful in dating manuscript inks. XPS in combination with sputtering is applied to establish degradation processes in reactive atmospheres of stones [23], the earliest material used by humans. Comparison with other techniques confirms the difficulties encountered when using sputter profiles to evaluate the thickness of layers.
310
X-ray photoelectron and Auger electron spectroscopy
This selection of examples of analysis of cultural heritage materials provides proof of the potential of XPS and AES/SAM, the main advantage being the possibilities for chemical state identification with high lateral resolution combined with surface sensitivity. It is obvious, however, that great care has to be taken to avoid alteration of the samples’ composition during investigation. We are convinced that, thanks to the continuous improvement of both techniques, their value in the field of art and archaeology will continue.
Acknowledgements Physical Electronics is greatly acknowledged for the help with the figures. Many thanks to O. Steenhaut for the spectra, and V. Hayez for the bibliography. REFERENCES 1 2 3 4 5 6 7
8
9 10 11 12
D. Briggs and M.P. Seah, Practical Surface Analysis. Volume 1—Auger and X-Ray Photoelectron Spectroscopy, 2nd ed., Wiley, Chichester, 1990. T.A. Carlson, Photoelectron and Auger Spectroscopy. Plenum Press, New York, 1975. J.F. Watts, An Introduction to Surface Analysis by Electron Spectroscopy. Oxford University Press, Oxford, 1990. G. Hollinger and P. Pertosa, In: B. Agius, et al. (Eds.), Surfaces Interfaces et Films Minces—Observation et Analyse. Dunod, Paris, 1990, Chapter 3. M. Thompson, M.D. Baker, A. Christie and J.F. Tyson, Auger Electron Spectroscopy. Wiley, New York, 1985. V.I. Nefedov, X-Ray Photoelectron Spectroscopy of Solid Surfaces. VSP BV, Utrecht, 1988, English Edition. J.F. Moulder, W.F. Stickle, P.E. Sobol and K.D. Bomben, In: J. Chastain and R.C. King Jr. (Eds.), Handbook of X-Ray Photoelectron Spectroscopy. Physical Electronics, USA, 1995. K.D. Childs, B.A. Carlson, L.A. LaVanier, J.F. Moulder, D.F. Paul, W.F. Stickle and D.G. Watson, In: C.L. Hedberg (Ed.), Handbook of Auger Electron Spectroscopy, 3rd ed., Physical Electronics, USA, 1995. G. Beamson and D. Briggs, High Resolution XPS of Organic Polymers. The Scienta ESCA300 Database. Wiley, Chichester, 1982. M.F. Chung and L.H. Jenkins, Surf. Sci., 21 (1970) 253. N.M. Glezos and A.G. Nassiopoulow, Surf. Sci., 254 (1991) 314. J.I. Goldstein, D.E. Newbury, P. Echlin, D.C. Joy, A.D. Romig Jr., C.E. Lyman, C. Fiori and E. Lifshin, Scanning Electron Microscopy and X-Ray Microanalysis, 2nd ed., Plenum Press, New York, 1992, Chapter 2.
311
A. Hubin and H. Terryn 13 14 15
16 17 18
19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37
312
J. De Laet, K. De Boeck, H. Terryn and J. Vereecken, Surf. Interface Anal., 22 (1994) 175. M.P. Seah and W.A. Dench, Surf. Interface Anal., 1 (1979) 2. M. Detroye, F. Reniers, C. Buess-Herman and J. Vereecken, In: I. Olefjord, L. Nyborg and D. Briggs (Eds.), Proceedings ECASIA 97, Seventh European Conference on Applications of Surface and Interface Analysis, 16 –20/06/97, Go¨teborg, Sweden, Wiley, New York, 1997, 995 pp. C.D. Wagner, L.H. Gale and R.H. Raymond, Anal. Chem., 51 (1979) 466 and references cited herein. F. Reniers, A. Hubin, H. Terryn and J. Vereecken, Surf. Interface Anal., 21 (1994) 483. G. Treiger, I. Bondarenko, P. Van Espen, G. Goeminne, N. Roose and H. Terryn, In: H.J. Mathieu, B. Reihl and D. Briggs (Eds.), Proceedings ECASIA 95, Sixth European Conference on Applications of Surface and Interface Analysis, 9–13/ 10/95, Montreux, Switzerland, Wiley, New York, 1996, 775 pp. J.B. Lambert and C.D. McLaughlin, Archaeometry, 18(2) (1976) 169. J.B. Lambert, C.D. McLaughlin and A. Leonard Jr., Archaeometry, 20(2) (1978) 107. J.B. Lambert, C.D. McLaughlin, C.E. Shawl and L. Xue, Anal. Chem. News Features, (1999) 614A, September 1. E. Ciliberto and G. Spoto, Modern Analytical Methods in Art and Archaeology. Wiley, New York, 2000. G. Spoto and E. Cilibert, In: E. Ciliberto and G. Spoto (Eds.), Modern Analytical Methods in Art and Archaeology. Wiley, New York, 2000, Chapter 13. P. Bruno, M. Caselli, M.L. Curri, P. Favia, R. Lamendola, A. Mangone, A. Traini and C. Laganara, Fresenius J. Anal. Chem., 350 (1994) 168. P. Bruno, M. Caselli, M.L. Curri, P. Favia, C. Laganara and A. Traini, Ann. Chim., 87 (1997) 539. M. Polak, J. Baram and J. Pelleg, Archaeometry, 25(1) (1983) 59. E. Paparazzo and L. Moretto, J. Electron Spectrosc. Relat. Phenom., 76 (1995) 653. E. Paparazzo, L. Moretto, J.P. Northover, C. D’Amato and A. Palmieri, J. Vac. Sci. Technol. A, 13(3) (1995) 1229. E. Paparazzo and L. Moretto, Vacuum, 55 (1999) 59. E. Paparazzo, A.S. Lea, D.R. Bear and J.P. Northover, J. Vac. Sci. Technol. A, 19(4) (2001) 1126. G. Spoto, E. Ciliberto, G.C. Allen, C.M. Younes, P. Piccardo, M.R. Pinasco, E. Stagno, M.G. Ienco and R. Maggi, Br. Corr. J., 35(1) (2000) 43. E. Paparazzo, Appl. Surf. Sci., 74 (1994) 61. E. Paparazzo and L. Moretto, Vacuum, 49(2) (1998) 125. G.M. Ingo, L. Scoppio, R. Bruno and G. Bultrini, Mikrochim. Acta, 109 (1992) 269. G.M. Ingo, S. Mazzoni, G. Bultrini, S. Fontana, G. Padeletti, G. Chiozzini and L. Scoppio, Surf. Interface Anal., 22 (1994) 614. G.M. Ingo, G. Padeletti, G. Chiozzini and G. Bultrini, J. Therm. Anal., 47 (1996) 263. K.M. Wilson-Yang and G. Burns, Can. J. Chem., 65 (1987) 1058.
Chapter 7
Laser ablation inductively coupled plasma mass spectrometry Teresa E. Jeffries
7.1
INTRODUCTION
Laser ablation inductively coupled plasma mass spectrometry (LA-ICP-MS) is a relatively new microbeam technique and is still undergoing rapid development. It has been applied to an extraordinarily diverse range of materials and applications. The technique provides spatially resolved determinations of trace element concentrations and isotope ratios in solids, complementing others, such as electron microprobe analysis and ion microprobe analysis. LA-ICP-MS is frequently referred to as a non-destructive analytical technique and with its inclusion in this book, the reader would be forgiven for believing this to be the case. Nevertheless, as the name implies, the process of laser ablation is one which does involve sample destruction (Fig. 7.1), albeit at the microscopic scale and local to the site of interest. It is worth remembering that most of the so-called non-destructive analytical techniques involve hugely damaging sample preparation steps, including cutting, sectioning, polishing and coating. LA-ICP-MS is atypical in that samples may be presented in an unprepared state. So for a self-confessed destructive technique, in some circumstances, LA-ICP-MS might be considered the least damaging of all. Indeed, it is often this aspect of the technique, which has led to its use in the study of rare or precious cultural heritage and other materials. The intention of this chapter is to provide a practical description of LAICP-MS, concentrating on the important instrumental characteristics and other factors which influence the data obtained. In recent years the acronym LA-ICP-MS has come to represent a number of techniques in which a laser is coupled to an ICP-MS, be it a quadrupole, time of flight, magnetic sector, high resolution, single or multicollector based ICP mass spectrometer. Comprehensive Analytical Chemistry XLII Janssens and Van Grieken (Eds.) q 2004 Elsevier B.V. All rights reserved
313
Teresa E. Jeffries
Fig. 7.1. Scanning electron microscope image of crater formed in Pele’s hair (volcanic glass) by laser ablation.
This chapter begins with a description of a basic, or generic quadrupole based ICP-MS instrument. The quadrupole ICP-MS was the first to be developed, is widely available for routine use and has thus far the widest application to cultural heritage material. The main components which make up a laser are discussed in the next section, concentrating again on the most commonly used lasers, the frequency quadrupled and quintupled Nd:YAG lasers operating at 266 and 213 nm, respectively. Following this, the laser and ICP-MS are brought together in a discussion of analytical concepts and factors affecting analysis. A brief literature review is presented next and the chapter concludes with a look at some recent developments in LA-ICP-MS such as the introduction of collision and reaction cells. 7.2 7.2.1
THE INDUCTIVELY COUPLED PLASMA MASS SPECTROMETER Historical account
In the mid-1960s, long before the advent of ICP-MS, the inductively coupled plasma was being developed and used by Greenfield et al. [1] and Wendt and Fassel [2] as an excitation source in optical emission spectrometry (ICPOES). The first commercial emission systems became available in the mid1970s and rapidly gained popularity. The technique, still popular today, is now more frequently termed inductively coupled plasma atomic emission
314
Laser ablation inductively coupled plasma mass spectrometry
spectrometry (ICP-AES). The ICP proved to be an excellent excitation source but whilst the advantages of ICP-AES included high precision and large dynamic range, the technique was limited by the complexity of the emission spectra produced. Thus the challenge was set to find a detection system which was free from complex spectral interferences and so arose the idea of using a mass spectrometer to detect the ions from the plasma source. The main obstacles in coupling a plasma source to the high vacuum system of a mass spectrometer are the high gas temperature (, 7000 K) and pressure (105 Pa) of the ICP. Using a direct current plasma (DCP), and a quadrupole based mass spectrometer, Gray [3] was the first to demonstrate that it was feasible to couple a plasma source to a high vacuum system. The plasma was allowed to impinge on to an aperture in the tip of a cone, mounted in the wall of a first vacuum stage (,0.1 Pa). Ions, created in the plasma, were focused into a second higher vacuum stage containing the quadrupole mass analyser. Further work was initiated following this, using alternative plasma-mass spectrometer combinations including the microwave induced plasma (MIP), but it was the earlier DCP work upon which Houk et al. [4] and Date and Gray [5] based their development of the ICP-MS. Both groups encountered problems with the small cone aperture at the interface between the plasma and first vacuum stage. However, at the same time Douglas and French [6] were having some success with a new sampling interface and larger aperture, using a MIP. In 1983, Date and Gray [7] and Douglas et al. [8] simultaneously reported the use of the new sampling interface in conjunction with an ICP and so the ICP-MS was born. By 1984 two companies had delivered the first commercial ICP-MS instruments. Huge improvements in instrument sensitivity have been made since the early days of the technique, with modern quadrupole instruments capable of delivering single parts per quadrillion (ppq—10215) detection limits. There are now several manufacturers offering quadrupole, magnetic sector and time of flight ICP-MS instruments, each of course, with their own variation of the once troublesome interface. 7.2.2
Operational rationale
The operation of a quadrupole ICP-MS may be summarized quite simply (Fig. 7.2): a sample is dispersed into a stream of gas (typically Ar, He or a mixture of both) and injected into the core of an inductively coupled plasma. In the plasma the sample is rapidly heated, vaporized and ionized. The plasma impinges on the aperture in the tip of a cone, through which
315
Teresa E. Jeffries
Fig. 7.2. Schematic of an inductively coupled plasma mass spectrometer with ion lens array. Drawing not to scale.
the sample ions are extracted into a reduced pressure region. Some of the extracted sample ions pass through a second aperture into a region of high vacuum. Here, either an electrostatic ion lens array or, in some recent instruments, a multipole collision and reaction cell is used to direct the ions to a quadrupole mass analyser. The quadrupole is supplied with DC and RF voltages. For a given DC and RF voltage, only ions of one mass/charge ðm=eÞ value are transmitted through the quadrupole. In operation, the DC and RF voltages are cycled rapidly and repeatedly, so that the quadrupole transmits the ions of each m=e of interest many times, in rapid succession. Transmitted ions are registered at an ion detector. Identification of the elements is straightforward as each has a unique pattern of near integer m=e values which correspond to the element’s natural (stable) isotopes. The number of ions registered at the detector for each isotope will depend on the concentration of the element in the sample, and this provides a system of quantification. The majority of ICP-MS instruments are used to determine trace element concentrations in liquids. Here, a fine mist of liquid is produced using a nebulizer and is introduced to the plasma. Some instruments are configured for the direct analysis of solid samples via laser ablation. In this case, a laser is used to remove material (ablate) from a small area of the sample. The tiny particles which are produced are flushed away from the sample in a flow of gas and injected into the plasma.
316
Laser ablation inductively coupled plasma mass spectrometry
Other methods of sample introduction are available and include specialist nebulizers such as the desolvating micro-concentric (MCN) or ultrasonic nebulizers (USN), electrothermal vaporization (ETVA), slurry nebulization and spark ablation. For elemental speciation studies samples may also be introduced via liquid chromatography (LC-ICP-MS). 7.2.3
The inductively coupled plasma
A plasma is an ionized gas containing approximately equal numbers of electrons and ions. It is electrically neutral, highly conducting and is affected by magnetic fields. Constituting 99% of matter, natural plasmas such as that found at the Sun’s surface, are the most abundant phase in the Universe. On Earth they exist transiently and dramatically in phenomena such as the auroras and lightening. In the laboratory, an ICP (Fig. 7.3) is generated by coupling an intense radio frequency (RF) electromagnetic field to a suitable gas. The radio frequency generator (typically with an output between 0.5 and 2 kW at 27 MHz) produces an oscillating current in an induction (load) coil, which wraps around a torch consisting of three concentric tubes composed of SiO2 (Fig. 7.4). The induction coil produces an oscillating magnetic field and this in turn promotes an oscillating current in the ions and electrons of the gas.
Fig. 7.3. An inductively coupled argon plasma. Note the induction or load coil wrapping around the end of the plasma torch.
317
Teresa E. Jeffries
Fig. 7.4. The inductively coupled plasma torch. Note the arrangement of gas flows, the induction coil, plasma and cones of the interface region.
To generate and sustain the plasma, gas is introduced tangentially into the two outer tubes of the torch. Initially the gas is seeded with free electrons by passing a high voltage spark along the inner wall of the outer tube. The electrons accelerate in the electromagnetic field, rapidly reaching ionizing energy. They collide with atoms in the gas, transferring energy to create a very high temperature, self-sustaining plasma (Fig. 7.5, adapted after Houk [9]). In addition to sustaining the plasma, the gas flowing through the outer tube of the torch removes the heat dissipated by the plasma to the inside torch walls. This gas flow, which is typically between 10 and 15 l min21, is therefore referred to as the coolant gas or “cool gas.” The gas flowing through the second tube, typically 0 to 1.5 l min21, is referred to as the auxiliary gas, “aux gas” or plasma gas. The flow is not required to maintain the plasma, but is used to control the position of the plasma within the torch. In the innermost tube of the torch, a gas flow of 0.7 to 1.5 l min21 is used to transfer the sample and inject it into the core of the plasma. For this reason it is referred to as the sample carrier, transport gas, or if solution nebulization is implied, the nebulizer gas flow. In the case of ICP-MS argon gas, better than 99.996% pure is used. Argon was discovered by Nobel laureates Sir William Ramsay and Lord
318
Laser ablation inductively coupled plasma mass spectrometry
Fig. 7.5. Temperature variation within an inductively coupled plasma torch.
(John William Strutt) Rayleigh in 1894. It is an inert gas comprising 0.94% of the Earth’s atmosphere. It has three naturally occurring isotopes and currently 12 known radioactive isotopes. However, it is the high ionization potential of argon (15.8 eV) which makes an argon plasma a good source of Mþ ions, with few molecular MOþ or Mþ þ species, and therefore appealing for use in ICP-MS. An approximation of the extent of ionization in an argon plasma can be made, even though the dynamic environment of the plasma is not at thermal equilibrium. Assuming local thermodynamic equilibrium and Maxwell –Boltzmann distributions, use is made of the Saha expression (7.1) (Boumans, [10]): ni ne ð2pme kTÞ3=2 2Zi 2Ei =kT ¼ e na Za h3
ð7:1Þ
where ni is the ion concentration, ne the free electron concentration, nathe atom concentration, me the mass of an electron, k Boltzmann’s constant, T the temperature in Kelvin, h Planck’s constant, Ei the ionization potential, Zi the partition function of the ion, and Za the partition function of the atom. Saha-based percent ion populations under typical plasma conditions for the naturally occurring elements can be found in the literature, e.g., Houk [9].
319
Teresa E. Jeffries
As the electron population is buffered by the amount of argon ionization, the ionization efficiency of the plasma is governed by the ionization potential of argon. Those elements with first ionization potentials greater than that of argon are not efficiently ionized (Table 7.1). Whilst elements with second ionization potentials less than the first ionization potential of argon can be doubly ionized (Table 7.1). All mass analysers separate ions on the basis of their mass/charge ðm=eÞ value (below). For singly charged ions this means the mass/charge value is equal to the mass of the ion. For doubly charged ions, the mass/charge value will be half the mass so that an ion of 138Baþ þ , e.g., will have a mass/charge value of approximately 69, which is close to the mass/charge value of 69Gaþ. For interfering species such as these, a highresolution mass spectrometer will often be able to separate the masses, but quadrupole based mass spectrometers have a low, unit mass resolution and are unable to separate these masses. However, in practice even for the most TABLE 7.1 First and second ionization potentials of the naturally occurring elements Ionization potential (eV)
Elements First ionization: Mþ ions
8–9 9–10 10– 11 11–12
Li, Na, Al, K, Ca, Sc, Ti, V, Cr, Ga, Rb, Sr, Y, Zr, Nb, In, Cs, Ba, La, Ce, Pr, Nd, Sm, Eu, Gd, Tb, Dy, Ho, Er, Tm, Yb, Lu, Hf, Tl, Th, U Mg, Mn, Fe, Co, Ni, Cu, Ge, Mo, Ru, Rh, Ag, Sn, Sb, Ta, W, Re, Pb, Bi B, Si, Pd, Cd, Os, Ir, Pt Be, Zn, As, Se, Au, Te P, S, I, Hg, Rn C, Br
12– 13
Xe
13– 14 14– 15 15– 16 .16
H, O, Cl, Kr N Ar He, F, Ne
,7
7–8
320
Second ionization: Mþ þ ions
Ba, La, Ce, Pr, Nd, Ra Ca, Sr, Sm, Eu, Tb, Dy, Ho, Er Sc, Y, Gd, Tm, Yb, Ac, Th, U Lu, Ti, Zr V, Nb, Hf Mg, Mn, Ge, Pb Rest
Laser ablation inductively coupled plasma mass spectrometry
susceptible elements, Ba and La, the doubly ionized species form only a few percent of their total ion populations in the plasma. From Table 7.1 it can be seen that, in addition to argon, plasmas of the inert gases helium or neon might also be advantageous in ICP-MS. However, quite apart from the higher cost of these gases, their higher ionization potentials give rise to an unacceptably greater population of Mþ þ ions in the plasma for routine use. 7.2.4
The plasma sampling interface
The plasma sampling interface is the region between the plasma and the ion focusing system (below). Figure 7.6 (inset adapted after Jarvis et al., [10]) is a schematic diagram of a typical plasma sampling interface together with an electrostatic ion lens array used for beam focusing. The plasma is sampled via an aperture in the centre of a cooled cone referred to as the “sampler cone.” The sampler cone has a large aperture (1– 2 mm). As the mean free ion path in the plasma (1026 m) is smaller than the aperture in the cone, continuum sampling of the plasma is achieved. Behind this cone, the plasma expands supersonically into a low vacuum area referred to as the expansion chamber. The second cone, known as the skimmer cone, is positioned in the expansion chamber behind the aperture of the sampler cone. The purpose of this cone is to sub-sample, or skim, the expanding plasma through its smaller aperture (0.5 –0.7 mm). As molecular interferences can result from the recombination of ions at the edge of the expanding plasma and in a turbulent downstream, the skimmer cone is closely positioned behind the sampler cone. Apart from the recombination reactions which occur in the supersonic gas jet, other problems arise at the plasma sampling interface. In the absence of a plasma potential, all ions should reach the terminal velocity of argon during expansion and their kinetic energy should be proportional to mass (ke ¼ 0.5 mv2). However, a plasma potential does develop and imparts an undesirable additional kinetic energy and greater range of energies to the ions. The load coil is grounded to mitigate this effect. In addition, the ions have a like charge and mutually repel one another. Light ions are deflected more readily than heavy ions, and this can be a source of matrix effects in ICP-MS. Most ICP-MS instruments have cones composed of high purity Ni for general use, although Cu, Al, ceramic, Pt (or Pt tipped) cones may also be used to assist with specific background and interference reduction.
321
322 Teresa E. Jeffries Fig. 7.6. Schematic diagram of the plasma extraction interface and electrostatic ion lens array used for ion focusing. Not to scale. Inset adapted after Jarvis et al. [10].
Laser ablation inductively coupled plasma mass spectrometry
7.2.5
Ion focusing
Ion focusing is the process of separating the ions from the neutral species, which are pumped away, and transmitting them to the (quadrupole) mass analyser. Ion focusing occurs in a region of high vacuum directly behind the skimmer cone. Until recently most ICP-MS manufacturers used an electrostatic ion lens array to achieve this (Fig. 7.6). The lenses comprise cylindrical or disc shaped electrodes to which a voltage is applied. By changing the voltages applied to each lens they are “tuned” to maximize or minimize the transmission of specific mass/charge values. Typically the lenses would be tuned to maximize the transmission of an ion in the middle of the mass range of interest, whilst minimizing the transmission of oxides or other molecular species. A problem with linear ion lens arrays, is the transmission of stray photons from the plasma, through to the detector. Photons that reach the detector are a source of background noise and decrease its life. Various methods exist to minimize photon transmission. These include offsetting the ion lenses from a straight path or including a barrier, often referred to as a ‘photon stop’, in the central channel and bending the ions around it. Recently, ICP-MS manufacturers have moved away from the electrostatic ion lens arrays and replaced them with multipole collision and reaction cells in their instruments. Multipoles are classified according to the number of rod pairs used, thus a quadrupole, with two pairs is second order, a hexapole is third order and an octopole is a fourth order collision and reaction cell. In operation a reactive or non-reactive gas is slowly added the cell to promote collisions and ion molecule reactions with the sample ions passing through it (Fig. 7.7). One consequence of this is to reduce the axial kinetic energy of the ions. The multipole itself is used as a stability field to prevent loss due to scattering by containing and focusing the products of the high-energy ion molecule reactions. This process is known as collisional focusing. The higher the order of the multipole device, the wider is the stability field for the ions passing through. Although used for ion focusing, the main reason for introducing collision and reaction cells to ICP-MS is to promote ion molecule reactions to reduce and remove interferences. This aspect of collision and reaction cell technology is discussed briefly in section 7.5.1 below. 7.2.6
Quadrupole mass analyser
It is worth noting that the ICP plasma sampling interface and ion focusing systems discussed in the preceding sections could be found on
323
Teresa E. Jeffries
Fig. 7.7. Typical layout of a multipole collision and reaction cell. Not to scale.
almost any ICP-MS instrument, irrespective of the mass analyser used. However, in the remainder of this chapter the discussion will focus on the components, technologies and techniques specific to a generic quadrupole based ICP-MS. The quadrupole mass analyser is used to separate the ions of different mass/charge values (m/e) formed in the ICP. It consists of four Mo rods, typically 12 mm diameter and 230 mm long. The rods are held in a square array such that the inscribed radius r0 produces a field which is a good approximation to the ideal hyperbolic quadrupole field (Fig. 7.8). The pair of rods positioned in the x plane are electrically connected and a positive DC voltage is applied. The rods in the y plane are similarly connected and a negative DC voltage is applied. RF voltages are applied to both pairs in equal amplitude, but 1808 out of phase. Ions reaching the quadrupole from the ion focusing system interact with the applied fields and oscillate. For a given combination of RF and DC voltage applied to the rods, ions of a single m=e value will follow a stable trajectory through the centre of the quadrupole whilst the others will spiral into the rods where they are neutralized. Ion motion in the quadrupole is difficult to imagine but may be considered to have three Cartesian components. The motion along the axis of the quadrupole (z) is constant, the velocity being determined by ion energy. The motion about x and y is complex but the ion may be thought of as spiralling about z, to give a blurred profile when looking
324
Laser ablation inductively coupled plasma mass spectrometry
Fig. 7.8. Quadrupole mass analyzer rod geometry and voltages applied.
end on. The ion beam is approximately symmetrical in x and y when the quadrupole is tuned to the mass peak (Fig. 7.9). When the quadrupole is tuned to a mass just less than the centre of the mass peak, the ion beam becomes flatter in x and wider in y. The opposite occurs on the high mass side of the peak. When the profile is wide enough in one direction, ions collide with the rods and are lost. The fall off each side of the peak is asymmetrical because the nature of ion motion differs in x and y. On the high mass side of the peak, ions pick up energy from RF excitation faster than on the low mass side. Thus the peaks in quadrupole based mass spectrometry are skewed and tail more on the low mass side. Before use, it is necessary to determine the required digital to analogue conversion (DAC) setting of the quadrupole control system, to set it at any given mass. This process is known as mass calibration and in practice, this is
Fig. 7.9. Schematic representation of ion motion in a quadrupole mass analyzer.
325
Teresa E. Jeffries
performed using approximately 10 points spanning the mass range (typically 5 – 250 amu). As already stated, change in the mass/charge value transmitted by the quadrupole is achieved by changing the DC and RF voltages applied to the rods. This allows data to be collected in several modes of operation (Fig. 7.10): Single ion monitoring. In this mode the DC and RF voltages are fixed to transmit the m=e value of a single ion of interest, (Fig. 7.10a). This mode is useful for tuning the instrument and gives the lowest possible detection limits. Scanning or surveying. A pre-determined mass range(s) is scanned rapidly (up to approximately 2500 amu s21) many times in succession (Fig. 7.10b). To achieve this, the DC and RF voltages are changed continuously under computer control. This method is used in laser ablation to ascertain the presence of the analytes of interest before analysis proper. It is also useful for the characterization of interferences. Peak jumping or hopping mode. This mode is sometimes referred to as multiple ion monitoring (Fig. 7.10c). Selected m=e values are transmitted in turn, jumping from peak to peak. This is a rapid method of analysis and
Fig. 7.10. Schematic representation of the main modes of operation using a quadrupole mass analyzer: (a) single ion monitoring; (b) scanning or surveying; (c) peak jumping or hopping; (d) time resolved analysis.
326
Laser ablation inductively coupled plasma mass spectrometry
as time is not wasted in detecting ions of no interest, it leads to low detection limits. It is useful for low sample volumes, however the elements of interest must be defined before the analysis can start. Time resolved analysis. The quadrupole is operated in any of the above modes, but the data are stored in time slices to give a real time component (Fig. 7.10d). Used in conjunction with peak jumping, this is the preferred mode of operation during analysis by laser ablation ICP-MS. 7.2.7
The vacuum system
A mass spectrometer will not function without a vacuum system because at atmospheric pressure (,100,000 Pa), the mean free path of an ion (the average distance an ion can travel before it hits another particle) is only 0.1 mm (see Table 7.2, adapted after Longerich and Diegor [11]). Quadrupole based ICP-MS instruments typically operate with a vacuum pressure of 1026 mBar (1024 Pa) giving rise to a reasonable mean free ion path length of 50 m. In a closed system there would be little difficulty in maintaining this level of vacuum. However, it should be remembered that the ICP-MS is an open system; the ICP resides outside the vacuum envelope but along with the sample ions, a large volume of gas flows through the apertures in the interface cones. Maintaining a vacuum under these conditions is difficult and for this reason instrument failure is often attributable to vacuum problems. TABLE 7.2 Mean free ion path length at pressure, adapted after Longerich and Diegor [11] Mean free ion patha
Pressure Pa
mBar
m
100,000 100 5 0.001 1024 1025 1026
1000 1 0.05 1025 1026 1027 1028
1027 (0.1 mm) 0.00005 (0.05 mm) 0.001 (1 mm) 5 50 500 5000
As most ICP-MS manufacturers give pressure readings in mBar, the pressures are given in both the SI unit (Pa) and in mBar. a Values for N2 or O2 approximated at 258C and molecular diameter of 370 pm.
327
Teresa E. Jeffries
Fig. 7.11. Schematic layout of a typical vacuum envelope, pressures attained and arrangement of pumps in a quadrupole ICP-MS.
Figure 7.11 shows a typical vacuum envelope, the pressures attained and the arrangement of pumps in a quadrupole ICP-MS. Most manufacturers use a combination of mechanical rotary (positive displacement) pumps and turbo-molecular pumps. In older instruments oil vapour diffusion pumps are used instead of turbo-molecular pumps. Rotary pumps are effective from one atmosphere to 1024 mBar. They consist of an eccentric wheel with vanes on the perimeter, lubricated with oil. As the wheel rotates, the vanes trap residual gas molecules from the vacuum vessel against the pump wall. The molecules are transported to an exhaust through a one-way valve. Rotary pumps are very noisy and the oil they use is a potential contaminant of the ICP-MS on system failure. Turbo-molecular (turbo) pumps are effective from 1023 to 1029 mBar and are used to turbo-charge the rotary pumps that back them up. They operate on the principle of momentum transfer. Vanes, mounted on bearings, rotate at between 30,000 and 100,000 rpm. As they spin they hit residual gas molecules and drive them to the intake of a rotary pump. Because the number of molecules increases at the intake of the rotary, so the pressure increases back to the normal working range of the rotary. Unlike rotary pumps, turbo-molecular pumps are clean and quiet.
328
Laser ablation inductively coupled plasma mass spectrometry
7.2.8
Ion detection and signal handling
ICP-MS manufacturers have employed a variety of ion detection systems in their instruments. Most modern quadrupole instruments use detectors based on the electron-multiplier. Electron-multiplier tubes are similar in design to photo-multiplier tubes, consisting of a series of biased dynodes that eject secondary electrons when they are struck by an ion, thereby multiplying the current. Electron-multipliers have a large dynamic range of 9 – 10 orders and a low noise gain. Instead of electron-multipliers, some manufacturers use Daly detectors in their quadrupole instruments. Daly detectors consist of an aluminium knob held at negative potential. When this is struck by a positive ion, secondary electrons are ejected and accelerated into a scintillator. Photons produced at the scintillator are detected by a sealed photo-multiplier. Both the Daly and the electron-multiplier detectors are capable of operation in pulse counting (or digital) mode and in analogue mode. The detectors used in modern ICP-MS instruments, may be operated simultaneously in both modes, a process referred to as dual mode detection. Ions are counted in pulse counting mode and this mode is the optimum for low intensity signals of between 0 and 1,000,000 counts per second (cps). Pulse counting detectors are not 100% efficient because at high ion count rates the detector cannot measure an incoming pulse when it is recording the previous pulse. This phenomenon is known as detector dead time. In analogue mode the ion beam current (amperage) is converted to a potential (voltage) and then to a number using analogue to digital conversion hardware. Analogue mode is best suited to signals between 1000 and 1,000,000,000 equivalent cps. It becomes preferable to pulse counting mode at around 1,000,000 cps, when dead time considerations are of significance. In dual mode operation a detector cross calibration is required to convert the analogue value into an equivalent digital count rate. In practice this is performed by taking measurements over the range where both modes are operational (100,000 – 1,000,000 cps) and, because of the mass dependence of this process, for several elements across the mass range. 7.3 7.3.1
LASER ABLATION: ESSENTIAL COMPONENTS Development of the laser
Lasers are everywhere. For most individuals in the developed world interaction with lasers is an often unconsidered part of everyday life.
329
Teresa E. Jeffries
We find them in our hi-fi systems, our computers and at the supermarket check-outs. Common surgical and dental procedures are aided by lasers. They are used to clean artefacts and monuments and remove graffiti from buildings. Yet, to many, the mention of the acronym LASER (light amplification by stimulated emission of radiation) is often also associated by the general public with images of (high tech) destruction. For the analytical chemist who could concoct many an explosive cocktail from the commonplace reagents in his laboratory, the laser is viewed as it should be; a tool used to assist in the analysis of solids. The intriguing story of the development of lasers centres on four characters [12]: Charles H. Townes, inventor of the microwave emitting maser at Columbia University, USA; his long time collaborator Arthur L. Schawlow, initially a postdoctoral fellow to Townes at Columbia who later moved to the Bell Laboratories; Gordon Gould, an PhD student at Columbia University and Theodore Maiman, working at the Hughes Research Laboratories in California, USA. Following the proliferation of (microwave emitting) masers invented by Townes in the mid 1950s Schawlow, now at the Bell Laboratories, and Townes turned their attention to the laser, or as they proposed, the optical maser. The physical differences between visible light and microwaves meant the transition from maser to laser was not simple. The pair spent many months investigating this problem, eventually detailing their proposal for building an optical maser in 1958, and filing for a US patent which was granted in 1960. With the backing of the Bell Laboratories, the achievements of the pair were soon famous amongst the scientific community. At the same time Gordon Gould was quietly considering the same problems and in 1957 he coined the term laser and its definition. He wrote down his ideas for the laser in his notebook and, in the hope of gaining a patent, had them witnessed by (an unrelated) sweet shop owner, Jack Gould. The race to build a working laser had begun. Most groups concentrated their efforts in producing laser action in alkali metal vapours because their energy level structures were already well characterized. With the exception of funding from the Bell Laboratories for Schawlow and Townes, most groups were working with limited budgets. Gould, knowing that he could not continue working on his PhD and his laser ideas at the same time, quit Columbia and joined TRG Inc, a small company on Long Island, NY. Using his ideas, the company was able to secure high level funding from the American Department of Defence who had visions of developing laser weapons. With such funding
330
Laser ablation inductively coupled plasma mass spectrometry
Gould was able to consider laser action in many materials, but could not obtain security clearance or details of the research the scientists at TRG were pursuing to build a laser, based on his ideas. Working at the Hughes Laboratories in California, Maiman, who had considerable knowledge of ruby masers, started to look at the possibility of laser action in ruby. Schawlow had dismissed this idea in 1959 based on erroneous data and continued with alkali metal vapours. But, continuing undeterred, Maiman succeeded in making the ruby laser work for the first time on 16th May 1960 and the race was won. Theodore Maiman’s laser was beautifully simple; a rod of ruby with silvered ends placed in a coiled flashlamp. Maiman went on to receive the prestigious Japan Prize. It was Townes who was awarded the Nobel prize in 1964 for his work on laser theory. Schawlow went on to win a Nobel prize in 1981 for his contribution to the development of laser spectroscopy. Meanwhile, Gould, whose contributions to laser theory had, in the main, been overlooked, quietly continued in pursuit of his patent. In 1973 the original Townes and Schawlow patent was invalidated and Gould was granted a number of patents on laser applications and theories. The royalties he collected made him a dollar multimillionaire.
7.3.2
The association of lasers with ICP-MS
Since the invention of the laser, laser action has been demonstrated in many materials. However, only a few types of two classes of laser have been used in conjunction with ICP-MS. Two solid state lasers have been used in LA-ICP-MS, the ruby laser, now obsolete for this use and the Nd:YAG laser (a rod of yttrium aluminium garnet doped with neodymium) the most commonly used laser. Gas lasers are represented by the excimer laser (excited dimer). Table 7.3 lists the wavelengths associated with the lasers used in LA-ICP-MS. The use of lasers to introduce solids to inductively coupled plasma mass spectrometers began 25 years after Maiman created the first working laser. Using a ruby laser, working in the infrared (IR), Gray [13] tested its potential to introduce geological materials to the ICP-MS, but found that the laser was unstable. Shortly afterwards, Arrowsmith [14] employed the more stable Nd:YAG laser to introduce metals to the ICP-MS. At first these were used only as bulk sampling devices [15,16], incapable of meeting the growing demand of micro-sampling with resolutions of .100 mm. Then, in 1992, two groups [17,18] simultaneously modified the bulk sampling Nd:YAG laser,
331
Teresa E. Jeffries TABLE 7.3 Lasers used in LA-ICP-MS and associated wavelengths Solid state lasers Ruby laser: Nd:YAG:
Gas lasers Excimer:
Fundamental Fundamental Second Harmonic Third harmonic Fourth Harmonic Fifth Harmonic
694 nm (IR) obsolete 1064 nm (IR) ICP-AESa 532 nm (visible) obsolete 355 nm (near UV) unused 266 nm (UV) 213 nm (far UV)
KrF ArF
248 nm (UV) rarely used 198 nm (deep UV)
a
The fundamental of the Nd:YAG laser, now seldom used with LA-ICP-MS, has recently been re-introduced as a solid sampling device for ICP-AES.
using beam and energy attenuation to produce a laser beam of approximately 40 mm. This attenuated IR laser soon proved to have a much wider application [19], but a number of other problems were soon apparent. Significant among them were poor ablation characteristics [20] and laser induced elemental fractionation [21 – 25]. A catastrophic ablation, when a large fragment of the sample, too large to be transported to the plasma, was removed in a single shot, often resulted from ablation using these IR lasers. IR laser ablation also led to chemical fractionation when, during the course of an analysis, some elements apparently increased or decreased in concentration relative to others, resulting in poor analytical precision. With these difficulties in mind, other lasers were investigated for use with ICP-MS. Shibata et al. [26], tried doubling the frequency (equivalent to halving the wavelength) of a Nd:YAG laser so that it operated in the visible at 532 nm (green). This, however, did not offer any advantage over IR lasers and was abandoned. It was considered by most investigators that lasers operating in the ultraviolet (UV) would provide much better ablation characteristics and show reduced fractionation effects. This was confirmed by the introduction of the frequency quadrupled Nd:YAG laser operating at 266 nm in the UV [27 – 28]. The 266 nm Nd:YAG laser is compact and inexpensive and is now very widely used with ICP-MS. Nevertheless, it is not without problems and for some materials that are transparent at 266 nm, this laser is difficult to use. For this reason, two other UV lasers have been introduced: the
332
Laser ablation inductively coupled plasma mass spectrometry
UV ArF excimer laser, operating at 193 nm, introduced in 1997 [29,30], and the 213 nm frequency quintupled Nd:YAG, introduced in 1998 [31]. The 193 nm excimer laser has superior coupling efficiency in most materials. The laser, adapted for use with ICP-MS, is available commercially but, because of its size and expense, it is not yet widely used. Similarly, the 213 nm frequency quintupled Nd:YAG laser is superior to the 266 nm laser in its coupling efficiency and, like the 266 nm laser, it is compact and inexpensive. Consequently this laser is finding wide application and has taken over from the 266 nm laser to become the industry standard. The process used to introduce a solid to an ICP-MS by laser ablation can be summarized quite simply. A pulsed laser beam is directed at, focused on and fired at a sample, which is contained in an airtight cell. A suitable carrier or transport gas (normally He, Ar or a mixture of both) flows through the cell. Energy from the laser is transferred to the sample and, through various processes including disaggregation and thermal erosion, particles are removed. Some of these are picked up and transported in the flow of gas to the inductively coupled plasma where they are ionized. A basic laser ablation system then should consist of a laser, together with a means of controlling its energy output, some optics for steering and focusing the beam, a sample cell with mounting and movement system, a sample transfer system and for convenience, a means of viewing the sample and ablation event. The popularity and widespread use of the frequency quadrupled and quintupled Nd:YAG lasers with ICP-MS has meant that these lasers have been used almost to the exclusion of others for the analysis of cultural heritage materials. Therefore, in the remainder of this discussion, the components which go to make up a typical Nd:YAG system are described. Whilst the major components in any frequency converted Nd:YAG system are broadly similar in function, there is considerable variation in, e.g., lasing rod geometry, ancillary optics and system layout. For this reason the discussion will concentrate on the function of each component. Figure 7.12 is a schematic showing typical components and layout used in a frequency quintupled Nd:YAG laser (l ¼ 213 nm). 7.3.3
Stimulated emission
The word “laser” is now part of everyday language, but it should be remembered that the name stems from the acronym coined by Gordon Gould: “light amplification by stimulated emission of radiation.” Before looking in
333
Teresa E. Jeffries
Fig. 7.12. Schematic showing typical components and layout used in a frequency quintupled Nd:YAG laser (l ¼ 213 nm).
detail at the components of Nd:YAG lasers used in ICP-MS then, it is worth noting what is meant by stimulated emission of radiation. Substances normally emit light by spontaneous emission, in which excited atoms lose energy in the form of electromagnetic radiation, returning to ground state with no external impetus. When a wave of electromagnetic radiation (photon) emitted by one excited atom strikes another, it stimulates the second atom to emit energy in the form of a second wave, travelling parallel to and in step with the first wave. This stimulated emission results in amplification of the first wave. If the two waves strike other excited atoms, a large population of photons, with the same energy, wavelength, direction, polarization and phase is formed. Or in other words, a laser pulse is generated. 7.3.4
Nd:YAG laser (resonator) cavity
A laser, or resonator cavity consists of the lasing material, a source of electromagnetic radiation and two reflective surfaces in a light-tight housing. For the Nd:YAG laser the lasing material consists of a rod of
334
Laser ablation inductively coupled plasma mass spectrometry
yttrium aluminium garnet (Y3Al5O15) doped with approximately 1% Nd3þ ions. The rod is fixed in a housing along with a suitable source of electromagnetic radiation (Fig. 7.13), typically a linear xenon flashlamp, which is used as an optical pump (below). The rod and flashlamp housing may also contain a filter to protect the rod from UV damage arising from the flashlamp. Typically the area is cooled by a pumped water supply. To complete the resonator cavity, two mirrors are placed at either end of the rod and flashlamp housing, the separation (d) between the two mirrors satisfying the equation: d¼
nl 2
ð7:2Þ
where l is the characteristic wavelength and n an integer. The first mirror, referred to as the rear mirror has a reflectivity of 100%, whilst the second mirror, the front mirror or optical coupler has a reduced reflectivity at the characteristic wavelength of the lasing material (1064 nm). The gas in the flashlamp is held in conduction by the application of an AC voltage, often referred to as the simmer voltage. When the flashlamp is triggered, the short pulse of broadband electromagnetic radiation emitted is coupled to the lasing rod, often by using close coupled ceramic reflectors. The light travels along the rod, reflected back and forth by the end mirrors. Specific wavelengths of the electromagnetic radiation cause electronic transitions to occur in the Nd atoms of the rod, so that some assume an excited state. This process is known as optical pumping. During optical pumping an overproduction of excited Nd atoms develops and, when there is a greater number of excited Nd atoms than those in the ground state, a population inversion is reached. Population inversion
Fig. 7.13. Schematic of a Nd:YAG laser or resonator cavity showing the arrangement of the lasing rod, flashlamp, Pockels cell and mirrors.
335
Teresa E. Jeffries
is an unstable condition so that as soon as it achieved a sudden cascade of electrons return to a lower energy level, with a radiative release of photons. Stimulated emission is induced as these photons encounter further unstable electrons and a laser pulse is generated. For the duration of the flashlamp discharge, this process is repeated many times. So, for a typical Nd:YAG laser a pulse length of some 150 ms is generated. It consists of many spikes, referred to as relaxation oscillations, each a few microseconds in duration. When a laser is operated in this mode, it is referred to as free running or fixed-Q, the “Q” referring to the ‘quality’ of the resonator cavity. In laser ablation ICP-MS it is important that the laser pulse couples well to the sample. Free running lasers exhibit poor coupling and for this reason, in LA-ICP-MS an alternative “Q-switched” mode of laser operation is used. To Q-switch a laser an electro-optical shutter, known as the Pockels cell or Q-switch, is placed between the rod and flashlamp housing and the rear mirror (Fig. 7.13) to alter the quality of the resonator cavity. Pockels cells are comprised of two phase plates (analogous to conventional polarizers) and a compensating block. They manipulate the linear electro optic effect or Pockels effect which describes the variation of the refractive index of an optical medium under the influence of an external electrical field. In this case, certain crystals become birefringent in the direction of the optical axis which, without an applied voltage, is isotropic. When linearly polarized light propagates along the direction of the optical axis of the crystal its state of polarization remains unchanged. However, if a voltage is applied the light exits the crystal in an elliptical state of polarization. In a sense the Pockels effect is analogous to manipulating conventional polarizers; the phase plates introduce a phase shift between the ordinary and extraordinary rays but, unlike conventional polarizing optics, the magnitude of the phase shift can be adjusted with an externally applied voltage. In a Q-switched laser, as the flashlamp is triggered, a voltage is applied to the Pockels cell which causes the two phase plates to be in opposition, i.e., their planes of polarization are at 908. This splits the polarized light into two vibration directions and velocities to cause a destructive interference of the laser pulse such that it does not reach the rear mirror of the resonator cavity. This suppresses laser operation and allows a larger overpopulation of excited Nd atoms to form. After a preset delay, the voltage is removed from one of the phase plates causing it to be “rotated” through 908. This allows oscillation between the two
336
Laser ablation inductively coupled plasma mass spectrometry
mirrors and produces a single, giant laser pulse of a few nanoseconds duration.
7.3.5
Harmonic generation
A Q-switched Nd:YAG laser produces a high intensity pulse train, typically between 1 and 20 Hz, at 1064 nm in the IR. However, coupling efficiency in the IR is poor for many materials and catastrophic ablation is common (Fig. 7.14). To produce an UV laser from the fundamental wavelength of a Nd:YAG laser, the beam must be converted by the process of harmonic generation. Not surprisingly, this process employs the use of harmonic generators, sometimes referred to as non-linear crystals. When high intensity electromagnetic radiation passes through a non-linear crystal, some of the dipole vibrations assume a higher frequency than the input radiation. Thus harmonic generators are sometimes referred to as frequency modifiers. Typically they are used to double the frequency of the input laser beam to produce, in the case of a Nd:YAG, the second and fourth harmonics (523 nm and 266 nm) or, in a process referred to as sum frequency mixing, they are used to mix frequencies to produce the third and fifth harmonics (e.g., mixing the first fundamental beam and the second harmonic to produce
Fig. 7.14. Scanning electron microscope image of a laser ablation crater in calcite, formed by ablation using an infrared laser.
337
Teresa E. Jeffries
Fig. 7.15. Schematic representation of the production of a frequency quadrupled 266 nm and a frequency quintupled 213 nm Nd:YAG laser beam using harmonic generators.
the third harmonic at 355 nm, or mixing the second and third harmonics to produce the fifth harmonic at 213 nm). Figure 7.15 is a schematic representation of this process in the production of a frequency quadrupled 266 nm and a frequency quintupled 213 nm Nd:YAG laser. Several materials are used to produce harmonic generators. Many are dielectric optical non-linear crystals, including KH2PO4 (KDP) and KD2PO4 (KD p P) typically used as frequency doublers and b-BaB2O4 (BBO) used for sum frequency mixing. These materials are hygroscopic, sensitive to input beam angle, expensive and temperature dependent and are normally housed in air-tight, temperature stabilized ovens. For successful harmonic generation the input beam must propagate through the crystal along a unique axis with respect to the crystallographic axis. The crystals are mounted such that they can be angle tuned with respect to the incoming beam—a process referred to as phase matching. However, conversion efficiency is usually poor at less than 50%.
7.3.6
Harmonic separation
The laser beam resulting from harmonic generation is a mixture of all the wavelengths created in addition to the residual fundamental wavelength of
338
Laser ablation inductively coupled plasma mass spectrometry
the laser. Thus it is necessary to separate the desired wavelength from the others, a process known as harmonic separation. There are two methods of harmonic separation commonly associated with Nd:YAG laser ablation instrumentation. The first uses a harmonic separating prism such as a Pellin Broca prism which, when placed in the beam path, causes an angle separation of the harmonics (Fig. 7.16a). In a Pellin Broca prism an ordinary dispersing prism is split in half along the bisector of the apex angle. Using a right angle prism, the two halves are joined to create a dispersing prism with a right angle bend obtained by total internal reflection. The advantages of this method of harmonic separation are that the resultant beam is 100% harmonically pure and the prism is wavelength independent. The disadvantage is that the angle of separation is usually quite small, particularly between the UV wavelengths, so the separated harmonics have to travel some distance before the required one can be isolated.
Fig. 7.16. The process of harmonic separation. (a) Separation of a 213 nm beam using a Pellin Broca prism and (b) separation of a 266 nm beam using dichroic mirrors.
339
Teresa E. Jeffries
The second method, used in the majority of commercial instruments, is to separate the harmonics using wavelength specific dichroic mirrors (Fig. 7.16b). The advantage of this method is that it requires little space. The disadvantage is that each mirror is only 95% efficient, so at least two are required and it is usual to have three or four to provide a beam of sufficient harmonic purity.
7.3.7
Energy attenuation and control
Most laser ablation systems are capable of delivering pulse energies far in excess of those required for ablation (0.01 – 2 mJ for most materials) despite the huge losses of energy associated with harmonic generation. The response of materials under laser ablation is extremely variable being dependent on the physical properties of the material and those of the incoming laser pulse. To achieve a controlled ablation in any material the laser pulse energy must be adjusted appropriately. Arguably then, a means of attenuating and controlling laser energy is the single most important beam modifying component of any laser ablation system. In the past a range of methods has been tried to control laser energy [17,18,32,33] with varying degrees of success. Now most workers in the field agree that the best approach to energy control is to use an optical attenuator that takes advantage of the laser’s natural linear polarization [22,31]. Such a device could consist of a rotatable half wave plate (typically quartz) and a polarizer manufactured from wavelength appropriate material (e.g., magnesium fluoride for the deep UV). The polarizer only allows transmission of radiation polarized in one direction, in a manner similar to a polarizing microscope. So that the remaining orthogonally polarized radiation is rejected. By placing a wave plate in front of the polarizer the polarization direction of the laser beam can be rotated so that more or less of the beam passes through the polarizer, thereby creating a sensitive variable optical attenuator. Good laser systems include a beam shutter which can be moved into and out of the beam path. This allows the user to fire the laser continuously and so maintain thermal equilibrium. In addition a means of monitoring output energy is an essential part of laser ablation instrumentation. Such monitoring allows the user to maintain a consistent standardized approach to their analysis and is an essential tool in the diagnosis of faults.
340
Laser ablation inductively coupled plasma mass spectrometry
7.3.8
Beam delivery and viewing optics
Important considerations in laser design are those of delivering the beam to the sample and viewing both the sample and the ablation event. The optics used to deliver the beam vary in their sophistication from one instrument to another. The process of getting the beam to the sample, beam steering, is fairly simple and one that is usually achieved with mirrors or steering prisms. A focusing objective is used for both viewing the sample and focusing the beam onto it. A simple plano-convex lens is really all that is necessary for bulk sampling but, for high resolution work, a higher powered reflecting or refracting objective is normally used. Refracting objectives in general have a lower damage threshold than reflecting objectives. Dielectric coated reflecting objectives are free from the chromatic aberrations of refracting objectives and give excellent image quality. However, they are prone to damage, expensive and are restricted to use at only one wavelength. Compound air-spaced refracting objectives are now being used on modern laser ablation systems. These use two different UV transmitting materials with different optical properties. The focal length of these objectives is independent of wavelength (achromatic) so, in addition to being usable over a large wavelength region, they focus the laser and give a good image. In addition to controlling pulse energy, it is equally important to have control over the spot size used for ablation. This is achieved on modern systems by imaging a variable or interchangeable aperture onto the sample surface. By changing the size of the aperture a proportional change in spot size is achieved. Most commercial laser ablation systems now have a flat-top beam profile. Nd:YAG lasers typically output a near Gaussian beam, but by using an output coupler with graduated transmission or beam homogenizing systems, such as two multielement lens arrays, a flat-top beam profile is created. This is of particular importance in the reduction of laser induced elemental fractionation and during depth profiling. Using aperture imaging with such a system has the advantage of maintaining a fixed energy density at the sample surface, irrespective of the spot size. Some commercial systems also incorporate a beam expanding telescope to assist with spot size control. When the spacing between the two lens elements of the beam expander is increased, the output beam becomes more convergent. The effect of this is to raise the focal point of the beam, effectively “defocusing” or spreading it. Similarly, the beam could be defocused with its focal point below the sample
341
Teresa E. Jeffries
surface but focusing the beam within the sample is more likely to cause a catastrophic ablation. Good laser ablation systems should include a CCD camera in their viewing optics to allow safe viewing of the sample during ablation. Normally, the output from the camera is configured such that the beam is correctly focused when the image is in focus. A zoom lens on the camera will assist the user in sample location. A variety of light sources can be a great aid in sample identification. Typically reflected, transmitted and general illumination are included. 7.3.9
Ablation cell and sample transport
The ablation cell fulfils the seemingly simple function of housing the sample during ablation, yet there is probably greater diversity of design in laser ablation cell construction than in any other component of a laser ablation system This must be, in part, a consequence of the array of applications for which LA-ICP-MS is used. It is quite common to find users abandoning the ablation cells provided with their system in favour of an in-house cell adapted for the specific application. An obvious design goal for an ablation cell is good gas flow dynamics resulting in efficient sample transport. Interestingly, although proposed several times, no clear relationship between cell volume and transport efficiency has yet been proven. Despite the variations in cell design, there is a number of features which all ablation cells must have. They must allow the laser energy to enter through a laser window transparent to the incoming wavelength. Often these windows are tilted or anti-reflection coated to reduce stray back reflections which damage the objective lens. The cell must include entry and exit ports for the sample carrier gas to transport ablated material to the plasma. Importantly, the cell must be gas tight to avoid loss of sample or the introduction of atmospheric gases. Often little attention is given to this issue during cell design and the consequences are both poor sensitivity and high background signals. Additionally, ablation cells may also have a second window in the base to allow sample viewing by transmitted light. They may have a facility for gas mixing (e.g., He or Ar þ He) situated either down- or upstream of the cell. A valve (solenoid) system to divert the gas flow and permit sample changeover without extinguishing the plasma may also be included. Finally, they may be mounted on a motorized XY or XYZ stage for sample movement.
342
Laser ablation inductively coupled plasma mass spectrometry
When a cell is opened to load/remove a sample, the plasma is disrupted, sometimes resulting in instrument drift. Therefore, a well designed cell should be of sufficient diameter to accommodate one or more samples together with all the calibration and reference materials necessary to complete an analytical run. Connecting the laser to the ICP-MS in order to transport the sample to the plasma is a simple process. All that is required is a length of plastic (e.g., polyurethane or similar) tubing connected to the exit of the ablation cell and the ICP torch injector. The tubing used is typically around 6 mm external diameter, 3 mm internal diameter and between 1 and 2 m in length.
7.4
ANALYTICAL CONCEPTS AND FACTORS AFFECTING ANALYSIS
In this section the laser ablation system and ICP-MS are brought together in a discussion of some of the important factors to be considered before undertaking an analysis. These factors are numerous and often dependent on the application and material to be analysed. For this reason, the discussion will be confined to those issues that are important and more or less general to all applications.
7.4.1
Why use the technique?
The first questions a user is likely to have when considering the use of LAICP-MS are often concerned with the sort of data likely to be obtained, how long the analysis will take, sample’s suitability and so on. In other words, why use this technique? The advantages of using LA-ICP-MS for the analysis of solids are fairly clear. Samples are analysed in the solid state and so there is no need for lengthy digestion and dissolution steps. Most inorganic solids can be analysed, along with some organic solids such as wood and body tissues. Spatial information can be obtained, at resolutions varying between five and a few hundred microns. The technique is extremely sensitive, providing lowlevel trace element data for almost all elements from Li to U, in a single analysis which could take just a minute or two. Isotope ratios may be determined. Samples, although almost always prepared, can on occasion be analysed without any preparative steps. The technique is less prone to
343
Teresa E. Jeffries
interferences than many other solid sampling techniques such as ion probe microanalysis or electron microprobe analysis. The disadvantages of the technique are not so immediately obvious except that it is sample destructive on a microscopic scale. It is not a primary analytical technique and requires the use of an internal standard to correct for variation in ablation yield between standards and samples. This internal standard must be measured beforehand by a technique such as electron microprobe analysis. With modern systems, phenomena such as laserinduced elemental fractionation are less of a problem than in the past, but remain a challenge. The physics and chemistry of laser– sample interaction are poorly understood. Finally, as much an advantage as a disadvantage, the technique is young and still in rapid development. 7.4.2
Sample preparation and mounting
It is true that for LA-ICP-MS no sample preparation techniques are necessary. All that is required is that the sample has a reasonably flat area at a similar scale to the sampling resolution required and that it fits the ablation cell. This would seem ideal for precious or rare cultural heritage material. However, it is also true that samples are rarely presented without some form of preparation. This is partly because it is probable that the samples will have been analysed for major and minor elements by electron microprobe prior to laser ablation. Indeed, without a prior determination of a major or minor element by a technique such as this, or using an assumed concentration for one, the data from LA-ICPMS will only be qualitative. Additionally accuracy is improved if samples and calibration standards are prepared in the same fashion. For most samples, the route to successful analysis by LA-ICP-MS begins with good sample preparation. For spatially resolved microbeam measurements or bulk analysis of homogeneous material, samples are usually presented as polished, approximately 100 mm thick, thin sections, or as polished slices, blocks or resin mounted pieces. Heterogeneous samples are usually crushed to a fine powder then pressed into a pellet to determine their bulk composition. In these cases a large spot size is used and often rastered over the sample in an attempt to resolve any residual heterogeneity. Samples do not need to be in electrically conducting and any previous carbon or other coatings should be removed prior to analysis. Care needs to be exercized during sample preparation and subsequent handling to avoid contamination.
344
Laser ablation inductively coupled plasma mass spectrometry
Whenever it is possible, all samples and standards are mounted flat and flush with the mounting surface in the ablation cell. Whilst this is not crucial to the analysis, it does mean that the material is easier to view and focus on and that the gas flow dynamics through the cell are preserved. Laser ablation sends a shock-wave through the ablation cell. This, coupled with the general vibration caused by pumps and so forth, means that it is usual to clamp or fix the samples in the cell. Any movement of the sample will degrade the sampling resolution that can be achieved. Volatile, “sticky putty” mounting media (e.g., blu-tace) are normally avoided as they warm and deform in the cell causing movement of the sample. Some of these products are known to outgas and contaminate the system. 7.4.3
Analysis of transient signals
In a typical analysis by laser ablation ICP-MS data are collected for approximately 2 min. The first minute is usually devoted to the collection of background signals arising from electronic noise, the carrier and plasma gases and so on, without the laser firing. These background data vary with time, but are reasonably constant throughout the minute of collection. In the second minute the laser fires at the sample, ablating further and further in with each shot. For this part of the analysis the data recorded at the detector can fluctuate greatly with time, i.e., the signals are transient. There are many reasons for fluctuations to occur including sensitivity fluctuations in the instrument, changes in laser pulse energy from shot to shot, variations in the volume of material ablated as focusing depth and physical changes occur at the laser sample interface, laser induced elemental fractionation, heterogeneity or zonation in the sample, surface contamination from sample preparation, previous ablation, fingerprints or remnants of conductive coating, encountering a phase change in the material at depth such as an inclusion or organic layer, and at structural changes in the sample such as cleavages or fractures. It is imperative, therefore, that signals are monitored, recorded, examined and processed as a function of time. This process is known as real time data acquisition or time resolved analysis. The importance of real time data acquisition cannot be overstated. It is frequently necessary, even for homogeneous samples, to examine the data as a function of time to exclude any heavily fractionated part. For heterogeneous samples real time data acquisition is essential. To illustrate this point, Fig. 7.17a shows a real time data plot of the analysis of an opaque
345
Teresa E. Jeffries
Fig. 7.17. Time resolved data acquisition. (a) Time resolved spectra obtained during the analysis of an opaque mineral. Inset is a schematic cross section through the sample; (b) Plane polarized light photomicrograph of the ablated mineral from which the data in (a) were obtained. Note the mineral apatite is underlying the magnetite to the left of this image. The apatite was not ablated, but calcite, directly underlying the magnetite was sampled in the latter part of the analysis.
346
Laser ablation inductively coupled plasma mass spectrometry
mineral (magnetite) in a rock. The sample, a carbonatite, has a mineral assemblage consisting chiefly of calcite, interspersed and included with grains of apatite, magnetite, pyroxene and garnet. A 40 mm thick section of the rock was prepared and mounted on a glass slide using a bonding resin which had been spiked with Bi. Laser ablation started after approximately 50 s of background signal collection. After 10 s of ablation, a second phase is encountered. The second phase (calcite) is underlying the opaque magnetite, but its presence could not be determined from visual examination (Fig. 7.17b). After a further 15 s, a rise in the signal for Bi, from the bonding material closely followed by a rise in signal for Zr, which was present in the supporting glass slide, indicates that the section has been ablated right through. Interestingly, high signals from Ca, Ce and La indicate that the crater walls continue to be ablated long after the section has been penetrated. With such a large difference in chemistry between the magnetite and the calcite, had the analyst relied on the visual appearance of the sample and integrated all the data from the analysis, assuming it all to come from the magnetite, a large error in the data would have arisen. Collecting and processing time resolved data from transient signals although essential to laser ablation ICP-MS, presents a number of problems for the analyst. The mass spectrometer must be capable of fast scanning and real time reporting, external calibration is not possible, internal measurement precision is difficult to determine, detection limits vary continuously and data reduction is reliant on the interpretative skills of the analyst. 7.4.4
Factors affecting analysis
Data acquisition by laser ablation ICP-MS is a complex procedure and there are several instrumental parameters and other factors that affect the outcome of the analysis. As one might expect, these usually fall into one of two categories: those arising at or affecting the ICP-MS and those attributable to the process of laser ablation. Choosing appropriate instrumental parameters and data acquisition protocols are amongst the most important decisions the analyst needs to make to be certain of a successful analysis. The choices will vary between applications and sample matrices. Compromises are common and there is often more than one right choice. There are no recipes to follow and often the analyst will frequently rely on the intuition developed by experience. The ICP-MS parameters which must be considered include “sweep time,” “quadrupole settling time,” “dwell time” and “number of sweeps per reading.”
347
Teresa E. Jeffries
Sweep time is the time required for the acquisition of intensity data from the complete list of isotopes (elements) to be determined. Quadrupole mass spectrometers are sequential instruments and measure from one mass to the next in turn. Sweep time reflects a single pass across the mass range of interest. Sweep time is made up principally from quadrupole settling time and dwell time. The greater the number of isotopes to be determined, the longer the sweep time. Sweep time is not a user set parameter, but the components that make it up are, however, it is important that the time is as short as possible, as real variation in sample chemistry could be missed. During sample acquisition several sweeps across the mass range are made. The number of sweeps may be set or calculated as a function of a user set acquisition time. The greater the acquisition time or larger the number of sweeps in the sample, the better the counting statistics. A minute of background data acquisition followed by a minute of sample data acquisition is usually more than sufficient. Quadrupole or spectrometer settling time is the time allowed for the detector and quadrupole to stabilize following a jump from one selected mass to the next. No ions are counted during this jump, so it is essentially wasted time and because of this it needs to be kept to a minimum, but not so short that settling does not take place fully. The settling time required after each jump depends on the size of the jump made and is typically 1 ms but varies between approximately 0.1 and 10 ms. The largest jump to make is often between the last isotope measured and the first in the sequence, e.g., between 238U and 7Li. On modern instruments a settling time can be set for each jump. On older instruments this may not be possible and in this case rather than using the settling time required for the longest jump for all the jumps, a shorter time is used and unused masses are placed in the sequence where long jumps would otherwise occur. For example, the jump between 238 U and 7Li would be replaced with a jump between 238U and mass 5 where it does not matter that the quadrupole has not settled fully, and then a small jump from mass 5 to 7Li would follow. When dealing with transient data, the time spent measuring at each isotope should be as short as possible, but long enough to maintain an acceptable count rate. This time is referred to as dwell time. Dwell times set for laser ablation are typically in the region of 10 ms, so that with a 1 ms quadrupole settling time, 90% of the time spent at each isotope is spent counting. Longer dwell times would not significantly increase the proportion of time spent counting. Shorter dwell times might seem desirable as they would increase the number of sweeps, but they would undesirably decrease the proportion of time spent counting. It is common practice to keep dwell
348
Laser ablation inductively coupled plasma mass spectrometry
times synchronous with the AC cycle time to reduce noise from the power supply. Thus, in the UK a dwell time of 10 ms is preferred, it being a half cycle at 50 Hz, i.e., one 100th of a second. In the US a dwell time of 8.33 ms might be preferred, that being a half cycle of 60 Hz, or one 120th of a second. It is worth noting that with a dwell time of 10 ms, 1 ion is 100 cps, a value far greater than the continuum background of approximately 10 cps for a quadrupole instrument. For an element list of 30 isotopes using a 10 ms dwell time and a 1 ms quadrupole settling time, the counting efficiency is 90%. The total sweep time is 330 ms resulting in approximately three sweeps per second. For a shorter list of four isotopes, as might be the case for isotope ratio determinations, the total sweep time is 44 ms, giving approximately 23 sweeps per second. To reduce the noise during data reduction, it is usual to combine or integrate together a number of sweeps into a single reading. The number of sweeps per reading ought to reflect a careful consideration of counting statistics in both the background noise and sample signal and the effect of dwell time. In practice it is usual to combine at approximately 1 s intervals. In addition to those parameters and considerations associated with the ICP-MS, there are several more related to the laser. These include laser pulse energy, beam diameter or spot size, the repetition or firing rate of the laser, power density which is the energy over a given beam size, the focus position of the laser, depth penetration rate, transport gas and whether the sample remains in a fixed position relative to the beam or is moved during the analysis. In setting these parameters the analyst must consider that the best spatial resolution (smallest craters) yield less ablated material and therefore give poorer detection limits, accuracy and precision. Higher energy densities may improve detection limits in a small crater, but laser induced elemental fractionation could be worse, the depth penetration rate will be higher and for thin materials, the signal duration could be short. It might also be difficult to relate material removed from deep within a crater to internal standards. Much of the goal in choosing appropriate laser parameters is in minimizing laser induced elemental fractionation. There has been much written on the subject of laser induced elemental fractionation and although still far from understood, it is probably the most discussed subject at LA-ICP-MS conferences and meetings. It is beyond the scope of this chapter to discuss the phenomenon fully, but the reader is encouraged to look at the recent literature on this subject [22– 24,30 – 31,34 –45].
349
Teresa E. Jeffries
The following generalizations regarding fractionation summarize these discussions: fractionation starts as the result of thermal processes at the ablation site [23]. It is the process by which non-representative sampling occurs during the course of an analysis. Volatile elements such as Pb, Cd and Zn are fractionated more readily that the refractory elements such as Ca, Zr, Y and the REE [22– 24,31]. Fractionation occurs at the ablation site, during sample transport and in the ICP. It is both time and matrix dependent but the matrix dependency is decreased when using lasers of shorter wavelengths and pulse widths [45]. A small change in laser focusing can have a large effect on fractionation and using large diameter shallow craters minimizes time-dependent fractionation [42]. Using a high velocity jet of gas at the point of ablation reduces the effect of fractionation [35]. Transport gas can have a significant effect on fractionation behaviour and whilst ablation in He does not lessen the magnitude of the fractionation, it improves the reproducibility of the fractionation trends [45]. 7.4.5
Optimization and calibration
For new users, optimizing or tuning an ICP-MS for laser ablation can seem at first an impossible task. Without any points of reference a de-tuned instrument may take several hours to tune. Tuning is an iterative process, in which each parameter affects the others. Instrument tuning is both mass and matrix dependent. Several parameters require daily optimization, including torch position in X, Y and Z, and the gas flows (coolant, auxiliary and the carrier gas or gases). In an ion lens system, each ion lens will need tuning. The RF output (forward power) and quadrupole bias must be set. Less often, perhaps weekly to monthly depending on the diversity of the applications and the age of the detector, a current detector plateau needs to be found, a detector cross calibration and mass calibration need to be performed. On most modern systems, several isotopes may be observed and tuned simultaneously in addition to interference masses such as doubly charged or molecular species. This allows the analyst to attempt to minimize the fractionation, interferences and the signals for elements in a mass range which is of no interest or where the elemental concentration is expected to be high. At the same time the analyst will attempt to increase the sensitivity in the mass range containing elements of interest whose concentration is expected to be low. For example, an analyst may wish to provenance a glass vessel by considering the pattern of trace elements
350
Laser ablation inductively coupled plasma mass spectrometry
and low level REE present. Calcium, present at 10 wt% CaO has been chosen as the internal standard. The analyst will attempt to maximize the signal for the REE at the expense of the signal for Ca. At the same time the interference from BaO on the REE and from Baþ þ on 69Ga will be monitored to keep this at a lowest level possible. The Th/U ratio might also be monitored to determine the level of laser induced elemental fractionation. Tuning an ICP-MS frequently results in a compromise. The most sensitive signal is not necessarily the best tuned. In the case of isotope ratio determinations, e.g., it may be of greater importance to achieve good stability rather than sensitivity. It might be of greater importance to tune out an interference at the sacrifice of sensitivity. Probably the greatest obstacle overcome in the development of the LAICP-MS analytical technique was that of calibration. In the early days of the technique it was thought that finding a suitable calibration strategy for solids, whose ablation characteristics varied greatly, would be difficult, particularly in view of a lack of homogeneous, well characterized trace element reference materials. In the end and to the surprise of many, calibration was found to be quite straightforward for many materials. Three basic methods of calibration are in current use for LA-ICP-MS. All require that the ablation signal of interest from the sample, with background subtracted, is referenced to the background subtracted ablation signal from a standard reference material. In the simplest of the three calibration methods an external calibration from a solid reference material is performed [13]. Data are background subtracted but are otherwise untreated. The difficulty with this approach is that it is not really suited to fluctuating, transient signals that are frequently a feature of LA-ICP-MS, because differences in ablation volume (yield) within an analysis or between reference materials and samples are uncorrected. To give any approximation to accuracy, samples and reference materials need to be closely matrix matched as must ablation conditions. This technique is seldom used now, but it does provide a useful and rapid sample reconnaissance without the need for internal standardization. The most commonly used approach to calibration involves the use of internal standardization to correct for ablation yield variations and matrix differences. Typically, the internal standard used is a major component of the sample, whose concentration has been determined by an alternative technique such as electron microprobe analysis, or has been assumed. For many materials the major components are elements of comparatively low
351
Teresa E. Jeffries
atomic mass relative to the elements of interest. Typically a minor isotope of these elements is used, so as not to saturate the detector. Using internal standards and analytes of similar ablation behaviour, even in the absence of matrix matched standards and samples, produces remarkably accurate results. The most widely used calibration materials for this purpose are the National Institute of Standards and Technology (NIST) standard reference glasses 612 and 610. In the final of the three basic calibration methods, internal standardization is essential, but in addition the approach requires introduction of reference and other materials using solution nebulization along with the ablated material [27,46,47]. In this approach the sample gas flow from the ablation cell is mixed with aspirated solution from a nebulizer, prior to the ICP torch. During tuning and calibration a reference material is aspirated, but during sample ablation a blank solution is used to maintain similar plasma conditions. The introduction of reference materials in this way facilitates tuning and allows flexibility in the choice of reference material. The drawback of this technique is that by introducing water to an ICP, the advantages of a dry plasma (e.g., low interferences and molecular ion backgrounds) are lost. Some advantage can be regained by desolvating the wet material [48]. The greater problem with this approach is that the difference in the way reference materials and samples enter the ICP will worsen matrix effects, which in turn make correction of elemental fractionation extremely difficult. In addition, mixing the ablated volume with the gas stream coming from the nebulizer can dilute the sample and result in a loss of sensitivity. For these reasons and because highly accurate results can be obtained simply with solid reference materials, this approach to calibration has not been adopted widely. Individual laboratories usually develop their own analytical protocols to suit the range of applications in hand. Typically a calibration standard will be analysed twice at the start of the analytical run and then twice again at the end. A further reference material may be analysed mid-run as an unknown, to check on the accuracy of the procedure. A run is often limited to around 20 determinations including standards, in order to avoid long gaps between calibration standards and the possibility of instrumental drift. 7.4.6
Figures of merit and analytical performance targets
Along with a table of concentrations, it is usual for the analyst to present a number of other statistics and figures of merit pertaining to the analysis
352
Laser ablation inductively coupled plasma mass spectrometry
including details of detection limit, precision and accuracy. These figures are variable, dependent on a number of instrumental and operating parameters and are often difficult to calculate for LA-ICP-MS. For example, precision and detection limits are affected by ablation yield and sample transfer efficiency, the number of isotopes determined, their (isotopic) abundance and ionization potential, the counting time used, the instrument sensitivity at each isotope and the background noise. There are, in fact, so many variables that quoting a set of typical figures of merit is unreasonable. Nevertheless, at a recent conference on laser ablation ICP-MS, (European Laser Workshop, held at The British Geological Survey, 1999) the delegates discussed and agreed a set of analytical performance targets. So for guidance, these are presented here: 7.4.6.1 Detection limit, (3 sigma) Using a 100 mm crater, the limit of detection for a mono-isotopic, easily ionized element in an easily ablated matrix should be better than 10 ppb (parts per billion). Most modern quadrupole based ICP-MS instruments currently achieve unit parts per quadrillion (1 ppq ¼ 1 part in 1015) detection limits in solution analysis. This is the equivalent of detecting in an aliquot of a few millilitres the presence of a single drop of a pure substance (e.g., mercury) homogeneously dispersed in a body of water the volume of loch Ness (Scotland), or a standard (alcohol) measure of mercury (0.025 l) dispersed in lake Baikal, Southern Siberia. 7.4.6.2 Precision The analytical performance target for precision in laser ablation ICP-MS is defined as better than 5% RSD (relative standard deviation) for concentrations of greater than 1 ppm determined using a 100 mm crater. Instrumental drift for a modern instrument in a temperature and humidity stabilized laboratory should be low. Extremely good precision of better than 2% RSD for general work at all concentrations, and 0.5% RSD for isotope ratio determinations can be achieved with careful choice of acquisition parameters. 7.4.6.3 Accuracy The recognized levels of accuracy for LA-ICP-MS are dependent on concentration. For analytes present at less than 1 ppm, an accuracy of less than ^20% is defined as good and greater than ^40% as poor. At concentrations between 1 and 10 ppm, accuracy is good when below ^15%
353
Teresa E. Jeffries
and poor when above ^ 30%. For elements present at concentration levels greater than 10 ppm, accuracy is good when below ^10% and poor at above ^20%. With care, the accuracy of the technique can be improved to approximately ^2% at all concentration levels. To put these statistics in the context of a real example, Table 7.4 and Fig. 7.18 document a case study of the analysis of the United States Geological Survey standard reference material, BCR2-G, a basaltic rock
TABLE 7.4 Comparison of measured concentrations together with errors and certified or working values for USGS BCR2-G Element
Certified value (ppm)
Measured value (ppm)
Error (%)
Cr Co Sr Y Zr La Ce Pr Nd Sm Eu Gd Tb Dy Ho Er Tm Yb Lu Hf Ta Pb Th U
16 37 337 32.51 184.4 25.3 53.6 6.83 28.63 6.67 2 6.8 1.04 6.83 1.29 3.66 0.54 3.34 0.51 5.17 0.78 10.34 5.98 1.7
15.53 37.86 338.78 33.51 190.36 26.33 53.68 6.74 28.82 6.69 1.98 6.83 1.04 6.77 1.3 3.65 0.54 3.33 0.5 5.09 0.8 10.53 6 1.65
3.05 2.27 0.53 2.99 3.13 3.92 0.15 1.3 0.67 0.25 1.1 0.38 0.35 5.78 0.77 0.27 0.68 0.3 2 1.57 2.5 1.83 0.33 3.03
Data were obtained using a single point calibration and internal standardization. The calibration standard used was NIST 612. Data obtained at The Natural History Museum, London, using a 213 nm Nd:YAG laser with a crater diameter of 50 mm.
354
Laser ablation inductively coupled plasma mass spectrometry
Fig. 7.18. Comparison of measured concentrations and certified working values for USGS BCR2-G together with limits of quantification (10 sigma). Data were obtained using a single point calibration and internal standardization. The calibration standard used was NIST 612. Data obtained at The Natural History Museum, London, using a 213 nm Nd:YAG laser with a crater diameter of 50 mm.
which has fused to form a glass. This reference material is frequently used as an accuracy check standard, so data from repeated analyses readily accumulate in many LA-ICP-MS laboratories. 7.5 7.5.1
CONTINUING DEVELOPMENTS AND FINAL REMARKS Continuing developments
Laser ablation ICP-MS is a remarkable technique for the analysis of trace elements in solids, but it is not free from difficulties. Its continued development will include closer study of fundamental problems including laser-induced elemental fractionation and calibration issues, such as matrix effects. Lasers are beginning to be routinely connected to multicollector and time of flight ICP-MS instruments and with this has come the development of new applications particularly in the field of isotope ratio measurements. Perhaps the most significant development in LA-ICP-MS in recent years has been the introduction of collision and reaction cells. Although in their infancy
355
Teresa E. Jeffries
these devices have already had an enormous impact in ICP-MS and it would be a gross omission not to give a brief discussion of their use here. It has been long known that in the vacuum of a mass spectrometer, a charged ion can induce a dipole moment in a non-polar gas molecule or interact with the dipole moment of a polar gas molecule. It is the exploitation of these phenomena that led to the introduction of collision and reaction cells in ICP-MS. In section 7.2.5, collision and reaction cells were discussed in the context of their use for collisional ion focusing but, the main reason for introducing collision and reaction cells to ICP-MS is to promote ion molecule reactions in order to reduce and remove interferences. Reduction of spectroscopic interferences improves the accuracy, precision and detection limits in ICP-MS. In a collision and reaction cell, ions collide and react with gas molecules and in so doing lose kinetic energy, a process referred to as thermalization. The ion molecule gas phase reactions which can occur include neutralization, association and condensation and these reactions continue whilst the ions and molecules remain within the stability field of the cell. The consequence of these reactions is that some polyatomic interferences are broken down. The advantage of using this method for interference reduction (rather than the use of a high resolution mass spectrometer) is that in many cases, reactions proceed without the loss of sensitivity. There are some drawbacks associated with the use of collision and reaction cells. Sometimes unwanted reactions will suppress analyte ion transmission and lead to a loss of sensitivity. They may also give rise to the production of additional interferences. It is beyond the scope of this chapter to give a full discussion of collision and reaction cell technology. Mason [49] provides an in-depth and accessible discussion, and the interested reader is encouraged to start here. The use of multipoles in ICP-MS is developing rapidly. At the time of writing, of those ICP-MS manufacturers using multipoles in their instruments, one is using a quadrupole, two are using hexapoles and one is using an octopole. This technology has revolutionized ICP-MS and its continued development is eagerly awaited. 7.5.2
Final remarks
Laser ablation ICP-MS has undergone a rapid development since its introduction in 1985. The advantages of the technique are clear: its speed, flexibility, sensitivity, accuracy, and the diversity of applications for which it
356
Laser ablation inductively coupled plasma mass spectrometry
may be used. Of course the technique cannot alone satisfy the needs of every application, but its introduction has served to expand the arsenal of the analytical chemist and open the door to an increasingly diverse array of scientific research.
REFERENCES 1 2 3 4 5 6 7 8 9 10 11
12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27
S. Greenfield, I.Ll. Jones and C.T. Berry, Analyst, 89 (1964) 713. R.H. Wendt and V.A. Fassel, Anal. Chem., 37 (1965) 920. A.L. Gray, Proc. Soc. Anal. Chem., 11 (1974) 182. R.S. Houk, V.A. Fassel, G.D. Flesch, H.J. Svec, A.L. Gray and C.E. Taylor, Anal. Chem., 52 (1980) 2283. A.R. Date and A.L. Gray, Analyst, 106 (1981) 1255. D.J. Douglas and J.B. French, Anal. Chem., 53 (1981) 37. A.R. Date and A.L. Gray, Spectrochim. Acta B, 38 (1983) 29. D.J. Douglas, E.S.K. Quan and R.G. Smith, Spectrochim. Acta B, 38 (1983) 39. R.S. Houk, Anal. Chem., 58 (1986) 97. K.E. Jarvis, A.L. Gray and R.S. Houk, Handbook of Inductively Coupled Plasma Mass Spectrometry. Blackie, Glasgow, 1992, 380 pp. H.P. Longerich and W. Diegor, Diegor, Introduction to mass spectrometry. In: P. Sylvester (Ed.), Laser Ablation ICPMS in the Earth Sciences, Short course series volume 29, Mineralogical Association of Canada, 2001, 243 pp. J. Hecht, Laser Pioneers. Academic Press, Boston, 1992, 298 pp. A.L. Gray, Analyst, 110 (1985) 551. P. Arrowsmith, Anal. Chem., 59 (1987) 1437. W.T. Perkins, R. Fuge and N.J.G. Pearce, J. Anal. At. Spectrom., 6 (1991) 445. W.T. Perkins, N.J.G. Pearce and T.E. Jeffries, Geochim. Cosmochim. Acta, 57 (1993) 475. N.J.G. Pearce, W.T. Perkins, I. Abell, G.A.T. Duller and R. Fuge, Anal. Chem., 7 (1992) 53. S.E. Jackson, H.P. Longerich, G.R. Dunning and B.J. Fryer, Can. Mineral., 30 (1992) 1049. T.E. Jeffries, W.T. Perkins and N.J.G. Pearce, Chem. Geol., 121 (1995) 131. T.E. Jeffries, W.T. Perkins and N.J.G. Pearce, Analyst, 120 (1995) 1365. S. Chenery, A. Hunt and M. Thompson, J. Anal. At. Spectrom., 7 (1992) 647. B.J. Fryer, S.E. Jackson and H.P. Longerich, Can. Mineral., 33 (1995) 303. T.E. Jeffries, N.J.G. Pearce, W.T. Perkins and A. Raith, Anal. Commun., 33 (1996) 35. H.P. Longerich, D. Gu¨nther and S.E. Jackson, Fresenius J. Anal. Chem., 355 (1996) 538. D.J. Figg, J.B. Cross and C. Brink, Appl. Surf. Sci., 129 (1998) 287. Y. Shibata, J. Yoshinaga and M. Morita, Anal. Sci., 9 (1993) 129. S. Chenery and J.M. Cook, J. Anal. At. Spectrom., 8 (1993) 299.
357
Teresa E. Jeffries 28
29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44
45
46 47 48 49
358
T.E. Jeffries, Investigations of Mineral Analysis by Laser Ablation Inductively Coupled Plasma Mass Spectrometry, PhD. Thesis, University of Wales, Unpublished, 1996, 246 pp. D. Gu¨nther, R. Frischknecht, C.A. Heinrich and H.J. Kahlert, J. Anal. At. Spectrom., 12 (1997) 939. D. Gu¨nther and C.A. Heinrich, J. Anal. At. Spectrom., 14 (1999) 1369. T.E. Jeffries, S.E. Jackson and H.P. Longerich, J. Anal. At. Spectrom., 13 (1998) 935. Y. Huang, Y. Shibata and M. Morita, Anal. Chem., 65 (1993) 2999. E.F. Cromwell and P. Arrowsmith, Anal. Chem., 67 (1995) 131. M.D. Norman, N.J. Pearson, A. Sharma and W.L. Griffin, Geostand. Newslett., 20 (1996) 247. S.E. Jackson, I. Horn, H.P. Longerich, G.R. Dunning and V.M. Goldschmidt, Conf. J. Conf. Abstracts, 1 (1996) 283. P.J. Sylvester and M. Ghaderi, Chem. Geol., 141 (1997) 49. P.M. Outridge, W. Doherty and D.C. Gregoire, Spectrochim. Acta B, 52 (1997) 2093. D. Figg and M.S. Kahr, Appl. Spectrosc., 51 (1997) 1185. S.M. Eggins, L.P.J. Kinsley and J.M.G. Shelley, Appl. Surf. Sci., 129 (1998) 278. G. Gutie´rrez Alonso, J. Ferna´ndez Sua´rez, G.A. Jenner and S.E. Jackson, Can. J. Earth Sci., 35 (1998) 1. Z. Chen, J. Anal. At. Spectrom., 14 (1999) 1823. A.J.G. Mank and P.R.D. Mason, J. Anal. At. Spectrom., 14 (1999) 1143. R.E. Russo, X.L. Mao, O.V. Borisov and H. Liu, J. Anal. At. Spectrom., 15 (2001) 1115. D. Gu¨nther and B. Hattendorf, Elemental fractionation in LA-ICP-MS. In: P. Sylvester (Ed.), Laser Ablation ICPMS in the Earth Sciences, Short course series volume 29, Mineralogical Association of Canada, 2001, 243 pp. S.E. Jackson, The application of Nd:YAG lasers in LA-ICP-MS. In: P. Sylvester (Ed.), Laser Ablation ICPMS in the Earth Sciences, Short course series volume 29, Mineralogical Association of Canada, 2001, 243 pp. L. Moenke-blankenburg, T. Schumann, D. Gu¨nther, H.M. Kuss and M. Paul, J. Anal. At. Spectrom, 7 (1992) 251. B.J. Masters and B.L. Sharp, Anal. Commun., 34 (1987) 237. D. Gu¨nther, H. Cousin, B. Magyar and I. Leopold, J. Anal. At. Spectrom., 12 (1997) 165. P.R.D. Mason, Expanding the capabilities of laser ablation ICP-MS with collision and reaction cells. In: P. Sylvester (Ed.), Laser Ablation ICPMS in the Earth Sciences, Short course series volume 29, Mineralogical Association of Canada, 2001, 243 pp.
Chapter 8
Infrared, Raman microscopy and fibre-optic Raman spectroscopy (FORS) Howell G.M. Edwards and Dalva L.A. de Faria
8.1
INTRODUCTION
The scientific study of works of art and the materials that were used in their creation has provided an impetus for the application and development of non-destructive, microsampling analytical techniques. However, the specimens themselves often provide some of the most challenging problems for analysis because of the complexity in composition of the source materials used in their construction. The problem is often exacerbated by the lack of precise knowledge about the ancient technologies used in the production of the artwork. Ancient recipes were sometimes mis-translated and generic names of pigments, e.g., minium, were changed over the centuries that the materials were in use. Specific details of materials which were added to pigments or pigment mixtures to aid their bonding to the substrate materials are often lacking and these may well have been experimental at the time, which creates problems for conservation scientists and art restorers. Also, most artwork is likely to have been repaired or restored at some time and it is often by no means obvious what restorative materials were used for this purpose. Later generations of conservators now have the problem of defining areas of restoration as well as dealing with new problems of interaction of non-compatible materials, which may have been unconsciously created in this process. Environmental damage is a key feature to be addressed for art works which are exposed to the elements; in particular, the aggressive colonization by microbes and lichens of exposed stone, rock and plaster surfaces found in wall-paintings and frescoes, as well as the obvious substratal damage caused by acid rain or hydrocarbons from vehicles and habitation combustion processes, is an urgent problem which requires some rapid solutions. In these cases, therefore, analytical techniques are required not only simply to Comprehensive Analytical Chemistry XLII Janssens and Van Grieken (Eds.) q 2004 Elsevier B.V. All rights reserved
359
Howell G.M. Edwards and Dalva L.A. de Faria
furnish information about the original construction of an art work, but also to identify the sources and extent of chemical or biological deterioration therein and, if possible, to provide some “early warning” signals to alert conservators to major potential problems. Hence, the arresting of the degradation of an artwork by external or internal agencies is now assuming importance before the restoration can be effected and in this area IR and Raman microscopy have been particularly relevant in their applications as well as providing information about pigment composition, resins used in varnishes and the original sourcing of materials.
8.2
COMPARISON OF THE POTENTIAL USE OF IR AND RAMAN SPECTROSCOPIES FOR THE NON-DESTRUCTIVE ANALYSIS OF ART WORKS
Both Fourier-Transform infrared (FT-IR) and Raman (visible and FT) spectroscopic techniques have been applied to the analysis of art work for some years, and it is perhaps timely to consider the advantages and disadvantages of each, which often is resolved by the type of material or artwork that is to be examined. IR spectroscopy, along with X-ray examination, was applied to art materials and art work over 50 years ago; in conjunction with the adoption of IR spectroscopy as a rapid analytical technique for industrial and forensic applications, there now exist several comprehensive databases for the identification of materials such as fabrics, plastics, fibres (natural and synthetic), minerals, paints and pigments. Raman spectroscopy did not make inroads into the art historical field until much later, because of the problems associated with fluorescence emission swamping the lower intensity Raman signals excited by the blue radiation, commonly used before the advent of laser sources with different wavelengths. Even simple mineral pigments, which might have been expected to realize good Raman data often caused problems because of the presence of highly fluorescent varnishes, degraded resins and binders used in their application. Substrates such as horn, bone and skin (parchment, vellum) also fluoresce strongly at short wavelengths. The advent of laser Raman microscopy, with a choice of excitation wavelengths, has provided an impetus to the applications of Raman spectroscopy to the analysis of ancient materials. The low scattering intensity of water and hydroxyl groups in the Raman effect contrasts with their very strong absorption in the IR region and this generally favours the use of Raman spectroscopy for the analysis of ancient
360
Infrared, Raman microscopy and fibre-optic Raman spectroscopy (FORS)
biomaterials, possibly hydrated, such as linen, wools, leather and skins. An example of this is shown in Figure 8.1, which shows a Raman spectrum (excited in the near IR at 1064 nm) and an IR absorption spectrum of the same mummy wrapping from a XIIth Dynasty ancient Egyptian tomb, dating from ca. 4000 years BP [1]. It is evident that the Raman spectrum provides a clearer analytical picture in the numbers and relative intensities of the vibrational data than does the IR spectrum, which is characterized by broad, featureless absorptions not lending themselves to detailed interpretation. Despite this, there is an extensive literature in the area of IR analyses of ancient cloths and fabrics, whereas Raman spectroscopy is only now being applied here. Also, in Fig. 8.1 are shown the Raman spectra of a modern damask linen and of an untreated cotton boll (Boone Hall Plantation, South Carolina, USA), which indicates that the 4000-year-old cellulose specimen has hardly suffered in spectral quality compared with the modern examples. An important difference between the data provided by Raman spectroscopy and by IR absorption relates to the wavenumber range of instrumental coverage. Much IR data for art work analysis spans the range 4000 – 600 cm21; whereas this is adequate for the characterization of organic components in a mixture, it is sadly deficient for many minerals. Hence, if the objective of the analysis is the assessment of the presence of a heavy metal oxide or sulphide pigment and its interaction with an organic substrate, such as vellum, it is quite likely that the mineral pigment will not have an IR absorption band above 400 cm21. We have noticed this in our analysis of ancient pigmented beeswax excavated archaeologically from a marine environment [2], in which the Raman data clearly show the presence of cinnabar, mercury(II) sulphide, but the IR spectrum does not (Fig. 8.2). However, the degraded beeswax bands are better defined in the 860 – 1600 cm21 region in the IR than they are in the Raman data. Another example of the power of long-wavelength excitation in Raman spectroscopy is provided by the comparison in Fig. 8.3 of the Raman spectra of two biodegradative lichens growing on a medieval fresco [3]; clearly, the Raman spectral differences (1064 nm excitation) enable us to differentiate between these lichen species and to confirm their presence on the artwork. However, in Fig. 8.4 the results are shown of a similar analysis involving 785 nm (top) and 633 nm (bottom) excitation, in which the fluorescent emission background is significant for the swamping of the Raman signals. The discrimination ability of Raman spectroscopy for the identification of ancient materials is provided in Fig. 8.5, which gives in the top specimen (a) the Raman spectrum of a mediaeval vellum manuscript—the protein bands at 1650, 1460 and 1240 cm21 are clearly present, along with a broad feature
361
Howell G.M. Edwards and Dalva L.A. de Faria
Fig. 8.1. (a) Raman spectra of ancient and modern cotton fibres; (b) IR spectrum of a cotton and linen mummy wrappings from the XIIth Dynasty tomb of Nekht-Ankh.
362
Infrared, Raman microscopy and fibre-optic Raman spectroscopy (FORS)
Fig. 8.2. (a) and (b) Infrared and Raman spectra of ancient pigmented beeswax excavated from a marine environment, showing the presence of cinnabar, mercury (II) sulphide.
cited at 790 cm21 characteristic of calcium hydroxide which was used as a preservative for parchment or vellum in its pre-treatment for use in scriptoria [4]. Spectrum (b), in Fig. 8.5, on the other hand was obtained from another region of the manuscript; this is clearly not vellum, and
363
Howell G.M. Edwards and Dalva L.A. de Faria
Fig. 8.3. Raman spectra of two biodegradative lichens growing on a medieval fresco: Xanthoria elegans and Acarospora gwynii obtained using a 1064 nm laser.
Fig. 8.4. Raman spectra of the same species as shown in Fig. 8.3, but using 785 nm (top) and 633 nm (bottom) excitation wavelengths.
364
Infrared, Raman microscopy and fibre-optic Raman spectroscopy (FORS)
Fig. 8.5. Raman spectra of (a) a mediaeval vellum manuscript, showing protein bands at 1650, 1460 and 1240 cm21 along with a broad feature at 790 cm21 characteristic of calcium hydroxide; (b) another region of the manuscript; (c) cellulosic paper.
matches exactly the spectrum c for cellulosic paper! Other details of the manuscript led conservators and art historians to conclude that the “repair” was carried out in medieval times, possibly to conserve a poor surface on the vellum at the time of its decoration. The spectroscopic analysis of art materials which have been discovered in archaeological excavations, rather than being located in art galleries or museums, poses a special problem for analytical interpretation since there is a strong likelihood that they have either absorbed fluorescent materials from the burial environment or otherwise been subjected to environmental degradation. An example of the latter is provided in Fig. 8.6 by three ivory artefacts [5]; the bottom spectrum is of modern elephant ivory—the strongest band occurs at 960 cm21 and is the n (PO) phosphate stretching mode of the calcium hydroxyapatite matrix. The bands near 400 and 600 cm21 are the d (PO) phosphate deformations and a complex feature near 1070 cm21 represents carbonate/phosphate stretching modes. The central spectrum is of an ivory comb from the Viking occupation at York (ca. 1000 years BP); the collagen proteinaceous modes at 1650, 1460 and 1240 cm21 have decreased significantly in intensity due to a leaching-out of the organic component in
365
Howell G.M. Edwards and Dalva L.A. de Faria
Fig. 8.6. Raman spectra of (top) a Roman ivory die (ca. 1800 years BP); (middle) an ivory comb from the Viking occupation at York (ca. 1000 years BP). The collagen proteinaceous modes at 1650, 1460 and 1240 cm21 have decreased significantly in intensity; degradation of the collagen is also indicated by an increase in the bandwidths of the organic bands and a loss of the phenylalanine residue, defined by the sharp peak at 1002 cm21 in the lower spectrum; (bottom) modern elephant ivory—the strongest band occurs at 960 cm21 and is the n (PO) phosphate stretching mode of the calcium hydroxyapatite matrix. The bands near 400 and 600 cm21 and a complex feature near 1070 cm21 represents carbonate/phosphate stretching modes.
the burial environment. Degradation of the collagen is also indicated by an increase in the bandwidths of the organic bands and a loss of the phenylalanine residue, defined by the sharp peak at 1002 cm21 in the lower spectrum. Finally, the upper spectrum shows the Raman data for a Roman ivory die (ca. 1800 years BP); the collagen component has disappeared completely and only the inorganic matrix remains. At this point it would be interesting to consider some basic aspects on the theory of Raman and IR spectroscopies.
366
Infrared, Raman microscopy and fibre-optic Raman spectroscopy (FORS)
8.3
SOME THEORETICAL ASPECTS OF IR AND RAMAN SPECTROSCOPIES
Spectroscopic techniques are based on the interaction of some sort of radiation (UV, visible, IR, microwave etc) with matter. The energy of such radiation is directly proportional to its frequency (E ¼ hn; where h is the Planck’s constant) and defines its effect on molecules. Thus, the more energetic UV and visible radiation promotes electronic transitions whereas the less energetic IR and microwave (MW) causes vibrational and rotational movements, respectively. Even at very low temperatures the atoms in the molecular structure are in an oscillatory movement with the frequency depending on the bond strength and on their masses. It was recognized since the very first systematic measurements with IR radiation carried out by Coblentz, that absorption only occurs for specific frequencies, characteristic of atomic groups, thus making vibrational spectroscopy a very powerful tool for compound identification and chemical structure investigations. The “ball and spring” model gives the simplest theoretical description for the atomic vibrational movement. In the so-called harmonic oscillation, when a mass is displaced from its equilibrium position by Dx (Fig. 8.7), a restoration force ðF ¼ 2kD xÞ appears, which is directly proportional to the spring constant (Hooke’s law). It is possible to combine this expression with a very basic movement equation ðF ¼ maÞ to obtain: 2kD x ¼ ma or 2 kDx ¼ mD x00
D x00 2 ðk=mÞD x ¼ 0
ð8:1Þ
where D x00 is the second derivative of D x with respect to time
Fig. 8.7. A restoration force (F ¼ 2kD x) appears when the ball is moved by D x.
367
Howell G.M. Edwards and Dalva L.A. de Faria
The solution for such an equation is D x ¼ D x0 cosð2pn0 tÞ with n0 ¼ (1/2p)(k/m)1/2. The same rationale applies to molecules and in this case the spring constant and mass is replaced by the bond force constant and the reduced mass ðm ¼ m1 m2 =m1 þ m2 Þ; respectively; the vibrational frequency is thus expressed as n ¼ ð1=2pÞðk=mÞ1=2 : A classical harmonic oscillator, however, cannot describe conveniently the atomic movement and quantum mechanics must then be used. In this case, the same expression for frequency is obtained but energy can now only assume discrete values, depending on the vibrational quantum number v, instead of varying freely as in the classic harmonic oscillator: E ¼ hnðv þ
1 2
Þ
v ¼ 0; 1; 2; 3; …
ð8:2Þ
According to the quantum –mechanical description, the vibrational levels of the harmonic oscillator are equally spaced and only transitions to adjacent levels are allowed ðDv ¼ ^1Þ; originating absorption (þ1) or emission (2 1) bands at defined positions; these are the fundamentals. Unfortunately, molecules do not behave as harmonic oscillators and deviations from ideality are observed; this includes the observation of harmonics ðDv ¼ ^2; ^3 etc.) and combination bands (sum or subtraction of two fundamentals of proper symmetry). Furthermore, in the anharmonic oscillator the energy levels are not equally spaced and thus harmonic vibrations do not appear at exact multiples of a fundamental frequency but rather at a slightly lower frequency. Matching a vibrational frequency with the energy of the IR radiation is not, however, the only factor to determine whether an absorption band will be observed or not in the IR spectrum. Energy transfer from the IR radiation to the molecule proceeds through the coupling of the oscillating electric field of the radiation with the electric field generated when the molecular dipole moment changes with the vibrational movement. Hence, IR absorption will take place whenever the vibration leads to a change in the dipole moment (m) of the molecule. Whereas IR spectroscopy is based on the absorption of IR radiation by a vibrating molecule, similar information (Fig. 8.1) can be obtained using a visible and monochromatic light source (NIR and UV radiations can also be used). In this case it is quite obvious that a resonance mechanism cannot be operative because of the energy difference (vibrations occur in the IR region and the laser radiation is in the visible), instead, it was shown nearly 80 years ago that light could be inelastically scattered by vibrating molecules with the energy difference corresponding to a vibrational transition in the molecule [6].
368
Infrared, Raman microscopy and fibre-optic Raman spectroscopy (FORS)
The simplest description for light scattering is the classical picture by which scattering is described as a secondary emission caused by an induced dipole created in the molecule by light. The oscillating electrical field of the radiation (E) creates an induced dipole moment (P) on the molecule, since electron clouds are polarizable to a variable degree, depending on the nature of the atoms and on the chemical bond order (single, double etc). Thus, the magnitude of P (and the intensity of the scattered light) depends on a molecular property called polarizability (a) and on the incident electrical field (E): P ¼ aE
ð8:3Þ
Considering that E ¼ E0 cosð2pntÞ; P can be expressed as: P ¼ aE0 cosð2pntÞ
ð8:4Þ
where n is the frequency of the incident radiation According to the laws of classic electromagnetism, an oscillating dipole irradiates at the same oscillatory frequency and at all directions, except along the dipole axis; the total energy irradiated by such a dipole is: I ¼ ð16=3Þðp4 n4 =c3 Þa2 E20
ð8:5Þ
The dependence of the Raman intensity on the fourth-power of n makes the excitation in the NIR much less favourable than in the visible as will be discussed opportunely. It also emerges from Eq. (8.5) that the more intense is the incident radiation and the more polarizable the chemical species, the more intense the Raman band is, due to the dependence of I on E20 and a2, respectively. On the other hand, the polarizability in a vibrating molecule is likely to change as a function of the vibrational frequency and this modulation can be expressed as a ¼ a0 þ Da cosð2pnv tÞ
ð8:6Þ
where a0 is the polarizability at the equilibrium position, Da is the maximum variation in polarizability and nv is the vibrational frequency. Combining these equations, it is possible to express P as a function of the vibrational frequency: P ¼ aE0 cosð2pntÞ ! P ¼ ða0 þ Da cosð2pnv tÞÞE0 cosð2pntÞ P ¼ a0 E0 cosð2pntÞ þ note that cosu cosf ¼
1 2
or ð8:7Þ
DaE0 ½cos2pðn þ nv Þt þ cos2pðn 2 nv Þt
1 2
½cosðu þ fÞ þ cosðu 2 fÞ
369
Howell G.M. Edwards and Dalva L.A. de Faria
The first term on the right hand side of this equation, ða0 E0 cosð2pntÞÞ; corresponds to the elastic (Rayleigh) light scattering since the scattered frequency is the same as the incident radiation whereas the second term 1 shows that n changes by ^ nv. 2 DaE0 ½cos2pðn þ nv Þt þ cos2pðn 2 nv Þt Inelastic light scattering is known as the Raman effect (named after C.V. Raman) and the frequency of the scattered radiation can be lower (n 2 nv) or higher (n þ nv) than the frequency of the excitation energy: these are the Stokes and anti-Stokes regions of a Raman spectrum (Fig. 8.8). It must be noted, however, that the observation of a Raman band for a particular vibration depends on the change in the molecular polarizability with such vibration, in the same way as the observation of an IR absorption band depends on the change of the dipole moment with the vibration.
Fig. 8.8. (a) Scheme for IR and Raman spectroscopies; (b) Raman scattering can produce photons with either higher or lower energies than the incident energy.
370
Infrared, Raman microscopy and fibre-optic Raman spectroscopy (FORS)
Of course, the classical model cannot be used when a quantitative description is necessary. In this case a quantum-mechanical model must be considered and molecular polarizability is expressed as described by Kramers and Heisenberg, taking into account all excited electronic states of the molecule. A more detailed description of such model is beyond the scope of this chapter and can be found in the literature [7]. Due to the very different nature of Raman and IR spectroscopies (scattering and absorption, respectively) there are some specific points that must be considered. First of all, because the selection rules are different, IR and Raman spectroscopies are complementary techniques, helping in the assignment and interpretation of a vibrational spectrum, mainly considering the absence of harmonics and combination bands in the Raman spectrum (except under resonance conditions, as will be discussed later) whereas they appear with appreciable intensity in the IR. A major advantage of Raman spectroscopy is the weak Raman scattering from water molecules which are very poor light scatterers; hence, for reactions in aqueous solution, biomolecules and biopolymeric studies, little sample preparation in the form of desiccation or freeze drying is necessary. This is particularly crucial for the monitoring of living biological systems by vibrational spectroscopy, where IR spectroscopic absorption due to water is strong, swamping the important vibrational features and interfering with the spectrum interpretation. However, as shown in Fig. 8.9, the moving of the excitation wavelength further into the IR severely restricts the Raman data that can be acquired in the presence of water, because of increased absorption. This will present a severe limitation to the longer-wavelength capability for FT-Raman (FTR) excitation. Even at 1064 nm, the effect of higher-overtone water based absorption on the higher wavenumber shifts will be manifest. For this reason, the substitution of heavy water (D2O) has been noted to alleviate partially the absorption problem with 1064 nm excitation in the near IR. Concerning sampling (which will be discussed below); IR samples are usually dispersed in a support medium and commonly prepared as halide salt pellets (KBr, NaCl, CsI, etc.) or oil dispersions (Nujolw, Fluorolubew and Kel-Fw are generally employed); in the case of Raman spectroscopy, the samples can be investigated as received and eventually minimal manipulation is necessary. From the point of view of instrumentation, IR equipment is generally much cheaper than Raman systems (despite a significant drop in prices for portable Raman equipment observed in recent years) and the maintenance is
371
Howell G.M. Edwards and Dalva L.A. de Faria
Fig. 8.9. Absorption spectrum of water in the NIR region, showing the wave-number shifts (0 –3500 cm21) from Stokes Raman spectra excited with 800-, 1064- and 1339-nm laser wavelengths.
proportionally cheaper too. Low cost IR instruments usually have KBr optics, which limit the lower wavenumber range to ca. 400 cm21, whereas a Raman spectrum can easily start at 100 cm21, covering an important spectral region for inorganic compounds, as mentioned before. Finally, fluorescent samples are not a problem for IR spectroscopy because emission is occurring in another spectral region. This is not the case with Raman spectroscopy; inelastic light scattering is a very weak effect and the Raman spectrum is generally excited in the visible, where fluorescence usually appears. It is a major problem for Raman spectroscopy even if coming from low concentration impurities; as typically 1 out of 108 photons are inelastically scattered, an impurity at 1 ppm (1 part in 106) level with quantum efficiency of 0.1 will produce 10 times more photons than Raman scattering does. To cope with fluorescence some special procedures are frequently used and the most frequent option is to change the laser line wavelength
372
Infrared, Raman microscopy and fibre-optic Raman spectroscopy (FORS)
(and reduce the excitation energy) to the red or NIR (633, 780 and 830 nm are the commonest frequencies) but UV lines at higher energy (244 and 325 nm, e.g.) are alternative options [8]. Generally, laser power must be carefully controlled to avoid sample degradation or black body (thermal) emission (in the NIR) due to sample heating, but sometimes it is possible to bleach fluorescence to acceptable levels by over-exposing the sample to the laser light, provided the emission comes from impurities and the sample is not affected by the procedure. In confocal Raman microscopes the high axial resolution (ca. 3 mm) can also be used to minimize the fluorescence contribution to the spectrum [9]. Baseline correction [10] or subtraction using shifted spectrum [11] may lead to some improvement in the data quality in favourable cases. Surface enhanced Raman scattering (also denoted as the SERS effect) is becoming more popular. Another point worth mentioning is that in the case of badly degraded materials, such as the samples collected in excavations in archaeological sites, e.g., fluorescence may be a problem even with excitation at 1064 nm.
8.4
INSTRUMENTATION
Raman and IR instrumentation can be either dispersive or interferometric depending on the method used to separate each spectral component. In dispersive equipment, radiation light coming from the sample is dispersed by a grating or prism whereas an interferometer is used with the same purpose in interferometric systems. Basically, IR instrumentation consists of a source of continuum IR radiation, a discrimination component and a detector, as shown in Fig. 8.10. Commonly, the light source is a ceramic rod (cerium and zirconium oxides) heated at 1500 – 2000 K (Nernst glower). KBr or CsI lenses focus the radiation on the sample and the transmitted light is directed to a single or double beam interferometer. Interferometers are optical systems in which interference patterns are created as the optical path changes with the movement of a mirror (Fig. 8.11). If there is no difference path between the two interfering beams (A and B), a constructive pattern (IA þ IB) is observed. If, however, the movable mirror is displaced by d a difference path of 2d appears, eventually leading to a destructive interference, as schematically represented in Fig. 8.12.
373
Howell G.M. Edwards and Dalva L.A. de Faria
Fig. 8.10. Scheme for FT-IR spectroscopy.
Fig. 8.11. Scheme for a Michelson interferometer. LS, light source; S, slit; D, detector; FM, fixed mirror; MM, movable mirror.
374
Infrared, Raman microscopy and fibre-optic Raman spectroscopy (FORS)
Fig. 8.12. Interferogram patterns for monochromatic radiation: in-phase (a) and outof-phase (b); polychromatic radiation (different frequencies) originate a more complex interferogram (c).
Thus, the intensity recorded at the interferometer exit represents the interference of all the spectral components and can be written as: IðdÞ ¼ I0 cos2 ðf=2Þ with f ¼ ðd=lÞ2p or ðd=cÞ2pn IðdÞ ¼ IðdÞ ¼
ð
1 2
BðnÞ cos2 ½ðd=cÞ2pndn ¼ ð
BðnÞdn þ
1 2
ð
1 2
ð
BðnÞ{1 þ cos½ðd=cÞ2pn}dn
ð8:8Þ
BðnÞ cos½ðd=cÞ2pndn
The resolution of such an interferogram is 1=x; where x is the path length of the recorded interferogram; hence, a 0.5 cm path difference gives an effective spectral resolution of 2 cm21. The mathematical procedure used to express the information contained in an interferogram in terms of frequency instead of distance is performed by the Fast Fourier Transform algorithm developed ca. 50 years ago [12]. In the integrals above (Eq. (8.8)) the integration limit is from zero to infinite, but of course the mirror can only move within a limited distance, thus integration must be performed over a finite interval and truncation originates side bands. To avoid this artefact, the integrand on Eq. (8.8) is multiplied by a convenient function named apodization function. The advantages of FT-IR equipment over the dispersive analogues are: (i) all spectral bands are recorded at the same time since every point in the interferogram contains information from the whole spectrum (multiplex or Felgett advantage); (ii) there is no need of slits since resolution is given by the path distance in the movable mirror arm (Jacquinot or throughput advantage), and (iii) the movement of the movable mirror is controlled by the fringes of a He – Ne laser thus band positions are known with a much greater precision than in dispersive systems (Connes advantage). Concerning detectors, for the mid-IR region MCT (mercury and cadmium telluride) or DTGS (deuterium triglycine sulphate) are more common
375
Howell G.M. Edwards and Dalva L.A. de Faria
whereas in the NIR photoconducting detectors, such as PbS, InAs, InGaAs or InSb, can be employed. FT-IR microscopes generally use MCT detectors and the microscope operates with mirror lenses and condensers, which are more efficient in the IR than other materials. Raman spectroscopic equipment can be interferometric or dispersive and dispersive systems can be either single (scanning) or multichannel. As mentioned before, Raman spectroscopy is particularly sensitive to fluorescence, a very common problem in ancient samples where degradation usually leads to an increase in the emission background under laser illumination. One approach to avoid fluorescence is using an exciting line with less energy than that necessary to populate the excited states from which emission occurs (Fig. 8.13). Excitation at 1064 nm (NIR) has proved to be very convenient to this purpose, but here the problem is the n 4 dependence of the Raman intensities (Eq. (8.5)), which makes Raman effect particularly weak in this region. To circumvent this problem the use of FTR equipment is mandatory: since radiation from all spectral elements over the desired wavelength range is measured in one interferogram scan, the signal-to-noise ratio (S/N) can be increased interferometrically (Fellgett advantage) by the rapid coaddition of scans. For example, modern interferometers can achieve 60 interferogram scans per minute and calculation reveals that a 30 min spectral accumulation for a 2000 cm21 range therefore gives a S=N improvement of about 50 £ over the single-scan dispersive system at 1 cm s21 [13]. Additionally, because of the longer wavelengths in the IR region the precision required for the mirror movements is readily achievable. However, because of the sensitivity of the photomultiplier and CCD detectors in the
Fig. 8.13. NIR radiation (1064 nm) generally can not populate the excited states responsible for fluorescence (NRD, non-radiactive decay and RD, radiactive decay).
376
Infrared, Raman microscopy and fibre-optic Raman spectroscopy (FORS)
visible region, FT methods are not so advantageous and UV/visible spectrometers still tend to be dispersive-grating systems. The general concepts used in FTR instrumentation (Fig. 8.14) are obviously the same as those described for FT-IR but some differences are worth mentioning. The source is now a monochromatic Nd3þ/ YAG laser that emits at 1064 nm. This line is filtered, focused on the sample and the radiation inelastically scattered from the sample is filtered to remove the Rayleigh line and directed to an interferometer. Some FT-IR spectrometers are designed to accommodate a Raman accessory and in this case a wide range beam splitter must be used to cover both the IR and NIR regions. FTR detectors are generally liquid nitrogen cooled Ge or indium doped with gallium arsenide (InGaAs); the later can be operated at room temperature but there is a compromise of the signal-to-noise ratio. For Raman spectroscopy, interferometers have several advantages over monochromators of similar aperture where radiation must pass through narrow slits (typically 50 –100 mm cm21 spectral bandpass) compared with
Fig. 8.14. Scheme for an FT-Raman spectrometer (reproduced with permission of Bruker).
377
Howell G.M. Edwards and Dalva L.A. de Faria
the larger optical conductance of an interferometer (the Jacquinot advantage). For a spectral resolution of 4 cm21, for example, the e´tendue (radiation throughput) of an interferometer may be 100 £ better than that of the monochromator [14]. Other interferometric properties such as the Connes advantage, which arises from the coaddition reproducibility and superimposability of data accumulation, means that in the IR region the interferometer is supreme for Raman-scattering experiments. The feasibility of using interferometry and near-IR laser-excitation sources for Raman spectroscopy was suggested by Chantry et al. as long ago as 1964 [15] but interest was renewed through the work of Chase and Hirschfeld (1986) [16], which was quickly followed with the first commercial spectrometer in 1987/88. For Raman-spectroscopic studies the operation of a system with a near-IR excitation wavelength (1064 nm) obviates the problems associated with fluorescence excitation experienced in the blue and green regions of the visible [17]. This is an inestimable advantage for the examination of the Raman spectra of fluorescent species such as those commonly found in archaeological excavations [18]. On the negative side, however, the n 4 intensity law for Raman scattering means that the intensity of Raman scattering decreases with increase in wavelength ðln ¼ cÞ from the visible to the IR; hence, the Raman scattering from 457.9 nm Arþ laser excitation is about 40 £ as intense as that from a Nd3þ/ YAG laser operating at 1064 nm, power for power. With generally lower powers for IR lasers compared with the argon-ion system, this difference is compounded even further. Despite this intrinsic disadvantage, however, the speed and ease of sampling occasioned by the FT spectrometers now make these attractive instruments for both Raman and IR-spectroscopic studies of molecular systems. The advent of FTR spectroscopy and CCD-Raman spectroscopy has truly provided a renaissance for the Raman technique through the wider range of applications now available. The detailed comparison of conventional disperse-Raman with the FTR spectroscopic technique is difficult because of the correlation of the many instrumental factors involved in each technique, however, a summary of the major differences is as follows: 1. Moving to near-IR excitation at 1.06 mm from the visible reduces the Raman scattering intensity (dependent on the fourth power of the excitation wavelength) by up to 40 £ on 457.9 nm excitation. 2. An advantage of near-IR excitation is the reduction in energy of the lasing transition from about 25000 – 10000 cm21, which therefore inhibits the onset of fluorescence; this is of supreme importance in the application
378
Infrared, Raman microscopy and fibre-optic Raman spectroscopy (FORS)
of FTR techniques to biopolymers (such as tissues and bones), varnishes and resins. 3. In a scanning Raman monochromator system, and to some extent also in spectrographs, only a small portion of the dispersed spectrum is analysed at any one time. With the interferometer, however, the Fellgett advantage applies; this is based on the acquisition of all spectral resolution elements simultaneously. This greater “speed” of acquisition by an interferometer over a dispersing spectrometer is pffiffi quantified as, r where r represents the number of resolution elements in the spectrum. 4. The Jacquinot throughput is much greater for the interferometer than for scanning monochromators in any spectral range since the limitation of the spectral slit is removed. The J stop in a typical interferometer is about 8 mm in diameter compared with a 0.1 mm slit width in a spectrometer system. 5. Internal calibration of the interferometer against a helium-neon laser (Connes advantage) provides exceptional wave-number reproducibility, which facilitates the superposition of spectral data (accumulation) and data subtraction (background, solvent etc.). Hence, although all the spectral data in an FTR experiment may be provided in one scan of about 2 s, multiple spectral-data accumulation to improve the signal-tonoise ratio is desirable. In cases of very weak Raman spectra, especially from biological materials for which the incident laser power has been reduced to less than 20 mW from 1 W to minimize sample degradation, accumulations of many hours duration are acceptably realistic; e.g., the first FTR spectroscopic analysis of archaeological human tissue of the “Ice-man,” a 5200-year-old corpse, has been provided with up to 12000 accumulated spectral scans at a resolution of 4 cm21 from a 1 mg sample without deterioration of spectral resolution or quality or evidence of sample degradation [19]. The increased speed of data acquisition and ease of sampling have made the near-IR excitation of Raman spectra also attractive for industrial remotesensing applications. Extension of FTR spectroscopy into the realm of fast, real-time data acquisition in nanoseconds is limited by the relatively slow frequency response of the detectors, but time resolutions of 50 ms through step scanning have been shown to be possible. A critical parameter here, too, is the rather poor signal-to-noise ratio generally experienced from single-scan or singleshot Raman spectra [14].
379
Howell G.M. Edwards and Dalva L.A. de Faria
The advent of near IR excitation has provided a new dimension for the application of Raman spectroscopy to art objects, despite the disadvantage of the inherent weakness in the scattering in the NIR. The lower energy of the excitation, equivalent to 9600 cm21 is useful to minimize the onset of fluorescence, especially for biomaterials such as tortoiseshell [20], horn [21], ivory [22], skin [23] and fabrics [24]. In addition to FTR spectroscopy using Nd3þ/ YAG laser excitation at 1064 nm, there are now several commercially available options in the near-IR using CCD detectors: laser lines at 785 and 830 nm are now offered as alternatives to the standard 1064 nm excitation. In 2001, a report of a miniature Raman spectrometer, weighing only 1 kg, using a 1064 nm diode laser and CCD-detection was announced [25]. Although this system is being developed for a NASA flight mission for the remote Raman spectroscopic analysis of planetary surfaces, the compact size and research grade quality spectra provided so far immediately lend themselves to adaptation for in situ art studies. In this way, one of the major criticism of FTR spectroscopy, that in some cases it is time-consuming and requiring often several hours of accumulation of spectral data for adequate spectral signal-to-noise ratios, is largely overcome by the rapid response of the CCD detector instrumentation. The dispersive systems employed in the UV and visible regions use gratings (or less commonly prisms) to resolve the scattered light in its spectral components. Several optical arrangements are possible, but are generally based in Ebert or Czerny-Turner arrangement (Fig. 8.15). A grating is a surface ruled with adjacent lines which currently correspond to interference fringes from two crossed laser beams (holographic gratings); the number of lines per mm and the illuminated area define the resolution that can be achieved by a particular grating. If a single channel detector is coupled to the monochromator it is necessary to scan the spectrum over the wavenumber region of interest. Such scanning equipment usually has a photomultiplier tube (PMT) as detector. In a PMT, light strikes the photocathode surface removing electrons from the metal surface; such electrons are accelerated towards secondary electrodes (photodynodes) and more electrons are ejected in a multiplying cascade effect. Multichannel equipment registers a whole spectral region (window) in one shot, with the range depending on the grating used and on the wavelength of the radiation. The detector in this case is usually a charge coupled device (CCD) which consists essentially in a two-dimensional array of photosensitive elements on a silicon chip [26].
380
Infrared, Raman microscopy and fibre-optic Raman spectroscopy (FORS)
Fig. 8.15. Diagram for Czerny-Turner (upper) and Ebert (lower) monochromators. S1 and S2, slits; G, grating; M, M1 and M2, mirrors.
8.5
SAMPLING
As mentioned before, samples for IR spectroscopy are usually prepared as a dispersion in a medium transparent in the IR region. Pellets in KBr or CsI and dispersions in oil are the usual procedures. In the first case, a small amount of sample (few milligrams) is finely powdered with dried KBr (or CsI) and pressed at ca. 10 tons for ca. 5 min, after which a transparent disk is obtained (depending on the sample and on its water content). The dispersion in oil is prepared in a similar way: small amounts of the oil are added to the finely powdered sample until a butter-like consistency is obtained. This paste is then placed between IR transparent windows (KBr, CsI, KRS-5 etc.). Valuable or rare specimens cannot be studied through destructive procedures and other (more expensive) options must be considered. Using microsampling accessories or a microscope, the amount of sample required decreases to the microgram level and can be considered virtually nondestructive. Despite this, some manipulation (often compression) is still necessary.
381
Howell G.M. Edwards and Dalva L.A. de Faria
In the first case, the more popular or appealing IR microspectroscopy techniques involve attenuated total reflectance (ATR), which is in general much cheaper than an IR microscope. ATR accessories for minute samples require a pressure anvil cell (diamond cell) in order to promote a good contact with the diamond window and the optical focusing is generally made by ZnSe or KRS-5 prisms (Fig. 8.16); currently single and multi-reflection configurations are commercially available and samples as small as 20 mm can be studied [27]. In an FT-IR microscope, IR and white light paths match what allows the IR spectrum to be obtained from specific spots previously chosen during the white light inspection. Cassegrain (mirror) lenses are used and an aperture defines which area in the field of view will be studied. The same type of sample preparation used in ATR microspectrometry is required for IR microscopy, thus better quality transmission spectra can only be obtained from thin samples which demand a pressure cell. The samples can also be studied without manipulation by reflectance microscopy, using ATR or grazing angle objectives [28].
Fig. 8.16. (a) Single and multi-reflection ATR accessories; (b) Cassegrain lens.
382
Infrared, Raman microscopy and fibre-optic Raman spectroscopy (FORS)
The size, shape and surface properties of specimens are also very important for valuable artefacts or art works which cannot be subjected to chemical or mechanical pre-treatment—the taking of small samples by drilling, scraping or excision is often prohibited. This has led to the development of novel pieces of equipment to facilitate the recording of IR or Raman spectra from large objects such as easel paintings, mediaeval vellum manuscripts, large sculptures and miniature paintings set under glass or rock crystal covers. Curved surfaces are notoriously poor for IR examination and specimens are usually better analysed after crushing between the poles of a diamond anvil cell; this may be “de rigeur” for the forensic examination of fibres, but is not favoured for art objects generally. The use of remote fibre-optic sensing probes is now advocated for several applications in art object analysis; however, the interpretation of the spectral data obtained from such devices is critically dependent on the accurate spectral subtraction of the probe (usually silica) background, particularly for the important mineral “fingerprinting” region below 1000 cm21, such as that required for artists’ pigments. Likewise, the construction of miniaturized spectrometers with a remote probe or microscope attachment for the recording of IR or Raman spectra of artefacts in situ in art gallery or museum environments has been accomplished in answer to an observation of the need to remove fragile and valuable objects from secure locations. Clearly, there will be opportunities here for the examination in situ of historic paintings in churches and prehistoric rock art in otherwise inaccessible locations. In the case of Raman spectroscopy generally there is no need of sample preparation and in Raman microscopy the specimen can be investigated directly on the microscope stage or using on a glass slide. In Raman microscopes (Fig. 8.17) the spectrum is obtained directly from the sample in the back-scattering arrangement: the same objective used to focus the laser beam on the sample also collects the scattered light. The specimen does not need to be pressed, as thickness and opacity are not relevant for light scattering. The objectives allow the laser radiation to be focused on a 1 – 3 mm spot typically in the visible region (in the IR the spot size is 10 mm and over), what is particularly useful for the identification of specific components in heterogeneous mixtures (Fig. 8.18). This is the case, for example, in investigations of egyptian pigments [29] and pigments in drawings [30] and manuscripts [31]. Metallurgical microscopes are commonly used since the laser lines are in the visible and glass optics can be used without any restriction. UV-Raman microscopes are available [32] and in this case special lenses
383
Howell G.M. Edwards and Dalva L.A. de Faria
Fig. 8.17. Scheme for a Raman microscope (reproduced with permission of Renishaw plc.).
are necessary because glass is not transparent in the UV; the detector is a CCD coated with a fluorescent compound to permit the detection of ultraviolet radiation (an uncoated CCD is responsive only in the 400 – 1100 nm range). Another interesting feature to be explored in Raman microscopes is the confocality, which means that the larger contribution to the recorded spectrum comes from the focused spot (Fig. 8.19). As mentioned before, a confocal set up is an additional tool to overcome the problems caused by fluorescence or degradation as can be seen in Fig. 8.20: photobleached skin obtained using a too large laser exposure caused sample degradation and the Raman spectrum obtained under confocal conditions from the layer immediately below the damaged surface (a) is compared with a spectrum from the burned area (b). Concerning FTR microscopy, despite the advantage of long-wavelength excitation for the recording of Raman spectra of potentially fluorescent materials, the disadvantage still maintains in the relatively large spatial volumes involved compared with confocal instruments using visible excitation. Hence, the lower limit of FTR microscopy using a 100 £ objective lens is realistically a cylinder of about 8 – 10 mm diameter and, depending on the refractive index of the material, up to several hundred mm in depth. This should be contrasted with confocal CCD instruments where 1 – 2 mm diameter “footprints” and claimed depths of about 1 – 2 mm are routinely advanced. This focal dimension is essential when layers topped with transparent coating are being examined, e.g., pigments coated varnish or resin. The ability to depth profile artwork and probe pigments, which are
384
Infrared, Raman microscopy and fibre-optic Raman spectroscopy (FORS)
Fig. 8.18. Raman spectrum from different components can be obtained from heterogeneous samples. In the figure, the red pigment is hematite and the white one is calcite (sub gypsum).
situated beneath clear layers of varnish or glass etc. is a useful advantage of confocal microscopic techniques. Application of the FT technique has found success in Raman microscopy in the near IR region of the electromagnetic spectrum especially for biological materials. The limit of sample size for analysis with this technique is similar to that of IR microscopy and is essentially that of the diffraction limit of the IR radiation used. Serious problems can arise when the sample size or aperture approaches the wavelength of this radiation—resulting in a loss of focal clarity. For the near IR region using 1064 nm excitation, the limit of resolution of an FTR microscopic arrangement is , 8 mm.
385
Howell G.M. Edwards and Dalva L.A. de Faria
Fig. 8.19. Diagram for a confocal arrangement.
Fig. 8.20. Raman spectra of skin after photobleaching: excessive power caused sample degradation and the contribution from the non-degraded spot (a) can be separated from the degraded one (b) by a confocal arrangement.
386
Infrared, Raman microscopy and fibre-optic Raman spectroscopy (FORS)
NIR radiation is imaged onto the sample using objectives ranging from about 5 £ to 100 £ ; the larger the objective magnification, the smaller the size of the spot being illuminated, i.e. the “footprint” from which the spectral information is being obtained. Using a beam splitter, the scattered radiation in the IR region is analysed in the spectrometer and comparison made with the visual image as viewed directly through the eyepiece or as recorded from a screen using a video photograph. Special objectives are required for high focussing, long-distance working; the latter are necessary for extensively uneven surfaces or for probing cracks or hollows in carved objects. Similar to IR microscopy, white light and the laser beam paths are aligned to permit the Raman spectrum to be obtained from specific areas on the sample surface, previously inspected using white light. 8.6
RESONANCE RAMAN
Excitation in the NIR is mandatory when fluorescence swamps the Raman bands and its advantages and disadvantages have already been considered. Visible excitation has, however, an aspect that has not yet been discussed: the possibility of using the Resonance Raman (RR) effect, which is particularly useful in the case of pigments or dyes. The introduction of lasers as excitation sources permitted the observation of non-linear effects (e.g. hyper-Raman and CARS) and a detailed study of the RR effect. The latter corresponds to an enhancement of 4 –6 orders of magnitude in the Raman cross section for some vibrational modes, when the laser line used to excite the spectrum is within or close to an absorption band (Fig. 8.21). The enhanced bands are those associated with the chromophoric group and from the analytical point of view, it is a very important tool in the detection of coloured compounds in complex mixtures, or when the pigment (or dye) is the minor component in a mixture or even when a background swamps non-enhanced bands. The theory for Raman scattering uses the transition moments involving vibronic levels and the excitation frequency to define the molecular polarizability components, ðaij Þxy ; it must be remembered that the intensity of a Raman band is proportional to a2ij : ðaij Þmn ¼ ð1=hÞSe {A þ B} A ¼ Mme Men =ðnem 2 n0 þ iGe Þ B ¼ Mme Men =ðnen þ n0 þ iGe Þ
ð8:9Þ
387
Howell G.M. Edwards and Dalva L.A. de Faria
Fig. 8.21. Resonance Raman effect: vibrational bands are enhanced when the Raman spectrum is excited within an allowed electronic transition band; on the absorption spectrum are superimposed Arþ ion laser line wavelengths.
where Mme and Men are the moments for the m ! e and e ! n transitions, respectively (m and nare vibronic levels of the electronic ground state and e is a vibronic level of an electronic excited state). As the excitation frequency (n0) approaches the frequency of an electronic transition, the denominator of the first term (A) goes to zero (in fact, it tends to iGe, which is a dumping factor and depends on the excited state lifetime) and aij is largely enhanced. RR effects are only observed for highly allowed transitions and the enhancement of totally symmetric modes are the most common behaviour (for more details on the RR please refer to Ref. [33]). In conventional Raman spectroscopy, combination bands and overtones are not present, but this is not the case under resonance conditions. A good example is the Raman spectrum of lazurite (a blue pigment, with absorption maxima at 650 nm) excited at 632.8 nm, where at least four harmonics are easily observed (Fig. 8.22). The inherent advantage of RR spectroscopy (RRS) is an enhanced spectrum strength and simplification of the spectra; this has found particular application recently in biological studies of site-specific drugDNA interactions [34] and in the monitoring of the early stages of polymer degradation, which produces sites of Raman active unsaturation.
388
Infrared, Raman microscopy and fibre-optic Raman spectroscopy (FORS)
Fig. 8.22. Resonance Raman spectrum of lazurite excited at 632.8 nm, showing harmonics of the 547 cm21 band.
A critical limitation of RRS is the presence of an electronic absorption in the visible region that is accessible to probing by appropriate tunable dye lasers; of course, visible electronic absorption bands also confer colour on chemical materials and many systems worthy of study are colourless. To address the last point, the development of UV RR spectroscopy has been undertaken, since most molecules, particularly those of biological significance, have absorption bands in the 200 –400 nm region of the electromagnetic spectrum. In addition, it is realized that for laser excitation at wavelengths less than about 250 nm fluorescence processes are reduced because the excited singlet states relax through fast radiationless transitions before onset of slower fluorescence occurs [35]. Hence, UV RRS is seen to have great potential as a sensitive technique for the probing of the molecular structures of a range of interesting systems; the major drawback, however, is the lack of appropriate lasing wavelengths, particularly in the 200 – 300 nm region. Most UV RRS studies reported in the literature have adopted a frequency-quadrupled Nd/YAG-pulsed laser, operating at 266 nm but, because of small pulse rates and high peak pulse powers, there are associated problems related to sample exposure for unacceptably long times, which can result in photodegradation and chemical deterioration. Argon ion laser lines can also be frequency-doubled to provide UV radiation; a particularly good example is the use of 244 nm (frequency-doubled 488 nm) in UV Raman microscopes (see Ref. [8]).
389
Howell G.M. Edwards and Dalva L.A. de Faria
8.7
SERS
Another useful technique to improve the signal-to-noise ratio in a Raman spectrum that presents the additional advantage of quenching fluorescence is the SERS. Discovered in the mid-1970s [36], the SERS effect corresponds to a significant increase in the intensity of some bands in the Raman spectrum of a molecule which is on a metal surface or close to it [37]. Generally, rough copper, silver or gold surfaces are expected to be SERS active surfaces and not all molecules will experience the enhancement; it seems that the higher the polarizability the larger the enhancement, which can easily reach six orders or magnitude or more if combined with a RR effect. In this case, the enhancement factors combine to provide an overall 1010 2 1012 enhancement. Adequate roughness can be achieved in colloids [38], electrodes [39] and vacuum evaporated metal films [40] but other options are available such as chemically deposited metal films [41] and acid etched copper foil surfaces [42]. A very simple description for the surface enhanced effect, uses the fact that the Raman intensity depends on the square of the induced dipole moment ðPÞ; which depends on the molecular polarizability (a) and on the radiation electric field ðEÞ: Thus, any factor increasing E or a will increase P (or the band intensity). A theoretical approach which considers the amplification of the local electric field is usually described as an “electromagnetic” model [43], whereas the “chemical” model considers the molecular polarizability and charge transfer processes involving the metal surface and the adsorbate [44]. The chemical contribution only occurs for molecules directly adsorbed on the metal surface and for this reason it is generally classified as a short range contribution whereas the electromagnetic contribution decreases smoothly with the surface-molecule distance and does not depend on a specific interaction with the metal (long range contribution). It is currently accepted that both contributions are important to the overall enhancement. The SERS effect has not been explored in the art and archaeology context but the range of possible applications in these fields is wide, with the advantage that it can be used to investigate minute amounts of sample. 8.8
INTENSITY MEASUREMENTS IN RAMAN SCATTERING
A major advantage of Raman spectroscopy, whether conventional or using the FT technique, is the linear dependence on concentration of molecular species and the observed band intensity [45]. Unlike IR absorption
390
Infrared, Raman microscopy and fibre-optic Raman spectroscopy (FORS)
spectroscopy, where the Beer-Lambert law dictates a logarithmic dependency of concentration on absorption intensity, which demands knowledge of the molar absorptivity 1 for the quantification of spectral-intensity data, the linearity of concentration/intensity dependence in the Raman spectrum provides an excellent means of following the growth or depletion of molecular species during a reaction. In this way, Raman spectroscopy has been developed recently as a technique for monitoring reaction kinetics of polymerizations, such as the living anionic polymerization of styrene and butadiene in hydrocarbon solvents by a butyl-lithium initiator. The information about kinetic and thermodynamic quantities from Raman-spectroscopic measurements in solutions over various temperatures has thus provided detailed information about molecular and species equilibria, the presence of dimers or oligomers in a reaction system, hydrogen bonding in alcohol and water systems, nitrating species in acidic media, and tautomeric stability. All of these applications are of importance in commercial systems and synthetic procedures. Biologically, changes in distribution and composition of amino acid residues in large polypeptides and sulphur bridges in keratotic samples studied by Raman spectroscopy have provided an insight into drug-targeting systems and tissue degeneration [46], both of which are vital for biomedical diagnostic applications. Likewise, the absorption of drugs across the human skin barrier, transdermal drug delivery, of importance in pharmaceutical therapeutic formulations has been studied successfully by the FTR spectroscopic technique and the influence of chemical enhancers on drug-delivery systems quantified [47]. However, the linearity between the observed intensity and the concentration of scattering species in the Raman spectrum is not always applicable, especially when the wavelength of laser excitation approaches that of an electronic absorption band in the scattering species under investigation. This is the situation that occurs in RR spectroscopy; the signal strength of Raman scattering from a molecular species is increased by as much as 108 and the spectrum is simplified because only the vibrations associated with the chromophore under excitation are able to generate the signals in resonance. For example, in the case of hemoglobin, laser excitation within the heme absorption (400 – 600 nm) produces a RRS that contains only the enhanced signals of the heme ring vibrational modes and no bands arising from the protein component are observed. Semi-quantitative applications of Raman intensity measurements in the art field are quite few to date, but generally, relative intensity comparisons made between the component pigment bands in a spectrum can provide some
391
Howell G.M. Edwards and Dalva L.A. de Faria
information about mixture composition. Interpretation should be made with caution, however, as other factors such as mineral crystal orientation and laser absorption can affect the results for quantitative purposes. 8.9
RAMAN SPECTROSCOPY WITH FIBRE OPTICS
The capability of Raman instruments to identify and to determine the concentrations of chemical species remotely has been realized, and applications have been forthcoming in many diverse areas—for example, in reactor and polymer processing and in quality control, as well as in the remote sensing of art objects. The use of optical fibres to image the incident laser radiation has resulted in applications of Raman spectroscopic instrumentation to difficult sampling situations. Some recent examples include diagnostic studies of human skin tissue [48], in mummies in museum collections [49], the non-destructive identification of genuine elephant ivory samples embedded in wooden furniture [50], and the monitoring of environmental biodeterioration in cave art [51]. 8.9.1
Sampling considerations
A major factor in the development of remote Raman systems is the ease of sampling, particularly where stringent requirements are made in terms of non-destructive analysis in which chemical or mechanical pre-treatment of the sample is unfavoured or impossible. The coupling of the near IR diode laser to a CCD detector (with extended red wavelength sensitivity) provides a versatile system for remote Raman-sensing experiments. 8.9.2
Probe design
The probes used in remote Raman spectroscopy have been the subject of much technical innovation and discussion, particularly with regard to the number and arrangement of the optical filters used for probe background spectral rejection. Most unfiltered probes currently in use incorporate a single fibre for laser excitation (input) and up to 18 fibres for collection of the scattered radiation. In simple probe construction systems, the fibres for Raman excitation are all aligned and have ends that are square-cut and polished. Probe sensitivity does not increase linearly with the number of fibres used, but it does increase with fibre diameter. Increased collection angles can be provided by means of bevelled tip arrangement or by incorporation of a field lens in the system. Routine laboratory measurements
392
Infrared, Raman microscopy and fibre-optic Raman spectroscopy (FORS)
can be accomplished with probes of about 1 – 2 m in length, normally unfiltered; the most important is spectrometer calibration to facilitate probe background subtraction. Probes may be of the contact type, where the lens touches the surface of the specimen, and non-contact type, where the lens focuses the laser beam up to 5 cm away from the lens surface. The latter design clearly has advantages in art examination through transparent glass or covers without necessitating the removal of the glass or cover. 8.9.3
Probe background
Despite the fact the Raman spectra of glass or silica are weak, the probe background is still often much stronger than the Raman spectrum of the material under study. Often, the fibre-optic probe background, which may be considerable even in short, unfiltered probes, can compromise the study of several important classes of material, especially weakly scattering biological materials such as textiles, parchment, bone and vellum. However, it is still possible to execute spectral substraction routines for elimination of probe background for biomaterials, since the vibrational modes of interest are often in the functionality region between 1000 and 2000 cm21 wavenumber shift. For inorganic species, with vibrational modes occurring below 1000 cm21 shift, the problem of probe background is often not so easily addressed, e.g., metal-oxide stretching modes near 500 cm21 interfere with Si – O frequencies and probe background bands. The problem of probe background in remote Raman-sensing experiments is exacerbated by longer probe fibres; there are now reports of probes in use up to 100 m or more. In such cases, the use of a filtered probe is recommended with a combination of a narrow bandpass and long-pass filters for rejection of silica Raman scattering in the collection fibres and Rayleigh scattering in the excitation fibre.
REFERENCES 1 2 3 4
H.G.M. Edwards, E. Ellis, D.W. Farwell and R.C. Janaway, J. Raman Spectrosc., 27 (1996) 663. H.G.M. Edwards, D.W. Farwell, P.M. Fredericks and J.S. Lee, Spectrochim. Acta A, 59 (2003) 2311. H.G.M. Edwards and F.R. Perez, Biospectroscopy, 5 (1999) 47. H.G.M. Edwards, D.W. Farwell, F. Rull Perez and J. Medina Garcia, Analyst, 126 (2001) 383.
393
Howell G.M. Edwards and Dalva L.A. de Faria 5 6 7 8 9 10 11 12 13 14
15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33
394
H.G.M. Edwards and D.W. Farwell, Spectrochim. Acta A, 51 (1995) 2073. C.V. Raman and K.S. Krishnan, Nature, 121 (1928) 501. H.A. Szymanki (Ed.), Raman Spectroscopy: Theory and Practice, Vol.1 (1967) and Vol. 2 (1970), Plenum Press, New York. H.S. Sands, I.P. Hayward, T.E. Kirkbride, R. Bennett, R.J. Lacey and D.N. Batchelder, J. Forensic Sci., 43 (1998) 509. G.J. Puppels, F.F.M. de Mul, C. Otto, J. Greve, D.J. Arndt-Jovin and T.M. Jovin, Nature, 347 (1991) 301. T.M. Loehr, W.E. Keyes and P.A. Pincus, Anal. Biochem., 96 (1979) 456. S.E.J. Bell, E.S.O. Bourguignon, A.C. Dennis, J.A. Fields, J.J. McGarvey and K.R. Seddon, Anal. Chem., 72 (2000) 234. J.W. Cooley and J.W. Tukey, Math. Comput., 19 (1965) 297. T.N. Day, P.J. Hendra, A.J. Rest and A.J. Rowlands, Spectrochim. Acta A, 47 (1991) 1251. D.B. Chase and J.F. Rabolt, in: D.B. Chase and J.F. Rabolt (Eds.), Fourier Transform Raman Spectroscopy: From Concept to Experiment. Academic Press, San Diego, 1994. G.W. Chantry, H.A. Gebbie and C. Helsum, Nature, 203 (1964) 1052. T. Hirschfeld and D.B. Chase, Appl. Spectrosc., 40 (1986) 133. J.G. Grasselli and B.J. Bulkin (Eds.), Analytical Raman Spectroscopy. Chichester, Wiley, 1991. H.G.M. Edwards, D.W. Farwell, D.L.A. de Faria, A.M.F. Monteiro, M.C. Afonso, P. De Blasis and S. Eggers, J. Raman Spectrosc., 32 (2001) 17. H.G.M. Edwards, D.W. Farwell, A.C. Williams, B.W. Barry and F. Rull, J. Chem. Soc. Faraday Trans., 91 (1995) 3883. H.G.M. Edwards, D.E. Hunt and M.G. Sibley, Spectrochim. Acta A, 54 (1998) 745. W. Akhtar and H.G.M. Edwards, Spectrochim. Acta A, 53 (1997) 81. H.G.M. Edwards and D.W. Farwell, Spectrochim. Acta A, 51 (1995) 2073. A.C. Williams, H.G.M. Edwards and B.W. Barry, Biochim. Biophys. Acta— Protein Struct. Mol. Enzymol., 1246 (1995) 98. H.G.M. Edwards, E. Ellis, D.W. Farwell and R.C. Janaway, J. Raman Spectrosc., 27 (1996) 663. D.L. Dickensheets, D.D. Wynn-Williams, H.G.M. Edwards, C. Schoen, C. Crowder and E.M. Newton, J. Raman Spectrosc., 31 (2000) 633. J.E. Pemberton, R.L. Sobocinski, M.A. Bryant and D.A. Carter, Spectroscopy, 5 (1990) 26. D. Coombs, Int. J. Vib. Spectrosc., 2(2) (1998) 3 www.ijvs.com. J.E. Katon, Micron, 27 (1996) 303. A.R. David, H.G.M. Edwards, D.W. Farwell and D.L.A. De Faria, Archaeometry, 43 (2001) 461. C. Andalo, M. Bicchieri, P. Bocchini, G. Casu, G.C. Galletti, P.A. Mando, M. Nardone, A. Sodo and M.P. Zappala, Anal. Chim. Acta, 429 (2001) 279. L. Burgio, R.J.H. Clark and P.J. Gibbs, J. Raman Spectrosc., 30 (1999) 181. G. Adamopoulos, K.W.R. Gilkes, J. Robertson, N.M.J. Conway, B.Y. Kleinsorge, A. Buckley and D.N. Batchelder, Diamond and Related Mat., 8 (1999) 541. R.J.H. Clark and T.J. Dines, Angew. Chem. Int. Ed. Eng., 25 (1986) 131.
Infrared, Raman microscopy and fibre-optic Raman spectroscopy (FORS) 34 35 36 37 38
39 40 41 42 43 44 45 46 47 48 49 50 51
G. Smulevich, A.R. Mantini, A. Feis and M.P. Marzocchi, J. Raman Spectrosc., 32 (2001) 565. S.A. Asher, Anal. Chem., 65 (1993) 59A. M. Fleischmann, P.J. Hendra and A.J. McQuillan, Chem. Phys. Lett., 26 (1974) 163. A. Campion and P. Kambhampati, Chem. Soc. Rev., 27 (1998) 241. (a) J.A. Creighton, M.G Albrecht, R.E. Hester and J.A.D. Mathew, Chem. Phys. Lett., 55 (1978) 55; (b) R. Keir, D. Sadler and W.E. Smith, Appl. Spectrosc., 56 (2002) 551. H. Saito, Bull. Chem. Soc. Jpn, 66 (1993) 963. M. Moskovits and D.P. DiLella, J. Chem. Phys., 73 (1982) 4408. F. Ni and T.M. Cotton, Anal. Chem., 58 (1986) 3159. F.T. Li, Y. Lu, G. Xue and Q. Cao, Chem. Phys. Lett., 264 (1997) 376. M. Moskovits, Rev. Mod. Phys., 57 (1985) 783. A. Otto, J. Raman Spectrosc., 22 (1991) 743. D.A. Long, Raman Spectroscopy. McGraw-Hill, New York, 1977. E.E. Lawson, H.G.M. Edwards, B.W. Barry and A.C. Williams, J. Drug Targeting, 5 (1998) 343. G.L. Armstrong, H.G.M. Edwards, D.W. Farwell and A.C. Williams, Vib. Spectrosc., 11 (1996) 105. L.P. Choo-Smith, H.G.M. Edwards, H.P. Endtz, J.M. Kros, F. Heule, H. Barr, J.S. Robinson, H.A. Bruining and G.J. Puppels, Biopolymers, 67 (2002) 1. H.G.M. Edwards, M. Gniadecka, S. Petersen, J.P.H. Hansen, O.F. Nielsen, D.H. Christensen and H.C. Wulf, Vib. Spectrosc., 28 (2002) 3. H.G.M. Edwards, D.W. Farwell, J.M. Holder and E.E. Lawson, Stud. Conserv., 43 (1998) 9. H.G.M. Edwards, E.M. Newton and J. Russ, J. Mol. Struct., 550 (2000) 245.
395
This page is intentionally left blank
Chapter 9
Secondary ion mass spectrometry. Application to archaeology and art objects Evelyne Darque-Ceretti and Marc Aucouturier
9.1
INTRODUCTION
Secondary ion mass spectrometry (SIMS) is one of the material analysis methods based on ion beam bombardment. It can be qualified as a microanalysis method in the sense that the analysed elemental volume has one of its dimensions smaller than 1 mm. The principle of the method [1– 6] is centred around the irradiation of a small area of the object surface by heavy ions (possibly by neutral atoms) of low to medium energy (1– 50 keV). The interaction of the incident beam with the material induces atomic collision cascades, leading to primary particles implantation and to sputtering of the target components as charged particles (secondary ions) or neutral atoms and aggregates. Those charged particles (ionized by the sputtering process or sometimes obtained by post-ionization) are filtered in a mass spectrometer in order to reach the target elemental composition. The general performance of that method for solid materials may be summarized as follows: – a very high sensitivity (very low detection limits) for nearly all elements of the periodical classification, including hydrogen (trace analysis), – possible access to elemental isotopic information (allowing the use of isotopic tracers), – the measurement, in dynamic rating, of concentration profiles from the surface over a limited depth (thin film analysis, diffusion profile), – a possible access, in static rating, to the elemental composition, optionally the molecular composition, of the very first atomic or molecular layers, – the determination, with a good spatial resolution, of the lateral and in-depth distribution of elements and sometimes chemical species, Comprehensive Analytical Chemistry XLII Janssens and Van Grieken (Eds.) q 2004 Elsevier B.V. All rights reserved
397
E. Darque-Ceretti and M. Aucouturier
– the use of the “matrix chemical effect” to identify and possibly quantify, chemical compounds. As a result of the complex nature of the obtained mass spectra and of the difficulties of the quantification procedure, that method is not designed for quantitative analysis of alloying elements at high concentrations, for which more precise, more rapid and less expensive methods should be preferred, as X-ray fluorescence (Chapter 4), high energy ion beam analysis (Chapter 5), electron probe microanalysis (Chapter 3), analytical scanning electron microscopy (Chapter 3), laser-ablation inductively coupled mass spectrometry (Chapter 7), wet-chemical or spectro-chemical methods of analysis of sampled material. SIMS belongs primarily to the methods that are suited for the analysis of the uppermost surface layers, the interfaces between surface layers and substrates and optionally the extreme surface. Amongst microanalysis methods for cultural heritage materials, its application can be found in: – the identification and interpretation of the causes of pollution, corrosion or degradation phenomena of metallic, mineral (glass, ceramic, gems, etc.) and even organic objects and artefacts, – providing detailed knowledge and understanding on surface treatments as patination, decoration, protection, corrosion inhibition, and – identification of the mechanisms and process of interface adhesion in multilayered materials as surface-coated metals, glasses and ceramics. The ability to perform isotopic analysis with a high sensitivity and mass resolution also leads to applications such as artefact dating or the determination of the provenance of mineral materials. In dynamic SIMS analysis of solid materials, the primary ion beam creates a small hole (often called the sputtering crater) on the object surface. This is unavoidable because the principle of the method itself applies to sputtered matter analysis. Thus the method cannot be strictly qualified as a non-destructive method. It is moreover necessary to introduce the specimen in a high vacuum chamber of limited size. For small and flat artefacts which may be introduced as such into the vacuum chamber, and are able to support a high vacuum holding, one may consider that, owing to the very small quantity of sputtered matter (a crater smaller than 100 £ 100 £ 10 mm), SIMS is a micro-sampling analysis method, and that the integrity of the object is not affected. This is even more true for static SIMS analysis (SSIMS), where the sputtered matter during one experiment may be less than one or optionally a few atomic or molecular surface layers.
398
Secondary ion mass spectrometry. Application to archaeology and art objects
This chapter will first summarize some basic principles about secondary emission, a phenomenon that needs to be understood for a correct interpretation of the analyses, followed by a simplified description of the available equipment. The various analytical procedures that can be applied in SIMS and SSIMS analysis will be reviewed with the aim to guide the reader to the choice of an optimal analytical process for a given purpose. Some applications to the specific field of cultural heritage objects and materials will be developed in the last section. 9.2 9.2.1
PRINCIPLES AND EQUIPMENT Principles
The collision cascade process that leads to ion sputtering under ion bombardment is described by the Sigmund model [7], which has been developed for atomic elementary isotropic solids. A part of the deposited energy is reflected toward the surface through recoiled atoms (target atoms mobilized by the impact of incident ions or through cascade collisions). When the energy of the recoiled atoms near the surface is larger than the binding energy of the surface atoms, these are ejected into the vacuum environment. The remaining deposited energy is dissipated through creation of point defects. Subsequent slowing down of the primary particles leads to their implantation. In summary, the interaction volume of the primary particles with the solid is of the order of their implantation depth (from a few nanometers to some tens of nanometers, depending on the primary beam energy and the target composition), but the sputtered atoms are originated from not more than the two first atomic layers (100% of the first one and ca. 30% of the second one). It is important to note that the primary ions energy dissipation in the interaction volume (sometimes called “perturbed” or “modified” layer) may lead to composition modifications, creation of defects and even structure amorphization, uncontrolled displacements of the components by recoil effects or accelerated diffusion. The analytical consequences of these perturbations may be important and will be discussed further. Another point is that ion irradiation induces the emission of large amounts of secondary electrons, and this may lead, for insulating materials, to large shifts of the surface electric potential (charging effects). In the Sigmund model, the time delay between the impact of a primary ion with an energy E0 and emission of the secondary particles is of the order 399
E. Darque-Ceretti and M. Aucouturier
of 10212 – 10213 second. This point is important to evaluate the distinction between the dynamic and static mode of analysis (section 9.2.2.1). The energy distribution of the secondary particles is described by the equation: dn 3 aSn ðE0 Þ E ¼ dE C0 2p2 ðE þ Us Þ3
ð9:1Þ
where Sn ðE0 Þ is the nuclear stopping power, C0 a factor based on ion – atom interaction potentials, a a dimensionless factor depending on the mass ratio M1/M2 between the incident and target atoms and on the incidence angle and Us the surface binding energy. This distribution has a maximum at E ¼ Us =2 (between 1 and a few eV) and a spread of a few hundreds of electron-volts. 9.2.1.1 Analytical aspect The analytical process is governed, apart from the particle collection efficiency of the equipment, by two fundamental efficiencies: the sputtering yield and the ionization yield (see further: quantitative analysis). Theoretical interpretations of secondary ion emission are presently not developed enough to allow a comprehensive formulation of these yields directly applicable to quantitative analysis. So, the following assessments should be considered only as a guide for the analyst. Sputtering yield The sputtering yield ST is defined as the ratio of the total number of emitted particles nT from the target to the number of incident particles N0 from the incident primary beam. ST ¼
nT n þ n B þ nC þ … ¼ A N0 N0
ð9:2Þ
where nA, nB, nC,…are the respective numbers of particles, issued from the target elementary components (A, B, C,…) that are simultaneously emitted. The Sigmund formula [7] expresses the sputtering yield of a pure substance, assumed to be non-crystalline and isotropic, of atomic number Z2, under irradiation with particles of an inert gas of atomic number Z1 and E0 energy: ST ¼
400
3 FD ðE0 ; uÞ 4p2 C0 NUs
ð9:3Þ
Secondary ion mass spectrometry. Application to archaeology and art objects
where C0 ¼ 12pa 2, with: a ¼ 0.885a0Z21/2, a0 ¼ 0.0529 nm (Bohr radius), 2=3 2=3 Z ¼ Z1 þ Z2 ; N is the target atom number per unit volume; Us the surface binding energy, FD the fraction of the deposited E0 energy, per unit of length, under u incidence (angle between the bombardment direction and the target surface normal). Since the sputtering mechanism is purely a collision mechanism, apart from experimental parameters as the incident energy and angle, the only factors which affect the sputtering yield are the atomic number and mass of the incident and target elements, and the surface binding energies of the target components. Compilations of the sputtering yields for numerous pure elements as a function of the nature and energy of the incident ions may be found in the literature [8]. Figure 9.1 [9] shows an example of the ST variations for silver as a function of the primary beam energy for various primary ions. The shape of the curves is similar for most of elements in this energy range. For alloys or compounds, the binding energy varies with the atomic environment of each component, and this means that the sputtering yield
Fig. 9.1. Silver target. Variation with energy of the sputtering yield ST for various primary beams of increasing energy [9].
401
E. Darque-Ceretti and M. Aucouturier
does not vary linearly with concentration (sputtering “chemical” effect). The individual sputtering yield of each component SA, SB, SC, … is defined by: n Si ¼ i ¼ Cpi ST ð9:4Þ N0 where Cpi ¼ ni =nT is the concentration of the element i in the sputtered matter. For a non-crystalline solid, the influence of the primary beam incident angle u on the sputtering yields theoretically follows the simple law: ST ðuÞ ¼ S0 ðcos uÞ2f
ð9:5Þ
The f exponent is mostly a function of the M2/M1 ratio. It ranges from 5/3 for M2/M1 , 3 to a value slightly smaller than 1 for M2/M1 . 3. Another parameter is the crystalline anisotropy of most materials. The atomic collision process in crystalline solids is affected by canalization phenomena. As a summarized image, the solid exhibits “transparent” crystallographic directions (low Miller index) for which the interaction with incoming ions is low or absent, and “opaque” directions with a highly efficient sputtering. Ionization yield The ionization yield of a given species (A) is defined, from a phenomenological viewpoint, as the proportion of ions (A^) of that species generated out of one sputtered particle of (A), or also as the ionization probability of the (A) species as (A^) ions. n ^ ð9:6Þ PðA^ Þ ¼ A nA In a theoretical description, its expression should take into account the chemical nature of the primary ions, the atomic environment of the particles before their emission, and the possible recombination – ionization occurring after sputtering [10]. The general behaviour may be summarized by means of a few simple semi-quantitative rules, valid for sputtering under non-reactive incident beam (rare gas) (Table 9.1): – ionization yields are always very small (less than 1023 under rare gas primary beam); – elements with a low ionization potential (for instance metals) are good positive ions emitters; – elements with a high electron affinity (for instance metalloids, non-metals or noble metals) are good negative ions emitters;
402
Secondary ion mass spectrometry. Application to archaeology and art objects TABLE 9.1A Emission of positive ions
TABLE 9.1B Emission of negative ions
403
E. Darque-Ceretti and M. Aucouturier
– some elements, such as N, Zn, Hg, and Cd, are poor emitters of monoatomic ions of both polarity; – as a consequence of both sputtering and ionization process, the emitted ions exhibit continuous energy spectra, presenting a maximum at low energy and spreading more or less to higher energies, depending on their nature and their monoatomic or polyatomic character. These rules, even qualitative, are only predictive for the emission of monoatomic ions from a pure target, under bombardment by a non-reactive primary beam. Indeed, the ionization yield is strongly dependent on the atomic environment of the given species before its sputtering, i.e., on the chemical nature and composition of the target (“chemical or matrix effect”). It depends furthermore on the gaseous environment of the material. For example, the detected ion current of a metal under rare gas sputtering may vary by several orders of magnitude when comparing an un-oxidized and an oxidized metallic target (Fig. 9.2) [6]. In some alloys, these “exaltation effects” may be extremely dependent on the composition of the alloy itself. A known method to enhance the ionization yield of electropositive elements (minimizing at the same time the crystallographic effect) is to submit the specimen to an oxygen partial pressure, which leads to the covering of the surface by a “pre-ionized” adsorbed layer. The improvement may reach a factor 102 for an oxygen partial pressure of 1023 Pa (Fig. 9.3 [11]). Reactive sputtering Reactive sputtering takes advantage of the exaltation effect. The primary beam, which undergoes implantation into the specimen during the analysis, is chosen in such a manner as to significantly modify the chemical
Fig. 9.2. Variation of the ion yield during sputtering of an oxide layer NiO, under argon bombardment [6].
404
Secondary ion mass spectrometry. Application to archaeology and art objects
Fig. 9.3. Evolution of the Cuþ and Znþ ion yields under argon bombardment, as a function of the oxygen partial pressure, for pure copper, pure zinc, brass with 5 and 23% Zn [11].
environment of the target elements or the surface potential barrier (electron extraction energy). This modification leads to an enhancement of the ionization yields towards better sensitivity or detection limits. Sputtering under an oxygen ion beam leads to a much better detection of elements with a low ionization energy (electropositive metals). Sputtering under an alkaline ion beam (for instance Csþ) favours the detection of elements with a high electron affinity (metalloids, noble metals). Reactive sputtering is now a routine method in most SIMS analyses. 9.2.1.2 Nature of the emitted ions Secondary ion emission generates monoatomic and polyatomic ions, singlecharge and multiple-charge ions. Multicharged ions are positive ions, for instance Oþ þ , Nþ þ , Alþ þ or þþþ Al , Siþ þ or Siþ þ þ , Caþ þ , etc. The emission rules are the same as for single-charge ions but with more complex mechanism because of intervention of several ionization energies.
405
E. Darque-Ceretti and M. Aucouturier
Polyatomic ions (also called molecular ions), both positive or negative, are very important components of the emission spectra. Some examples are given for pure aluminium (Fig. 9.4), some oxides (Fig. 9.5) [12], or a polymer (Fig. 9.6) [13]. The series of polyatomic ions emitted from a compound under given analytical conditions may be considered as its “analytical signature”. One must nevertheless keep in mind a basic concept: unless under particular conditions, the polyatomic ions are generally not representative of the chemical species present in the analysed specimen. It is indeed demonstrated [14] that, at least in dynamic SIMS rating, most of the polyatomic ions are generated by recombination of emitted monoatomic ions on the top of the surface, after sputtering. Figure 9.5 is an illustration of that difference between the actual composition of the target and the nature of the detected polyatomic ions. In static SIMS rating (without overlaying of the successive primary ion impacts), the mechanism is more complex (section 9.2.1.3). Polyatomic ions may also originate from contaminating elements of the surrounding residual atmosphere (CmH^ n ), and are, from this viewpoint, a source of analytical error. They may be generated by the combination of one þ target component and the primary beam element (MxO^ y , MCs , etc.). þ þ For instance, it has been shown that MCs and MCs2 ions are formed by recombination of the target M neutral atoms with the Csþ or Csþ 2 ions simultaneously sputtered; the consequence is a minimization of the matrix effect. The fact that the polyatomic ions are generated by recombination of monoatomic ions that are emitted simultaneously strongly changes their
Fig. 9.4. Yield of Alþ n polyatomic ions for polycrystalline aluminium under argon bombardment [12].
406
Secondary ion mass spectrometry. Application to archaeology and art objects
þ Fig. 9.5. Yield of MmOþ (6 keV) n ions emitted for various oxides under Ar bombardment [12].
energy distribution. Their energy spectrum is logically narrowed on its large energy side (by consumption of the kinetic energy for the recombination process) (Fig. 9.7 [15]). This feature may be used to minimize the contribution of the polyatomic ions to the mass spectrum by energy filtering. 9.2.1.3 Static and dynamic sputtering ratings As described in the next section, the primary beam is usually focused and rastered (scanned) over a limited surface of the analysed specimen (104 – 105 mm2). Knowing that the sputtering yield is always (and even largely) higher than 1, a primary ion flux of 1015 ions cm22 s21 (i.e., 1.6 mA mm22) is sufficient to sputter more than one atomic layer per second from the specimen. If one takes into account the acquisition time necessary for ensure an adequate level of accuracy, which is much higher than 1 s in standard equipment employing sequential detection, analysis under a continuous primary beam provides information on a surface continuously renewed by the sputtering process. This operating mode, the most commonly applied, is called dynamic SIMS. Analysis in static circumstances, or SSIMS [16,17,18] tends to decrease by several orders of magnitude the primary ion flux so that the secondary ion acquisition concerns only a fraction of one atomic layer sputtered from the surface. Since the secondary particles are emitted from the two first atomic
407
E. Darque-Ceretti and M. Aucouturier
Fig. 9.6. TOF-SIMS mass spectrum of secondary ions emitted by four polymethacrylate polymers differing by their ester group: PEMA, Poly(ethylmethacrylate); PnPMA, Poly(n-propylmethacrylate); PsBMA, Poly(s-butylmethacrylate); PcHMA, Poly(cyclohexymethacrylate) [13].
layers of the target, SSIMS truly is a method for analysis of the “extreme surface”. From the preceding argumentation, the integrated primary particle flux (dose) in this case must be much smaller than 1014 ions cm22. This aim is realized either by lowering the primary current (a few pA for a rastered area of 100 £ 100 mm2), or by chopping the primary beam as very short duration
408
Secondary ion mass spectrometry. Application to archaeology and art objects
þ Fig. 9.7. Emission energy distribution for Cuþ, Cuþ 2 and Cu3 ions emitted from a þ copper polycrystal under Ar bombardment (6 keV) [15].
pulses while keeping the raster running. In this way, the impact zone of one particle pulse does not overlap with the impact zone of the preceding pulse, and the dose received by a unit area remains very low. From the viewpoint of specimen modification induced by the primary beam, this mode of operation can be subdivided into two more cases: – Either, the dose received by the material during the entire analysis is such that none of the ejected secondary particle can originate from a zone already modified by the preceding primary impacts. This could be called “true static mode”. One considers that this condition is fulfilled when the total dose received is smaller than 1013 ions cm22, but it is clear that this limit depends on the sputtering conditions and on the nature of the detected secondary particles (size, surface occupation ratio). – Or, it is the dose received during one analysis sequence (one measurement cycle of all chosen elements) that fulfils the above condition (1013 – 1014 ions cm22 per sequence), but the following sequences will impinge upon a surface already modified by former sequences. The specimen is
409
E. Darque-Ceretti and M. Aucouturier
indeed analysed layer by layer, but it is not really an extreme surface analysis. This rating is sometimes called “quasi-static rating”. Experiment shows that in the latter case, in contradiction with the abovementioned hypotheses about polyatomic ion emission, the surface bombardment induces, particularly for molecular solids, the emission of intact molecular species or polyatomic fragments of these molecular species. This sputtering process cannot be described by the Sigmund model. Several models have been developed to explain this experimental observation [19]. All are based on the fact that the molecular ions are emitted from the periphery of the primary particle impact, and this explains why it happens only in static rating, where each impact is spatially separated from the others. This occurrence of molecular ion emission brings a very powerful tool for the identification of extreme surface compounds (pollution species, adsorption, etc.), and for polymer surface analysis (section 9.3.4). 9.2.2
Equipment and choice of analytical parameters
Secondary ion emission analysis of solid materials is based on the irradiation of specimen by a primary ion beam focused on the surface as a probe. The primary beam undergoes a periodical deflection which ensures a continuous rastering of the surface along two perpendicular directions ðx – yÞ: The prime aim of rastering is to ensure, in dynamic mode, an homogeneous erosion of the surface that recedes parallel to itself during the analysis. The primary beam rastering defines the rastered area or irradiated area (for instance 100 £ 100 mm2 to 500 £ 500 mm2), which limits, together with the perturbed depth defined above, the specimen volume modified by the primary beam (i.e., 102 –5.103 mm3). The ionized fraction of the sputtered species from this rastered area (also called the secondary ion beam) is lead into a mass spectrometer where they are identified ( filtering) and their intensity measured. In dynamic mode, the intensities are recorded as a function of sputtering time, which is a measure of the eroded depth from the surface onwards. An (empirical) relation between the erosion time and the depth reached into the specimen must be established. 9.2.2.1 Microscope mode and microprobe mode The collection and treatment of the secondary ion beam is such that for all ions emitted by the rastered area, a part is selected that originates from a smaller area, the analysed area, which is not perturbed by the edges of the rastered area. It is also possible to obtain images of the space distribution
410
Secondary ion mass spectrometry. Application to archaeology and art objects
of the ion species in an imaged area, also smaller than the rastered area. Two processes are used to define the analysed and imaged areas and to obtain the expected ion images (Fig. 9.8). The first method, defined as the microscope mode, is to focus “optically” the secondary ion beam in order to obtain an image or successive images of the surface on beam course, and to introduce in one of these image plans a “field aperture” that allows a delimitation of the analysed area. The analysis and the production of the final image concerns the entire area limited by the aperture and the final lateral resolution of the image is limited by the “optical” aberrations of the system (chromatic aberration, i.e., energy dispersion; spherical aberration; astigmatism; and diffusion at the aperture edge).
Fig. 9.8. Microscope mode and microprobe mode.
411
E. Darque-Ceretti and M. Aucouturier
The second method, defined as the microprobe mode, is based on the detection of the mass-filtered secondary ions in synchronization with the primary beam raster (after correction for the time of flight of these ions in the analyser). The signal detected at a given instant corresponds to one specific position of the primary probe on the specimen surface. The secondary ions distribution image may be then reconstructed by feeding the ion intensity signal to a CRT tube whose rastering is itself in synchronization (with time-of-flight correction) with the primary raster. The analysed or imaged area is defined by electronic selection of the corresponding signal. The lateral resolution depends on the quality of that selection, on the probe size and on the detection time unit or image area unit (pixel).
9.2.2.2 Equipment components The main components of a secondary ion analyser are (Fig. 9.9): – one or several primary ion sources; – a “primary beam column” where primary ions are filtered, focused and rastered; – a specimen chamber; – a “transfer column” where the secondary ion beam is optically treated before the filter; – a mass spectrometer, often coupled with an energy filter; – a detection and imaging system; – an electronic system for signal treatment. The various fittings available in laboratories or on the market belong to three main categories, depending on the principle used for mass filtering (Figs. 9.9 – 9.11): – analysers with “triple focusing” (Fig. 9.9 [20]), in which filtering is ensured by a magnetic sector spectrometer, preceded by a electrostatic sector. The aim of the electrostatic sector is to compensate the energy distribution of the secondary ions. Detection is either sequential (magnetic field variation in the magnetic sector) or parallel by multiple collection; – analysers with a quadrupole spectrometer (Fig. 9.10 [20]); – time-of-flight analysers (ToF-SIMS analysers, Fig. 9.11 [20,21]), for which the energy compensation may be realized either by a “reflectron” system or by three electrostatic sectors.
412
Secondary ion mass spectrometry. Application to archaeology and art objects
Fig. 9.9. General scheme (ion path) for a “triple focusing” secondary ion analyser: (a) with sequential detection (CAMECA IMS 4F), (b) with parallel multidetection (CAMECA IMS-1270) [20].
413
E. Darque-Ceretti and M. Aucouturier
Fig. 9.10. General scheme (ion path) for a quadrupole analyser (RIBER-CAMECA MIQ 256) [20].
Fig. 9.11. General scheme (ion path) for a time-of-flight analyser: (a) with “reflectron” [20], (b) with a triple electrostatic deflector [21].
414
Secondary ion mass spectrometry. Application to archaeology and art objects
Complementary equipment may be added, such as device for primary ion neutralization (FAB, for fast atom bombardment), or for post-ionization of the secondary particles (SNMS for secondary neutral mass spectrometer). In the most recently developed equipment [20,22] “nano-SIMS”, a “triple focusing” analyser is used with a computerized control of all coaxial ionic and electronic beams and a liquid metal primary source, which provides a lateral resolution of 20 – 30 nm. In the following section a short description of the different component will be given, taking as example the most widely used equipment (Figs. 9.9 and 9.11b) and emphasizing the influence on the final analysis of the various experimental parameters.
9.2.2.3 Primary beam production and optics. Choice of parameters The choice of the primary ion source is driven by the nature of the elements to be analysed and by the desired lateral resolution. Rare-gas sources are universal from the viewpoint of the produced secondary ions but the ionization yield for the secondary particles is poor and the lateral resolution (probe size) is not very good. Oxygen sources are very widely used for analysis of elements with a low energy ionization energy (positive secondary ions). A cesium source is chosen for analysis of the species with a high electron affinity (negative secondary ions) or for analysis of species detected as secondary MCsþ ions (minimization of mass interference, minimization of matrix effect, detection of hardly ionizable species). Liquid metal (Gaþ) sources have the highest brilliancy and give rise to the best lateral resolution. Electronic impact sources are easy to be “pulsed” but provide a poor lateral resolution. They are only used in ToF-SIMS. The pulsed source for ToF-SIMS may also be a chopped Csþ or Gaþ source. The choice of the primary acceleration voltage determines the value of the energy of the primary ions, equal in eV to this voltage increased or diminished by the specimen polarization voltage. This energy has a decisive importance for monitoring the sputtering yield and the specimen thickness concerned by the primary beam implantation and associated effects. The primary beam is focused on the specimen. The minimum probe size is 0.5 –1 mm for gas sources, 0.2 mm for microbeam cesium sources, and much less than 0.1 mm for liquid gallium sources. This size will determine the lateral resolution in microprobe mode. The primary ion current varies from some pA to hundreds of mA.
415
E. Darque-Ceretti and M. Aucouturier
The beam is positioned and rastered on the specimen surface by deflection plates. The rastered surface, which fixes the size of the sputtering crater may be varied between the probe size and a square of 500 £ 500 mm2. In FAB mode, the primary beam is neutralized by crossing a chamber with a neutral gas at partial pressure. In continuous settings, the primary flux on the specimen varies between 1 nA mm22 (some 1011 ions cm22 s21) and hundreds of mA mm22 (some 1015 ions cm22 s21). Thus, one can continuously switch from dynamic to static SIMS conditions. In pulsed settings (ToF-SIMS), an electronic impact argon source provides 1000 ions, for a 0.8 ns pulse and a diameter of 100 mm while a liquid gallium source provides 2 primary ions in a 2 mm spot for a pulse of 3 ns. In this manner, the primary ion dose for a repetition rate of 10 kHz becomes 2 1010 – 1011 ions cm22 s21. Primary beam incidence angle The angle between the specimen surface and the primary beam is either adjustable (with quadrupole apparatus) or fixed (with deflection or ToF equipment). Due to the specimen polarization, necessary to ensure extraction of the secondary ions, the beam-is deflected and the final impact angle u (relative to the surface normal) is given by the formula: sffiffiffiffiffiffiffiffiffiffiffiffi HT sin u ¼ sin u0 for primary and secondary ions with ð9:7Þ HT þ V different polarities sffiffiffiffiffiffiffiffiffiffiffiffi HT for primary and secondary ions with sin u ¼ sin u0 HT 2 V
ð9:8Þ
the same polarity Here, u0 is the geometrical angle between the primary column and the surface normal, HT is the primary acceleration voltage and ^V is the specimen polarization voltage. Specimen chamber vacuum The vacuum inside the specimen chamber is responsible for the specimen surface “cleanness”. In dynamic setting, the surface is continuously sputtered and its ability to reconstruct a pollution layer depends on the sputtering rate and the chamber vacuum. A vacuum of 1025 –1027 Pa usually is sufficient to avoid pollution-related disturbing effects. Static SIMS measurements evidently need a vacuum of much better quality (better than 1027 – 1028 Pa). It is sometimes useful to provide a controlled partial
416
Secondary ion mass spectrometry. Application to archaeology and art objects
pressure of oxygen (1022 – 1024 Pa) in order to improve the ionization yield of some species (metals). 9.2.2.4 Optics and spectrometry of secondary ions. Choice of parameters A X – Y movement of the specimen and sometimes a rotation around the surface normal are usually possible. To obtain secondary ion extraction, the specimen is polarized positively (e.g., þ 4500 V) for positive secondary ion detection and negatively (24500 V) for negative secondary ion detection. In microscope mode, the extracted secondary beam is treated by electrostatic lenses to produce an enlarged real image of the surface where a field aperture delimits the analysed area. It is focused on a “cross-over” which is the spectrometer entrance aperture. The enlargement of the transfer optics and the field aperture diameter define the lateral resolution in microscope mode of the analysis. The size of the entrance aperture or slit defines the mass resolution of the spectrometer. The combination of these parameters defines the collection efficiency of the equipment, i.e., the final sensitivity. Energy dispersion compensation As already stated, secondary ions are not emitted with a uniform energy. This energy dispersion will lead to a parasite dispersion in the mass spectrometer and a chromatic aberration of the ionic images. To avoid this, an energy compensation is provided by the electrostatic sector in the magnetic deflection equipment, by three electrostatic sectors or a retarding tube in ToF-SIMS equipment. Energy filtering of the secondary ions is possible thanks to such complements and is useful to eliminate interfering polyatomic ions. Insulating specimens The bombardment by ions, the extraction of secondary ions, and overall the resulting expulsion of secondary electrons produce strong charge effects in insulating specimens, changing their surface potential. To avoid such effects, modern equipment includes a compensating electron gun that provides an electron beam on the surface during the analysis [23]. Post-ionization In order to minimize matrix effects, it may be useful to add a device that ionizes the sputtered neutral particles, for instance by laser irradiation, electron bombardment, or in a heated furnace. The method is then denoted SNMS (secondary neutrals mass spectrometry) or STIMS (secondary thermally ionized mass spectrometry) [24– 27]. In resonance laser postionization [28], the laser wavelength is tuned to ionize selected species.
417
E. Darque-Ceretti and M. Aucouturier þ Analysis by detection of MCsþ or MCsþ 2 ions under Cs bombardment is assimilated to a post-ionization process, as these secondary ions are generated by recombination of the neutral matrix particles with Csþ ions after sputtering.
Spectrometer, mass resolution In the “triple focusing” equipment (Fig. 9.9), mass filtering is obtained by applying a adjustable magnetic field in a 908 magnetic sector. As stated above, the energy dispersion of the secondary ion beam is compensated before mass filtering by an electrostatic sector and an intermediary focusing lens. A selection slit detects the ions for a given magnetic field value, yielding a sequential mode of data acquisition. The width DM of the peak representative of a given M=q value (M, atomic or molecular mass, q, charge) is proportional to M. The mass resolution is quantitatively defined as the ratio M=DM: Its value may be adjusted between 300 and 10,000 (DM defined at 10% of the peak maximum). In microscope mode, the available spectrometer parameters are: – the enlargement of the ionic image, which controls, with the field and contrast apertures, the collection efficiency and the lateral resolution; – the diameter of the field selection aperture, which defines the analysed area; – the diameter of the contrast aperture or the width of the entrance slit, which controls directly the mass resolution; an improvement of the mass resolution (narrowing the entrance slit) results in a degradation of the collection efficiency (sensitivity); the width of the selection slit is adapted to the width of the entrance slit; – optionally, the width of the energy selection slit, which allows the selection of secondary ions of a given energy, for instance to eliminate interference. In microprobe mode, the limitation of the analysed area is done by electronic selection. The other parameters have the same role as stated above. Some “triple focusing” equipment are modified to ensure a parallel detection (a larger magnetic sector, followed by a sextupole selector), which allows the simultaneous detection of secondary ions of different mass. Quadrupole spectrometers make use of four parallel bars polarized by a direct voltage ^U plus a radio-frequency voltage. The mass peak width DM can be kept constant for all masses. The explored mass range and the mass resolution are more limited than with magnetic sector spectrometers, but the
418
Secondary ion mass spectrometry. Application to archaeology and art objects
specimen geometry is more flexible and the chamber vacuum is usually better. The available parameters are essentially the same as for magnetic sector spectrometers. The time-of-flight spectrometers are based on measurement of the time evolved between the impact of the chopped or pulsed primary beam (ca. 10 kHz frequency) and the detection of the secondary ions. The mass separation is obtained by recording the arrival time of the different secondary ions after one pulse. Each individual pulse (of nanosecond order) defines the “zero” and the flight time is larger than 1 ms. The secondary ion energy dispersion is compensated by an electrostatic mirror (“reflectron”) or by a set of three electrostatic sectors. The mass resolution M/DM varies as M1/2 and is thus better for the large masses (up to 11,000 at the mass 100 for primary pulses of 800 ps). The available parameters are, apart from the already described optical parameters, the pulse duration and frequency, and the time resolution of the recording electronics, which define the mass resolution and the sensitivity (number of detected ions). Time-of-flight spectrometers have an intrinsic transmission of the same order as magnetic spectrometers, bur allow sensitivity improvement thanks to their parallel detection and to the fact that the transmission is not limited by the mass resolution. Detection and recording of spectra, profiles and images The ion detection is done either by an electron multiplier, by a channeltron or by a channel-plate followed by a fluorescent screen or a CCD camera. The individual signals detected for each ion at each moment are sorted in order to obtain either a mass spectrum (signal versus atomic mass) or a signal versus time diagram for a given mass. In static mode (ToF-SIMS), the mass spectrum is directly obtained by recording the signal as a function of the time of arrival of the ions to the detector. In dynamic mode, the sputtering time is representative of the eroded depth and the signal versus time curve may be converted into a concentration versus depth profile by adequate calibration of the erosion time-eroded depth relation and a quantification of the ion yield. The depth-scale calibration is obtained by measuring with a mechanical profilometer the depth of the eroded crater at the end of the analysis. The concentration calibration (quantification) will be discussed further. The channel-plate/screen or camera set allows, in microscope mode, a direct ion image recording. The lateral resolution in this mode is limited by the optic aberrations (chromatic, i.e., energy dispersion-induced, and spherical aberrations) of the ion transfer optics.
419
E. Darque-Ceretti and M. Aucouturier
In microprobe mode, the image is reconstructed by synchronization of the detection with the primary probe rastering, each detection pixel corresponding to one position of the probe on the specimen surface. The lateral resolution is limited by the probe size and the detection time unit (pixel). In dynamic setting, when the image is quantified as a function of the position of the emitting point and of the sputtering time (CCD camera in microscope mode, synchronization in microprobe mode), an adequate signal treatment permits to construct a 3D distribution of each detected ion emitted by the specimen. 9.3
ANALYSIS PROCEDURES
It is not possible to describe secondary ion emission from solid materials in a quantitative manner by a simple theory. In addition, several analytical procedures may be used, and each procedure contains its own set of analytical difficulties and specific experimental artefacts. A universal analytical procedure therefore does not exist. In the following, the various analytical procedures adopted to answer to given questions are described, going from simple to complex, taking into account the analytical questions raised and the performances expected to solve a given problem. 9.3.1
Elemental identification, sensitivity
9.3.1.1 Selectivity The basic method for chemical elements or species identification is mass spectra recording. The expected performances are an exact identification of the species (selectivity) and the lowest possible detection limit (best sensitivity). The principal problem in the interpretation of the spectra is the existence of a large number of ion species, monoatomic and polyatomic, leading to a great number of possible interferences for a given mass number. Possible answers to that problem are as follows: – the choice of the primary ions and of the polarity of the detected secondary ions is essential: Oþ 2 ions (or partial pressure of oxygen) and detection of positive secondary ions are suitable for analysis of low ionization energy species as current metals, silicon, carbon; Csþ ions and detection of negative secondary ions for analysis of high electron affinity species as metalloids, non-metals, noble metals; Csþ ions and detection of MCsþ (or MCsþ 2 ) secondary ions for analysis of specific species as N, Zn, Cd, Hg, etc; 420
Secondary ion mass spectrometry. Application to archaeology and art objects
– the comparison of spectra of both polarity gives interesting information on the occurrence of parasitic species; – pure element spectra exhibit “spectral fingerprints”, that is series of natural isotopes with well-defined intensity ratios. For polyatomic ions, the “rule of the polynomial development” must be used: if a molecule contains n atoms, each with p isotopes of abundance ai, each contribution is deduced from the terms of a ða1 þ a2 þ … þ ap Þn polynomial development. – some molecular compounds present in the specimen may also produce spectral fingerprints resulting from their splitting up under irradiation. This is particularly true for organic compounds analysed under static SIMS setting, as already mentioned. Special care must be taken in the interpretation of such molecular fingerprints by comparison with fingerprints proposed by various published databases [29,30], valid only if calibrated with standards analysed with the same equipment in the same conditions. Figure 9.12 [31] shows an example of the mass spectrum (positive ions) obtained by ToF-SIMS from a complicated organic compound adsorbed on a silver substrate; – the same procedure may be applied for mineral species, for instance oxides (Fig. 9.5). It must be emphasized again that, especially in dynamic
Fig. 9.12. ToF-SIMS mass spectrum by of a vitamin B12 deposit (mass number 1356) on a silver substrate [31].
421
E. Darque-Ceretti and M. Aucouturier
mode, such polyatomic ions are generally produced by recombination above the surface after sputtering and are not at all representative of the stoechiometry of the analysed compounds; – secondary ion energy selection may also be a help for eliminating undesired ions from the spectrum, specially polyatomic ions. This procedure is often used to eliminate the hydrocarbon ions CxH^ y , very often present in the negative ion spectra; – but the most efficient method to identify and eliminate interferences in mass spectra is the improvement of the spectrometer mass resolution to high resolution spectrometry. Well-known examples concern the separation of the 32S peak from the 16O2 peak and the analysis of 31P in a silicon-containing material where hydrogenated ions (28SiH3, 29SiH2, 30SiH) are easily formed with the same mass number. In this last case, the mass resolution needs to be larger than M/DM ¼ 5000 to ensure a correct separation. Another example [31] is the separation in ToF-SIMS of the polluting species on the surface of a silicon-containing material, as 28Si2, 56Fe, 28SiCO, 29SiCNH, 23NaO2H, 28 SiCNH2, C3OH4, C2NH4, etc. that all have mass number 56. An increase of the mass resolution inevitably leads to a loss of transmission, i.e., a loss in sensitivity. 9.3.1.2 Space resolution, detection limit, precision As for all microanalytical methods, the detection limits (minimal volume fraction or surface fraction detectable for a given species) and the precision (possible separation between neighbouring values or uncertainty on a given measurement) are directly related to the analysed volume. Any limitation of this volume to improve the spatial resolution leads to a degradation of both figures-of-merit. The trueness (i.e., the extent to which an analysis result is representative of the true value) may be affected, because of the lack of accuracy about the analysed volume. Figure 9.13 summarizes this relation between detection sensitivity and the analysed volume, which itself is dependent on the beam radius and the primary ion current density, i.e., the number of analysed atomic layers per unit of time. The detection limit of a given element in a given matrix may be expressed either as a concentration (%, atom ppm, at. cm23, etc.) or as a number of atoms per surface unit (at. cm22, fraction of atomic layer, etc.) when the analysed depth is small (static SIMS). When detection limits are given, the analysed volume or area should always be mentioned as well. The detection limit values given in Table 9.2 concern analysed areas of some 105 mm2, i.e., analysed volumes of a few tens of mm3. The quoted detection limit thus are of
422
Secondary ion mass spectrometry. Application to archaeology and art objects
Fig. 9.13. Relation between the detection sensitivity and the analysed volume per unit time.
the order of about 1 ppm for 1 mm3 analysed volume, that is 104 atoms or 10218 g in absolute terms. The detection limits of SIMS are strongly dependent on the element and on analysis conditions. Some elements are hardly detected as monoatomic ions (Zn, N) while others are extremely easy to detect (Al, alkaline, alkalineearth). When the presence of an element is disturbed by the pollution-species deriving from the analysis atmosphere (oxygen, hydrogen, carbon), also poor detection limit will be achieved for that specimen. Provided it is possible to introduce them into the specimen, it is useful to use natural isotopes with a small natural abundance (18O, 2H, 13C) as tracers. In such cases, the detection limit of the analyses can be greatly improved.
9.3.1.3 Irradiation damage, charge effect As stated previously, the energy dissipation of the primary ion beam occurs in a volume much larger than the volume actually analysed at a given instant. This leads to implantation of the primary ions, modification of the composition and of defect concentration, increase of the local atom diffusion rate. An estimation of the volume implied in this modification, as well as the number of created defects may be obtained thanks to Monte-Carlo models, as the TRIM code [32]. It may have important consequences on the validity of the final analysis (e.g., depth resolution, see further). For insulating materials, the primary ion bombardment, the secondary ion extraction, and over all the emission of irradiation-induced secondary electrons induces a modification of the surface potential which hinders the analysis liability.
423
424 TABLE 9.2 Detection limits, in routine condition, of some impurities (mass resolution for 10% of maximum signal) Analysed ion
Primary ions
Primary current (nA)
Rastered area (mm2)
Analysed area (mm2)
Mass resolution (M/DM)
Detection limit (at. ppm)
H D C O S B P in Si P in Si As in Si Ag, Mg, Mn Al, Cr Si in GaAs
1
Csþ Csþ Csþ Csþ Csþ Oþ 2 Csþ Csþ Csþ Oþ 2 Oþ 2 Csþ
200 –400 200 –400 200 –400 200 –400 400 1000 400 400 400 1000 1000 400
250 £ 250 250 £ 250 250 £ 250 250 £ 250 250 £ 250 250 £ 250 250 £ 250 250 £ 250 250 £ 250 250 £ 250 250 £ 250 250 £ 250
3600 3600 3600 3600 3600 3600 3600 3600 3600 3600 3600 3600
300 300 300 300 3000 300 300 4000 300 300 300 300
2–20 0.1 –1 1–20 1–50 2 £ 1022 1023 50 1 0.2 (energy offset) 5 £ 1024 a` 1023 1024 a` 5 £ 1024 2 £ 1022
H2 H2 12 2 C 16 2 O 32 2 S 11 þ B 31 2 P 31 2 P 75 As2 Mþ Mþ 28 2 Si 2
E. Darque-Ceretti and M. Aucouturier
Element
Secondary ion mass spectrometry. Application to archaeology and art objects
The effect is lower but still exists for static SIMS analysis. The use of a compensating electron gun minimizes it. 9.3.2
Quantitative analysis
The conversion of the secondary ionic emission signals to actual concentrations in the analysed material is one of the weakest points of the method. Selectivity and sensitivity aspects evidently have to be treated first as described above. 9.3.2.1 General expressions The phenomenological relation between the ion intensity (number of particles per unit time) for a given ionic species A^ can be expressed as the convolution of the collection, ionization and sputtering functions of the system: ^
^
IðA^ Þ ¼ F1A ðcollectionÞ ^ F2A ðionizationÞ ^ F3A ðsputteringÞ
ð9:9Þ
It is generally considered that the three factors are independent, so that the convolutions (denoted with the ^-signs) are in fact multiplications. The collection factor F1, often written h(A) is mainly a function of the experimental conditions, but may also vary with the ion species and the specimen texture (mono- or polycrystalline). It is hard to evaluate and is generally eliminated in quantitative analysis by employing ratios of signals. The sputtering factor F3 for a given primary current Ip (number of particles received per unit time over the analysed area) is: F3A ¼ ST Ip CA
ð9:10Þ
if it is assumed that the total sputtering yield ST is the sum of the sputtering yields of all individual species (A, B, C, etc.) weighted by their concentrations (CA, CB, CC, etc.). This is true for steady state sputtering, when the composition of the sputtered matter is identical to the composition of the material, but not for the transitory stage at the beginning of an analysis. Moreover, the sputtering yields are concentration-dependent. The ionization factor F2 is the ionization probability defined in section 9.2.1. It is strongly related to the nature and composition of the analysed specimen (“spec”). Thus, the ion intensity can be written as: IðA^ Þ ¼ hðA^ ÞPspec ðA^ ÞST Ip CA
ð9:11Þ
425
E. Darque-Ceretti and M. Aucouturier
which would correspond to a straight calibration line for the analysis, if the factors h, P and ST were concentration-independent. Unfortunately this is not the case for any of them. One may define a “relative sensitivity factor” between two species: kAB ¼
hðAÞ Pspec ðA^ Þ ; hðBÞ Pspec ðB^ Þ
with
IðA^ Þ C ¼ kAB A IðB^ Þ CB
ð9:12Þ
The use of such relative sensitivity factors, only valid for a given composition (because ST depends on the composition) is the only way to treat quantitatively the results of a measurement and compare them from one laboratory to another. Their evaluation requires that the calibration is performed in perfectly defined conditions (material with known structure and composition near the analysed material, identical equipment settings, i.e., transmission factors, etc.). Another approach is to consider the quantity rðA^ Þ ¼ ðPspec ðA^ ÞÞ=ððPA ðA^ ÞÞ; often called “exaltation factor”, which expresses the ionization yield ratio between the specimen and a pure species A. If the transmission factor is the same, one finds: Ispec ðA^ Þ Pe ch ðA^ ÞST;spec ST;spec ¼ CA ¼ rðA^ ÞCA ^ ^ PA ðA ÞSA SA IA ðA Þ
ð9:13Þ
It may be sometimes assumed that, for a given alloy category, the ionization probability is only function of the atom environment. For example, for a binary alloy: Pspec ðA^ Þ ¼ PA ðA^ ÞCA þ PB ðA^ ÞCB
ð9:14Þ
The exaltation factor is then a linear function of the concentration and the sensitivity factor can be written as: kAB ¼
PA ðAþ Þ CA þ ð1 2 CA Þ½PB ðAþ Þ=PA ðAþ Þ PB ðAþ Þ ð1 2 CA Þ þ CA ½PA ðAþ Þ=PB ðAþ Þ
ð9:15Þ
This model has been used for the quantification of positive ions emission when the surface is saturated with oxygen (all species are emitted from their oxidized compound). When its validity has been demonstrated, it is useful to interpolate the values of the sensitivity factors over a larger concentration range. 9.3.2.2 Iteration method For materials containing a large number of always the same elements, as is the case for mineralogical materials, an iterative method may be
426
Secondary ion mass spectrometry. Application to archaeology and art objects
proposed [33]. In a linear model, the ionization probability for n components is: n X Pspec ðiþ Þ ¼ Pj ðiþ ÞCj ð9:16Þ j¼1
One element (for instance Si in geological material, in ceramics or in glasses) is taken as reference to define a relative ionization coefficient: n X Pj ðiþ ÞCj n X j¼1 RSi ðiþ Þ ¼ n ¼ Xj ðiþ ÞCj ð9:17Þ X þ j¼1 Pj ðSi ÞCj j¼1
if one assumes, which is generally true, that the ionization yield of Si is not concentration-dependent, and with Xj the ionization probability of the j species normalized to that of Si in the same environment. The concentration of a i element is calculated from the normalized I(iþ) intensity: Ci ¼
Iðiþ Þ=RSi ðiþ Þ n X Iðiþ Þ=RSi ðiþ Þ
ð9:18Þ
i¼j
The Xj coefficients are measured from standard specimens of the same chemical family as the analysed material (for instance silicates containing Si, Al, Na, K, Ca, Fe, Mg, Ti). Then the R coefficients are computed for the analysed material from a supposed set of concentrations. The composition obtained from the above formulas is used to calculate a new set of R coefficients, and so on. Three or four iterations are generally necessary to reach a stable result. One should however recall that when the analysed volume (thickness) is large enough, a quantitative analysis using electron probe microanalyser gives, for the major elements, more precise results by using a smaller number of standards. 9.3.2.3 Diluted alloys (trace analysis) When the concentration of the analysed element is lower than 1 at.%, the problem is highly simplified, because the environment can be considered as constant. The ionization yield and the sputtering yield become independent from the concentrations. The sensitivity factor then becomes (M is the matrix): kiM ¼
hðiÞ PM ði^ Þ hðMÞ PM ðM ^ Þ
so that
Ci ¼ kiM
Iði^ Þ IðM ^ Þ
ð9:19Þ
427
E. Darque-Ceretti and M. Aucouturier
These factors can be obtained through only one calibration measurement for each impurity. Precise calibration standards may be manufactured by implantation of one or more impurities. Figure 9.14 summarizes the calibration procedure in the ideal case of the measurement of the concentration profile of a dopant (11B) in pure silicon. The experimental profile is shown in Fig. 9.14a. The standard is a silicon monocrystal implanted with 11B (150 keV energy, dose 1015at. cm22). Starting from the experimental profile in the implanted standard (Fig. 9.14b), knowing the thickness of the sputtered crater (by profilometry) and the implantation dose, the true concentration profile is obtained (Fig. 9.14c). The sensitivity factor kBSi ¼ CB ðIðSiþ Þ=IðBþ ÞÞ can be calculated for each point of the curve. This factor can be applied to the analysed specimen profile whose thickness scale is calibrated also by crater profilometry, by inversion of the same formula: CB ¼ kBSi ðIðBþ Þ=IðSiþ ÞÞ (Fig. 9.14d). It is necessary to perform the analysis of the
Fig. 9.14. Concentration calibration for a diluted impurity, here boron in silicon: (a) raw profile of the 11Bþ signal in the specimen as a function of time, (b) raw profile of the 11Bþ signal in the implanted standard (1015 at. cm22, 150 keV), (c) boron profile calibrated in depth and concentration, (d) calibrated profile in the specimen.
428
Secondary ion mass spectrometry. Application to archaeology and art objects
standard and of the unknown specimen exactly in the same conditions. Most of the modern computer systems perform this calibration automatically. 9.3.2.4 Use of post-ionization or MCsþ ions Post-ionization (SNMS) [34] or MCsþ ion detection [35] minimizes the influence of ionization exaltation, because only the emitted neutral particles contribute to the signal. The quantification is greatly facilitated, and it is only necessary to calibrate the sputtering yield and its possible variation with composition. SNMS analysis is particularly useful for thin film analysis, where an abrupt concentration variation or contamination at the interfaces may induce uncontrolled exaltation effects in common SIMS. 9.3.2.5 Isotope analysis The quantitative determination of isotope-ratios of a specific element (or the concentration ratio of the corresponding parent elements in a fission reaction) can be obtained by SIMS with a high accuracy, provided the spectrometer reliability is sufficiently improved (stability and reproducibility of the magnetic field). Applications are numerous in geology, astrophysics, datation, provenance determination. Several examples will be given in the application section of this chapter. 9.3.3
In-depth analysis and depth resolution
Because of the continuous erosion by sputtering during the analysis, dynamic SIMS is well suited for the determination of compositional variations as a function of the distance from the surface (in-depth profiling). All aspects concerning identification of the species, detection limits, quantitative expression of the concentration have been developed above and can be used also in the present case. The aim is to obtain a precise relation between concentration (or ion yield) and depth, characterized by the depth resolution. One should take care of the possible confusion between the following distinct, although not independent, notions: In some analysis methods (e.g., RBS, ERDA, electron beam analysis under variable electron energy), the signal treatment gives access to the depth from which the signal has been emitted. The depth resolution is here the accuracy of that calculated depth. The second notion is the analysed depth that is the thickness of the specimen which emits signals indistinguishable from each-others. For secondary ion analysis, this depth is of one to two atomic layers; it is the
429
E. Darque-Ceretti and M. Aucouturier
ideal lowest limit of the depth resolution, if one would be able to exploit instantaneously all the collected signals (this is theoretically the case in static SIMS). In dynamic SIMS, the signals are accumulated during some time, and the analysed depth attached to a given signal is much larger than these one or two atomic layers. The third notion, specific to the methods using ion sputtering as SIMS, accounts for the precision of the knowledge of the position of the actually analysed surface at a given moment, referred to the initial surface, as a function of the time elapsed since the beginning of the analysis (see the following: depth measurement). Finally, the concentration distribution in the analysed specimen may be perturbed by the analysis itself (implantation effects, etc.), leading to a falsification of the concentration profiles. The depth resolution for ion analysis is in fact a combination of the last three factors. It also is a function of various factors described below. 9.3.3.1 Crater edge effects, memory effects, re-deposition effects As explained above, primary beam rastering is applied in order to obtain a flat crater bottom; the analysed area has to be chosen much smaller than the rastered area in order to avoid distortion of the analysis by “crater edge effects”. When the concentration variations are large over small distances, a pollution of the analysis chamber (extraction lens) may occur at the beginning, leading to re-deposition of the elements in later phases of the analysis (“memory effect”). This memory effect is sometimes avoided by changing the front face of the extraction lens during the analysis. It may also happen that volatile elements that have escaped from regions neighbouring the analysed area are re-deposited inside the crater during the analysis. This phenomenon is particularly dangerous for organic compounds, easily decomposed by the ion beam. The faster the analysis (high primary current), the smaller is the re-deposition probability. 9.3.3.2 Depth measurement, influence of the initial roughness The conversion of the erosion duration to the eroded depth is presently only possible after the analysis. The most accurate means of doing this is mechanical profilometry by a small tip (1– 5 mm curvature) with a precision of some nanometers. But the accuracy of this crater depth measurement evidently cannot be better than the initial surface roughness, as the initial surface is the depth reference. This evident statement is often forgotten by experimentalists. In any profilometer, only a limited range of spatial
430
Secondary ion mass spectrometry. Application to archaeology and art objects
wavelengths is accounted for in the measurement, which means that the accuracy of the crater depth value may be overestimated. The conversion of erosion time to eroded depth supposes that the erosion rate, that is the sputtering yield, remains constant during the whole analysis. Knowing that the sputtering yield varies with the composition, it is very important to take this hypothesis into account in the calibration of the profile depth scale. Multiple calibration at different depth values is often necessary, for instance in the case of multilayered specimen analysis. 9.3.3.3 Primary beam induced roughness Ion bombardment induces a surface roughness that damages continuously and increasingly the surface during the analysis. The result is a progressive degradation of the depth resolution. This roughness originates from the random character of sputtering and from material-induced effects as the role of the crystallographic orientation on sputtering. It is clearly the most deleterious effect on depth resolution, and also the most difficult to overcome. Figure 9.15 [36] shows an example of induced roughness on a polycrystalline Cu – Zn alloy. The imprecision on depth may reach several mm. Whatever the origin of the beam-induced roughening, the quantitative expression of its consequence on the depth resolution may be expressed through a probabilistic analysis. The “SLS” (sequential layer sputtering) model [37] concludes that an ideally perfect concentration stepfunction of the depth leads to a measured profile with an error function (erfc) shape, whose standard deviation Dz84 16 (profile width measured as the distance between the points corresponding to 16% and 84% of the stepheight) is proportional to the square root of the reached depth z as: 1=2 Dz84 16 ¼ aðzÞ
ð9:20Þ
Experimental means to avoid this kind of defect are not universal. Reactive sputtering with an oxygen beam, together with surface leaching by an oxygen partial pressure, minimizes the crystallographic effects. Rotating the specimen holder (azimutal rotation) allows to avoid shadowing effects which tend to stabilize and worsen the induced roughening. 9.3.3.4 Direct primary beam implantation effects The primary beam implantation induces a perturbed region under the surface of about 10 – 20 nm depth. In that region, a large number of defects are created and accelerated element diffusion can take place. Implantation itself induces atomic displacements by cascade collisions (ion collision mixing). Moreover, the change in environment produced by the implanted
431
E. Darque-Ceretti and M. Aucouturier
Fig. 9.15. Cu-30% Zn brass. Centre of a sputtering crater (Arþ, 5.5 keV) [36]: (a) level curves by roughness measurement (0.25 mm between two curves), (b) secondary electron image of the same area.
primary ions modifies the ionization yield of the emitted particles. The influences of the primary ion nature and energy are decisive for these three effects. One should distinguish between two steps: – a transitory step, when the concentration of the incident species in the perturbed layer varies continuously with erosion; – a following steady state step, when competition between surface receding and implantation leads to a composition steady state in the perturbed layer.
432
Secondary ion mass spectrometry. Application to archaeology and art objects
Quantitative modelling shows that the steady state is reached after sputtering of a depth equal to about 2Rp, and the perturbed depth is of the order of 3sp, Rp and sp being respectively the mean range of the implanted particles and the straggling of their distribution. The Rp and sp values may be obtained through modelling codes (e.g., TRIM [32]). For example, for Csþ ions at 14.5 keV energy in silicon Rp is 15 nm and sp ¼ 4.3 nm. Collision mixing have been described by semi-theoretical models. These predict that for an ideal decreasing concentration step function (from c1 to c2 , c1), the signal decrease is exponential as a function of the depth. The slope of that decrease on a logarithmic scale (decay length) is inversely proportional to Rp, that is to the primary ion energy and the square root of their mass. Thus, it is possible to attempt a deconvolution of the experimental profile, thanks to preliminary measurements on model specimens under variable incident ion energy. 9.3.3.5 Induced segregation effect Primary beam implantation may also lead to long distance segregation of chemical elements. This effect is a result of the surface composition heterogeneity induced by the primary species, and is specific for each analysis case. 9.3.3.6 Preferential sputtering When the sputtering yields of the various elements present in the material are different, preferential sputtering of the elements with the highest sputtering yield induces, at the beginning of the analysis or for steep concentration variation at a given depth (multilayer), a local transitory modification of the concentrations (Fig. 9.16 [36]). For a given A – B alloy (50% each), the more rapid sputtering of the A element leads to an overconcentration of A in the sputtered matter and a depletion of the same element at the surface. It is only after a sufficient sputtering time that the composition of the sputtered matter becomes equivalent to that of the target. So, during a transitory stage, the analysis is not representative. One should note that this effect exists also for surface analysis methods using an ionetching beam (AES, X-ray induced photoelectron spectroscopy, XPS), but in this latter case, the effect and the error are permanent. Modelling of the preferential sputtering effect is not presently done and would need a precise knowledge of the individual element sputtering yields or simultaneous measurement of the surface composition and of the sputtered matter composition on the same area at the same moment.
433
E. Darque-Ceretti and M. Aucouturier
Fig. 9.16. Schematic concentration variation in the sputtered matter CpA and CpB and on the surface CsA and CsB, during the first analysis stage of an equiatomic alloy A –B, when the sputtering yield of the A species is larger than that of the B species [36].
9.3.3.7 Matrix effects, interest of SNMS The fact that ionization yields are strongly dependent on composition is a serious drawback of in-depth analysis, particularly for the analysis of thin layers where the concentration variations are very steep. Post-ionization analysis (SNMS or STIMS) is able to overcome that handicap. 9.3.3.8 Summary In-depth analysis must, in addition to the condition of identification and correct quantification already stated, fulfil the following conditions: – minimization of edge effects by choosing the analysed area much smaller than the rastered area; – minimization of the re-deposition and memory effects thanks to a high erosion rate; – good knowledge of the eroded depth and small initial roughness; – accounting for, and if possible deconvolution, of the induced roughening effect; – accounting for, measurement (by variable incident primary energy), and if possible deconvolution, of the primary beam implantation effects; – accounting for differential sputtering effect, and measure of the individual sputtering yields; – accounting for the matrix effects on the ionization yields.
434
Secondary ion mass spectrometry. Application to archaeology and art objects
9.3.4
Surface analysis
Because it is in principle a “destructive” method, SIMS can only be called a surface microanalysis method when the quantity of analysed matter during one single analysis is smaller or equal to the emission depth, i.e., about one atom layer. This is the definition of static SIMS. In time-of-flight SIMS (ToF-SIMS), the condition is fulfilled thanks to the pulsed source and the parallel detection of secondary ions, but this can also be the case with conventional equipment, provided the primary current is sufficiently small and the detection sufficiently rapid. Table 9.3 compares some average performances of different spectrometers. Among the already discussed performances, the problems of detectability, of species identification and of quantitative analysis exist in the same terms. Sensitivity (detection limit) is an important parameter because the analysed volume is here very small. This is why point detection (imaging) is generally only possible for elements with a high concentration and in the most sensitive equipment (time-of-flight). For quantification, the use of relative sensitivity coefficients is necessary, but the choice of a reference signal is somewhat delicate because one has to choose a species representative of the studied phenomenon (substrate for adsorbed layers, simple fragment for polymers, etc.). 9.3.4.1 Organic materials The field of organic materials has been the most widely explored by surface SIMS (SSIMS) [26], thanks to the ability to detect polyatomic (molecular) ions, which can be useful for the identification of molecular species. Figure 9.17 shows an example of metal surface contamination by an organic compound (here a glycol-based oil) [38]. Imaging of the sputtered polyatomic fragments allows location of the main contaminating species with a good spatial resolution. TABLE 9.3 Compared typical performances of some spectrometers in static setting [26] Resolution Masse range Transmission Detection M/DM mode Quadrupole Magnetic sector
102 2 103 104
,103 .103
0.01–0.1 0.05–0.5
Time of flight
104
103 2 104
0.5
Relative sensitivity
Sequential 1 Sequential 10 –104 or parallel Parallel 104
435
E. Darque-Ceretti and M. Aucouturier
Fig. 9.17. Ion images by ToF-SIMS of a polyethylene-terephtalate polymer coated by an aluminium film which exhibits covering defects due to traces of a glycol-containing oil (molecules labelled with an asterisk (p)). The images correspond to the ions: 27 Al, 28(AlH), 31(CH3O)p, 39(C3H3), 43(AlO), 45(C2H5O)p, 54(Al2), 55(C4H7), 59(C5H7O)p, 69 (C5H9), 73(C4H9O)p, 87(C4H7O2)p [38].
As already mentioned, primary beam-induced surface damaging is very limited in truely static conditions because the beam “sees” the surface only once during the analysis. So, SSIMS is a very effective method for the study of polymer surface evolution, for instance ageing or surface treatment (irradiation, oxidation, etc.). It is also adapted to the study of surface segregation. 9.3.4.2 Inorganic materials Application of static SIMS to surface analysis of inorganic materials is somewhat recent [19]. The main applications are found in investigations on adsorption phenomena (first stage of oxidation, catalysis) and surface pollution. Figure 9.18 [39] shows how, by using at the same time the
436
Secondary ion mass spectrometry. Application to archaeology and art objects
Fig. 9.18. Variation of the ion yields in static SIMS for various metals during the oxidation first stage, as a function of the oxygen dose (in langmuir) [39]: (a) positive secondary ions, (b) negative secondary ions. 437
E. Darque-Ceretti and M. Aucouturier
exaltation effect on positive ion emission and the emission of molecular ions (here negative ions), one is able to characterize the first stages of metal oxidation. 9.3.5
Imaging, lateral resolution
In order to obtain a spatial distribution of elements (or phases, see following section), the analysis has to be confined to a small volume. This limitation of the analysed volume leads to a loss of analysis sensitivity and precision. Moreover, the measured concentration at one “point” must be referred to the analysed volume: a very high local segregation may not be detected if the individual analysed volume which includes it is large compared to the segregation extension. For instance a 100% segregation on a single atomic layer of a grain boundary (equilibrium segregation) is measured as an average atomic fraction of 1024 if the analysed volume around the boundary is 2.1022 mm3 (4 mm2 £ 5 nm), a common figure for a well-focused ion probe. Lateral distributions can be obtained either in microscope mode or in microprobe mode. The lateral resolution is defined as the shortest observable separation distance between two points with different concentration. In microscope mode, for quantitative measurements on the entire analysed area, the minimum size of the analysed area is limited by the optical properties of the beam, of secondary optics and of the apertures. It is of the order of 2 mm. When a CCD camera is used to quantify the image, the local measurement is only limited by the optical aberrations and becomes smaller than 1 mm. However, additional contrast effects may appear, due to local variations of the emission probability. For instance, in Fig. 9.19 [6] an atomic grain boundary segregation of carbon in a nickel alloy is observed only on the 24C2 2 image and not on the 12C2 image. In this case, the contrast is amplified for the biatomic ions because the simultaneous emission probability of two carbon atoms ions is much larger from the grain boundary carbon monolayer than far from the boundary, even if the average carbon concentration referred to the analysed volume remains small. In microprobe mode the ultimate lateral resolution is limited by the size of the primary probe and by the individual analysis or image pixel. The best lateral resolution (a few tens of nanometers) is obtained with “nano-SIMS” instruments. In microprobe mode, or in CCD recorded microscope mode, the signals may be saved as electronic files as a function of the sputtering time.
438
Secondary ion mass spectrometry. Application to archaeology and art objects
Fig. 9.19. Ion images 63Ni2 (a), 12C2 (b) and 24C2 2 (c), obtained on a Ni-based alloy with an intergranular segregation of carbon. Contrast effect in ion imaging [6].
Because of the relation between the sputtering time and the depth, it becomes possible to reconstruct the 3D distribution of one given species in the specimen.
9.3.6
Chemical compound analysis and distribution
The ionization yields of the species are dependent on the chemical nature of the phase which contains them. This is a serious drawback for quantitative analysis, but it may be used as a tool to identify and measure the phase content of a material. The qualitative (imaging) or quantitative exploitation of the matrix effects or polyatomic ion emission is an important application of secondary ion analysis. The procedure should be adapted to the specific problem, but a few rules have to be fulfilled. The choice of polyatomic ions must take into account their emission mechanism: for instance, in order to analyse silicon carbide inside a silicon matrix, it is better advised ^ ions, because the formation of SiC^ to choose SiC^ 2 ions than SiC 2 ions by recombination of three emitted atoms is totally unlikely from the carbon-poor matrix, whereas formation of SiC^ ions is possible. In a solid solution A – B with respective concentrations CA and CB, the emission ^ ^ are respectively proportional to C2A, C2B probabilities for A^ 2 , B2 and AB and 2CACB, whereas for B aggregates in the same alloy, the emission probability of B^ 2 ions is proportional to CB. As a general rule, if polyatomic ions are identified as characteristic of given phases, their quantitative evaluation allows, by application of the phase diagram rules, a determination of the volume fraction of the respective phases.
439
E. Darque-Ceretti and M. Aucouturier
9.4
EXAMPLES OF APPLICATIONS FOR CULTURAL HERITAGE
The following examples are chosen from the recent literature in order to illustrate the specific performances of the SIMS method in cultural heritage applications. The aim is not to present an exhaustive enumeration of various SIMS investigations, described in recently published review articles [40,41]. The examples are here attempting to emphasize the cases were the use of that somewhat elaborated method cannot be avoided to solve specific characterization problems, sometimes in combination with other more common microanalysis methods. The limitations of SIMS will also appear in some examples. 9.4.1
Dating and/or provenance studies based on isotopic analysis
Being sensitive to the atomic mass of the analysed elements, SIMS is suited for performing isotopic analyses. The determination of isotopic ratios of the same element is unfortunately not precise enough for accurate dating, except in the astro-physics field. An interesting application has nevertheless been experienced for the dating of an archaeological natural nuclear reactor (OKLO), by quantitative measurement of the isotopic ratio between the nuclear fuel (235U and 238U) and its fission products (206Pb and 207Pb) (Fig. 9.20 [42]). The date of occurrence of the nuclear reactions in the analysed area has been established to (1968 ^ 50) million years before present. The determination of atomic isotopic ratios may be used to estimate the provenance of cultural heritage items. One should be aware of the fact that the measurement of isotopic ratios by SIMS may be affected by various effects: – first, in all commonly used magnetic sequential spectrometers, the reproducibility of the magnetic field is not precisely reliable. The use of parallel detection is necessary to perform an accurate measurement of the isotopic ratios; – secondly, the ion yield of an element depends on its chemical form, and the distribution of the isotopic ratio may also be affected by the chemical form. If one compares isotopic ratios of an element for items in which this element does not have the same chemical form, an error may be committed by not taking into account the chemical isotopic effect. The following examples from recent literature illustrate the problems and success of the use of isotopic ratio measurements for provenance studies.
440
Secondary ion mass spectrometry. Application to archaeology and art objects
Fig. 9.20. Dating of a region of the OKLO nuclear reactor by point analysis (20 mm 235 206 probe) of the 238UOþ UOþ Pbþ, 207Pbþ ions. The intersection of the lines 2, 2, corresponding to the corrected measurements (black experimental points) with the diagram calculated by the “concordia” simulation gives the age of the analysed area (here 1968 million years) [42].
9.4.1.1 Lead in metallic artefacts Lead is a current component of the bronze alloys and this element has four naturally occurring isotopes, namely 204Pb (1.5%), 206Pb (23.6%), 207Pb (22.6%) and 208Pb (52.3%). The variation of the isotopic abundance of these isotopes from one geological origin to another is large enough to be used to indicate the origin of a lead-containing bronze by measurement of the isotopic ratios. A further favourable point is that lead is always precipitated as metallic particles in solid bronze (its solid solubility is nearly zero). Then the comparison between different objects is less affected by the above-mentioned chemical effect, provided care is taken to control (for instance by secondary ion imaging) that the analysis is actually done on an homogeneous field. The comparison with oxidized lead-containing compounds (e.g., mineral ores or corrosion products) has to de done with care.
441
E. Darque-Ceretti and M. Aucouturier
Lead isotopic analysis for provenance study purposes is usually done by conventional mass spectrometry on dissolved samples [43,44]. In a recent study [45] devoted to provenance investigation on Chinese bronze artefacts, a comparison was made between two procedures: conventional (thermally ionized) mass spectrometry on dissolved samples and (less destructive) SIMS analysis. It could be shown that although less precise, the SIMS procedure may be considered as satisfactory in a number of cases. The same study proved indeed that it is important to perform the SIMS analyses on a homogeneous area. No comparison was reported for the case of oxidized species. 9.4.1.2 Investigations based on 18O analysis Oxygen is the most abundant element of the earth crust. It has three natural isotopes, namely 16O (99.756%), 17O (0.039%) and 18O (0.205%). The abundance rate of a minor given isotope (here 18O) referred to the principal isotope is generally evaluated as the “d” ratio (expressed in ‰): d18 O ¼ ðð18 O=16 OÞspecimen =ð18 O=16 OÞstandard Þ 2 1 £ 1000: ð9:21Þ For oxygen, the international standard corresponds to the mean sea water (standard mean ocean water, SMOW), and the d ratio of most of the geological minerals is positive (they contain more 18O than SMOW) and amounts 5 – 30‰. That isotopic fractionation is a consequence of isotopic effects in the chemical or phase-change reactions. It is usually measured by conventional mass spectrometry of gaseous CO2 obtained by decomposition and carbonization of a sampled part of the object. The possible use of SIMS for the determination of isotopic fractionation in solid specimens has been extensively discussed in literature, because this method introduces nonnegligible instrumental mass fractionation by itself. The instrumental fractionation (IMF) is expressed as: IMF ¼ ðð18 O=16 OÞmeasured 2 ð18 O=16 OÞactual Þ=ð18 O=16 OÞactual £ 1000 ð9:22Þ The origin of such fractionation may be found either in true instrumental inaccuracy (magnetic field shift in sequential measurement) or in chemical effects due to the variety of chemical bounds which exist in oxygenated compounds. For instance, it has been demonstrated [46] that the “instrumental fractionation”, measured for a very large variety of minerals with a sequential CAMECA IMS 4F conventional analyser, reaches 235 to 270‰ and is correlated to the mean atomic mass of the analysed minerals. This deviation
442
Secondary ion mass spectrometry. Application to archaeology and art objects
could be a consequence of matrix (chemical) effects. This means that the d values may be instrumentally strongly falsified. Correction procedures are proposed. More recent equipment (e.g., CAMECA IMS 1270) provide more reliable isotopic ratio measurements, thanks to parallel secondary ion multiple collection, and calibration using standards of the same chemical nature as the investigated items is necessary to avoid the chemical effects. Application of SIMS analysis to attempt, through 18O isotopic abundance measurement, a localization of the origin of oxygen-containing gemstones is clearly a great advantage as it avoids sampling (usually a few 1023 g) from rare specimens and replaces it by ion sputtering of a very small amount of matter (2.10211 g) from a micron-scale crater (Fig. 9.21) [47]. The example quoted here concerns an investigation of the trade routes of emerald since antiquity. Calibration is done with emerald specimens. The authors showed how the ancient Silk Route and Austrian stones supply was progressively replaced by the importation from South-America (Colombia) after the 16th century (Fig. 9.22) [48].
Fig. 9.21. Example of craters (c) sputtered from the surface of an emerald during SIMS analysis. The size of the craters may be compared to the size of fluid inclusion (if ) trapped in the crystal [47].
443
E. Darque-Ceretti and M. Aucouturier
Fig. 9.22. Values of d18O for nine analysed emeralds (white boxes). The mining areas are indicated (black or grey fields), and samples are ordered chronologically. (1) galloroman earrings; (2) Holy Crown of France; (3) Hauy’s emeralds; (4) Spanish galleon wreck; (5) old mine emeralds [48].
9.4.2
Dating (not based on isotopic analysis)
It is known that most of the vitreous silica-based materials suffer a surface degradation in aqueous media or in wet atmosphere, and this is an important conservation problem (section 9.4.4). That degradation is accompanied by a hydration of the surface layers and the depth of water diffusion has been proposed to be a way to measure the age of the stone since its last polishing. Obsidian is from that viewpoint an ideal material because it is a glassy natural stone used for artefacts since archaeological periods. SIMS is
444
Secondary ion mass spectrometry. Application to archaeology and art objects
used to evaluate the penetration (measurements of OH2 ion profile). The interpretation and modelling of the measurements is however difficult [49], because the lixiviation process of the surface depends on various parameters: the diffusion mechanism of water is not clearly established and depends on the water concentration; the position of the original surface is moreover not always ascertained. It is clear that the influence of the environment has to be taken into account. Leaching of the surface may have happened, leading to a non-negligible disappearance of a non-controlled amount of the surface layers. Attempts to use the diffusion profile of other elements (e.g., nitrogen [50]) seem also to be hindered by uncontrolled artefacts, as the existence of surface micro-cracks. As stated previously it is evident that absolute penetration depth measurements by SIMS profiling over small distances (less than 1 mm here for nitrogen) needs a perfect knowledge of the surface position and integrity. Another recent attempt to use diffusion of environmental element into minerals to monitor the age of ancient objects is based on the combination of SIMS and nuclear analysis to quantify the diffusion of hydrogen into quartz or other mineral crystals [51]. The investigation is still in progress, and it is clear that a very careful modelling will be necessary to interpret the results. Combined SIMS and nuclear analyses (resonant nuclear reaction and/or ERDA) are the unique mean to obtain accurate hydrogen profiles in materials. Another interesting application is the study of the diffusion profile of specific elements, for instance fluorine, in biologic materials (teeth) [52]. The distribution of elements between the surface enamel and the underlying layers of a tooth is known to be a function of the age of the owner and of his alimentation habits. The fluorine concentration in the inner enamel varies between 0 and 200 ppm during the life of an average adult, but the surface concentration saturates around 2000 ppm. SIMS analyses done on chosen regions of teeth found in a site of Cyprus (Fig. 9.23) [52] have shown that the inner enamel fluorine concentration is related to the time of burial. It is assumed that the surface fluorine concentration remains constant at 2000 ppm and fluorine diffuses from this reservoir into the enamel bulk during the time of burial. The bulk fluorine content is then a measure of the time of burial. The use of SIMS is justified by its exceptional sensitivity for fluorine local analysis and by the fact that localized analysis is easily performed. In that same field of biological materials, the SIMS ability to detect with a high sensitivity local variations of trace elements has been also used for the study of archaeological teeth staining. The specimens were excavated from a
445
E. Darque-Ceretti and M. Aucouturier
Fig. 9.23. Cross section of a tooth and SIMS line scans across the cross-section. The dashed curves are from a tooth dating from 1800 BC ; the others are typical from a recently extracted tooth [52].
medieval churchyard in Trondheim, Norway. Localized black staining is observed on the teeth. SIMS analyses by step-scanning were performed, with the purpose to identify if such staining had a relation with chewing habits of ancient men, or if it was related to a contamination by the burying medium [53]. The presence of manganese, calcium and strontium traces in the observed black stains is clearly attributed to a contamination of the residues by environment. EDS analysis had a too low sensitivity to detect the observed variations, but atomic absorption spectrometry on sampled pieces from the stained and unstained regions of the teeth confirmed quantitatively the observed variations. Unfortunately, the analysis of the burial soil could not be performed to confirm the conclusions. Such information is useful for historical and provenance studies.
446
Secondary ion mass spectrometry. Application to archaeology and art objects
9.4.3
Provenance studies not based on isotopic analysis
SIMS may be used in provenance studies by the exploitation of performances other than the isotopic analysis. SIMS is able to measure on very small amounts of matter, directly on the object, the relative concentration of trace elements. Mass spectra are also very useful for fingerprint identification of chemical compounds. An example is the determination of the Se/Te ratio in inclusions of Late Bronze Age copper ingots [40]. The small size of the inclusions (a few mm), the small amount of the impurities (less than 0.1%), and the possible application to preserved items justify the use of SIMS. When smelting sulphurcontaining copper ore, Se and Te tend to concentrate in the sulphide inclusions. Determination of the Se/Te ratio in inclusions may bring information on the origin of the initial ore. In the quoted investigation, the Se/Te ratio was constant throughout each ingot but the variations were very large from one ingot to the other. The authors concluded to the evidence of different sources for copper supplying of the studied archaeological site (situated in Essex, UK). The fingerprint technique using mass spectra may be a powerful tool for origin investigation. For example, analyses on Cypriot potteries [54] revealed that the mass spectra obtained on the clay used for their fabrication may be sorted in five different groups, revealing five possible origins of the raw material. Such classification is evidently of great use for historical studies. In this case, it is the quasi-non-destructive ability of SIMS to obtain a rapid overall mass spectrum description of the object which is exploited. The use of 3D impurity imaging, now possible in most SIMS instruments, opens improved possibilities for material characterization. Figure 9.24 [55] is an example of the distribution, obtained by recording the ion images on a CCD camera, of minor elements (Fe, Mn, Ni, Sb) in the bulk of a brass (Cu – Zn-based) alloy with an archaeological composition. Here, the specimens are simulated alloys elaborated by powder metallurgy, with compositions supposed to be representative of archaeological alloys. The knowledge of the distribution of impurities throughout the microstructure and amongst the different inclusions is a precious tool for understanding the ancient elaboration techniques and ore provenance. SIMS is the only method able to bring in a single experiment such 3D distributions of low-level impurities.
447
E. Darque-Ceretti and M. Aucouturier
Fig. 9.24. SIMS 3D distribution of the minor elements Fe, Mn, Sb, Ni in a Cu –Zn alloy with archaeological composition. Fe and Sb show the same distribution as Sb is probably originated from contamination of the Fe particles [55].
9.4.4
Surface layer analysis on artefacts
9.4.4.1 Degradation of ancient glass Because it concerns the outermost surface layers of a very flat material, the degradation of glass by environmental action is a subject perfectly adapted to SIMS analysis. A large number of investigation have been conducted on that subject. An example is shown in Fig. 9.25 [56]. In this ancient medieval glass item rich in potash, the lixiviation process is well evidenced, and the degraded layer thickness may be measured with accuracy. The depletion of the glass elements (K, Ca, but also Ba and Pb), as well as surface enrichment, specially for lead, are visible. This is known as a common weathering process of glass: the network modifiers (alkaline elements and calcium) are known to
448
Secondary ion mass spectrometry. Application to archaeology and art objects
Fig. 9.25. Depth distribution of various elements in a medieval glass. Layer 1 is a gold film coated to minimize charging; layer 2 is the outermost film with corrosion products; layer 3 is the alliterated glass layer; layer 4 is the unaffected bulk material [56].
diffuse out from the glass. Lead, which may be considered also as a network modifier in certain circumstances, diffuses also to the neighbouring medium, but has a tendency to redeposit from that medium. Another example may be found in a study of Roman glass degradation completed by a laboratory simulation of the process by an accelerated test in a HCl solution [57]. This accelerated test on modern glass in HCl solution (0.1% HCl, 708C) proved to lead to a distribution of the interesting elements or composite ions (Na, K, Si, OH) very similar to that observed on ancient items. This is an interesting information on the role of the environment acidity on glass alliteration. Simulation experiments with model glasses buried in actual soil may be also mentioned [58]. The mechanism of weathering of Middle Age stained church windows appears to be the same, although concerning a much larger alliteration depth (Fig. 9.26 [59]). This difference, from about 1.5 mm for the example of Fig. 9.25 to 60 mm here, may be attributed to different glass compositions and different environments (acidity). But the possible occurrence of a leaching of
449
E. Darque-Ceretti and M. Aucouturier
Fig. 9.26. Depth distribution of elements in a part of a glass window from St Gatien de Tours church [59].
the surface layer cannot be also excluded in the first case. In that case, the observed surface would not be the original glass surface, as the hydrated silica layer could have been leached out at an unknown period. Moreover, micrographic observations [59] showed that the surface alliterated layer exhibits often microcracks due to the corrosion-induced stresses. These microcracks offer a preferential path to the aggressive medium and to the propagation of the alliteration. The role of potassium and calcium, which were the common network modifiers used at those periods, is shown to be very important in the lixiviation process. They are the most depleted elements. A penetration of water in the lixiviated layer is always observed. Potassium and calcium seems to be more sensitive than sodium to lixiviation and the low amount of SiO2 of ancient glass, as compared to modern glass is clearly one of the reason of their poor resistance to water-containing aggressive agent. The lixiviated
450
Secondary ion mass spectrometry. Application to archaeology and art objects
potassium and calcium are converted at the surface into carbonates and sulphates which may precipitate and lead to the formation of opacifying stains. One may mention here that the combination of SIMS analysis with nondestructive high energy ion-beam analyses (as RBS, see Chapter 5) appears very rich in that case, because the latter brings quantitative information and is less sensitive to surface exaltation phenomena. SIMS, on its side has better depth resolution and trace detection limit. 9.4.4.2 Patina on bronze artefacts. Study by combined methods As stated in the first two section of this chapter, the high sensitivity, high depth resolution, good lateral resolution, and isotopic resolution of SIMS are counterbalanced by several drawbacks as a poor quantitativity and strong matrix effects in analysis. The combination of various microanalysis methods for the resolution of given problems appears thus as necessary process. An example of such combination will be given here for the characterization of patina on copper-based objects, and in the next section for the study of glass –metal interfaces. The case of “black bronze” or “black copper” has been extensively studied in literature [60]. This problem concerns the elaboration by chemical process of a black patina on the surface of copper– gold alloys (2 – 8% Au) used in art objects of very various civilizations (ancient Egypt, Greece, Rome, Japan, etc.). That black surface coating is known to be constituted of cuprite Cu2O, but the exact role of the gold alloying addition is not clearly elucidated. In a recent study [61], using nuclear analyses (particle-induced X-ray emission, PIXE; Rutherford backscattering spectrometry, RBS; and nuclear reaction analysis, NRA), and conventional SIMS, the exact distribution of gold between the patina outermost layer and the bulk substrate could be evidenced quantitatively. Authentic ancient Japanese “Shakudo” copper alloys (an ancient blackpatinated material containing ca. 4% Au) and simulated modern Cu –Au alloys patinated by chemical or thermal processes were studied. PIXE was used for an overall analysis of the surface region including the patina and the first substrate layer. RBS under protons or alpha particles and NRA were used to measure the thickness of the surface layers and obtain the element distribution over the first micrometers from the surface (Fig. 9.27) [61]. SIMS completed the study and showed the exact shape of the gold profile with a high accuracy (Fig. 9.28). As modern simulation specimens were used, it was also possible to validate the study by cross-section microanalysis using analytical scanning electron
451
E. Darque-Ceretti and M. Aucouturier
Fig. 9.27. Observed and simulated spectra obtained by RBS under protons (a), NRA under deuterons (b), and RBS under a particles (c), for a chemical patina on a Cu –Au alloy [61].
microscopy. It was shown that in all patinas elaborated by chemical process, including the authentic “Shakudo”, the surface patina, specially the Cu2O layer, is enriched in gold as compared to the substrate. On the contrary, black coloration obtained by thermal oxidation is constituted of
452
Secondary ion mass spectrometry. Application to archaeology and art objects
Fig. 9.28. SIMS profiles (Csþ primary beam, negative ion detection) on the same chemical patina as Fig. 9.27 [61].
two layers Cu2O and CuO without any gold content, and the outermost substrate region is enriched in gold over a few mm by inward diffusion. Thermo-chemistry specifies indeed the formation of that double Cu2O/ CuO layer in high temperature oxidation of copper and the inward diffusion of non-oxidized gold. But, the exact role of gold on the chemical black patination remains difficult to understand: the natural colour of Cu2O is red, and the reason why it becomes black on Cu –Au alloys is still not elucidated. In this investigation, the specific interest of using SIMS analysis and its complementarity with other microanalysis method are clear: RBS, NRA and EDS microanalyzes bring quantitative information, but are unable to reach the depth resolution and sensitivity of SIMS. On the other hand, SIMS is not quantitative enough to measure the exact amount of gold enrichment in the surface layer. 9.4.5
Interface studies on coated layers
Because of the continuous sputtering on a small area which occurs during SIMS analysis, this method is able to reach the embedded interfaces under the free surface and leads to a chemical characterization of the interface region. Here also, SIMS can be combined with other analysis methods, as nuclear analyses and/or cross-section microanalysis in the scanning electron microscope. Such studies are very scarce and should be developed in the future.
453
E. Darque-Ceretti and M. Aucouturier
An example is a recent study of the gold/glass interface of gilded glass objects [62]. The combination of SIMS, RBS and GDOES (glow discharge optical emission spectrometry) was able to clarify and quantify the mechanism of formation of the interface between a sodo-calcique glass and a gold film deposited by the “liquid gold” process (deposition of an organo-metallic compound followed by high-temperature annealing). A diffusion of gold into the substrate was evidenced and the diffusion coefficient was measured (Fig. 9.29). The influence of an intermediate
Fig. 9.29. Gold diffusion profiles into a glass substrate without (a) or with (b) an intermediate TiO2 or SnO2 film. The penetration depth and diffusion coefficient depends on the presence of the intermediate film [62].
454
Secondary ion mass spectrometry. Application to archaeology and art objects
TiO2 or SnO2 film pre-coated on the substrate is shown: this film hinders gold penetration into the substrate. Because of its good depth resolution, SIMS was the only method able to obtain a value of the gold diffusion coefficient in glass in the different cases. But, only RBS could quantify the gold content at the interface. SIMS has moreover shown an important penetration of some glass components (Na, K, Ca) into the surface gold film, without any hindering action of the intermediate film. Thanks to RBS and GDOES analysis, the amount of this penetration could be estimated. 9.4.6
ToF-SIMS applications
The relatively recent development of static ToF-SIMS explains the small number of investigations done with this kind of equipment on cultural heritage artefacts. The fact that static SIMS is a true surface, nearly nondestructive, analysis method should, however, lead many researchers on cultural heritage materials to a wide use of this method. The examples given here are therefore at the boundary of the cultural heritage material science. 9.4.6.1 Smalt discoloration in 16th century paintings Smalt is a ground cobalt glass used as a blue pigment in some oil paintings since the middle of the 16th century. In order to explain the reason of local degradation of the blue colour on some paintings, an investigation was conducted by local microanalysis of embedded samples from these paintings [63]. Several methods were again used in this study. The main result is that the degradation of the smalt glass is not due to a loss of the cobalt salt responsible for the blue colour, but is primarily attributed to a degradation of the glass itself in the presence of hydrolysed painting oils. This happens because the glass is rich in potassium, an element already proved to favour glass lixiviation (section 9.4.4.1). The weathering of the oil leads to the formation of fatty acids (FA) which themselves accelerate the glass dissolution; Figure 9.30 [63] shows distribution maps of elements or compounds around a smalt particle. The correlation between the location of the lead element issued from the glass degradation and of the fatty acid is an evidence of the role of the latter on glass alliteration. Cobalt itself remains unaffected. Static SIMS was in that case the only method to identify and show the space distribution of an organic compound as palmitic acid. Its combination with other microanalysis methods brought a solution to the problem.
455
E. Darque-Ceretti and M. Aucouturier
Fig. 9.30. ToF-SIMS images of a smalt particle in a cross-section sampled from a 16th century painting. The distribution of the fatty acid (FA, m/z ¼ 255, parent ion of palmitic acid) and lead (m/z ¼ 208) is an evidence of the glass alteration by the acid [63].
9.4.6.2 Study of keratin fibre bleaching process In that study [64], it is again the ability of static ToF-SIMS to identify organic compounds adsorbed on extreme surface which is taken into advantage. The bleaching process of cashmere wool is investigated. This process is applied to induce a whitening of melanin pigmented wool. The analytical study examines the chemical degradation of the wool fibre surface, the adsorption of textile auxiliaries (surfactants) and the characterization of a complex conditioner mixture on the modified fibre surface. The investigation is conducted by identification of the different organic compounds on the mass spectra during the different steps of the process. It is shown that, although the traditional bleaching process is effective in destroying the melanin pigment, it also severely degrades the cashmere fibre by oxidation of the major lipid component 18-methyleicsanoic acid (18-MEA). A more selective treatment process is proposed, proved to be less aggressive for the 18-MEA component. The effect of bleaching is moreover to increase the inter-yarn friction, leading to a harsher handle. Adsorption of cationic alkyl proteins onto the surface during the softening treatment by a specific agent is demonstrated by the ToF-SIMS analysis. The selective adsorption mechanism is detailed.
456
Secondary ion mass spectrometry. Application to archaeology and art objects
The study is completed by atomic force microscope observations showing the evolution of the fibre surface micro-roughness. The quoted investigation was conducted on modern wool fibres, but one can easily imagine that its application to ancient preserved wool could be of interest to evidence old treatment process.
9.4.6.3 Adsorption mechanism of a corrosion inhibitor on metals Corrosion inhibitors are widely used in conservation of metallic cultural heritage items. The most prevalent inhibitor for copper-based alloys is benzotriazole (BTA) C6H5N3, and this compound seems to be specific for these alloys. In a recent investigation combining ToF-SIMS and XPS [65,66], the adsorption mechanism of BTA on Si, Fe, Ag, Cu and brass substrates have been studied extensively. Results obtained with XPS confirm exactly those obtained in literature on copper and alloys. BTA reactivity towards copper is a particular case: BTA molecules take the place of all the hydrocarbon molecules initially present at the copper surface, all the copper atoms detected have an oxidation degree of þ1 and Cu/N/C atomic ratios are 1/3/6 as for an organo-cuprous polymer having a 2D structure covering the copper surface. Results obtained with brass are very similar and the first monolayers of the brass surface are constituted with cuprous oxides and do not contain zinc atoms. Tof-SIMS analysis show particularities of copper and alloys versus other surfaces. Only copper and brass exhibit intense quasi-polymeric ions. Positive and negative spectra on copper surface show not only copper ions and characteristic ions of BTA molecule but also ions at upper mass (150 – 1000), identified as series CunBTAm and CunBTAmCN. These polymeric ions are observed only on copper and alloys and never on silicon and iron surfaces. Another possibility of Tof-SIMS is to follow the different steps of cleaning procedures of metal surfaces which can affect the reactivity of subsequent surfaces; as water and organic contaminants are eliminated, intensities of copper and other non-organic ions (Na, K, Cl, O) are increasing. Molecular ion imaging has been used to map the distribution of the inhibitor film on the copper surface. Tof-SIMS images (Fig. 9.31) [65] obtained under static conditions on the BTA-treated copper surface show clearly composition heterogeneity in the uppermost layer: the regions rich in Cl2 and Naþ ions hinder BTA adsorption on the surface.
457
E. Darque-Ceretti and M. Aucouturier
Fig. 9.31. Spatial distributions of BTA2, Cuþ, Naþ and Cl2 ions on a copper surface treated with a BTA solution in methanol for 10 min [65].
9.5
CONCLUSION
The specific performances of SIMS microanalysis make this instrument necessary in the following cases: – quantitative analysis of trace impurities (less than 1 at.% and down to 0.1 ppb) in solids, – analysis of concentration depth profiles from the surface for distances much smaller than 1 mm, – analysis of thin films (from 1 nm to a few mm thick) and corresponding interfaces, – molecular analysis of organic materials, specially polymers, – true surface analysis (first atomic or molecular layer) in static setting for organic and inorganic materials, – spatial distribution, with a resolution better than 1 mm, of elements, particularly trace elements, or of molecular compounds in complex materials. One should however be aware of the fact that bulk quantitative analysis of elements in noticeable concentration is not in the field of SIMS. Accuracy and precision for such measurement are much better with more common microanalysis methods, some of them being described in other chapters of this book (Chapters 3 –7).
458
Secondary ion mass spectrometry. Application to archaeology and art objects
Applications of SIMS analysis for investigation on cultural heritage items are not numerous. This is partly due to the fact that SIMS is not usually considered to be a non-destructive method. The very small size of the sputtered craters (often less than a few micrometers) in dynamic settings and even the absence of crater in static settings allow in fact quasi-nondestructive investigations on preserved objects, provided they are sufficiently flat and of sufficiently small size to be introduced into the vacuum chamber.
REFERENCES 1 2 3
4 5 6 7 8 9 10
11 12 13 14 15 16 17 18 19
20
G. Slodzian, Thesis, Doctorate of Science, Orsay, 1963. R. Castaing and G. Slodzian, J. Microsc., 1 (1962) 395. M. Aucouturier, G. Blaise and E. Darque-Ceretti, In: J. Ammou (Ed.), Microcaracte´risation des solides. Centre Re´gional d’Analyse des Mate´riaux, CNRS, Sophia-Antipolis, France, 1989, 295 pp. G. Blaise and G. Slodzian, J. Phys., 31 (1970) 93. R. Castaing and G. Slodzian, J. Phys. E. Sci. Instrum., 14 (1981) 1119. E. Darque-Ceretti, J.-N. Migeon, M. Aucouturier, Emission ionique secondaire SIMS, In: Les Techniques de l’Inge´nieur, P 2 618, P 2 619, Paris, 1998. P. Sigmund, In: R. Behrish (Ed.), Sputtering by Particle Bombardment I, Topics in Applied Physics, Vol. 47. Springer, Berlin, 1981, 9 pp. H.H. Andersen and H.I. Bay, In: R. Behrish (Ed.), Sputtering by Particle Bombardment I, Topics in Applied Physics, Vol. 47. Springer, Berlin, 1981, 145 pp. D. Rosenberg and G.K. Wehner, J. Appl. Phys., 33(5) (1962) 1842. G. Blaise, Ele´ments de choix d’une me´thode de microanalyse. Analyse des surfaces et des couches minces, In: Les Techniques de l’Inge´nieur, P 3 795, Paris, 1990. E. Darque-Ceretti, F. Delamare and G. Blaise, Surface Interface Anal., 13 (1988) 14. G. Blaise, O. Lyon and C. Roques-Carmes, Surface Sci., 71 (1978) 630. D. Briggs, Surface Interface Anal., 14 (1989) 209. N. Winograd, D.E. Harrison and B.J. Garrison, Surface Sci., 78 (1978) 467. G. Blaise and G. Slodzian, Rev. Phys. Appl., 8 (1973) 105. N.M. Reed and J.C. Vickermann, In: L. Sabbatini and P.G. Tambonin (Eds.), Surface Characterization of Advanced Polymers. VCH, Weinheim, 1993, 83 pp. H.N. Migeon, In: J. Ruste and J.F. Bresse (Eds.), Nouvelles Techniques de Micro et Nanoanalyse, E1. ANRT, Paris, 1995. D. Briggs, In: D. Briggs and M.P. Seah (Eds.), Practical Surface Analysis, Vol. 2, Ion and Neutral Spectroscopy. Wiley, New York, 1992, 367 pp. N.M. Reed and J.C. Vickermann, In: D. Briggs and M.P. Seah (Eds.), Practical Surface Analysis, Vol. 2, Ion and Neutral Spectroscopy. Wiley, New York, 1992, 303 pp. Documents kindly provided by the CAMECA company, Courbevoie, 1997.
459
E. Darque-Ceretti and M. Aucouturier 21 22 23 24 25
26
27 28 29 30 31 32 33 34
35 36 37 38 39 40 41 42 43 44 45 46
460
Documents provided by Physical Electronics company, Cachan, 1996. G. Slodzian, B. Daigne, F. Girard, F. Boust and F. Hillion, CR Acad. Sci. Paris, 311-II (1990) 57. M. Bernheim, M. Chaintreau, R. Dennebouy and G. Slodzian, Surface Sci., 126 (1983) 610. H. Oechsner, Thin films and depth profile analysis. In: H. Oechsner (Ed.), Topics in Applied Physics, Vol. 37. Springer, Berlin, 1984, 63 pp. J.M. Schroer, H. Gnaser and H. Hoechsner, In: A. Benninghoven, Y. Nihei, R. Shimizu and H.W. Werner (Eds.), Secondary Ion Mass Spectroscopy, SIMS IX. Wiley, Chichester, 1994, 386 pp. N.M. Reed and J.C. Vickermann, In: L. Sabbatini and P.G. Tambonin (Eds.), Surface Characterization of Advanced Polymers. VCH, Weinheim, 1993, 83 pp. G. Blaise, R. Castaing and R. Queltier, J. de Phys., 45(Suppl. 2) (1984) 125. D.M. Gruen, M.J. Pellin, C.E. Young and M.H. Mendelsohn, Phys. Scripta, T6 (1983) 42. D. Briggs, A. Brown and J.C. Vickermann, Handbook of SSIMS. Wiley, Chichester, 1988. J.G. Newman, B.A. Carlson, R.S. Michael, J.F. Moulder and T.A. Hohlt, Static SIMS Handbook of Polymer Analysis. Perkin Elmer, Eden Prairie, MN 1981. A. Benninghoven, J. Vac. Sci. Technol., 3 (1985) 451. J.F. Ziegler and J.P. Biersack, The Stopping and Range of Ins in Solids. Wiley, London, 1999, Code SRIM 2000 available on: http://www.SRIM.org A. Havette and G. Slodzian, CR Acad. Sci. Paris, 209-B (1980) 51. R. Jede, O. Granschow and U. Kaiser, In: D. Briggs and M.P. Seah (Eds.), Practical Surface Analysis, Vol. 2, Ion and Neutral Spectroscopy. Wiley, New York, 1992, 425 pp. Y. Marie, Thesis, INPL Nancy, 1995. E. Darque-Ceretti, Thesis, Doctorate of Science, Besanc¸on, 1986. A. Benninghoven, F.G. Ru¨denauer and H.W. Werner, Secondary Ion Mass Specrometry. Wiley, New York, 1987. H.-N. Migeon, In: J. Ruste and J.F. Bresse (Eds.), Nouvelles Techniques de Micro et Nanoanalyse, E1. ANRT, Paris, 1995. N.M. Reed and J.C. Vickerman, Secondary Ion Mass Spectroscopy, SIMS VII. Wiley, Chichester, 1990, 793 pp. A. Adriaens, In: D.C. Creagh and D.A. Bradley (Eds.), Radiation in Art and Archaeometry. Elsevier, Amsterdam, 2000, 180 pp. G. Spoto, Thermochim. Acta, 365 (2000) 157. P. Holliger, In: A. Benninghoven, Y. Nihei, R. Shimizu and H.W. Werner (Eds.), Secondary Ion Mass Spectrometry, SIMS IX. Wiley, Chichester, 1994, p. 906. G. Faure, Principles of Isotope Geology. Wiley, New-York, 1977, Chapters 13 and 14. N. Gale and Z.A. Stos-Gale, Science, 216 (1982) 11. C.S.L. Yeung, R.W.M. Kwok, P.Y.K. Lam, S.P. Wong, S.K. Hark, Jin Zhengyao, P.C.L. Wongt, M.L. Yu and K.K. Mark, Surface Interface Anal., 29 (2000) 487. J.M. Eiler, C. Graham and J.W. Valley, Chem. Geol., 138 (1997) 221.
Secondary ion mass spectrometry. Application to archaeology and art objects 47 48
49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66
G. Giuliani, M. Chaussidon, C. France-Lanord, C. Rollion, D. Mangin and P. Coget, Analusis, 27 (1999) 203. G. Giuliani, M. Chaussidon, H.-J. Schubnel, D.H. Piat, C. Rollion-Bard, C. France-Lanord, D. Giard, D. de Narvaez and B. Rondeau, Science, 287 (2000) 631. L.M. Anovitz, J.M. Elam, L.R. Riciputi and D.R. Cole, Archaeol. J. Sci., 26 (1999) 735. S.B. Patel, R.E.M. Hedges and J.A. Kilner, Archaeol. J. Sci., 25 (1999) 1047. O. Dersch, F. Rauch and J.E. Eicson, 16th ICXOM Conference, Vienna, July 2001. P.M. Fischer, A. Lodding and J. Nore´n, In: Y. Maniatis (Ed.), Archaeometry. Elsevier, Amsterdam, 1989, 109 pp. E.M. Sterner, S. Risnes and P.M. Fischer, Eur. J. Oral Sci, 104 (1996) 253. P.M. Fischer, In: P. Fischer (Ed.), Archaeology and Natural Science (ANS), Vol. 1. Paul Astro¨ms Fo¨rlag, Go¨tegorg, 1993. I. Constantinides, M. Gritsch, A. Adriaens, H. Hutter and F. Adams, Anal. Chim. Acta, 440 (2001) 189. M. Schreiner, Mikrochim. Acta, II (1991) 255. S. Dalio, C. Piccirillo, C. Pagura, B. Faccin, S. Zecchin and M. Verita`, Rapid Commun. Mass Spectrom., 10 (1996) 1286. A. Aerts, Microscopic analysis of Roman vessel glass, Dissertation, University of Antwerp, 1998. G. Libourel, P. Barbey and M. Chaussidon, La Recherche, 25 (1994) 169. P. Craddock and A. Giumla-Mair, In: S. La Niece and P. Craddok (Eds.), Metal Plating and Patination. Butterworth, London, 1993, 101 pp. M. Aucouturier, SMI Rev. Art Technol., 2 (2002) 69. D. He´lary, E. Darque-Ceretti, Seventh International Conference on Adhesion and Surface Analysis, Loughborough, March 2002. J.J. Boon, K. Keune, J. van der Weerd, M. Gelfold and J.R.J. van Asperen, de Boer, Chimia, 55 (2001) 952. S. Volooj, C.M. Carr, R. Mitchell and J.C. Vickermann, Surface Interface Anal., 29 (2000) 422. R. Combarieu, G. Dauchot and F. Delamare, Metal 98. James and James Publisher, 1998, 223 pp. E. Darque-Ceretti, R. Combarieu and M. Aucouturier, Vide-Sci. Technique et Appl., 54 (1999) 141.
461
This page is intentionally left blank
Color plates
Color plates
Plate 4.I. Left panel: car paint multilayer stack, consisting of a series of nine (possibly 10) differently pigmented or transparent layers with thicknesses in the range 20 – 100 mm; right panel: X-ray intensity depth profiles of various elements. The sample was analysed with the blue layer on top. Layer 1 (blue) can be observed to contain Ti, some Co, Cu and Pb; layer 2 (red) contains Cr, Fe, Cu, Zn and some Sr; layer 3 (black) is not associated with any strong XRF intensity; layer 4 (orange) contains Ti, Fe, Co and Zn, but less Fe and more Zn than layer 2, layers 5–7 (black) also do not show strong XRF signals, except layer 6 that is associated with Sr and some Co; finally layer 8 (orange) shows a similar pattern as layer 4. Some of the signals originating from this deep layer (ca. 350 mm deeper than the surface) are strongly absorbed by the layers on top of it. Layer 9 is a transparent varnish layer. (See Fig. 4.42.)
Color plates
Plate 10.I. Overall view of Greek inscribed copper plaque, dated to the 8th– 7th century BC , which preserves two rare letters in the Greek script which only occurred during the 8th–7th centuries BC as Greek evolved from Phoenician. Dimensions 212 £ 137 £ 1.2 mm3 thick. (See Fig. 10.1.)
Color plates
Plate 10.II. Photomicrograph of lettering inscribed into the original copper surface of the Greek plaque shown in Fig. 10.1, now preserved within the cuprite corrosion layer, and not in the metallic surface itself. Magnification £45. (See Fig. 10.2.)
Color plates
Plate 10.III. Egyptian bronze solid cast statuette of the God Osiris, inlaid with gold and blue glass. The surface has been extensively altered to massive light blue and dark green corrosion, identified as an overall patina of atacamite, with patches of chalconatronite. Frontal view. (See Fig. 10.11.)
Plate 10.IV. Egyptian bronze solid cast statuette of the God Osiris, inlaid with gold and blue glass. The surface on the reverse shows massive regions of alteration to chalconatronite. Reverse view. (See Fig. 10.12.)
Color plates
Plate 11.I. Small sphere of a soldering alloy introducing cadmium sulphide in gold –silver–copper alloy as made by metal workers in antiquity: (a) just after melting; please note the black deposit around the cadmium gold alloy; (b) after a gentle compression. (See Fig. 11.11.)
Plate 11.II. The Byzantine cross of the treasure of the Cathe´drale Notre-Dame de Tournai. (See Fig. 11.24.)
Color plates
Plate 11.III. Zoomorphic pendant (Columbia, 500 – 1500 (See Fig. 11.28.)
AD ;
length: 71 mm).
Color plates
Plate 11.IV. Eagle warrior pendant (San Francisco Caxonos, Oaxaca, 1500 AD ; length 8.0 cm), Oaxaca’s Regional Museum, Mexico. (See Fig. 11.29.)
Color plates
Plate 12.I. The Entrance of the Mother of God into the Temple, Manuel Panselinos. Wall painting, 2 m £ 3.5 m. Prior to cleaning and with the positions of the colour measurements indicated. (See Fig. 12.1.)
Plate 12.II. Fluorescence photography of the wall painting shown in Plate 12.I. (See Fig. 12.5.)
Color plates
Plate 12.III. Cross-section from the background. Photography via a microscope. Reflected light. (a) Lime and carbon black; (b) carbon black and lime. (See Fig. 12.8A.)
Plate 12.IV. Cross-section of a letter from the inscription. Photography via a microscope. Reflected light. (a) Lime and carbon black; (b) carbon black and lime; (c) azurite; (d) lime. (See Fig. 12.9A.)
Color plates
Plate 12.V. Cross-section of a light in the Virgin’s mantle. Photography via a microscope. Reflected light. (a) Caput mortuum; (b) azurite. (See Fig. 12.10A.)
Plate 12.VI. Cross-section of a principal line at far left of the maiden’s tunic. Photography via a microscope. Reflected light. (a) Lime, carbon black and yellow ochre; (b) warm ochre and lime; (c) red ochre. (See Fig. 12.11A.)
Plate 12.VII. Cross-section of a light in Zacharias’ tunic. Photography via a microscope. Reflected light. (a) Red ochre; (b) cinnabar; (c) minium. (See Fig. 12.12A.)
Color plates
Plate 12.VIII. Cross-section of a light in tunic of central maiden. Photography via a microscope. Reflected light. (a) Red ochre; (b) cinnabar; (c) altered minium. (See Fig. 12.13A.)
Plate 12.IX. Cross-section of a light in cloak of the central maiden. Photography via a microscope. Reflected light. (a) Principal line: yellow ochre and grains of red ochre; (b) underlay: green earth and lime; (c) light: lime and green earth; (d) highlight: lime. (See Fig. 12.14A.)
Plate 12.X. Cross-section from flesh tone in St Joakeim’s right foot. Photography via a microscope. Reflected light. (a) Principal line: red ochre and grains of carbon black; (b) underlay: green earth and yellow ochre; (c) flesh tone: yellow ochre and lime. (See Fig. 12.15A.)
Color plates
Plate 12.XI. (a) ap bp diagram (CIELAB 1976) of the colour measurements in The Entry of the Mother of God; (b) ap bp diagram (CIELAB 1976) of the total of colour measurements. (See Fig. 12.16.)
Color plates
Plate 12.XII. Mother of God Hodegetria. Egg-tempera on wood, 94 £ 69.5 £ 2.5 cm. Before cleaning. (See Fig. 12.18.)
Color plates
Plate 12.XIII. (a) Rear side. Visible are the joints of the three wooden panels, the three battens and the white layer of the ground that was applied on the surface. The small rectangles with broken black lines demarcate the positions from which the fragments of wood were removed in order for the iron nails to be inserted between the joints. (b) Upper (left) and lower (right) sides of the icon. The texture of the wood at the upper end was exposed because of the excision of a narrow horizontal strip, which also included the wooden frame. The lower end was preserved unharmed. Visible are the remains of the white ground, which was applied on the sides. (See Fig. 12.19a,b.)
Color plates
Plate 12.XIV. (A) Cross-section from a light in the Virgin’s mantle. Photography via a microscope. Reflected light. (a) Gesso ground; (b) underlay: caput mortuum, red cochineal lake (?) and carbon black; (c) 1st light: caput mortuum, cochineal lake (?), lead white and grains of carbon black; (d) 2nd light: lead white; (e) tinning: a glaze of red cochineal lake (?); (f) sandarac varnish. (B) Fluorescence under ultraviolet light. (See Fig. 12.23A,B.)
Plate 12.XV. (A) Cross-section from a light in the Virgin’s tunic. Photography via a microscope. Reflected light. (a) Gesso ground; (b) underlay: indigo and grains of lead white; (c) light: azurite, grains of yellow ochre, of red ochre and of lead white; (d) sandarac varnish. (B) Cross-section from a light in Christ’s tunic. Photography via a microscope. Reflected light. (a) Gesso ground; (b) underlay of the mantle: caput mortuum, red cochineal lake (?) and carbon black; (c) underlay of the tunic: azurite and lead white; (d) 1st light: lead white and azurite; (e) sandarac varnish. (See Fig. 12.24A,B.)
Color plates
Plate 12.XVI. (A) Cross-section from a gilt line in Christ’s cloak. Photography via a microscope. Reflected light. (a) Gesso ground; (b) underlay: yellow ochre and cinnabar; (c) mordant: boiled linseed oil (?) and carbon black; (d) gilt line: gold leaf; (e) sandarac varnish. (B) Fluorescence under ultraviolet light. (See Fig. 12.25A,B.)
Plate 12.XVII. Cross-section from a flesh tone in Christ’s right foot. Photography via a microscope. Reflected light. (a) Gesso ground; (b) underlay: yellow ochre, carbon black, cinnabar, grains of green earth and of lead white; (c) flesh tone: lead white, cinnabar and yellow ochre; (d) highlight: lead white; (e) sandarac varnish. (See Fig. 12.26A.)
Color plates
Plate 16.I. Fragment of an archaeological glass with an iridescent surface layer formed during the corrosion of the glass in soil.
Color plates
Plate 16.II. Medieval glass painting: The Birth of Christ, North window of the former monastery church at Viktring/Austria.
Plate 16.III. Exterior surface of a medieval glass painting with a weathering crust completely covering the glass surface.
Color plates
Plate 16.IV. Medieval glass painting with a weathering crust on the exterior surface reducing the transparency so that the composition is barely recognizable.
Plate 16.V. The medieval Goblet with Lid of the Kunsthistorisches Museum Vienna, Inv. No. PL 85, enamelled on the outside and inside with “e´mail en ronde bosse.” The artefact is decorated with stars and moons made of gilt silver and beasts painted with opaque white paint enamel.
Color plates
Plate 16.VI. The inside of the Goblet with Lid of the Kunsthistorisches Museum Vienna, Inv. No. PL 49, is decorated with ultramarine blue enamel showing white crystals (weathering products) on the surface.
Plate 17.I. Meditationes passionis Domini Nostri Iesu Christi (reproduced with kind permission of National Library of Poland). (See Fig. 17.1.)
Color plates
Plate 17.II. Fe(II) and Fe(III) distribution maps within the character area (data from m-XANES). (See Fig. 17.10.)
Part II: Case Studies Section
Chapter 10
The non-destructive investigation of copper alloy patinas David A. Scott
10.1
A BRIEF HISTORICAL ACCOUNT
Bronze vessels that have been interred under the earth a thousand years appear pure green the colour of kingfisher feathers, those that have been immersed in water a thousand years are pure emerald in colour with a jade-like lustre. Those that have not been immersed as long as a thousand years are emerald green but lack the lustre, those that have been transmitted down from antiquity, not under water or earth but through the hands of men, have the colour of purple cloth and a red mottling like sand, which protrudes when excessive, and looks like first-quality cinnabar. When boiled in a pan of hot water the mottling becomes more pronounced. Zhao Xigu, 1230 AD .
This quotation from the Song Dynasty (960 – 1279 AD ) writer Zhao Xigu [1], who reports here on his observations of bronze patinas, is germane to the subject of this chapter which concerns the observation of bronze patinas and the methods of their non-destructive analysis. The scientific examination of ancient bronzes has a long tradition, through the studies of many eminent chemists, such as the French scientist, Sage in 1779, [2] and John Davy, the brother of Humphry Davy, who in 1826, described his examination of a bronze helmet recovered from the sea near Corfu. Davy determined, by means of wet chemical analysis, both the chemical composition of the metal, as an alloy of copper and tin, and the identity of the corrosion products. He was able to identify the ruby-red suboxide of copper (cuprous oxide: cuprite); the green rust of the carbonate (basic cupric carbonate: malachite), submuriate of copper (one of the copper trihydroxychlorides: probably paratacamite, clinoatacamite or atacamite), crystals of metallic copper (which can be redeposited from solution during the corrosion of bronze objects) and a dirty white material, tin oxide (stannic oxide: cassiterite). This was an impressive study for the time; Davy was also Comprehensive Analytical Chemistry XLII Janssens and Van Grieken (Eds.) q 2004 Elsevier B.V. All rights reserved
465
D.A. Scott
aware that some of his work was potentially important for the authentication of ancient bronzes. He writes [3]: Permit me to lay before the Royal Society, the results of some experiments and observations on the incrustations of certain ancient alloys of copper, which I trust may not be undeserving of notice, whether considered in connexion with the arts of Ancient Greece, or in relation to the slow play of chemical affinities acting during a long period of time.
Similar information regarding metallic composition and corrosion is still sought today, when advances in instrumentation have resulted in an array of methods for non-destructive investigation. The techniques of environmental scanning electron microscopy (ESEM), Raman spectroscopy (RS), Fourier transform-infra red spectroscopy (FTIR), X-ray fluorescence (XRF) spectrometry, X-ray diffraction (XRD) and optical examination represent some of the laboratory techniques which are now commonly available for the investigation of copper alloy patinas in a completely non-destructive manner. RS, in particular, has shown considerable development in the last few years, with the result that several applications of the technique in the non-destructive analysis of art objects and patinas have been published [4– 6]. Some of these techniques do have limitations on object size, such as the ESEM which may be able to accommodate only small artefacts, a few cubic centimetres in dimensions, but most of the other techniques can now be used to examine copper objects in situ, regardless of size. The advent of non-destructive examination of patinas has had to wait until the development of suitable analytical instrumentation to address this need. If we think about the methods that were available, e.g., in 1975, there were relatively few appropriate methods suitable for the non-destructive examination of patinas, such as XRF, X-radiography, and visual examination using the optical light microscope. Today, 27 years later, there are a plethora of possible techniques, some of which are described in the other chapters of this book. Some important questions regarding copper alloy patinas still require destructive sampling, e.g., the electron microprobe analysis of polished sections is still necessary in order to determine precise elemental composition of the patina, to investigate the absorption of elements from burial, such as silicon, phosphorus, aluminium, oxygen, chlorine and sulphur, and to determine the percentages of remaining metallic elements such as copper, tin, arsenic, lead, zinc and so on. However, the quality and quantity of information that may now be determined non-destructively has brought exciting new possibilities of artefact studies within reach. The aim of
466
The non-destructive investigation of copper alloy patinas
the present chapter is to describe how some of these methods may be used in the characterization of copper alloy patinas. There are two commonly used definitions of “non-destructive” examination as applied to cultural materials. The first definition, not uncommon in scientific circles, is to regard “non-destructive examination” as applying to a sample removed from the object for purposes of analysis. The sample which is taken is not consumed by the analytical technique, and can be stored in a vial, or in some cases reattached to the object after examination. The second definition of “non-destructive” implies that the integrity of the object itself is not affected in any way by the analysis being performed, and that no sample is removed from the object under study. It is this latter approach to the subject that the present chapter is concerned with, since with this non-invasive analysis the integrity of the object is not compromised in any way during the study. This approach to the concept of “nondestructive” represents the best possible scenario for the conservation of the object itself; the meaning of “non-destructive” within the conservation community is normally taken to mean that no sampling is permitted from the object under any circumstances. In the past, this restriction was a source of great frustration to the scientist investigating cultural materials, but now we can satiate much of our curiosity without sampling and still gain invaluable knowledge. 10.2
OPTICAL EXAMINATION
Bench binocular microscopy and the detailed study of the overall surface patina of ancient copper alloys should not be overlooked in the process of examination and non-destructive analysis. The overall description of the copper patina is heavily dependent on this phase of the examination, during which the colour, coherence, surface characteristics, variability, associated materials, pseudomorphic remnants of organic materials, and the extent of penetration of the patina into the sound remaining metal can all be gauged and recorded, as a first step in the process of non-destructive investigation of copper and bronze patinas. An example of this initial phase of study is the description of the surface patina of an important inscribed Greek copper plaque, with an unusual version of the Greek language on both obverse and reverse sides of the plaque, illustrated in Fig. 10.1. Indeed, because of the strange nature of the Greek text inscribed on both sides of the plaque, and the fact that the corrosion crust on one of the four known plaques of this type, in the collections of the Johann von Wagner Museum, Wurzburg, had been described as a forgery, it was especially important to determine if the patina
467
D.A. Scott
Fig. 10.1. Overall view of Greek inscribed copper plaque, dated to the 8th– 7th century BC , which preserves two rare letters in the Greek script which only occurred during the 8th–7th centuries BC as Greek evolved from Phoenician. Dimensions 212 £ 137 £ 1.2 mm3 thick. ( For a colored version of this figure, see Plate 10.I.)
of the plaque examined here appeared to be authentic or not. Under binocular magnification at 40 £ , the patina was determined to be multilayered and complex, with associated pseudomorphic organic remnants, small fragments of charcoal and a complex structure to the corrosion interface between metal and corrosion products. Some of the lettering of the plaque is preserved within this corrosion layer, as can be seen from Fig. 10.2, which illustrates part of the cuprite surface in which epitactic growth of the cuprite has resulted in preservation of the inscribed lettering.
468
The non-destructive investigation of copper alloy patinas
Fig. 10.2. Photomicrograph of lettering inscribed into the original copper surface of the Greek plaque shown in Fig. 10.1, now preserved within the cuprite corrosion layer, and not in the metallic surface itself. Magnification £ 45. ( For a colored version of this figure, see Plate 10.II.)
Cemented to the surface are a number of different mineral phases, highly rounded clear quartz grains ranging in size from about 0.1 – 0.4 mm, occasional mineralized textile fibres, small fragments of wood, both mineralized and only lightly stained with copper corrosion products, and a hard, dark green corrosion crust that was determined later by in situ XRD studies to be of malachite with a little atacamite, overlying a cuprite patina, all of which appears to be perfectly acceptable for an ancient copper object. X-radiographic examination of the plaque at 100 kV, 5 mA, showed the clear presence of hammering marks within the copper plaque, showing that
469
D.A. Scott
the plaque had been worked and annealed to shape, and was not made of rolled copper sheet. Studies are currently in progress to describe these unique objects in greater detail, which will be published in the epigraphic and scientific literature [7]. 10.3
ENVIRONMENTAL SCANNING ELECTRON MICROSCOPY
The principles of SEM examination are well known and will not be described here. The principal drawback of conventional SEM examination is that the sample or object must be coated in order to image the surface and obtain an analysis. Use of the ESEM allows examination of uncoated surfaces, which is potentially very important for patina studies, since the fragility or integrity of the surface of the object is entirely unaffected by the ESEM examination. The ESEM utilizes a sophisticated differential pumping system and a series of pressure-limiting apertures to create a pressure gradient between the sample chamber of about 900 Pa and the electron gun compartment of about 1025 Pa. To maintain the high resolution of an SEM and to minimize scattering of the primary electrons by gas molecules, the distance between the sample and the final lens assembly in an ESEM is kept as short as possible. The high pressure in the sample chamber precludes the use of the classical Everhart– Thornley detector and therefore a new, environmental secondary detector (ESD) has been developed for the use in ESEM instrumentation [8]. The principals of the final operational path of the ESEM are shown in Fig. 10.3. ESEM examination was employed to study the patina of a number of small bronze objects and fragments, excavated from the southern Italian site of Francavilla Marittima, dating to the 6th – 5th centuries BC . The results of the examination of a heavily corroded phiale fragment are shown in Figs. 10.4 – 10.6. The patina of the phiale reveals several morphological variants of patina preservation that were studied in order to understand some of the corrosive events that had been responsible for the severe deterioration of some of the bronze objects from this site. The scanning electron photomicrograph shown in Fig. 10.4 was taken from a dark green area of the surface and reveals exceptionally well-crystallized patina components in this location, which were later characterized (qv) as atacamite, Cu2(OH)3Cl and clinoatacamite, another recently identified isomer of Cu2(OH)3Cl. The clinoatacamite crystals in Fig. 10.4 are revealed as occurring as acicular sheaths of fine needles between the patina of atacamite. This is an unusual patina, but research has shown that it is possible for a copper patina to be formed overall from atacamite, overlying the cuprite that occurs contiguous with the metallic substrate. In this case,
470
The non-destructive investigation of copper alloy patinas
Fig. 10.3. Diagram showing principles of operation of the environmental secondary detector, utilized in the ESEM, allowing in situ examination of metallic surfaces without coating the sample or touching it in any way. (After Stulik and Doehne, 1991.)
the phiale fragment was so heavily corroded that little metallic remnants remain; the whole being effectively mineralized. Figure 10.5 illustrates the in situ examination of the green patina in an area where only well-developed, euhedral crystals of atacamite occur within the patina. The ability of the patina to develop such well-characterized crystals is unusual, but further non-destructive examination of copper alloy patinas should enable investigators to document further examples in the future. The broken fragment revealed regions showing the underlying cuprite, Cu2O, below the atacamite layer, and an example is shown in Fig. 10.6 of a cuprite crystal which is undergoing transformation to clinoatacamite. It is interesting to observe that the basic copper chloride is growing from the cuprite directly from
471
D.A. Scott
Fig. 10.4. ESEM photomicrograph at 600 Torr pressure of part of the surface of a small Phiale fragment from the site of Francavilla Marittima, 6th –5th centuries BC , showing that the patina consists of large crystals of atacamite, with smaller acicular growths of clinoatacamite. Scale bar represents 50 m.
472
The non-destructive investigation of copper alloy patinas
Fig. 10.5. ESEM photomicrograph at 600 Torr of part of the surface of the small Phiale fragment from the 6th–5th centuries BC , showing the presence of finely developed euhedral crystals of atacamite which comprise most of the patina. Same magnification as Fig. 10.4.
different planes of the crystal face. The cubic nature of the cuprite crystals themselves can also be clearly seen in Fig. 10.6. This study was particularly revealing in being able to characterize the nature of the copper alloy patina on the phiale fragment without the need to coat the sample or interfere with it in any way. Since it is often assumed that the copper trihydroxychlorides do not form a uniform patina on ancient copper alloy objects, and that the patina must be based on malachite, this example is an important one in showing that such patinas are genuine occurrences. The elemental composition of the different phases present on the surface can also be analysed in the ESEM, using energy dispersive X-ray analysis methods which are well-known and which will not be described further here.
473
D.A. Scott
Fig. 10.6. ESEM photomicrograph at 600 Torr of part of the surface of the small Phiale fragment, revealing the patina underneath the atacamite layer where it has broken away. The large cubic crystals are of cuprite, and within this layer, the clinoatacamite crystals can be seen to be growing directly from the cuprite layer. The magnification is twice that of Fig. 10.4.
10.4
X-RAY FLUORESCENCE ANALYSIS
The application of XRF analysis to the elemental characterization of works of art is well known and has been, for the last 30 years, one of the most powerful tools for the non-destructive investigation of cultural materials. From studies of 17th century dance costumes [9], to platinum and palladium photographic prints [10], from arsenic and lead in violin varnishes [11] to the development of Chinese overglaze enamels [12], the use of XRF analysis range over a remarkably wide territory which encompasses objects made
474
The non-destructive investigation of copper alloy patinas
primarily from inorganic materials to those where the inorganic components are less obvious, such as costumes for the dance, and in many of these varied materials, no samples whatever can be removed for analysis. Most of our own XRF analyses of works of art have been carried out using a Kevex 0750A spectrometer, operating in a large lead-lined room, which enables objects of practically any size to be examined. The advantages of the Kevex 0750A instrument are that it can be operated in air or helium, employing secondary targets (STXRF), which can be selected specifically for certain kinds of analytical investigations, and which provide optimum detection limits for the elements under study. An interesting example of the use of XRF analysis for patina studies is that of the 18th century bronze sculpture, Medea Rejuvenating Aeson in the J. Paul Getty Museum (Acc. no. 74.SB.6) shown in Fig. 10.7, which is attributed to the French sculptor Louis-Simon Boizot (1743– 1809). The bronze exemplifies a common neo-classical theme of vestal virgins and rejuvenating scenes much loved by French artists in the 18th century. The sculpture is of uncertain date and is not signed. It was examined for technological evidence in part to establish whether this could indeed be an 18th century cast. In the course of this work, surface analysis of the metal composition by secondary target XRF analysis showed that the sculpture was cast in a leaded brass alloy, rather than a bronze, and that the thin, dark brown over steel-grey coloured gloss surface consisted of a platinum coating. The XRF analysis of unpatinated regions of the inner surface or base rim revealed the following composition for the metallic substrate. Only copper, zinc, lead, tin and nickel were detected and the analytical results for different areas are as indicated in Table 10.1. No arsenic, antimony, silver or gold was detected in these analyses, which show that the Boizot group has been assembled from a leaded brass containing a small amount of tin, a composition similar to the leaded brasses still used in France today to cast decorative brass objects [13]. Most Renaissance and later bronzes show detectable amounts of nickel [14], which is also present here. The accuracy of the results is limited by the difficulty of analysing a flat surface of the brass as it is presented to the X-ray beam. The predominantly rounded surfaces of the sculpture account for the variations to be seen in the results shown above in which totals less than 100% are to be expected. No major differences in composition could be found for the individual components on surface XRF analysis, suggesting that a similar alloy had been used to make the separately cast parts of the object.
475
D.A. Scott
Fig. 10.7. Large brass sculptural group. Attributed to Louis Simon Boizot (1743 – 1809), Medea Rejuvenating Aeson. Brass group, Height 26 3/8 in. (67 cm), frontal view. The J. Paul Getty Museum, 1974.SB.6. The patina of this brass was shown to be coated with platinum creating the dark steel-grey patina.
When the sculpture was analysed by XRF on the patinated surface of the brass casting, prominent peaks for the presence of platinum could be discerned, which are clearly not derived from the composition of the copper alloy itself, but must be a component of the patina which had been applied over the brass-coloured alloy in order to create a dark chocolate-brown coloured patina.
476
The non-destructive investigation of copper alloy patinas TABLE 10.1 Local compositions observed in an 18th century bronze sculpture by means of XRF Spectra
Area
Cu
Zn
Pb
Sn
Ni
Total
Boizot 7 Boizot 10 Boizot 11
Cauldron handle Polished base rim Inner surface
61.5 51.8 69.5
26.6 17.4 14.3
5.4 2.5 2.6
0.7 0.5 0.4
tr tr tr
94.2 71.7 86.8
According to the literature, the platinum patina could have been applied either by: an electrolytic technique; by use of a mercury – platinum amalgam; or by electrochemical replacement plating from a platinum salt solution. The electrolytic coating of platinum would require immersion of the entire object, which would leave a deposit on the inner surface of the hollow sculpture as well. It was not possible to take the sculpture apart and analyse the inner surface of each component part, however, there is no trace of platinum on the inner surface of the base or on the rim of the soldier’s chest that was removed and re-fastened for conservation purposes. The component parts are probably coated in a similar fashion since the surface coloration is relatively uniform. The surface treatment of the base would, therefore, be representative of the other pieces. A mercury –platinum amalgam would have left a trace of mercury at the surface of the object, but again no evidence of the presence of mercury could be found by analysis of the inner or outer surfaces. The outer coating is very thin in some areas and has none of the features associated with a mercuryamalgam process. For example, there are no areas of apparently thicker deposits, pooling of the amalgam, or residues of mercury. The visual and analytical evidence suggests that a patination solution was neatly applied to coat only the outer surfaces with platinum. Whether or not the coating was applied to the separately cast parts or to the assembled sculpture is not certain. This leads to the conclusion that the surface coating was created by the third alternative method of application by brush or swab of a chemical solution. The platinum patina is, therefore, deposited by electrochemical replacement plating and is protected by a wax finish. For further details of the platinum patina study the reader is referred to the research carried out on this object by Bewer and Scott [15]. 10.5
SCANNING X-RAY FLUORESCENCE MICROANALYSIS
The idea for scanning acquisition of XRF elemental images began to become feasible little more than a decade ago. Schreiner et al. [16], described a
477
D.A. Scott
system that was developed at the Technical University of Vienna. This system, rather like a flat-bed scanner, is equipped with a Si(Li) detector and a spectroscopic X-ray tube with a maximum power of 2800 W. The primary beam is collimated to a diameter of 1.2 mm and DC motors are used to move the support for detector and X-ray tube in horizontal and vertical directions, parallel to the object and above it. The selected pixels for analysis can be spread over an area of up to 1000 £ 800 mm2. At about the same time as the system in Vienna was being built, engineers at Kevex Corporation began to work on a concept designed to enable wafers for the electronic industry to be mapped for elemental distribution by X-ray microfluorescence analysis, employing a fine-focussed X-ray tube capable of delivering enough flux to perform a satisfactory analysis over diameters as small as 10 – 50 m [17,18]. This instrument is the Kevex Omicron X-ray microfluorescence spectrometer. The X-ray unit is self-contained and can be operated in air, under helium or in vacuum. The maximum size of objects which can be placed in the chamber are 240 mm long £ 230 mm wide £ 110 mm in height. For scanning, the objects must be held or placed in a holding device which limits the useable size to about 220 £ 190 £ 100 mm3: about the size of a large cigar box. The X-ray tube uses a molybdenum target, operating at 50 kV, 1 mA, and the X-ray beam can be collimated to diameters from 50 – 500 m. An image of the object surface can be viewed in colour at all times on the video monitor attached to the machine, since a miniature video camera is focussed at the same area on the object as the X-ray beam and detector. While observing the object on the monitor, a joy-stick can be used to move the stage remotely, allowing any area of the object to be selected and brought into view. Three different angles of lighting of the object are available which correspond roughly to bright field, dark field, and oblique illumination. Usually one of these is satisfactory for optimal viewing of the sample or object while the spot analysis or X-ray scan is being performed. After setting regions of interest (ROI) for the scanning function, under Xmap/Chemimagec, a number of variables concerning the scanning regime can be selected, e.g., map size, number of frames, motor speed in steps/s, and dwell time variation (time/pixel). The maximum map size is 256 £ 256 pixels. Stage co-ordinates can be precisely set with a joystick, and mapping can then commence. Depending on the motor speed and map size, the acquisition time may vary between some minutes to several hours. Longer scans can be set up to run overnight, which frees up the instrument during the day. Scans or simple XRF spot analysis of an object can be run at any time in air, helium or under vacuum. The Greek copper plaque illustrated in Fig. 10.1 was scanned
478
The non-destructive investigation of copper alloy patinas
in this instrument, and some of the results of the surface studies are shown in Fig. 10.8. Figure 10.8 reveals that there are occasional tiny particles of gold present, which must have been embedded in the corrosion crust during burial, since the plaques themselves are not gilded. Also present are occasional calcite grains, as shown by the high calcium concentration in the elemental map. The presence of calcite was later confirmed by in situ XRD studies.
Fig. 10.8. Scanning elemental maps obtained by in situ X-ray microfluorescence spectroscopy showing part of the surface of the Greek plaque illustrated in Fig. 10.1. The elemental maps for gold and silver reveal that a small flake of gold is embedded in the patina, as well as some calcium-rich mineral grains, which are probably calcite. The overall area examined by the elemental scanning is 2.3 £ 2.4 mm2.
479
D.A. Scott
10.6
XRD ANALYSIS
There are two recent developments in XRD analysis, both of which are very significant for the non-destructive examination of copper alloy patinas. These developments are: (a) Conventional X-ray diffractometry employing a Go¨bel mirror assembly for the analysis of whole objects: the surface does not need to be flat. (b) General area X-ray microdiffraction (GADDS) for the characterization of selected regions or surface patinas of small objects. Advances in X-ray physics has enabled the production of the Go¨bel mirror, which produces an intense parallel beam of X-rays allowing the in situ XRD analysis of many museum objects to be undertaken. The machine that we are using for this type of examination is a Siemens (Bruker) D5005 theta –theta instrument with attached Go¨bel mirror, employing detector scan mode typically through 5 – 808 2-theta. We have arranged two small lasers to be co-incident at the correct focal height for the instrument, so that the object to be examined can be simply raised into position on a laboratory scissors jack stand. In situ studies of the plaque shown in Fig. 10.1, allowed a depth profile XRD scan to be acquired, which is shown in Fig. 10.9. The advantage of this technique for the examination of copper patinas is that it allows some of the components of the patina or the metallic phase to be determined remotely, without sampling. The depth profile in the case of the plaque shows that there are relatively minor variations in XRD data within a few microns of the surface. Individual scans from this threedimensional array can be isolated for characterization, as shown in the diffractogram of Fig. 10.10, which is the 29th scan from the immediate surface, which reveals the presence of cuprite, malachite, Cu2(CO3)(OH), and atacamite within the patina. If the first scan is examined, only quartz and malachite can be seen, which illustrates the utility of the technique in the characterization of copper alloy patinas, since it allows in situ probing of the surface, entirely non-destructively. The example shown in Figs. 10.11 and 10.12 is that of a gilded and inlaid Egyptian bronze statuette of the God Osiris, which was thought to be suffering from bronze disease, particularly in the regions where the light blue excrescences occur on the surface, as in the reverse, shown in Fig. 10.12. Surface patina studies of the different corrosion products of the Osiris were carried out in order to investigate the corrosion products and to ascertain if the light blue corrosion which can be seen interspersed in the green corrosion crust of the bronze is an example of bronze disease or not. The area of the
480
The non-destructive investigation of copper alloy patinas
Fig. 10.9. Three-dimensional X-ray diffractogram for the surface examination of the Plaque shown in Fig. 10.1. The x-axis represents the d-spacings, y-axis the peak intensities, and z-axis, the depth profiling, which extends a few microns into the patina of the object under study.
surface chosen for examination contains several excrescences, and is rough and uneven, being veined with light blue corrosion contiguous with the rough dark green surface. The surface XRD study obtained from the dark green patina of the Osiris, produced the diffractogram shown in Fig. 10.13, which reveals that the dark green crust is composed of atacamite. Following this determination, the object was re-positioned in order to characterize the light blue surface, which produced the diffractogram shown in Fig. 10.14, revealing that this part of the patina is composed of chalconatronite, Na2Cu(CO3)2·3H2O, a mixed sodium copper carbonate. The chalconatronite has formed in this case, not from conservation treatment, which is commonly thought to be the aetiology of chalconatronite in many instances, but from natural corrosion processes in the natron-rich soils of Egypt. The investigation was, therefore, able to conclude that the Osiris is not suffering from active bronze disease corrosion, as the light blue surface was only suffering from mechanical powdering and not from chemical instability due to bronze disease or cuprous chloride, CuCl. Once again, the overall patina of
481
D.A. Scott
Fig. 10.10. The 29th scan extracted from the three-dimensional array shown in Fig. 10.9, which reveals the presence of cuprite, malachite and atacamite within the patina.
the bronze Osiris is interestingly revealed to be of atacamite, with no malachite present at all. There appears to be an association between the occurrence of chalconatronite as a natural corrosion product and atacamite and cuprite which is exemplified by this example. The second type of XRD equipment now available, uses general area diffraction detector geometry (GADDS), which allows for small objects to be analysed by XRD, or for selected areas of the surface to be analysed. The object is positioned on the stage, of which several variants are available, and the area to be analysed is chosen by manipulation of the sample and alignment using a laser and video microscope positioned at the correct axial angle. Acquisition of the XRD scan can be spread over two or more frames, allowing selected angle diffraction peaks to be searched, or an overall XRD pattern of an unknown to be acquired. The patina of the Francavilla Marittima material, already introduced in the section on ESEM studies, shown in Figs. 10.4 – 10.6, was characterized using this technique, employing a fragment of the broken and corroded phiale. The surface patina of this object can be visually seen to contain two different corrosion products which were analysed in two different
482
The non-destructive investigation of copper alloy patinas
Fig. 10.11. Egyptian bronze solid cast statuette of the God Osiris, inlaid with gold and blue glass. The surface has been extensively altered to massive light blue and dark green corrosion, identified as an overall patina of atacamite, with patches of chalconatronite. Frontal view. ( For a colored version of this figure, see Plate 10.III.)
regions of the surface, one producing the X-ray diffractogram shown in Fig. 10.15. The diffractogram for Fig. 10.15, reveals that the darker green surface of the bronze is covered with a layer of malachite, whilst Fig. 10.16 shows that the lighter green –blue surface is primarily a mixture of atacamite and clinoatacamite; the long acicular needles in the ESEM photomicrograph of Fig. 10.6, represent the clinoatacamite phase. Also present in the patina are small amounts of gypsum and malachite. Recent research has shown that some of the products of “bronze disease” may be one of the four copper
483
D.A. Scott
Fig. 10.12. Egyptian bronze solid cast statuette of the God Osiris, inlaid with gold and blue glass. The surface on the reverse shows massive regions of alteration to chalconatronite. Reverse view. ( For a colored version of this figure, see Plate 10.IV.)
trihydroxychlorides, particularly paratacamite or clinoatacamite, the latter being a recently characterized polymorph which may be present in many situations within bronze patinas, which were previously characterized as paratacamite [19,20]. In the patina of the fragment from Francavilla, that was certainly the case, as the match of the XRD data is closer to clinoatacamite rather than paratacamite. With conventional X-ray diffractometry, there are limitations on the size and surface features of the object to be studied that have meant relatively few non-destructive examinations have been reported in the literature.
484
The non-destructive investigation of copper alloy patinas
Fig. 10.13. X-ray diffractogram obtained, in situ, for the dark green area of the patina of the Osiris shown in Fig. 10.11, on the front of the figurine. The principal constituent of this patina is atacamite.
One such study is that of Lee [21] who reported on a study of several corroded coins in the collection of the British Museum. A pale brown and compact patina on Chinese lead coins from the Kaiyuan tong bao period, of about 850 AD , was shown by in situ XRD to be composed of cerussite, PbCO3 with minor amounts of PbO. Poor storage conditions were suspected to have contributed to the corrosion of a copper alloy coin from the Xi Ning tong bao period, dated to 1068 – 1077 AD , where black patches showed the presence of copper sulphides, probably of digenite, Cu1.8S and chalcocite, CuS. Green corrosion on a group of United States coins of 1813 and 1808, together with a Russian coin of Nicholas I, showed evidence of copper formates, acetates and chlorides, while a strong smell of acetic acid was associated with a group of coins from the Xian Jeng tong bao period, minted in 1854 AD . With the new XRD techniques mentioned above, it would now be possible to analyse these surface patinas without difficulty. 10.7
FTIR SPECTROSCOPY
The use of the remote arm on the FTIR microscope, allows for the in situ analysis of a wide array of objects and materials, both inorganic and organic,
485
D.A. Scott
Fig. 10.14. X-ray diffractogram obtained, in situ, for the light blue area of the patina of the Osiris shown in Fig. 10.12, on the back. The principal constituent of this patina is chalconatronite.
even when the surface is not completely flat, although if the surface of the patina to be examined is extremely rough, e.g., with the chalconatronite crust examined on the Egyptian Osiris shown in Fig. 10.12, then it may not be possible to obtain an FTIR spectrum from either total reflection examination or from attenuated total reflection (ATR). The surface of even the green patina, best shown in Fig. 10.11, of the Osiris proved to be difficult to characterize using FTIR because of the very uneven and crystalline nature of the patina. For the study of the Egyptian Osiris bronze statue, a side port reflectance accessory attached to a Nicolet “Nic-Plan” infrared microscope was utilized. Both Reflachromat and ATR objectives were tested to determine the best objective for the study. Due to the rough and uneven surface of the statue, the ATR objective was unsuccessful in collecting clear and interpretable spectra. However, the use of the Reflachromat objective resulted in spectra that enabled the identification of two differently corroded areas on the surface of the statue. The spectrum contained major bands that matched well with the reference data for atacamite. The second spectrum obtained, showed a mixture of atacamite and paratacamite. The presence of this mixture was evident in the fingerprint region between 989 and 831 cm21.
486
The non-destructive investigation of copper alloy patinas
Fig. 10.15. XRD in situ using the GADDS microdiffraction equipment for a corroded phiale fragment from the site of Francavilla Marittima, Southern Italy, dated to the 6th–5th century BC , showing the patina in this area to be principally of malachite. A small amount of cuprite, from the underlying layer, and atacamite were also detected in this region of the surface.
One of the spectra is shown in Fig. 10.17, and essentially confirmed the determination of atacamite as the overall patina component of the Osiris. Strong peaks at 3450.7 and 3362.4 cm21 and the sextet of peaks from 989.8 to 850.4 cm21 are quite diagnostic for atacamite. Examination of the reverse side of the phiale fragment (Fig. 10.18) that gave the ESEM results shown in Figs. 10.4 – 10.6, and the XRD data shown in Figs. 10.15 and 10.16, resulted in the determination of malachite as the patina constituent, which was later confirmed by in situ XRD studies of the reverse of the phiale fragment. The technique of in situ FTIR microscopy is a very useful confirmatory method for the XRD results, also obtained in situ, although the XRD using the Go¨bel mirrors is not so affected by surface roughness, since there is no contact, in this case, between the instrument components and the sample, and only the XRD technique could satisfactorily characterize the light blue corrosion crust as chalconatronite. All of these FTIR analyses were performed using the 15 £ Reflachromat objective, with
487
D.A. Scott
Fig. 10.16. XRD in situ using the GADDS microdiffraction equipment for the corroded phiale fragment from the site of Francavilla Marittima, Southern Italy, dated to the 6th–5th century BC , showing the patina in this area to be a mixture of atacamite and clinoatacamite to be of atacamite.
sample analysis areas apertured to 100 £ 100 mm. The collected spectra are the sum of 200 scans, at a resolution of 4 cm21. 10.8
CONCLUSIONS
The availability of a number of different totally non-destructive methodologies for the investigation of copper alloy patinas and corrosion products has been demonstrated by the studies used as examples of the different techniques described here. There is now considerable scope, not available even 10 years ago, to carry out detailed studies of patina components on Museum objects without the removal of a sample from the object. This is an important advance in the study of ancient and historic artefacts, and will greatly extend and expand our knowledge of copper alloy patinas in the research that can be carried out in the years ahead. A recent review of some aspects of the study of copper alloy patinas and corrosion has been published by Scott [20], which should be consulted for further literature on this
488
The non-destructive investigation of copper alloy patinas
Fig. 10.17. FTIR Spectrum obtained, in situ, from the dark green surface of the Egyptian bronze Osiris shown in Fig. 10.11. The patina was shown to be of atacamite, but the result was only obtained with some difficulty and the chalconatronite patina could not be identified by this technique without sampling.
diverse topic. Some of the analytical advances described in the present volume are especially exciting, particularly RS, the depth profiling of copper patinas using XRD with Go¨bel mirrors, and the use of the ESEM, which allows patinas to be observed directly, without coating of the object. It is important to attack the problem of identification of patina components and morphology from a number of different perspectives and to use complementary methods of examination wherever possible, in order to assemble as complete a picture as possible of the patina under study. Reliance on only one particular analytical method cannot be recommended as the desired approach to the study of ancient art objects, particularly when complex surface characteristics have to be identified and described; this approach tends to emphasize only what the individual technique in question is capable of achieving, which often loses sight of the importance of a thorough study of the object itself, including its context, history, conservation, and detailed scientific investigation.
489
D.A. Scott
Fig. 10.18. FTIR Spectrum obtained, in situ, from the reverse side of the Phiale fragment examined in this chapter. The result of the FTIR examination surprisingly shows the presence of malachite, which had not been observed in the XRD study.
Acknowledgements Thanks are due to Dr Bruce Zuckerman, University of Southern California, for allowing photographs of the Greek copper plaque to be reproduced; to Dr Lynn Schwartz-Dodd of the University of Southern California for allowing work on the Egyptian bronze Osiris, to David Carson and Herant Khanjian, Department of Science, Getty Conservation Institute, for arranging the ESEM examination, and for carrying out the FTIR examination; to Jane Bassett, of the Department of Decorative Arts and Sculpture Conservation, J. Paul Getty Museum, for access to the Boizot sculpture. REFERENCES 1 2
490
R. Kerr, Chinese Bronzes. Victoria and Albert Museum, London, 1990, 112 pp. M. Sage, Observations sur la mine rouge de Cuivre, Observations sur la Physique, sur L’histoire Naturelle et sur les Arts, 14 (1779) 155 –157.
The non-destructive investigation of copper alloy patinas 3
4 5
6
7
8
9
10
11 12
13 14 15 16
17
J. Davy, Observations on the changes which have taken place in some ancient alloys of copper. Pt. 2, Philosophical Transactions of the Royal Society of London, 116 (1826) 55 –59. S.P. Best, R.J.H. Clark and R. Withnall, Non-destructive pigment analysis of artefacts by Raman microscopy, Endeavour, 16 (1992) 66–73. L. Burgio, D.A. Ciomartin and R.J.H. Clark, Pigment identification on mediaeval manuscripts, paintings and other artefacts by Raman microscopy: applications to the study of three German manuscripts, J. Mol. Struct., 405 (1997) 1 –11. R.A. David, H.G.M. Edwards, D.W. Farwell and D.L.A. DeFaria, Raman spectroscopic analysis of ancient Egyptian pigments, Archaeometry, 43 (2001) 461 –473. D.A. Scott and R. Woodard, Technical examination of an inscribed Greek copper plaque. In: I.D. MacLeod, S. Pennec and J. Theile (Eds.), Metal 2000, ICOM-CC Metals Conference, Santiago de Chile, 2002, in preparation. D. Stulik and E. Doehne, Applications of environmental scanning electron microscopy in art conservation and archaeology. In: P.B. Vandiver, J. Druzik and G.S. Wheeler (Eds.), Materials Issues in Art and Archaeology II. Materials Research Society, Pittsburg, 1991, pp. 23–29. H.E.A. Dalrymple, Seventeenth century dance costume: results of analysis. In: T. Bryce and J. Tate (Eds.), The Laboratories of the National Museum of Antiquities of Scotland. National Museum of Antiquities of Scotland, Scotland, 1984, pp. 111–117, 2 [n.d.]. C. McCabe and L.D. Glinsman, Understanding Alfred Stieglitz’ platinum and paladium prints: examination by X-ray fluorescence spectrometry, Research Techniques in Photographic Conservation: Proceedings of the Conference in Copenhagen, 14 –19th May 1995, Vol. 3. Royal Danish Academy of Fine Arts. School of Conservation, Kobenhavn, 1996, pp. 1–40. A. von Bohlen and F. Meyer, Arsen und Blei im Geigenlack: zur Elementanalyse in kleisten Lackproben, Restauro, 102 (1996) 472 –478. R. Scott, N. Wood and R. Kerr, The development of Chinese overglaze enamels: part 1. In: P. Vincenzini (Ed.), The Ceramics Cultural Heritage: Proceedings of the International Symposium “The Ceramics Heritage” of the 8th CIMTECWorld Ceramics Congress and Forum on New Materials, Florence, Italy, June 28th –July 2nd 1994, TECHNA, Faenza, 1995, pp. 153 –160. J.P. Rama, Le bronze d’art et ses techniques. Editions H. Vial, Paris, 1988, 45 pp. J. Riederer, Metallanalysen von Statuetten der Wurzelbauer-Werkstatt in Nurnberg, Berliner Beitrage zur Archaometrie, 5 (1980) 43–49. F. Bewer and D.A. Scott, A bronze sculpture attributed to Louis-Simon Boizot and platinum coating methods, Archaeometry, 37 (1995) 351 –361. M. Schreiner and O. Wachter, Ein erwitertes System der Rontgenfluoreszenzanlyse zur zerstorungsfreien Analse von Kunstwerken und Anwendung bei tintenfrassgeschadigten Papierobjekten. In: M. Koller and R. Prandtstetten (Eds.), Zum Thema Papier und Graphik. Verlag Mayer and Co, Vienna, 1994, pp. 43–56. Kevex Corporation, Omicron Spectrometer User’s Guide. Fisons Instruments Manufacturing Inc, Valencia, CA, 1996.
491
D.A. Scott 18 19 20 21
492
D.A. Scott, The application of scanning X-ray fluorescence microanalysis in the examination of cultural materials, Archaeometry, 43 (2001) 475 –482. D.A. Scott, A review of copper chlorides and related salts in bronze corrosion and as painting pigments, Stud. Conserv., 45 (2000) 39– 54. D.A. Scott, Copper and Bronze in Art: Corrosion, Colorants, Conservation. Getty Publications, Los Angeles, 2002. L.R. Lee, Scientific methods for investigation of coin surfaces for conservation. In: D. Goodburn-Brown and J. Jones (Eds.), Look After the Pennies: Numismatics and Conservation in the 1990s. Archetype Publications, London, 1998, pp. 1–8.
Chapter 11
Precious metals artefacts G. Demortier
11.1
INTRODUCTION
Quantitative elemental analyses of archaeological materials are now used to explain the provenance of artefacts and understand the workmanship of ancient craftsmen. When applied to gold jewellery these analytical methods should be mainly non-destructive and able to give information on the chemical composition of different parts of the artefact without any sampling, even at a microscopic level. Particle-induced X-ray emission (PIXE) [1], particle-induced gamma-ray emission (PIGE) and nuclear reactions involving the detection of a charged particle are three methods (which are sometimes used simultaneously) that offer most of these qualities [2]. PIXE and PIGE may be applied to artefacts of large dimensions, kept at atmospheric pressure during the irradiation. If necessary, the investigations may be undertaken in a microprobe assembly (beam diameter down to 3 mm). Most of the studies on archaeological materials performed at LARN (Namur) during the last 20 years concern gold jewellery and mainly solders on gold jewellery artefacts, and more recently the depletion gilding performed by Mesoamerican goldsmiths. Additional information using methods allowing to perform analyses down to the trace level will also be presented briefly [3]. 11.2
NON-DESTRUCTIVE ANALYSIS OF GOLD JEWELLERY ITEMS
11.2.1 Contribution of atomic and nuclear (but non-radioactive) methods to the analysis of ancient gold jewellery items Where magnifiers and optical microscopes are indispensable and powerful instruments for the examination of gold jewellery items, non-destructive physical methods of analysis of the chemical elements are also useful to provide information about the actual composition of the objects. Both types of Comprehensive Analytical Chemistry XLII Janssens and Van Grieken (Eds.) q 2004 Elsevier B.V. All rights reserved
493
G. Demortier
observations feature a valuable complementarity when they apply to the same size of regions under investigation. Topographical analyses using particle irradiation of narrow surfaces on artefacts are now possible on areas down to several thousandths of a square millimetre, on objects whose size may be up to 30 cm in each direction of space. This irradiation must be done in vacuum. In usual cases, numerous analyses on several surfaces of about 1 mm2 of one single object may provide enough information on the materials used to compose a jewel. In-air (also called external beam-) particle-induced X-ray emission (PIXE) has been successfully applied to numerous jewellery items. The incident particle is very often a proton. In this non-vacuum milliprobe arrangement (Fig. 11.1) the incident protons’ (2.5 – 3 MeV) beam crosses a thin foil before reaching the sample situated at a distance of less than 1 cm in air. The diameter of the proton beam is about 700 mm at the target position. For the analysis of Au artefacts, X-ray analysis is performed using a foil of zinc (20– 40 mm thick), which is a selective absorber of gold X-rays, inserted between the target and the detector in order to reduce the mean counting rate; thus the contribution of the most abundant information from gold is reduced. This procedure enhances the sensitivity for the detection of Cu, Ag and other chemical elements less abundant in a gold jewellery artefact. Figure 11.2 illustrates the effect of the Zn absorber on the ratios of detected X-rays. In a Au matrix, the thickness of the analysed surface layer is less than 10 mm. For gold items this thickness is generally representative for the bulk material as demonstrated on several items where scratchings have been made and measurements have been repeated.
Fig. 11.1. The non-vacuum PIXE set-up used at LARN.
494
Precious metals artefacts
Fig. 11.2. Effect of Zn filters of different thicknesses to improve the sensitivity for Cu, Ag and Cd analysis in a reference sample (75% Au, 6% Cu, 4% Zn, 4% Ag and 1% Cd).
The Si(Li) detector is set at 1358 with respect to the incident proton beam. The material of the carbon collimator (Fig. 11.1) plays the role of an X-ray absorber of all elements present on (and in) the Al foil. For the elemental analysis of materials whose components are heavier than Ti, the Si(Li) detector is positioned at about 20 cm from the sample. Sometimes a second Si(Li) detector is used to monitor the incident proton beam by detection of the Ar X-rays induced in the air in a narrow region (5 mm long) between the target and the sample. Another way to monitor the incident beam is to use a Ge(Li) detector collecting g-rays from the Al exit foil (883, 1013 and 1778 keV). The size and position of the filters may give rise to very different detected secondary X-rays. The filter is ideally put between the target and the detector on a collimator (1 cm thick aluminium drilled with a 4 mm hole) at mid-distance. This collimator limits the incident X-ray flux on
495
G. Demortier
the filter. A second collimator, in front of the detector, defines the solid angle of detection. Alloys of known composition are used as reference materials for the quantitative determination of elements. The full X-ray spectra containing the information on all the elements of interest in the material are recorded and then processed by computer in order to subtract the actual background and isolate the peaks of neighbouring energies. As the peak intensities are not simply related to the actual concentrations of the elements in the sample, the results obtained by this first treatment of the spectra are then computerized for the physical parameters governing the emission, the absorption in the sample itself and the efficiency of the detection process; the latter depends on the type of energy-dispersive X-ray detector employed [4]. 11.2.2 Illustration of the analytical performances of non-vacuum PIXE for gold artefacts Table 11.1 illustrates the ability of the full treatment for the analysis, by PIXE with 2.8 MeV protons, of two quaternary Au – Cu – Ag – Cd samples for which the actual concentrations are known. The reference material contains
TABLE 11.1 Matrix effects and secondary fluorescent yield corrections on the quantitative analysis of a sample whose concentrations are very different from those of the reference material, using PIXE with 2.8 MeV protons Character Reference material Analyzed sample Claimed concentration Measured without account of matrix effects Difference (in %) (absolute) Difference (in %) (relative) Measured values using matrix effect Without secondary fluorescence Difference With secondary fluorescence Difference Absolute errors from statistics Relative difference with the claimed values (in %)
496
Au (%) Cu (%) Ag (%) Cd (%) 25.00
31.12
27.08
16.80
75.9 78.35 þ 2.45 þ 4.5
12.1 11.32 2 0.78 2 7.5
6.1 4.96 21.14 29.3
5.9 5.37 2 0.53 9.0
74.50 2 1.40 75.15 2 0.75 0.15 1
13.48 þ1.38 12.66 þ0.56 0.15 4.6
5.84 20.26 5.92 20.18 0.07 2.9
6.18 þ 0.28 6.27 þ 0.37 0.10 6.2
Precious metals artefacts
Au, Cu, Ag and Cd in concentrations that are very different from those of the second alloy. The relative concentrations of each element in the samples are different by a factor of 3 or more. Various corrections for matrix effects are necessary to achieve a sufficient accuracy for the determination of the concentration of all the elements. When these corrections for matrix effects have been done, the computation of secondary fluorescence still increases the accuracy, but concerns the copper concentration only. In conclusion, even with reference materials different by about 300%, the complete analysis of the X-ray spectra provides a result whose accuracy is in the range of 5% or better. This property of PIXE cannot be reached with most non-destructive methods of analysis. Gold coins of high fineness are expected to be homogeneous materials and the surface concentration determined by PIXE (characteristic depth of about 7 mm for 2.8 MeV protons) has been compared with the analysis of the bulk by proton activation using 12 MeV incident protons; these particles may probe the samples down to 400 mm. Table 11.2 illustrates the comparison (by both methods) on coins of Roman and Byzantine emperors.
TABLE 11.2 Analysis of gold coins (Roman and Byzantine emperors) Gold coins
Arcadius (a) Leo (b) Zeno (c) Zeno (b) Anastasius (b) Justinien (d) Maurice Tibere Phocas (c) Heraclius (b) Constantin IV (c) Constantin VII Michel IV (c)
PIXE in non-vacuum
Proton activation analysis
Cu
Ag
Au
Number of measurements
Cu
Ag
Au
0.32 0.27 0.37 0.40 0.39 0.43 0.25 0.43 0.30 0.33 0.31 1.75
1.16 1.14 1.01 1.32 1.24 1.49 1.00 1.54 1.10 1.40 2.80 8.10
98.5 98.6 98.6 98.3 98.4 98.1 98.8 98.0 98.6 98.3 96.9 90.1
(5) (4) (5) (4) (5) (5) (4) (4) (5) (2) (7) (12)
0.20
1.00
98.8
0.5 0.35 0.35 0.4 0.30 n.d. 0.41
1.40 1.35 1.30 1.60 1.50 n.d. 3.3
98.3 98.3 98.35 98.0 98.2 98.2 96.3
n.d.: not determined; (a) Semissis; (b) Tremissis; (c) solidus; (d) 1/2 solidus. All data for PIXE and PAA are mean values on several determinations; all concentrations are in wt%. PAA are taken from Refs. [5– 7].
497
G. Demortier
11.3 11.3.1
THE SOLDERING OF GOLD Ancient recipes for gold soldering
Two recipes for the preparation of gold solders are known from antiquity, one derived from “natural” chrysocolla (from the Greek krysoz: meaning gold and kolla: glue), the other obtained by mixing Cu, Ag and Au. As reported in the Natural History of Pliny (1st century AD ) “natural chrysocolla is an exudation found in the shafts … The colour of this natural product is yellow” [8]. The preparation of artificial chrysocolla, as also described in the Leyden Papyrus X (another old metallurgical handbook), includes “copper of Cyprus: 4 parts; assem: 2 parts; gold: 1 part.” The nature of the “assem” used in this preparation has not been certified yet. Amongst others, the term assem has been used in old manuscripts to describe any substance whose properties (colour, brightness, weight, etc.) counterfeit those of precious metals [9]. Another gold solder recipe known from the earliest times is “gold: 2 parts; copper: 1 part, melt and shake up. If one wants a brilliant colour melt with a little silver.” This last procedure provides an excellent brazing alloy very comparable to those used by goldsmiths today. Since the second half of the 19th century, brazing alloys containing cadmium have been used extensively. A small amount of cadmium in a gold matrix is able to lower the melting point of the brazing alloy far below that obtained by introducing equivalent concentrations of copper, zinc or silver in a gold matrix. An extended discussion of different techniques of soldering in ancient jewellery has been published by Thouvenin [10]. In addition to these discussions and analyses, Thouvenin himself reproduced several pieces of jewellery of ancient workmanship. He concluded that the goldsmiths of antiquity had skilfully created very fine and tiny solders, not by using brazing alloys (a technique only useful for “macrosolders”) but by diffusion at about 9008C of a very fine powder of deoxidized copper, giving rise to in situ binary or ternary alloys at “microsites.” The soldering process of fine granulations was achieved in antiquity (e.g., in Etruscan jewellery) by a process of diffusion bonding rediscovered by Littledale in 1936 at the British Museum [11]. In the next section, we will review several ways to understand the procedures used by ancient goldsmiths to realise tiny solders on gold jewellery items by using only the results of quantitative analysis of Cu and Ag in typical regions of solderings. We will also comment on the possible use of cadmium ores (mainly greenockite) in ancient times. This is already a subject of current debate in archaeological journals [12,13].
498
Precious metals artefacts
11.3.2
Iranian goldsmithery from the 4th century
BC
The high level of workmanship of ancient goldsmiths can hardly be equalled even today. Among the items of Iranian and Syrian jewellery that we have studied, the wonderful Achemenide pendant of the 4th century BC , which belongs to the Department of Iranian Antiquities of the Muse´e du Louvre, clearly illustrates the skill of ancient Iranian goldsmiths. The pendant is shown in Fig. 11.3. The total width of the disk decorated in repousse´ is 5.5 cm. The surrounding ornaments were made with 28 identical motifs. Each of them includes a hollow cylinder and two hemispherical caps. The detail of the top right part (Fig. 11.4) clearly shows the regions of solders. These areas were non-destructively analysed for composition, and the results are summarized in Table 11.3 and Fig. 11.5. All measurements have been performed with PIXE in a microprobe assembly [14]. The PIXE microprobe facility of the LARN has been used to irradiate small regions (15 mm wide) of the sample put in a vacuum chamber. The Achemenide pendant was fixed on a X – Y frame that can be moved in the proton beam by means of computer-controlled stepping motors (with a reproducibility of better than 2.5 microns after a translation of several centimetres). The same computer also coordinates to the collection of data from the photon and electron detectors. Electron detectors allow us to obtain a simultaneous image of the irradiated region. Four joins were encountered in regions 2 – 8 of Fig. 11.4, over a distance extending less than 5 mm. The elemental compositions at solder sites, 2, 4, 6 and 8 show that three different joining procedures were performed. At site 6,
Fig. 11.3. The Achemenide pendant. Muse´e du Louvre (AO 3171).
499
G. Demortier
Fig. 11.4. Detail of the top right part of the Achemenide pendant.
the gold content is significantly greater than in the neighbouring regions and indicates that the join was made by welding without adding any external material. The end of the cylinder and the bottom of the first hemispherical cap were simultaneously heated to an early stage of fusion. In this process, metals like copper and silver were selectively eliminated, mainly by oxidation, so that the gold content was enhanced. The temperature was around 10008C. TABLE 11.3 Composition of different regions of the Achemenide pendant [14] Region
Cu (%)
Ag (%)
Au (%)
Description
1 2 3 4 5 6 7 8 9
5.6 ^ 0.3 " 9.8 ^ 0.6 5.2 ^ 0.3 " 7.5 ^ 0.5 5.7 ^ 0.4 # 3.1 ^ 0.3 7.0 ^ 0.4 " 12.7 ^ 0.7 7.2 ^ 0.4
11.6 ^ 0.6 " 15.6 ^ 0.8 12.6 ^ 0.7 " 13.9 ^ 0.8 11.7 ^ 0.7 6.8 ^ 0.3 16.6 ^ 0.9 # 14.6 ^ 0.8 15.6 ^ 0.9
82.8 ^ 1.2 # 74.6 ^ 1.5 82.2 ^ 1.0 # 78.6 ^ 1.1 82.6 ^ 0.9 " 90.1 ^ 1.5 76.4 ^ 1.3 # 72.7 ^ 1.3 77.2 ^ 1.4
Repousse´ area Brazing Vertical sheet Brazing Cylinder Local melting Internal hemisphere Copper diffusion External hemisphere
The arrows indicate increased (or decreased) concentrations at solders. Impact regions are given in Fig. 11.4.
500
Precious metals artefacts
Fig. 11.5. Scan over 6 mm using a microbeam (15 mm wide) in the top right part of the pendant.
At site 8, an increase in copper concentration was observed simultaneously with a decrease in gold and silver concentrations. The two caps, previously well fitted, were joined by a process known as solid-state diffusion bonding with copper salts in a reducing atmosphere. The process took place at about 8908C [10,11]. This relatively low temperature allowed to make a join at site 8 without desoldering the join at site 6, only 1.5 mm away. This joining method was widely used by the Etruscans in their famous granulation work. An increase in copper and silver and a decrease in gold were simultaneously observed at sites 2 and 4, a result indicating that at both sites a brazing alloy has been used. The lower concentration of gold in region 2 indicated that the temperature of fusion of the alloy used at that site was lower than that at site 4 (possibly around 820 and 8608C, respectively). These temperatures are close to 8908C, the soldering temperature at site 8. However, the latter join did not desolder. The process of diffusion bonding produces a join that cannot be desoldered. Reheating after joining results in further copper diffusion. That diffusion induces a local decrease of the copper content with a consequent increase in the local melting temperature. The small temperature difference of sites 2 and 4 seems to indicate that the 28 attached elements were soldered to the gold sheet before the sheet was joined to the decorated disk, because the alloy used at site 4 had a higher melting point than that at site 2.
501
G. Demortier
Fig. 11.6. Scan over a few millimetre in the region of a suspension ring (modern addition?).
The high degree of workmanship displayed in a piece as complex as the Achemenide pendant shows that, 2500 years ago, the goldsmith recognized and used different temperatures for various joining processes. The regions 11 and 12 of solders between the suspension rings and the main body (Fig. 11.6) do not contain traces of metals: no gold, no silver and no copper. An organic glue could have been used here. These elements are of low atomic weight and are not detectable by PIXE. After these analyses the hypothesis that the four additional rings could be of modern origin was considered to be plausible by the curator of the Department of Oriental Antiquities of the Muse´e du Louvre. 11.3.3
Tartesic gold artefacts
The jewellery item (P1 of Fig. 11.7a) comes from a fortuitous discovery in the “El Pedroso” area (Sevilla, Spain). This small piece was found with another one, almost perfectly preserved [15], that helped the curators rebuild the original shape of P1 and date it in the Oriental period of the “Valle del Guadalquivir.” Because the sample does not have technical details, like filigrees or granulations characteristic of earlier periods, it can be dated to the beginning of that period, possibly in the 5th century BC . Each of the three elements of the P1 item is almost exclusively made by repousse´. The three parts seem to have been soldered together afterwards. This technique enables the modelling of the different elements around the central piece, the
502
Precious metals artefacts
Fig. 11.7. (a) Sample found in “El Pedroso,” 5th Century BC ; (b) bead of Ebora made by strips attached on a hollow cylinder. Arrow indicates that the analysis position corresponds to the symmetric region at the rear. Impacts 13, 14, 15 refer to the three points [ represented at the bottom left of part b; (c) biconical bead of Ebora (partially broken). Analysis positions indicated by an “i” refer to inner part of the artefact. Point 3 refers to the opposite; (d) details of a region of the granulations in the artefact of part (c). Brazing is clearly exhibited. The composition at the various positions is given in Table 11.4.
shape of which is of typical Oriental style. Although the different components were directly soldered to one another, the solder seems to have been reinforced by adding an inner wire to bind them. The other items (Fig. 11.7b – d) are from the “Ebora” treasure and they were found with a large group of Oriental jewels, during agricultural works in the area of “Sanlucar de Barrameda,” a village near Cadiz (Spain). These samples are conserved at the Archaeological Museum of Sevilla and have already been widely studied [16 – 18]. The composition of the alloy in various impact regions on the three jewellery artefacts is given in Table 11.4 [19]. Highly accurate measurements with a narrow beam (350 mm) have been made in order to be able to suggest an explanation of technological interest.
503
G. Demortier TABLE 11.4 Copper and silver contents (in wt%) at impacts of Fig. 11.7 [19] El Pedroso 1
Ebora 3
Ebora 5
Impact no.
Cu
Ag
Impact no.
Cu
Ag
Impact no.
Cu
Ag
1 2 3 4 5 !6 !7 !8 !9 ! 10 ! 11 ! 12 ! 13 ! 14 ! 15 16 17 18 19 20 21
3.1 2.9 2.3 3.2 2.4 3.0 6.6 4.3 4.2 4.0 5.2 4.7 3.6 3.9 3.5 2.6 2.8 2.6 2.9 3.1 3.2
18.5 19.4 18.5 18.7 17.5 18.0 16.4 22.0 18.7 15.5 16.4 16.1 15.7 19.5 15.4 17.0 19.1 17.9 19.9 18.6 19.4
1 2 !3 4 5 6(0) 7 8 9 10(0) ! 11(0) ! 12 ! 13 ! 14 ! 15 ! 16 17 18 – – –
8.1 7.5 8.2 7.6 6.4 7.5 7.1 7.2 7.1 7.5 6.5 6.3 6.6 6.0 7.2 7.2 7.5 7.5 – – –
8.1 7.4 8.8 7.8 7.3 7.9 7.8 7.8 7.6 7.4 7.3 8.2 8.6 8.8 8.1 8.3 6.8 8.4 – – –
1 2 3(0) 4 5 6(i) 7(i) 8(i) 9(i) 10(i) ! 11(g) ! 12(g) ! 13(g) ! 14(g) ! 15(g) – – – – – –
2.3 3.5 2.0 2.4 2.2 4.5 4.4 4.3 4.5 3.7 3.2 2.9 3.2 3.0 3.0 – – – – – –
14.6 16.8 13.5 16.0 13.3 20.5 20.4 20.1 17.6 20.1 14.8 14.7 14.5 15.1 16.0 – – – – – –
The balance to give 100% concerns the gold content. ! refers to soldering regions; (i) inner part of the artefact; (g) grazing incidence of incident proton beam.
The maximum statistical error on the reported data is ^1% for Cu, ^3% for Ag and ^0.5% for Au. Special attention has been paid to regions of expected solders (indicated by an arrow in the table). It is clear that the concentration of Cu and Ag in those regions is very close to that measured outside the solders. To extract a more pertinent interpretation, compositional data are displayed in the classical ternary diagrams of Fig. 11.8a – c. For sample P1, the results outside the solders indicate that the relative Ag/Cu concentration ratio is 6.5, with gold contents ranging from 77 up to 81%. This range of variation clearly indicates that the material (a natural electrum) is not fully homogeneous: the reported values vary by much more than the expected statistical error. In the solder regions, the Cu content is systematically higher and the results are largely scattered. This situation may be explained
504
Precious metals artefacts
Fig. 11.8. (a) Ternary compositional diagram of sample P1. Small dots refer to analysis positions in the bulk; large dots refer to soldering regions. (b) Ternary compositional diagram of sample E3. Dots refer to analysis positions on the strips outside solders, “open circles” refer to regions of the binding of strips to bottom and upper rings, the star refers to the soldering of the outer horizontal strip. (c) Ternary compositional diagram of sample E5. Small dots refer to the outside bulk, “open circles” to inside bulk, large dots to granulations.
by a brazing procedure to solder the elements together: a liquid alloy with a temperature of fusion of 50 – 808C below the melting point of the bulk material was used. The scattering of the results in the ternary diagram is partially due to the width of the incident proton beam which may extend beyond the soldering regions. In sample E3, points are closely distributed around 84.5% Au, 8% Ag and 7.5% Cu. A closer look indicates that at impacts outside the solders (dashed area in the ternary diagram) the copper concentration is a bit higher than in almost all soldering regions. The lower Cu concentrations at solders may be understood if we consider that some forging procedure has been used to bind the strips together. No additional material is used: parts of the item to be
505
G. Demortier
bound are simply locally heated at a temperature slightly higher than the solidus temperature. The metal is partially fused and copper, the less noble metal in the alloy, is selectively eliminated by forming oxides that are lost during the final procedure of brushing. At point 3, both copper and silver are slightly more abundant than on the other locations; probably a brazing procedure has been used to bind both ends of the central circular ring. The central ring is then soldered by a brazing procedure while the strips on the bottom and upper rings are bound by forging. For sample E5, the experimental values may be distributed in three groups: (a) outside any solder in the outer part of the item, (b) outside any solder in the inner part the item and (c) into the granulations. The gold concentration inside is lower than outside. Furthermore, if we compute the Ag/Cu ratio, we obtain 5 at locations inside the hollow sphere and 6.1 outside. These observations may be interpreted by some enhancement of both Au and Ag at the external surface produced, either by an original surface treatment in order to increase the golden appearance or a chemical treatment to clean (after the excavation) the external surface and to give the object a “better look.” Eliminating the less noble metals was the aim of this cleaning process. In granulations of dimension 200 mm, i.e., lower than the beam size (350 mm), we observe that some brazing procedure involving the melting of a fusible alloy containing more copper was used. The embedded granulations are clearly shown in the micrograph of Fig. 11.7d. Other Tartesic items have also been studied [20] and the conclusion was that the soldering procedures were forging (or casting) and brazing, but there is no indication of solid-state bonding as is the case in Etruscan jewellery. 11.3.4
Later Iranian goldsmithery
In any human activity, alternate means are sought to achieve a given end. Early goldsmiths must have tested and used a number of different joining materials and methods. Among Iranian and Syrian items from the 1st to the 9th century AD , we have occasionally found significant amounts of cadmium in solders. For most people concerned with archaeological jewellery, the presence of cadmium suggests forgery or modern repair [12,13]. However, in several investigations conducted since 1983, we have shown that ancient solders containing cadmium may sometimes be differentiated from modern solders [21,22]. Brazing alloys that contain cadmium have been used extensively since mid-19th century. A small amount of cadmium in a gold matrix can lower the melting point below that obtained by adding equivalent concentrations of
506
Precious metals artefacts
copper silver (or) to the same gold matrix. It is commonly believed that cadmium or cadmium compounds were not used in antiquity (cadmium ores are indeed rare in nature), and that objects containing cadmium, either in the solder or as an impurity in the gold alloy, were recently manufactured or restored. Nevertheless, cadmium was detected in a solder on a monetary medallion excavated several years ago in Houmeau, France, by an official staff of French archaeologists [23]. Several experts in ancient jewellery question the authenticity of the gold jewellery that we analysed because the artefacts were not found during official excavations and several of the artefacts have been declared by us as fakes or recent restorations on the basis of our analyses and our criteria of authentification. Although it is usually present at lower concentrations, cadmium is sometimes present in concentrations up to several percent in regions where soldering was necessary to join elements in pieces of apparently ancient jewellery. We have shown [2,21] that the amount of cadmium found in such jewellery (generally from Iran, Syria or Southern Italy) is related to the amounts of copper and silver present, but in completely different proportions from those observed in modern soldering alloys. Analytical results [2,21,22,24] collected during analyses of numerous artefacts, from museums as well as from private owners, have led us to propose several criteria useful for distinguishing modern joining procedures from ancient that used cadmium-based materials, and to indicate a new interpretation of old metallurgical descriptions. We suggest that ancient Iranian goldsmiths smelted greenockite (a natural yellow cadmium ore) simultaneously with copper ores and gold to obtain an alloy suitable as a solder at low temperatures. If several closely spaced joins are present on a piece of jewellery, solders with different melting points may be necessary, depending on the order of construction. Modern commercial solders are characterized by compositions in which silver and copper concentrations are retained in a constant proportion (direct correlation) during the soldering procedure (Fig. 11.9). Metallic cadmium introduced by modern goldsmiths in their basic alloy results in an alloy with a low melting point. When this alloy is used for brazing, the highly volatile cadmium is eliminated. On the other hand, for ancient solders, a strong correlation exists between copper and cadmium concentrations (Fig. 11.10). The main criterion to distinguish modern from ancient solders containing cadmium applies only to jewellery artefacts that show several regions of solders. The artefact is considered to be a forgery or a restored piece if these relative concentrations are correlated like those in Fig. 11.9. No criterion for authenticity is available if the number of solders is not sufficient to draw a line across the points. The detection of cadmium in one
507
G. Demortier
Fig. 11.9. Ternary composition diagram for Cu–Ag –Cd in modern soldering alloys. Analytical results indicate that when the cadmium amount varies, the ratio of copper on silver remains constant.
single isolated region of an artefact cannot be used to reject or attest the authenticity. Other criteria may be used to distinguish ancient from modern cadmiumcontaining solders. These criteria include: (a) The frequent presence of iron in ancient solders. Modern brazing alloys (for fine gold jewellery) sold by precious metal suppliers or made by jewellery craftsmen in their own shops contain only gold, silver, copper, and sometimes zinc and cadmium. These metals are alloyed from pure metal ingots, whereas ancient solders were made with rough materials, generally by direct smelting of ores like chalcopyrite (an iron – copper ore with a colour close to that of gold). Iron is therefore expected to be present in ancient artefacts but not in modern jewellery. (b) Luminescence can be induced by the proton beam used in PIXE – PIGE measurements on ancient solders. This effect is probably due to the
508
Precious metals artefacts
Fig. 11.10. Ternary composition diagram for Cu –Ag –Cd in regions of solders on an Iranian necklace of the 1st century AD (see Ref. [2] for details). Here, the copper concentration increases with the cadmium concentration.
presence of microinclusions of slag retained by the metal produced by a crude smelting technique. This optical luminescence is only induced in insulators, not in metals, and does not appear in modern goldsmithery in which only metals are involved. (c) Differences in relative concentrations of zinc and cadmium [25]. (d) The presence of sulphur [26]. 11.3.5
Preparations of low-melting brazing alloys
Early attempts to prepare low-melting alloys with cadmium sulphide (its natural form, greenockite, is orange-yellow) encouraged further investigation to check whether this process might have been used in antiquity.
509
G. Demortier
Our results [21,22] indicate the production of brazing alloys by a procedure of “soft metallurgy,” which can give a soldering whose composition can be more easily controlled than that frequently used by modern jewellers. A few milligrams of gold was melted to form a small sphere. Powdered cadmium sulphide was then poured onto the melted gold. The powder dissolved rapidly and gave a small sphere of alloy that was then analysed microscopically. Each sphere of alloy appeared approximately homogeneous. The centres of the spheres contained less cadmium than the surfaces, but the difference in concentration between the centre and the surface was less than 20%. Depending on several conditions (i.e., temperature, crucible shape, addition of copper and iron ores), the solubility of cadmium from cadmium sulphide reached saturation when the alloyed cadmium content was between 1 and 10%. In all cases, the addition of copper or copper ores to molten gold enhanced the solubility of cadmium. This observation leads to a direct correlation between the copper and cadmium contents. When powdered cadmium sulphide is mixed with chips of gold and the mixture is heated to melting, the dissolution takes longer than when greenockite is added to liquid gold. Furthermore, the alloy obtained by this last procedure is less homogeneous than the alloy made with cadmium sulphide poured onto the molten gold. The sphere of gold– cadmium alloy sometimes appeared to consist of two different parts: a gold-rich alloy formed at the bottom of the crucible and a black coating that appeared at the top (Fig. 11.11). Cadmium and copper correlations were observed in the gold-rich material as well as in the black earthy coating (Fig. 11.12 and Table 11.5). Cadmium sulphide was used in this work as part of an archaeological argument: the mixture by ancients of materials of nearly the same colour. Direct alloying of metallic cadmium with melted gold, a modern procedure, results in a violent interaction when the metallic cadmium is placed in the melt because the boiling point of cadmium (7658C) is far below the melting point of gold (10638C). The melting point of cadmium sulphide is so high (above 17508C) that its incorporation in the molten gold is not violent. No cadmium vapour is produced when the amount of greenockite added to gold is below the saturation concentration. Above this limit, wisps with the colour of cadmium sulphide appear floating above the crucible. The application of the four criteria given above is illustrated in the following study of a hollow gold pearl (Fig. 11.13) found in Syria along with other more prestigious objects (necklaces, pendants, phials). Most of these artefacts were found during excavations in Hauran for the draining of Lake Orantes in the 1950s. Some of this jewellery was acquired by museums, such as the Staatliche Museen in Berlin. Several pieces of the
510
Precious metals artefacts
Fig. 11.11. Small sphere of a soldering alloy introducing cadmium sulphide in gold – silver–copper alloy as made by metal workers in antiquity. (a) Just after melting; please note the black deposit around the cadmium gold alloy; (b) after a gentle compression. ( For a colored version of this figure, see Plate 11.I.)
treasure, first considered as authentic by curators and archaeologists, have then been classified as fakes or restored items after cadmium had been detected. An electron micrograph (Fig. 11.14) of a region near two soldered granules clearly shows that a brazing procedure has been used to join both granules to a thin gold base that was then soldered to the main hollow sphere. Scans across large areas were performed with a proton probe (50 mm wide); narrower scans in characteristic regions were also obtained. Maps of copper, gold, silver and cadmium (Fig. 11.15) clearly indicate the presence of cadmium only in solders between the granules and between the main hollow sphere and each granule. The local decrease of silver and gold at each site
511
G. Demortier
Fig. 11.12. Ternary compositional diagram of this reproduced “ancient” soldering alloy showing the correlation of cadmium and copper concentrations (similar to the distribution of results of Fig. 11.10).
where cadmium was present, and the increase of copper at that site, point out a direct correlation between copper and cadmium contents. An additional solder characterization was accomplished by using the selective excitation of Zn K X-rays without excitation of gold L X-rays. It was found that zinc is also present, at trace levels (about 300 ppm) [25]. Unresolvable interferences arise when PIXE is used to measure very low concentrations of Zn in Au: Ka and Kb lines of Zn interfere with Ll and La lines of Au, respectively (see also: differential PIXE). Due to these interferences, the limits of detection for Zn in Au-rich matrices cannot be lowered below 200 ppm with PIXE. Selective X-ray fluorescence induced by X-rays produced by proton bombardment of a primary target is expected to solve problems of sensitivity for Zn at the trace level. Both Ka and Kb energies of Ge are lower than the three L absorption edges of Au, but they are situated just above the K absorption edge of Zn. Thus, the cross-section for fluorescence in Zn is extremely favourable while
512
Precious metals artefacts TABLE 11.5 Analytical results from different areas of a gold solder obtained by dissolving CdS in a liquid gold–copper–silver alloy Element concentration (%)
Basic gold Gold-rich regions of soldera
“Argenteous” regions of soldersa
Dust
Cu
Ag
Cd
Au
12.05 4.85 5.20 4.95 5.00 7.55 7.20 9.20 16.40 16.15 19.40 25.05 29.35 26.80 29.50 32.80
16.35 14.20 14.50 15.30 13.95 12.55 14.70 15.65 12.30 14.15 14.75 12.25 11.45 16.70 10.50 13.95
– 8.90 8.55 8.45 9.15 12.95 10.75 14.25 43.20 26.85 28.55 45.35 46.70 21.40 50.45 48.75
71.60 72.05 71.75 71.30 71.90 66.95 67.35 60.90 28.10 42.85 37.30 17.35 12.50 35.10 9.65 4.50
a
See also the colour plate at the end of the book.
L X-rays of Au are not produced. Rayleigh and Compton scattered primary Ge X-rays (also present in the detected X-ray spectrum) are used to monitor the measurements for samples of similar compositions. The analytical procedure (Fig. 11.16) has been optimized on the basis of the following arguments: (a) Proton irradiation of a very pure target (such as Ge) in order to avoid Rayleigh and Compton scattering of primary X-rays generated from impurities in the irradiated target. K X-rays of As would be optimal for Zn determinations but ultrapure targets of As are not easily available. (b) Irradiation of the primary target in vacuum. The material surrounding the Ge target is pure graphite, in order to avoid as much as possible the production of photons giving rise to a continuum background in the final X-ray spectrum. (c) The presence of a thin window of mylar between the evacuated beam pipe and the region of the sample (at atmospheric pressure) to stop scattered protons.
513
G. Demortier
Fig. 11.13. Hollow pearl of the 6th century AD excavated with other more prestigious objects of Byzantine jewellery. Total width: 8.5 mm.
(d) The use of a tube collimator between the Ge target and the sample made of pure Au. This is a 1 cm long cylinder; the size of the collimated beam may vary from 0.5 to 5 mm in diameter depending on the need for localized analysis. (e) The use of an open geometry between the sample and the solid-state detector. The collimator between the sample and the detector is pure Si. (f) The use of an Si(Li) and not Ge detector, since the latter is too sensitive to g-rays. (g) The use of the minimum proton energy (compatible with time of irradiations and desired sensitivity) to avoid neutron and g-ray emission as much as possible. The intensity of primary X-rays can be regulated by the intensity of the incident beam (several mA) of 2.2 MeV protons. Cooling of the Ge target is not necessary. This process induced by proton irradiation is far better than that induced by electron bombardment. Electron irradiation of a Ge target would give rise to a bremsstrahlung continuum extending beyond the Kb 514
Precious metals artefacts
Fig. 11.14. Electron micrograph of a detail on the pearl in Fig. 11.13. Brazing procedure involving a liquid phase is clearly visible.
line of Ge, and those irradiations would then excite L X-rays of gold which are undesirable. With this experimental arrangement, the limit of detection of Zn in gold may be lowered down to 50 ppm. The possibility to determine low
Fig. 11.15. Maps of Cu, Ag, Au and Cd obtained by PIXE microprobe analysis of the hollow pearl of Fig. 11.13. Correlation of copper and cadmium concentrations is obvious.
515
G. Demortier
Fig. 11.16. The experimental arrangement for the detection of zinc in narrow regions of gold by X-ray fluorescence induced by Ge lines generated by proton bombardment.
concentrations of zinc in solders of ancient gold jewellery items is extremely useful for the detection of fakes and/or repairs. Sulphur and zinc determinations may be added to the main criteria of characterization of ancient solders: the copper– cadmium correlation and the presence of iron to give a total of four differences between solders on the ancient hollow pearl and solders made with commercial brazing alloys containing cadmium. Sulphur at a 30 ppm level in a region where cadmium is present at 3% indicates that if CdS has been used as alloying material with gold, at least 99% of the sulphur would have been eliminated during the alloying process. Fine modern cadmium-based brazing alloys are zinc-free or contain zinc and cadmium in a ratio of about 1:5. When Cu –Ag – Zn –Cd – Au alloys are melted to make solders, cadmium is selectively eliminated because of its high volatility. Then the final relative concentration of zinc to cadmium in modern joins is still zero or higher than 1/5. In the hollow pearl (and also in other jewellery), this ratio is situated between 1/25 and 1/300. The range of these ratios is either too low to be attributed to the use of modern alloys containing zinc or too high to be interpreted as impurities in metals used for alloying. These amounts of zinc in ancient jewellery items are understandable because greenockite (a natural cadmium ore) appears as a yellow coating on zinc blendes. The natural content in ancient soldering processes is attributed to the imperfect procedure of separation of greenockite from the zinc blende. The use of a CdS ore as additional ingredient to reduce the melting point of gold alloys may be checked by the measurement of residual sulphur. This was realized by using a nuclear reaction induced by a deuteron beam [26].
516
Precious metals artefacts
11.3.6
A new reading of Elder Pliny’s Natural History
Early artisans used several recipes to prepare gold solders These recipes included natural chrysocolla as well as other compositions prepared from alloys containing metals such as copper, silver and gold. The use of chrysocolla was described with many details in the 1st century AD by Pliny the Elder in his Natural History. As can be expected from its etymological origin, the name chrysocolla meant, in antiquity, any material used to join parts of gold jewellery. At the beginning of his 33rd volume, Pliny wrote: “Gold is dug out and, with it, also chrysocolla, which continues to bear a name derived from the term gold in order that its value may appear greater.” This special attention to chrysocolla indicates its importance in ancient goldsmithery. None of the original manuscripts of Pliny is now available. The modern translation of Pliny’s book often refers to the Bambergensis manuscript (10th century) and less often to the Parisinus Latinus version (12th century). The two versions are very similar, but several terms are different. All comments in the English [27] and French [28] translations refer to the Bambergensis manuscript, for which the English chemist K.C. Bailey gave technical comments on all scientific subjects [29]. The 26th paragraph of the 33rd book in both versions (Bambergensis and Parisinus Latinus) describes the origin of chrysocolla: Gold solder is a liquid found in the shafts we spoke of, flowing down along a vein of gold, with a slime that is solidified by the cold of winter event to the hardness of pumice stone. A more highly spoken of variety of the same metal has been ascertained to be formed in copper mines, and the next best in silver mines. A less valuable sort also with an element of gold is also found in lead mines. In all these mines however an artificial variety is produced that is much inferior to the natural kind referred to: the method is to introduce a gentle flow of water into the vein all winter and go on till the beginning of June and then to dry it off in June and July, clearly showing that gold solder is nothing else than the putrefaction of a vein of metal.
The next sentence in the Bambergensis is, in the Latin version, “Nativa duritia maxime distat; uvam vocant,” which is translated as “Natural gold solder, known as grapes differs very greatly from the artificial in hardness” [27]. On the other hand, in the Parisinus Latinus one finds “Nativa duritia maxime distat, luteam vocant,” which is now translated as “Natural gold solder is the best as far as its hardness is concerned; it is called yellow chrysocolla.”
517
G. Demortier
The next sentence in both versions is “Et tamen illa quoque herba, quam lutum appellant, tingitur,” which is translated as “Furthermore, it is still dyed with a plant called yellow-weed.” If, in the translation of the Bambergensis, uvam refers to the shape of grapes, one cannot understand why Pliny’s text continues with Et tamen. This sentence refers to the colour of the material, and “Et tamen” does not make sense, if the reference to grapes concerns shape. This interpretation has led translators and chemical archaeologists [30] to identify chrysocolla with green malachite, which sometimes occurs as green “grapes.” In the Parisinus Latinus version, uvam becomes luteam, and luteam means yellow, yellow-like gold [30]. In the Bambergensis version, “uvam” can be connected not with the shape but with the colour of grapes, specifically the colour of Italian grapes that are golden yellow in full maturity. Chrysocolla is then yellow (like gold). Colour seems to be one of the reasons for the use of chrysocolla not only in gold jewellery, but also as a dye and in medicine. This use of a yellow plant to improve the yellow colour of an alloy is difficult to understand with our present knowledge of materials science. Even ancient goldsmiths had observed that vegetables submitted to temperature treatments would undergo complete change in their shape and colour. The usefulness of this yellow weed in the soldering procedure may probably be better understood in the context of the following paragraph of Pliny’s description of urine being added to a recipe of goldsmithery. The reasons for adding urine may have been: (a) to liberate carbon as a reductor; (b) to liberate nitrogen, which increases the solubility of cadmium from CdS in gold; (c) to emit a yellow substance in non-metallurgical applications (27th and 28th paragraphs of Pliny’s book). According to the foregoing interpretation, chrysocolla of ancient times is neither malachite nor a blue copper silicate (the mineral now called chrysocolla) but it is a yellow substance, possibly the yellow mineral cadmium sulphide, which appears as a coating on other minerals, chiefly zinc sulphide. This description fits Pliny’s text, which describes gold solder (chrysocolla) as a liquid that “flows from several mines to give a solid deposit.” The 29th paragraph of Pliny’s book deals with the use of chrysocolla by goldsmiths. In the Bambergensis version: “Chrysocollam et aurifices sibi vindicant adglutinando auro, et inde omnes appellatas similiter virentes dicunt,” or “The goldsmiths also use a special gold solder of their own for
518
Precious metals artefacts
soldering gold, and according to them it is from this that all the other substances with a similar green colour take the name.” The translation of “virentes” as green is the most evident, but “virentes” may also be understood not as a colour, but as the property of brightness that can be attributed to all metals that reflect light. In the Parisinus Latinus, this sentence is “Chrysocollam et aurifices sibi vindicant adglutinando auro: et inde omnes appellatam similiter utentes dicunt” or “The goldsmiths also use gold solder for soldering gold, and all the users (goldsmiths) call with the same name any substance suitable for this purpose.” This sentence gives no new information, but rather reminds us that the meaning of chrysocolla refers to its use as gold solder. On this use of chrysocolla by goldsmiths, both versions are identical at the end of the paragraph: Temperatur autem ea cypria aerugine, et pueri impubis urina, addito nitro. Teritur cyprio aere in cypriis mortariis: santernam vocant nostri. Ita ferruminatur aurum, quod argentosum vocant. Signumque est, si addita santerna nitescit. E diverso aerosum contrahit se, hebetaturque, et difficulter ferruminatur. Ad id glutinum fit, auro, et septima parte argenti ad supradicta additis, unaque contritis
which is translated as: They (goldsmiths) make the mixture with copper verdigris and with urine of a boy who has not yet reached puberty and some soda (sodium carbonate). It is ground with a copper pestle in a copper mortar. They call this preparation santerna. In this way they can solder argenteous gold. A sign of its having been so treated if the application of santerna gives a brilliant colour. On the other hand coppery gold shrinks in size and becomes dull and is difficult to solder. For this purpose a solder is made by adding gold with one seventh of silver into the above material and they grind them together.
Surprisingly, we have not found any comment on this sentence. If chrysocolla is malachite (copper carbonate), why can cuprous gold not be soldered with a brazing alloy containing copper, when copper can be dissolved in gold in any proportion? If chrysocolla is malachite, why this dual use of the copper minerals verdigris and malachite? If chrysocolla is greenockite (cadmium sulphide), however, this situation is explained by our results on soldering alloys prepared with copper salts, cadmium sulphide and gold showing a black earthy deposit in addition to the golden alloy [4,31].
519
G. Demortier
11.3.7
Italian jewellery
One element of a pair of earrings of the Askos type (see Fig. 11.17) belonging to a well-known family in the Tarento region of Southern Italy, and dating from the 4th century BC , was studied by PIXE. Each earring is a composition of two crescent-shaped gold leaves soldered together to give a hollow body, which is covered by individual granules of regular shape and size (0.7 mm). A beaded wire surrounds vertically the
Fig. 11.17. Askos type earrings from Tarento. The irradiated regions are indicated. For additional measurements on one single pair, see Table 11.6. TABLE 11.6 Analytical results for different areas of the Tarento earring (Askos type; see Fig. 11.17) Impact no.
1 2 3 4 5 6 7 8 9 10 11
Element concentration (%) Cu
Ag
Au
1.85 8.75 0.90 2.85 1.20 1.20 1.10 1.35 2.10 2.15 1.55
3.75 4.25 4.35 3.20 4.20 3.95 3.60 3.80 6.45 6.05 6.65
94.40 87.00 94.75 93.95 94.60 94.85 95.30 94.85 91.45 91.80 91.80
Impacts 6, 7, 8 are obtained on a separate granule (side applied on the body); 9, 10, 11 on the outside part of this separate granule.
520
Precious metals artefacts
maximum diameter of the hollow body. The maximum sizes of the two pieces are, respectively, 3.7 and 3.9 cm, their masses 8.77 and 9.30 g. The loop is a forged wire bulged at the soldered end. The regions where no solder is expected (impacts 3, 5, 6, 7, 8) contain 1.15% of copper and 4.00% of silver, the balance being gold (Table 11.6). The alloy of the loop (impact 2) contains more copper and that of the beaded wire more copper and less silver. The results of measurements on a separate granule (impacts 6, 7 and 8 for the side attached to the main body; 9, 10 and 11 for the outside surface) indicate that a material containing more copper and more silver has been used; it can then be concluded that it is a typical brazing procedure. 11.3.8
Gold artefacts from Slovenia
Gold artefacts are exciting archaeological finds In Slovenia, two finds of that type caused an animated debate. In 1859, a pair of gold Norico-Pannonian brooches (2nd century) was discovered on the south-eastern Alpine pass of Kranjski Rak. This find is interpreted by archaeologists as a votive deposit [32]. The second find is related to the early Slavic period. A female grave 355 in Ptuj contained, among its inventory, a pair of gold earrings and a ring. The manufacturing technique and the material aspect suggested that the earrings and the ring were contemporary, belonging to the Moravian jewellery of the 9th century [33]. The ring culturally differed from the other objects, since it was a male ring, pressed into a rectangular form to fit the hand of a woman. This and other observations led Koroec [33] to attribute the ring to a known historical person, related to the Byzantine court. In the opposing view [34], the presence of the Moravian jewellery is considered as a mere evidence for the migration of the Moravian nobility into the Slavic state of Blatograd. These questions and the quality of the objects provided sufficient motivation for further studies of their technical details. The external millibeam of the University of Ljubljana was used for the PIXE analysis. 11.3.8.1 Norico-Pannonian brooch One brooch of the pair (National Museum of Slovenia, no. R6913) was analysed at 14 characteristic regions (Fig. 11.18a). The respective concentrations of the elements are shown in Table 11.7. Not only the body measured at head [9], bow [12] and leg [14] (.98%) but also the expanding hook [8] and the spring [5] are made of very fine gold (more than 98%).
521
G. Demortier
Fig. 11.18. Gold samples and selection of the regions analyzed by PIXE. (a) NoricoPannonian brooch; (b) Slavic ring; (c) and (d) earrings. The corresponding concentrations are listed in Table 11.7.
The top knobs of the wings [10] and putto heads (points 1 – 3), the wings [11] and the filigree wire [4,13] contain more copper (2.2%), while the end knobs on the spring axis [6,7] also contain more silver and therefore markedly differ from the other parts of the brooch. These data suggest the following manufacturing procedure: the brooch body, including the part from which the spring and the needle were forged, was cast in one piece from a very fine gold. A similar procedure was used for making brooches of copper alloys [35]. The wings [11] were made of an alloy, which has a melting point about 208C lower than the brooch body; the bonding of the two parts was made by local heating at the junction region. The scanning of this region reveals a slightly increased concentration of gold and a reduced concentration of copper and silver, which resulted from selective oxidation during the heating process. The filigree wire [4,13] was attached similarly. The bonding region here is less than 0.1 mm thick; we
522
Precious metals artefacts TABLE 11.7 The Norico-Pannonian brooch, Slavic ring and earrings analyzed in points of Fig. 18a –d, respectively Impact
Cu
Norico-Pannonian brooch (Fig. 11.18a) 1 1.37 2 1.36 3 1.37 4 0.84 5 0.14 6 0.61 7 0.76 8 0.20 9 0.29 10 0.44 11 2.21 12 0.56 13 2.13 14 0.34
Ag
Au
2.6 2.5 2.6 2.9 1.7 7.4 6.5 1.2 1.5 2.6 0.8 1.0 2.4 1.3
96.0 96.1 96.0 96.2 98.1 91.9 92.7 98.6 98.2 96.9 96.9 98.4 95.4 98.3
Slavic ring (Fig. 11.18b) 1 2 3 4 5 6 7 8 9
3.17 3.14 2.25 2.88 3.16 3.57 2.99 2.97 3.08
18.8 20.3 21.8 24.8 23.8 20.7 21.9 18.2 22.3
78.0 76.5 75.9 72.3 73.0 75.7 75.1 78.8 74.6
Earring (Fig. 11.18c) 1 2 3 4 5 6 7 8 9 10
5.62 5.08 6.24 5.64 7.05 7.43 5.57 6.69 6.42 5.70
15.5 12.6 12.1 10.6 10.1 10.2 11.9 10.0 10.8 11.2
74.8 82.3 81.6 83.7 82.8 82.3 82.5 83.3 82.7 83.1 continued 523
G. Demortier
TABLE 11.7 (continuation)
Impact
Cu
Ag
Au
11 12 13 14 15 16 17 18
5.97 5.83 6.14 5.59 5.15 5.10 5.09 2.93
10.1 9.9 9.4 10.8 11.0 11.5 11.0 10.9
83.9 84.2 84.4 83.6 83.8 83.4 83.9 86.1
Earring (Fig. 11.18d) 1 2 3 4 5 6 7 8 9 10
5.83 3.42 3.04 5.17 3.27 4.92 4.68 4.69 4.85 5.13
9.5 8.1 10.4 9.5 10.6 10.5 9.5 10.5 37.8 34.4
84.6 88.4 86.5 85.3 86.1 84.5 85.8 84.8 57.3 60.4
Concentrations are given in wt%. The statistical errors are 1– 3% for copper, about 10% for silver and ,1% for gold [36].
were not able to detect it in a cross-scan of the brooch bow, as the resolution of the beam was about 0.5 mm. The composition of the brooch spring reveals the function of the brooch. Made of very pure and soft gold, the brooch was probably not intended to be worn. Should it be, the spring would soon break. It appears that it was made as a votive object. 11.3.8.2 Slavic jewellery Both the earrings and the ring (Regional Museum of Ptuj, S551, S553) are composed of hollow spheres made of gold foil and covered with granulation in triangular form. The analysed zones (Fig. 11.18b – d) were chosen in smooth and granular regions, as well as in junctions of the spheres with small (probably solid) beads and the connecting belts. No systematic differences were observed between different parts. The reason may be sought in a
524
Precious metals artefacts
relatively broad beam spot (0.5 mm) and in the non-homogeneity of the raw material, both of which screened the small differences between the soldered parts and the primary material. The soldering technique [11,14,19,20] cannot be identified with such a large beam. A microprobe could probably give more information. Nevertheless, the results permit the following explanations: the earrings were made of the same material, though probably not from the same ingot. One earring was repaired by a new wire of a less precious alloy (containing 60% gold). Furthermore, the craftsman was not particularly meticulous, leaving the wire end blunted. The ring, reportedly of different cultural group than the earrings, is indeed made of a different material. Though the copper content in the ring remains approximately constant (about 3%), the silver content varies irregularly between 18 and 25%. Such a variation is characteristic of natural electrum. This suggests that the ring was made locally, using local sources of raw materials. 11.3.9
The Guarrazar treasure
The Guarrazar treasure is an ensemble of Visigothic votive crowns and crosses made of fine gold alloy It was found in 1859 at Guarrazar, near Toledo (Spain) and it is dated to the 7th century AD . It was probably hidden at the beginning of the Muslim invasion in 711. The treasure is now spread over several places, the two main portions of it kept at the Museo Arqueologico Nacional in Madrid (six votive crowns and five crosses) and in the Cluny Museum in Paris (three votive crowns, three crosses and other small pieces, including the “R” letter from the Reccesvintus’ crown) [37,38]. PIXE analyses of more than 40 samples from the Guarrazar treasure held in the Museo Arqueologico Nacional in Madrid were performed. A welldecorated crown belonging to the treasure is shown in Fig. 11.19. Besides gold, the main element in the alloy, two other elements were systematically found in all the samples: silver (1 –30% in weight) and copper (0.3 – 7% in weight). The composition of the crown of Reccesvintus (detail in Fig. 11.20) is Au 94%, Ag 5% and Cu 1%. On the contrary, in a piece near the rim of the crown shown in Fig. 11.21, the composition is Au 64%, Ag 33% and Cu 3%. This high compositional variability can be explained by considering separately the results of each individual object. The analysis of each one shows a more homogeneous composition, especially in the silver content. Nevertheless, as for the crown and cross of Reccesvintus, both objects have the same composition (Ag 4 – 7% and Cu . 1%); this confirms the right association established between both items.
525
G. Demortier
Fig. 11.19. The Reccesvintus crown of the Visigothic treasure of Guarrazar (Toledo, Spain) kept in the Museo Arqueologico Nacional in Madrid [38].
Contrary to what is found in the crown and cross of Reccesvintus, another object, a crown with 10 open-work sections, is completely different from the cross with which it has been linked. In this case, they were probably incorrectly joined after the discovery of the treasure and restoration. Another object, the processional cross (Fig. 11.22), is the only one containing less than 1% Cu; this low copper concentration and the silver content between 10 and 15% correspond to the composition of the alluvial gold of the Iberian Peninsula and suggests the use of non-refined gold.
526
Precious metals artefacts
Fig. 11.20. Detail of the Reccesvintus crown showing the settings of gems in the gold main sheet [38].
Fig. 11.21. Small crown with gems [38].
Fig. 11.22. Horizontal arms of the processional cross of the Guarrazar treasure [38].
527
G. Demortier
The modern items added during the restoration at the beginning of the 20th century were also identified. The alloy used for these items is perfectly recognized by its standard and uniform concentrations of Ag and Cu (about 10 and 2%, respectively), although the restorers tried to reproduce the original colour of each restored area. Finally, these analyses confirm the difference of the quality of gold used in jewels and that used in coinage and goldsmithing at this time. The latter objects always contain less than 80% Au, a lower concentration than the one used for jewels. 11.3.10
Merovingian and late Byzantine jewellery
The study of a plate-buckle assembly excavated in 1842 at Pouan (Aube, France), with 13 other objects including two swords with gold ornaments on their sheaths and among others pommels, bracelet, torques, ring, gives an example of less sophisticated workmanship. These objects are now in the Muse´e Arche´ologique de Troyes (France). They are attributed to the tomb of Theodoric, King of the Wisigoths killed in 451 during the battle of Champs Catalauniques [39,40]. All the analytical data, except at location P5, indicate that the material at the solder sites as well as that in the bulk of the piece has nearly the same composition. No increase of copper is observed at the solder joins. The different parts of the artefact seem to have been bound together by local heating without any addition of fusible alloy. Bonding by local diffusion of copper powder is not expected here: diffusion bonding is of no use when the joins are so large and the soldering of the inner components to the external body is not meticulously carried out (Fig. 11.23). The solders are actually shaped by the interior and have to be hidden by the setting of the stones (garnets). The object was designed to be more utilitarian than decorative. Nevertheless, the quality of the material indicates that it was worn by a prince or a king. The tongue (P11) is of the same composition as the rest of the object (Table 11.8). While seven garnets are missing on this plate-buckle, a second piece of the same workmanship is still complete, the three rivets for the fastening on the belt included. The total length of the artefact is 4.6 cm. The differences in the visual outlook and composition at location P5 indicate that the ring housing and the articulation of the buckle are of modern origin. The richly decorated gold Byzantine cross (Fig. 11.24) containing a piece of wood, which is said to originate from the cross of Jesus Christ, has also been studied [41]. The front side of the relic is decorated with 48 stones and
528
Precious metals artefacts
Fig. 11.23. Plate-buckle from Pouan (France), 5th century AD . (a) Front view; (b) back view. The analyzed positions are indicated.
68 pearls; on the backside, the number of stones is 24, the number of pearls remains 68. The edges are also decorated with pearls, stones, filigrees, … The interior of the box contains a lot of partitions attesting that the jewel has been arranged for several purposes during centuries. This cross is venerated since the 13th century at Tournai and belongs to the treasure of the Cathe´drale Notre-Dame in this town. This Byzantine cross is known as a piece of a treasure originating from Constantinopolis, sacked in 1204. It may have been constructed in the 7th century AD [42]. The elemental composition at about 70 sites allows us to appreciate the bulk composition of the basic gold, wires and granulations, repairs and additional elements. Analyses at selected proton impacts on one main side of the jewel are reported in Table 11.9. The homogeneity of the
529
G. Demortier TABLE 11.8 Analytical results for different areas on the plate-buckle of Pouan (Fig. 11.23) Impact no.
1 2 3 4 5 6 7 8 9 10 11 12 13
Elemental concentration (%) Cu
Ag
Au
1.25 1.00 0.85 1.20 14.70 0.70 1.10 0.80 1.15 1.25 0.80 1.95 1.50
5.55 8.90 7.35 9.50 13.65 7.70 6.65 6.80 7.00 6.90 7.50 7.85 7.95
93.20 90.10 91.80 89.30 71.65 91.60 92.25 92.40 91.85 91.85 91.70 90.20 90.55
concentration in nearly all regions (in flat gold and in wires, into and out of the solders) indicates that no sophisticated procedure of soldering has been used in the construction of the box: the joined parts are sintered (without any additional alloying element) by heating the surfaces to be bound up to nearly the melting temperature. We have observed the lack of sophistication in the soldering procedure during the analysis of several elements inside the box. Decorations on the outside mask the rudimentary procedure of soldering. Repairs are detected in regions of impacts 9– 14: copper and silver wires, a silver support (97% Ag – 3% Cu) coated with a Au – Hg amalgam (90% Au – 10% Hg) next to additional caps enclosing substitution stones. 11.4 11.4.1
PRE-HISPANIC GOLD ARTEFACTS OF MESOAMERICA Archaeological context
Gold metallurgy in America started about 1500 BC in Peru. Some centuries later, casting (such as the lost-wax technique), solderings and various processes for gilding were invented and developed to manufacture artefacts and ornaments. The alloy used is known as tumbaga, a copper-rich gold alloy. The reasons for the use of tumbaga do not only concern technical aspects
530
Precious metals artefacts
Fig. 11.24. The Byzantine cross of the treasure of the Cathe´drale Notre-Dame de Tournai. ( For a colored version of this figure, see Plate 11.II.)
such as fusion point, malleability, hardness and metal economy. Gold and other precious metals were associated with religion, social status and political power. The colour of the ornament belongs to religious and cultural aspects of the Amerindian cosmos. In that respect, goldsmiths of ancient America learned to modify the colour of their artefacts, i.e., they learned to modify the surface of the items. Gold metallurgy in America was specially directed to surface modifications [43].
531
G. Demortier TABLE 11.9 Composition of several sites on the Byzantine cross of the Cathedral of Tournai (Belgium) (Fig. 11.24) [41] Impact no.
Cu
Ag
Hg
Au
Analysed region
Original basic gold 1 1.8 ^ 0.1 2 1.7 ^ 0.1 3 3.0 ^ 0.2 4 3.0 ^ 0.2
5.6 ^ 0.5 6.0 ^ 0.5 6.6 ^ 0.6 7.1 ^ 0.6
92.6 ^ 2.5 92.3 ^ 2.5 90.4 ^ 2.5 89.9 ^ 2.5
Flat part Flat part Edge of a lost pearl Top cover
Original wires 5 2.5 ^ 0.2 6 3.4 ^ 0.2 7 2.8 ^ 0.2 8 3.0 ^ 0.2
7.2 ^ 0.6 7.4 ^ 0.6 9.8 ^ 0.6 8.7 ^ 0.6
90.5 ^ 2.5 89.2 ^ 2.5 87.4 ^ 2.5 88.3 ^ 2.5
Thick wire between pearls Thick wire between pearls Thin wire across a pearl Thick wire across a terminal pearl
Repairs and additions 9 8.1 ^ 0.4 10 4.6 ^ 0.2
91.9 ^ 1.5 4.5 ^ 0.5
– 90.9 ^ 2.5
11
100
–
–
12
5.5 ^ 0.3
13
(0.6)
(20)
14
4.2 ^ 0.2
94.2 ^ 1.5
Silver wire Gold wire (may also be original) Copper wire to bind a new pearl Cap of an additional stone Central element in silver coated with amalgam Contains also 1% Pb
9.4 ^ 0.7
85.1 ^ 2.5 (7.2)
(72.2) 0.6 ^ 2.5
All concentrations are in wt%.
In pre-Hispanic Mexico, metallurgy was developed only a few centuries before the Spanish conquest, in a period called “post-classic” (900 – 1530 AD ). Indeed, the basic skills of gold metallurgy were imported from South America, where the first methods of copper and gold folding were developed before the early horizon (about 500 BC ). Casting, gilding, silvering and soldering, among other techniques, were invented some centuries later [44]. Despite the foreign origin of this technology, the level of performance reached by the goldsmiths of Mesoamerica, and especially those in the south of Mexico (Oaxaca) was remarkable. People from this region, the Mixtecs, achieved a skill for casting and gilding gold-based alloy artefacts that was recognized by other Mesoamerican people and the Spanish conquerors.
532
Precious metals artefacts
Gold alloy artefacts were used by religious officials and members of the leading areas of society. Their colour was very important for religious and cultural reasons. In ancient America, metallurgy was developed following the concept of the manipulation of the colour of items [45]; this technology was not known in the Old World at that time. The gold alloys used in Mesoamerica contained considerable amounts of copper and/or silver and their colours varied from red to white-yellow, depending on the alloy composition. The lost-wax and false-filigree techniques for casting and depletion gilding were the most developed techniques in this cultural area. In South America, soldering was frequently used before the Spanish conquest [46] whereas it appeared to be unknown in Mesoamerica. To manufacture an artefact by means of the lost-wax technique, first a core of clay and charcoal needs to be prepared; this core closely resembles the final figure of the artefact. Then, the core is covered with a wax coating, the thickness of the wax layer determining the thickness of the metal to be cast. The artefact is modelled with all its details in wax. Some wax rods are added to the model to provide air vents; a wax cone is also added to be used as a casting tunnel. A mould of the artefact is prepared by covering the wax model with clay; some wooden pegs are added to hold the core in place within the mould. The artefact mould is dried and heated so that the wax melts and runs out of the mould. Then, the hot mould is inverted and the molten metal poured in through the tunnel. When it is cold, the mould is broken and the core supports are removed. Sometimes the core is broken up and extracted when it is possible. Finally, the artefact is cleaned and polished. The false-filigree technique is based on the application of wax wires to model the artefact. In this way, after the lost-wax process, very fine and delicate features can be achieved without soldering wires onto the artefact [47]. The items produced in this manner could undergo successive processes of surface oxidation and elimination of Cu and Ag oxides to get a gilded surface. To oxidize the surface, the object is heated at a temperature above 5008C. To eliminate oxides, the artefacts were dipped in a solution of water, salts and acids from plants [48]. After several steps, the surface is “enriched with gold” and the surface alloy has a more “golden” appearance than the original alloy. This process affects a depth up to some microns depending on the number of steps applied and the alloy composition. Other gilding techniques in ancient America consisted of plating copper alloys with gold alloys by using “electrochemical” processes [49].
533
G. Demortier
The analysed objects include finger rings, chin and ear ornaments as well as zoomorphic and solar pendants from the Mixtec culture corresponding to the post-classic period from Oaxaca, Mexico. All these gold artefacts were shown during the exhibition “Mobilier Fune´raires des Zapote`ques et Mixte`ques,” presented during the Europalia Mexico 93 Festival in Brussels, Belgium [50]. The analyses that were performed helped us establish the metallurgical techniques used at that time and to complement the information on the ancient gold metallurgy of Mesoamerica. 11.4.2
A selection of typical artefacts
Amongst numerous jewellery items from Mexico, Columbia that have been studied by PIXE [51,52], the serpent chin ornament (Fig. 11.25 and Table 11.10) has almost the same composition at the measuring positions 1 – 4, 7– 12 and 21, i.e., 65.2 ^ 2.6% of Au, 31.5 ^ 3.0% of Ag and 3.3 ^ 0.7% of Cu. The largest deviations from the mean values are observed at positions 5 and 6. The enhanced gold concentrations at those positions, 70.4 ^ 1.4% of Au, 26.5 ^ 1.5% of Ag, 3.1 ^ 0.2% of Cu, may be the result of a soldering process without any brazing alloy. In that case, local fusion could have been used to solder the rings. During heating, the less noble metals (Ag and Cu) oxidize, and these oxides are removed after cleaning. From all those measurements one may deduce that the artefact was fabricated by the lost-wax technique. Measurements performed at various incident proton energies indicate that the material is homogeneous: the same concentrations
Fig. 11.25. Serpent chin ornament (the analytical results of the measuring positions are listed in Table 11.10). Length: 4.5 cm; height 2.5 cm; width: 2.7 cm.
534
Precious metals artefacts TABLE 11.10 Elemental composition of the serpent chin ornament (Fig. 11.25) [51] Impacts
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21
Au-Lb Au (%)
Ag (%)
Cu (%)
63.8 64.3 65.0 65.9 69.0 70.9 65.0 65.6 65.2 66.6 67.0 66.3 67.9 67.1 65.4 67.9 64.3 64.9 67.1 69.8 66.7
32.9 32.5 31.7 30.4 28.1 25.7 32.0 31.4 31.5 30.3 29.4 30.4 28.4 29.4 31.3 28.9 32.5 31.8 29.9 27.9 30.7
3.3 3.2 3.3 3.7 2.9 3.4 3.0 3.0 3.3 3.1 3.6 3.3 3.7 3.5 3.3 3.2 3.2 3.3 3.0 2.3 2.6
are found not only at the outer surface (0.1 mm) but also down to 10 mm in depth. Very often PIXE analyses are performed at one single fixed energy. In almost all analyses of objects of antiquity, discrepancies are observed between the compositional data obtained by various methods. Variations obtained by the same method but at various projectile energies and/or in different geometrical arrangements or variations from point to point are generally attributed to the lack of local homogeneity, or to the presence of a corrosion layer or patina at the surface. However, the possibility of corrosion or depletion of less noble components of a gold alloy has been demonstrated. We have shown [53] that for gold jewellery items of Mesoamerican origin, the assumption of homogeneity in depth may not be valid.
535
G. Demortier
Gold jewellery of Columbia and Peru were often made with tumbaga, a gold – silver– copper alloy containing a large proportion of copper in the bulk. An improper preservation of such artefacts would show a visible alteration of the surface by burial over a long period of time due to the transformation of copper in copper oxides. These oxides may be eliminated by a mechanical or a light chemical process but these processes would completely modify the relative local concentrations. These local modifications could easily be observed by using a scanning PIXE microprobe [54]. The pre-Columbian goldsmiths were aware of this imperfect preservation of tumbaga and they experimented with procedures leading to the removal of the excess of copper at the surface by various recipes [48,49]. In this way, they produced artefacts with a colour very close to that of pure gold. These improvements comprised two aspects: (a) the aesthetics of the artefact were increased: the colour became closer to the colour of pure gold; and (b) the material obtained had a better resistance to corrosion. The copper depletion and the consequent gold enhancement at the surface may be identified by RBS but the quantitative analysis of the depth profiles of copper, silver and gold is not easily determined by using a-particles or proton spectroscopy of elastically scattered projectiles [52,56]. PIXE at various energies from 600 keV to 3 MeV and with various incidences of the proton beam with respect to the irradiated surface of the jewellery item may give a more quantitative depth profiling. 11.4.3
Differential PIXE
As was mentioned above, the relative intensities of Cu and Au X-ray lines induced by protons on gold-plated copper may be differentiated from copper gold alloys, by using X-ray intensity ratios [55]: the Cu-Ka/Au-La intensity ratio has been used for that purpose. Practically, in the study of gold artefacts, the possible presence of zinc in the material would give rise to line interferences: the Ka and Kb lines of zinc (at 8.62 and 9.57 keV) nearly coincide with the Ll and La lines of gold (8.49 and 9.6 – 9.7 keV). It is not possible to use peak deconvolution to separate these lines since the natural Ka/Kb and La/Lb ratios are far from the standard ones: the large absorption coefficient of the La line of gold is indeed very close to the absorption edge of Zn and the secondary fluorescence of all gold L lines in zinc would completely modify the relative intensities of all X-ray lines of interest. One way to avoid this selective effect is to use the Lb line of gold. Figure 11.26 illustrates the calculations. The calculated ratio of Cu-Ka /Au-Lb intensities for several incident proton energies in the range 536
Precious metals artefacts
Fig. 11.26. Cu-Ka/Au-Lb intensity ratios for various incident energies of protons and various thicknesses of Au layers on a pure copper substrate. The corresponding variations of this ratio for different concentrations of copper in homogeneous alloys.
537
G. Demortier
0.6 –3 MeV on a layer of pure gold on a pure copper substrate decreases by four orders of magnitude when the Au layer thickness increases from 0.5 to 15 mm. The data indicated along the dashed lines on these calculated curves represent the Cu contents (in percent) of Au –Cu homogeneous alloys giving the same Cu-Ka /Au-Lb intensity ratio. In particular (point A in Fig. 11.26) for 1.8 MeV protons, a gold layer (4 mm thick) on a pure copper substrate is equivalent to a homogeneous sample (92% Au þ 8% Cu) as far as the Cu-Ka/ Au-Lb ratio is concerned (0.86 for point A). It is very easy to distinguish between the presence of a layer or a homogeneous alloy by changing the proton incident energy. So, if one increases the incident energy up to 2.6 MeV the Cu-Ka/Au-Lb ratio would increase up to 2.05 at point B if the sample is a 4 mm thick gold layer on a pure copper substrate, but the same ratio would decrease down to 0.71 (point C) if the sample is the homogeneous bulk material containing 8% of copper. On the other hand, if one decreases the proton energy down to 1.2 MeV, the same ratio would become 0.17 (point D) in the case of a 4 mm thick gold layer on copper but it would increase up to 1.2 (point E) in the case of the homogeneous copper– gold alloy with 8% of gold. Any other composition (i.e., of layered copper– gold alloy) would give a variation of the ratio A at 1.8 MeV between these limits B – E when the incident proton energy is varied in 0.4 MeV steps. Around the starting energy at 1.8 MeV, n measurements at n different energies would be sufficient to determine the depth profile in (n 2 1) layers on a pure copper bulk material. 11.4.4 Application to the measurement of the gold enhancement at the surface of tumbaga By using differential PIXE, the depth profile of gold at the surface of an alloy may be determined. As the measurements only concern ratios of X-ray intensities of major elements, the counting rate must be high enough to achieve statistically accurate ratios: the error on each measurement would be 1% or lower. It is not complicated to choose the actual pair of parameters (composition and layer thickness) by collecting this X-ray ratio at various energies (1.2 – 2.2 MeV in this example). As described in the previous paragraph, the choice of the actual composition at the surface to start the computation of the homogeneous bulk composition or some depletion in depth of copper may be determined experimentally by using 1 MeV a-particles (having a much lower range than protons) as incident projectiles and by detection of the K lines of Cu, the K- and L lines of Ag and the M- and L lines of gold. The M/L intensity
538
Precious metals artefacts
Fig. 11.27. Solar disk (Peru, 500 –1500 AD ; diameter: 44 mm).
Fig. 11.28. Zoomorphic pendant (Columbia, 500 – 1500 AD ; length: 71 mm). ( For a colored version of this figure, see Plate 11.III.)
539
G. Demortier
ratio of gold and the L/K ratio of silver are completely different when a homogeneous sample or an artefact exhibiting a sharp depth profile is irradiated in the same geometrical conditions. Those ratios are particularly sensitive to the depth profile as a result of the high absorption coefficient of the low-energy X-ray lines of gold and silver in the material. A solar disk (Fig. 11.27) and a zoomorphic pendant (Fig. 11.28) have been studied in this way. Cu-Ka/Au-Lb intensity ratios were measured at 10 different incident proton energies (0.6 – 2.6 MeV). The variation of the Cu-Ka/Au-Lb intensity ratio indicates that the artefacts contain more gold at the surface than in the bulk. The depth profile for each artefact has been calculated in order to fit with the PIXE data. One observes that the copper depletion at the surface is evident and that the depleted layer is in the micron range. By using the top surface composition given by PIXE with 1 MeV a-particles one obtains the depth profile reported in Table 11.11. The comparison of these fits with those obtained by computing the RBS spectra, which are simultaneously recorded in a very good agreement and are reported elsewhere [56]. A control of the actual depth profile by PIXE may also be done by tilting the irradiated surface with respect to the incident proton beam while avoiding to modify the position of the detector as reported in another study [51]. The pendant of Fig. 11.29 was recently discovered in Oaxaca, Mexico. It corresponds to the late post-Classic period (1500 AD ). This artefact was nondestructively analysed at the Pelletron Laboratory of the IFUNAM [57]. The results of a non-vacuum PIXE analysis of the various sections of the pendant are shown in Fig. 11.30. The bells of the pendant have lower concentrations of gold than the other parts of the artefact, but in some regions (positions 7 – 8, 13 – 15), the amounts of gold are higher and quite similar to the rest of the pendant. It is observed that the alloy composition is rich in gold but it is not uniform in the rest of the pendant (feathers and main body). There is a clear difference of colour between the bells and the other pendant sections. The gold pendant has been irradiated with protons of energy ranging from 1.3 – 3 MeV and the relative intensities of the signals of Cu and Au have been measured. For the feathers sections, results at 3 MeV could be compatible with a uniform distribution of Au, Ag and Cu over the entire depth at, respectively, 68, 12 and 20%. When the same region is analysed at lower proton energies, the calculation of concentrations from the corresponding PIXE spectra give completely different results.
540
Precious metals artefacts TABLE 11.11 Concentration depth profile of Cu, Au and Ag in the Amerindian artefacts [53] Depth below the surface (mm)
Concentration (wt%) Cu
Au
Ag
Solar disk (Fig. 11.27) 0.28 0.28 0.28 0.28 0.28 0.28 Bulk
0 4 8 12 16 24 28
94 90 86 82 74 70 68
6 6 6 6 6 6 6
Zoomorphic pendant (Fig. 11.28) 0.12 0.09 0.09 0.06 0.06 Bulk
15 22 27 31 35 41
80 70 65 60 55 45
5 8 9 9 10 14
The gold content seems to increase when the proton energy decreases (Table 11.12). The reason is that the material is not homogeneous with respect to its depth. For 1.3 MeV protons, the analysed layer is much less than for 3 MeV protons, and the analysis thus more sensitive to the surface concentration. Some measurements at 1.3, 1.7, 2.2, 2.6 and 3 MeV have been performed to give the actual distribution of Au and Cu in four successive layers and also the bulk concentration (Table 11.13). Similar results have been found for the main body section of the pendant (warrior head). Only a very smooth profile was found for the bells. In this case the gold alloy can be considered to be quasi-homogeneous. Then, the variations observed for the elemental composition by PIXE at 3 MeV at different analysis points are due to the thickness of the enriched gold surface and the gold depth profile. A depletion gilding process was used to enrich the surface of gold at a mean depth of 3 –5 mm. Therefore, the differences observed at positions 7 –8 and 13 – 15 of the bells must correspond to gilded regions. These results are in agreement with a surface examination by optical microscopy of these regions. The determination of the depth profile is also possible by using Rutherford scattering of a or proton particles. Any substrate containing a larger proportion of medium or light elements in the bulk in comparison
541
G. Demortier
Fig. 11.29. Eagle warrior pendant (San Francisco Caxonos, Oaxaca, 1500 AD ; length 8.0 cm), Oaxaca’s Regional Museum, Mexico. ( For a colored version of this figure, see Plate 11.IV.)
with the surface may be differentiated from the gold-plating thickness. Applications may also be made for gold plating of wood, ivory and ceramic items, but RBS is generally not sufficiently powerful to identify the entire depth profile of Cu, Ag and Au down to several micrometre. The contribution of particles backscattered by medium and light element at the surface is indeed superimposed on the signals derived from gold in the deeper layers. A combination of differential PIXE and RBS is recommended to solve the problem. By starting the computation of PIXE spectra at various energies by using the gold surface concentration determined by RBS one reduces the number of iterations necessary to determine the complete depth profile. This RBS procedure is complementary to PIXE with low-energy a-particles.
542
Precious metals artefacts
Fig. 11.30. Apparent elemental composition determined by PIXE at the impacts points indicated in Fig. 11.29. For these analyses a 3 MeV proton beam and an external beam setup were used. For the calculation of this apparent composition we suppose that the material is homogeneous in depth.
TABLE 11.12 Apparent elemental composition (with the hypothesis that the material is homogeneous) of the middle region (impact 40 of Fig. 11.29) of the warrior pendant as a function of the proton beam energy Proton beam energy (MeV)
3.0 2.6 2.2 1.7 1.3
Apparent elemental composition Au (%)
Ag (%)
Cu (%)
68.0 ^ 2.0 68.0 ^ 2.0 70.0 ^ 2.1 72.5 ^ 2.2 75.0 ^ 2.3
12.0 ^ 0.6 12.0 ^ 0.6 12.0 ^ 0.6 12.0 ^ 0.6 12.0 ^ 0.6
20.0 ^ 1.0 20.0 ^ 1.0 18.0 ^ 0.9 15.5 ^ 0.8 13.0 ^ 0.7
Concentrations in wt% [57].
543
G. Demortier TABLE 11.13 Elemental depth profile determined by PIXE at various proton energies (impact 40 of Fig. 11.29) Layer
1 2 3 4 5
Thickness (mm)
1.0 0.7 0.5 0.5 Bulk
Elemental composition Au (%)
Ag (%)
Cu (%)
100 ^ 3 85.0 ^ 2.6 70.0 ^ 2.1 55.0 ^ 1.6 40 ^ 1.2
– 5.0 ^ 0.3 10.0 ^ 0.5 10.0 ^ 0.5 10.0 ^ 0.5
– 10.0 ^ 0.5 20.0 ^ 1.0 35.0 ^ 1.8 50.0 ^ 2.5
Concentration in wt% [57].
11.5
CHARACTERIZATION OF COMPLEX ITEMS
PIXE, RBS and PIGE are sometimes not sufficient to study complex objects Additional non-destructive techniques must then be used to complement these results. 11.5.1
XRF induced by a g-ray source
Gamma-rays from a 241Am source (59.54 keV) can be collimated to provide a narrow beam through a regular hole pierced in a 2 cm thick lead shield. The incident g-ray beam is at 908 with respect to the detection of secondary X-rays. The sample surface is tilted at 458 with respect to both incident and outgoing directions and can be finely adjusted so as to always observe the Compton peak at the same location in the XRF spectrum. This arrangement is most convenient to irradiate narrow regions only and to avoid large shadowing effects if the surface of the artefact is not flat. The relative intensity of the K lines of gold (induced by the low-intensity g-rays of energy greater than 110 keV from the 241Am source) to that of the L lines is qualitatively used to recognize thin gold layers (20 mm or less) from samples containing gold in the bulk as well (100 mm or more). Since L X-rays are more absorbed than K lines, the K/L intensity ratio increases with the increasing of the gold layer thickness, provided the geometry of the experiment is not modified. 11.5.2
Gamma-ray transmission measurements
For jewellery items of thickness greater than a few mm, the estimate of the core density (initially evaluated from weight and volume measurements)
544
Precious metals artefacts
was locally checked by using a collimated g-ray beam (1 mm in diameter drilled in a 7 cm thick lead shield) from a 137Cs source (Eg ¼ 662 keV). Measurement of the intensity of the transmitted g-rays while scanning the specimen in front of the collimator permits to identify regions of various density. Other g-ray sources giving lower energy g-rays (such as 226Ra) or production of g-rays by PIGE on materials producing a high flux of photons such as Al (844 and 1013 keV), Na (439 keV), F (110, 197 keV) and Li (429 keV) could be alternative (but more expensive) methods to cover a larger region of attenuation coefficients in order to investigate materials of various thicknesses and compositions. The transmitted intensity is described by the attenuation law Itrans ¼ Io e2mx where Itrans is the transmitted and Io the original beam intensities, m denotes the linear absorption coefficient of the material (expressed in cm21) and x the thickness of the absorbing layer. The choice of the product mx in the range 0.3 – 3 can give rise to convenient intensity ratios with and without the specimen in front of the source. This procedure cannot be used to identify the elements in the core region (several millimetres below the surface) of the specimen (no secondary fluorescence from the core may be detected due to a large absorption) but is only meant to qualitatively verify that the core is not empty.
11.5.3
Study of a composite gold jewellery artefact
A bracelet of Dacian origin (Hellenistic) (Fig. 11.31) was studied by using differential PIXE (Ep being varied between 2.2 and 3.2 MeV), XRF and gamma-ray transmission (GRT) [58]. The weight and volume measurements gave a mean density of ca 5.2 g/cm3, i.e., only 35% of the density of the surface alloy whose composition was confirmed to be homogeneous down to a depth of minimum 10 mm (by differential PIXE) and at least down to 100 mm (by XRF), as far as the high gold content is concerned (Table 11.14). GRT indicated that the external surface of the toroı¨dal bracelet was of much higher density (as observed by a larger absorption of 662 keV g-rays, shown in Fig. 11.32). The item is made out of a sheet of a gold-rich alloy surrounding a core of lower density. The transmission in region B indicates that the thickness of the gold sheet is of maximum 0.35 mm, an estimate which is qualitatively compatible with measurements at A and C (the g-ray beam, 1 mm in diameter, was too wide to give a precise estimate). The soldering of the heads of rams is probably of modern origin (possibly a repair): the alloy in this region contains not only larger concentrations of Cu and Ag but also traces of Cd. The number of different solder compositions does not allow us
545
G. Demortier
Fig. 11.31. Hellenistic bracelet (Dacia) (3rd Century BC ?) in the experimental setup for g-ray transmission measurements. Weight: 66.8 g; volume: 12.9 cm3.
to determine if the soldering procedure is modern or not, as discussed in section 2.5. The determination of the centre of gravity (G in Fig. 11.31) also indicates that the composition of the core of the item in the regions of both heads is not fundamentally different from that of the toroı¨dal region. The Ka /La intensity ratio of the gold lines observed by XRF is nearly twice the value observable for a flat thick gold sample. The high-energy g-rays of 241 Am have indeed the possibility to excite the X-ray lines of gold on both sides of the toroı¨dal structure, but only K lines induced in the second face may reach the detector due to the complete absorption of the corresponding L lines.
546
Precious metals artefacts TABLE 11.14 PIXE and XRF analyses of Dacian bracelet shown in Fig. 11.31 [58] Impact PIXE (2.8 MeV) 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 XRF x1 x2 x3
Cu (%) 0.5 2 0.7 0.2 0.2 0.2 0.2 2.0 2.2 0.9 5.1 4.1 6.6 3.0 4.5 #1.8 #1.0 #1.1
Ag (%)
Au (%)
,0.1 0.1 0.1 0.2 0.2 0.1 0.1 0.1 0.1 0.1 1.9 3.5 6.8 1.8 4.8
99.5 97.9 99.2 99.6 99.4 99.7 99.7 97.5 97.3 99.0 93.0 92.4 86.6 95.2 90.7
0.05 0.02 0.05
Remark
Possibly 0.2% Cd
0.4% Cd 0.4% Cd Granule Granule Soldering Granule Granule
98.2 98.9 98.8
Fig. 11.32. Transmission of 662 keV g-rays when a (1 mm2 diameter) photon beam is scanned across regions A to C. The increase of absorption in the external regions of the torque (A and C) clearly indicates that the core is empty or contains a large quantity of light elements.
547
G. Demortier
11.6
GOLD COINS
As demonstrated in the first paragraph of the present chapter, PIXE and proton-induced activation (PAA) analyses of Byzantine gold coins give comparative analytical results for major elements (see Table 11.1). The gold content of those coins is indeed very high. Contrary to silver and copper coins, the surface corrosion of those gold coins is expected to be very low. 11.6.1
Fineness measurements of gold coins
For coins of lower gold content, various surface techniques could give very different results. In Table 11.15 we show the comparison of XRF, PIXE, scanning electron microprobe (SEM) of data on coins of the Bishop-Princes of Lie`ge. The incident photon beam for XRF was produced by a 241Am source emitting monoenergetic g-rays of 59 keV: the estimated thickness of the analysed layer is around 40 mm. For PIXE with 3 MeV protons the effective analysed layer is less than 10 mm [59]. The SEM results concern only the first 1 –2 mm below the surface. PIXE is more efficient for the copper determination but XRF is more sensitive for silver: both elements are quantitatively analysed using their K lines. A comparison of spectra on the same sample by those techniques is given in Fig. 11.33. SEM results are extracted from the book of Scheers et al. [60]. None of the above reported techniques can give the actual composition. One can only conclude that the gold enrichment at the surface is very important: as the thickness of the analysed layer increases, the apparent gold content decreases. Since less noble metals were selectively eliminated from the surface, bulk concentrations are more representative than surface or subsurface content for determination of the original fineness of the coin. Nevertheless, several groups refer to PIXE analyses of a large variety of gold coins using PIXE at low [62] or high [63] incident proton energies. 11.6.2
Gold coins from the ancient world
For the numismatists, the fineness of gold coins is not the only useful parameter. If they intend to look for the origin of the raw material in which the coin is struck, they also need to know the concentrations of a number of trace elements and sometimes isotopic ratios of key elements. Depending on the geological formation of the ores, trace elements present patterns that may be used for “fingerprinting.” As far as the provenance of the material is to be studied, the number of elements to be detected must be as high as
548
Precious metals artefacts TABLE 11.15 Composition of gold coins by XRF, PIXE and SEM [61] Bishop of Lie`ge
Type, date
Louis de Bourbon (1450– 1482)
Postulat
Jean de Horne (1) (1484– 1505)
Postulat
Jean de Horne (2) (1484– 1505)
Postulat
Jean de Horne (3) (1484– 1505)
Postulat
Englebert de la Marck (1506– 1558)
Postulat
Ferdinand de Bavie`re (1) (1612– 1650) Ferdinand de Bavie`re (2) (1612– 1650) Ferdinand de Bavie`re (3) (1612– 1650) Ferdinand de Bavie`re (4) (1612– 1650) Ferdinand de Bavie`re (5) (1612– 1650)
XRF
PIXE
SEM
Theoretical fineness
64.4 32.5 3.1 60.0 35.2 4.8 57.6 36.4 6.0 48.8 44.5 6.7 58.9 40.0 1.1 81.0 5.0 14.0 79.8 4.8 15.4 89.2 5.3 5.5 79.0 8.2 12.8 86.7
70.2 27.5 2.3 64.2 33.3 2.5 60.1 35.5 4.4 40.8 52.6 6.6 59.0 38.9 2.1 78.0 4.4 17.6 82.6 4.8 12.6 89.8 5.1 5.1 82.1 7.9 10.0 Not measured
(85–94)
?
(68–78)
(38.9– 41.6)
(91–55)
(38.9– 41.6)
(31–35) (54–56)
(38.9– 41.6)
(49–60)
(38.9– 41.6)
Forint Bouillon, 1613 Forint Bouillon, 1613 Crown Bouillon, 1613 Forint Hasselt, 1614 Postulat
Au Ag Cu Au Ag Cu Au Ag Cu Au Ag Cu Au Ag Cu Au Ag Cu Au Ag Cu Au Ag Cu Au Ag Cu Au
Lie`ge, 1637
Ag Cu
3.6 9.7
(88–94) 77.1 (84–94) 77.1 (3– 73) (92–96) 88.2 (4– 5) (92–94) (3– 5)
77.1
(95–98)
88.2
(2– 3)
possible and the analytical equipment employed should provide the lowest limit of detection for the determination of these traces. Few techniques allow the determination of trace elements in metals down to very low levels [64]: two highly sensitive techniques are proton activation analysis [65 – 68] and ICP-MS [3,69,70]. Pioneer investigations in that respect were made intensively by Dr M.F. Guerra and her co-workers. Charged particle beams are used to determine
549
G. Demortier
Fig. 11.33. Comparison of X-ray spectra induced by PIXE (a) and RIXRF (b) on a gold coin of Louis de Bourbon.
550
Precious metals artefacts
specific elements in specific matrices. Nevertheless, their use is limited because of the slowing down of the incident particle during the penetration of the sample; thus, the thickness of the analysed layer is in the range 300 – 400 mm, i.e., far better than the layer that is sampled by means of PIXE but sometimes still too small. Inductively coupled-plasma atomic emission spectrometry (ICP-AES) has been widely used since the 1970s, but the (optical) measurement of the atomic spectra was only recently replaced by mass spectrometric forms of detection (ICP-MS). ICP-MS has been used to determine a very wide range of trace elements at a very low level either in solutions or in solid samples, the latter by using a laser ablation system [69]. This technique is able to determine the concentration of a large number of elements simultaneously; these not only include major and minor elements but also trace elements. Volatile elements may be lost during the sample preparation or during the interaction laser/matter and the transportation to the plasma. Guerra’s team developed a technique of solid analysis for gold with a UV laser that produces craters of 40 mm diameter and 130 mm depth (sampling of about 3 mg per crater) with a limit of detection from 10 ppb to 1 ppm. A sampling of 2 mg of a coin to be brought into solution is also sufficient to obtain confident results. This technique can be considered as virtually non-destructive and has been applied to various problems of archaeometry [69 – 71]. A complementary technique for provenancing is based on isotope analysis, usually produced by thermal ionization mass spectrometry (TIMS) but sometimes by ICP-MS. The isotopic composition of Pb is dependent on the ore deposit from which the Pb originates since the Pb isotopes are the stable end-members of the radioactive U and Th decay series. Thus the ratios 208Pb/206Pb, 207Pb/206Pb, 206Pb/204Pb can be used to differentiate between ores of different provenance [72]. The quantitative aspects of ICP-MS analysis for the study of gold coins were described by Guerra and Gondonneau [73], They used gold coins with certified composition as reference materials. The main components (Au, Ag, Cu) and 10 trace elements with various levels of concentrations were used as standards (see Table 11.16). NBS reference samples were also used in order to check the best experimental procedures to obtain confident results with ICP-MS. They conclude that the use of UV laser ablation was satisfactory for the study of major, minor and trace elements, if a well-established analytical protocol is maintained. Laser ablation is a complex process and difficult to model at a fundamental level. The properties of the crater (depth, size, shape) and the amount of ablated material highly depend on the properties of the sample, its surface and also on the properties of the laser, all
551
G. Demortier TABLE 11.16 Composition determined by proton activation analysis of the gold-certified coins used as standards [73] M2 Au (%) Ag (%) Cu (%) Traces (in ppm) Pb Pt Pd Fe Sn Sb As Zn Te Ti
98.4 1.5 0.04 33 98 7 550 3 0.4 0.1 9 0.5 2
Britain 92.9 3.0 3.9 146 562 191 662 56 59 26 74 4 1.5
France 90.1 6.3 2.9 196 2924 149 1998 272 112 82 146 15 2
Russian 91.2 0.23 8.5 79 307 16 103 1 28 63 5 1 0.3
of which undergo complex time-dependent changes during the laser ablation [74]. It is recommended to use a standard material with a composition very close to that of the unknown artefact. ICP-MS and PAA have been applied to the study of Muslim North African gold. The dinar was one of the most important means of commercial exchange in the Mediterranean basin during the Middle Ages. A number of elements were determined to shed light on problems such as (a) debasement, (b) manufacturing technology and (c) provenance of this coinage. The first Muslim dinars were struck with the same gold as Byzantine solidi. However, under the Aghlabids (807– 1063 AD ), three different “sources” of gold appear. Under the Fatimids there are certainly two different types of gold, one similar to the gold of the Aghlabids and another closer to that of the Almoravids (1075 – 1114 AD ) and the Almohads (1150 –1217 AD ). Coins struck under the latter dynasties are made in yet another type of gold. The analysis of a few Spanish Almoravid coins suggests that they were made out of trans-Saharian Sudanese gold. A similar study on Spanish coins and Oriental coins is in progress; its aims are to investigate the importance of, respectively, Sudanese gold and Red Sea and Middle East gold sources during the Middle Ages [73].
552
Precious metals artefacts
The Almoravid event was a real revolution: not a political, social or even religious revolution, but rather an economic one. The key stone of this power, which gave a new political revival to the far Maghreb during the second half of the 11th century and the first half of the 12th century, is in fact of commercial nature: the control on trans-Saharian trade brought back to the Muslim north, the gold from more southern regions. The Almoravid coinage is the brightest expression of this success. So, the monetary tool appears like one of the most significant signs of the Berber emirate health. The combined use of LA-ICP-MS and proton activation gave information on the monetary devaluation phenomena and the origin of the ores. These analytical results together with the written sources (which sometimes contain insufficient information) are new sources of understanding of the economical, social and politico-military history of the Almoravid dynasty [76]. Gondonneau and Guerra have shown that until about 750 AD , all the Islamic world re-cycled previous coins: Byzantine and Sassanian coinage in the Orient, Byzantine coins in North Africa and Visigothic coins in Spain (Fig. 11.34). The transition between the end of the Ummayad dynasty (i.e., Arab caliphs who reigned in Damascus from 661 to 750 AD and started the expansion of the Muslim Empire) and the arrival of the Abbasids brought a change in the silver and gold supplies to mints issuing dinars and dirhams all over the Arabian Empire. These phenomena certainly
Fig. 11.34. Variation of the fineness of Iberian coins: X, Byzantine; A, Latin; K, Bilingual; W, Arabic.
553
G. Demortier
correspond to the exploitation of deeper silver veins and of new West African gold ores. Changes in the Pt, Ga and Sb contents in the Near and Middle Eastern dinars showed that under the Abbasids, gold from Egypt—probably from Nubia, the Red Sea area or Arabia—was used together with a second gold ore from the North-Eastern Empire. In Egypt, the Fatimid coins have nonhomogeneous compositions that may be the result from the 969 reform and the re-cycle of the Abbasid (Byzantine gold type) dinars. In addition, another gold ore was used in Alexandria with a high Sn content. Considering the gold ore used to issue the Western dinars, they demonstrated that under the Aghlabids “a short-lived gold” was used, probably of Ifriqhiyan origin. In the mints located on the caravans’ gold
Fig. 11.35. Croeside.
554
Precious metals artefacts
routes, a West African gold was used. This gold also reached Spain and Sicily, but no signs of the distribution of Eastern gold within the Western Empire were observed. Dinars seem to have been made with a mixture of Egyptian and West African gold ores [75]. In all the measurements (on creseides, daric, double daric, staters of Philippe II and Alexander), a high fineness was observed. In their study of 70 gold coins of the Persian world from the reign of Croesus (560 –546 BC ) (see Fig. 11.35) to Alexander the Great (356– 323 BC ) (Fig. 11.36) special attention was devoted not only to the fineness but also to the trace content. Typical concentration levels of Pb (up to 1260 ppm), Pt (up to 2800 ppm), Pd (up to 350 ppm), As (up to 150 ppm), Sb (up to 20 ppm), Sn (up to 1960 ppm) were observed. Their limit of detection was around 0.1 ppm for all those trace elements.
Fig. 11.36. Coin of Alexander the Great.
555
G. Demortier
It is generally accepted that the first gold coins are attributed to Croesus King of Lydia whose fabulous wealth created one of the most legendary characters of history. One century later the kingdom of Lydia became a province of the Persian Empire that spread from Asia Minor to the Indus and from Egypt to the Danube. The hanker towards this empire having a rich mining resource gave rise to a golden age in antiquity under Alexander the Great. Only one coinage struck in this period (560 – 323 BC ) presented a debasement: the double darics attributed to the Babylonian mint. From the relative concentrations of these trace elements the re-use of Persian treasures to strike the Alexander’s type staters was investigated; this study showed the use of different sources of gold in the Macedonian mints. For the Near and Middle-Eastern Muslim dinars, the use of three different gold ores was suggested. 11.6.3
Gold coins from the new world
Gold coins from the new world have also been studied using PAA by the groups of Barrandon [77] and Guerra [78]; this last one completed the PAA measurements with LA-ICP-MS analyses. The trade in American precious metals is one of the most important aspects of the commercial contacts between Europe and the Americas from the 16th to the late 18th century, and even later. The fabulous quantities of silver and gold brought to Europe by the legendary Spanish “Flotas de Indias,” and later by the Portuguese, led to an extraordinary increase in demand for consumer and capital goods, and a notable accumulation of capital as the profits of overseas trade were invested in agriculture, mining and manufacture. Through the discovery of reliable ore tracers the impact of American gold production on European monetary stocks and issues could be studied in a quantitative manner [77]. For the analysis of Brazilian gold, the first step was to identify the ores’ tracers; it appeared that for the ore from Minas Gerais the element of interest is palladium; it is present at the 0.1% level in American ores but only at the 10 ppm level in European coinages before the arrival of the noble metals from America. Thus, the level of palladium traces in gold coins struck in Brazil is approximately 100 times higher than the pieces struck in Portugal before 1700 (Fig. 11.37). A coin from Lisbon struck in 1702 already shows a 10-fold increase from the earlier average palladium content of 11.4 ppm in 15 analysed pieces dated to 1690. It confirms the presence of American gold in Portuguese issues corresponding to the increase in gold
556
Precious metals artefacts
Fig. 11.37. Palladium in gold coins minted in Brazil and Portugal: X, Portugal; W, Brazil.
supply registered in Lisbon. The European nation mostly involved in trade with Brazil was Britain; this trade was a primary source of Britain’s prosperity. It has been estimated that Britain imported about 30% of Brazilian gold between 1700 and 1760 (Fig. 11.38).
Fig. 11.38. Palladium in gold coins minted in England: X, Britain; W, Brazil.
557
G. Demortier
In her study of Brazilian gold of the 16th century, M.F. Guerra [78] was not only interested in the Pt content but also in the abundance of other trace elements. South American artefacts typically contain characteristic elements such as platinum (in Colombia) and palladium – platinum (in Brazil). Thus, the most important elements to characterize the gold ores appear to be those that do not show precipitation behaviour in gold [79]. Except for coins struck in the Bahia mint, most of the analysed coins and ingots were made during the 18th and 19th century, but the gold ores were of the same geological type as those used in earlier periods. In coins from Mexico (four coins), Colombia (five coins) and Brazil (two coins), Guerra observed that the Colombian gold (mainly Popayan) has the highest concentrations of Pt, while Peruvian and Chilean gold (except for one coin) have low contents of Pt and fit with one Mexican group. The other Mexican group may also correspond to a Colombian ore or to a mixture of the first group with a Colombian ore. All these coins have a low content of Pd while a higher concentration was found for Brazilian ingots. Two coins struck in Bahia in the 18th century were made with a typical Brazilian gold while the gold struck at the end of the 17th century resembles the material used to mint Colombian coins. The fact that in Ouro Preto (black gold), native gold is associated with oxygen-bearing compounds of Pt, Pd and Fe explains the high content of Pt found in Brazilian gold ingots [80]. By using LA-ICP-MS technique previously employed for study of Muslim gold, other trace elements in Mexican, Peruvian, Colombian and Chilean coins were observed and revealed new information about the gold sources. Colombian gold contains more Rh and Ir but Ga, Ge and Cd may be used to characterize differences between Mexican, Colombian and Peruvian gold. Colombian gold contains more Sn and Sb. Coins from Mexico and Peru feature low quantities of platinum group elements but the content of Sn and Sb is higher. Brazilian coins of the 17th century are similar to Colombian gold if one refers to Pd and Pt contents, but they do not show a much higher content of Sn and Sb. In this respect, Guerra suggests that a mixture of Colombian and Mexican Peruvian ores was used [78,80,81].
11.7
CONCLUSIONS
In addition to the usual physical and chemical methods of analysis of precious metal artefacts available in laboratories belonging to museums and
558
Precious metals artefacts
restoration workshops, high-energy ion beams to induce X-ray and nondestructive tools using g-ray signals are now available. Proton accelerators to reach an energy ranging from 2 to 4 MeV are used today in about 20 laboratories over the world to investigate works of art and archaeological items. These new techniques have been applied at LARN to characterize gold items in their elemental composition and proven to be outstanding for the identification of soldering processes applied to Mesopotamian, Tartesic, Italian, Slovenian, Merovingian and Byzantine items. The analytical results have been interpreted on the basis of a new reading of ancient technological books. As gold items suffer low surface corrosion, even on periods of millennia, one can expect that the surface composition reflects the bulk composition, but this aspect has to be checked by using incident particles of different energies in order to probe several depths. Using incident protons of various energies we have shown that the depth profile of Au and Cu may be achieved over depths of several microns. X-ray emission induced by charged particle (PIXE) is more sensitive to light elements than to heavy ones and may be considered to be very complementary to XRF, which achieves a better sensitivity for elements giving rise to X-rays lower (but close) to the primary exciting X-rays. PIXE and XRF are then panoramic methods for a global approach of the composition of the artefact but are less impressive than selective analytical methods like ICP for the identification of traces. Nevertheless, proton irradiation of a secondary target could be used to selectively excite X-ray lines of trace elements without exciting X-rays of the matrix. PIXE and XRF are completely non-destructive in comparison with any ablation method, which requires sampling of a few milligram out of the item. By comparison with neutron activation analysis (NRA) or charged particle activation analysis (CPAA), PIXE and XRF do not introduce residual activity.
Acknowledgements The permanent collaboration of Y. Morciaux and Ch. Honhon at LARN is highly appreciated. This work has been made possible with the participation of many Ph.D. and graduate students. The enthusiastic contributions of B. Van Oystaeyen, D. Decroupet, Ch. Moore, S. Mathot, A. Gilson, N. Vandesteeme, F. De Cuyper, J.-L. Ruralcaba, D. Dozot, B. Simon, A.-M. Meyer and P. Coquay were essential to allow me to write this chapter.
559
G. Demortier
Data, documents and fruitful discussions are also acknowledged from M.-F. Guerra, A. Perea, F. Fernandez-Gomez, M. Respaldiza, Z. Gabelica, Z. Smit, N. Trampuz-Orel, J.-N. Barrandon, G. Nicolini, I. Montero, J.-C. Dran, Th. Calligaro, R. Van Laere, A. Gondonneav. I spare a thought for my late lamented colleague Th. Hachens, who introduced me to the field of Archaeometry in the late 1970’s. REFERENCES 1
2 3
4
5
6
7
8 9
10
11 12
13
560
J.C. Dran, Th. Calligaro and J. Salomon, Ion beam microanalysis. In: K. Janssens and R. Van Grieken (Eds.), Non-destructive Microanalysis of Cultural Heritage Materials. Elsevier, Amsterdam, 2004, Chapter 5. G. Demortier and T. Hackens, Milliprobe and microprobe analysis of gold items of ancient jewelry, Nucl. Instr. Meth., 197 (1982) 223 –236. Th. Jeffries, LA-ICP-MS and other-based micro-beam methods. In: K. Janssens and R. Van Grieken (Eds.), Non-destructive Microanalysis of Cultural Heritage Materials. Elsevier, Amsterdam, 2004, Chapter 7. B. Van Oystaeyen and G. Demortier, Matrix effects in PIXE evaluation of the major components in thick homogeneous samples, Nucl. Instr. Meth., 215 (1983) 299 –313. M. Amandry, J.N. Barrandon, C. Brenot, J.P. Callu and J. Poirier, L’affinage des me´taux monnaye´s au Bas-Empire: les re´formes valentiniennes de 364– 368, Estratto Dalla Rivista: Quaderni Ticinesi di Numismatica e Antichita Classiche (Lugano), 11 (1982) 279–285. M. Amandry, M. Dhe´nin, J. Lafaurie and C. Morrisson, Catalogue des Monnaies d’Or; Journe´es de la Socie´te´ Franc¸aise de Numismatique, Saint-Omer, 4–5 Juin 1983. Bulletin de la Commission De´partementale d’Histoire et d’Arche´ologie du Pas de Calais, nume´ro spe´cial, Arras, 1983, 141 pp. C. Morrisson, C. Brenot, J.N. Barrandon, J.P. Callu, J. Poirier and R. Halleux, L’or Monnaye´ I—Orifications de Rome a` Byzance, Cahiers E. Babelon No. 2, CNRS, France, 1985. Histoire Naturelle de Pline. Traduction et commentaires d’Ajasson de Grandsagne e´dite´ par C.L.F. Panchoncke en 1833. M. Berthelot, Introduction a` l’e´tude de la chimie des anciens et du Moyen-Age, Georges Steinheil, Paris (1889). Culture et Civilisation, Nouvelle e´dition, Bruxelles, 1966. A. Thouvenin, La soudure dans la construction des oeuvres d’orfe`vrerie antique et ancienne, Revue Arche´ologique de l’Est et du Centre-Est, France, 24 (1973) 11–68. H.A.P. Littledale, UK Patent 415 181, 1936. N.D. Meeks and P.T. Craddock, The detection of cadmium in gold/silver alloys and its alleged occurrence in ancient gold artefacts, Archaeometry, 33 (1991) 95 –107. see also Archaeometry, 34 (1992) 310 –312. G. Demortier, The detection of cadmium in ancient gold/silver alloys and its alleged occurrence in ancient gold solders, Archaeometry, 34 (1992) 305–310.
Precious metals artefacts 14 15 16 17 18 19
20
21 22 23
24 25
26
27 28 29 30 31
32 33
G. Demortier, LARN experience in non destructive analysis of gold artefacts, Nucl. Instr. Meth. B, 14 (1986) 152 –155. F. Fernandez Gomez, Tesoros de la Antiguedad en el Valle del Guadalquivir. Caja Sur Publications, Cordoba, 1997, Fig. 101. C. Blanco de Torrecillas, El tesoro del Cortijo de Ebora, Arch. Esp. Arqueologia, XXXII (1959) 50 –57. G. Nicolini, Technique des Ors Antiques. La Bijouterie Ibe´rique du IV Sie`cle, Paris, 1990. A. Perea, Orfebreria Prerromana. Arqueologia del oro. Caja Sur Publications, Madrid, 1991. M.A. Ontalba Salamanca, G. Demortier, F. Fernandez Gomez, P. Coquay, J.-L. Ruvalcaba-Sil and M.A. Respaldiza, PIXE and SEM studies of Tartesic gold artefacts, Nucl. Instr. Meth. B, 136– 138 (1998) 851 –857. G. Demortier, F. Fernandez-Gomez, M.A. Ontalba-Salamanca and P. Coquay, PIXE in an external microbeam arrangement for the study of finely decorated Tartesic gold jewelry items, Nucl. Instr. Meth. Phys. Res. B, 158 (1999) 275 –280. G. Demortier, La chrysocolle des orfe`vres est-elle jaune?, Archaeometry, 29(2) (1987) 275–288. G. Demortier, Ancient gold solder: what was chrysocolla?, Archaeological Chemistry IV. American Chemical Society, Washington, 1989, pp. 249 –263. G. Flouret, G. Nicolini and C. Metzger, Les bijoux Gallo-Romans de l’Houmeau (Charente-Maritime), Gallia, Fouilles et Monuments Arche´ologiques en France Me´tropolitaine, 39 (1981) 85–101. G. Demortier, Analysis of gold jewelry artefacts, Gold Bull., 17 (1984) 27–38. G. Demortier, Trace analysis of medium Z elements in narrow regions of a heavy matrix by XRF induced by PIXE, Proceedings of the XIth ICXOM conference, London, Canada. University of Western Ontario Press, London, Ontario, 1987, pp. 147–150. G. Demortier and A. Gilson, Determination of traces of light elements in gold artefacts using nuclear reactions, Nucl. Instr. Meth. Phys. Res. B, 18 (1987) 286 –290. H. Racklam, in: Loeb (Ed.), Pliny’s Natural History, Book 33, Heineman, London, 1968, pp. 4–72. H. Zehnacker, Pline l’Ancien, Histoire Naturelle, Livre 33, Les Belles Lettres, Paris, 1983, pp. 47– 85. K.C. Bailey, The Elder Pliny’s Chapter on Chemical Subjects. Arnold, London, 1929, pp. 59–109, see also pp. 175 –209. P.G.N. Glare (Ed.), Oxford Latin Dictionary. Clarendon, Oxford, 1976. G. Demortier, D. Decroupet and S. Mathot, Brazing and soldering of gold by soft metallurgy. In: G. Vermeylen and R. Verbeeck (Eds.), Proceedings of the International Precious Metals Institute, Brussels. IPMI Press, Allentown, PA, 1987, pp. 335– 352. L. Pauli, The Alps, Archaeology and Early History. Thames and Hudson, London, 1984, 161 pp. P. Koroec, A contribution to the understanding of the man’s big golden ring from the castle necropolis in Ptuj, Dissertations of the Faculty for Philosophic Sciences, Ljubljana, 1997.
561
G. Demortier 34 35 36
37
38 39 40 41
42
43
44 45
46
47
48 49 50
51 52
562
A. Pleterski, in: J. Dular and B. Aubelj (Eds.), Zakladi Tisoletij. Modrijan, Ljubljana, 1999, 370 pp. R. Koevi, Die Werksta¨tte kra¨ftig profilirter Fibeln in Siscia, Archaeologia Iugosl., 16 (1979) 51. Z. Smit, M. Budnar, P. Pelicon, B. Torko, T. Knific, J. Istenic, N. Trampuz-Orel and G. Demortier, Analyses of gold artefacts from Slovenia, Nucl. Instr. Meth. Phys. Res. B, 161 –163 (2000) 753 –757. A. Perea, I. Montero, G. Demortier and A. Climent-Font, Analysis of the Guarrazar treasure by PIXE, Ion Beam Study of Art and Archaeological Objects, EUR 19218, 2000, pp. 99 –101. A. Perea (Ed.), EL Tesoro Visigoto de Guarrazar. CSIC, Madrid, 2001, 403 pp. J. Bienaime´, Le tre´sor de Pouan au Muse´e de Troyes, Les cahiers du Muse´e de Troyes, no. 1. J. Vlaeminck, Childeric-Clovis, 1500e Anniversaire, Catalogue, Tournai, 1982. F. De Cuyper, G. Demortier, J. Dumoulin and J. Pycke, La croix Byzantine du Tre´sor de la Cathe´drale de Tournai, Aurifex 7, UCL Press, Louvain-la-Neuve, 1987, pp. 1–88. Charles de Liras, Orfe`vrerie Byzantine. Le reliquaire de la Sainte Croix au tre´sor de la Cathe´drale de Tournai. Magasin Pittoresque, Paris, 1885, pp. 100–102. J.L. Ruvalcaba-Sil and G. Demortier, Scanning RBS-PIXE study of ancient artefacts from South America using a microbeam, Nucl. Instr. Meth. Phys. Res. B, 130 (1997) 297 –302. W. Bray, Gold working in ancient America, Gold Bull., 11(4) (1978) 136 –143. W. Bray, Techniques of gilding and surface enrichment in pre-Hispanic American metallurgy. In: S. La Niece and P. Craddock (Eds.), Metal Plating and Patination. Butterworth Heinemann, Oxford, 1993, pp. 182 –192. D.A. Scott and E. Doehne, Soldering with gold alloys in ancient south America: examination of two small gold studs from Ecuador, Archaeometry, 32(2) (1990) 183 –190. A.M. Miller, Gold techniques in the new world in pre-Christian Era 2000 BC – 1500 AD . In: C. Elue`re (comp.) (Ed.), Outils et Ateliers D’orfe`vres des temps Anciens, Muse´e des Antiquite´s Nationales, Me´moire 2, Saint-Germain en Laye, France, 1993, pp. 255 –258. D.A. Scott, Depletion gilding and surface treatment of gold alloys from the Narino area of ancient Colombia, J. Historical Metall. Soc., 17(2) (1983) 99 –115. H. Lechtmann, A pre-Columbian technique for electrochemical replacement plating of gold and silver on copper objects, J. Metals, December (1979) 154 –160. F. Solis, M. Carmona-Macias, E. Carmona-Ortiz, Mobilier fune´raire des Zapote`ques et Mixte`ques, Catalogue of Exhibition, Europalia 93 Mexico, Brussels, 1993. J.L. Ruvalcaba-Sil, G. Demortier and A. Oliver, External beam PIXE analysis of gold prehispanic Mexican jewelry, Int. J. PIXE, 5(4) (1995) 273 –288. J.L. Ruvalcaba-Sil and G. Demortier, Non destructive analysis of American gold jewellery items by PIXE, RBS and PIGE, Applications of Particle and Laser Beams in Materials Technology, NATO-ASI series; Serie E: Applied Science, Vol. 283. Kluwer Academic Publishers, The Netherlands, 1995, pp. 463– 470.
Precious metals artefacts 53 54 55
56 57
58
59 60 61 62 63 64 65 66 67
68
69
70
71
G. Demortier and J.L. Ruvalcaba-Sil, Differential PIXE analysis of Mesoamerican jewelry items, Nucl. Instr. Meth. Phys. Res. B, 118 (1996) 352 –358. G. Demortier, Review of recent applications of high energy microprobes in arts and archaeology, Nucl. Instr. Meth. B, 54 (1991) 334. G. Demortier, Complementarity of RBS, PIGE and PIXE for the determination of surface layers of thicknesses up to 25 microns, Nucl. Instr. Meth. B, 49 (1990) 46. J.L. Ruvalcaba-Sil and G. Demortier, Elemental concentration profile of ancient gold artefacts by ion beam scattering, Nucl. Instr. Meth. B, 113 (1996) 275 –278. J.L. Ruvalcaba-Sil and E. Ortiz, Non destructive study of a gold pendant from Oaxaca, Mexico, using PIXE and RBS techniques. Br. Archaeological Rep. (in press). G. Demortier, Y. Morciaux and D. Dozot, PIXE, XRF and GRT for the global investigation of ancient gold artefacts, Nucl. Instr. Meth. Phys. Res. B, 150 (1999) 640–644. B. Simon, Graduate thesis, LARN, 1994. S. Scheers, J. Van Heesh and R. Van Laere, Franc Comme l’or, Imprimerie Banque Nationale de Belgique, 1991. G. Demortier, Analyse non destructive de monnaies par les techniques PIXE et XRF, Bull. l’Inst. Arche´ologique Lie´geois, t. CVII (1995) 371– 394. G. Ferreira and F.B. Gil, Elemental analysis of gold coins by PIXE, Archaeometry, 23(2) (1981) 189 –199. H. Mommsen and T. Schmittinger, Test analysis of ancient Au and Ag coins using high energy PIXE, Archaeometry, 23(1) (1981) 71 –76. G. Ambrosino and P. Pindrus, Non destructive analysis of ancient metal objects, Rev. Metall., 50 (1953) 136 –138. E.V. Sayre and P. Meyers, Nuclear activation applied to materials of art and archaeology, AATA, 8(4) (1971) 115 –150. P. Meyers, Non destructive activation analysis of ancient coins using charged particles and fast neutrons, Archaeometry, 11 (1969) 67 –84. F. Beauchesne and J.N. Barrandon, Analyse globale et non destructive des objets arche´ologiques cuivreux par activation avec des neutrons rapides de cyclotron, Rev. d’Arche´ome´trie, 10 (1986) 75 –85. J.N. Barrandon, J.L. Debrun and M. Hours, Application des reactions nucle´aires induites par des protons de moyenne e´nergie a` la caracte´risation des ors anciens, Academia Nazionale dei Lincei, Roma, (1976) 77– 85. A. Gondonneau and M.F. Guerra, The analysis of gold by ICP-MS with an UV laser. Application to the study of the gold currency of the Muslim West Africa. In: S.M.M. Young, A.M. Pollard, P. Budd and R.A. Ixer (Eds.), Proceedings of the Conference Metals in Antiquity, BAR international Series, Vol. 792. 1999, pp. 262 –270. R.J. Waltin, H.K. Herbert, D. Delev and I.D. Abell, Gold fingerprinting by laser ablation-inductively coupled plasma-mass spectrometry, Spectrochim. Acta, 49B(2) (1994) 205 –219. M.F. Guerra, C.O. Sarthre, A. Gondonneau and J.N. Barrandon, Precious metals and provenance enquiries using LA-ICP-MS, J. Archaeol. Sci., 26 (1999) 1101–1110.
563
G. Demortier 72
73
74 75 76 77
78 79 80
81
564
N.H. Gale and Z.A. Stos-Gale, Lead isotope studies in the Aegean. In: A.M. Pollard (Ed.), New Developments in Archaeological Science. Oxford University Press, The British Academy, Oxford, 1992, pp. 63 –108. A. Gondonneau and M.F. Guerra, The gold from Ghana and the Muslim expansion. A scientific enquiry into the Middle Ages using ICP-MS combined with an UV laser. In: S.M.M. Young, A.M. Pollard, P. Budd and R.A. Ixer (Eds.), Metals in Antiquity, BAR International Series, Vol. 792. 1992, pp. 202 –270. C. Roux and M.F. Guerra, La Monnaie Almoravide: de l’Afrique a` l’Espagne, Revue d’Arche´ome´trie, 24 (2000) 39 –51. M.F. Guerra and A.Gondonneau, Private communication. A. Gondonneau and M.F. Guerra, L’or perse a` travers l’analyse de monnayages antiques et me´die´vaux, Revue d’Arche´ome´trie, 24 (2000) 27–38. Ch. Morrisson, J.N. Barrandon and J.N. Morrisson, Or du Bre´sil, Monnaie et Croissance en France au 18e`me Sie`cle, Cahiers Ernest Babelon 7, CNRS Editions, Paris, 1999, pp. 1–220. M.F. Guerra, The circulation of South American precious metals in Brazil in the end of the 17th century (to be published). N.D. Meeks and M.S. Tite, The analysis of platinum-group element inclusions in gold, J. Archaeol. Sci., 7 (1980) 267 –275. J. Jedwab and J. Cassedane, Historical observations on oxygen-bearing compounds of platinum and palladium in Minas Gerais, Brazil, Can. Mineralogist, 36 (1998) 887 –893. M.F. Guerra, Characterisation of Gold Coins by Charged Particles and Mass Spectrometry Analysis. Founders Smiths and Platters, University of Oxford, Oxford, 2001, in press.
Chapter 12
Diagnostic methodology for the examination of Byzantine frescoes and icons. Non-destructive investigation and pigment identification Sister Daniilia, Sophia Sotiropoulou, Dimitrios Bikiaris, Christos Salpistis, Georgios Karagiannis and Yannis Chryssoulakis
12.1
INTRODUCTION
This chapter comprises two case studies coming from the Byzantine iconographic tradition. The first one is the scene of the Entry of the Mother of God into the Temple, which belongs to the wall paintings of the Protaton Church at Karyes, the capital of the Holy Mountain of Athos. The entire wall-painting programme of this church, still intact, was painted at the end of the 13th century (1295) by a legendary artist, Manuel Panselinos, chief representative of the Macedonian School of iconography [1– 4]. The second one is a portable icon of the Mother of God of Hodegetria type, which appertains to the iconostasis in the small church of Saint Modestos, at Kalamitsi, in the region of Chalkidiki and is representative of the Cretan style of iconography. Both paintings are of high artistic and historic value. The employment of non-destructive analytical methods has made feasible the revelation of the painting technique in its entirety, including the approach to drawing, the choice of materials, and their technique of application. At the same time, analytical methods have also brought to light whatsoever damage and change may have befallen over the passage of time. The findings of these investigations fill the scholarly lacunae in any number of issues related to traditional techniques of Byzantine iconography. Simultaneously, they provide essential information for art historians and conservators who are involved in tasks of restoration and cleaning. Comprehensive Analytical Chemistry XLII Janssens and Van Grieken (Eds.) q 2004 Elsevier B.V. All rights reserved
565
Sister Daniilia et al.
12.2 12.2.1
THE ENTRY OF THE MOTHER OF GOD INTO THE TEMPLE Description
The scene of the Entry of the Mother of God into the Temple (Fig. 12.1) occupies the second zone of the eastern wall in the central area of the church’s southern aisle. Its subject matter, depicted in a sequential and narrative fashion, allows one to distinguish, from left to right, the following groups: (a) Ss Joakeim and Anna, the parents of the Blessed Virgin; (b) a gathering of the seven maidens who lead the infant Mary into the Temple; (c) the Prophet Zacharias, standing before the Temple ready to receive the 3-year-old Mary; and, above and to the right, (d) the Archangel Gabriel who brings food to the child in the Holy of Holies. It is, therefore, a generously populated composition, one that sustains this particularly intense chain of events with expressive power, perfection in design, and colour harmony. As such, it has been held up as an exemplar in that it comprises the sum total of possible, available, pictorial data. Thus, this scene is considered to be representative of all other subjects in the edifice’s entire iconographic programme [5].
Fig. 12.1. The Entry of the Mother of God into the Temple, Manuel Panselinos. Wall painting, 2 m £ 3.5 m. Prior to cleaning and with the positions of the colour measurements indicated. ( For a colored version of this figure, see Plate 12.I.)
566
Diagnostic methodology for the examination of Byzantine frescoes and icons
12.2.2
The preparation of the plaster: materials and technique
According to common wisdom, Byzantine images on church walls were traditionally painted directly onto wet plaster. However, in view of the fact that plaster retains moisture for only a restricted amount of time, special care has been exercised to discover the manner in which Manuel Panselinos confronted this practical problem. Indeed, what came to light was his combined use of fresco and secco painting technique; the latter following the former once the plaster had dehydrated. In the second stage—that of dry work—it was noticed that slaked lime had been used both as white pigment and as a substance for binding. The application of analytical methods proved the existence of a modest quantity of a protein-based organic binder [6]. This offered a superior and more secure adhesion of the topmost paint layers to those below. The latter had been deposited onto the wet plaster with the aid of diluted lime wash. The overall thickness of the plaster was seen to vary from place to place, depending on the bulkiness of the masonry or on the existence of cavities therein. In the primary, coarsely granulated, rough layer of plaster, large crystals of marble scattered in lime were detected. By contrast, the second layer, finely granulated (0.5 –3 cm thick), was typically low in density but of high porosity. Its principal components were found to be lime (CaCO3) and straw; other secondary ingredients included quartz (SiO2) and albite (NaAlSi3O8), (Fig. 12.2). A high content of straw rendered the plaster more
Fig. 12.2. XRD Spectrum of the middle layer of plaster. (1) quartz (SiO2), (2) albite (NaAlSi3O8), (3) calcite: (CaCO3).
567
Sister Daniilia et al.
hydrophilic and resulted in a retardation of the drying process. The final layer, 0.5 –1 mm thick and consisting of lime, was finely granulated and of low porosity. Observation has confirmed that the Byzantine artists prepared this topmost layer with great care in order to create an extremely smooth and sleek surface.
12.2.3
The drawing
Careful eye observation reveals several drawing incisions, which the artist had used as a guide for his subsequent brushwork (Fig. 12.3). It is important to emphasise that these incisions had been cut into the wet plaster and did not constitute a completed preparatory design. Moreover, in that the artist declined to respect them precisely in the subsequent, final brush strokes, he has demonstrated impressive freedom. Notwithstanding the application of infrared reflectography [7 – 9], it was impossible to isolate the brushstrokes of the underdrawing; they were concealed by the finishing touches owing to the excessive use of carbon black, which absorbs infrared rays intensively (Figs. 12.4 and 12.6). That a primary drawing existed and was executed by brush is, however, beyond doubt. In rare instances where it is possible to verify small changes through the infrared reflectography, the artist’s preliminary underdrawing in brush was also registered [5].
Fig. 12.3. Detail of St Anna’s cloak. Photography with raking light. The incised lines of the preliminary drawing are visible.
568
Diagnostic methodology for the examination of Byzantine frescoes and icons
Fig. 12.4. Infrared reflectograms. Details of the faces.
The conduct of further examination in several cross-sections has affirmed the above hypothesis. Over the white plaster was discovered a thin paint layer, consisting chiefly of ochres (yellow ochre or red ochre, with or without a small quantity of carbon black granules). The artist’s selection of this shade had been determined by the hue of the underlay, which must not have hidden the original brush strokes of the drawing entirely (Fig. 12.14 and 12.15). With respect to the final stage of the drawing, its perceptible linear subtlety is a sign of that sovereign freedom and assurance of hand, which has infused Panselinos’ work with a wealth of movement and expression. In each figure, powerful expressions, concentrated in the gaze and in the contractions of the mouth, are also highlighted in bodily postures and in the movements of the folds of clothing (Fig. 12.4). This manner of drawing the features of the face and of the garment folds was a decisive element in the formation of the style of the Panselinos School in which the hallmarks of the master painter’s personal genius continually prevailed. Attention must be drawn to the fact that, beyond the artist’s general traits in portraying and shaping figures, his choice between conferring additional detail on the one hand, or to indicating a preference for stylistic simplicity in figural depictions on the other, depended upon two details: the location of the composition in the church building; and the role that each character performed in the specific scene.
569
Sister Daniilia et al.
12.2.4
Materials and painting techniques
12.2.4.1 Non-destructive investigation of pigments In identifying the nature of certain pigments, the contribution made by techniques of infrared reflectography and of fluorescence under ultraviolet light has been significant. Figure 12.5 demonstrates with extraordinary clarity the areas where use had been made of (a) minium recognized by virtue of its characteristic orange fluorescence, as a final light for the red garments, (b) azurite identified by its deep blue fluorescence, in the highlights of the garments of the Mother of God, of St Anna, and of the Prophet Zacharias, (c) yellow ochre noted for its characteristic yellow – brown hue under ultraviolet light, in the halos of the saints and in the maidens’ yellow garments. Infrared reflectograms furnish indications about the nature of those pigments that absorb characteristic amounts of infrared rays, whereas in some cases the gradations of lights and shadows are clearly revealed. Iron oxides, such as yellow and red ochres, can be identified owing to their very low absorption of infrared rays (Fig. 12.6). Yellow ochre was located in the Virgin’s halo and caput mortuum in the mantle. The clear reading of the gradations of the mantle’s shadows was possible because of their variegated amount of carbon black which amply absorbs infrared rays. The characteristic grey levels in the infrared reflectogram of the central maiden’s purple tunic indicate the use of: (a) carbon black in combination with lime for the lights; (b) a kind of iron oxide (probably caput mortuum) for the underlay and (c) the same pigment mixed with carbon black for the final shadows (Fig. 12.7).
Fig. 12.5. Fluorescence photography of the wall painting shown in Fig. 12.1. ( For a colored version of this figure, see Plate 12.II.)
570
Diagnostic methodology for the examination of Byzantine frescoes and icons
Fig. 12.6. Detail of the Mother of God. Infrared reflectogram.
12.2.4.2 Micro-analytical identification of pigments and study of layer structure (Table 12.1) Background and earth surface From what can be perceived in the painting today, dark grey has been rendered over the greater part of the background by means of two successive layers of differing shades of grey. Figure 12.8 shows the primary (light grey) layer, consisting of carbon black blended with lime, and the second (dark grey) layer, where carbon black dominates. Investigations into the wall painting’s cross-sections (from the outlines of the garments and from the letters of the inscriptions) have
571
Sister Daniilia et al.
Fig. 12.7. Centre: Detail of the Virgin’s tunic. Infrared reflectogram.
revealed the authentic and original layer structure of the background. Figure 12.9 exposes, below the white layer of lime used for the calligraphy of the inscription, a blue layer of heavily granulated azurite. This is a discovery of some significance since it leads to the conclusion that onto the dark grey (upper) coating over the background, a paint layer of pure azurite had been superimposed. Owing to the latter’s high grain size and deficient cohesion with the layers below; it became detached from the wall surface. This tactic of applying azurite over the grey layers served two purposes: first, it yielded a blue surface finish of high saturation, and secondly, it substantially reduced the overall cost of this expensive pigment which was otherwise used extensively over the entire compositional programme. The technique of painting employed for the green earth surface is analogous. The aforementioned two, grey-shaded background coats were also applied on the areas of the earth surface and, above them, coarsely granulated green earth was painted. Blue Predictably, the painting technique for the blue garments is the same as that for the background. A shade of grey (carbon black mixed with lime) formed the underlay while azurite was used exclusively for the lights; the principal lines were rendered in carbon black. Garments of light grey tonality do not contain azurite. Instead, all of the primary gradations were
572
Diagnostic methodology for the examination of Byzantine frescoes and icons TABLE 12.1 Pigments and layer structure of the Entry of the Mother of God into the Temple Paint area
Layer structure and pigments detected
Background
(i) Lime and carbon black, (ii) carbon black and lime, (iii) azurite (i) Lime and carbon black, (ii) carbon black and lime, (iii) green earth Underlay: (i) lime and carbon black, (ii) carbon black and lime; Light: azurite; Shadows: carbon black Underlay: red ochre; 1st light: cinnabar; 2nd light: minium; Shadows: red ochre and carbon black Underlay: warm ochre; Lights: warm ochre and lime; Highlight: lime; Shadows: red ochre Underlay: yellow ochre; Lights: yellow ochre and lime or carbon black and lime; Highlight: lime; Shadows: red ochre or yellow ochre and carbon black Underlay: green earth or green earth and lime or green earth and yellow ochre or carbon black and lime and green earth; Lights: green earth and lime or green earth and yellow ochre and lime or carbon black and lime or yellow ochre and lime; Highlight: lime; Shadows: green earth and carbon black or green earth and red ochre Underlay: caput mortuum or caput mortuum and lime; Lights: azurite or caput mortuum and lime or carbon black and lime or caput mortuum and carbon black and lime; Highlight: lime; Shadows: caput mortuum or caput mortuum and carbon black Underlay: green earth and yellow ochre and grains of lime; Flesh tones: yellow ochre and lime; Highlights: lime; Light transitory shade: green earth and lime and grains of yellow ochre; Warm transitory shade: red and yellow ochre and/or cinnabar; Shadows: (i) red ochre, (ii) red ochre and carbon black
Earth surface Blue garments
Red garments
Orange garments Yellow garments
Green garments
Purple garments
Flesh tones
573
Sister Daniilia et al.
Fig. 12.8. (A) Cross-section from the background. Photography via a microscope. Reflected light. (a) Lime and carbon black; (b) carbon black and lime ( For a colored version of this figure, see Plate 12.III); (B) Micro-Raman spectrum of carbon black.
rendered by means of a mixture of lime and varying amounts of carbon black. Noteworthy is the complete absence of pure blue apparel in the iconographic programme of the Protaton. Purple Examination of representative samples taken from the purple garments, themselves of diverse shades, has revealed mixtures that vary in the layers from the underlay up to the highlights. The advantage for the artist in employing this technique was that he could achieve a rich palette and
574
Diagnostic methodology for the examination of Byzantine frescoes and icons
Fig. 12.9. (A) Cross-section of a letter from the inscription. Photography via a microscope. Reflected light. (a) Lime and carbon black; (b) carbon black and lime; (c) azurite; (d) lime ( For a colored version of this figure, see Plate 12.IV ). (B) MicroRaman spectrum of azurite.
harmony of colour. The basic pigment was found to be caput mortuum, which emits a subtle, purple hue. In the underlay of the Virgin’s mantle caput mortuum has been recorded in its pure form, though in the highlights azurite was preferred (Fig. 12.10). For St Joakeim’s clothing, the underlay of caput mortuum was mixed with lime and the progressive gradations of light were rendered with ever-increasing quantities of lime. By contrast, the underlay and principal lines of the central maiden’s tunic were of a shade equivalent to that of St Joakeim, except for the lights which are now grey (a blend of carbon black and lime).
575
Sister Daniilia et al.
Fig. 12.10. (A) Cross-section of a light in the Virgin’s mantle. Photography via a microscope. Reflected light. (a) Caput mortuum; (b) azurite ( For a colored version of this figure, see Plate 12.V ). (B) Micro-Raman spectrum of caput mortuum.
Yellow and orange Yellow ochre has been the only yellow pigment to be found in the wall paintings of the Protaton. To achieve successfully an assortment of hues for the lights, the artist, in certain cases, drew on a mixture of yellow ochre and white, and in others, carbon black together with lime. Warm ochre could be detected in the orange robe of one of the maidens (Fig. 12.1, left hand corner). Its reddish principal lines revealed red ochre and in the gradations of the lights warm ochre was mixed with lime. Reds The painting of the red garments bears interest not only because of its relevance to the subsequent paint layers but also to the information that it
576
Diagnostic methodology for the examination of Byzantine frescoes and icons
Fig. 12.11. (A) Cross-section of a principal line at far left of the maiden’s tunic. Photography via a microscope. Reflected light. (a) Lime, carbon black and yellow ochre; (b) warm ochre and lime; (c) red ochre ( For a colored version of this figure, see Plate 12.VI ). (B) Micro-Raman spectrum of warm ochre.
gives concerning the garments’ current state of preservation. The incidence of flaking in the paint layers of the red garments immediately suggests deterioration in the layers’ proper coherence with the underlying plaster, something not to be seen in paintings set in wet plaster. In the cross-section of a sample removed from the region of a highlight (Fig. 12.12), observation has revealed that: (a) the underlay has red ochre; though quite thin this did not sink into the plaster; (b) for the first gradation of the lights, pure cinnabar was used while in the second and final lights, the artist employed minium; (c) the absence of lime from these layers indicates the existence of an organic binder to ensure proper adherence to the plaster. Thus, with respect to the red garments, it became clear that in order to preserve the saturation of their different shades, a commixture with lime had
577
Sister Daniilia et al.
Fig. 12.12. (A) Cross-section of a light in Zacharias’ tunic. Photography via a microscope. Reflected light. (a) Red ochre; (b) cinnabar; (c) minium ( For a colored version of this figure, see Plate 12.VII). (B) Micro-Raman spectrum of red ochre, (C) Micro-Raman spectrum of cinnabar. 578
Diagnostic methodology for the examination of Byzantine frescoes and icons
Fig. 12.13. (A) Cross-section of a light in tunic of central maiden. Photography via a microscope. Reflected light. (a) red ochre; (b) cinnabar; (c) altered minium ( For a colored version of this figure, see Plate 12.VIII). (B) Micro-Raman spectrum of minium, (C) Micro-Raman spectrum of altered minium.
579
Sister Daniilia et al.
been avoided in favour of the secco technique assisted by the application of a proteinaceous binder. The alteration of the minium’s final light from an orange to a grey –pink tone prompted more assiduous investigation and the removal of a sample from a region with pronounced alterations (Fig. 12.13). Indeed, it was possible to observe that the two first layers of red ochre and cinnabar had been preserved unspoiled, whereas the final exhibited an almost entire alteration of the minium into grey –brown granules. Since the peaks of the Raman spectrum corresponded with those of PbO2, this suggested a chemical change from Pb3O4 (orange-coloured minium) to PbO2 (black-coloured), a change provoked by the protracted effect of sunlight on the minium. Greens The only green pigment used by Panselinos was green earth, which he employed either in a pure form or mixed with lime and/or yellow ochre. The gradations of the lights in the garments progressed as far as lime white (Fig. 12.14), whereas in the principal lines, green earth was combined with carbon black. Dominating the grey – green shades was carbon black mixed with lime, in addition to a small quantity of green earth (as in the cloak of the first maiden to the right). Flesh tones Deserving of mention is the outstanding simplicity of the procedure implemented by the artist to draw the faces (Fig. 12.15). Their oil-green coloured underlay consisted of a mixture of thickly granulated green earth and yellow ochre. The flesh tones have been rendered in only one shade consisting of yellow ochre blended with lime. Here, however, the colour was applied in successive layers in such a fashion that the artist was able to build up to the desired density for the protruding points of the face. The effect, in these instances, is a sensation of light—a feeling that is ultimately reinforced by the white highlights of paint. Consistently in paintings by Panselinos, one can witness his extensive use of a light blue – green transitory paint layer between the shades of the underlay and the flesh tones. This paint layer was typically a mix of green earth and lime with a few grains of yellow ochre. In the red blushes of the cheeks and the lips, besides red ochre mixed with yellow and lime, a subtle touch of cinnabar was also introduced. Panselinos’ palette, aside from lime (a white pigment) and carbon black, included eight pigments in all [5]: (i) (ii)
580
Yellow ochre (Fe2O3·H2O), silica and clay; Warm ochre (Fe2O3·H2O and Fe2O3);
Diagnostic methodology for the examination of Byzantine frescoes and icons
(iii) (iv) (v)
Red ochre (Fe2O3), clay and silica; Caput mortuum (pure Fe2O3); Green earth (seladonite: hydrous iron, magnesium and aluminium potassium silicate); (vi) Azurite (2CuCO3·Cu[OH]2); (vii) Cinnabar (HgS); (viii) Minium (Pb3O4). 12.2.5
Study of the colour palette
This method has made it possible to account for Panselinos’ palette and to observe the harmony and complementarity of the various shades. In the ap bp colour diagrams (Fig. 12.16a and b), it may be observed that the white of
Fig. 12.14. (A) Cross-section of a light in cloak of the central maiden. Photography via a microscope. Reflected light. (a) Underdrawing: yellow ochre and grains of red ochre; (b) underlay: green earth and lime; (c) light: lime and green earth; (d) highlight: lime ( For a colored version of this figure, see Plate 12.IX ). (B) Micro-Raman spectrum of yellow ochre.
581
Sister Daniilia et al.
Fig. 12.15. (A) Cross-section from flesh tone in St Joakeim’s right foot. Photography via a microscope. Reflected light. (a) Underdrawing: red ochre and grains of carbon black; (b) underlay: green earth and yellow ochre; (c) flesh tone: yellow ochre and lime ( For a colored version of this figure, see Plate 12.X ). (B) Micro-Raman spectrum of green earth.
the plaster (pure calcium carbonate) had been selected for white. Its components are x ¼ 0:3466; y ¼ 0:3614 and Y ¼ 39:59 in the xyY CIE 1931 colour system and ap ¼ 1:13; bp ¼ 13:89; Lp ¼ 69:17 in the CIELAB 1976 colour system, for the light source D65. The features of Panselinos’ palette are: (a) the almost entire absence of colour measurements within the field Y–G – C (Yellow – Green – Cyan hues). It is worth mentioning that in the ap bp colour diagram (Fig. 12.16a), the palette in the Protaton painting of the Entry is similar with that of all 15 studied scenes in the church (Fig. 12.16b) [5]; (b) that the sum total of colours exhibits a low level of saturation; (c) the complementarity in several
582
Diagnostic methodology for the examination of Byzantine frescoes and icons
Fig. 12.16. (a) ap bp diagram (CIELAB 1976) of the colour measurements in The Entry of the Mother of God; (b) ap bp diagram (CIELAB 1976) of the colour measurements of all the 15 studied scenes in the same church. ( For a colored version of this figure, see Plate 12.XI.)
583
Sister Daniilia et al.
Fig. 12.17. x, y chromaticity diagram CIE 1931 of complementary colour combinations.
of the colour combinations (Fig. 12.17): between yellow –orange for halos and blue background; between yellow and blue garments; and, between green and purple garments. Briefly put, it may be proposed that colour analyses of Panselinos’ densely populated, painted scenes, such as this, bring to the surface the stylistic fingerprints of the artist. The characteristic measurements of the colour space within which the tonalities of a specific subject are located can with confidence and accuracy be considered as representative of the entire wall painting programme. 12.2.6
Conclusions
The subjection to scientific analysis of Manuel Panselinos’ representation of the Entry of the Mother of God into the Temple in the Protaton Church has yielded results representative of the artistic vocabulary of this great painter. This is an iconographic style embodying combination of fresco and secco technique. In the first phase, paint was applied to wet plaster for the length of time that it remained moist. Once dry, the details were completed
584
Diagnostic methodology for the examination of Byzantine frescoes and icons
with the aid of lime and the addition of a small quantity of an organic, proteinaceous binder. The preparatory drawing was initially scored by a few drawing incisions into the wet plaster, while the ensuing detailed underdrawing was rendered by brush and with pigments that varied from field to field in the wall paintings. Characteristic is the extraordinary facility of the artist, who completed his drawing with finishing brushstrokes that betray not even the slightest change of sweep. In its final stage, the drawing, typified by frugal but uninhibited and confident lines that span a variety of expressions and movements, created a sumptuous aesthetic effect. Ten pigments: yellow, warm and red ochres, caput mortuum, cinnabar, minium, green earth, azurite, lime white and carbon black were used over the entire painting. Predominantly minerals, they were applied either in a pure form or in simple mixtures with lime or black. On rare occasions, they can be seen combining two or more pigments. Initial paint layers were applied using the fresco technique: direct application onto the wet plaster and mixing the pigments with diluted lime wash. The painting was then completed by means of secco technique, this time with the use of lime and an organic, proteinaceous binder. Competence in balancing cool with warm tones and an elevated sense of colour harmony distinguish the deftness of the painter, who has shown himself capable of assembling a remarkably generous gamut of colour combinations. Panselinos’ complementary pairing-relationship of shades reveals his developed aesthetic sensibility. Relatively low levels of saturation are the result of the nature of the pigments employed and their frequent combination with white. A 7-hundred-year-old deposit of film, dating from the establishment of the monument, has diminished the lustre of the colours but without causing significant change in their hues. If one were to overlook the damage caused by constructional problems and some arbitrary flaking in the paint layers, the Protaton Entry of the Mother of God by Manuel Panselinos displays admirable resistance and cohesion by virtue of the technique of its painting. This monumental work impresses one by its perfection of design, the wealth of its palette, and the spirituality exuded by the illustrated figures. 12.3 12.3.1
MOTHER OF GOD HODEGETRIA Description
The icon of the Mother of God (Fig. 12.18) is one of the principal images on the iconostasis in the small church of Saint Modestos, at Kalamitsi,
585
Sister Daniilia et al.
Fig. 12.18. Mother of God Hodegetria. Egg-tempera on wood, 94 cm £ 69.5 cm £ 2.5 cm. Before cleaning. ( For a colored version of this figure, see Plate 12.XII.)
Chalkidiki, Greece. In keeping with the more austere portraits of this Hodegetria type, the Virgin is depicted on a gold background from the waist up, holding the Christ Child and turning slightly towards the right. Christ is portrayed in an erect frontal pose, making a gesture of blessing with his right hand and holding a closed scroll in the left. In the upper corners are miniatures of the archangels Gabriel and Michael in a suppliant pose. The Virgin wears a purple mantle edged with gold; Christ a deep bluish tunic and an orange himation with gilded lines.
586
Diagnostic methodology for the examination of Byzantine frescoes and icons
12.3.2
Construction and state of preservation of the support
The icon, a work typical of the Cretan style, is painted on a wooden panel 94 cm high, 69.5 cm wide and 2.5 cm thick. This support consists of three boards of similar height whose right and middle panels are of approximately the same width (31 –32 cm), while the left is 6 – 6.5 cm wide and narrows slightly at the upper end (Fig. 12.19a). For the more secure adhesion of these three wooden panels, four iron nails were used to grip the left to the middle while four others attach the
Fig. 12.19. (a) Rear side. Visible are the joints of the three wooden panels, the three battens and the white layer of the ground that was applied on the surface. The small rectangles with broken black lines demarcate the positions from which the fragments of wood were removed in order for the iron nails to be inserted between the joints. (b) Upper (left) and lower (right) sides of the icon. The texture of the wood at the upper end was exposed because of the excision of a narrow horizontal strip, which also included the wooden frame. The lower end was preserved unharmed. Visible are the remains of the white ground, which was applied on the sides. ( For a colored version of this figure, see Plate 12.XIII.)
587
Sister Daniilia et al.
middle panel to that on its right. The nails, inserted from left to right, incline slightly towards the facing side of the icon. Noteworthy is the fact that the nail heads cannot be seen from the rear. Close inspection of an X-radiograph (Fig. 12.20) reveals that a small piece of wood to the left of each nail head had been removed, a circumstance which reduced the total X-ray absorption in these places. It was not possible to locate these points simply by inspecting the board’s rear side, since a white layer of ground concealed any trace of detail. They were finally detected once information from both the X-radiograph and the macroscopic examination of the rear had been assembled. Small cracks in the white ground at the perimeter of the holes contributed to their recognition. This remarkable structural detail in the carrier had been introduced to prevent the formation of cracks along the joints of the three wooden panels. Three transverse battens were fixed with iron nails onto the icon’s rear side. The battens gripped the three wooden panels together and at the same time prevented warping, which is frequently provoked by changing climatic conditions. Minor damage to the paint layers has been created by virtue of very small vertical cracks along the joints of the board. Another vertical crack in the centre of the icon’s paint surface runs along virtually 2/3 of its total length and is visible to the naked eye. This crack coincides with the imaginary vertical line that descends from the nail heads that secure the middle and right wooden panels (Figs. 12.20 and 12.21), an observation that leads one to assume that it was probably made during the insertion of these nails. Of course, it is not out of the question that this fissure could have been created after the upper, horizontal strip of wood had been excised. A wooden frame, , 2.5 cm wide and 1 cm thick, was fixed onto the icon’s board with wooden nails whose heads are visible even to the naked eye. In its present state of preservation, these wooden nails hold the frame very delicately. The upper horizontal section of the frame has not survived since a piece, , 3.5 cm long, severed from the icon’s upper end (Fig. 12.21), has caused the most significant injury. Aside from the small cracks, the board is extremely well preserved; it exhibits neither warping nor woodworm damage. 12.3.3
State of preservation of the surface
Extensive damage and abrasions on the gold background, brought about during an attempt to clean off its soil, have significantly affected the icon’s aesthetic value. The flaking of red pigment in the inscriptions IC-XC, O VN
588
Diagnostic methodology for the examination of Byzantine frescoes and icons
Fig. 12.20. Composite X-radiograph (computerized synthesis with image enhancement). The carrier’s structural details are clearly recorded: battens, nails and fissures. Note the use of lead white in areas of high X-ray absorption.
589
Sister Daniilia et al.
Fig. 12.21. Schematic representation of the support’s structural details. The upper narrow horizontal band with the low grey tone demarcates the severed part of the icon. Two vertical lines (—) mark the joints of the three wooden panels while the vertical line that runs approximately along 2/3 of the height of the icon identifies the length of the crack from the central section of the wood. Visible on either side of the joints are the eight iron nails used to fix together the wooden sections. To the left of the nail heads are the small fragments of wood that were removed from the rear of the icon so that the iron nails could be inserted. On the upper batten, it is possible to detect the position of the two holes used to hang the icon. The texture of the canvas over the entire icon suggests its presence between the wood and the ground. Broken lines (- - - -) mark the inner contours of the wooden frame. Horizontal broad grey bands ( ) show the positions of the three battens over which the iron nails used for securing the joints are marked in black. The smaller, narrow grey band of the central piece of wood shows the spot from which the small piece from the rear of the board was removed (5 cm thick). Heavy, oblique lines ( ) denote the areas where, for unknown reasons, a thin slice of wood was removed from the upper and lower battens.
590
Diagnostic methodology for the examination of Byzantine frescoes and icons
has made them indecipherable, while the arms of the red cross in Christ’s halo are only barely visible. Damage to the surface paint layers of the faces, the right hand of the Mother of God and of Christ, and the mantle has been identified while extensive deterioration has also occurred to the paint layers of the Archangel Gabriel’s face and to a part of his cloak. This had probably been caused by some kind of solvent used to clean the gold background. The varnish layer has dissolved and at the same time the surface paint layers were removed, thereby making the underlying brushstrokes of the drawing of the eyes, nose and mouth on the Archangel’s face visible. This damage has been clearly recorded in the infrared reflectogram, owing to the minimal absorption of infrared rays from the remains of the paint layers (Fig. 12.22).
Fig. 12.22. Composite infrared reflectogram (computerized synthesis with image enhancement). Carbon black was used for the brushstrokes of the shadows, the principal lines of the mantle and, to a lesser extent, the folds of the other garments. Visible are the brushstrokes of the underlay of the faces and hands owing to the presence of some carbon black in their colour mixtures.
591
Sister Daniilia et al.
A thick layer of surface varnish, greatly contracted and yellowing, has significantly depreciated the icon’s aesthetic worth. In particular, it has altered the original blue hue of the tunics which, before the icon’s cleaning, had exhibited a dark greenish tone. The layer of accumulated soil has, to a great extent, reduced the degree of brightness both of the gold background and of the entire gamut of the icon’s colourings in general.
12.3.4
The ground
Because flaking and cracks formed on the ground during the removal of the frame and of the wooden strip at the upper end of the icon, the weave of the canvas could be seen even by the naked eye. This tells us that once the icon’s support had been constructed and the wooden frame fixed to it on the facing side, a linen canvas of medium texture was attached and sized with animal glue. Multiple layers of a mixture of gesso and animal glue, with a combined thickness of about 1 mm, had been applied to the facing side. Staining with Amidoblack (pH7) was used to trace the animal glue and it has also determined its distribution in the mixture: the quantity of glue is highest in the original layers, moderate in the middle ones, while in the final, finishing layers it increases significantly. A thin layer of gesso had been spread over the rear and on the sides of the wood (Fig. 12.19) in order to first smooth out the board’s irregularities, and secondly, to counteract the forces of contraction and expansion that had developed on the facing side during the application of the gesso. Finally, the ground of the facing side had been polished and burnished in order to create a suitable surface for the gilding and painting that followed. 12.3.5
The drawing
Naked-eye inspection of the icon reveals strong incised lines both in the contours of the figures and in the folds of the garments. On the faces, hands and feet the incised lines are smooth and can only be seen with difficulty. It becomes evident that the drawing had been transferred onto the white ground from a certain, prepared cartoon by incising its basic lines. These lines had been drawn with confidence and stability by the hand of an experienced and skilful artist.
592
Diagnostic methodology for the examination of Byzantine frescoes and icons
The fact that these incised lines could not be registered on the radiographic film indicates the limited use of X-ray absorbent pigments in the painting of the icon. Indeed, an examination of the X-radiograph (Fig. 12.20) makes it clear that lead white, which absorbs X-rays intensely, is absent from the folds of the garments and from the principal lines of the faces where drawing incisions were located. However, it had been used generously in the lights of the garments and of the faces, to a lesser degree in the underlays of the tunics of Christ and of the Virgin, as well as in the garments of the two Archangels. In the infrared reflectogram (Fig. 12.22) the folds of the Virgin’s mantle were recorded with surprising clarity and no significant changes in the drawing could be observed. The gradations of the tinnings and the principal lines are noticeable because of the differing amounts of carbon black used in their brushstrokes. Shadows had been rendered normally with broad brushstrokes and only in two areas of the mantle, on the right arm and below the Virgin’s left hand, could the presence of fine parallel brushstrokes (hatching) be identified. For the principal lines of the remainder of the garments, a small quantity of carbon black had been used. As a result the folds were faintly registered in the infrared reflectogram. For the drawing of the faces and hands a small quantity of carbon black had been employed and mixed with ochres which do not absorb infrared rays. On the other hand, the existence in the underlay of a moderate amount of carbon black in the mixture resulted in the rays’ absorption. As a result, the brushstrokes of the modelling were clearly marked on the infrared reflectogram. Despite the fact that the principal lines of the faces are not heavy, it is evident that the skilful painter rendered the drawing by thin single brushstrokes without changing the features of the faces in the final stage of the painting. A small change is observable only in the principal lines of the middle fingers of Christ’s left hand. Study of the X-radiograph and the infrared reflectogram does not reveal the slightest evidence of subsequent intervention in the painting of the icon. Rather, what has been revealed is the skill of the painter, who succeeded with few and certain strokes to create an inspired work. 12.3.6
Materials and technique of the painting
The results obtained from non-destructive methods of analysis—X-radiography, infrared reflectography, photography and macrophotography of the fluorescence under ultraviolet light—provide documentation for the authentication of the icon-painting of the Hodegetria Mother of God. As such, a
593
Sister Daniilia et al.
study and spectroscopic analysis of selected cross-sections has been made in order to determine the materials used and to reveal the entire structure of its paint layers [10 – 13]. The value of this information lies precisely in the fact that the icon under examination constitutes an excellent example of sacred portraiture from the Cretan period. Its select palette mostly comprises earth pigments: yellow ochre, caput mortuum, and green earth. Moreover, natural cinnabar, azurite, and blue indigo (the latter of plant origin used widely in that period) were also employed. The presence of a red lake (most probably cochineal) was found in the Virgin’s mantle as a glaze. Finally, carbon black and lead white complete the simple palette of the seven basic pigments in the icon (Table 12.2).
TABLE 12.2 Pigments and layer structure of the original painting of the icon the Mother of God Hodegetria Paint area
Layer structure and pigments detected
Ground Gold background Red border Virgin’s mantle
Gesso and animal glue (i) Yellow bole, (ii) gold leaf Cinnabar Underlay: caput mortuum, red cochineal(?) lake and carbon black; Lights: (i) caput mortuum, red cochineal(?) lake, lead white and grains of carbon black, (ii) lead white; Tinning: a glaze of red cochineal(?) lake Underlay: indigo and grains of lead white; Lights: azurite and grains of lead white Underlay: azurite and lead white; Lights: (i) lead white and azurite, (ii) lead white Underlay: yellow ochre and cinnabar; Gilt lights: (i) mordant: boiled oil (?) and carbon black, (ii) gold leaf Underlay: red cochineal(?) lake lead white and grains of azurite; Lights: lead white and red cochineal(?) lake Underlay: yellow ochre, carbon black, cinnabar, grains of green earth and of lead white; Flesh tones: lead white, cinnabar and yellow ochre
Virgin’s tunic Christ’s tunic Cloak of Christ—Braids of the Virgin’s mantle Cloak of Archangel Michael
Flesh tones
594
Diagnostic methodology for the examination of Byzantine frescoes and icons
12.3.6.1 Gold background and red border Once the drawing had been transferred to the white ground, it was deeply incised with a pointed tool in order to demarcate the contours of the figures from the background. The circles of the halos were engraved and sprung on the gesso by means of a compass. Following this, a thin layer of yellow bole mixed with animal glue was applied onto the ground, after which gold leaf was attached to the bole and burnished with agate. A fine layer of cinnabar had been employed for the red inscriptions and for the border, where a subsequent intervention in the form of a new layer of cinnabar and varnish was detected. Because no overpainting was found on the figures, this small intervention on the icon’s red border seems unjustifiable. Nevertheless, it did not alter significantly the original painting. 12.3.6.2 The Virgin’s mantle The oxidized varnish and the dirt layer have significantly altered the various tonalities of the lights and, as a result, the intermediary subtle gradations cannot be distinguished by the naked eye. An underlay of caput mortuum, with the characteristic peaks in its Raman spectrum at 224, 291, 409, 490 and 611 cm21 (Fig. 12.23C-1), a small quantity of carbon black, with two broad and strong peaks at 1313 and 1590 cm21 (Fig. 12.23C-2), and a red lake (probably cochineal) had been applied over the white ground (Fig. 12.23A). A mixture of caput mortuum, red lake and lead white with its characteristic peak of – CO22 3 vibration had been used for the lights, [14]. Since the grains of lead white are very small, the peaks of caput mortuum in the Raman spectrum are also recorded (Fig. 12.23C-3). The final heavy brushstrokes of the highlights are of pure lead white. Once all of the gradations in the mantle had been painted, the artist applied a glaze of red lake over the entire garment in order to achieve the desired violet hue and to unite the strong tonal contrasts between the lights and the principal lines. This red lake coating was exposed after microscopic examination of the cross-section under a source of ultraviolet light. The difference between the fluorescence of the red lake (layer e) and that of the superimposed sandarac varnish (layer f) facilitated the detection of both. This technique is encountered, almost exclusively, in the painting of the Virgin’s mantle during the acme of Cretan art. Like all organic materials, red lake is very difficult to identify through the use of micro-Raman spectroscopy due to its high fluorescence, but such problems do not arise in micro-FTIR spectroscopy where this paint layer produces a spectrum with characteristic peaks (Fig. 12.23D-2). However, the most intense are those of the methylene groups at 2917 and
595
Sister Daniilia et al.
Fig. 12.23. Cross-section from a light in the Virgin’s mantle. (A) Photography via a microscope. Reflected light. (a) Gesso ground; (b) underlay: caput mortuum, red cochineal(?) lake and carbon black; (c) 1st light: caput mortuum, cochineal(?) lake, lead white and grains of carbon black; (d) 2nd light: lead white; (e) tinning: a glaze of red cochineal(?) lake; (f ) sandarac varnish. (B) Fluorescence under ultraviolet light ( For a colored version of this figure, see Plate 12.XIV ), (C) Micro-Raman spectra. (D) Micro-FTIR spectra.
2845 cm21, the amide carbonyl group (– CO –NH – ) at 1640 cm21, and the stretching of the -NH- group at 3288 cm21, attributed to a protein binder [15]. Egg yolk had undoubtedly been used for this purpose. Two other peaks do not belong to the binder: a stronger at 1032 and the shoulder at 1110 cm21. These two peaks are believed to correspond to the red lake. FTIR spectrum of the icon’s preparation layer can be seen in Fig. 12.23D-1. Gypsum, with its highly characteristic peaks at 1112 cm21 (corresponding to sulphate groups [ – (SO4)2]) and at 3550, 3397 and 1620 cm21 (corresponding to water molecules), had also been used [16]. Methylene peaks at 2919 and 2844 cm21 (as compared with others of low intensity: between 1300 – 1800 cm21) originated from an organic material in the preparation layer, probably animal glue. 12.3.6.3 The tunics of Christ and of the Virgin In the surface image, before its cleaning, the blue shade of the tunics of Christ and of the Virgin demonstrates great change due to yellowing from the
596
Diagnostic methodology for the examination of Byzantine frescoes and icons
oxidized varnish. Moreover, the dirt layer has significantly diminished the brightness of the colours and, as a result, they have assumed an almost neutral, slightly greenish tone. For the underlay of the Virgin’s blue tunic, the artist had made use of a mixture of blue indigo with a small amount of lead white (Fig. 12.24A), pigments which absorb infrared rays minimally (Fig. 12.22). The indigo’s existence was also identified by micro-Raman spectroscopy and in its spectrum can be seen two strong peaks at 250 and 545 cm21 (Fig. 12.24C-1). For the lights, instead of indigo, azurite had been employed mixed with a small amount of lead white. The few scattered grains of yellow and red ochre are natural admixtures of azurite. On the sample’s surface, an extremely thick layer of sandarac varnish can be seen. In the course of the examination, the layer structure of a cross-section from Christ’s blue tunic was exposed (Fig. 12.24B). Since the sample had come from a border position, the underlay of the adjacent mantle could also be detected in the cross-section (layer b). In the underlay of the blue tunic, a mixture of azurite, with its characteristic peak at 398 cm21 (Fig. 12.24C-2), and lead white is present. For the gradations of the lights the content of lead white had been increased and in the final brushstrokes of the highlights it is found in a pure form.
12.3.6.4 Christ’s cloak and the gold braids of the Virgin’s mantle The distinctive use of gilt lines in Christ’s garments and on the gold braids and stars of the Virgin’s mantle as well as on the wings of the two supplicating angels is remarkable. Here, our talented painter achieved a tonal balance between the gold background and the painted surface. The orange hue of the gilded garments’ underlay has resulted from a mixture of yellow ochre, with a strong peak at 403 cm,21 and a high quantity of cinnabar, with two characteristic peaks at 253 and 343 cm21 (Fig. 12.25). For the adhesion of leaf for the gilded lines a particular kind of mordant (probably boiled oil), in which grains of carbon black are mixed, have been applied. Thanks to the characteristic fluorescence exhibited by this mordant, the layer has been recorded in the sample’s photograph under ultraviolet light. Its composition, however, could not be determined because it was applied very thinly. Grains of carbon black had not been added to this substance in order to accelerate the drying process, given that this pigment has no siccative quality. It may have acted as a guide for the painter, since it
597
Sister Daniilia et al.
Fig. 12.24. (A) Cross-section from a light in the Virgin’s tunic. Photography via a microscope. Reflected light. (a) Gesso ground; (b) underlay: indigo and grains of lead white; (c) light: azurite, grains of yellow ochre, of red ochre and of lead white; (d) sandarac varnish. (B) Cross-section from a light in Christ’s tunic. Photography via a microscope. Reflected light. (a) Gesso ground; (b) underlay of the mantle: caput mortuum, red cochineal lake (?) and carbon black; (c) underlay of the tunic: azurite and lead white; (d) 1st light: lead white and azurite; (e) sandarac varnish ( For a colored version of this figure, see Plate 12.XV ). (C) Micro-Raman spectra.
rendered visible the brushstrokes of the mordant, which is otherwise known for its high transparency. In the photograph of the same cross-section under ultraviolet light, the gold leaf appears as a thin, dark line over which had been coated an
598
Diagnostic methodology for the examination of Byzantine frescoes and icons
Fig. 12.25. (A) Cross-section from a gilt line in Christ’s cloak. Photography via a microscope. Reflected light. (a) Gesso ground; (b) underlay: yellow ochre and cinnabar; (c) mordant: boiled linseed oil (?) and carbon black; (d) gilt line: gold leaf; (e) sandarac varnish. (B) Fluorescence under ultraviolet light ( For a colored version of this figure, see Plate 12.XVI). (C) Micro-Raman spectra.
599
Sister Daniilia et al.
extremely heavy layer of sandarac varnish, itself displaying its typically high fluorescence. 12.3.6.5 Flesh tones The warm, dark, olive-green colour of the underlay was the result of mixing yellow ochre, carbon black, cinnabar and green earth (Fig. 12.26A). Yellow ochre is hydrous ferric oxide with high amounts of aluminium-silicate
Fig. 12.26. (A) Cross-section from a flesh tone in Christ’s right foot. Photography via a microscope. Reflected light. (a) Gesso ground; (b) underlay: yellow ochre, carbon black, cinnabar, grains of green earth and of lead white; (c) flesh tone: lead white, cinnabar and yellow ochre; (d) highlight: lead white; (e) sandarac varnish ( For a colored version of this figure, see Plate 12.XVII). (B) Micro-FTIR spectra.
600
Diagnostic methodology for the examination of Byzantine frescoes and icons
kaolinite. The characteristic – OH peaks of yellow ochre at 3689 and 3618 cm21, with the strongest at 1101 (Si – O), 1030 (Si – O – Si), 1005 (Si – O – Al) and 910 cm21, are recorded in its FTIR spectrum (Fig. 12.26B) [17,18]. In the same spectrum can be seen the peaks of the green earth at 956 and 3560 cm21 and the characteristic peaks of the proteinaceous binder. A few grains of lead white can also be detected in this mixture. For the initial gradations of the flesh tones a mixture of lead white, cinnabar and yellow ochre had been employed, while in the final highlights there was evidence of pure lead white. Conspicuous in this cross-sectional layer structure is the use of a high quantity of cinnabar in the paint layers both of the underlay and of the flesh tones. Furthermore, the difference in thickness between the underlay and the superimposed gradations of the flesh tones is characteristic. In these ways the painter was able to create an opaque, basic colour tone over which he painted the flesh tones in fine, tight layers. Volume in the faces had been achieved by the application of thin white brushstrokes at the most protruding points; these were marked spectacularly in the X-radiograph. 12.3.7
Conclusions
The importance of the present study rests on the documentation which it provides authenticating the icon’s painting. As such, an examination of its materials and detailed readings of the painting’s cross-sectional layer structure are of great importance, given that the icon constitutes a representative example of ecclesiastical artwork in the Cretan style. The exquisite drawing, its astonishing dynamism in the faces and in the movements reveals a mature and inspired artist. One is also impressed by the distinctive use of gilded lines in the garments that creates tonal harmony with the gold background and is counterbalanced by the thin white brushstrokes of the lights and highlights in the faces. Characteristic of works from this period are the finely wrought brushstrokes of the final lights and of the principal lines, the limited coverage of flesh tones in the faces, as well as the strong tonal contrasts between the underlays and the lights. The palette consists of natural pigments of medium granularity. The mixtures are simple and the paint layers have been evenly applied. Lead white is dominant in the finely worked highlights. Undoubtedly, this rare work of great value belongs to an artist from the peak period of Cretan art.
601
Sister Daniilia et al.
APPENDIX A12.1
EXPERIMENTAL DETAILS
A12.1.1 Non-destructive analysis Photographs in the visible area of the spectrum were taken through the use of two different acquisition systems: (A)
(B)
In the case of the Protaton wall-painting a SINAR 4 £ 500 camera with two electronic flashtubes ,5500 K was used on a KODAK EPR 64 ASA daylight, code N8 6117, colour reversal film. Colour slides were digitized using a SCANVIEW Scanmate 5000 scanner, supplied with the 3.33 colour Quartet acquisition and processing software. In the case of the portable icon a HASSELBLAD 205 FCC camera coupled with a digital back and a illuminating system comprising 2 £ 6 daylight lamps were used.
Moreover, for colour control requirements during the capture and digitizing procedure, the paintings were photographed together with reference tables comprising square-shaped (2 cm £ 2 cm) and uniformly painted colour patches. These tables were prepared in the laboratory with a wide range of pigments, frequently used in Byzantine iconography, and an egg yolk binder. Their colorimetric co-ordinates were measured and well known. X-radiographs of the icon were obtained by means of an ANDREX BW SMART 160 X-ray portable unit (0.5 – 6 mA, 10 –160 kV), supplied with a EP-100 dosimeter, on a KODAK MX-Industrex film, which was then developed using a KODAK M35 X-ray film processor. X-radiographs were taken at 33 kV, 6 mA and exposure times of 1 min 50 s. Infrared reflectograms were taken using a HAMAMATSU 2400-03D Precision TV Infrared camera equipped with a Vidicon infrared tube and a NIKON F1,2/50 mm lens, in front of which an interference filter at 1800 nm was placed. Image capture and digitization procedure was controlled through special software. Ultraviolet fluorescence photographs were taken using two different systems: (A) (B)
602
In the case of the Protaton wall-painting: a SINAR 4 £ 500 camera and an illuminating system of 2 £ 4 ultra-violet lamps. In the case of the icon: a HASSELBLAD 205 FCC camera and an illuminating system comprising four high-pressure mercury lamps.
Diagnostic methodology for the examination of Byzantine frescoes and icons
In both cases a KODAK E-100S, 100 ASA-daylight colour reversal film was used. In front of the camera lens a KODAK 2E WRATTEN gelatine filter was placed. Colour measurements were taken on homogeneous spot areas of 4 mm in diameter, under diffuse reflectance and d/28 geometry, using a MINOLTA CM-2022 portable spectrophotometer. Colour measurement representation was realised on (a) Reflectance percentage (R%) spectra, and (b) ap ; bp ; CIELAB 1976 colour diagrams.
A12.1.2 Micro-sampling analysis Special attention was paid to obtaining a representative sampling of the artefact under consideration and to avoiding further irreversible damage to it. Samples were mounted in polyesteric transparent resin so that their cross-sections would provide, after grinding and polishing using a STRUERS PLANOPOL-V machine, all relevant information from the existing stratigraphy. Cross-sections were observed under a LEICA DM RXP research polarising microscope equipped with a quartz halogen and an ultraviolet excitation light source as well as an automatic photography device. Raman spectra were collected on a RENISHAW System 1000 Raman spectrometer comprising an OLYMPUS BH-2 imaging microscope, a grating monochromator and a charge-coupled device (CCD) Peltier-cooled detector (576 £ 384 pixels). The incident laser excitation was provided by an aircooled He – Ne laser source, operating at 632.8 nm. The power at the exit of a £100 objective lens was varied from 1 to 3 mW, depending upon the stability of the pigments identified. Spectra were recorded with a resolution of 4 cm21, at a collection time of 30 s, and after an accumulation of 10 scans. In order to avoid undesirable Rayleigh scattering, two 100 cm21 notch-filters were employed to cut off the laser line. Pure silica was used for the calibration of the instrument. FTIR spectra were obtained using a Biorad FTS-45A FTIR spectrometer, connected to an UMA 500 microscope and equipped with a mercury cadmium telluride (MCT) detector cooled by liquid nitrogen. For the purpose of analysing the different components in every paint layer of the samples collected, a small amount from each was removed using a micro-scalpel. The particle was then placed on the surface of a freshly prepared KBr pellet. For each spectrum 250 consecutive scans were recorded with a resolution of 4 cm21 using a £15 objective lens. The sample collection area was adjusted
603
Sister Daniilia et al.
with the upper aperture of the microscope. All spectra were collected in the transmission mode and converted afterwards into absorbance spectra. As background, the spectrum of the KBr pellet was used. X-Ray diffraction analysis was performed trough the use of a GANTOLFI chamber with a Cu tube.
REFERENCES 1 2 3 4 5 6
7 8 9 10 11 12 13 14 15 16 17 18
604
M. Achimastou-Potamianou, Ekdotiki Athinon, (1994). I.M. Chatzifoti, Ethniko Idruma Neotitos, (1995). Dionysius of Fourna, Oakwood Publications, 1996. D.C. Winfield, Dumbarton Oaks Papers, 1968. Sister Daniilia, S. Sotiropoulou, D. Bikiaris, Y. Chryssoulakis, C. Salpistis, G. Karagiannis, B.A. Price and J.H. Carlson, J. Cultural Heritage, 1 (2000) 91. A. Tsakalof, K. Bairachtari and Y. Chryssoulakis, 2nd Conference on Instrumental Methods of Analysis (IMA—2001), 5 –8 September 2001, Ioannina, Greece. A.I. Kosolapov, ICOM Committee for Conservation, 6th Triennial Meeting, Ottawa, 1981. Van Ansperen De Boer, Studies Conserv., 14 (1969) 96. J.K. Delaney, C. Metzger, E. Wamsley and C. Fletcher, 10th Triennial Meeting, Washington DC, USA, 22-27/08/1993, (1993), 15. I.M. Bell, R.J.H. Clark and P.J. Gibbs, Spectrochim. Acta, 53A (1997) 2159. J. Pilc and R. White, Natl Gallery Technical Bull., 16 (1995) 73. R.J.H. Clark, Chem. Soc. Rev., 24 (1995) 187. D. Bikiaris, Sister Daniilia, S. Sotiropoulou, O. Katsimbiri, E. Pavlidou, A.P. Moutsatsou and Y. Chryssoulakis, Spectrochim. Acta, 56A (1999) 3. A. Degen and G.A. Newman, Spectrochim. Acta, 49A (1993) 859. R.J. Meilunas, J.G. Bentsen and A. Steinberg, Studies Conserv., 35 (1990) 33. S. Sotiropoulou, Sister Daniilia, D. Bikiaris and Y. Chryssoulakis, Revue d’Archeometrie, 23 (1999) 79. R. Prost, A. Damene, E. Huard, J. Driard and J.P. Leydecker, Clays Clay Miner., 37 (1989) 464. S. Shoval, B. Champagnon and G. Panczer, J. Thermal Anal., 50 (1977) 203.
Chapter 13
The provenance of medieval silver coins: analysis with EDXRF, SEM/EDX and PIXE Robert Linke, Manfred Schreiner, Guy Demortier, Michael Alram and Heinz Winter
13.1
INTRODUCTION
For the analysis of cultural-heritage materials, X-ray fluorescence analysis (XRF), scanning electron microscopy (SEM) with energy dispersive X-ray spectrometry (SEM/EDX) and proton induced X-ray emission spectroscopy (PIXE) hold an important position in the field of non-destructive techniques. The chemical characterization of the material composition facilitates the determination of their provenance [1], age [2], technology of production [3], intended purpose and, therefore, also their authenticity [4,5]. For objects made from precious metal(s), the concentration levels of the basic constituents can be used for drawing conclusions concerning the economic situation of an area or a period [6]. Additionally, these techniques have a great potential in the field of conservation science. The identification of the chemical composition of an object or the corrosion products enables the development of conservation procedures and helps to preserve objects that are endangered by decay [7,8]. Within the last decades a great number of papers have been published dealing with the analysis of metals [9,10], pigments [11– 13], paper [14], glass [15,16] or ceramics [17,18] usually by applying a combination of these techniques. In the field of coin analysis, non-destructive techniques frequently applied are neutron activation analysis (NAA) [19], energy dispersive X-ray fluorescence analysis (EDXRF) [20– 23], PIXE [24 –28] and SEM/EDX [29– 31]. Apart from the fact that the analyses do not result in any visible damage of the object being investigated, XRF, SEM/EDX and PIXE have also the considerable advantage of being multi-elemental, which enables a simultaneous determination of all elements present. Compared to techniques such Comprehensive Analytical Chemistry XLII Janssens and Van Grieken (Eds.) q 2004 Elsevier B.V. All rights reserved
605
Robert Linke et al.
as inductively coupled plasma mass spectrometry (ICP-MS) or atomic absorption spectrometry (AAS) also elements that are not expected can be detected. Further, trace elements as well as main components may be determined within one single measurement if excited with photons (XRF), electrons (SEM/EDX), or protons (PIXE). In general, such analyses are fast, partly because nearly no sample preparation is required. The detection power, precision and accuracy of these methods have already been compared by Klockenka¨mper et al. [32]. In this chapter, we highlight the (dis)advantages of XRF, SEM/EDX and PIXE applied to the analysis of objects of art and archaeology by using medieval silver coins as examples. In these techniques, the atoms of the objects are excited with particles of varying energies and physical properties. Thus, the main difference among these techniques becomes apparent when considering the penetration depth of the primary radiation as well as the dependence of the information depth on the chemical composition of the sample. Therefore, when quantitative information needs to be obtained from objects of art or archaeology, the state of preservation is of great importance. In what follows, objects of investigation are the “Friesacher Pfennig” and the “Tiroler Kreuzer,” both Austrian medieval silver coins that are part of the collection of the Mu¨nzkabinett (Coin Cabinet) of the Kunsthistorisches ¨ sterreichische Museum Vienna (KHM) and the Currency Museum of the O Nationalbank, respectively. 13.2 13.2.1
THE FRIESACHER PFENNIG Introduction
Increasing trade at the beginning of the 12th century AD in Europe led to the introduction of new currencies with a wider geographical range of validity. One of these new coinages in the area of Carinthia, a southern part of Austria, was the Friesacher Pfennig. As a result of its economic success, its area of circulation extended from Carinthia to North Italy, over Croatia and Hungary up to Slovakia and southern parts of Germany. The considerable demand for this new currency led to the establishment of a number of different mints and even to the production of contemporary imitations. Mints at that time were located in the towns of Laufen, Salzburg, Friesach, St. Veit and Aquileia. Figure 13.1 shows a map of the area of circulation and the mints within the European borders of today. The term “Friesacher Pfennig” (denarius Frisacenses) refers to all coins that were produced in the same style by different sovereigns and in different locations.
606
The provenance of medieval silver coins
Fig. 13.1. The mints of Laufen, Salzburg, Friesach, St. Veit and Aquileia within today’s borders of Central Europe.
Coins of the early period (approx. 1125/1130 – 1168), which were struck under the archbishops of Salzburg, the dukes of Carinthia, and the patriarchs of Aquileia [33] are particularly unique or at least very rare, and of a high value but have no marks or legends, by which numismatists can determine the coins’ mint of origin. Therefore, non-destructive analysis by EDXRF was carried out in order to enable geographical or chronological assignments and to support and confirm the numismatic findings carried out so far. By comparing the results of already classified coins with those of uncertain origin, conclusions could be drawn concerning the mint. As the analyses had to be carried out completely non-destructively it also was not allowed to induce radioactivity temporarily, trace element analysis by NAA or lead isotope ratio determinations by means of ICP-MS could not be performed. An essential requirement for the geographical assignment of coins according to their elemental composition is that a clear material distinction exists between the mints that are involved. Therefore, in the case of the Friesacher Pfennig, it was of considerable advantage that nearly each of the mints that needed to be taken into consideration had its own adjoining silver or lead mine. As will be shown below, also differences in the technology of
607
Robert Linke et al.
silver production and of silver refinement between the mints could be used for the identification and clustering. In similar investigations, carried out on mintage of later periods, e.g., the “Wiener Pfennig,” an Austrian coinage from the 12th to the 14th centuries, difficulties were encountered with the geographical and chronological assignment. These difficulties were caused by the fact that this mintage was produced from silver from different sources and therefore did not fulfil the requirements mentioned previously. The Friesacher Pfennig coins investigated here belong to the collection of the KHM, where a new inventory has been under preparation [34,35]. Coins from the mints of Salzburg, Laufen, Friesach, St Veit, and Aquileia were examined. Additionally, coins from the mint of Cologne were compared with Austrian coins due to similarities in style and fabrics (see section 13.2.3.1). Figure 13.2 shows the obverse and reverse sides of coins minted by Archduke Bernhard, Archbishop Adalbert II(?), and Archduke Ulrich II. 13.2.2
Experimental
13.2.2.1 Energy dispersive X-ray fluorescence analysis Most of the coins investigated were in a used but cleaned condition. In order to remove traces of microcrystalline wax that was applied for conservation reasons by the conservator of the KHM, the objects were cleaned in white
Fig. 13.2. Three examples of a “Friesacher Pfennig.” From the left: Bernhard (1202 –1256), Adalbert II (1168 –1177)(?), Ulrich II (1181 –1202).
608
The provenance of medieval silver coins
spirit and ethanol in an ultrasonic bath prior to analysis. Investigations were carried out by using a SPECTRACE 5000 (TRACOR XRAY Inc.) instrument, equipped with a Si(Li)-detector of 155 eV resolution at the Mn Ka line and 1000 cps. For the excitation a Rh-tube with an acceleration voltage of 50 kV and a primary filter of palladium with 0.05 mm thickness was used. The acquisition time for one spectrum was usually 100 s real time. During the analysis the coins were rotated with a speed of 6 rpm in order to suppress effects caused by inhomogeneities of the bulk composition [21] or the surface structure of the objects [36]. As shown in Table 13.1, it was possible to increase the precision of the examinations by the use of a spinner for sample rotation. This table lists the standard deviations of the Ag and Cu X-ray intensities derived from 10 repeated measurements under different sample orientations with and without the use of the spinner. Experiments were carried out on a current Austrian silver/copper coin with a silver content of 80%. A reduction of the two-fold relative standard deviation 2s (95%) to 1.1% for silver and 4.0% for copper was observed. The area of irradiation was of circular shape with a diameter of approximately 1 cm. A trend in modern applications of X-ray and g-ray techniques is to use primary beam diameters in the range of several micrometers [37,38]. In view of the inhomogeneities of the chemical composition and/or the presence of corrosion products on the surface of the coins, the use of these microscopic equivalents is not very relevant in this case. In practice, the investigation of a great number of samples requires a fast “single-measurement”-technique, whereas microscopic X-ray and g-ray techniques require a number of measurements on one object in order to achieve a statistically representative determination of the chemical composition. Analyses of the medieval silver coins were performed on both sides of the coins and mean values were calculated for each element. In most cases the relative differences of the silver intensities between the obverse and reverse sides were lower than ^ 1%, with a maximum of ^ 3.1%. For gold, lead, bismuth, mercury, and especially copper and iron, the relative deviations TABLE 13.1 Comparison of the two-fold relative standard deviation 2s (95%) of the silver and copper X-ray intensity measured in a silver–copper coin (80% Ag) with and without sample rotation during repeated measurements ðn ¼ 10Þ With spinner 2s
Ag 1.7%
Without spinner Cu 3.2%
Ag 2.8%
Cu 7.2%
609
Robert Linke et al.
were situated between ^ 5 and ^ 15%. Coins with an exceptionally high deviation between the obverse and reverse sides (differences .10% were observed in approximately 5% of the coins) were cleaned for a second time with ethanol in an ultrasonic bath prior to analysis. Repeated analyses revealed a lower standard deviation between the obverse and reverse sides in 60% of the cases due to the removal of corrosion products on the coins’ surfaces. Silver and copper were quantified by using appropriate silver/copper ¨ GUSSA [39] with silver concentrations of 500/1000, standards produced by O 800/1000, 900/1000, 925/1000, 950/1000, and 999/1000. As XRF is an analytical technique for main components and investigations could not be carried out on polished surfaces for conservation reasons, which is an essential prerequisite for a precise quantitative analysis, the X-ray intensities of the trace elements lead, gold, bismuth, iron and mercury were directly introduced into the statistical evaluation program WinStatw after normalization to the total spectrum intensity. For determination of the elements silver, copper, and iron the Ka-line intensity was employed, while for gold, mercury, bismuth, and lead the Lb-line were used in order to minimize peak overlap. Figure 13.3 shows an X-ray fluorescence
Fig. 13.3. X-ray fluorescence spectrum of a “Friesacher Pfennig” showing traces of iron, gold, and lead.
610
The provenance of medieval silver coins
spectrum of a Friesacher Pfennig, indicating the presence of traces of iron, gold, and lead. A main disadvantage occurring during XRF analysis is the low penetration depth, which is mainly a function of the energy of the primary X-ray radiation, the angle of incidence, and the matrix itself [13]. Referring to formula 13.1, D represents the thickness of a layer that emits 63.2% of the fluorescence intensity of an infinitely thick bulk material. This formula applies to the characteristic primary radiation and neglects secondary and tertiary excitation that sligthly alters the value of D. D¼
1 mðEÞ mðFÞ r þ sin C1 sin C2
ð13:1Þ
where D is the information depth (63.2%) (cm); r the density (g cm23); m(E) the mass absorption coefficient for the incident photons (cm2 g21); m(F) the mass absorption coefficient for the observed fluorescent photons (cm2 g21); C1 the angle between the primary beam and the specimen surface (in our case 458); and C2 the take-off angle between the fluorescent beam and the specimen surface (in our case 458). For silver/copper coins with a silver content of 80% an information depth of approximately up to 100 mm can be calculated for Ag Karadiation, whereas for Cu Ka due to the higher absorption of the specific radiation, an information depth of ca. 15 mm in the same matrix is obtained [13,40]. As can be seen in Fig. 13.4, the information depth of Ag Ka radiation increases in a series of silver/copper alloys with increasing silver content, while the Cu Ka information depth decreases. Instead of D, the five-fold information depth 5D is plotted in Fig. 13.4. A sample of thickness 5D will yield a fluorescent intensity that is 99.3% of the intensity that can be derived from an infinitely thick specimen when irradiated at the same parameters. As archaeological objects are usually affected by corrosion, the chemical composition of the surface layer does not correspond to that of the bulk. The less noble component in the alloy will be oxidized and dissolved due to corrosion processes. This results in a relative enrichment of silver at the surface of the coins while copper is depleted. When examining corroded objects on their surfaces without sampling by means of XRF, the question may arise whether reliable results concerning the chemical composition of excavated and apparently corroded objects can be obtained by XRF or not. If the thickness of the corrosion layer exceeds the information depth of the X-ray fluorescence radiation, a qualitative and quantitative evaluation of the data is even impeded [41,42]. Therefore, it should be considered, to which
611
Robert Linke et al.
Fig. 13.4. The five-fold information depth 5D (99.33% intensity) of Ag K a radiation increases in a silver/copper alloy in dependence of the silver content, while Cu K a radiation decreases in dependence of the matrix composition.
extent the chemical composition in the present state and of the surface layer corresponds to that of the time of production and that of the bulk of the coin. By considering the Ag – K/Ag –L intensity ratios, it is possible to establish whether the surface of a coin is corroded or not [43]. This is based on the fact that the Ag Ka radiation (energy: 22.1 keV) originates from a depth of approx. up to 100 mm (depending on the matrix composition) while the Ag La radiation (energy: 3 keV) stems from a depth of at most 2 mm. Ag – K/Ag – L ratios obtained from unknown coins can be compared to those derived from polished silver/copper standards of similar composition. Coins with a lower ratio indicate a relative enrichment of silver at the surface, whereas higher ratios reveal the deposition of corrosion products with an elemental composition different from the bulk. For the Friesacher Pfennig coins, Ag – K/Ag – L ratios in the range 50.0 ^ 2.9 were obtained, whereas the corresponding mean value for standards was 51.6 ^ 0.6. This result indicates that the majority of the Friesacher Pfennig coins investigated show no significant deviations from the data obtained from the standard specimen and that the compositions obtained by EDXRF are representative for the chemical composition of the coins.
612
The provenance of medieval silver coins
13.2.2.2 Energy dispersive X-ray microanalysis in the scanning electron microscope In order to validate the results obtained by EDXRF, one coin was polished at its edge (by using SiC paper of 4.000 mesh grain size) and analysed by means of SEM. An SEM, type JSM 6400 of Jeol Inc., with an energy dispersive X-ray microanalyzer, type Link eXL, was available for the measurements at the Institute of Chemical Technologies and Analytics of the Vienna University of Technology. An accelerating voltage of 20 kV was used for the analyses. Compared to EDXRF, SEM/EDX investigations only achieve surface-specific information due to the low information depth (1 –5 mm, depending on the matrix composition). Therefore, SEM/EDX seems to be an inadequate technique for studying the chemical bulk composition of archaeological objects non-destructively. Another main disadvantage compared to EDXRF are the high detection limits due to the low energy of the electrons. In the case of the silver/copper coins, trace elements in the concentration range below 0.5% could not be detected, whereas EDXRF enabled a detection limit down to approximately 50 – 100 ppm depending on the matrix composition [21]. While wavelength dispersive detectors, in general, enable better detection limits, analyses can only be performed on spots, which results in time consuming analysis procedures, if inhomogeneous sample material has to be analysed. Comparative examinations of archaeological objects by means of SEM/EDX and XRF were already published among others, by Klockenka¨mper et al. [4], Love et al. [44], and Stern [45]. As can be seen in Fig. 13.5, the backscattered electron image reveals a homogeneous elemental distribution of silver and copper and no corrosion layers at all. The quantitative results, obtained by ZAF calculations, revealed a good correspondence with the data obtained by EDXRF. On the other hand, it must be kept in mind that the SEM/EDX investigations of one single object cannot be representative for all coins, as the objects were excavated at various places and buried for different periods of time under varying environmental conditions. 13.2.3
Results
219 medieval coins as well as two modern fakes were analysed by EDXRF in order to characterize any possible decay of the silver standard and to enable geographical assignments. In what follows below, a survey of the results will be presented. The individual data can be found elsewhere [46], including a more detailed discussion of the numismatic aspects of the investigations [47].
613
Robert Linke et al.
Fig. 13.5. The backscattered electron image of the polished edge of a “Friesacher Pfennig” reveals a homogeneous structure of the coin due to the homogeneous distribution of the elements silver and copper.
Multivariate statistical methods such as cluster and factor analysis were employed in order to facilitate the interpretation of the data. For a statistical evaluation of the data, a representative number of coins is required because untypical values can cause significant problems during the interpretation of the analytical results. However, this requirement could not be complied with in all cases considering the limited number of coins available. Especially coins of the early period of the “Friesacher Pfennig” (12th century AD ) are extremely rare and even though the investigations were carried out nondestructively, only single objects were available for scientific analysis. Prior to the analysis of the unknown coins, objects definitely dated and assigned to a specific sovereign or a certain mint by the numismatists were characterized in their chemical composition and distinctive marks were defined. This first phase was followed by the analysis of coins of unknown provenance. 13.2.3.1 Coins of known provenance An essential element in the identification of the mint where the coins were produced are the lead and copper-rich ore deposits the metals were derived from, especially in Carinthia in former times. It can be concluded from the archives that in the 12th century each mint in the area of Carinthia had its
614
The provenance of medieval silver coins
own affiliated silver mine. The technology used for the silver production was cupellation. The high content of lead in the coins combined with isolated and low mercury concentrations observed, exclude a silver production by amalgamation. Corresponding results were obtained by geological, archaeometrical [48] and historical [49] examinations. In addition to silver, copper, lead, and mercury, also traces of iron, gold, and bismuth were detected by EDXRF. It can be assumed that lead, gold, mercury, and bismuth are components of the ores, whereas iron appears to be mainly a contamination from the soil. This results from the fact that iron is a less noble metal than copper and can be easily oxidized during the cupellation process. Additionally, no correlation could be found between on the one hand the coins’ mint or the period of mintage and on the other hand the iron concentration. Coins showing extremely high amounts of iron were cleaned by the conservator of the KHM; after this treatment, additional analyses revealed a reduction of the iron X-ray intensities derived from the coins’ surfaces. Therefore, iron was not included in the statistical evaluation of the data. The large variation in silver contents in specific coins mainly resulted from corrosion effects. Varying silver concentrations only partially refer to an unreproducible mintage technology, which can be detected by the Ag – K/ Ag – L ratio method described above. By studying the archives [49], it can be concluded that in the medieval times metallurgists were able to produce alloys with a high precision and accuracy. The evaluation of the EDXRF data revealed that the contents of silver and copper slightly varied during the decades, but that no geographical assignments can be inferred. Only coins struck in the mints of St Veit and Aquileia feature a higher copper content than coins minted in Friesach or Laufen. Clearer geographical differences were obtained for gold, lead, bismuth, and mercury. Coins from the mint of Aquileia show high concentrations of mercury compared to all objects investigated. The reason, therefore, might be the use of mercury enriched lead ores for the silver production or the use of recycled silver, which was originally produced by amalgamation. An enrichment of mercury in the surface of the coins by the storage in a mercury rich soil can be excluded due to the fact that all coins of Aquileia showed the high mercury content independent from their excavation sites. By comparing the trace element pattern of gold, lead, bismuth, and mercury a definite local assignment is possible. Figure 13.6 shows that the coins of the six different mints are clustered in groups corresponding to their origin. The factor analysis included the elements gold, lead, bismuth, and mercury. Coins that show an atypical trace element pattern are indicated by their analysis number. In these cases, a
615
Robert Linke et al.
Fig. 13.6. Scatterplot of the coins of known provenance (£, Friesach; X, Salzburg; þ, Cologne; B, Laufen; W, Aquileia; A, St Veit).
mintage in a different mint than stated can be excluded from the numismatist’s point of view. An explanation could be that old silver was recycled in order to produce new coins. During these times, it was also common to restrike old coins with a new die, although no traces of an old mintage could be observed on the coins investigated. These coins demonstrate the limitations of scientific investigations and show that reliable results concerning the authenticity, chronology or provenance should only be obtained in an interdisciplinary co-operation. Comparing the mints of Friesach and Cologne Although there are great differences in the quality of mintage, style and weight of the coins, according to the numismatic literature [33] the Friesach mintage may have been influenced by the technology used in Cologne/ Germany, an important minting centre during the medieval period. Thus, it was supposed that in the 12th century the mintage technology was transferred by mint workers from Cologne to Friesach. In order to validate this assumption, nine coins of each mint were analysed by EDXRF. Although this number of coins is not representative for
616
The provenance of medieval silver coins
a statistical evaluation, it was not possible to extend the number of objects due to their rarity. Investigations were carried out in order to discover similarities or differences in the chemical compositions of the alloys and, therefore, also in the technologies used in Friesach and Cologne. A number of problems were encountered during the interpretation of the data since it was unclear whether the observed differences were caused by differences in technology or by the use of different kinds of silver ores. When comparing the silver content of coins from both mints it is remarkable that the silver concentration of the Friesacher coins within one series varies by about ^ 5% relative, while coins from Cologne tend to be more precise in their alloy composition. The silver content only varies in the range of ^ 2.5%, which corresponds approximately to the error of measurement. The mean value of the silver concentration from Friesach ranges from approximately 75.0 ^ 2.5 to 85.0 ^ 2.5% silver, while coins from the mint of Cologne have a more homogeneous silver content ranging from 84.7 ^ 2.5 to 86.7 ^ 2.5%. Characteristic for the Cologne coins is the presence of the trace elements lead and gold in higher concentrations, while silver from Friesach mainly contains low amounts of lead and bismuth (Fig. 13.7). Although the silver content of the coins from Cologne is very homogeneous, it is evident that the
Fig. 13.7. X-ray intensity scatterplot of coins from the mints of Friesach (open symbols) and Cologne (filled symbols). The coins minted in Friesach contain lower amounts of mercury, gold, and lead compared to those from the mintage of Cologne.
617
Robert Linke et al.
trace element contents vary within all coins. These results lead to the conclusion that in Cologne no silver mining industry was available so that either recycled or imported silver had to be used as starting material. Although the die was cut of better quality and the fineness is of higher precision compared to the “Friesacher Pfennig,” the silver of Friesach has a higher degree of purity. The high amount of gold in Cologne silver coins results from the fact that in the mint of Friesach workers were able to purify silver in a better quality. In summary it can be said that no similarities could be found between the mintage of Friesach and Cologne and therefore, no positive conclusion could be drawn concerning a transfer of mining technology in the 12th century. 13.2.3.2 Comparison and assignment of coins with unknown provenance As already mentioned previously, the following scientific investigations should support the numismatic assignments. According to the extensive and complex formulation of the question only a selection of the analyses will be presented in the following. As the coins of unknown provenance are similar in their silver content as well as in the trace element pattern, it was not possible to treat all coins in one single statistical evaluation procedure. Therefore, the coins were evaluated according to their period of mintage. Archbishop Konrad I versus Archduke Engelbert The first coins of the “Friesacher Pfennig” were struck by Archbishop Konrad I (1106 – 1147) between 1125/30 and 1135. Coins of this period are represented by three variants, which are supposed of having been minted simultaneously or at least within a short period by numismatic criteria. The average silver contents of all variants was found to be, respectively, 84.4 ^ 3.4, 84.1 ^ 3.4, and 83.2 ^ 3.4% and featured a homogeneous trace elemental pattern. During this time also Archduke Engelbert (1124 – 1135) minted coins of the same style and silver alloy. Due to the rarity of these coins, only two objects (analysis number M7 and M190) were available for investigations. It can be derived from the archives that the Episcopal mintage of Konrad I was produced in Friesach, whereas no information is available concerning the provenance of the coins of Archduke Engelbert. Although it is known that during the 12th century the ducal mintage was transferred to St Veit, a common mintage of Engelbert and Konrad I in the town of Friesach cannot be excluded from the numismatic point of view. Therefore, investigations were carried out in order to compare the unknown coins with Episcopal coinage of Konrad I and coinage of the
618
The provenance of medieval silver coins
Archdukes Ulrich I (1135 – 1144) and Heinrich V (1144 – 1156), which are supposed of being produced in the mint of St Veit. As can be seen in Fig. 13.8 the coins minted in Friesach (indicated with the letter “F” in the scatterplot) and those minted in St Veit (indicated with “V”) are clustered together according to their origin. Differences mainly occur in the contents of lead, gold, and bismuth. Coins minted in Friesach are characterized by higher amounts of lead and bismuth, while coins from St Veit show a more homogeneous trace element pattern (Fig. 13.8). Also coin M7 shows a trace element pattern similar to the silver alloy of St Veit, whereas coin M190 is characterized by a striking low content of silver (71.0 ^ 2.8%) and bismuth. Since this coin is directly connected with coin M7 due to numismatic criteria, a mint different from coin M7 can be excluded. Therefore, this coin serves as an excellent example of all the difficulties that might occur during data evaluation. The mintage of the Archdukes Heinrich V, Hermann, and Ulrich II As a second example the mintage of the ducal coinage of Heinrich V (1144 – 1161), Hermann (1161 – 1181), and Ulrich II (1181 – 1202) was object of investigation. Due to the absence of indications in the archives or on the coins itself, their place of production was unknown. From the historic point
Fig. 13.8. Comparison of the ducal coins M7 and M190 of unknown provenance with coins from Friesach (“F”) and St Veit (“V”) in the plane of the first two principal factor derived from to the X-ray intensities of lead, gold, bismuth, and mercury.
619
Robert Linke et al.
of view it is supposed that they were produced in the mints of Friesach or St Veit, but there might also exist another, still unknown mint. Comparing these objects with coins from the mint of St Veit, the problem occurs that similar coins were minted by Duke Bernhard (1202 –1256) approximately 50 years later. As can be seen in the scatterplot in Fig. 13.9, main differences occur in the content of bismuth and lead. While coins minted by Heinrich V contain no traces of bismuth but a varying content of lead, those minted by Heinrich’s followers Hermann, Ulrich II, and Bernhard are characterized, in general, by higher concentrations of bismuth. Untypical values from the mintage of Heinrich V are indicated with the analysis number (35a, 36b, and 123) in Fig. 13.9. Concerning the coins’ silver concentrations no significant chronological variations could be observed within the coinage of Duke Heinrich V (75.8 ^ 3.0%) and Duke Bernhard (80.0 ^ 3.0%). Assuming that the technology of mintage did not change significantly within the last 50 years, conclusions can be drawn that the coins minted by Hermann and Ulrich II were produced in St Veit, while those minted by Heinrich V stem
Fig. 13.9. The ducal coinage of Heinrich V, Hermann, and Ulrich II of unknown provenance compared to Duke Bernhard, whose mint was located in St Veit (þ , Duke Heinrich V (1144 – 1161); W, Duke Ulrich II (1181 – 1202); X, Duke Herman (1161 –1181); B, Duke Bernhard (1202 –1256), mint: St Veit).
620
The provenance of medieval silver coins
from another mint, probably Friesach. Therefore, from the metallurgical point of view the change of the mint occurred in 1161 on occasion of the accession to the throne of Duke Hermann. Contemporary and modern fakes Additionally, contemporary and modern fakes were analysed in order to characterize their chemical compositions. Although there exists no difficulty in distinguishing these coins from the numismatic point of view, analyses were carried out in order to characterize the chemical composition of the coins. Due to its economic success, the “Friesacher Pfennig” was imitated in medieval as well as in modern times. Seven forged coins dating from the period after the Episcopal mintage were objects of investigation. Six of these coins were produced from a copper/tin alloy with a low silver content. As it is known from the literature [6,22,50], the debasement of the coins’ noble metal content was used since ancient periods as a means of compensating for economic difficulties. Therefore, coins with a low content of silver or gold were minted and blanched in diluted acids. During this procedure, the base components of the alloy (e.g. copper or tin) were dissolved while the more noble component was relatively enriched at the surface. This procedure could be applied to coins with a silver content down to approximately 1 – 5%. During circulation the silver enriched surface layer was worn off and the copper rich core of the coin appeared. These coins contain a surface layer of silver of just a few micrometers thick, which could be detected by means of the Ag – K/Ag –L intensity ratio method. Thus, a non-destructive quantification of the chemical composition of the coins’ core by means of EDXRF was not meaningful. As forged coins of this period are very rare and therefore of high value, it was not possible to carry out SEM/EDX analyses on cross-sections or polished edges at least. Another Episcopal forgery from 1160/1180 contained solely copper and tin, but no traces of silver could be detected by EDXRF analysis. In this case, the high amount of tin gave the coin a silvery and bright appearance. Furthermore, two modern copies were analysed. Compared to the genuine coins, differences in the trace elemental composition as well as in the silver standard were observed. While the two modern coins are characterized by a silver content of approximately 90.0 ^ 3.0%, genuine coins from the 12th century contain a silver concentration of approx. 80 ^ 4.0%. Additionally, no trace elements could be observed by EDXRF analysis, which indicates a 20th century production. As mentioned
621
Robert Linke et al.
previously, the process of silver production in the 12th century was cupellation followed by an oxidizing refining step leaving a typical trace element pattern. In the second half of the 19th century this technique of silver production was replaced by electrochemical methods, yielding to silver of higher purity. Silver of this period contains no trace elements detectable with EDXRF (see Ref. [51] for details).
13.3 13.3.1
THE TIROLER KREUZER Introduction
An important Austrian coinage of the 15th century was the so-called “Tiroler Kreuzer” minted by Archduke Sigismund of Tyrol (1439 – 1490, († 1496) (Fig. 13.10). The coins investigated are part of a hoard, which was found in Lower Austria in 1992 and now belong to the collection of the Currency ¨ sterreichische Nationalbank [52]. Compared to the coins of Museum of the O the “Friesacher Pfennig,” these coins were in a bad state of preservation after excavation, as can be seen in Fig. 13.11. The complete hoard consists of 2797 silver and 14 gold coins, which were stored in a jar buried in the soil for approximately 500 years. 157 of the silver coins, which were chosen according to numismatic criteria, were analysed. According to the iconography and stylistic variations, the mintage of Archduke Sigismund can be separated into 10 main groups, labelled I – X. While coins of the groups I – VII were minted in the town of Merano (South Tyrol), coins from groups IX and X were produced in Hall (North Tyrol). Within group VIII, which consists of three variants, the transfer of the mint took place (in December 1477). It was supposed by numismatists that subgroup VIII/1 was still produced in Merano and subgroup VIII/3 already in Hall. However, there existed no evidence for this postulation and it was therefore uncertain, whether the coins of subgroup VIII/2 were still minted in Merano or already in Hall. Nevertheless, the silver content
Fig. 13.10. “Tiroler Kreuzer” by Archduke Sigismund of Tyrol (1439 –1490, † 1496).
622
The provenance of medieval silver coins
Fig. 13.11. “Tiroler Kreuzer” in the state of excavation.
of these coins is of major importance for historians in order to draw conclusions of the economic situation in medieval Tyrol and Austria in the second half of the 15th century. In this respect, Wieser [53] and Rizzolli [54] reported on an initial silver concentration of 75%, which was reduced to 50% not later than 1473. However, the precise date of this depreciation was unknown. Accordingly, scientific investigations were carried out with the aim of determining the provenance of the coins belonging to subgroups VIII/1 to VIII/3 as well as their silver contents. As demonstrated in section 13.2, the chemical composition of the coins can be characteristic for the various mints of the medieval period, due to differences in either the ores or the silver production process. In the 15th century the lead ore deposits near Schwaz in North Tyrol produced silver for both the Merano and Hall mints. Additionally, silver was also imported by merchants and old coins were recycled, which makes an assignment of the coins regarding their material composition more difficult. Due to the degree of corrosion it must be also assumed that the silver coins contain high amounts of copper, which increases the number and amount of trace elements and yields to a change of the trace element pattern of the silver alloys. Although the majority of the coins was in a bad state of preservation, the analyses were required to be non-destructive in view of the unique character and high value of the coins. Therefore, a combination of EDXRF, SEM/EDX and PIXE was applied in order to determine the chemical compositions of the coins.
623
Robert Linke et al.
13.3.2
Experimental
13.3.2.1 Energy dispersive X-ray fluorescence analysis EDXRF analyses were performed under the same conditions as described in section 13.2.2.1. After the excavation in 1992, all coins were cleaned by the conservator of the KHM, using a solution of complexone-IIIw (mainly EDTA) and ethanol. It was observed that the silver content of the coins was situated in the range from 55.5 ^ 6.1 to 88.3 ^ 1.2%, but no conclusions could be drawn concerning the chronology of their emission. The Ag – K/Ag – L ratios indicated that this high variation of the chemical composition resulted from corrosion processes, which had mainly enhanced the silver concentrations at the coins’ surfaces. Indeed, from non-corroded, polished silver/copper standards of similar composition (50 – 80% Ag), Ag – K/Ag – L ratios that were approx. 20% lower ratios were observed. Apart from copper, all coins contained detectable amounts of lead and bismuth. In a few coins, also traces of iron, gold, and mercury were observed. Comparison of the contents of bismuth and gold revealed that a number of the coins from the mint of Merano contain higher amounts of bismuth while coins from Hall show higher concentrations of gold. However, a clear distinction between both mints could not be drawn (Fig. 13.12). While coins of subgroup VIII/1 are similar in their chemical composition to coins of the groups I to VII (Merano), those of subgroup VIII/3 are similar to those of the groups IX and X (Hall). In the scatterplot coins of subgroup VIII/2 are located between the coins of known provenance, which may suggest a continuous succession. However, a clear assignment could not be performed on the basis of these EDXRF data alone. 13.3.2.2 Energy dispersive X-ray microanalysis in the scanning electron microscope Due to the fact that EDXRF did not allow to make a clear distinction between the mints and a quantitative analysis of the silver content was not meaningful due to surface corrosion, additional investigations were carried out using SEM on cross-sections of specific coins by means of the instrument described in section 13.2.2.2. 52 coins were selected according to their state of preservation, cut with a scalpel, embedded in resin, ground and polished with SiC-paper up to 4.000 mesh perpendicular to their surfaces so that the SEM/EDX investigations could be performed on the polished cross-sections. For quantitative analysis, areas of approximately 200 £ 160 mm2 (magnification 500 £ ) to
624
The provenance of medieval silver coins
Fig. 13.12. Concentrations (in mg/g) of the elements bismuth and gold in a series of “Tiroler Kreuzer” coins of groups (X, coins from the mint of Merano; W, coins, for which Merano is supposed as mint (group VIII/1); £ , coins of unknown provenance (group VIII/2); A, coins, for which Hall is supposed as mint (group VIII/3); B, coins from the mint of Hall).
100 £ 80 mm2 (magnification 1000 £ ) were irradiated. In general, five analyses were performed at each cross-section and mean values were calculated. In the backscattered electron image as well as in X-ray mapping mode the relative enrichment of silver and the depletion of copper due to corrosion processes could be observed. Quantitative analyses were restricted to areas unaffected by corrosion. SEM investigations on cross-sections of several coins confirmed that silver was relatively enriched at the surface, while copper, the base component of the alloy, was depleted. In Fig. 13.13, the dark domains correspond to the copper-rich phase, whereas the bright domains on the coin’s surface corresponds to the silver-rich phase. The cavities (black areas) in-between the bright silver-rich phase near the surface are caused by leaching and diffusion of copper during the corrosion process. The thickness of the corroded layers varied between 50 and 200 mm, rendering non-destructive SEM/EDX investigations of the coins’ surfaces without purpose in this case. Due to the low information depth of EDXRF the apparent silver content is higher than that obtained by SEM/EDX from the cross-sections in the non-corroded core of the coins. The relative
625
Robert Linke et al.
Fig. 13.13. Backscattered electron image of a cross-section of a “Tiroler Kreuzer.”
differences range from 3 to 105%. Although all coins were buried under the same conditions and for the same time, the objects revealed significantly different degrees of corrosion. Both, completely corroded coins as well as objects with a corrosion layer of just a few micrometers could be observed. The observed silver concentrations were situated in the range from 35.3 ^ 1.5 to 76.0 ^ 1.6%. In contrast to EDXRF, however, no trace elements could be detected. In summary, it was concluded that in this specific case, SEM/EDX appeared to be a non-sufficient technique for non-destructive analysis of the chemical composition of corroded objects due to the unfavourable detection limits of the EDX method and the low penetration depth of electrons combined with an advanced corrosion process in the surface layers of the coins. 13.3.2.3 Proton induced X-ray emission spectroscopy As the investigations using EDXRF and SEM/EDX did not yield a clear answer concerning the origin of the coins of group VIII/1 to VIII/3, additional analyses by PIXE were undertaken at the Laboratoire d’Analyses par Re´actions Nucle´aires (LARN) in Namur (Belgium).
626
The provenance of medieval silver coins
PIXE represents a quasi non-destructive method for chemical analysis and can be regarded at present as one of the most popular ion beam techniques for material analysis of archaeological artefacts [55,56,57]. Compared to other techniques, the information depth of PIXE (5 – 10 mm) exceeds that of SEM/EDX slightly, but is definitely lower compared to that of XRF. Due to the fact that the analyses in the SEM revealed that the thickness of the corroded layer on the coins was approx. 100 mm, also non-destructive PIXE was considered to be unsuitable for the determination of the coins’ silver content. On the other hand, when appropriate absorbers are applied, this technique features better detection limits than EDXRF and SEM/EDX, Measurements were performed at ambient atmosphere using an external proton beam of 2.9 MeV energy with a beam diameter of less than 1 mm [55]. A 10 mm Co absorber (with the Co – K-absorption edge situated at 7.7 keV) was placed between the coin and the Si(Li)-detector in order to improve the detection conditions for the elements with Z , 29 (such as Ni, with Ni Ka line at 7.5 keV) and to simultaneously reduce the intensities of the Ka- and Kb-lines of Cu (Cu Ka: 8.0 keV, Cu Kb: 8.9 keV), the most abundant element in the coins apart from silver (Fig. 13.14). As a result of the well-defined geometry of the experimental set-up [58],
Fig. 13.14. Mass absorption characteristics of the Co K-edge related to the elements Zn, Cu, Ni, Co, and Fe. The apparent concentration of Cu has been decreased by the use of a Co absorber.
627
Robert Linke et al.
no signal of secondary Co fluorescence was visible in the resulting PIXE spectra. Each coin was mounted on a frame in order to keep the geometric conditions constant during the analysis. The coins were analysed on both the obverse and reverse and mean values were calculated. The experimental set-up has already been described in the literature [58]. For the quantitative evaluation of the data, an iterative program was used that takes several parameters such as X-ray cross sections, X-ray absorption, sum and escape peaks, secondary fluorescence etc. into account [59,60]. The program enables a quantitative determination of the alloy by comparison with signals from their respective neighbouring elements using the appropriate corrections. All measurements are normalized. The elements silver, copper, lead, bismuth, and iron were detected in all coins; several coins also contained traces of nickel, antimony, gold, and mercury. The quantitative evaluation covered the elements silver (mean ^ s: 91.4 ^ 2.8%), copper (6.8 ^ 2.9%), lead (0.7 ^ 0.3%), bismuth (0.6 ^ 0.3%), antimony (0.2 ^ 0.2%), and nickel (0.03 ^ 0.03%). Compared to XRF and SEM/EDX the results obtained for silver showed higher values due to the surface corrosion layer and the low penetration depth of the protons compared to photons. The evaluation of the trace and minor components revealed that mainly the elements nickel and bismuth allow a distinction between coins minted in Merano and Hall. As can be seen in Fig. 13.15, apart from differences in the intensities of lead and bismuth, the coin from Hall contains no nickel, while in the coin produced in Merano this element could be clearly observed.
Fig. 13.15. PIXE spectra of coins from the mints of Merano and Hall.
628
The provenance of medieval silver coins
Fig. 13.16. Ni/Ag and Bi/Ag X-ray intensity ratios for a series of “Tiroler Kreuzer” coins (X, coins from the mint of Merano; W, coins for which Merano is supposed as mint (group VIII/1); £ , coins of unknown provenance (group VIII/2); A, coins, for which Hall is supposed as mint (group VIII/3); B, coins from the mint of Hall).
Comparison of the X-ray intensity ratios of Ni/Ag and Bi/Ag revealed that most coins minted in Merano contain higher amounts of nickel while the majority of those minted in Hall are characterized by higher levels of bismuth (Fig. 13.16). According to these findings, coins of the subgroups VIII/1 and VIII/2 can be related to the mint of Merano. According to their nickel and bismuth contents, coins of subgroup VIII/3 are more similar to the mint of Hall. In conclusion it can be said that the existence of subgroups VIII/1 and VIII/3 corresponds to coins originating from the mints of Merano and Hall. This was expected from a numismatic point of view and could be confirmed by means of PIXE. The similarity of the chemical composition of the coins of the subgroup VIII/2 with coins of the mint of Merano can be interpreted as evidence for a change of the mint between the subgroups VIII/2 and VIII/3. 13.4
CONCLUSION
Among the non-destructive techniques applied in the field of art and archaeology, EDXRF, SEM/EDX, and PIXE are valuable tools for
629
Robert Linke et al.
characterizing the provenance of metallic coins by means of their chemical compositions. As could be demonstrated, the state of preservation of the coins is of major importance for quantitative analysis. As EDXRF, SEM/EDX, and PIXE have different information depths according to their primary energy, the thickness of the corrosion layer has to be taken into consideration. Therefore, single investigations of cross-sections in the SEM are mandatory for characterizing the microstructure of the alloys and for determining the presence of corrosion layers on the surface. If the analyses need to be carried out without sampling, the Ag – K/Ag – L X-ray intensity ratio obtained by EDXRF is a very useful measure for estimating the degree of depletion. Additionally, EDXRF as well as PIXE have the significant advantage that the analysis is not as time consuming as with other analytical techniques. If the objects are clean, no sample preparation is necessary. However, as quantitative XRF requires polished specimen, the resulting X-ray intensities only convey information on the surface of the coins. Due to the fact that in the medieval period most mints had their own silver mine or at least their own technology of silver refinement, the content of the minor and trace elements in the alloy can be used for assigning the coins to their origin. Even for a small number of objects available for analysis, it is possible to determine their provenance.
Acknowledgements ¨ sterreichische Nationalbank” (Project No. The Jubila¨umsfonds of the “O 7026: “Materialanalytische Untersuchungen am Friesacher Pfennig”) is gratefully acknowledged for financial support of the investigations on the Friesacher Pfennig. PIXE analysis were carried out within the frame of COST G1 Action: ion beam analysis applied in art and archaeology. REFERENCES 1 2 3 4 5 6
630
E. Aloupi, A.G. Karydas and T. Paradellis, X-Ray Spectrom., 29 (2000) 18. P.L. Leung and H. Luo, X-Ray Spectrom., 29 (2000) 34. P. Hoffmann, St. Bichlmeier, M. Heck, C. Theune and J. Callmer, X-Ray Spectrom., 29 (2000) 92. R. Klockenka¨mper, M. Becker and H. Otto, Spectrochim. Acta, 45B(9) (1990) 1043. F. Schweizer, Arbeitsbla¨tter fu¨r Restauratoren, 19(1) (1985) 126. R. Klockenka¨mper, H. Bubert and K. Hasler, Archaeometry, 41(2) (1999) 311.
The provenance of medieval silver coins 7 8 9
10
11 12
13 14
15 16
17 18 19
20 21
22 23
24 25 26
C. Pearson, Conservation of Marine Archaeological Objects. Bodmin, 1987. T. Stambolov, The Corrosion and Conservation of Metallic Antiquities and Works of Art. Amsterdam, 1985. H.J.M. Wouters, Celtic goldwork found in north-east Belgium. In: W. Mourey and L. Robbiola (Eds.), Metal 98, Proceedings of the International Conference on Metals Conservation. London, 1998, 64. M. Ferretti and P. Moioli, The use of portable XRF systems for preliminary compositional surveys on large bronze objects, In: W. Mourey and L. Robbiola (Eds.), Metal 98, Proceedings of the International Conference on Metals Conservation London, 1998, 39. A. Longoni, C. Fiorini, P. Leutenegger, S. Sciuti, G. Fronterotta, L. Stru¨der and P. Lechner, Nucl. Instr. Meth. Phys. Res., A409 (1998) 407. C. Neelmeijer, I. Brissaud, T. Calligaro, G. Demortier, A. Hautoja¨rvi, M. Ma¨der, L. Martinot, M. Schreiner, T. Tuurnala and G. Weber, X-Ray Spectrom., 29 (2000) 101. M. Mantler and M. Schreiner, X-Ray Spectrom., 29 (2000) 3. H. Mommsen, Th. Beier, H. Dittmann, D. Heimermann, A. Hein, A. Rosenberg, M. Boghardt, E.-M. Hanebutt-Benz and H. Halbey, Archaeometry, 38(2) (1996) 347. P. Wobrauschek, G. Halmetschlager, S. Zamini, C. Jokubonis, G. Trnka and M. Karwowski, X-Ray Spectrom., 29 (2000) 25. K. Janssens, G. Vittiglio, I. Deraedt, A. Aerts, B. Vekemans, L. Vincze, F. Wei, I. Deryck, O. Schalm, F. Adams, A. Rindby, A. Kno¨chel, A. Simionovici and A. Snigirev, X-Ray Spectrom., 29 (2000) 73. A.E. Pillay, C. Punyadeera, L. Jacobson and J. Eriksen, X-Ray Spectrom., 29 (2000) 53. P. Mirti, X-Ray Spectrom., 29 (2000) 63. A.A. Gordus and D.M. Metcalf, Neutron activation analysis of the gold coinages of the Crusader states. In: D.M. Metcalf and W.A. Oddy (Eds.), Metallurgy in Numismatics, Special Publication Nr. 13, Vol. 1. The Royal Numismatic Society, London, 1980, 119 pp. M.M. Al-Kofahi and K.F. Al-Tarawneh, X-Ray Spectrom., 29 (2000) 39. W.B. Stern, Die naturwissenschaftliche Untersuchung der Mu¨nzen. In: A. Burkhard, W.B. Stern and G. Helmig (Eds.), Keltische Mu¨nzen aus Basel. Antiqua, 25. Verlag Schweizerische Gesellschaft fu¨r Ur- und Fru¨hgeschichte, Basle, 1994, 79. L. Daraban, C. Cosma and T. Fiat, J. Radioanal. Nucl. Chem., 201(5) (1995) 447. W.B. Stern, On non-destructive analysis of gold objects, In: G. Morteani and J.P. Northover (Eds.), Prehistoric Gold in Europe. Kluwer Academic Publishers, The Netherlands, 1995, 317. M.-A. Meyer and G. Demortier, Nucl. Instr. Meth. Phys. Res., B49 (1990) 300. B. Constantinescu, V.J. Kennedy and G. Demortier, Int. J. PIXE, 9(3/4) (1999) 487. H. Mommsen and T. Schmittinger, Archaeometry, 23(1) (1981) 71.
631
Robert Linke et al. 27 28 29
30 31
32 33
34 35
36 37 38 39 40 41 42 43 44 45
46
47
632
J.A. Buckley, Archaeometry, 27(1) (1985) 102. G.P. Ferreira and F.B. Gil, Archaeometry, 23(2) (1981) 189. M.R. Cowell and M. Ponting, British Museum Analyses. In: W. Hollstein (Ed.), Metallanalytische Untersuchungen an Mu¨nzen der Ro¨mischen Republik, Berliner Numismatische Forschungen, Neue Folge Band 6, Gebr. Mann Verlag, Berlin, 2000, 49. I. Calliari, M. Magrini, A. Zambon, P. Guerriero and R. Martini, X-Ray Spectrom., 28 (1999) 86. J.P. Northover, Analysis in the electron microprobe and scanning electron microscope. In: W.A. Oddy and M.R. Cowell (Eds.), Metallurgy in Numismatics, Special Publication Nr. 30, Vol. 4, London. 1999, 94. R. Klockenka¨mper, B. Raith, S. Divoux, B. Gonsior, S. Bru¨ggerhoff and E. Jackwerth, Fres. Z. Anal. Chem., 326 (1987) 105. M. Alram, Die Friesacher Mu¨nze im Alpen-Adria-Raum, Grazer Grundwissenschaftliche Forschungen, Vol. 2. Schriftenreihe der Akademie Friesach, Graz, 1996, 97. H. Winter, Mitteilungen der o¨sterreichischen numismatischen Gesellschaft, 37(1) (1997) 10. M. Alram, R. Ha¨rtel and M. Schreiner (Eds.), Die Fru¨hzeit des Friesacher ¨ sterreichische Akademie der Pfennigs (etwa 1125/30 – etwa 1166), O Wissenschaften, Philosophisch – historische Klasse, Denkschriften 300, ( ¼ Vero¨ffentlichungen der Numismatischen Kommission 36), Wien 2002. W. Stankiewicz, A. Fudal and M. Wojtowicz, X-Ray Spectrom., 12(3) (1983) 92. K.H.A. Janssens, F.C.V. Adams and A. Rindby (Eds.), Microscopic X-ray Fluorescence Analysis, Chichester, 2000. H. Bronk, S. Ro¨hrs, A. Bjeoumikhov, N. Langhoff, J. Schmalz, R. Wedell, H.-E. Gorny, A. Herold and U. Waldschla¨ger, Fres. J. Anal. Chem., 371(3) (2001) 307. ¨ sterreichische Gold- und Silberscheideanstalt. O R. Linke, M. Schreiner, G. Demortier and M. Alram, X-Ray Spectrom, 32 (2003) 373. J. Condamin and M. Picon, Archaeometry, 7 (1964) 98. E.T. Hall, Archaeometry, 4 (1961) 62. R. Linke and M. Schreiner, Mikrochim. Acta, 133 (2000) 165. L.J.C. Love, L. Soto and B.T. Reagor, Appl. Spectrosc., 34(2) (1980) 131. W.B. Stern, Zur zersto¨rungsfreien Zustandsdiagnose: Metallkundliche Untersuchungen an antiken Silberlegierungen. In: F. Schweizer and V. Villiger (Eds.), Methoden zur Erhaltung von Kulturgu¨tern. Bern and Stuttgart, 1989, 181. R. Linke, Materialanalytische Untersuchungen von Bronze-, Silber- und Goldmu¨nzen mittels EDXRF, REM/EDS, XRD, FTIR-Mikroskopie und ICP-MS, PhD Thesis, Vienna University of Technology, Vienna, 2000, 51. R. Linke and M. Schreiner, Materialanalytische Untersuchungen am Friesacher Pfennig mittels energiedispersiver Ro¨ntgenfluoreszenzanalyse. In: M. Alram, R. Ha¨rtel and M. Schreiner (Eds.), Die Fru¨hzeit des Friesacher Pfennigs
The provenance of medieval silver coins
48 49 50
51 52 53 54
55 56 57 58 59 60
¨ sterreichische Akademie der Wissenschaften, (etwa 1125/30 – etwa 1166), O Philosophisch –historische Klasse, Denkschriften 300, ( ¼ Vero¨ffentlichungen der Numismatischen Kommission 36), Wien 2002. G. Sperl, Akten der Friesacher Sommerakademie. Grazer Grundwissenschaftliche Forschungen Vol. 2, Schriftenreihe der Akademie Friesach, Graz 1, 1994, 75. G. Agricola, Zwo¨lf Bu¨cher vom Berg- und Hu¨ttenwesen. Reprint from 1556, Munich 1994. C.E. King and J.P. Northover, The analyses, In: H. von Kaenel (Ed.), Der Mu¨nzhort aus dem Gutshof in Neftenbach, Zu¨rich Denkmalpflege Archa¨ologische Monographien 16, 1993, 110. ¨ sterreichischen Gesellschaft fu¨r R. Linke and M. Schreiner, Zeitschrift der O Ordenskunde, 34 (1999) 15. M. Alram, H. Winter and M. Metlich, Numismatische Zeitschrift, 104/105 (1997) 109. F. Wieser, Numismatische Zeitschrift, 81 (1965) 14. H. Rizzolli, Die Tiroler Mu¨nzpra¨gung in Meran. Beitra¨ge zur Wirtschaftsgeschichte Su¨dtirols. Festschrift zum 125ja¨hrigen Bestehen der Su¨dtiroler Sparkasse, Bozen, 1979, 359. T. Calligaro, J.C. Dran and J. Salomon, this volume, Chapter 5. F. Beauchesne, J.N. Barrandon, L. Alves, F.B. Gil and M.F. Guerra, Archaeometry, 30(2) (1988) 187. G. Demortier, Nucl. Instr. Meth. Phys. Res., B64 (1992) 481. G. Demortier and Y. Morciaux, Nucl. Instr. Meth. Phys. Res., B85 (1994) 112. B. Van Oystaeyen and G. Demortier, Nucl. Instr. Meth., 215 (1983) 299. G. Demortier, S. Mathot and B. Van Oystaeyen, Nucl. Instr. Meth. Phys. Res., B49 (1990) 46.
633
This page is intentionally left blank
Chapter 14
Pigment identification in illuminated manuscripts Peter Vandenabeele and Luc Moens
14.1
INTRODUCTION
Throughout history, books, being symbols of wisdom, have always attracted people. Numerous mediaeval paintings show marvellous illuminated manuscripts, expressing the importance of their owner. Only the wealthiest people in the society could afford books. Indeed, apart from the time invested to manufacture a manuscript, even the raw materials needed were highly expensive [1,2]. The number of bifolia that could be obtained from a single sheet of skin, though depending on the size of the book, was limited; parchment and pigments were very expensive. Some manuscripts contain colourful miniatures, capitals or border illuminations, raising the cost of the manuscript. While admiring these codices, one almost automatically wonders about the production process of these artefacts. How were these works manufactured? How many people worked on this book? Who were they and how long did it take to make such a manuscript? These questions are not always simple to answer, and the solutions might depend on the work under consideration. Manuscript production starts with the purchase of different raw materials: parchment, pigments, binding media, etc. Depending on the book and its size as ordered, an estimation of the number of folios was made. Decisions were taken on the number, size and position of the different miniatures and the layout of the work was discussed. Subsequently the sheets of parchment were cut into the desired format and folded into a bifolium (or diploma). A number of these were gathered together into a quire. Often a quire consisted of four bifolia, thus containing eight folios (or leaves), i.e., 16 pages. The next stage in manuscript production was the ruling of the pages. Lines were drawn in order to determine the layout of the page. Then the scribes copied the text by hand. Spare room was left for later insertion of Comprehensive Analytical Chemistry XLII Janssens and Van Grieken (Eds.) q 2004 Elsevier B.V. All rights reserved
635
P. Vandenabeele and L. Moens
miniatures, capitals, paragraph marks, etc. During the long and labourintensive process of copying the whole treatise, occasionally mistakes were made. In this case the scribe marked the alterations in the margin. Similarly, the scribe made markings for the rubricator, who inserted capitals and headings in the next stage of manuscript production. The word “rubricator” originates from the Latin “rubrum” which means “red,” a colour that is often used for capitals, titles, etc. At the end of many manuscripts a “colophon” can be found: this often indicates the year of manufacture or purchase and the patron. The whole manufacture of mediaeval manuscripts was a very labourintensive and time-consuming process. Occasionally, in some manuscripts a “lamentatio” can be found in which the scribe complained about the work it took to copy the whole manuscript. Examples of these are [2]: Explicit hic totum, frater Jacobe, da michi potum (This is the end of my duty, brother Jacob, bring me a mug of wine) or Finito libro, frangamus ossa magistri (As the copying of the manuscript is finished, we will break the bones of the master, who composed it)
Generally, many mediaeval manuscripts have survived the ravages of time quite well. One reason for this phenomenon is that these works of art have always been considered as very valuable and thus have been handled with care. Often manuscripts were status symbols instead of utilitarian objects. Codices were kept in libraries, and thus were protected at least from some external influences. Manuscripts were stored closed, shielding the illuminations from harmful influences, such as light. On the other hand, one can note that there is also a kind of historical selection: works of art of bad technical quality are often eliminated in the course of time, so that mainly works of good technical quality remain. 14.2
COMBINED METHOD APPROACH
There are several possible reasons to perform art investigations [3]. Aside from the fundamental interest in ancient materials and techniques, art analysis can help in solving specific questions of (art-) historians, curators and restorers. These professionals often encounter problems concerning degradation, authenticity or provenance of an artefact. Degradation studies are made by looking at changes in the structure and the spectroscopic characteristics of the different components. The question of authenticity is closely related to that of dating the artefact. Painted objects of art can be
636
Pigment identification in illuminated manuscripts
approximately dated by examination of their pigments. Certain pigments have well-established dates of invention and finding them indicates a later creation or modification of the artefact. A well-known example is Prussian blue, a pigment that was invented in 1704 by Diesbach in Berlin [4,5]. A positive identification of this material enables dating post quem. On the other hand, certain pigments fell into disuse, as they were substituted by better pigments or by pigments that were less harmful or less expensive. Finding such pigments makes it possible to date an object of art ante quem. Natural Indian yellow is such a pigment that disappeared from the palette around 1900. This pigment was obtained from urine of cows that were fed with mango leaves. Although a good pigment was obtained, this practice harmed the animals, as the salts formed needle-shaped crystals in their kidneys. Therefore, under pressure of animal welfare activists, this practice was forbidden by law and thus, the identification of natural Indian yellow in an object of art indicates that the work almost certainly dates from before that period [6]. Anachronistic use of materials is a strong indication for an artefact to be a fake. In making these conclusions, care has to be taken that no restorations are analysed. Another way to recognize forgeries by spectroscopic means is by comparing the pigments in a specific object of art with those in a large set of other artefacts by the same artist. By analysing an extended collection of his paintings, the artist’s palette can be reconstructed as a function of time. The discovery of apparent contradictions can be indicative of a fake. Another important reason for art analysis is to obtain answers to specific questions of conservators, curators or (art-)historians. Such questions include the determination of the origin of degradation. An example was the identification of copper in a green pigment in a mediaeval manuscript of the Ghent University Library [7]. This pigment caused the deterioration of the manuscript as the parchment that served as substrate was corroded. For each examination of artefacts, the information obtained and the risk of possible damage should be balanced. The art investigator should aim to obtain as much information as possible on the artwork, while damaging it as little as possible (and if possible not damaging it at all). Therefore, it is a good idea to apply several non- or micro-destructive methods, in order to obtain complementary information [3,8,9]. In our case, two sensitive microanalytical techniques, namely Total-reflection X-Ray Fluorescence analysis (TXRF) and Micro-Raman Spectroscopy (MRS) were used, in combination with a gentle micro-sampling method [10– 12].
637
P. Vandenabeele and L. Moens
TXRF [13,14] reveals the elemental composition of the sample. As only the elements with Z . 14 are detected, the analysis is limited to the inorganic pigments. Most of the pigments can be identified by means of their “key elements.” For example, the presence of mercury in a red sample indicates the presence of vermilion. Despite this, in some cases TXRF is not able to identify the pigments unambiguously. The presence of copper in a green sample might indicate the presence of copper resinate, malachite, verdigris, posnjakite, brochantite, another green copper pigment or even the blue pigment azurite, mixed with a yellow pigment. On the other hand, TXRF determines the relative amounts of the detected elements, and thus the relative amounts of the pigments in the sample can be identified. Moreover, TXRF provides an idea of the (elemental) impurities in the sample. MRS provides complementary information on the sample. Whereas TXRF is not able to distinguish between several pigments with the same key element, a molecular technique such as MRS provides information on the chemical environment of these elements and (if any) on their crystalline structure [15 –17]. Moreover, MRS is not restricted to inorganic materials, but it is possible to record spectra of organic dyes and binding media as well. By using microscope optics, it is possible to record spectra from samples with a diameter of ca. 1 mm and thus all the pigment grains can be identified individually. Despite these advantages, by focussing on different pigment grains, the technique cannot be used for quantitative analysis of this type of sample. If no useful Raman spectrum can be recorded because of fluorescence effect giving rise to an excessive spectral background, a laser with another wavelength can be chosen, if available. In some cases, Raman spectroscopy may be applied in a direct way [18 –20]. This approach has the advantage that no sampling is required. Several groups have positioned loose leaves or whole bound manuscripts under the Raman microscope [21,22]. By changing the position of the manuscript or the folio, it was possible to analyse different painted areas. During these investigations, care has to be taken not to damage the fragile artefact, either manipulating it or by applying too much laser power. With standard Raman microscopes, unless small manuscripts are examined, it is often impossible to reach the central parts of the manuscript, because of spatial limitations. In this case, an alternative is the use of a tilted microscope [19] or a mobile probe equipped with fibre-optics [20]. As two complementary analytical methods are involved [23], it may be helpful to make use of a convenient sampling method. In art analysis it is critical to damage the object as little as possible, especially when delicate manuscripts are involved. For TXRF investigations, it is necessary that the
638
Pigment identification in illuminated manuscripts
sample can be spread out homogeneously, as a fine (mono-) layer of grains, over a flat sample carrier. For these reasons, a gentle micro-sampling method was developed [10 – 12,24]. This procedure consists of gently rubbing a cotton bud over the painted surface, thus transferring a small amount of paint (ca. , 1 mg) from the artefact to the Q-tip. This sampling method does not leave any visible trace on the artefact. In order to avoid loss of material and to eliminate contamination of the sample, the cotton bud was mounted in a plastic sample container. In the laboratory, the Q-tip is tapped on to a suitable sample carrier. An important disadvantage of this method is that the information is limited to the surface layer. By using this sampling procedure it is not possible to examine a paint layer that is covered with varnish. For the examination of illuminated manuscripts, this is barely a drawback, but for the analysis of easel paintings sampling has to be done when the varnish is removed, e.g., when the painting is being restored. This Q-tip sampling method has been applied for the present study. 14.2.1
Analysis of manuscripts
The spectroscopic examination of mediaeval manuscripts focuses on the different materials in these objects of art: paint (consisting of pigment and binding medium), parchment, dyes and ink. 14.2.1.1 Inorganic pigments One of the most striking aspects when admiring mediaeval illuminated manuscripts is the brightness of the painted colours. Paint consists of a mixture of a binding medium and a colouring agent. If the colouring agent is insoluble in the medium, it is called a “pigment”; otherwise the term “dye” is used. See Table 4.6 for an overview of inorganic pigments. Red pigments During the middle ages, red was an important colour, and was the colour most often used for capital letters, paragraph marks, running titles, etc. In that period several red and orange pigments were in use. XRF key elements of the most important red pigments are tabulated in Table 14.1; the corresponding Raman spectra are shown in Fig. 14.1. Vermilion (HgS) is a pigment that was known already to the Romans. They called it “minium,” a name that caused confusion in later periods, as this term was also used for red lead (Pb3O4) [25]. The high intensity Raman band at 254 cm21, in the spectrum of vermilion, can be assigned to the n(Hg – S) stretching vibration [25]. Different lead oxide pigments, such as red lead and massicot
639
P. Vandenabeele and L. Moens
TABLE 14.1 Overview of some important red pigments, their chemical composition, TXRF key elements and Raman band positions Pigment
Other names
Chemical composition
TXRF key elements
Raman wavenumbers (cm21)
Vermilion
Cinnabar, minium
HgS
Hg
Red lead
Minium
Pb3O4
Pb
Hematite
Main component of red ochre
Fe2O3
Fe
343(m), 283(w), 254(vs), 144(vw), 112(vw) 549(s), 477(vw), 457(vw), 391(m), 314(w), 231(w), 224(w), 152(m), 122(vs) 659(vw), 613(w), 499(w), 412(m), 394(vs), 246(w), 227(vs)
Indications of relative peak intensities are provided (vs, very strong; s, strong; m, medium intensity; w, weak; vw, very weak; sh, shoulder).
(see section “Yellow Pigments”) can be distinguished by their Raman spectra, as both the crystal symmetry and the oxidation state of Pb are different. The intense Raman band at 547 cm21 wavenumber shift can be assigned to the n(Pb – O) stretch vibration, while the band at 224 cm21 is related to the d(O – Pb – O) deformation. Hematite is the main component of different forms of red ochre, i.e., earth pigments that have been in use since prehistory. Using Raman spectroscopy their presence was demonstrated in prehistoric rock art in the southwest of France [26]. Raman and FT-IR spectra of different types of ochres were recorded by Bikiaris et al. [27].
Fig. 14.1. Raman spectra of red inorganic pigments: (a) vermilion, (b) red lead and (c) hematite (baseline-corrected). Raman spectra were recorded with a Renishaw System-1000 spectrometer with a laser of 780 nm.
640
Pigment identification in illuminated manuscripts
Blue pigments In the middle ages, two important blue mineral pigments were available, viz. azurite (2CuCO3·Cu(OH)2) and lapis lazuli (natural ultramarine, Na8···10 Al6Si6O24S2···4) [4,5,28]. Azurite is a basic copper carbonate, that is found as a mineral together with the green pigment malachite (CuCO3·Cu(OH)2). In Western Europe, it was the most important blue pigment until the middle of the 17th century. The natural form of ultramarine is lapis lazuli, a semiprecious stone that prior to this period (?) was only found in Badakshan (in present Afghanistan). Purification of the milled material was more complex and time-consuming than for most minerals and this, combined with its rareness and the long distance it had to travel, made it the most expensive mediaeval pigment. Only around 1830, three scientists (J.B. Guimet, C. Gmelin and F.A. Ko¨ttig) independently succeeded in synthesizing ultramarine [5]. Lapis lazuli and synthetic ultramarine are chemically identical and they may only be distinguished from each other by the presence of impurities. Smalt (CoO·nSiO2) is the earliest cobalt pigment. It is suspected to originate from glass production. According to Schramm and Hering [4] it was in use from the 15th century onwards in Western Europe. Raman spectra of some blue pigments are shown in Fig. 14.2 and are tabulated, along with TXRF key elements in Table 14.2. As azurite is anisotropic, the spectrum is strongly orientationdependent. The intense band at 1096 cm21 derives from the symmetrical
Fig. 14.2. Baseline-corrected Raman spectra of blue inorganic pigments: (a) azurite, (b) lapis lazuli and (c) smalt. Spectra recorded under similar conditions as for Fig. 14.1.
641
642 TABLE 14.2
Pigment
Other names
Chemical composition
TXRF key elements
Raman wavenumbers (cm21)
Azurite
Mountain blue
2CuCO3·Cu(OH)2
Cu
Lapis lazuli Smalt
Ultramarine
Na8···10Al6Si6O24S2···4 CoO·nSiO2
– Co
1578(w), 1460(w), 1430(m), 1419(m), 1344(w), 1259(w), 1096(s), 939(w), 837(w), 816(w), 764(w), 739(w), 701(w), 540(w), 401(vs), 385(m), 333(w), 283(w), 249(vs), 216(vw), 195(w), 172(m), 156(vw), 145(vw), 131(vw), 114(m) 1086(w), 796(w), 545(vs), 251(vw) 886(vw), 692(m), 620(w), 522(w), 483(w), 460(s), 445(vw), 236(vw), 197(m), 124(vw)
Indications of relative peak intensities are provided (vs, very strong; s, strong; m, medium intensity; w, weak; vw, very weak; sh, shoulder).
P. Vandenabeele and L. Moens
Overview of some important blue pigments, their chemical composition, TXRF key elements and Raman band positions
Pigment identification in illuminated manuscripts
stretch vibration of the carbonate ion. At lower wave numbers different Cu – O lattice vibrations are found. The Raman spectrum of ultramarine is characterized by the very strong band at 545 cm21, which is assigned to the n([S – S]2) stretch vibration [29,30]. The other bands in this spectrum are weak to very weak. The Raman spectrum of a smalt sample shows three medium to strong Raman bands (at 692, 460 and 197 cm21). Lapis lazuli has no TXRF key elements, as the characteristic X-rays of low-Z elements are absorbed by the air and by the Be-window of the detector. Additionally, the sulphur-K lines overlap with Pb –Ma radiation. Moreover, as sulphur is present in several traditional binding media, this element is not a conclusive indicator for the presence of lapis lazuli.
Yellow pigments Lead – tin yellow, type I (Pb2SnO4) was in use between ca. 1300 and ca. 1750 AD . The pigment lead –tin yellow, type II (PbSn2SiO7) was a pigment that probably originated from Venetian or Bohemian glass production. It became available at the beginning of the 14th century [31]. Massicot (PbO) is a yellow lead(II)oxide. Yellow ochre (Fe2O3·nH2O) is an important earth pigment that was in use since prehistory. The main component is limonite. Raman spectra are shown in Fig. 14.3 and are summarized together with the TXRF key elements in Table 14.3. In the spectrum of massicot the very strong band at 143 cm21 can be assigned to the n(Pb – O) stretching vibration [30]. In the same region (130 cm21) in the spectrum of lead – tin yellow (type I) a very
Fig. 14.3. Raman spectra of yellow inorganic pigments: (a) lead –tin yellow (type I), (b) massicot and (c) limonite (baseline-corrected). Spectra recorded under similar conditions as for Fig. 14.1.
643
P. Vandenabeele and L. Moens TABLE 14.3 Overview of some important yellow pigments, their chemical composition, TXRF key elements and Raman band positions Pigment
Other names
Lead –tin yellow (type I) Massicot Limonite
Main component of yellow ochre
Chemical composition
TXRF key elements
Raman wavenumbers (cm21)
Pb2SnO4
Pb, Sn
PbO
Pb
Fe2O3·nH2O
Fe
613(vw), 525(vw), 456(w), 379(vw), 336(vw), 292(w), 273(w), 196(m), 130(vs) 384(vw), 289(m), 217(vw), 143(vs), 124(vw) 552(w), 484(w), 420(w), 400(m), 389(s), 302(m), 247(w), 206(vw)
Indications of relative peak intensities are provided (vs, very strong; s, strong; m, medium intensity; w, weak; vw, very weak; sh, shoulder).
intense Raman band is found. The Raman spectrum of red ochre is easily distinguished from the spectrum of yellow ochre, since many peak positions and intensities are changed.
Green pigments Many different green pigments have been used by artists. Copper pigments in particular were commonly used in the middle ages. In the literature often verdigris (Cu(CH3COO)2·nCu(OH)2) or malachite (CuCO3·Cu(OH)2) are cited [4], but in reality many more green copper pigments were in use [32,33], several of these being of mineral origin. Common examples are the basic copper sulphates, such as posnjakite (Cu 4SO4(OH)6 ·H2O) and brochantite (Cu4SO4(OH)6). In the middle ages, verdigris was synthesized by the reaction of vinegar with copper plates and, depending on the recipe, different shades and types of verdigris were obtained. Malachite, on the other hand, was a natural mineral. Copper resinate results from the dissolution of copper salts in turpentine and resinous material. Chrysocolla is a green copper silicate of mineral origin that, after grinding, was used as a pigment, while the main component of green earth is an iron silicate. When performing dispersive Raman spectroscopy with a near infrared laser (l ¼ 780 nm), green pigments in general give rise to rather weak Raman spectra with low signal-to-noise ratios, as a result of the absorption of the laser light (Fig. 14.4, Table 14.4).
644
Pigment identification in illuminated manuscripts
Fig. 14.4. Baseline-corrected Raman spectra of green inorganic pigments: (a) verdigris, (b) malachite, (c) copper phosphate, (d) basic copper sulphate (Brochantite), (e) copper resinate, (f ) veronese green earth and (g) chrysocolla. Spectra recorded under similar conditions as for Fig. 14.1.
Black pigments For centuries, the most important black pigment was carbon, produced by the incomplete combustion of lamp oil (lamp black), bones (bone black), plants (plant black), etc. It is often hard to distinguish between these different types [5]. In bone black, next to carbon, calcium phosphate is detected. Another black pigment, of synthetic origin, is iron oxide black (FeO·Fe2O3). It was produced since ca. 1920; according to Schramm and Hering [4] natural magnetite has never been applied as an artist’s pigment. Raman spectra of black pigments feature rather broad bands. Those of iron oxide black are clearly different in position and relative intensity from those of hematite. Carbon black produces two broad Raman bands (Fig. 14.5 and Table 14.5) in which the position as well as the band width is a measure of the crystallinity of the graphite [34]. White pigments In the middle ages, chalk (CaCO3) was in use for the preparation of parchment, but sometimes it was also used as whitener. In Northern mediaeval and renaissance easel paintings, the material frequently served
645
646 TABLE 14.4 Overview of some important green pigments, their chemical composition, TXRF key elements and Raman band positions TXRF key elements
Raman wavenumbers (cm21)
Verdigris
Cu(CH3COO)2·nCu(OH)2
Cu
Malachite
CuCO3·Cu(OH)2
Cu
Copper phosphate
Cu3(PO4)2
Cu
Basic copper sulphate
CuSO4·nCu(OH)2
Cu
Copper resinate
Cu-salt, dissolved in resin
Cu
Veronese green earth
FeSiO3·nH2O
Fe
Chrysocolla
CuSiO3·nH2O
Cu
948(s), 700(vw), 398(vw), 324(w), 233(w), 179(m) 1488(w), 1205(vw), 1093(w), 1055(w), 718(vw), 536(vw), 508(vw), 431(m), 351(w), 268(w), 221(m), 180(s), 169(m), 153(s), 119(w) 961(m), 741(vw), 609(m), 578(vw), 482(m), 448(s), 410(vw), 365(w) 971(s), 802(vw), 621(w), 479(w), 444(w) 1652(m), 1430(w), 1202(w), 948(s), 714(w), 577(w), 413(w), 316(m), 232(w), 180(m) 1509(vw), 1327(vw), 1077(w), 689(w), 560(w), 464(s), 206(m) 972(s), 619(w), 502(w), 482(m), 447(w), 421(w), 389(m), 319(w), 239(vw), 169(w)
Other names
Indications of relative peak intensities are provided (vs, very strong; s, strong; m, medium intensity; w, weak; vw, very weak; sh, shoulder).
P. Vandenabeele and L. Moens
Chemical composition
Pigment
Pigment identification in illuminated manuscripts
Fig. 14.5. Raman spectra of black pigments: (a) carbon black and (b) iron black. Spectra recorded under similar conditions as for Fig. 14.1.
as preparation layer, mixed with animal glue, while in Italy gypsum (CaSO4·2H2O) was more commonly used for the preparation layer, using a technique called “gesso.” For the white pigment, anhydrite (CaSO4) was frequently applied, a natural mineral that is found together with gypsum, but that often was synthesized by calcination of gypsum. One of the most important white pigments is white lead (2PbCO3·Pb(OH)2); beside its colouring properties this material is appreciated as a siccative. Permanent white (BaSO4) is a synthetic white pigment that has been available since the beginning of the 19th century [5]. The natural form, which is called barite, can occasionally be observed in mediaeval manuscripts, often together with gypsum [35]. Normally it is not used as white pigment as it lacks masking power. The Raman spectra of different white materials are easily distinguished (Fig. 14.6). These spectra all have a very intense Raman band, which TABLE 14.5 Overview of some important black pigments, their chemical composition, TXRF key elements and Raman band positions Pigment
Other names
Chemical composition
TXRF key elements
Raman wavenumbers (cm21)
Carbon black Iron black
Lamp black
C FeO·Fe2O3
– Fe
1594(m), 1297(s) 667(m), 529(w), 492(w), 306(m)
Indications of relative peak intensities are provided (vs, very strong; s, strong; m, medium intensity; w, weak; vw, very weak; sh, shoulder).
647
P. Vandenabeele and L. Moens
Fig. 14.6. Baseline-corrected Raman spectra of white pigments: (a) chalk, (b) gypsum, (c) white lead and (d) permanent white. Spectra recorded under similar conditions as for Fig. 14.1.
is caused by the symmetric stretching vibration of the carbonate or sulphate ions. The exact band position is mainly determined by the nature of the cation. These cations are of use as TXRF key elements (Table 14.6) for the identification of the whitener, although in practice it is hard to discriminate between calcium carbonate and calcium sulphate by means of TXRF. 14.2.1.2 Binding media In the middle ages a pigment that was synthesized or purchased was mixed with a binding medium in the workshop in order to make paint. For the production of manuscripts, binders were usually of polysaccharide or proteinaceous nature. Binding media are not really identifiable using TXRF, because of their organic nature and the presence of inorganic pigments that are much more sensitive, although some progress has been made on the identification of violin varnishes with XRF [36]. MRS on the other hand, being a molecular spectroscopic method, is able to provide information on the composition of these organic products [37]. Proteinaceous binding media [38], frequently used for tempera painting, are mainly of animal origin. Casein, egg glair and egg yolk (either mixed
648
Pigment identification in illuminated manuscripts TABLE 14.6 Overview of some important white pigments, their chemical composition, TXRF key elements and Raman band positions Chemical composition
TXRF key elements
Raman wavenumbers (cm21)
Chalk
CaCO3
Ca
Gypsum
CaSO4·2H2O
Ca
White lead
2PbCO3·Pb(OH)2
Pb
BaSO4
Ba
1086(vs), 712(m), 282(m), 156(m) 1372(vw), 1134(w), 1006(vs), 669(w), 618(w), 492(m), 413(m), 317(w), 210(w), 181(w),122(w) 1056(vs), 418(w), 353(vw), 260(m), 203(m), 154(m) 1165(vw), 1137(vw), 986(vs), 645(w), 615(w), 460(m), 451(m), 188(vw), 153(vw)
Pigment
Permanent white
Other names
Barite
Indications of relative peak intensities are provided (vs, very strong; s, strong; m, medium intensity; w, weak; vw, very weak; sh, shoulder).
together or unmixed) and several animal glues have frequently been applied. Casein, a phosphoprotein, can easily be obtained from milk, as it precipitates after acidification and heating of skimmed milk. Albumen has been applied in different ways: glair has been used as such and the protein can be mixed with a small amount of water to adjust the viscosity. Albumen has even been applied together with egg yolk, the latter component acting as emulsifier. At the time animal glues (i.e., gelatin, bone glue, skin glue, fish glue, isinglass, etc.) were not often used in manuscripts as binding media in paint, but applied as fixative (e.g., to glue gold leaves). Raman spectra of proteins are characterized by the presence of intense amide I (ca. 1650 cm21) and amide III bands (ca. 1250 cm21). Often, a sharp Raman band is observed at ca. 1002 cm21, which is assigned to the aromatic ring breathing vibration of the phenylalanine amino acid. Another group of binding media that was often used in mediaeval manuscript production is of vegetal origin. Starch (poly-D -glucose) has been used as glue or substrate for organic dyes. Its Raman spectrum is dominated by an intense band at 477 cm21,which can be attributed to a deformation of the backbone of the polysaccharide chain. Unlike resins, gums are polysaccharide plant exudates that are soluble or swell in water (gel formation). Among these, gum Arabic is the most important. Being harvested from different acacia species, this polysaccharide consists of
649
P. Vandenabeele and L. Moens
different monosaccharides, among which L -arabinose and D -galactose. The Raman spectra of gums are composed of broad, overlapping bands, that can be assigned to d(C –H) bending vibrations and n(C –O) and n(C – C) stretching vibrations, all below 1500 cm21. Raman spectra of some binding media are shown in Fig. 14.7. 14.2.1.3 Iron gall ink Iron gall ink is one of the most frequently used mediaeval inks, besides charcoal [39]. Iron gall ink is prepared by adding vitriol (iron(II)sulphate, FeSO4) to the aqueous extract of, among others, gallnuts. In its pure form, the ink is the precipitated complex of the iron ion with four molecules of gallic acid, but in practice the extract also contains several esters such as di-gallic acid and tannin besides the pure gallic acid. As the ink is applied on the paper or parchment, one observes the darkening of the ink, a phenomenon that has been ascribed to the oxidation of Fe2þ to Fe3þ under the influence of oxygen in the air [40]. By performing TXRF analysis on mediaeval iron gall ink samples, large amounts of Fe are observed, often accompanied with smaller amounts of Cu and Zn. These may be ascribed as impurities in either the iron ore that was used for the production of vitriol or in the vessels that were used for the production or storage. Another possible explanation is the use of several admixtures in the vitriol that was used.
Fig. 14.7. Baseline-corrected Raman spectra of a proteinaceous binding medium: albumen (a) and two polysaccharide media: starch (b) and gum Arabic (c). Spectra recorded under similar conditions as for Fig. 14.1.
650
Pigment identification in illuminated manuscripts
In Fig. 14.8 Raman spectra of pure gallic acid, FeSO4, mediaeval iron gall ink and a laboratory analogue are presented. The laboratory sample has been synthesized by using pure FeSO4 and gallic acid powder, dissolved in de-mineralized water. It is readily observed that the spectrum of mediaeval ink is much more complex than that of the modern product; other compounds in the gallnut extract may form complexes as well and the crystallinity of the precipitate may be different. The Raman spectrum of gallic acid powder consists of several well-defined Raman bands. The doublet at ca. 1600 cm21 may be attributed to the n(CyO) stretching vibration and the benzene quadrant stretch vibration. The intense band at 961 cm21 may be attributed to the benzene ring breathing vibration, while the numerous and overlapping bands between 1550 and 1000 cm21 are correlated with d(C – H) deformations and n(C – O) and n(C – C) stretching vibrations. The spectrum of FeSO4·x H2O is dominated by the Raman band that can be assigned to the n(SO4) symmetric stretch vibration of the sulphate group (976 cm21).
Fig. 14.8. Baseline-corrected Raman spectra of (a) gallic acid, (b) FeSO4, (c) mediaeval gall ink and (d) a laboratory specimen. Experimental conditions for the preparation of the laboratory sample are given in the text. Spectra recorded under similar conditions as for Fig. 14.1.
651
P. Vandenabeele and L. Moens
14.2.1.4 Parchment Parchment was an important writing medium during the middle ages, although in the early middle ages papyrus was used as well, and from the 13th century onwards paper, already known for centuries in China and the Arab world, was introduced. Mediaeval paper was made from linen rags [1]. In mediaeval times, the percamentarius or parchmenter had the task of transforming animal skins into parchment. The first stage of this delicate process consisted of the selection of good skins from healthy animals. After skinning, the pelt had to be washed under cold running water. Often they were put into a vessel with lime for several days to promote a partial decomposition of the skin. As the skin started to rot, the hairs were loosened from it. In the next stage, the wet skins were taken out of the lime and the remaining hairs were scraped off with a curved knife with two handles. Then the remaining flesh and fat were scraped from the reversed side of the skin. In this phase care had to be taken not to cut the skin by accident. The wet skin was then stretched on a frame, where it was allowed to dry. While drying, the parchment tended to shrink. The dried and stretched material was then scraped again for several times in order to make it whiter and thinner. The pieces of skin that were removed could be boiled to produce animal glue. Often the parchment was treated with chalk to make it look paler and to prepare it for the writing process. Parchment was very expensive and therefore in several mediaeval manuscripts repairs can be found: little holes were sometimes sewn to prevent further damage. Parchment is a proteinaceous material, so its Raman spectrum (Fig. 14.9) contains well-defined amide I and III bands, similar to the spectra of proteinaceous binders. As the parchment is treated with chalk, the intense Raman band at 1086 cm21 (n(CO3) symmetrical stretch) is also observed. 14.2.2
Sources of impurities
When working with micro-samples, special care has to be taken to avoid any contamination, which may lead to false conclusions. One should always be careful to take a sample that is representative of the work or paint area under study. Sampling, sample transportation to the laboratory and preparation by the analyst should be done in such a way that contamination is avoided. But even when all these conditions are fulfilled, the nature of mediaeval manuscripts makes two forms of contamination hard to avoid; these types have been dubbed “internal” and “external” contamination [35]. Internal contamination originates from imprecise sampling: e.g., during the analysis of minute details in initials or border illuminations in mediaeval
652
Pigment identification in illuminated manuscripts
Fig. 14.9. Raman spectrum of a piece of calf parchment. Spectrum recorded under similar conditions as for Fig. 14.1.
manuscripts, it was observed that, though unintended, the surrounding areas were sampled as well. This is because normal cotton buds are relatively large in comparison to the area to sample. In order to reduce this internal contamination, an alternative for the Q-tips was found in biomedical practice. Small cotton buds were used, consisting of a metal needle that is covered with a tuft of cotton. The size of the head is 2 mm, which is considerably smaller than the diameter of the head of a normal cotton bud (ca. 5 mm). This needle is also mounted in a plastic container to avoid loss of material and contamination. The micro-chemical purity of the sampling tools was tested by investigating several blanks. TXRF analyses showed that these cotton buds did not cause any interferences. External contamination originates from coloured areas of facing folios in a volume. As over the centuries manuscripts were kept closed for most of the time, the coloured areas of facing folios were in intimate contact. By using the Q-tip sampling method, only the surface layer is sampled and thus the composition of the sample is influenced by the facing coloured area as well. Therefore it may be appropriate to use a first cotton bud to gently clean the area. Thereafter the actual sampling is done with a second cotton bud. Independent of these impurities caused either by the sampling procedure or by the nature of manuscripts, there is also the intrinsic impurity pattern of the paint. Some medieval pigments, such as azurite, were of mineral origin
653
P. Vandenabeele and L. Moens
and the mineral impurities contribute to the final impurity pattern. Moreover, during grinding more impurities could be introduced. Other pigments were synthesized from mineral sources; here again, during synthesis traces of impurities may be added. During transport and trade in pigments still more impurities could be introduced accidentally or intentionally (in the case of adulteration). Finally, in the workshop the pigments were ground and blended, in order to obtain the right hue. By mixing the pigments with binder, paint was made. At all these stages involved in pigment and paint production elemental impurities were introduced, resulting in the intrinsic impurity pattern of the paint layer, which is a reflection of the whole transformation that the pigment underwent from the mineral ore to the miniature. During this processing many different influences affect the impurity pattern, resulting in a combination of trace elements that is unique for a specific batch of paint. Thus, by analysing the elemental composition, it is possible to discriminate between different batches of paint, and thus between the workshops or miniaturists [28,41,42].
14.3
14.3.1
ANALYSIS OF THE MANUSCRIPTS FROM THE COLLECTION OF RAPHAEL DE MERCATELLIS Introduction
Raphael de Mercatellis (1437 – 1508) [43] was a natural son of Philip the Good, Duke of Burgundy, born in Bruges, in a Venetian family of merchants, the Mercatelli family. He studied theology at Paris University and became a monk in the Benedictine abbey of Saint Peter in Ghent. Later, he became abbot of the abbey of Saint Peter in Oudenburg, where the onset for his library was given. During his ecclesiastical career he also became abbot of the abbey of Saint Bavon in Ghent, suffragan of Tournai and nominal bishop of Rhosus. Whereas contemporary bibliophiles preferred literary, moralistic or historical texts written in French, Mercatellis’ library contained works written in Latin, covering a wide range of subjects. Together with works on theology and ecclesiastical law, his collection also contained works by classical authors (such as Aristotle and Plutarch) and manuscripts on astronomy, medicine and mathematics. Although book printing was invented during his lifetime (Gutenberg bible, ca. 1456), the books he commissioned were manuscripts. Moreover, there is a manuscript that is known to be copied from a printed version (Georgus Reish, Margarita Philosophica,
654
Pigment identification in illuminated manuscripts
Ghent University Library Ms. 7). The manuscripts were written on parchment and they have remarkably large dimensions. Often Mercatellis’ coat of arms or his monogram is found. Some manuscripts are illuminated with acanthus borders, historiated initials or miniatures. As no contemporary catalogue of his library survives, the extent of his collection at that time is unknown. The remaining lists, dating well after his death, mention 80 manuscripts; today, 60 of these are spread out over different libraries. The manuscripts have been studied on codicological [43] and stylistic grounds [44] and different groups of works can be distinguished, which can be related to different periods in Mercatellis’ life and to different workshops. It is very likely that the Mercatellis manuscripts were manufactured in private workshops, but little is known on the organization and way of working in these workshops. There is a suggestion that the provenance of the manuscripts is located near Ghent or Bruges. The latest art historical research tends to suggest a Bruges provenance, as there appears to be a relationship between the miniatures in the Ghent University Library Manuscripts 2 and 5 (both astrological treatises) and the Vrelant workshop [45]. The aim of this work is to illustrate how the spectroscopic examination of mediaeval manuscripts from the Mercatellis collection can provide information on the way the manuscripts were produced. Information will be gained on the relationships between samples from within a manuscript (intra-manuscript relationship). The spectroscopic examination of intermanuscript relationships has been examined elsewhere [35,46]. In what follows, we focus on the paint composition of border illuminations, initials and paragraph markings.
14.3.2
Pigment identification with TXRF and MRS
The pigments in the manuscripts were examined by means of MRS and TXRF analysis. All the samples in the manuscript contain a significant amount of Ca. This could originate from the treatment of parchment with chalk, in order to prepare it for writing. From Raman investigations, it turned out that this element was present in the form of calcium carbonate, as well as gypsum. Another white component that was detected in these manuscripts is barite (BaSO4). The synthetic form of this mineral was used as a white pigment since the beginning of the 19th century, namely permanent white [5]. As these manuscripts are mediaeval, and taking into
655
P. Vandenabeele and L. Moens
account that no signs of later overpaintings or restorations were found, it is likely that we found the natural form of this pigment, probably present as an impurity in gypsum. From the TXRF investigations it turned out that the blue samples contained an important amount of Cu, key element for azurite. These findings were confirmed by the Raman investigations. No lapis lazuli was found in the Mercatellis manuscripts. All the red samples contain vermilion. Some red samples feature an admixture of red lead. Figure 14.10 shows a Raman spectrum from the green pigment out of De viris clarissimis by Plutarch (Ghent University Library, Ms. 109). The quality of the Raman spectrum of many green pigments is relatively poor (low S/N ratio), as the 785 nm laser light is absorbed by the green pigment, and as the sample produces fluorescence in this region. Despite this, it was possible to identify the pigment. The spectrum is identical to the spectrum of a basic copper sulphate that was synthesized under laboratory conditions [47]. X-ray diffraction of this synthetic pigment showed that the material had the same composition as the natural mineral brochantite.
Fig. 14.10. Raman spectrum of a green pigment from De viris clarissimis by Plutarch (Ghent University Library, Ms. 109). Spectrum recorded under similar conditions as for Fig. 14.1.
656
Pigment identification in illuminated manuscripts
14.3.3 Intra-manuscript comparison of Expositio problematum Aristotelis Spectroscopic methods can help art historians to discriminate between different workshops. In the Mercatellis manuscript Expositio problematum Aristotelis (Ghent University Library, Ms. 72), the acanthus borders may be divided into three distinct groups (Fol. 1r-151v, 152r-233v and 234r-395v) on stylistic grounds. The difference between the first two groups is minimal and therefore these groups were previously considered as a single class. For this research, 123 micro-samples were taken from blue leaves in acanthus borders and blue initials. In Fig. 14.11 the relative amounts (by TXRF) of Ca, Pb, Fe, Zn and As are presented. These elements are all impurities in the blue paint, where the blue Cu-containing pigment azurite is present. From this representation, the Fe:Zn ratio appears to be mainly responsible for the differentiation between the first and second parts of the manuscript (B þ O) on the one hand, and the last part (X) on the other hand. The same groups are found for the blue samples from the initials. The discrimination between the first (B) and second (O) group becomes clear if
Fig. 14.11. TXRF analysis of 123 blue micro-samples, taken from the borders and initials in the Mercatellis manuscript Expositio problematum Aristotelis (Ghent University Library, Ms. 72). Legend: B fol. 1r-144v; O fol. 153r-230v and X fol. 234r-395v. Group marks are purely indicative.
657
P. Vandenabeele and L. Moens
the relative amounts of Fe:Ca:Pb are considered. When considering the samples from the initials, the relative amount of Pb for the first part of the manuscript (B) is clearly lower than for the second part (O). A similar discrimination is found when considering the Fe:Ca ratio in the blue samples from the acanthus borders. These results illustrate the possibilities of spectroscopic examination to assist in art historical research. During a primary stylistic examination, two distinct groups were found, but after conscientious evaluation of the TXRF results, this thesis had to be adjusted, as three groups were distinct. The need for close interdisciplinary consultation during this research is shown as well with this example. 14.3.4
Analysis of Decretum Gratiani
Decretum Gratiani (Ghent, University Library Ms. 3) is a medley of texts on canon law, collected by the Bolognese monk Gratianus. The manuscript is divided into three volumes (labelled 3-I, 3-II-1 and 3-II-2). At the end of volume 3-II-1 a colophon states that Raphael de Mercatellis acquired this book/volume in 1505: Revendus pater Raphael episcopus Rosensis, abbas Sancti Bavonis, comparavit hoc volumen anno XVcV
As this manuscript is divided into several volumes, one might wonder whether the manuscript was produced by several workshops. This was examined by analysing the impurity patterns in a very large number (128) of samples from the different volumes in the manuscript. In Fig. 14.12 the
Fig. 14.12. TXRF analysis of 128 micro-samples, taken from blue initials in three volumes of Decretum Gratiani (Ghent University Library, Mss. 3-I, 3-II-1 and 3-II-2). Legend: O Ms. 3-I; X Ms. 3-II-1 and B Ms. 3-II-2.
658
Pigment identification in illuminated manuscripts
ratios of the elemental composition of Fe:Ca:K and Fe:Zn:Hg is given for the blue initials over the manuscript. From these diagrams, it is seen that the classes overlap; the centroids for each class are almost coincident. Therefore, we should conclude that it is impossible to discriminate between the different volumes, based on their TXRF impurity pattern. Other elements gave similar results. These findings show that there is great similarity between the volumes and they support the thesis that only one workshop has been involved in the production of this manuscript. Such homogeneity is also confirmed by the analysis of ink samples. Fourteen samples throughout the different volumes of the manuscript have been analysed. Samples originated from the main text, from annotations as well as from corrections. All the samples contained iron and zinc, key elements for iron gall ink [48]. The presence of metallogallic ink was also confirmed by Raman spectroscopy. Comparison of the elemental composition of the ink samples may also give some indication on the way the manuscript was produced and of the allocation of tasks in the workshop. Italian researchers have determined the elemental composition of the ink in the notes of Gallileo Gallilei. By comparing the elemental composition of the ink in his undated letters with the composition of the ink in his diary, they aimed to order these writings [49]. By comparing the elemental composition of the ink samples in Decretum Gratiani, it was found that all the samples had a similar composition, which supports the theory that the whole manuscript was produced in a single workshop. Only the composition of the sample from the colophon was different from the others, but it is very likely that this difference is sooner due to a different preparation of the parchment of the last page than a real different composition of the ink: when omitting the difference in Ca content the sample has a composition similar to that of the others [50]. 14.4
CONCLUSION
In this work it was shown how a combined approach of MRS and TXRF analysis can be of interest for the examination of mediaeval manuscripts. Both methods can be applied in an almost non-destructive way by using a gentle micro-sampling method. By using TXRF, the inorganic pigments are identified by using their key elements. Raman spectroscopy provides information on the molecular composition, inorganic as well as organic, of the different grains in the sample. In addition to identification of pigments, it is possible by using TXRF to obtain an overview of the impurity pattern in the sample. This impurity
659
P. Vandenabeele and L. Moens
pattern is a reflection of all the stages in the pigment production, from the mineral source to the paint production in the workshop. Therefore, the analysis of the impurity patterns may be of use for the study of the provenance and the organization of the workshops where the manuscripts were produced. This approach has been applied for the investigation of two manuscripts from the former library of Raphael de Mercatellis. It was shown how the study of the elemental impurities can be of use for discriminating different hands in a manuscript or to indicate the same provenance of different volumes of a single manuscript.
Acknowledgements The authors wish to thank Martine De Reu, Reinhold Klockenka¨mper, Guido Van Hooydonk and Alex von Bohlen for their cooperation in this research. Mark Clarke is acknowledged for proofreading this text. This research has been supported by the fund for scientific research—Flanders (FWOVlaanderen) and by the Ghent University (Bijzonder onderzoeksfondsBOF). PV is especially grateful to the Fund of Scientific Research—Flanders (FWO-Vlaanderen) for his postdoctoral fellowship. REFERENCES 1 2 3 4 5 6
7 8
9 10
660
C. de Hamel, Medieval Craftsmen—Scribes and Illuminators. British Museum Press, London, 1992. M. Smeyers, Vlaamse Miniaturen van de 8ste tot het midden van de 16de eeuw— De middeleeuwse wereld op perkament. Davidsfonds, Leuven, 1998. E. Ciliberto and G. Spoto (Eds.), Modern Analytical Methods in Art and Archaeology, Chemical Analysis Series, Vol. 155. Wiley, New York, 2000. H.-P. Schramm and B. Hering, Historische Malmaterialen und ihre Identifizierung. Akademische Druck und Verlagsanstalt, Graz, 1988. R.J. Gettens and G.L. Stout, Painting Materials—A Short Encyclopaedia. D. Van Nostrand Company, Inc, New York, 1962. N.S. Baer, A. Joel, R.L. Feller and N. Indictor, In: R.L. Feller (Ed.), Indian Yellow, A Handbook of Their History and Characteristics, Vol. 1. National Gallery of Art—Cambridge University Press, Cambridge, 1986. W. Devos, L. Moens, A. von Bohlen and R. Klockenka¨mper, Stud. Conserv., 40 (1995) 153. CNRS, Pigments et Colorants de l’antiquite´ et du Moyen Age—Teinture, ´ tudes Historiques et Physico-chimiques. E ´ ditions du Peinture, Enluminure—E Centre National de la Recherche Scientifique (CNRS), Paris, 1990. M. Schreiner and M. Grasserbauer, Fresenius J. Anal. Chem., 322 (1985) 181. R. Klockenka¨mper, A. von Bohlen, L. Moens and W. Devos, Spectrochim. Acta B, 48B (1993) 239 –246.
Pigment identification in illuminated manuscripts 11 12
13 14 15 16 17 18 19 20 21 22 23
24 25 26 27
28 29 30 31 32
33 34
L. Moens, W. Devos, R. Klockenka¨mper and A. von Bohlen, Trends Anal. Chem., 13 (1994) 198 –205. W. Devos, L. Moens, A. von Bohlen and R. Klockenka¨mper, Archaeometry 94— The Proceedings of the 29th International Symposium on Archaeometry, 1996, pp. 243–251. R. Klockenka¨mper, Spectrosc. Int., 2(2) (1989) 26–33. R. Klockenka¨mper, Total Reflection X-Ray Analysis. Wiley, New York, 1997. D.J. Gardiner and P.R. Graves (Eds.), Practical Raman Spectroscopy. Springer, Berlin, 1989. I. Lewis and H.G.M. Edwards (Eds.), Handbook of Raman Spectroscopy. Marcel Dekker, New York, 2001. P.V. Huong, Vibr. Spectrosc., 11 (1996) 17. A. Derbyshire and R. Withnall, J. Raman Spectrosc., 30 (1999) 185 –188. D.C. Smith, Techn. Sci. Service Hist. Art Civil., 11 (2000) 69. P. Vandenabeele, F. Verpoort and L. Moens, J. Raman Spectrosc., 32 (2001) 263 –269. P. Dhamelincourt, J. Barbillat and M. Delhaye, Spectrosc. Eur., 5(2) (1993) 16 –26. R.J.H. Clark, Chem. Soc. Rev., 24(3) (1995) 187 –196. P. Vandenabeele, A. van Bohlen, L. Moens, R. Klockenka¨mper, F. Joukes and G. Dewispelaere, Spectroscopic examination of two Egyptian masks: a combined method approach, Anal. Lett., 33(15) (2000) 3315 –3332. A. von Bohlen, W. Devos, L. Moens and R. Klockenka¨mper, Restauro, 5 (1994) 328 –330. H.G.M. Edwards, D.W. Farwell, E.N. Newton and F. Rull Perez, Analyst, 124 (1999) 1323 –1326. D.C. Smith, M. Bouchard and M. Lorblanchet, J. Raman Spectrosc., 30 (1999) 347 –354. D. Bikiaris, Sister Daniilia, S. Soutiropoulou, O. Katsimbiri, E. Pavlidou, A.P. Moutsatsou and Y. Chryssoulakis, Proceedings of the Sixth International Conference on Non Destructive Testing and Microanalysis for the Diagnosis and Conservation of the Cultural and Environmental Heritage, Rome, 1999, pp. 483–498. P. Vandenabeele, B. Wehling, L. Moens, B. Dekeyzer, B. Cardon, A. von Bohlen and R. Klockenka¨mper, Analyst, 124 (1999) 169 –172. R.J.H. Clark and M.L. Franks, Chem. Phys. Lett., 34(1) (1975) 69 –72. K. Nakamoto, Infrared and Raman Spectra of Inorganic and Coordination Compounds, 4th ed., Wiley, New York, 1986. R.J.H. Clark, L. Cridland, B.M. Kariuki, K.D.M. Harris and R. Withnall, J. Chem. Soc. Dalton Trans., 16 (1995) 2577– 2582. F. Ellwanger-Eckel, Herstellung und Verwendung ku¨nstlerlicher gru¨ner und blauer Kupferpigmente in der Malerei. Diplomarbeit Institut fu¨r Technologie der Malerei, Staatliche Akademie der Bildenden Ku¨nste, Stuttgart, 1979. Ernst-Ludwig Richter, Zeitschrift fu¨r Kunsttechnologie und Konservierung, 2(1) (1988) 171–177. P.V. Huong, Spectrosc. Eur., 9 (1997) 24 –26.
661
P. Vandenabeele and L. Moens 35 36 37 38 39 40 41
42 43 44 45
46
47 48 49 50
662
G. Van Hooydonk, M. De Reu, L. Moens, J. Van Aelst and L. Milis, Eur. J. Inorg. Chem., 1 (1998) 639 –644. A. von Bohlen and F. Meyer, Spectrochim. Acta B, 52B (1997) 1053. P. Vandenabeele, B. Wehling, L. Moens, H. Edwards, M. De Reu and G. Van Hooydonk, Anal. Chim. Acta, 407 (2000) 261 –274. S.L. Vallance, Analyst, 122 (1997) 75R. M. Zerdoun Bat-Yehouda, Les encres noires au moyen aˆge ( jusqu’a` 1600). e´ditions du CNRS, Paris, 1983. C.-H. Wunderlich, Restauro, 6(94) (1994) 414 –421. B. Dekeyzer, P. Vandenabeele, L. Moens and B. Cardon, Actes du Colloque XII ´ tude du Dessin Sous-jacent et de la Technologie dans la peinture. La pour l’E peinture dans les Pays-Bas au 16e Sie`cle. Pratiques d’Atelier. Infrarouge et Autres Me´thodes d’Investigation, UCL, Louvain-La-Neuve, 1999, pp. 303 –316. M. De Reu, G. Van Hooydonk, P. Vandenabeele, L. Moens, A. von Bohlen and R. Klockenka¨mper, Scriptorium, 53 (1999) 352 –357. A. Derolez, The Library of Raphael de Marcatellis, Abbot of St. Bavon’s, Ghent (1437 – 1508). E. Story—Scientia Ltd, Gent, 1979. A. Arnould, The art historical context of the library of Raphael de Mercatellis, PhD Thesis, Ghent University, 1992. B. Bousmanne, “Item a Guillaume Wyelant aussi enlumineur” Willem Vrelant Un aspect de l’enluminure dans les Pays-Bas me´ridionaux sous le me´ce´nat des ducs de Bourgogne Philippe le Bon et Charles le Te´me´raire. Koninklijke Bibliotheek, Brussel en Brepols, Turnhout, 1997. P. Vandenabeele, B. Wehling, L. Moens, M. De Reu, G. Van Hooydonk, A. von Bohlen and R. Klockenka¨mper, In: E. Jerem and K.T. Biro´ (Eds.), Proceedings of the 31st International Symposium on Archaeometry (Vol. II), BAR International Series, Vol. 1043(II). Archaeopress-Archaeolingua, Oxford, 2002, pp. 825–827. B. Wehling, P. Vandenabeele, L. Moens, R. Klockenka¨mper, A. von Bohlen, G. Van Hooydonk and M. De Reu, Mikrochim. Acta, 130 (1999) 253 –260. I. Espadaler, M.C. Sistach, M. Cortina, E. Eljarrat, R. Alcaraz, J. Caban˜as and J. Rivera, Anales de Quimica, 91 (1995) 359 –364. L. Giuntini, F. Lucarelli, P.A. Mando, W. Hooper and P.H. Barker, Nucl. Instr. Meth. Phys. Res. B, 95 (1995) 389 –392. A. von Bohlen, P. Vandenabeele, M. De Reu, L. Moens, R. Klockenka¨mper, B. Dekeyzer and B. Cardon, Restauro, 2 (2003) 118– 122.
Chapter 15
Provenance analysis of glass artefacts Bernard Gratuze and Koen Janssens
15.1
INTRODUCTION
In this chapter, a number of different case studies are described, chosen from various historical contexts (from Protohistory to the end of the Post-Medieval period) and from various geographical areas (France, the Low Countries, Europe, the Indian world and the Pacific region) to illustrate the manner in which the chemical analysis of historical glasses can provide information on trade and provenance of glass artefacts in different historic periods. A long time before artificial glass was invented, pre-historic populations were using obsidian, a natural glass, to make tools. The importance of trace analysis to understand and reconstruct the trade and exchange patterns of this material during Neolithic times is described below. Later, during the Bronze Age, artificial glass in the form of glass beads was traded over long distances. Glass objects, found at Bronze and Iron Age archaeological sites in France, are used to show that the relationship between the chemical composition of an object, its chronology and the production area of the raw material can be used to build a distribution model of glass. By means of some examples, it is demonstrated that the composition of glass can also be used as a relative dating tool that can help to specify the stratigraphic attribution of an object. Finally, the importance of trace element determinations for identifying the provenance of some of the raw materials used in glass workshops is discussed. The international glass trade such as that existed between the Mediterranean region and India at the beginning of our era can also be studied through the analysis of glass. Here, trace element determination appears to be crucial to identify specific glass production types that have a very similar composition but a completely different origin. At the end of the 1st millennium AD , in Europe, the glass industry was subject to many important changes. In Carolingian times, i.e., at the Comprehensive Analytical Chemistry XLII Janssens and Van Grieken (Eds.) q 2004 Elsevier B.V. All rights reserved
663
B. Gratuze and K. Janssens
beginning of the European glass industry, some extraordinary glass recipes that emerged were in use. By using chemical analysis in combination with other complementary methods such as lead isotopes ratio analysis, it is possible to – identify the places of origin of these recipes, – follow the distribution of the resulting glass products throughout Europe. The difficulties associated with interpreting archaeometric data are illustrated by the study of the production of a 14th century French glass workshop. The great compositional variability encountered for this medieval glass workshop, situated between the production area of two different glass types, is interpreted in terms of glass recycling or plant ash mixing. In the 16th and 17th centuries, the transfer of technology from the North of Italy to the Low Countries can be documented by considering the different glass types and compositions used for making verre-a`-la-facon-de Venise (Venetian style luxury glass vessels). Finally as a second illustration of the increasing internationalization of glass trade during the Post-Medieval period, the composition of trade beads found on the small island of Cikobia-iRa (northeastern end of the Fiji Islands) is discussed. Glass is among the first artificial materials made by ancient civilizations. Although, the deliberate production of glass has really started in the middle of the 2nd millennium BC , its history dates back to the 4th millennium BC , with the invention of faı¨ence. At the beginning, this new material was used mainly to produce luxury objects (core formed vessels and jewellery) often linked to political and religious centres. However, these objects were rapidly exported far away from their production area. They may have been used as “trade beads” by the Near Eastern civilization centres towards the “European barbarians.” From the end of the Bronze Age onwards, the glass production gradually increases and glass items become more numerous. However, it is with the invention of glass blowing, during the 1st century BC , that glass will really become a common material. After that, glass trade develops rapidly, and glass items of diverse origin are found all along the ancient terrestrial and maritime trade roads. Nowadays, the term “glass” is mostly used to refer to the physical state of a material situated in between liquid and crystalline states; it can also be defined as a sort of polymerized state. Thus, one can speak of metallic glasses as well as of oxide and polymer glasses. In the recent past, glass was mainly referred to as a product obtained by melting silica and an alkaline compound (such as soda) together with lime, thus reducing its solubility in water. This combination of silica, soda and lime leads to the well-known soda-lime glass,
664
Provenance analysis of glass artefacts
which corresponds to an eutectic composition of this ternary mix (approximately 73% SiO2, 23% Na2O and 5% CaO). This particular soda-lime glass composition is frequently encountered in ancient glasses, with the notable difference that these glasses were probably made by mixing only two compounds: sand or crushed quartz pebbles (for silica) and plant ashes or mineral deposits (for soda or potash). The lime was probably introduced into the glass as impurities of these raw materials. Different additives, colourants and opacifiers (such as metallic oxides, salts, lead and antimony compounds) were also added to modify the physical properties of the material. All these components frequently feature relatively high levels of impurities, which reflect their origin or their manufacturing processes. These associations of minor and trace elements that characterize the production processes (recipes) and/or the origin of the materials can be used as fingerprints of the glass. Thus, chemical characterization of ancient glass items should allow to identify the different production workshops and accordingly, it should be possible to follow the objects from the workshop towards the consumption site. However, glass objects, similar to metallic artefacts, were relatively easy to reprocess. Glass recycling renders analytical data interpretation more complicated, and often imposes limits on the generality of the conclusions of glass provenance studies. Sometimes, the geochemical signature (i.e., mainly the isotopic composition) of some glass components (such as lead, strontium or oxygen) can help to identify the type and origin of the raw materials that were employed. Isotopic analysis can also be used to study the geographical diffusion of objects, starting from their point of manufacture. Finally, whereas the term “archaeological glass” usually refers to a deliberately produced material obtained by melting together sand and an alkaline compound, it can also be applied to another material: obsidian. Obsidian is a natural volcanic glass, which has been widely used to make tools or other types of artefacts, such as bowls, vases or parts of statues since pre-historic times. 15.2
OBSIDIAN, A NATURAL GLASS USED SINCE THE PALEOLITHIC ERA
Obsidian is a rhyolithic glass, formed when silica- and alumina-rich lavas cool rapidly at the surface of the earth. Although it consists almost entirely of glass, obsidian may contain some crystalline phases (less than 15%)
665
B. Gratuze and K. Janssens
and gas bubbles. The major, minor and trace element composition reflects its geological history, and obeys the relatively simple geochemical laws. While major components have been used to source the origin of this type of material [1– 3], more successful results were obtained by studying trace elements [4– 6] and especially the hygromagmaphile ones (Zr, Nb, Y, Ba, Sr, Th, U and the rare earth elements). Among all the different geological materials used by pre-historic population groups, obsidian is one of the most useful and successful to study ancient trade and exchange paths [7]. This is due to two important properties of that material: – its sources are relatively rare and well localized, – its chemical composition is characteristic of its origin. Various characterization methods are used to source obsidian: dating, isotopic composition, mineralogical studies and different methods of chemical analysis. Among them, the chemical analytical techniques are the most widely used. From an archaeological point of view, the results obtained could be interpreted in terms of exchange mechanisms and cultural contacts between pre-historic populations, as shown by the study of the obsidian artefacts found at different archaeological sites in upper Mesopotamia (Syria) and at Shillourokambos in Cyprus (see map in Fig. 15.1). Obsidian trade studies in the Near East have shown that during Neolithic times, worked obsidian mainly originated from two areas: Cappadocia and Taurus. In Cappadocia, the East Go¨llu¨dag sources were the most widely employed, while in Taurus, different outcrops around the town of Bingo¨l and around the Nemrut Dag volcano were exploited [8]. In Cyprus, the pre-ceramic site of Shillourokambos, not far from Limassol, has yielded 217 obsidian artefacts, in a context dated between the ninth and the 8th millennium BC . It is, for the moment, the oldest pre-ceramic Neolithic dwelling site in Cyprus [9]. In a study where 74 blades were selected for chemical characterization [10], 40 were analysed by fast neutron activation analysis (FNAA), 31 by laser-ablation inductively coupled plasma mass spectrometry (LA-ICP-MS), and the remaining three with both methods. The results obtained show that 71 artefacts come from the Cappadocian volcanic zone of the Go¨llu¨dag, while the other three come from the Nenezi Dag (Fig. 15.2). These results lead to important revelations concerning the early history of Cyprus. They show the importance of maritime trade between Cyprus and Turkey at the beginning of the Cyprus pre-ceramic period. The rarefaction of obsidian importation, which is observed during the next centuries at
666
Provenance analysis of glass artefacts
Fig. 15.1. Localization of the different obsidian sources and archaeological sites. Obsidian sources are represented by squares, archaeological sites by triangles.
Shillourokambos and at other important classical sites, of the 7th millennium BC , such as Khirokitia and Tenta [11], could be explained by a progressive cultural isolation of the island. From the first observation, it could be concluded that obsidian was probably traded in the form of pre-shaped artefacts rather than as raw material. Important obsidian knapping workshops are known in Cappadocia, and more especially in the Go¨llu¨dag area. The presence of their productions as early as the 9th millennium BC in Cyprus shows their importance at the beginning of the Neolithic period. A number of questions remain concerning the trade route followed by these obsidian artefacts: was there a direct maritime trade between Cyprus and the Anatolian coasts or was there a distribution from Levant? The importance of Cappadocian sources is also apparent in the compositions obtained from different Syrian archaeological sites, located in upper Mesopotamia: Cheikh Hassan, Dja’de, Halula, Jerf el Ahmar, and Mureybet [12 – 14]. These different sites are spread in date between the 11th
667
B. Gratuze and K. Janssens
Fig. 15.2. Simplified Ba/Zr versus Rb/Zr diagram for Shillourokambos obsidian artefacts and geological obsidian samples from East Go¨lludag and Nenezi Dag sources. Some data points overlap.
millennium and the 5th millennium BC . The most ancient, those before 7500 (Natoufian, Khiamian and PPNA periods), represent the transition between hunting/gathering and agriculture, at the beginning of the Neolithic (Mureybet phase I and II, Jerf el Ahmar), and then the beginning of an agricultural economy (Mureybet phase III and Cheikh Hassan). Others belong to the later pre-ceramic Neolithic period, the PPNB, between 7300 and 6000 BC : these are Mureybet phase IV, Dja’de and Halula. Some deposits of the Halula and Dja’de sites belong to very last phases, appropriate for the desert, of the PPNB (between 6000 and 5500 BC ), contemporary with the first ceramics in other regions. Other deposits belong to even later phases with ceramics at the end of the 6th millennium. It is also at Halula that, at the end of the Neolithic period, Halafian samples are encountered. Although in all these sites, obsidian objects remain exceptional in relation to the rest of the artefacts recovered, the distribution of the occurrence of the obsidian sources [12,13] for the different chronological phases shows that (Fig. 15.3): – during the hunting/gathering occupation phases, obsidian exclusively originated from Cappadocia, and more especially from the East Go¨llu¨dag sources,
668
Provenance analysis of glass artefacts
Fig. 15.3. Near Eastern obsidian sources’ distribution at different chronological phases, for five Middle Euphrates archaeological sites: Cheik Hassan, Dja’de, Halula, Jerf el Ahmar and Mureybet. Artefacts numbers are given at the bottom, from Ref. [12].
– from the late PPNA period onwards, i.e., at the beginning of agriculture in that area and its diffusion towards Taurus (East Anatolia), obsidian comes from both Cappadocia and Taurus. According to M.C. Cauvin, the neolithization of Taurus, which was introduced from the Middle Euphrates area, explains that an obsidian trade was observed between these regions. It is interesting to note that Cappadocian obsidian arrives in the Middle Euphrates at a period where, up to now, no such ancient installation phases were found in Cappadocia. This hiatus may be artificial and only new research will allow answering the following questions: – was obsidian imported as raw blocks or pre-shaped artefacts? – did the Middle Euphrates population bring obsidian themselves from Cappadocia? We can also observe that the East Go¨llu¨dag obsidian seems to have been extracted exclusively or nearly exclusively in Cappadocia until the middle PPNA. The exploitation of others sources, such as Acigol and Nenezi Dag, starts, or becomes more important, from the late PPNB period onwards (some Nenezi Dag obsidian was found in the pre-ceramic levels of Shillourokambos). Provenance studies show that the East Go¨llu¨dag obsidian is the most widely traded material in the western part of the Near East, from the upper Paleolithic to the Bronze Age. Its diffusion area spread towards the occidental part of Turkey (Troad and Marmara Sea), Cyprus, the Dead Sea,
669
B. Gratuze and K. Janssens
the northwest of Iran and the Eastern Taurus. Recent studies have shown that East Go¨llu¨dag obsidian is the most suitable for knapping [15]. 15.3
BRONZE AND IRON AGE GLASSES
15.3.1 Neolithic first artificial glassy materials and the discovery of glass during Bronze Age Until the middle of the Neolithic, obsidian will remain the only glassy material used by pre-historic populations. It is only around the middle of the 4th millennium BC that the first artificial glassy material was probably discovered in the Near East: glazed quartz and steatite, and faı¨ence [16]. Faı¨ence is a material made of sintered quartz, which usually presents a homogeneous core of powered quartz or sand, covered with an alkaline glaze (a type of glass layer that forms when quartz reacts with an alkaline mixture). When the quartz grains are fully embedded in the glass phase, the term “glassy faı¨ence” is used. The amount of glass in the core varies according to the method of production of faı¨ence (temperature, humidity…). It is also in Mesopotamia, two millennia later (in the second half of the 2nd millennium BC ) that glass production started. The real date of glass invention is still subject to controversy: the first half of the 3rd millennium BC (around 2500 BC ) is sometimes proposed, although the earliest glass may have been produced accidentally while firing faı¨ence. It is nowadays widely accepted that the actual glass production started around 1600 BC in Mesopotamia, and shortly after that in Egypt [17]. Ancient glass was obtained by melting together sand or crushed quartz pebbles, with a fluxing agent. The flux was either plants ash or mineral soda, extracted from natural deposits (such as those located near Wadi Natrun in Egypt, which gave its name to the natron). The glass obtained by melting silica and soda is a soluble product; addition of a stabilizer (lime or alumina) is needed to obtain a stable glass. In Protohistory and Antiquity, it is thought that these compounds were introduced into the mixture as impurities of sand and the fluxing agent. Some additives, metals, salts or oxides, were intentionally added to the glass in order to change its physical properties (colour, opacity). Different glass compositions could be obtained by changing the ingredients of the recipes, or their proportions: therefore, glass composition reflects its origin, and compositional study should allow following its diffusion. Nevertheless, we must not forget that glass is often recycled, and that the interpretation of data therefore is sometimes more complex.
670
Provenance analysis of glass artefacts
15.3.2
When trade beads reached Europe
The first glass objects, which were mainly beads, were used as ornaments. We will not discuss their function here (cultural, apotropaic…) but rather their geographical spread, from the eastern workshops toward the western regions. In western countries, faı¨ence and glass beads are found in many archaeological contexts, from the end of Neolithic period onwards. For the pre-historic and Protohistoric periods, these objects were rapidly interpreted as imported objects from the great civilization centres, such as Mycene, Egypt or Mesopotamia. A chronology was established using the cross-dating method, but before the 1970s, no archaeometric investigations of this material were undertaken. Due to the scarcity of these objects, the first analyses, which were destructive, were made on a small number of items only [18– 20]. It is only recently that non-destructive analytical tools have allowed the study of large artefact populations [21– 24]. The analytical method predominantly used for the studies described below, LA-ICP-MS [25], allows the determination of a large number of elements, from major to trace levels (see Chapter 7). It helps to characterize the different ingredients used during the glass-making process. The size of a typical ablation pit is below 80 mm, i.e., invisible to the naked eye. The trace element level sensitivity and the possibility to remove, in an invisible manner, the corrosion layer that covers the objects (see Chapter 16) are among the main advantages of this method, compared to X-ray microanalysis in a scanning electron microscope. This latter method was the most widely used for these types of studies in the recent past. As long as the studied material is homogeneous in composition, LA-ICP-MS is a perfectly suitable method, making it very suitable for use in glass studies. However, due to the heterogeneous structure of faı¨ence, this method is less suitable to study this type of objects. In France, a large research program on the origin and circulation of glass beads from the Bronze Age to the Iron Age was started in the middle of the 1990s [26]. It is the widest program ever undertaken in Europe: more than 600 glass beads, representing 150 archaeological sites, ranging in date from Chalcolithic levels to the early Iron Age, have been analysed. The Protohistoric glass beads that were studied belong to three main glass types (Fig. 15.4a and Table 15.1). For a few objects, other miscellaneous glass types are encountered, but some doubts exist about their dating.
671
B. Gratuze and K. Janssens
Fig. 15.4. (a) Ternary diagram of Na2O–MgO þ K2O –CaO, showing the classification of the main glass chemical groups according to the fluxing agent that was used. Groups 1, 2 and 3 characterize Bronze Age and Iron Age glass productions. (b) Diagram of CaO versus the K2O/MgO ratio for plant ash soda-lime glass allowing splitting up of plant ash soda-lime glasses into two groups: (i) Mycenaean [79] and Egyptian [27] glasses are below the line, on the left part of the diagram, together with Middle and late Bronze Age European glass beads (MgO . K2O); (ii) medieval plant ash soda-lime glasses ([65] and our laboratory data) lie above the line on the right part of the diagram.
15.3.3
Middle Bronze Age plant ash soda-lime glasses
The most ancient glasses appear to have a plant ash soda-lime composition. Annular, spherical and, in some cases, open ring beads are found. The glass is usually transparent turquoise green, transparent blue or opaque turquoise. Among the few other colours encountered, we can mention some black, brown, translucent pale yellow or more rarely colourless glass. Some beads are decorated with dots, circles, spiral or right lines of white, yellow or red glass.
672
Provenance analysis of glass artefacts
Fig. 15.4 (continued).
This glass type seems to appear in the Near East, around 1600 BC , and characterizes Near Eastern, Mycenian and Egyptian glass beads’ production [27]. Chemically, this glass is close to medieval plant ash soda-lime glass, which was produced in Mediterranean Europe during the (post) Medieval era. Both glasses are characterized by relatively high magnesia and potash contents. However, magnesia levels are superior to potash levels in the Near Eastern glass, while the reverse is observed for the Medieval glass (Fig. 15.4b). Higher concentrations of magnesia, compared with potash, are also observed for Islamic glasses. Therefore, this parameter can also be used for geographical discrimination of glass. Glass objects with this composition are mainly found in middle and late Bronze Age levels. Some of them are attributed to more ancient contexts, such as those of the early Bronze Age and Chalcolithic periods, but their datings are often not well established. In all the cases, the occurrence of glass objects remains exceptional in these contexts. For more than 90% of the archaeological sites, the number of glass items per site varies from one to a maximum of five.
673
B. Gratuze and K. Janssens TABLE 15.1 The three main glass groups encountered for Protohistoric beads, according to their flux Chemical groups
Na2O (%)
K2O (%)
CaO (%)
MgO (%)
Plant ash soda-lime glass Soda –potash glass or mixed alkaline glass Mineral soda-lime glass
12–18
.1.5 and , MgO
4 –9
. 1.5 and .K2O
4–13 ($Na2O)
2 –4
, 0.5
,1.5
4 –9
, 1.5
4–12
12–23
Concentrations of soda (Na2O), potash (K2O), lime (CaO), magnesia (MgO) are in wt%. The other main components for a copper blue glass are silica (SiO2) 55– 77%; alumina (Al2O3) 0.5–5%; phosphorus oxide (P2O5) 0.2– 3%; chlorine (Cl) 0.1–1.5%; iron oxide (Fe2O3) 0.2 –2%, manganese oxide (MnO2) 0.1–2% and copper oxide (CuO) 1– 5%. Special glass types, such as opacified glass, are not considered here.
15.3.4
Late Bronze Age mixed soda –potash glasses
During the late Bronze Age, large quantities (up to 70 beads or more for some archaeological sites) of small translucent turquoise annular beads, made with a new glass type, are found. This glass type was also used to make: – dark blue or red small annular beads, – barrel-shaped beads, decorated with a continuous spiral of opaque white glass, – ring beads, decorated with three or four horns, formed by two stratified layer of white and blue glass, and – ring beads decorated with three white circles. Another type of bead, made of clay covered by a thin outer skin of glass, decorated by three series of three concentric glass circles, also belongs to this group. This compositional type is characterized by high potash and soda levels (usually soda is a little bit lower than potash), low lime and magnesia contents, and high silica concentrations (ranging from 73 to 76%). Copper, which is the main colouring agent encountered, is also present in high concentrations (from 2 to 5%). Beads of this composition are found throughout the Western Europe: Ireland, England, France, Germany, Switzerland, and Italy [28,23]. This glass type characterizes late Bronze Age European glass objects.
674
Provenance analysis of glass artefacts
This group of glass probably originates from the site of Frattesina, in the province of Rovigo, in northern Italy. In this site, datable from the 11th century to the 8th century BC , traces of glass-working activities were discovered. The glass found on this site presents the same unusual mixed soda – potash composition [29 –31]. The study of early and middle Bronze Age faı¨ence beads reveals that, for some of them, the glaze layers’ composition is similar to the glass produced at Frattesina (Table 15.2). The composition differs only in the sand/flux ratios, silica levels being higher in the glaze than in the glass [21]. The transition from faı¨ence making to glass making arises probably at the beginning of the late Bronze Age. Note that the presence of quartz grains is frequently observed in the annular or barrel-shaped beads of this group. This led some authors to use the term “glassy faı¨ence” for those productions [32]. This term may be applied to some beads which may have been fashioned like faı¨ence objects (such as some barrel-shaped beads, decorated with white spiral line). However, it cannot be applied to the small annular beads, which were fashioned like glass beads, by stretching and winding a molten material around an axe. In some late Bronze Age archaeological sites, plant ash soda-lime glass beads are found together with mixed soda – potash glass. At the end of this period, both glass types disappeared and a new glass type occurred. 15.3.5
Iron Age and Antiquity natron soda-lime glasses
At the beginning of the early Iron Age, Hallstatt period, glass findings (which remain exceptional for all the Bronze Age phases) become more frequent. Bead typologies are more and more complex: larger beads and new types of decoration, by mixing different coloured glasses, appeared. The profusion of the glass finds indicates that glass has entered into a new period of its history: the material is rare until the end of the Bronze Age, and it becomes common during the Iron Age. At this period, a chemical change occurs in glass composition, which may explain the new status of glass. Iron Age glass also belongs to the soda-lime glass family, but is characterized by low magnesia and potash levels (MgO and K2O , 1.5%). Instead of soda – plant ash, natron is now used as a fluxing agent. Natron is a sodium carbonate compound (also containing sulphates and chlorides) extracted from a natural mineral deposit. The profusion of glass that is observed may be explained if we consider that, in opposition to plant ashes, natron was continuously available and did not require special preparation.
675
676 TABLE 15.2 Average composition of some late Bronze Age glass beads belonging to the mixed soda– potash glass group compared to the composition of some early Bronze Age faı¨ence found in France Concentration (wt%)
Some French faı¨ence beads, analysis of the glass phase [21,22,37,a]
United Kingdom and Ireland [23,80]
Germany [28]
Frattesina, Italy
FNAA and LA-ICP-MS
LA-ICP-MS
MEB
SEM, ICPAES, ICP-MS
LA-ICP-MS [a]
Turquoise glass
White glass
Dark blue glass
La Peyre`yre, Run ar Valferrie`res, Muret Justicou, Se´ranon Crozon
Turquoise glass
Dark blue glass
AAS, OES [31]
AAS, SEM [30]
75
SiO2
76
74.6
78.7
85.9
80.3
71.9
74.8
75.2
80.6
75.8
Al2O3
2
3.6
2.8
1.8
1
0.51
1.8
1.7
1.9
2.3
1.8
1.6
Fe2O3
0.7
1.65
0.89
0.41
0.11
0.08
0.6
0.56
0.59
0.74
0.61
0.65 9
85.1
K2O
8.3
7.5
9.4
4.8
4.7
5.4
8.5
10.6
9.1
6.5
9.9
Na2O
5.7
4.2
4.9
2.4
5.2
4
7.7
4.6
5.6
5.9
5.9
5.5
CaO
2.11
6.1
1.83
1.1
0.5
1.3
1.98
2.3
2.09
2.77
1.85
1.5
MgO
0.6
0.63
0.56
0.2
0.1
0.25
0.7
0.58
0.6
0.63
0.58
0.6
Cl
0.14
0.27
0.23
0.2
0.3
0.1
0.05
0.17
0.16
CuO
3.3
0.83
0.24
2.8
7.0
3.1
3.2
4.2
0.15
3.6
4.97
CoO
0.0037
0.0042
0.33
NiO
0.01
0.013
0.24
0.003
0.016
As2O3
0.01
0.008
0.11
SnO2
0.16
0.034
0.0078
0.36
0.46
0.44
0.81
(a) Gratuze, unpublished data.
0.24
4.2
0.23
0.0053
0.13
0.002
0.52
0.0033
0.26
0.67
0.0025
0.11
B. Gratuze and K. Janssens
France [21,22,39,a]
Provenance analysis of glass artefacts
For the next 16 centuries, i.e., until the 8th century AD , the natron sodalime glass would become the predominant glass in the Mediterranean regions and Western Europe. It spread throughout the entire Roman Empire and accordingly was erroneously called “Roman glass.” The terms “SyroPalestinian” or “Near Eastern” glass, which tend to be used nowadays, refer to its origin. If during the Iron Age, different composition types could still be distinguished inside this group of glass (variation of the silica/natron ratios, use of different colouring recipes…) [33], from the La Te`ne period onwards, the composition of this glass rapidly became very uniform, so that provenance studies are more difficult to carry out. This may point to the existence of large primary workshops and/or to an intense reprocessing of the glass. In some cases, subtle variations or trends in the trace level content can be observed, as is illustrated in Fig. 15.5, where the Ti and Zr content of 1st – 5th century AD glass excavated in the Roman border town of Maastricht is plotted [34]. The tendency to find more Ti and Zr in the glass of the later centuries suggests either a slow changeover from one silica source to another or the combined effect of the use of a Ti- and Zr-rich type of sand for making new glass and extensive remelting of old and new glasses.
Fig. 15.5. Relation between Ti and Zr in Roman glass excavated in Maastricht, a Roman border town, dated to the 1st –5th century AD , suggesting a change in the silica source employed for glass making.
677
B. Gratuze and K. Janssens
15.3.6
Protohistoric glass trade routes
The study of French Protohistoric glass objects brought new information on glass trade in Western Europe, during the Bronze Age and Iron Age. For the middle Bronze Age, glass beads mainly came from the eastern part of the Mediterranean Sea. Maritime trade routes were probably the commonest way to reach the western regions and the British Islands. The glass artefacts found on Bronze Age shipwrecks of Cap Gelidonya and Ulu Burun in Turkey have shown that maritime glass trade occurred in these early periods [35,36]. Some European middle Bronze Age glass beads are typologically and chemically very similar to those found on the Ulu Burun shipwreck. Glass beads with the same chemical composition were also found in French Britanny, and more especially on the Ouessant Island, which is on the Atlantic maritime road to the British Isles [37]. For the late Bronze Age, the results show that a large majority of objects comes from Northern Italy. Trade routes, passing through the Jura mountain cols, the Doubs valley, the Alpine cols and the Rhone valley, were highlighted. The presence, on the Ouessant Island, of one bead, made of clay covered by a thin outer layer of blue glass, decorated by series of concentric white glass circles, which is identical to those found at Rathgall in Ireland [38], illustrates the persistence of this maritime road. This unusual type of material was only found in four different archaeological sites, one in Ireland and three in France ([21] and unpublished results): Ouessant, Fort Harrouard and Gouaix (the last two sites are close to the Parisian region). This may illustrate the existence of particular trade axes throughout the different periods. 15.3.7
Glass chrono-typo-chemical models: a dating tool?
For these Protohistoric periods, glass compositional studies also have allowed to revise the dating of some beads. A few of the studied beads were found in dolmen, together with objects belonging to different periods, ranging sometimes from Neolithic to Antiquity, or even Middle Age or Modern times [39]. A model based on the typology, the chemical composition of both major, minor, and trace elements of the bead and on its chronology was built using the well-dated glass objects. This chrono-typo-chemical model provided the possibility to “date” more or less precisely the objects, depending on their chemical compositions and typologies. This is the case of three glass beads found in the Rancogne cave (Charente, southwest of France) in archaeological contexts, whose dating are
678
Provenance analysis of glass artefacts
spread from the early Bronze Age to the late Iron Age [40]. Traces of occupation until the actual period are also found. The typology of one of these beads (no. 52/7), made in a wound polygonal dark blue glass, is unusual for the Bronze Age periods, and led to an attribution to later periods, such as the Iron Age or Antiquity. The chemical analysis of these three beads (Table 15.3) gave the following results [21]: – the first one (no. 52/6), a copper blue spherical bead, belongs to the plant ash soda-lime glass type, – the second one (no. 52/8), a copper blue ring bead, decorated with three white “glassy faı¨ence” circles, is a mixed soda – potash glass, – the third one (no. 52/7), the wound polygonal dark blue bead, is made with a lime – potash glass. The composition of the first two beads is in perfect agreement with their attribution to Bronze Age periods. The second one belongs probably to the late Bronze Age phase. The last one has the composition of Medieval or TABLE 15.3 Composition, obtained using LA-ICP-MS, of the glass beads found at Rancogne Cave (oxides in wt% and elements in ppm; 1ppm ¼ 1 £ 1024%) Oxide
SiO2 Na2O K2O MgO CaO Al2O3 Fe2O3 Cl P2O5 CuO Mn Co As Bi Ni U Sn
No. 52/7
68.6 0.25 15.7 0.40 13.1 0.40 0.52 0.09 0.78 0.005 173 224 1350 206 563 7.4 115
No. 52/8
No. 52/6
Blue ring
White circle
77.3 4.51 9.42 0.64 2.50 1.94 0.57 0.035 0.51 2.55 155 50 50 4 60 0.4 1853
76.8 4.78 7.83 0.75 3.85 2.26 0.75 0.14 1.13 1.73 195 95 121 2 240 0.4 860
69.4 13.3 1.4 3.4 5.8 2.7 0.9 1.1 1.5 0.5 280 8.3 36 0.4 23 8 213
679
B. Gratuze and K. Janssens
Modern glasses, and the bead cannot belong to the Iron Age or Antiquity phase. By using trace elements, the production period can be defined more precisely. For that particular bead, cobalt, which is the main colouring agent used to obtain dark blue glass, is associated with arsenic, nickel, bismuth and uranium (Table 15.3). In a recent study, we show that, according to trace elements analysis, different cobalt pigments are successively found (Table 15.4), in Near Eastern, Mediterranean and European glasses, from the Bronze Age to the 18th century AD [41]. The Co – As – Ni – Bi – U association is typical of objects produced after the 16th century AD . Therefore, the bead can be dated to the Post-Medieval period. 15.4
GLASS TRADE TOWARDS AND FROM CENTRAL ASIA AND THE INDIAN WORLD DURING ANTIQUITY
For the Iron Age and Antiquity, maritime trade is well established: numerous shipwrecks, containing glass ingots and recycled waste glass, were found in the Mediterranean Sea [42]. However, the new organization of glass-making activities as primary workshop (where raw glass was produced) and secondary workshops (where objects were fashioned) renders provenance and trade route studies more difficult. Only the presence of some particular type of chemical compositions or typologies can help identify the origin and the diffusion of some productions. For example, some red and green millefiori objects, for which plant ash soda-lime glasses were used instead of the usual natron glass, are found in the 1st century AD contexts [43]. These objects were found in different sites located in France (Lattes in the south and Dijon in the east) and Italy (Aquileia in the northeast). Apart from these millefiori objects, some green glass ingots were found in Switzerland at Avenches [44]. This composition, which is unusual (for European objects, dated from the 1st century AD , this composition represents less than 1% of the used glass) shows that at this period some glass-making centres were still using plant ash recipes to make glass. The glasses with this composition were all coloured by copper (green and red copper glass). The plant ash recipe may still have been used by some Mesopotamian or Near Eastern glass workshops. It is known from recent studies that some glass workshops located in the Middle East and Central Asia (such as Bara in Pakistan) used this recipe during this period [45]. The presence of this type of glass in Western Europe might reflect the activities of some glass trade routes between Central Asia and the Mediterranean region. The anonymous author of
680
TABLE 15.4 Main chemical characteristic associations found for cobalt pigments used for glass and glaze productions in Western Europe and Mediterranean regions ([41] and unpublished results) Chemical association
Period
Provenance
1
Co –Ni –Zn–Mg– Al
Middle and late Bronze Age and beginning of Halstatt
2
Co –Ni –As
Late Bronze Age
3
Co, Co– Sb?, Co –Cu?, Co–Mn?
4
Co –As –Ni –Fe þ Pb –Sb–Sn –Zn
5
Co –As –Fe, and sometimes Cr and Zn
6
Co –Zn–Pb –In–Fe
7 8
Co –Ni –Mo –Fe Co –As –Ni –Bi –W–Mo– U–Fe
9
Co –Ni (less As)
From the beginning of Iron Age until the end of the 12th century in Europe From the end of the 7th or the beginning of the 8th century in the Near East From 12th century to the 15th century in Near Eastern glazed pottery From the end of the 12th century to the end of the 15th Around 1500 (transition group) From the beginning of the 16th century to the 18th century 19th century
Cobalt from alum deposits in Egyptian oasis, associated with plant ash soda-lime glass and first Halstatt natron soda-lime glass European (may be Erzgebirge in Germany), associated with mixed soda –potash glass from Frattesina Unknown?, Near East, associated with natron soda-lime glass May be Iranian ores from Kashan, associated with the last natron soda-lime Islamic glass May be Iranian ores from Kashan, associated with Medieval Islamic ceramics Erzgebirge (Germany), Freiberg? Erzgebirge (Germany)? Erzgebirge (Germany), Schneeberg
?
681
Groups 4 and 5 characterize Near East cobalt pigments, while the other groups are associated with glasses found in Europe from the Halstatt period until the 19th century.
Provenance analysis of glass artefacts
Group
B. Gratuze and K. Janssens
the Periple (1st century AD ) has described the exportation of glass items from the Mediterranean regions towards the Indian world [46]. Therefore, an occasional importation of glass from Near Eastern inland workshops or from Mesopotamia and Central Asia is not impossible. If we compare the composition of the different millefiori glasses, we notice that the red glasses used to make the millefiori artefacts from Lattes and Dijon have a similar major and minor element composition (Table 15.5). Furthermore, the colouring ingredients are also identical. This suggests a common origin of these glass items. Therefore, it should be possible to reconstruct the trade routes of this type of material, if we find other objects with the same composition throughout Europe and the Mediterranean regions. On the other hand, it is not possible to establish a relationship between the western and eastern glasses. The glass recipe from Bara, Pakistan, is far different from the one used to produce the red glasses found in France and Italy: – first, the glass from Bara contains less copper and a very low amount of tin, antimony and lead, – second, the sand used at Bara is richer in alumina and the glass contains a lower amount of lime. Then, if from a chemical point of view, it is not easy to follow the diffusion of “Near Eastern” glass throughout the Mediterranean area, it seems easier to study its diffusion outside this region. Dussubieux recently studied the glass trade between the Mediterranean world and the Indian sub-continent [45]. The author shows the importance of trace elements’ determination for provenance studies. By means of LA-ICPMS, 550 glass items from Afghanistan, Pakistan, South India, Sri Lanka, Thailand, Malaysia, Indonesia and Cambodia were studied. The samples are classified into nine different glass types. Some of these types represent local productions, while others reflect the importation of glass. Among them, two groups are particularly interesting, both belonging to the family of mineral soda-lime glass, and are hardly distinguishable as long as their major components are concerned (Fig. 15.6). The considered glass samples come from various archaeological sites located in India, Sri Lanka, Vietnam, Cambodia, Indonesia, Thailand and Afghanistan (see map in Fig. 15.7). All these sites are spread in time between the 3rd century BC and the 4th century AD . Among them, we find a majority of dark blue glasses and two gold-foiled glass beads. A classification using only major elements (silica, alumina, lime, potash, magnesia, chlorine, phosphorus, iron and manganese)
682
TABLE 15.5 Unusual plant ash soda-lime glass composition found for some European 1st century AD millefiori and tesserae glasses and a Pakistani glass workshop of the same period (Bara, average composition of 8 red glasses) Oxide France Dijon (Coˆte-d’Or)
Millefiori 253
Millefiori 503
Millefiori 504
Switzerland Pakistan
Aquileia
Avenches
Tesserae 2 Raw glass
Green glass Red glass Green glass Red glass Green glass Red glass Red SiO2 65.5 Na2O 14.6 1.93 K2O MgO 1.67 CaO 7.1 Al2O3 2.3 Fe2O3 1.66 Cl 0.72 P2O5 0.43 CuO 2.57 MnO 0.99 Sb2O3 0.03 PbO 0.09 SnO2 0.33
57.9 12.1 2.2 2.3 11.8 3.3 1.85 0.57 0.62 2.31 0.78 0.36 3.3 0.35
61.8 16.6 2.19 1.75 8.55 2.7 1.50 1.39 0.85 1.79 0.50 0.025 0.13 0.22
55.6 13.2 1.72 2.00 11.5 3.7 1.85 1.35 0.64 2.42 0.37 0.40 4.99 0.25
62.4 17.4 1.53 1.62 7.71 2.7 1.53 1.32 0.62 1.79 0.84 0.11 0.22 0.23
55.8 13.3 1.85 2.15 12.4 3.4 1.98 1.08 0.90 1.64 0.44 0.37 4.31 0.39
60.1 12.8 2.7 2,9 7.8 1.7 1.58 0.82 1.05 3.39 0.61 0.012 3.5 0.89
Bara Beads and rods
Green glass Red glass 63.3 15 1.3 1.7 6.3 3.9 1.8 0.55 3.8 0.42 0.10 0.63
60.8 14.6 4.7 4.0 6.8 4.9 0.79 0.78 0.4 0.99 0.059 0.0004 0.036 0.012
Provenance analysis of glass artefacts
Lattes (Herault)
Italy
683
B. Gratuze and K. Janssens
Fig. 15.6. Minimum, maximum and average contents for the mineral soda-lime glasses found in the Indian regions and in Southeast Asia (dotted line) and Near Eastern natron glass (full line).
Fig. 15.7. Localization of the different studied Indian and Southeast Asian archaeological sites. Chinese and Vietnamese provinces are reported in bold type. The archaeological site of Bara, mentioned earlier, and the Red Sea Egyptian port of Berenike are also reported.
684
Provenance analysis of glass artefacts
would have led to the definition of only one group. Based on the great similarity observed between these glasses and the Near Eastern glass type, one would have concluded that all these artefacts were imported from the Mediterranean regions. If major element studies do not bring any other useful information, it is not the case with trace elements, which allow dividing this group of glass into two sub-groups: one of them remains closed to Near Eastern glass, while the second presents higher level of concentration for vanadium, titanium, zirconium, rubidium and uranium (Fig. 15.8). Although the average trace elements’ level is slightly higher in the first sub-group than the one that is usually encountered for Near Eastern glasses, they matched the concentration levels found for dark blue Near Eastern glass ingots found on a caravan road in the Egyptian desert. Therefore, the origin of the glass samples of this group is probably the Mediterranean area. The origin of the second sub-group is a priori less clear. Two hypotheses are possible: either an Indian or a Chinese origin can be proposed. However, as we have noticed earlier, this sub-group contains a majority of dark blue glasses coloured by cobalt. According to the results of Dussubieux, the cobalt pigment used for the dark blue glass beads of this group is associated with manganese and comes probably from asbolane deposits. Manganese itself is associated with different trace elements, such as barium and cerium. These trace elements allow splitting up the soda-lime dark blue glasses into two groups (Fig. 15.9). In the first one, characterized by low
Fig. 15.8. Diagram of Zr versus U for the mineral soda-lime glasses found in the Indian regions and in Southeast Asia. The group splits up into two sub-groups: one that is similar to the Near Eastern natron soda-lime glass, and the second, which originates probably from the Indian or Chinese regions.
685
B. Gratuze and K. Janssens
Fig. 15.9. Diagram of Co versus Ce for the dark blue soda-lime glasses found in the Indian regions and Southeast Asia. The group splits up into two sub-groups: one that is similar to the Near Eastern dark blue natron soda-lime glass, and the second, which originates probably from the Chinese regions. This strange Co–Ce correlation is explained if one considers that cerium is correlated to manganese, which itself is correlated to cobalt in the asbolane.
cerium values, we find the glass with low zirconium and uranium concentrations. In the second, characterized by high cerium values, we find all glasses with high zirconium and uranium contents. This cobalt pigment extracted from asbolane ores features similarities to the one used by Chinese glass workers. Asbolane is found in the south of China, in the Yunnan, Zhejiang and Jiangxi provinces [47,48]. Therefore, a Chinese origin of this group is probable. These results on cobalt confirm also the Mediterranean origin of the first sub-group (low zirconium and uranium glasses). Glass items of this subgroup come from Afghanistan (Begram), India (Arikamedu and Karakaidu), Sri Lanka (Anuradhapura and Ridiyagama) and Thailand (Khlong Thom). It is important to notice here that except for Afghanistan, glass items from Sri Lanka, India and Thailand are all dark blue glass beads or gold-foiled glass beads. The two gold-foiled glass beads come from India (Arikamedu) and Thailand. Artefacts of the second sub-group come from India (Arikamedu), Sri Lanka (Anuradhapura and Ridiyagama), Vietnam (Lach Truong), Cambodia (Ankor Borei), Indonesia (Ulu Leang) and Thailand (Khlong Thom). In this group, we find a majority of dark blue beads, but other colours are also found (turquoise blue, yellow, green…). The importance of trace element studies is particularly highlighted in this example, where they allow characterizing both the sands and the
686
Provenance analysis of glass artefacts
colouring agents. It shows that their determination is often the only way to distinguish glass types with closed compositions. This example shows that during the 1st century BC and AD , different glass workshops, located in the Near East and in an area that corresponds nowadays to the south of China, were producing two glasses of closed chemical compositions. The raw glass or the beads, made with those glasses, have been traded far away from the original workshops. Nowadays, archaeologists are finding these two types of glass beads in Sri Lanka, South India or Thailand, in archaeological sites of the same periods or sometimes within the same archaeological sites. A too rapid examination of that material would lead to the conclusion of a Mediterranean origin of these beads, or of the raw glass. Careful study of these items shows the existence of two trade routes, one from the west, which brings Mediterranean goods, the other from the east, for Chinese products. According to Dussubieux [45], it appears that only Near Eastern cobalt blue glass and gold-foiled glass beads were imported from the Mediterranean regions. These Mediterranean artefacts are mainly found in South India and Sri Lanka. Some of them were probably redistributed from the South Indian regions, towards South East Asia: this is shown by the presence of a Mediterranean gold-foiled glass bead at Khlong Thom in Thailand. In the other group of glass, we also find a majority of cobalt blue glass items. We shall notice here that all the analysed cobalt blue glass items found in India are made of imported glass (soda-lime or potash glass). It seems, in fact, that cobalt as a colourant was not available for Indian glass workers, and that they were obliged to import cobalt blue glass from the Mediterranean, Southeast Asia or China. 15.5
CAROLINGIAN GLASS PRODUCTION: SOME UNUSUAL LEAD GLASS COMPOSITION SMOOTHERS
As stated earlier, the “Near Eastern” glass would remain the main glass type found in Western Europe until the end of the 8th century AD . Depending on the different European regions, the importation of this glass into Europe seems to break off progressively from the 9th century to the 11th or 13th century [49,50]. In Western Europe, the Carolingian period is one of transition for glass production: Near Eastern glass is partly replaced by new glass types, containing potash or lead. For example, potash –lime glass, also called wood-ash glass, appears in France at the beginning of the 9th century.
687
B. Gratuze and K. Janssens
Smoothers, as they are currently named, are among the glass objects produced during this period [51,52]. Their chemical characterization, together with windowpanes and vessels of this period, was studied to allow a better understanding of these innovations. Recent chemical characterization of smoothers, undertaken in England, Germany and France, using different analytical techniques, shows that two main glass types were used for their fabrication. More than 80% of the smoothers are wood-ash glass, while the remainder are a special silica – lime –lead glass [53]. This unusual composition seems to be specific to the smoothers: according to the data published by Wedepohl and Bezberodov [54,55], no other analysed lead glass objects of this period are similar. Within this period, lead glasses tend to characterize Eastern European glass productions. These glasses are divided into two main groups: – high lead silica glasses: in these glasses, lead concentration is superior to 50% and could reach 80%, the remaining constituent being mainly silica, – mixed lead– alkaline glasses: lead is used with an alkaline flux, such as potash or soda, but more rarely with lime (Fig. 15.10). The presence of a high amount of alumina, iron, barium and antimony is the other main characteristic of this silica –lime – lead glass (Fig. 15.11). Eastern European lead glasses of this period contain considerably less alumina and iron. Lead glass smoothers of this unusual type have been identified in different places (Fig. 15.21) in France (Aniane, Arles, Blois, Bordeaux, Distre´, Melle and Montours), England (York, J. Bayley personal communication) [56], and
Fig. 15.10. Diagram of Na2O þ K2O þ MgO versus CaO for lead smoothers and other medieval lead glasses and alkaline lead glasses.
688
Provenance analysis of glass artefacts
Fig. 15.11. Diagram of Al2O3 versus Fe2O3 for lead smoothers and other medieval lead glasses and alkaline lead glasses.
Germany (Haithabu) [50]. One of them has also been found in Novgorod, Russia [57]. Recently, while working on the Carolingian lead –silver mines of Melle (50 km southwest of Poitiers, France) we have shown that the glassy slag, produced in this site, has a composition which is similar to the one of the lead glass smoothers: i.e., a silica – lime – lead glass with high alumina and iron concentrations (Figs. 15.12 and 15.13). Although the slag appears to be quiet heterogeneous, lead, lime, iron and alumina concentrations may vary tremendously, its average composition being close to that of the lead smoothers. Lead isotope analyses (Fig. 15.14) were carried out on the smoothers found in Germany and France showing that their lead isotope signatures are identical to that of galena and slag from Melle [53] (Wedepohl, personal communication). For the French smoothers, strontium isotope analyses confirm these first results [58]. The chemical and isotopic relationships observed indicate that Melle slags were used to produce the lead glass smoothers. From an archaeological point of view, it also explains why the total amount of slag recovered on-site is lower than expected, according to the supposed lead production of the mine [59]. Future lead and strontium isotopic analyses will show whether this technical innovation (reprocessing of slag to produce glass objects) was specific for the Melle lead – silver mine, or was also used in other Carolingian lead production centres. The lead glass smoothers’ example illustrates the importance of isotopic analysis and explains provenance studies. However, this method is more an
689
B. Gratuze and K. Janssens
Fig. 15.12. Ternary diagram of PbO – Na2O þ K2O þ MgO – CaO for the lead smoothers, the Melle slag and other medieval lead glasses.
exclusive method than an inclusive one. Different mines can have the same isotopic signature. Thus, isotopic analysis allows to reject a hypothesis when different isotopic ratios are found; however, if the same ratios are obtained, it can be fairly difficult to prove the relationship between the mine and an object. Nevertheless, in this case, both chemical and isotopic correspondences are found. The composition of the slag depends on the rocks surrounding the galena. Since the probability to find, at two different places, the same geological and geochemical parameters (for both lead and strontium) is low, it is possible to conclude that all the Carolingian silica – lime – lead glass smoothers probably originate from Melle. The use or function of the Carolingian smoothers is still subject to debate (were these enigmatic objects apotropaic or tools?). The fact that this lead glass composition has not been found among other analysed Carolingian objects suggests that the smoothers were not glass ingots. Due to their characteristic chemical and geochemical signature, the distribution of these
690
Provenance analysis of glass artefacts
Fig. 15.13. Ternary diagram of PbO–Al2O3 –CaO for the lead smoothers, the Melle slag and other medieval lead glasses.
objects, traded by the Vikings together with other (wood-ash) glass smoothers, is easy to trace. The study of these particular smoothers improves our knowledge of Carolingian glass trade through Europe and highlights some technical innovations of that period. Perhaps it may also induce a better understanding of the function of the Carolingian smoothers. 15.6
LATE MIDDLE AGE RECYCLED GLASS
If the reprocessing of metallurgical slag is unusual in the glass industry, the recycling of glass itself is much more common. During Antiquity, Western Mediterranean glass workshops were commonly melting raw glass together with leftover glass fragments. The cargo found on some Mediterranean shipwrecks of this period attests to the trade of raw glass and recycled broken glass objects from the Near East towards Europe [42]. Some Roman authors have also described these recycling methods along with the use of imported
691
B. Gratuze and K. Janssens
Fig. 15.14. Lead isotopes of some Carolingian lead glass smoothers (Arles, Blois, Bordeaux and Haithabu), Melle galena and slag, and some Carolingian coins minted at Melle.
raw glass [60,61]. Later, during the medieval period, Theophilus has described how particular glasses, such as mosaic tesserae, were collected and melted with colourless glass to produce coloured windowpanes [62]. Recycled or broken glass is frequently cited in ancient texts, coming just after soda, but before frit and (de)colouring agents [49]. According to Foy, its trade was sometimes well organized: e.g., at Marseille in 1331, it was forbidden to export broken glass; instead this material was to be sold to a local glass workshop at a price fixed by the town authorities [49]. Later, in Venetian recipes, the Italian terms rotti or rottami referred to the use of recycled glass [63,64]. During Antiquity, where only natron soda-lime glass (also called Near Eastern glass), was produced, one of the most perverse effects of glass recycling (at least for provenance studies) is the standardization of the glass composition throughout the entire Roman Empire. In Europe, during the late medieval period, the situation is different. In France, e.g., in different regions, glass types of various compositions are produced [65]: – in the Mediterranean regions predominantly plant ash soda-lime glass (made with ashes from halophytic plants such as salicornia) was used, – in the centre of the country forest plant ashes were used as fluxing agent, yielding potash– lime glasses having various potash/lime ratios, depending on the burnt plants or trees,
692
Provenance analysis of glass artefacts
– in the east of the country, in a region named Argonne, high lime glasses, probably using oak ashes, were produced. In such a situation, analytical studies carried out on glass consumption sites allow to determine the origin of glass objects according to their composition. However, they do not provide any quantitative information on the importance of glass recycling. To obtain information on this aspect of the glass trade, the productions of glass workshops must be studied, such as that of the medieval French glass workshop called “La Verrie`re” (Saint-Che´ly d’Aubrac, in Aveyron, southwest of France). This 14th century workshop was located at the frontier between two different glass type areas [66]: the Mediterranean zone where halophytic plant ash soda-lime glass was made and the centre zone where forest plant ash potash – lime glass was produced. The recorded objects found at this site were mainly vessels’ fragments. The typologies of these objects are similar to those manufactured in the Languedoc area, a region producing plant ash soda-lime glass. However, compared to the Languedoc glass, the glass from La Verrie`re is different: it is more greenish and thicker. It was interesting to determine which type of recipe these glass workers used; this archaeometric investigation was undertaken in order to link this workshop to one of the glass recipes. Fourteen glass objects, recovered on this site, were analysed using LA-ICP-MS. Among them, eight were monochrome, greenish or bluish, while the remaining six were decorated with thin streaks and pastilles of cobalt blue glass. Although all the studied glass items were made with plant ash soda-lime glass, some objects presented unusually high concentrations of lime and potash together with lower soda values. Figure 15.15 shows the inverse
Fig. 15.15. Binary diagram of Na2O–MgO, Na2O–K2O and Na2O–CaO, showing the inverse relationships which exist between these oxides.
693
B. Gratuze and K. Janssens
Fig. 15.16. Binary diagram of Na2O– Cl and Na2O–P2O5, showing the different relationships which exist between these compounds.
relationships observed between those elements. We can also notice that potash and lime are correlated with magnesium, rubidium, strontium, phosphorus, and to some other trace elements (such as barium, cerium and uranium), while sodium is correlated with chlorine, lithium and zirconium (Figs. 15.16– 15.18). Two different hypotheses could explain the observed relationships: – both halophytic and forest plant ash were used by the glassworkers, soda plant ash remaining the main component, – recycled glasses from the surrounding areas, with both potash and soda compositions, were used for making glass.
Fig. 15.17. Binary diagram of Na2O – Rb2O and Na2O –SrO, showing the inverse relationships which exist between these oxides.
694
Provenance analysis of glass artefacts
Fig. 15.18. Binary diagram of Na2O –ZrO2 and Na2O – Li2O, showing the relationships which exist between these oxides.
From an analytical point of view, both hypotheses are equally valid, and it is possible to calculate the maximum amount of forest plant ash (perhaps recovered from the furnace) or recycled glass that was used by the glass workers. Figure 15.19 allows to identify the potash glasses, which may have been recycled, or by extrapolating, the types of the forest plant ashes that were used. The average composition of these compounds could then be calculated. Using simple calculation ratios, the maximum amount of recycled potash glass (or forest plant ash added to the glass batch) used by the glass workers from La Verrie`re (Table 15.6) was found to vary between 35 and 40%. These results illustrate the difficulties that sometimes arise when attempting to interpret archaeometric data. In this example, one can either draw conclusions on the importance of glass recycling practices or on the mixing of different types of plant ash. Neither major element nor trace element data can help to solve that problem. Both soda-ash and potash glass compositions could explain the different observed relationships. Oak and fern ash analysis, carried out by FNAA, shows that plant ashes contain many trace elements that are also present in sand. Their concentration levels are sometimes higher in plant ashes than in sands (Table 15.7). In addition, when crushed quartz is used, as it appeared to be the case at La Verrie`re, the trace elements encountered in the glass probably almost entirely originated from the ashes. These results also show the large composition variability of the glass produced by a single workshop, and the difficulties encountered to carry out
695
B. Gratuze and K. Janssens
Fig. 15.19. Ternary diagram of Na2O–MgO þ K2O –CaO showing that the plant ash soda-lime glass produced at Saint-Che´ly d’Aubrac is obtained (a) by mixing two types of glass of different compositions (type 2 and 3) or (b) by mixing soda littoral plant ash together with potash forest plant ash.
detailed studies on the distribution of glass objects from the production workshops towards the consumption sites. If the trade over short or medium distances of finished objects is sometimes difficult to characterize, long-distance trade of raw materials used in the glass industry is easier to document. The results obtained on cobalt blue glass, used to decorate the vessels, show that the cobalt of the blue colouring agent used at Saint-Che´ly d’Aubrac is mainly associated with zinc, lead and indium (Fig. 15.20). This chemical association is found in all the European medieval cobalt blue glasses and glazes, dated from the end of 12th century to the end of the 15th century [67,41]. It characterizes the colouring agent produced from the cobalt ore of the Freiberg area, Germany (Table 15.4).
696
TABLE 15.6 Estimated calculation of the maximum amount of recycled glasses used by the medieval glass workers at Saint-Che´ly d’Aubrac Oxides contents (wt%)
15.0 2.12 2.60 63.8 1.11 0.90 4.84 7.39 0.77 1.19
Average composition of the lime potash recycled glasses, which may have been used by the glass workers
1.88 6.51 1.22 56.6 3.08 0.20 14.3 13.2 0.98 0.57
Average composition for the mixed glasses, which were produced by the glass workers
10.6 3.87 1.82 60.4 2.34 0.74 8.78 9.3 0.93 1.05
Estimated composition of the mixed glass obtained by using different proportions of recycled glasses 30%
35%
40%
11.1 3.4 2.2 61.6 1.7 0.69 7.7 9.1 0.83 1.01
10.4 3.7 2.1 61.3 1.8 0.66 8.2 9.4 0.84 0.98
9.7 3.9 2.0 60.9 1.9 0.62 8.6 9.7 0.85 0.94
Provenance analysis of glass artefacts
Na2O MgO Al2O3 SiO2 P2O5 Cl K2O CaO MnO Fe2O3
Average composition of unaltered plant ash soda-lime glass produced directly from sand and soda plant ash
697
698 Major, minor and trace element levels measured in oak and fern ash compared to those found in two different French sand samples Sample name
SiO2
Al2O3
K2O
CaO
MgO
Fe2O3
MnO2
TiO2
Zr
Ce
Ba
Zn
Rb
Cs
Oak ash Fern ash Etampes sand Dourdan sand
31.4 43.4 98.1 99.2
3. 4 0.2 1.3 0.36
7.8 39.9 0.5 0.15
49.4 6.5 nd nd
2.3 3.3 nd nd
1.2 0.05 0.04 0.13
3.0 0.5 nd nd
0.27 0.01 0.07 0.08
223 nd 54 92
36 nd 2 4
5800 600 186 13
823 291 nd nd
268 1550 17 nd
7 19 nd nd
Concentrations are expressed as wt% for oxides and as ppm for elements; nd, not detected.
B. Gratuze and K. Janssens
TABLE 15.7
Provenance analysis of glass artefacts
Fig. 15.20. Binary diagram of CoO–PbO, CoO–ZnO and CoO –In*20, showing the relationships which exist between these oxides and elements.
15.7
GLASS TECHNOLOGY TRANSFER DURING THE 16TH – 17TH CENTURY TO AND FROM ANTWERP
In the early 16th century, the port of Antwerp (Belgium) was the most important one in Europe and an intense exchange of people and goods with many parts of the world took place. From 1541, an Italian dynasty of glass makers, originating from the North of Italy and in particular from Venice, started activities in Antwerp, importing and reproducing Venetian glass products. Deraedt et al. [69 – 71] investigated the composition of colourless “Fac¸on-de-Venise” glass vessels recovered from various cities in the Low Countries (such as Antwerp, Maastricht, Breda, Amsterdam) with the aim of determining the importance of the local glass production, relative to the import of vessels from Venice. Some of the vessels’ shapes are shown in Fig. 15.21.
Fig. 15.21. Photographs of some of the Fac¸on-de-Venise vessels (belonging to the VVB group) excavated in Antwerp.
699
B. Gratuze and K. Janssens
In Venice, two types of high quality transparent glass, called “Vitrum Blanchum” and “Cristallo” were produced. As can be seen in Fig. 15.22a, in Antwerp, four distinct compositional groups of soda-lime glass are encountered in the 16th –17th century, next to potash and wood-ash-based
Fig. 15.22. (a) CaO versus K2O content (in %w/w) of Fac¸on-de-Venise glass vessels excavated in Antwerp (Belgium); (b) a Hf versus Zr plot (in mg/g) allows to distinguish between Venetian and Antwerp production, resulting from differences in the silica source.
700
Provenance analysis of glass artefacts
glass (not shown in the figure) [69]. Two of the four groups (marked MA and FDV) have compositions richer in CaO than the Venetian glass types and can be considered to be local production [70]. One group (marked AVB) has a composition corresponding to that of Venetian Vitrum Blanchum and also shows the same trace element fingerprint. The group marked AC is similar in major composition to the Venetian Cristallo (VC) but on average is richer in K2O and shows a trace element fingerprint that more resembles that of the other Antwerp groups than that of the Venetian ones, as illustrated in Fig. 15.22b [71]. At the end of the 16th century, the Low Countries were split into an independent Northern part, with Amsterdam as its capital, while Antwerp remained part of the Habsburg empire and was cut off from trade. At this time, many artists and artisans, among whom glass makers, fled Antwerp to neighbouring urban centres such as London and Amsterdam. When the composition of the flux used for making the 16th – 17th century glass vessels, excavated in Antwerp, London and Amsterdam, is compared as in Fig. 15.23 (i.e., vessels from before and after the above-mentioned political changes), it is possible to distinguish two series of compositional groups, having variable Na2O and K2O contents but in which the sum of these two flux constituents is more or less constant. Furthermore, it is striking to note that the glass types introduced in 16th century Antwerp are found back in London and Amsterdam during the 17th century while in Antwerp, other compositions
Fig. 15.23. Flux composition plot of 16th –17th century glasses excavated in Antwerp and various neighbouring cities.
701
B. Gratuze and K. Janssens
were being used in that period, probably the result of restricted availability and/or increased prices of raw materials for glass making such as soda-rich ashes [72].
15.8
TRADE BEADS: THE GLASS TRADE INTERNATIONALIZATION, DURING THE POST-MEDIEVAL PERIOD
During the Post-Medieval period, trade beads were mainly exported to Africa and America by the Dutch, French Spanish, Portuguese and English. Recent studies published by Hancock and co-workers [68,73] have documented the importance of trade beads in North America: compositions of beads are used as chronological markers and sometimes allow for the identification of (Dutch, English or French) traders. Later, during the 18th and the 19th centuries, trade beads reached the Pacific region, as shown by the analytical results obtained on some glass beads found at Cikobia-i-Ra (a small island at the northeastern end of the Fiji Islands). The beads come from two different tombs of a burial place [74,75]. A cobalt blue cornerless hexagonal bead was found in the first one, while many small annular beads of different colours (white, opaque turquoise, cobalt blue, red and red with a white core) were found in the other. All the small annular beads were heavily corroded; also some powdery white beads were found. From a historical point of view, either European sailors during the 18th century or American whaleboat sailors during the 19th century may have imported these beads in the region [76,74]. Thus, a chemical study of the glass was undertaken to obtain information on the provenance of these beads and to find arguments favouring one of these two possibilities (Table 15.8). The blue hexagonal bead is often referenced as a Russian or cut bead, while the small annular red beads with a white core are named white heart or cornaline d’Aleppo [77]. According to Francis, the hexagonal beads were made in Czechia from 1820 to 1890, while Venetian, Czech or French artisans made the white heart beads from ca. 1830 onwards. These beads were largely traded by the Hudson Bay Company and other North American trade companies and usually are associated with Yankee and Canadian trading activities. Analytical results (Table 15.8) show that the cobalt blue hexagonal bead is a high silica – potash – lime glass. Its major element composition is similar to the one of some Bohemian glasses from the 19th century. These results are in good agreement with P. Francis’ model. High nickel values and low arsenic
702
TABLE 15.8 Some trace and major element contents measured for the studied Fijian beads Small annular turquoise blue beads
Small annular cobalt blue beads
Small annular red beads
White heart bead (red glass on the left, white glass on the right)
Small annular white beads
Cornerless hexagonal blue beads
MnO CoO CuO Sb2O3 As2O3 PbO Ni Ag Au Bi U
0.032 0.020 2.8 2.97 0.52 11.1 74 13.3 0.9 23 0.4
0.73 0.14 0.025 0.086 1.53 15.7 362 4.9 1.7 207 2.8
0.014 0.0001 0.0124 0.32 3.47 16.5 3.6 4.3 242 4.4 0.6
0.016 0.0000 0.0109 0.61 2.28 11.5 5.9 12.5 227 7.3 0.3
0.02 0.0002 0.0287 0.058 6.28 37.8 6.3 9.5 0.6 25 0.6
0.038 0.0313 0.0050 0.000 0.049 0.0081 284 0.2
0.078 0.0004 1.2 1.93 0.40 13.7 12 7.3 0.6 7.8 1.3
0.77 0.15 0.026 0.088 0.56 13.0 403 5.2 1.9 214 2.8
0.12 0.0003 0.0200 0.019 5.03 35.6 6.0 7.7 24 24 0.6
0.016 0.0004 0.0186 0.074 5.80 36.4 5.0 10.5 1.0 9.5 0.4
0.7 0.6
0.051 0.0337 0.0050 0.000 0.049 0.0063 283 0.2 0.3 0.9 0.6
The contents of oxides are in wt% and elements in ppm. The small annular cobalt blue beads contain higher levels of bismuth, uranium and arsenic than the hexagonal bead. The relative contents of nickel with regard to cobalt are lower for the annular beads than for the hexagonal one. Therefore, the annular beads could be attached to the group 8 of Table 15.4, while the hexagonal bead defines the group 9.
Provenance analysis of glass artefacts
Oxides/ elements
703
B. Gratuze and K. Janssens
levels characterize the cobalt pigment while bismuth and uranium are absent. According to the compositional cobalt pigment model that was developed up to the beginning of the 18th century [41,67], this bead defines a new group, probably pointing to the use of at least one new cobalt source from the middle/end of the 18th century or the beginning of the 19th century (Table 15.4). Different compositional types were found among the small annular beads (Table 15.8). The white glass from both the white heart beads and the white annular bead is a lead arsenate soda silica glass. All the other glasses found at Cikobia-i-Ra are mixed alkali – lime – lead –silica glasses. They contain various amounts of arsenic and antimony that probably act as opacifiers. Despite their heterogeneity, this group can be associated with the Venetian glass production. The white colourant and opacifier is lead arsenate. Venetian glassmakers used this chemical compound from the end of the 17th century; Murano “girasol” glass appears in Darduin recipes’ book in 1693 [78]. The red glass is gold ruby glass; Darduin also mentioned this recipe to be dated to the end of the 16th century. The blue colourants are copper and cobalt. Antimony is used as an opacifier in the copper opaque turquoise blue beads. In the dark blue beads, cobalt is associated with nickel, arsenic bismuth and uranium. According to our cobalt model, these beads may have been produced between the middle of the 16th century and the beginning of the 18th century (Table 15.4). According to the model developed by P. Francis, all these European beads are dated from the 19th century and were traded by American whaling companies. The analytical results obtained on the hexagonal bead agree with this model. However, the results obtained on the annular beads, which come from another tomb, could be interpreted differently. According to Venetian recipes, these glasses could be dated as far as the end of the 17th century. If we now consider the cobalt chemical associations, one can conclude that the dark blue annular beads are more ancient than the hexagonal beads, and are probably older than the beginning of the 19th century. The heavy corrosion of the annular beads, compared to the hexagonal one, may also be interpreted in the same way. Thus, a probability exists that these cobalt blue annular beads date back to the first contacts between Europe and Fiji during the 18th century.
704
Provenance analysis of glass artefacts
15.9
CONCLUSIONS
Chemical characterization of archaeological glass artefacts can be used for studying ancient technical processes and recipes, raw materials’ provenance or manufactured artefacts’ trade and diffusion. The preceding examples mainly focus on trade and provenance studies. They were chosen from various geographical and archaeological contexts spread in time from the end of the Paleolithic era to the 19th century, as shown in Fig. 15.24. They help in illustrating the possibilities and the limits offered by chemical analyses to gain information on the trade in natural and synthetic glasses and in provenance research work. In a few cases, information on the major element composition is sufficient to answer (in full or part) the archaeological questions. However, in the great
Fig. 15.24. Localization of the different European and Mediterranean archaeological sites mentioned in the text. Bronze Age sites are reported in bold type, Roman period archaeological sites are underlined. Other sites date from the early and late medieval periods.
705
B. Gratuze and K. Janssens
majority of cases, minor and trace elements’ determinations are absolutely necessary to distinguish between different hypotheses. Such is obviously the case for obsidian sourcing, where only the hygromagmaphile elements are useful for assigning artefacts to their geological source. Sometimes, additional methods, such as isotopic analysis, are needed for highlighting the difference between groups with closely related compositions or for confirmation of provenance, as was illustrated by means of the Carolingian lead glass smoothers. Chemical results could also lead to different hypotheses about production technologies between which it is not possible to choose from a pure analytical point of view. This is particularly the case for the glass production study of the 14th century glass workshop La Verrie`re. Glass recycling, which is one of the possible hypotheses here, is often a limiting factor for trade studies. In favourable situations, e.g., when a new technology or glass type is (re)introduced in specific geographical areas, the glass composition can help to improve the insight in trade and technology transfer patterns, as illustrated with the Fac¸on-de-Venise glass in the Low Countries. For synthetic glass, we have shown that trace element concentrations are useful for characterizing both the major components (sand and fluxing agents) and the additives, such as opacifiers and colourants. The chemical composition of these additives reflects their geological provenance and their manufacturing processes. Thus, when a chronological evolution of these parameters is observed, trace element concentrations can be used to build chrono-compositional models. This is the case for cobalt in the previous examples. Therefore, when possible, the most complete characterization of glass productions originating from identified workshops should be undertaken in order to build a reliable database. Glass, either as raw material or as manufactured item, has been traded far away from its production centres: thus, short-scale diffusion studies are likely to fail the provenance studies that are considered on an international or intercontinental scale. Finally, we should not forget that the aim of all these studies is to highlight parts of our economical and technical history, so that it is mandatory to undertake such efforts in close collaboration with archaeologists and historians.
Acknowledgements Many museums and archaeologists have provided material for analysis, and have allowed the use of unpublished analytical results. B.G. would like to thank more particularly: Marie-Genevie`ve Colin, Conservateur en Chef du
706
Provenance analysis of glass artefacts
Patrimoine, SRA Midi-Pyre´ne´es, for the material from la Verrie`re; Fre´de´rique Valentin and Christophe Sand, UMR 7041, Arche´ologies et Sciences de l’Antiquite´, Maison des Sciences de l’Homme, Nanterre, for the material from Cikobia (this study is part of a French/Fijian research programn, on the prehistory of Cikobia and Naqelelevu island, funded by the French Foreign Office); Yves Billaud, DRASSM, for pre-historic glass studies. B.G. is also grateful to Justine Bayley, English Heritage, and Karl Hans Wedepohl, Geochemical Institut, University Go¨ttingen, for their data on Carolingian glass. K.J. would like to thank I. Deraedt, J. Veeckman and B. Velde for their efforts during the study of the Fac¸on-de-Venise glass.
REFERENCES 1
2
3 4
5
6
7 8
P. Acquafreda, S. Lorenzoni and E. Zanetti, Discrimination of Melos, Lipari, ¨ zer and G.D. Palmarola obsidian using SEM-EDS. In: S. Demirci, A.M. O Summers (Eds.), The Proceedings of the 29th International Symposium on Archaeometry, Ankara, Tu¨bitak, Ankara, 1996, pp. 87 –97. J. Keller, R. Djerbashian, S.G. Karapetian, E. Pernicka and V. Nasedkin, Armenian and Caucasian obsidian occurrences and sources for the Neolithic trade: volcanological setting and chemical characteristics. In: S. Demirci, A.M. ¨ zer and G.D. Summers (Eds.), The Proceedings of the 29th International O Symposium on Archaeometry, Ankara, Tu¨bitak, Ankara, 1996, pp. 69 –86. J.W. Michels, Bulk element composition versus trace element composition in the reconstruction of an obsidian source system, J. Archaeol. Sci., 9 (1982) 113. M.-C. Cauvin, Y. Besnus, J. Tripier and R. Montigny, Nouvelles analyses d’obsidiennes du Proche Orient: mode`le de ge´ochimie des magmas utilise´s pour la recherche arche´ologique, Pale´orient, 12(2) (1991) 89–97. B. Gratuze, Obsidian characterisation by laser ablation ICP-MS and its application to prehistoric trade in the Mediterranean and the Near East: sources and distribution of obsidian within the Aegean and Anatolia, J. Archaeol. Sci., 26 (1999) 869–881. E. Pernicka, J. Keller, G. Rapp Jr. and T. Ercan, Provenance of Late Neolithic and Early Bronze Age obsidian artefacts from the Troad. In: S. Demirci, A.M. ¨ zer and G.D. Summers (Eds.), The Proceedings of the 29th International O Symposium on Archaeometry, Ankara, Tu¨bitak, Ankara, 1996, pp. 515 –519. O. William-Thorpes, Obsidian in the Mediterranean and the Near East: a provenancing success story, Archaeometry, 37(2) (1995) 217 –248. C. Chataigner, Sources des artefacts du Proche Orient d’apre`s leur caracte´risation ge´ochimique. In: M.-C. Cauvin, A. Gourgaud, B. Gratuze, N. Arnaud, G. Poupeau, J.-L. Poidevin and C. Chataigner (Eds.), L’obsidienne au Proche et Moyen Orient. Du Volcan a` l’outil, Oxford B.A.R. I.S., Vol. 738. Maison de l’Orient Mediterrane´en, Lyon, 1998, pp. 273–324.
707
B. Gratuze and K. Janssens 9
10 11
12
13
14
15
16
17
18 19 20 21 22
23
708
J. Guilaine, F. Briois, J. Coularou and I. Carre`re, L’e´tablissement ne´olithique de Shillourokambos (Parekklisha, Chypre). Premiers re´sultats, Report of the Departement of Antiquities, Cyprus, 1995, pp. 11–32. F. Briois, B. Gratuze and J. Guilaine, Obsidiennes du site ne´olithique pre´ce´ramique de Shillourokambos (Chypre), Pale´orient, 23(1) (1997) 95. E. Peltenburg, S. Colledge, P. Croft, A. Jackson, C. McCartney and M.-A. Murray, Agro-pastoralist colonization of Cyprus in the 10th millennium BP: initial assessments, Antiquity, 74 (2000) 844. L. Bellot-Gurlet, Caracte´risation par analyse e´le´mentaire (PIXE et ICP-MS/ AES) d’un verre naturel: l’obsidienne. Application a` l’e´tude de provenance d’objets arche´ologiques, The`se, Universite´ Joseph Fourier, Grenoble I, 1998. M.-C. Cauvin and C. Chataigner, Distribution de l’obsidienne dans les sites arche´ologiques. In: M.-C. Cauvin, A. Gourgaud, B. Gratuze, N. Arnaud, G. Poupeau, J.-L. Poidevin and C. Chataigner (Eds.), L’obsidienne au Proche et Moyen Orient. Du Volcan a` l’outil, Oxford B.A.R. I.S., Vol. 738. Maison de l’Orient Mediterrane´en, Lyon, 1998, pp. 325 –350. B. Gratuze, J.-N. Barrandon, K. Al Isa and M.-C. Cauvin, Non destructive analysis of obsidian artefacts using nuclear techniques: Investigation of provenance of Near Eastern Artefacts, Archaeometry, 35(1) (1993) 11–21. N. Balkan-Atli and G. Der Aprahamian, Les nucle´us de Kalatepe et deux ateliers de taille en Cappadoce. In: M.-C. Cauvin, A. Gourgaud, B. Gratuze, N. Arnaud, G. Poupeau, J.-L. Poidevin and C. Chataigner (Eds.), L’obsidienne au Proche et Moyen Orient. Du Volcan a` l’outil, Oxford B.A.R. I.S., Vol. 738. Maison de l’Orient Mediterrane´en, Lyon, 1998, pp. 241–257. E.J. Peltenburg, Early faı¨ence: recent studies, origin and relation with glass. In: M. Bimson and I.C. Freestone (Eds.), Early Vitreous Materials, British Occasional Paper 56. 1987, pp. 5–30. A.J. Shortland and M.S. Tite, The interdependence of glass and vitreous faı¨ence production at Amarna. In: P. McCray and W.D. Kingery (Eds.), Ceramics and Civilization, Vol. 8. American Ceramic Society, Westerville, OH, 1998, pp. 251–265. A. Aspinal, S.E. Warren, J.G. Crummett and R.G. Newton, Neuron activation analysis of faı¨ence beads, Achaeometry, 14(1) (1972) 27 –40. J. Briard, Les perles de faı¨ence du Bronze ancien en Bretagne Me´diterrane´ennes ou Occidentales?, Rev. Arche´ol. Ouest, 1 (1984) 55 –62. J. Frana, A. Mastalka and N. Venlova, Neuron activation analysis of some ancient glasses from Bohemia, Archaeometry, 29(1) (1987) 69 –89. B. Gratuze, C. Louboutin and Y. Billaud, Les perles Protohistoriques en verre du Muse´e des Antiquite´s Nationales, Antiquite´s Nationales, 30 (1998) 13–24. J. Guilaine, B. Gratuze and J.-N. Barrandon, Les perles en verre du Chalcolithique et de l’Age du Bronze. Analyse d’exemplaires trouve´s en France. In: C. Chevillot and A. Coffyn (Eds.), L’Age du Bronze Atlantique, Actes du 1er Colloque du parc Arche´ologique de Beynac, Association des muse´es du Sarladais, Beynac et Cazenac, 1991, pp. 255 –266. J. Henderson, The earliest glass in Britain and Ireland. In: M. Feuge`re (Ed.), Le verre pre´romain en Europe occidentale. Mergoil, Montagnac, 1989, pp. 13 –24.
Provenance analysis of glass artefacts 24
25
26
27 28
29 30
31 32 33
34 35 36
37
38
39 40
J. Henderson, Chemical analysis of the glass and faı¨ence from HauteriveChampre´veyres, Switzerland. In: A.-M. Richner Faraggi (Ed.), HauteriveChampre´veyres 9, Me´tal et Parure au Bronze Final, Arche´ologie Neuchaˆteloise, Vol. 17. Muse´e Cantonal d’Arche´ologie, Neuchaˆtel, 1993, pp. 111–117. B. Gratuze, M. Blet-Lemarquand and J.-N. Barrandon, Mass spectrometry with laser sampling: A new tool to characterize archaeological materials, J. Radioanal. Nucl. Chem., 247(3) (2001) 645 –656. Y. Billaud and B. Gratuze, Les perles en verre et en faı¨ence de la Protohistoire franc¸aise. In: J. Guilaine (Ed.), Mate´riaux, Productions, Circulations, du Ne´olithique a` l’Age du Bronze. Se´minaires du colleg`e de France, Errance, Paris, 2002, pp. 193 –212. C. Lilyquist and R.H. Brill, Study in Early Egyptian Glass. Metropolitan Museum of Art, New York, 1993, 80 pp. G. Hartmann, I. Kappel, K. Grote and B. Arndt, Chemistry and technology of prehistoric glass from Lower Saxony and Hesse, J. Archaeol. Sci., 24 (1997) 547 –559. P. Bellintani, Frattesina: l’ambra e la produzione vitrea nel contesto delle relazioni transalpine, Ori Delle Alpi, 6 (1997) 116 –129. A. Biaviati and M. Verita`, The glass from Frattesina, a glass making center in the late Bronze Age, Rivista Della Stazione Sperimentale del Vetro, 19(4) (1989) 295 –299. R.H. Brill, Chemical analysis of some glass from Frattesina, J. Glass Studies, 34 (1992) 11–22. P. Santopadre and M. Verita`, Analysis of the production technologies of Italian vitreous material of the Bronze Age, J. Glass Studies, 42 (2000) 25–40. C. Jokubonis, P. Wobrauschek, S. Zamini, M. Karwowski, G. Trnka and P. Stadler, Results of quantitative analysis of Celtic glass artefacts by energy dispersive X-ray fluorescence spectrometry, X-ray Spectrom., 58 (2003) 627–633. A. Aerts, K. Janssens and B. Velde, Changes in silica source in Roman and postRoman glass, Spectrochim. Acta B, 58 (2003) 659 –667. J.P. Delgado, Encyclopedia of Underwater and Maritime Archaeology. British Museum Press, London, 1997, 494 pp. K. Demakopoulou, Les e´paves de navires en Me´diterrane´e orientale, L’Europe au Temps d’Ulysse. Dieux et He´ros a` l’Age du Bronze, Re´union des Muse´es Nationaux, 1999, pp. 35 –37. J.P. Le Bihan, In: J.-P. Le Bihan and J.-F. Villard (Eds.), Arche´ologie d’une ıˆle a` la Pointe de l’Europe: Ouessant, tome 1 Le site Arche´ologique de Mez-Notariou et le Village du Premier aˆge du Fer, Revue Arche´ologique de L’Ouest, 2000, 351 pp. B. Raftery and J. Henderson, Some glass beads of the Later Bronze Age in Ireland. In: O.-H. Frey, H. Roth and C. Dobiat (Eds.), Glasperlen der Vorro¨mischen Eisenzeit II. Mayence, 1987, pp. 39–53. R. Azemar, Y. Billaud, G. Costantini and B. Gratuze, Les perles protohistoriques de l’Aveyron, Cahier d’Arche´ologie Aveyronnaise, 14 (2000) 75–88. J. Roussot-Larroque, Parures et objets divers non me´talliques. In: M. Gruet, J. Roussot-Larroque and C. Burnez (Eds.), L’Age du Bronze de la Grotte de
709
B. Gratuze and K. Janssens
41 42
43
44
45
46
47
48 49
50
51 52 53
54
710
Rancogne (Charente), Antiquite´s Nationales, Me´moire, Vol. 3. Re´union des Muse´es Nationaux, Paris, 1997, pp. 113 –120. B. Gratuze, I. Soulier, M. Blet and L. Vallauri, De l’origine du cobalt: du verre a` la ce´ramique, Revue d’Arche´ome´trie, 20 (1996) 77 –94. D. Foy and M.P. Jezegou, Une e´pave charge´e de lingots et de vaisselle de verre, un te´moin exceptionnel du commerce et de la technologie du verre en Me´diterrane´e antique, Verre, 3(3) (1997) 65 –70. C. Moretti and B. Gratuze, Vetri rossi al rame e avventurina. Confronto di analisi e ricette, Rivista della Stazione Sperimentale del Vetro, 29(3) (1999) 147 –160. H. Amrein, A. Burkhardt and W.B. Stern, Analysen von Gla¨sern aus der fru¨hkaiserzeitlichen Glaswerkstatt von Aventicum (Schweiz), Bulletin de l’Association Pro Aventico, 37 (1995) 189 –201. L. Dussubieux, L’apport de l’ablation laser couple´e a` l’ICP-MS a` la caracte´risation des verres: application a` l’e´tude du verre arche´ologique de l’Oce´an Indien, The`se, Universite´ d’Orle´ans, 2001. W.H. Schoff (Tr. and Ed.), The Periplus of the Erythraean Sea: Travel and Trade in the Indian Ocean by a Merchant of the First Century, London, Bombay and Calcutta, 1912, http://www.und.ac.za/und/classics/india/periplus.htm Y. Porter, Origine du cobalt utilise´ en ce´ramique a` l’e´poque me´die´vale. Etude pre´liminaire, La ce´ramique me´die´vale en Me´diterrane´e, Actes du 6e congre`s de l’AIECM2, Aix-en-Provence, 13 – 18 Nov. 1995, Narron Edition, 1997, pp. 505–512. C. Yaocheng, G. Yanyi and C. Hong, Source of cobalt pigment used in Yuan blue and white porcelain wares, Oriental Art, 40(1) (1995) 14 –19. D. Foy, Technologie, ge´ographie, e´conomie. Les Ateliers de verriers primaires et secondaires en Occident, esquisse d’une e´volution de l’Antiquite´ au Moyen-Age. In: M.D. Nenna (Ed.), La Route du Verre Ateliers Primaires et Secondaires du Second Mille´naire av. J.-C. au Moyen-Age, Travaux de la Maison de l’Orient Me´diterrane´en, Vol. 33. 2000, pp. 147 –170. K.H. Wedepohl, The composition of glass from the Carolingian and PostCarolingian period in Central Europe, Il Colore nel Medioevo Art Simbolo Tecnica la Vetrata in Occidente dal IV all’XI Secolo, Atti delle Giornate di Studi Lucca Settembre 1999, Lucca, 2001, pp. 257– 270. A. Ferdie`re, Le travail du textile en Re´gion Centre de l’Age du Fer au Haut Moyen-Age, Revue Arche´ologique du Centre de la France, 23(2) (1984) 209 –243. C. Macquet, Les lissoirs de verre, approche technologique et bibliographique, Arche´ologie Me´die´vale, 20 (1990) 319–334. B. Gratuze, D. Foy, F. Tereygeol and J. Lancelot, Les “lissoirs” carolingiens en verre au plomb: mise en e´vidence de la valorisation des scories issues du traitement des gale`nes argentife`res de Melle (Deux Se`vres). In: D. Foy and M.D. Nenna (Eds.), Echanges et Commerce du verre dans le monde antique, Actes du Colloque de C’AFAV, Aix-en-Provence, Marseille, 7–9 June 2001. Monographie Instrumentum 24, Monique Mergoil ed, Montagnac, 2003, pp. 101 –107. M.A. Bezborodov, Chimie und Technologie der Antiken und Mittelalterlichen Gla¨ser. Mainz, Von Zabern, 1975, 327 pp.
Provenance analysis of glass artefacts 55 56 57 58 59
60 61 62
63
64
65 66
67
68
69
70
71
K.H. Wedepohl, I. Krueger and G. Hartmann, Medieval lead glass from Northwestern Europe, J. Glass Studies, 37 (1999) 65– 82. C. Mortimer, Glass Linen Smoothers from 16-22 Coppergate, York, Ancient Monuments Laboratory Reports 22/95, 1995, 3 pp. J.L. Scapova, Un lissoir de Novgorod, re´flexions sur la verrerie me´die´vale, Acta Arche´ologica, 62 (1992) 231– 243. Lancelot et al., to be published. F. Te´reygeol, B. Gratuze, D. Foy and J. Lancelot. Les scories de plomb argentife`re: une source d’innovation technique carolingienne?, dans les actes du colloque, Artisans, Industrie, Nouvelles Re´volutions, tenu au CNAM en Juin 2000, to be published. Martialus, Epigrammes, Book I, XLI. Stacius, Silvae, I, VI, 73–74. Theophilus, The´ophile preˆtre et moine Essai sur Divers Arts, In: C. L’Escalopier, J. Laget and P. Daviaud (Tr and Eds.), Librairie des Arts et Me´tiers, Nogent-leRoi, 1977. C. Moretti and S. Moretti, Le materie prime dei vetrai veneziani Natura, lessico e fonti di approvvigionamento rilivate dai ricettari dal XIV al XIX secolo, Rivista della Stazione Sperimentale del Vetro, 29(1) (1999) 31 –42. C. Moretti, Le materie prime dei vetrai veneziani rilivate dai ricettari dal XIV alla prima meta` del XX secolo, II parte: elenco materi prime, meterie sussidiarie e semilavorati, Rivista della Stazione Sperimentale del Vetro, 31(3) (2001) 17 –32. J. Barrera and B. Velde, A study of French medieval glass composition, Arche´ologie Me´die´vale, 19 (1989) 81–130. M.-G. Colin, L’atelier de production de verres creux de la Verrie`re de SaintChe´ly d’Aubrac. In: L. Fau (Ed.), Les Monts d’Aubrac au Moyen Age, Gene`se d’un monde agro-pastoral, DAF, pp. 185 –191, to be published. B. Gratuze, I. Soulier, J.-N. Barrandon and D. Foy, The origin of cobalt blue pigments in French glass from the 13th to the 18th centuries. In: D.R. Hook and D.R.M. Gaimster (Eds.), Trade and Discovery: The Scientific Study of Artefacts from Post-Medieval Europe and Beyond, British Museum Occasional Paper 109, B.M. Press, 1995, pp. 123 –133. R.G.V. Hancock, J. McKechnie, S. Aufreiter, K. Karklins, M. Kapches, M. Sempowski, J.-F. Moreau and I. Kenyon, The non-destructive analysis of European cobalt blue glass trade beads, J. Radioanal. Nucl. Chem., 244(3) (2000) 567 –573. I. Deraedt, K. Janssens and J. Veeckman, Compositional distinctions between 16th century “Fac¸on-de-Venise” and Venetian Glass Vessels, excavated in Antwerp, Belgium, J. Anal. At. Spectrom., 14 (1999) 493 –498. I. De Raedt, K. Janssens, J. Veeckman and F. Adams, Composition of facon-devenise and Venetian glass from Antwerp and the Southern Netherlands, Annales du 14e Congre`s de l’Association Internationale pour l’histoire du Verre, Lochem, AIHV, 2000, pp. 346–350. I. De Raedt, K. Janssens, J. Veeckman, L. Vincze, B. Vekemans and T. Jeffries, Trace analysis for distinguishing between Venetian and fac¸on-de-Venise glass
711
B. Gratuze and K. Janssens
72
73
74
75
76 77
78 79
80
712
vessels of the 16th and 17th century, J. Anal. At. Spectrom., 16(9) (2001) 1012 – 1017. I. De Raedt, Composition of 16 –17th Century Facon-de-Venise glass excavated in Antwerp and neighbouring cities, PhD thesis, University of Antwerp, Antwerp, Belgium, 2002. M. Sempowski, A.W. Nohe, R.G.V. Hancock, J.-F. Moreau, F. Kwok, S. Aufreiter, K. Karklins, J. Baart, C. Garrad and I. Kenyon, Chemical analysis of 17th century red glass trade beads from Northeastern North America and Amsterdam, Archaeometry, 43(4) (2001) 503–515. F. Valentin, C. Sand, I. Le Goff, T. Sorovi Vunidilo, S. Matararaba, A. Ouetcho, J. Bole, D. Baret and J. Naucabalavu, Burial practices at the end of the prehistoric period in Cikobia-i-ra (Macuata, Fiji). In: G.R. Clark, A.J. Anderson and T. Vunidilo (Eds.), The Archaeology of Lapita Dispersal in Oceania: Papers from the Fourth Lapita Conference, June 2000, Canberra, Australia, Terra Australis 17, Published for CAR and ANH by Pandanus Books, 2001. F. Valentin, C. Sand and I. Le Goff, Cikobia et Naqelelevu: deux ˆıles fidjiennes a` la charnie`re entre cultures me´lane´sienne et polyne´sienne. Une collaboration arche´ologique franco/fidjienne pour reconstruire les relations inter-insulaires pre´historiques dans le Pacifique Sud. Etude du fort de Rukunikoro et fouille d’un ensemble de se´pultures dans le village fortifie´ de Korotuku dans l’ıˆle de Cikobia (Macuata, Fidji), Rapport no. 4, 1999, 123 pp. et annexes. R.A. Derrick, A History of Fiji. Government Press, Suva, Fiji, 1950. P. Francis Jr., Beads at the Crossroads of Continents. In: W.W. Fitzgerald and V. Chaussonnet (Eds.), Anthropology of the North Pacific Rim. Smithsonian Institution, Washington, 1994, pp. 281 –305. L. Zecchin, Il Ricettario Darduin, un Codice Vetrario del Seicento Trascritto e commentato. Stazione Sperimentale del Vetro, Venezia, 1986, 265 pp. E.V. Sayre, Some ancient glass specimens with composition of particular archaeological significance. Brookhaven National Laboratory Report, New-York, B.N.L. 879 (T-534). J. Henderson, Electron probe analysis of mixed-alkali glasses, Archaeometry, 30(1) (1988) 79 –91.
Chapter 16
Corrosion of historic glass and enamels Manfred Schreiner
16.1
INTRODUCTION
Objects of art and archaeology can be considered as relicts of the past; however, they are rarely found in their original state. Their condition may have been altered depending on the nature of the material used and on the environment the artefacts were subjected to during their lifetime. Objects made of glass or enamels are no exception. By those artefacts, we usually have in mind a hard amorphous (non-crystalline) silicate material with high chemical stability, mainly transparent or translucent, which has been made by heating a mixture of raw materials such as sand, soda or potash and lime at a high temperature to form a liquid. When this heated material is taken out of the furnace, it stiffens as it cools until it resembles the glass we see in windows or hollow glass objects or the enamelled layer we admire as decoration on metallic artefacts made of gold, silver or copper [1– 4]. The process of corrosion (deterioration) sets in as soon as the glass or enamel is made. When exposed to the ambient atmosphere or soil a so-called leached layer is formed immediately on the surface of the silicate materials. In the case of archaeological glass from the ancient periods this surface layer can have a thickness of several tens up to several hundreds of micrometers and yields an iridescent effect due to the different optical properties of this surface layer and the bulk glass. Generally, this phenomenon is regarded as aesthetically pleasing and is not removed during conservation or preservation of such artefacts (Plate 16.I, Fig. 16.1). It is mentioned in annals from the period of the Roman emperor Hadrianus (117 – 138 AD ) that the emperor himself enjoyed such iridescent glass objects called “allassontes versicollores”. But only in the 19th century procedures could be developed and patented to produce the splendid rainbow coloured gleam of glass surfaces. The famous iridescent Art Nouveau glass artefacts of Luis Comfort Tiffany in New York or Johann Loetz in Vienna are typical examples therefore. Comprehensive Analytical Chemistry XLII Janssens and Van Grieken (Eds.) q 2004 Elsevier B.V. All rights reserved
713
M. Schreiner
Plate 16.I. Fragment of an archaeological glass with an iridescent surface layer formed during the corrosion of the glass in soil. ( For a colored version of this figure, see Plate 16.I.)
In principle, the corrosion mechanisms leading to a leached surface layer are well known, although the atmospheric attack as well as the corrosion of archaeological objects in moist soil are very complex due to the numerous factors involved [5– 9]. These factors depend on the characteristics of the
Fig. 16.1. Surface of the archaeological glass with an iridescent surface layer in Plate 16.I, seen with SEM. The leached layer has split off the glass due to the vacuum in the sample chamber of the SEM.
714
Corrosion of historic glass and enamels
silicate materials (e.g., composition, colour, thermal history, surface roughness) as well as on the external conditions such as microclimate, exposure and pollutants [10 – 15]. In a simplified presentation, the phenomenon of leaching can be explained similar to the corrosion of silicate glass in aqueous solutions [5,10], where in neutral and acidic solutions the so-called network modifier ions of the glass such as sodium, potassium or calcium are replaced by protons or hydrogen-bearing species from the aqueous solutions according to Eq. (16.1) – Si – O2 Mþ þ H3 Oþ X – Si – OH·H2 O þ Mþ
ð16:1Þ
This ion exchange mechanism is leading to the formation of the leached surface layer, where the silicate network is left more or less intact but depleted on the network modifier ions of the glass and enriched on hydrogen. Contrary to Eq. (16.1), in alkaline solutions an attack of the hydroxyl ions occurs, where a breakdown of the silicate network must be observed according to Eq. (16.2) – Si – O – Si – þ OH2 Y – Si – OH þ Si – O2
ð16:2Þ
It is the purpose of this chapter to demonstrate the applicability of microanalytical techniques delivering information on the corrosion of medieval stained glass and medieval enamels exposed to the ambient atmosphere for centuries. This requires analytical methods adequate for the morphological and chemical characterization of the silicate surfaces and the interfaces formed during weathering. Therefore, the common tool of scanning electron microscopy (SEM) in combination with energy dispersive X-ray microanalysis (SEM/EDX) and the recent techniques of secondary ion mass spectrometry (SIMS) and nuclear reaction analysis (NRA) were used for the analysis of the corrosion products and the layers built up on the glass surfaces over nearly a millennium. Additionally, infrared reflection absorption spectrometry (IRRAS) and tapping mode atomic force microscopy (TM-AFM), a special technique of scanning probe microscopy (SPM) could be applied for systematic investigations of the initial stages of the corrosion processes. During these studies sample glass with a chemical composition similar to that of medieval stained glass were investigated in situ while being exposed to a controlled environment (i.e., with known relative humidity and a specific content of an acidifying gas such as SO2). 715
M. Schreiner
16.2
THE WEATHERING OF MEDIEVAL STAINED GLASS
Medieval stained glass has been used as window panes in churches, cathedrals or other historic buildings of the 10th to the 15th century in Europe. Their alarming condition has resulted already in the 1950s in a multidisciplinary co-operation among curators, restorers, conservators and scientists. These historic objects consisting of numerous pieces of glass coloured with metal oxides (Plate 16.II) show accelerated deterioration, particularly on their exterior surfaces (Plate 16.III). Although many Romanesque and Gothic artefacts have escaped major damage by environmental influences for centuries or, in case of the earliest stained glass windows, even for a millennium, in some cases today’s air pollution threatens to cause total destruction. The continuous corrosion of the exterior surface of the glass reduces its thickness and a so-called weathering crust is built up on the decomposed surface. Consequently, the transparency of the glass painting
Plate 16.II. Medieval glass painting: The Birth of Christ, North window of the former monastery church at Viktring/Austria. ( For a colored version of this figure, see Plate 16.II.)
716
Corrosion of historic glass and enamels
Plate 16.III. Exterior surface of a medieval glass painting with a weathering crust completely covering the glass surface. ( For a colored version of this figure, see Plate 16.III.)
is reduced and in some cases the entire composition is barely recognizable (Plate 16.IV) [16]. Analytical investigations carried out by Geilmann [17] as early as 1960 showed that the weathering crust consists of non-crystalline products such as hydrated silica, with gypsum (Ca2SO4·2H2O) and syngenite (K2SO4· CaSO4·H2O) as crystalline components. Geilmann [17] concluded that these corrosion products are formed by reaction of glass components with SO2 from the ambient atmosphere, as the sulphur content of the glass investigated was rather low. Systematic investigations carried out on objects all over Europe yielded similar results [6,12,18 – 21]. The main corrosion products observed are a mixture of silica with potassium and calcium sulphates. Furthermore, Newton [6] could show that in some cases also the storage conditions of several objects during World War II initiated the accelerated deterioration of medieval glass paintings. Nevertheless, different corrosion and weathering phenomena could be observed on in situ objects [16]. The results of analytical investigations
717
M. Schreiner
Plate 16.IV. Medieval glass painting with a weathering crust on the exterior surface reducing the transparency so that the composition is barely recognizable. ( For a colored version of this figure, see Plate 16.IV.)
carried out on artefacts from Great Britain, France, Italy, and the Germanspeaking countries revealed that medieval glass was made from local raw materials. A mixture of one part sand and two parts beechwood or fern ash and chalk was common for that period. The reason for using ash instead of soda, which is typical for ancient and modern glasses, is not yet fully understood. It was, however, recommended also by Theophilus [22]. In fact, because beech wood ash contains iron and manganese in the range of 0.1 – 0.5 wt%, most colours except red could be obtained merely by manipulating the atmospheric conditions in the furnace without adding any other ingredients. As a result of this manufacturing process, the medieval glasses are at a serious disadvantage, as their chemical composition differs so profoundly from that of modern or ancient glasses. The total amount of silica and other so-called network formers such as aluminium, which are the main components responsible for the chemical durability [7,10,14,15], is very low in glass of the Middle Ages. In addition, potassium instead of sodium was
718
Corrosion of historic glass and enamels
introduced into the silicate network by using plant ash. It is a known fact that sodium-containing glass is approximately twice as durable as glass with comparable amounts of potassium [10,23,24]. Several studies showed that medieval glass weathers only slightly if the SiO2 content is higher than 60 mol% [20,25]. For such glasses it appears to be unimportant whether the network modifier is an alkali (e.g., K) or an alkaline earth (e.g., Ca). A marked change in the chemical behaviour was observed for glass panels with less than 60 mol% silica. Here, the alkali ions seem to influence the weathering durability of the glass in a much higher degree than the alkaline earths do. Window glasses with less than 60 mol% SiO2 and more than 13 mol% K2O usually show a tremendously corroded, crusted surface. Pitting corrosion is observed on glasses containing both low SiO2 and low potassium levels in the bulk [12,25]. 16.2.1 SEM investigations of the corrosion phenomena on naturally weathered medieval glass For the SEM investigations presented a JEOL (1418 Nakagami Akishima, Tokyo, Japan) instrument, type SEM 6400, and an energy dispersive X-ray analysis system of Link (High Wycombe, Buckinghamshire, UK), type eXL, could be used. The glass samples had to be coated with a thin layer of carbon prior to analysis in order to avoid charging effects on the electrically non-conducting corroded glass surfaces during the electron bombardment. Samples of Austrian medieval stained glass objects exposed to the ambient atmosphere for nearly 700 years were available. A morphological characterization of their surfaces as well as a chemical identification of the corrosion products present on the glass could be carried out. In a few cases small splinters of the glass could be cut off by using a low speed diamond saw, embedded in resin and cross-sectioned perpendicular to the surface in order to study the elemental distribution in the near surface region and get an overview of the components leached from the materials. The backscattered electron signal (BE images) and X-ray mapping (elemental distribution images) were used therefore. On the medieval glass samples investigated so far, three types of deterioration could be observed depending on the chemical composition of the bulk glass [13,26]: glass with a crusted, a pitted or an apparently unweathered surface. The most corroded glass samples are covered with a more or less uniform weathering crust shown in Fig. 16.2. Usually this crust was extremely hard and could not be separated mechanically from the glass.
719
M. Schreiner
Fig. 16.2. SEM micrograph of the surface of a naturally weathered medieval glass with a thick weathering crust consisting of flat, plate-like gypsum crystals and hydrated silica.
Calcium and sulphur could be determined in this layer with the energy dispersive microanalyzer in the SEM and investigations carried out by X-ray diffraction analysis have shown that in most cases gypsum (CaSO4·2H2O) and syngenite (K2SO4·CaSO4·H2O) has been formed as crystalline corrosion products [13]. In Fig. 16.2 gypsum can be recognized by its flat, plate-like structure. Furthermore, the SEM micrograph (Fig. 16.2) shows a fine-grained corrosion product, which consists of silica and small amounts of iron and manganese, indicating that amorphous hydrated silica also was formed during the weathering process. In only a few cases the weathering crust was quite loose and powdery and could be separated by a scalpel or a brush from the glass beneath, which had suffered a great deal of damage. Numerous small craters could be observed in the SEM, as shown in Fig. 16.3. In general, such heavily corroded surfaces leading to a thick weathering crust, which reduces the transparency of the glass, occur most frequently on medieval objects with a very low silica content and large amounts of potassium and calcium oxides [13,20,25]. Glasses with greater amounts of silica are particularly susceptible to pitting corrosion, a type of weathering typical for medieval glass (Figs. 16.4 – 16.6). Although the glass surfaces appear to be intact, microdomains of 50 – 150 mm in diameter show definite material deterioration. Numerous cracks are formed within these domains
720
Corrosion of historic glass and enamels
Fig. 16.3. The SEM micrograph of the glass surface beneath the removed weathering crust shows a high amount of pits and a great deal of damage.
(Fig. 16.6) and weathering products are located in the cavities. Since corrosion is still relatively low, identification of the corrosion products could not be carried out by X-ray diffraction analysis. Nevertheless, the micrograph in Fig. 16.5 and the microanalysis in the SEM/EDX revealed flat plates consisting of calcium and sulphur indicating the formation of gypsum. Although the reason for this type of weathering is not yet fully
Fig. 16.4. SEM micrograph of medieval glass with pitting corrosion.
721
M. Schreiner
Fig. 16.5. SEM micrograph of a pit filled with weathering products such as gypsum (plate-like structure) and hydrated silica (powdery).
understood, it is assumed that pitting corrosion is an initial stage of weathering for medieval window glass. It has been observed on objects from all over the European continent that, where the pits are enlarged and corrosion proceeds, in time a weathering crust is formed over the entire glass surface. The enlargement of not just the crater diameter but predominantly
Fig. 16.6. Initial pitting corrosion, seen with SEM.
722
Corrosion of historic glass and enamels
the crater depth leading to perforated window glass has been recognized on objects from Great Britain [6]. A smooth and generally unaltered surface is observed for the third type of medieval objects, the apparently unweathered glass. Such window panes show just a little iridescence due to the prolonged exposure to the ambient atmosphere. As already mentioned above, the chemical analyses published in the literature [6] revealed that glasses with a silica content above 60 mol% SiO2 and potassium and calcium as network modifiers are of high weathering stability as well as dark green coloured window glass with considerable lead and phosphorous oxides and copper as a colouring agent [13,21]. However, by applying SEM to such objects some incipient pitting and the formation of concentric microcracks, as shown in Fig. 16.7, can be observed. Additional results concerning the formation of such cracks could be obtained from the cross-section of a medieval glass. As shown in Fig. 16.8, the microcracks range several microns beneath the surface and the glass appears almost completely decayed within a region of 20 – 40 mm from the surface leading to weathering phenomena, as shown in Fig. 16.9. In the region of the microcracks in Figs. 16.7 and 16.8 a remarkable depletion of potassium and calcium could be determined with energy dispersive X-ray microanalysis (Fig. 16.10). These results indicate that the weathering of medieval glass must be governed by the ion exchange process mentioned above in Eq. (16.1). It can be determined by means of SEM/EDX on cross-sections of glass samples taken
Fig. 16.7. SEM micrograph of the surface of a naturally weathered medieval glass showing concentric cracks.
723
M. Schreiner
Fig. 16.8. Cross-section of medieval glass with pitting corrosion.
from scarcely weathered glass as well as glass samples with a thick weathering crust. In Fig. 16.11 the cross-section of a medieval glass is presented with the corresponding X-ray mappings for K, Ca and Si. A marked depletion of the mono- and bivalent network modifiers potassium and calcium must be observed in the near surface region, whereas silicon is more or less homogeneously distributed in the bulk glass as well as in the
Fig. 16.9. SEM micrograph of a pit formed on the surface of a medieval glass. The glass seems to be intact.
724
Corrosion of historic glass and enamels
Fig. 16.10. Results of the energy dispersive X-ray microanalysis in the SEM obtained for the intact glass and the domains near the microcracks in Figs. 16.7–16.9 that appear dark in the BE-image.
surface domain. Even a slight “enrichment” of Si can be concluded from Fig. 16.11d, but it has to be kept in mind that the analysed domain in energy dispersive X-ray microanalysis depends on the chemical composition as well as the density of the materials. As the leached layer is depleted on K and Ca and hydrogen incorporated into the silicate structure, the average atomic number of this surface layer as well as its density is lower compared to the bulk. Consequently, the penetration depth of the primary electrons is higher than in the bulk material, which also increases the volume of X-ray analysis. 16.2.2 The determination of hydrogen in the leached surface layer by SIMS and NRA As hydrogen cannot be detected by X-ray microanalysis, SIMS and NRA were applied in order to verify the ion exchange mechanism as the most important corrosion process occurring during the weathering of medieval stained glass. SIMS plays a unique role in corrosion science and surface analysis due to its high detection power for most elements in the periodic
725
M. Schreiner
Fig. 16.11. Cross-section of a strongly weathered medieval glass as seen with SEM and the corresponding elemental mappings of K (b), Ca (c) and Si (d).
table including hydrogen and its use for depth profiling. However, a strong matrix effect (high variation of the ion yields of the various elements and for the same element in different materials or layers) must be considered, when interpreting the depth profiles quantitatively. Quantitative distribution analysis of hydrogen is further hampered by the fact that the deposition of hydrogen-bearing molecules from the residual gas in the sample chamber represents a constant source of hydrogen to the sputtered sample. Additionally, applying SIMS to the analysis of glasses is a complex procedure because of sample charging occurring during the analysis of insulators under energetic ion bombardment. Using a CAMECA (Courbevoie, France) IMS-3f ion mass spectrometer the specimen, first coated with gold, were bombarded with primary mass filtered 16O2 ions at 14.5 keV. A beam current of 100 nA, a beam diameter of approximately 60 mm, a raster size of 250 £ 250 mm2, and an analysed area with a diameter of 10 mm were selected for the analysis. The secondary ions were accelerated into the double-focusing mass spectrometer by an actual accelerating voltage of þ 4500 V, which was controlled and corrected at each
726
Corrosion of historic glass and enamels
measurement cycle by a high-voltage autocontrol computer routine for further minimization of the residual charging effects. This routine, already described in the literature [27], checks the position of the energy distribution of a reference mass (for glass analysis 29Siþ) and normalizes the position of the energy distribution of all other masses to that reference element. Due to the large number of elements present in original medieval glass, a computer routine was used which enabled the simultaneous analysis of up to 29 masses. The counting time for each mass was 1 s but almost no depth information was lost during the analysis because of the rather low erosion rate used. The latter was determined from depth measurements of SIMS craters obtained by shorter runs using a DEKTAK profilometer (Sloan Technology Division, Veeco Instruments, Inc, Santa Barbara, CA, USA) [28]. For the NRA of hydrogen the resonant nuclear reaction 1H(15N,ag)12C was used. This resonance reaction yields characteristic g-rays of 4.43 MeV. As shown in Fig. 16.12, the hydrogen depth distribution in the medieval glass can be obtained by increasing the 15N energy from the resonance energy of 6.385 MeV stepwise to higher values. As the ion beam penetrates the material the ions lose energy until they reach the resonant energy for
Fig. 16.12. Scheme of the NRA technique for profiling hydrogen in glass. By increasing the energy of the incident ion beam the “resonance window” can be moved progressively deeper. The depth resolution diminishes with increasing depth due to energy straggling.
727
M. Schreiner
Fig. 16.13. (a) SIMS raw depth profile of the elements present in a medieval glass. Layer 1, Au and outermost glass surface; Layer 2, leached layer; Layer 3, bulk glass. (b) Quantified SIMS depth profile of Fig. 16.13a using relative sensitivity factors [30].
hydrogen. The yield of 4.43 MeV g-rays is proportional to the amount of hydrogen in the resonance window. The depth resolution is about 3 nm at the sample surface and decreases slowly with increasing depth, e.g., 10 nm at 400 nm depth [29,30]. The energy loss in the target can be calculated with the aid of energy loss tables [31]. Therefore, by increasing the energy from 6.385 MeV and measuring the yield of 4.43 MeV g-rays the concentration of hydrogen versus depth can be determined. The 15N beam was delivered by the 7 MV Van de Graaff accelerator at the Institute of Nuclear Physics of the J.W. Goethe-University in Frankfurt/Germany. The 4.43 MeV g-radiation emitted from the sample was measured with a NaI scintillation detector. The vacuum in the sample chamber was of the order of 1027 mbar. Electrical charging of the glass samples could be prevented by an electron emitting heated filament. NH4Cl was used as sample of known hydrogen concentration for a more accurate
728
Corrosion of historic glass and enamels
Fig. 16.13 (continued).
calibration, although the scale of absolute hydrogen concentration in a NRA profile can be determined in principle without a calibration standard from the peak cross section, the resonance width and the detector efficiency. Stopping power values necessary for the calibration of the depth scale and for the conversion of g-ray counts to hydrogen concentration were calculated from the glass composition. Figure 16.13 summarizes the results obtained from SIMS measurements on apparently unweathered (scarcely weathered) medieval glass samples without any corrosion products on the surface. The raw depth profile in Fig. 16.13a was obtained by recording ion intensities measured by cyclic switching between the masses. Already from this raw profile a low intensity for the main glass constituents K and Ca (mass 41 and 42, respectively) and the minor components Na and Ba (mass 23 and 138, respectively) in the near surface region (Layer 2 in Fig. 16.13a) can be deduced, whereas the intensities, e.g., at the masses one for hydrogen and 29 for silicon, are increased compared to the bulk glass (Layer 3 in Fig. 16.13a). The conversion
729
M. Schreiner
of the measured secondary ion currents versus analysing time into a concentration-depth profile in Fig. 16.13b was carried out by applying relative sensitivity factors (RFS) [32] determined at cross-sections of the intact medieval glass using Si as reference element. It is obvious from Fig. 16.13b that the network modifiers K, Ca, Na, Ba and also Mg are strongly depleted in the leached layer compared to the bulk. The DEKTAK measurements revealed that the higher intensity of Si in Fig. 16.13a (mass 29) in Layer 2 (leached layer) is caused predominantly by the higher erosion rates of the negative primary ions in this surface domain compared to the bulk. Although hydrogen could be detected by SIMS (Fig. 16.13a), quantification was not carried out due to the hydrogen-containing residual gas in the sample chamber. Comparative measurements with NRA on medieval glass samples have revealed that up to 40 at.% H is present in the leached
Fig. 16.14. Quantitative depth distribution of hydrogen in medieval glass exposed to the ambient atmosphere for nearly 700 years. The measurements were carried out by using NRA.
730
Corrosion of historic glass and enamels
layer [33], as shown in Fig. 16.14. Usually a pronounced enrichment of hydrogen within a region of 1 – 3 mm from the surface could be determined by SIMS as well as NRA at apparently unweathered (scarcely weathered) glass samples. However, for samples with pitting corrosion and especially specimen with a crusted surface (Glass 2 in Fig. 16.14), where the weathering crust was removed prior to analysis, the depth of the complete leached layers could not be determined by NRA and SIMS although the energy of 15N ions from the accelerator was increased to about 12 MeV. 16.2.3
Leaching studies of glass with medieval composition
Based on the results presented in previous sections, model glasses with medieval glass composition (Table 16.1) were prepared at the FraunhoferInstitut fu¨r Silicatforschung in Wu¨rzburg/Germany and used for leaching experiments in hydrochloric, nitric, sulphuric and oxalic acids. Due to the chemical composition Glass M1 is typical for medieval glass with a low weathering stability, whereas the chemical durability of Glass M3 is much higher, as the amount of network formers is higher. Glass M3 contains simply K2O, CaO and SiO2. The results of these systematic investigations concerning the leaching behaviour of the model glasses are already published in the literature [34], where a remarkable influence of the concentration and nature of the acidic solution on the amount of potassium and calcium leached was obtained. The glass samples treated in the aqueous solutions were used for SIMS analysis. Figures 16.15 and 16.16 present the SIMS raw depth profiles obtained for Glass M3 treated in sulphuric and hydrochloric acids, respectively. In these depth profiles (secondary ion intensity versus time) TABLE 16.1 Chemical composition (mol% and wt%) of the model glasses with medieval composition used for the leaching and weathering experiments Mol%
Na2O K2O MgO CaO P2O5 Al2O3 SiO2
Wt%
Glass M1
Glass M3
Glass M1
Glass M3
3.20 18.90 5.00 17.80 1.77 0.46 53.10
– 8.88 – 29.24 – – 61.70
3.0 25.5 3.0 15.0 4.0 1.5 48.0
– 25.0 – 15.0 – – 60.0
731
M. Schreiner
Fig. 16.15. SIMS depth profile measured for Glass M3 after treatment in 0.1 N H2SO4 for 1 h. Layer 1, Au; Layer 2, leached layer; Layer 3, bulk glass. In the leached layer the intensity for H (mass 1) is increased compared to the bulk and the ratio of the ion intensities measured at mass 32 and 34 corresponds to the natural ratio of the sulphur isotopes.
the leached layer (Layer 2 in Figs. 16.15 and 16.16) can be clearly seen by the increased intensity of silicon (masses 29 and 30) and hydrogen (mass 1) as compared to the bulk (Layer 3). Shorter runs and subsequent crater depth measurements with the DEKTAK profilometer have proved that this increased intensities at the masses 29 and 30 are caused also mainly by a higher erosion rate of that layer similar to the medieval glass samples in Fig. 16.13. Additionally, decreased intensities for potassium (mass 41) and Ca (mass 42) were obtained in Layer 2 due to leaching in the acidic solutions. The most surprising result concerns the masses 32 and 35 of the glass treated in sulphuric and hydrochloric acid, respectively. As can be seen in Fig. 16.15, the depth profile of the specimen treated in H2SO4 reveals secondary ion intensities at the masses 32 and 34, which are the
732
Corrosion of historic glass and enamels
Fig. 16.16. SIMS depth profile obtained for Glass M3 after treatment in 0.1 N HCl for 1 h. Layer 1, Au; Layer 2, leached layer; Layer 3, bulk glass. The ratio of the ion intensities measured at mass 35 and 37 corresponds to the natural ratio of the chlorine isotopes.
characteristic isotopes for S. The measurements
