Synchrotron Radiation and Nanostructures: Papers in Honour of Paolo Perfetti

Proceedings of the Workshop on

Synch

on Radiation and Nanostructures Papers in Honour of Paolo Perfetti

This page intentionally left blank

Proceedings of the Workshop on

Synchrotron Radiation and Nanostructures Papers in Honour of Paolo Perfetti

Antonio Cricenti \stituto di Struttura della Materia, Italy

Giorgio Margaritondo Ecole Politechnique F6d6rale de Lausanne, Switzerland

World Scientific NEW JERSEY· LONDON· SINGAPORE· BEIJING· SHANGHAI· HONG KONG· TAIPEI· CHENNAI

Published by

World Scientific Publishing Co. Pte. Ltd. 5 Toh Tuck Link, Singapore 596224 USA office: 27 Warren Street, Suite 401-402, Hackensack, NJ 07601 UK office: 57 Shelton Street, Covent Garden, London WC2H 9HE

British Library Cataloguing-in-Publication Data A catalogue record for this book is available from the British Library.

SYNCHROTRON RADIATION AND NANOSTRUCTVRES Proceedings of the Workshop Copyright © 2009 by World Scientific Publishing Co. Pte. Ltd. All rights reserved. This book, or parts thereof, may not be reproduced in any form or by any means, electronic or mechanical, including photocopying, recording or any information storage and retrieval system now known or to be invented, without written permission from the Publisher.

For photocopying of material in this volume, please pay a copying fee through the Copyright Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923, USA. In this case permission to photocopy is not required from the publisher.

ISBN-13 978-981-4280-83-9 ISBN-I0 981-4280-83-6

Printed in Singapore by 8 & JO Enterprise

PREFACE This book is dedicated to Paolo Perfetti an outstanding scientist and a wonderful friend. It is a collection of articles that were presented at the Workshop on Synchrotron Radiation and Nanostructures, held in Rome in November 2008 - that was also in the honor of Paolo and attracted many of his friends worldwide. Paolo Perfetti, in a long and illustrious career, made fundamental contributions to the development of synchrotron radiation and scanning probe microscopy and to their applications in materials science and biology. It would take a whole book to describe in detail his many results in a variety of domains. Hence we could only note some of the most important ones here. As early as 1971, Paolo pioneered molecular beam epitaxy (MBE), then a technique in its infancy and almost unknown outside the USA. Not even the name was universally known and used, so Paolo developed one of the first systems worldwide under a different name: "Bis technique" after R. F. Bis, author of early work in the epitaxy of lead chalcogenides (R. F. Bis et al., J. Vac. Sci. Technol. 9, 226 (1972)). During the same period, he became interested in semiconductor heterojunctions and their interface properties. One of the earliest works witness the shift of a then very young Federico Capasso from laser research to a bright career in interface engineering (P. Perfetti, M. Antichi, F. Capasso and G. Margaritondo, Infrared. Phys. 14, 255 (1974)). Working in Berkeley in collaboration with David Shirley, Paolo invented a new and very productive line of research: the use of photoemission to measure heterojunction band discontinuities (P. Perfetti, D. Denley, K.A. Mills and D.A. Shirley, Appl. Phys. Letters 33, 66 (1978)). Many scientists adopted this approach in the following decades. This was also one of the milestones in the development of synchrotron radiation and specifically of synchrotron-based photoemission.

v

vi

a formal appearance at his office (left) and a more version as a "fraschetta" with the editors book

r""""'U'l" ,,,,tEl....,ftf

radiation remained a element of and made him famous throughout contributions include and diffraction, X-ray absorption and EXAFS Su[loomams from solid interface formation to continuation of Paolo's work in synchrotron 11'lI.lll(1L:lVH more recent activity in free electron laser (FEL) SCll::::nCC. to use of in scanning near-field ,-"u,",,,,,,,, et al., Appl. Phys. Letters 73, 151 (1998)). He was a new giant Italian FEL projects work of Paolo was not limited to synchrotron ll'll.ll!',",,'vu his search for new phenomena VL~.H<'u'''''''VU brought him to master different tools such as energy lTn"pn,.. photoemission, neutron experiments, laser rlp~"""lti In recent years, he increasingly on VUO,"ll.l'''' the traditional boundaries of physics: neurobiology, "'V'1V/,;.1' and other biomedical subdomains. h.,.".h-n..,

vii

Paolo is not only an outstanding scientist and a pioneer of new techniques, but also a science leader. He was a key person in the development of the Frascati synchrotron project (PULS). For many years (1987 -2008), he was the Director of the Istituto di Struttura della Materia (ISM) of the Consiglio Nazionale delle Ricerche (CNR) in Rome, Italy - a reference worldwide for solid state physics, interface science, novel microscopies and other domains. The workshop that led to this book was also to celebrate his retirement after 21 years of leadership. We are very happy to honor him with this volume, containing a collection of contributions from his present and former co-workers, as well as from friends in the scientific community. This is our way to thank him for his friendship, enthusiasm and human wisdom. He is an example of rigor, honesty and dedication to science. The workshop took place in November 21, 2008 at the CNR Research Area of Tor Vergata, Rome, Italy. It was attended by 75 scientists from seven countries. Others who could not personally attend sent personal messages, greetings and best wishes to Paolo. The day was crowned by wine and "porchetta" at a Frascati "fraschetta", accompanied by bawdy Roman songs to which Paolo participated with great enthusiasm (offsetting his objectively limited singing skills). The workshop's objective was to present the capabilities of state-ofthe-art synchrotron radiation and scanning probe microscopy techniques, together with general theory work, in elucidating the fundamental electronic and structural properties of semiconductor and metal surfaces, interfaces, nanostructures, layers and diverse biological systems. We are grateful to the workshop sponsors, the Department of Materials of CNR and the Ecole Poly technique Federale de Lausanne (EPFL). We also wish to thank all the staff members of the ISM-CNR for the excellent support, organization and hospitality. Antonio Cricenti ISM-CNR, Rome. Italy Giorgio Margaritondo EPFL, Lausanne, Switzerland

This page intentionally left blank

CONTENTS

Preface Adsorption of cyclopentene on GaAs(OO 1) and InP(OO 1), a comparative study by synchrotron-based core level spectroscopy R. Passmann, T. Bruhn, B.o. Fimland, W. Richter, M. Kneissl, NEsser, P. Vogt

1

Time resolved energy dispersive X-ray reflectometry as a tool for material science studies: The case of organic solar cells B. Paci, A. Generosi, V Rossi Albertini, P. Perfetti, R. De Bettignies, C. Sentein

12

Novel oxidation process on chiral silicon nanowires P. De Padova, B. Aufray, G. Le Lay, C. Quaresima, P. Perfetti, B. Olivieri

24

From Bequerel to nanotechnology G. Margaritondo

34

High-resolved fluorescence imaging of X-ray micro-radiographics on novel LiF detectors F. Bonfigli, S. Almaviva, G. Baldacchini, F. Flora, A. Lai, R.M. Montereali, M.A. Vincenti, A. Cricenti, C. Oliva, A. Ustione, A. Faenov, T. Pikuz, L. Reale, P. Gaudio, S. Martellucci, M Richetta

40

Growth mechanisms of tin oxide and zinc oxide nanostructures from vapour phase L. Zanotti, M Zha, D. Calestani, R. Mosca, A. Zappettini

48

Broken-symmetry states at surfaces: The (tr)ARPES view L. Perfetti, M. Grioni

56

ix

x

Nanostructuring through laser manipulation F. Tantussi, N. Porffido, F. Prescimone, F. Fuso, E. Arimondo, M. Allegrini

68

Localization and diffusive processes in the electronic transport in quasi one-dimensional nanostructures

F. Flores, B. Biel, P. Sundqvist, F.J Garcia-Vidal Optical and electron energy loss spectra of liquid water: An ab-initio study 0. Pulci, V Garbuio, R. Del Sale

76

90

Electronic confinement of silver nanocluster in Er3+-activate silicate and phosphate glasses L. Minati, G. Speranza, A. Chiappini, A. Chiasera, M. Ferrari, S Berneschi, S Pelli, G. C. Righini

102

Dynamics at metal/semiconductor interfaces and exotic phenomena through the looking glass G. Le Lay

110

Supramolecular interaction of chiral molecules at the surface G. Contini, N. Zema, P. Gori, A. Palma, F. Ronci, S Colonna, S Turchini, D. Catone, A. Cricenti, T. Prosperi

115

Microradiology imaging of the biodistribution of polyethylene glycol (PEG) modified gold nanoparticles in cancer bearing mice c.-c. Chien, c.-J. Liu, H.-S Chen, c.-H. Wang, S-T. Chen, w.-H. Leng, Y. Hwu

132

The Frascati experiment G. Faraci

147

AFM and SNOM techniques at ISM: An overview M. Girasole, G. Longo, G. Pompeo, A. Cricenti

155

xi

a-SnlGe(111) and a-SnlSi(111) surfaces studied by STM measurements and ab-initio calculations S. Colonna, A. Cricenti, P. Gori, F. Ronci, 0. Pulci, G. LeLay

171

Development of Scanning Probe Microscopy Electronic Control at ISM M Luce, M Rinaldi, R. Generosi, A. Cricenti

197

Nanostructures induced by the absorption of Fullerenes: Structural and electronic properties R. Fetici, M Pedio

209

This page intentionally left blank

ADSORPTION OF CYCLOPENTENE ON GaAs(OOl) AND InP(OOl), A COMPARATIVE STUDY BY SYNCHROTRON-BASED CORE LEVEL SPECTROSCOPY R. PASSMANNa,b, T. BRUHNa,b, B. O. FIMLANDc, W. RICHTERd,a,

M. KNEISSLa, N. ESSERa,b, P. VOGTa a Institut

fur Festkorperphysik, Technische Universitat Berlin, Hardenbergstr. 36, D-10623 Berlin, Germany, b ISAS - Institute for Analytical Sciences - Department Berlin, AlbeTt-Einstein-Str. 9, 12489 Berlin, Germany, C Department of Electronics and Telecomunication, Norwegian University of Science and Technology, 7491 Trondheim, Norway d Universita degli Studi Roma 'Tor Vergata', Via della Ricerca Scientifica 1, 00133 Roma, Italy

Keywords: Synchrotron x-ray photoelectron spectroscopy, SXPS, organic molecules, InP(OOl), GaAs(OOl), (2 x 4) III-V semiconductors, cyclopentene

1. Abstract

The interface formation between cyclopentene and the (2 x 4) reconstructed GaAs(OOI) and InP(OOI) surfaces has been studied by soft X-ray photoemission spectroscopy (SXPS). After preparation of an uncontaminated (2x4) reconstruction under ultra-high vacuum (UHV) conditions the surfaces were exposed to cyclopentene. The changes in the In 4d and P 2p, as well as the Ga 3d, As 3d and C Is, core level emission lines indicate a covalent bonding of cyc10pentene to the topmost atoms of these surfaces. Based on these results and due to the comparison between the results for the different III-V semiconductors, two different adsorption structures of cyc1opentene are found. Our results suggest that cyc10pentene adsorbs onto the InP(OOI)(2 x 4) surface by the formation of two bonds, in a cycloadditionlike reaction. For the adsorption of cyclopentene on the GaAs(OOI)(2 x 4) surface, only a single bond formation is found. The difference in the adsorption structures are explained by the different dimer configurations of the substrate surfaces.

2

2. Introduction Synchrotron based photoemission spectroscopy (SXPS) is suitable for the investigation of clean reconstructed semiconductor surfaces as well as for the characterization of the interface formation between organic molecules and such reconstructions. This method can help to elucidate bonding configurations based on the analysis of core level emission line shapes, the identification of surface core level components and the chemical composition of the surface. Therefore, adsorption structures of organic molecules on semiconductor surfaces are presented based on SXPS data. By the understanding of the interface formation of such heterostructures, new applications can be developed or improved. Since organic molecules can be synthesized chemically, we can chose from a large variety of organic molecules in terms of structtire and functionality.1-3 Until now, most results concern the interface formation between organic molecules and the Si(OOl)(2 x 1) reconstructed surface. 1,2,4-12 In that context, it was shown that the surface dimer structure is a basic prerequisite for the bonding mechanism. For example, a bonding via a [2 + 2J-cycloaddition reaction is only allowed for a bonding on an asymmetric surface dimer, which is the case for the Si(OOl)(2 xl) surface. 2,13,14 In order to clarif:y the role of the surface dimer structure on the bonding mechanism, tions of molecule adsorption on different dimer configurations are necessary. Such different dimer configurations are provided in our investigations by GaAs(OOl) and InP(OOl) surfaces,15,17 as can be seen in Fig. 1.

tip

e

Ga

As

In3fold 'n4fO'd

(110] [110J

Fig. 1. On the left: topview of the InP(OOl)(2 x 4) surface reconstruction. On the right: topview of the GaAs(OOl)(2 x 4) surface reconstruction.

The atomic structures of the clean reconstructed InP(OOl)(2 x 4) and the GaAs(OOl)(2 x 4) surfaces have been investigated in recent years. The InP(OOl)(2 x 4) surface is well described by the so-called asymmetric 'mixed-dimer' model, a structure that consists of an In-P hetero-dimer with a filled dangling bond at the P atom and an empty dangling bond at

3

the In atom. 16 ,17 These dimer atoms are bonded to fourfold coordinated second layer In atoms. The outer In atoms of the second layer are threefold coordinated. For the GaAs(001)(2 x 4) the topmost layer consists of As atoms forming dimers with double occupied dangling bonds which are bonded two second layer Ga atoms. 18 Core level spectroscopy was performed to investigate the interface formation between the III-V (001) surfaces and cyclopentene. Numerical analysis of the In 4d, P 2p, Ga 3d, As 3d and C Is core level emission lines by best fit reveals information on the interface formation. Additionally, reflectance anisotropy spectroscopy (RAS) measurements were performed to monitor the deposition process. Based on the experimental results, structure models for the cyclopentene-surface linkage on the two (2 x 4) reconstructed surfaces, the InP(001)(2 x 4) and the GaAs(001)(2 x 4) surfaces, were developed.

3. Experimental The Sb-doped InP(OOl) samples, investigated here, were grown by metalorganic vapour phase epitaxy (MOVPE) using phosphine (PR 3 ) and trimethylindium (TMIn) as precursors. Directly after growth, the samples were capped with an amorphous phosphorous/arsenic double layer by photo-decomposition of PR3 and AsR 3 in the MOVPE reactor.19 The GaAs(OOl) samples used in this work were Si-doped (nominal n = 5 x 10 17 cm- 3 ), grown by molecular beam epitaxy (MBE) and capped with amorphous arsenic directly after growth using an AS 2 flux. 2o After transfer of the samples to URV, contamination free and well ordered (001) surfaces were prepared by heating to approximately 400°C(±20°C) for 15 min for InP and to 420°C(±20°C) for 15 min for the GaAs(001)(2 x 4). After this procedures the surface reconstructions were determined by low energy electron diffraction (LEED) showing a clear (2 x 4) pattern in both cases. The base pressure throughout all experiments was below 2 x 10- 10 mbar. Cyclopentene with a purity higher than 97% was introduced from gas phase into the chamber using a variable gas-inlet valve. During the deposition, the samples were held at room temperature (RT). In order to avoid decomposition of the molecules, all filaments inside the chamber, e.g. ion gauges, were switched off during the exposure. The effective cyclopentene layer thickness was estimated from SXPS measurements to be approximately one monolayer. The whole deposition process was monitored by

4

RAS. Synchrotron based photo emission measurements were performed at the Russian-German beamline (RGBL) at the synchrotron facility BESSY II. The spectra were taken in normal emission with a total instrumental resolution (beamline plus analyser) of 120 meV at an excitation energy of 75 eV. Kinetic energies (binding energies) of the In 4d, P 2p, As 3d, Ga 3d and C Is core level emission lines refer to the Fermi edge determined by photoemission from a molybdenum sample holder in electrical contact with the samples. The core level spectra were analyzed by numerical deconvolution into pairs of spin orbit-split doublets each of which consists of convoluted Lorenzian and Gaussian line shapes corresponding to lifetime and experimental broadening, respectively. The measured data points (spheres in Fig. 2-4) are shown together with best-fit from numerical analysis (full lines). For all measurements at the InP samples, a lifetime broadening for the In4d (P2p) core levels of 0.1 eV (0.06 eV), a branching ratio of 1.5 (2.0), an experimental broadening of 0.46 eV (0.41 eV), and a spin-orbit splitting of 0.86 eV (0.87 eV) were found. For the measurements at the GaAs samples, a lifetime broadening for the As 3d (Ga 3d) core levels of 0.1 eV (0.1 eV), a branching ratio of 1.5 (1.7), an experimental broadening of 0.50 eV (0.39 eV), and a spin orbit-splitting of 0.69 eV (0.43 eV) were found. The resulting residuum is shown below each fit.

4. Results and Discussion

4.1. Results for the adsorption of cyclopentene on the InP(OOl){2 X 4) For the adsorption of cyclopentene on the InP(00l)(2 x 4) reconstructed surface, it is well known that the two different surface bonding sites lead to a change for the surface core level components in In 4d and P 2p core levels, as seen in Fig 2. These changes are explained by an adsorption of cyclopentene on the 'mixed-dimer' as the main surface adsorption site. Additionally, subsequently bonding to the second layer In-In bonds is observed as we could show in our previous work. 21 Beside the two surface components in the In 4d core level emission line 22 ,23 (shaded) which are revealed by numerical analysis, an additional component after cyclopentene saturation is found. This component shifted towards lower kinetic (higher binding) energies, In-C, is assigned to a bond formation of cyclopentene to the topmost In atoms. 21 For the P 2p core level a second surface related component shifted towards lower kinetic (higher

and core level emission lines cy,elope:nt€:ne saturated (right) InP(OOl)(2 x 4) reconstructed

6

binding) energies, P-C, is found too. This component is assigned to a bond formation of cyclopentene to the topmost P atoms. 21 These components are explained to stem from In and P atoms which bond to carbon atoms of the cyclopentene molecules. The shift towards higher binding energies results from a charge transfer from In and P (1.8 and 2.2) to C (2.5) due to the higher electro negativity of the latter. These observations are supported by the C Is core level, as shown on the left graph of Fig. 3. Three main components have been revealed by the numerical analysis, C-C, C-In and C-P, and the energy shifts of the latter components are given with respect to the C - C component. These components are assigned to carbon atoms with C-C single bonds, to carbon atoms involved in a C-In (shift of +1.54 eV) bond and to C atoms involved in a bond formation to P atoms (shifted by +0.50 eV). The shifts in the energy level are in agreement with the different electronegativity values of phosphorus, indium and carbon. The fourth small component is shifted by -1.17 eV towards lower kinetic (higher binding) energies with respect to the C-C component. This component is believed to result from carbon atoms participating in C=C double bonds of cyclopentene as observed by Liu and Hamers. 24 These observations are explained by the adsorption of cyclopentene on the 'mixed-dimer' as the main surface adsorption site and subsequently bonding to the second layer In-In bonds as discussed before and are supported by theoretical DFT calculations. 21 The resulting adsorption structures are based mainly on the analysis of the C Is core level emission lines. The interpretation and comparison of these results to the ones found for the adsorption of cyclopentene on the GaAs(001)(2 x 4) surface will help discussing the resulting adsorption structures for the latter.

4.2. Results for the adsorption of cyclopentene on the GaAs(OOl)(2 X 4) The clean (2 x 4) reconstructed GaAs (001) surface (see Fig. 4 left), two surface related components are found in the emission line shape analysis. One component AS 1 is shifted towards lower kinetic (higher binging) energies with respect to the bulk component which is not yet clearly assigned, and another component shifted towards higher kinetic (lower binding) energies are related to the As dimer atoms of the topmost layer of the surface. In Fig. 4 (middle) the As 3d core level emission line taken after the adsorption of cyclopentene is shown. An additional component (As-C) is revealed

7

binding energy I eV

binding energy I eV

Fig. 3. The C Is core level for the cyclopentene covered InP(OOl)(2 4) (left) and the cyclopentene covered GaAs(OOl)(2 x 4) (right) reconstructed surfaces. The shifts are given in eV with respect to the C - C components.

in the fit shifted towards lower kinetic (higher binding) ~"'~LE>"~U, the different electronegativities of C and As and assilgrled to a bond formation between C and As atoms of the In the Ga 3d core level emission line shape no could be found after the of cyclopentene. In 4 (right) the Ga3d core level emission line of the cyclopentene covered surface is Three comclearly be """""F>''''''''', ponellts are evaluated two of which can not the shifted towards lower kinetic (higher UUiUWL1'. wards kinetic (lower binding) is related to Ga atoms of the second In 3 side) the C Is core level emission line taken after the saturation with is shown. Before deposition, no traces of carbon could be detected by SXPS. In the core level line three comPOnell1ts are found. The first pronounced component is to C atoms

9

At the cyclopentene\GaAs(OOl)(2 x 4) surface, the C=C component stems from cyclopentene molecules that are covalently attached to the surface. In contrast, on the cyclopentene\InP(OOl )(2 x 4) surface the C=C component stems from cyclopentene molecules that are only physisorbed and not covalently attached to the surface. The different energy shifts of the respective component could possibly be explained by the different bonding mechanisms of those molecules containing the C=C bonds. The difference in the intensities of the C-C, the C=C and the C-As contribution for the adsorption of cyclopentene on the GaAs(OOl) surface shows that cyclopentene adsorbs by a dissociation of one hydrogen atom as well as by a splitting of the double bond. This is in contrast to the adsorption of cyclopentene on the InP(OOl )(2 x 4) surface where the C=C component is less pronounced with respect to the C-C component. Therefrom, it can be concluded that during the adsorption of cyclopentene on the GaAs(OOl)(2 x 4) surface, the double bond remains intact while a splitting of the double bond occurs for the adsorption of cyclopentene on the InP(OOl)(2 x 4) surface. These observations result in two different adsorption structures, as shown in Fig. 5, which reveal the influence of the different surface dimer configurations. In case of the InP(OOl)(2 x 4) surface, the asymmetric dimer configuration leads to a [2 + 2]-cycloaddition-like reaction with a formation of two bonds between cyclopentene and the surface. Contrary to this, on the GaAs(OOl)(2 x 4) surface cyclopentene forms only one bond to As dimer atoms of the surface which have a symmetric dimer arrangement. This single bond formation could involve a charge transfer to the second Ga dimer atoms, or it could also be the case that dissociated hydrogen atoms of the cyclopentene molecules bond to the second As dimer atoms, which are not directly involved in the covalent bonding to the adsorbed molecules, as indicated in Fig. 5. 5. Summary In this paper, we have shown a comparison between the adsorption configuration of cyclopentene on two different III-V surfaces, the InP(OOl)(2 x 4) and GaAs(OOl)(2 x 4) reconstructions. The dimer configurations on these surfaces are different and result in two different adsorption configurations for cyclopentene. In the case of the InP(OOl)(2 x 4) surface, two bonds to the surface are formed. This is possible due to the asymmetric arrangement of the 'mixed-dimer' atoms in the topmost layer of the surface. The underlying bonding mechanism could therefore be described by a [2 + 2]-cycloaddition reaction as observed for cyclopentene adsorbed on the Si(OOl)(2 x 1) sur-

10

Fig. 5. On the left: schematic representation of the adsorption of cyclopentene at the 'mixed-dimer' of the InP(001) (2 x 4) surface. On the right: scheme for the adsorption configuration of cyclopentene on the GaAs(001)(2 x 4) surface. The light balls represent the group-V (P As) and the dark one the group-III (In, Ga) elements.

face.2 In the case of the GaAs(001)(2 x 4) surface, the topmost layer consists of symmetric As dimers where a 'cycloaddition' reaction is not allowed. Consistently, only a single bond formation to one of the topmost As dimer atoms could be observed. Hence, we can conclude that the asymmetry of the 'mixed-dimer' of the InP(001)(2 x 4) surface supports the formation of two bonds, as found for the adsorption of cyclopentene on the Si(001) (2 x 1) surface. This is not the case for the symmetric dimer of the GaAs(001)(2x4) surface. Acknowledgments We would like to thank S. Weeke and M. Pristovsek for providing the InP substrates. We would also like to adcnc'wl;ed~[e financial support from the DAAD (Deutscher Akademischer Austauschdienst), VIGONI and the Deutsche Forschungsgemeinschaft (DFG).

References 1. 2. 3. 4. 5.

R. Hamers, Nature 412, 489 (2001) Nature 412, 489 (2001). S.F. Bent, Surf. Sci. 500, 879 (2002). Z. Liu, A. A. Yasseri, J. S. Liadsey, D. F. Bocian, Science 302, 1543 (2003). J.S, Hovis, H. Liu, R.J. Hamers, Sur. Sci. 402-404, 1 (1998) J.W. Kim, M. Carbone, M. Tallarida, J.H. Dil, K. Horn, M.P. Casaletto, R. Flammini, M.N. Piancastelli, Surf. Sci. 559, 179 (2004)

11 6. C. A. Hacker, R. J. Hamers, J. Phys. Chern. B 107 (31), 7689 (2003). 7. A. Hermann, W. G. Schmidt, F. Bechstedt, Phys. Rev. B 71,153311 (2005). 8. W. G. Schmidt, F. Fuchs, A. Hermann, K. Seino, F. Bechstedt, R. Passmann, M. Wahl, K. Hinrichs, N. Esser, S. Wang, W. Lu, J. Bernholc, J. Phys. Can. Mat. 16, S4323 (2004). 9. L. Mantese, U. Rossow, D.E. Aspnes, Appl. Surf. Sci. 107, 35 (1996). 10. R. J. Cole, S. Tanaka, P. Gerber, J. R. Power, T. Farrell, P. Weightman, Phys. Rev B 54, 13444 (1996). 11. J. R. Power, T. Farrell, P. Gerber, S. Chandola, P. Weightman, J. F. McGilp, Sur. Sci. textbf372, 83 (1997). 12. S. Wang, W. Lu, W.G. Schmidt, J. Bernholc, Phys. Rev. B68, 035329 (2003). 13. R.J. Hamers, J.S. Hovis, S. Lee, H. Liu, J. Shan J. Phys. Chern. B 101, 1489 (1997). 14. H.Q. Yang, C.X. Zhu, Z.Q. Xue, S.J. Pang, Surf. Sci. 448, 225 (2000). 15. R. Passmann, M. Kropp, T. Bruhn, B.O. Fimland, F.L. Bloom, A.C. Gossard, W. Richter, N. Esser, P. Vogt, Appl. Phys. A 87, 469 (2007) 16. W.G. Schmidt, P.H. Hahn, F. Bechstedt, N. Esser, P. Vogt, A. Wange, W. Richter, Phys. Rev. Lett. 90 (12), 126101 (2003) 17. W.G. Schmidt, N. Esser, A.M. Frisch, P. Vogt, J. Bernholc, F. Bechstedt, M. Zorn, T. Hannappel, S. Visbeck, F. Willig, W. Richter, Phy. Rev. B 61, R16335 (2000) 18. W. G. Schmidt, F. Bechstedt, J. Bernholc, Appl. Surf. Sci. 190,264 (2002). 19. K. Knorr, M.Pristovsek, U. Resch-Esser, N. Esser, M. Zorn, W. Richter, J. Chryst. Growth 170, 230 (1997). 20. R.W. Bernstein, A. Borg, H. Husby, B.O. Fimland, J.K. Grepstad, Appl. Surf. Sci. 56-58, 74 (1992) 21. R. Passmann, P. Favero, W. G. Schmidt, R. Miotto, W. Braun, W. Richter, M. Kneissl, N. Esser, P. Vogt Phys. Rev. B, submitted, (2008) 22. W.G. Schmidt, F. Bechstedt, N. Esser, M. Pristovsek, Ch. Schultz, and W. Richter, Phys. Rev. B 57, 14596 (1998). 23. C. Schultz, A.M. Frisch, K. Hinrichs, J. Kinsky, T. Hermann, U. Rossow, N. Esser, W. Richter, J. Vac. Sci. Technol. B 15, 1260 (1997). 24. H. Liu, R. J. Hamersr, Surf. Sci. 416, 354 (1998).

TIME-RESOLVED ENERGY DISPERSIVE X-RAY REFLECTOMETRY AS A TOOL FOR MATERIAL SCIENCE STUDIES: THE CASE OF ORGANIC SOLAR CELLS BARBARA PACI t , AMANDA GENEROSI, VALERIO ROSSI ALBERTINI, PAOLO PERFETTI

ISM-C.N.R, Area di Ricerca di Tor Vergata, Via del Fosso del Cavaliere 100, 00133 Rome, Italy REMI DE BETTIGNIES

Laboratoire Composants Solaires CEA INES-RDI, Savoie Technolac, BP 332, 50 avenue du lac Leman 73377 LE BOURGET DU LAC, France

CAROLE SENTEIN

CEA Sa clay, DRT-LITEN-DSEN-GENEC-L2C, F91191 Gij-sur-Yvette, France

The original Energy Dispersive X-ray Reflectivity (EDXR) technique is applied in-situ, to study the real-time morphological changes of organic photovoltaic devices in working conditions. The reported time-resolved EDXR measurements allowed the crucial requirement, in the development of organic electronics, of device morphological stability to be addressed.

t

Corresponding author: [email protected]

12

13

1. Introduction

In the development of organic devices a crucial goal is the system structural/morphological stability in working conditions. This problem is particularly important in the case of plastic photovoltaic (PV) cells, in order to bring these very promising devices to a level at which commercialisation may be a realistic prospect. At present organic solar cells, although much less efficient than silicon cells, exhibit a unique combination of interesting properties, including: low cost, flexibility and the ability to cover large surfaces. Recently, a considerable advance in the field was obtained demonstrating that ultra-fast electron transfer may be photoinduced from a donor material to fullerene [1]. Therefore, nowadays, photo-induced electron transfer from a conjugated polymer (donor) to buckminsterfullerene C60 (acceptor) is the basic mechanism utilised in polymer-based photovoltaic cells, providing a molecular approach to high efficiency photovoltaic conversion. Notable progress has been made in recent years, using either poly[2-methoxy-5-(3',7' -dimethyloctyloxy)-I,4phenylenevinylene (MDMO-PPV) [2,3] or poly(3-hexyl thiophene (P3HT) blended with methano-fullerene [6,6]-phenyl C61 -butyric acid methyl ester (PCBM). In particular, an increasing number of research groups have recently moved towards poly thiophene and its derivatives since devices based on this polymer combine good PV performances and promising stability [4-7] and efficiencies exceeding 6% were obtained [8-12]. Nevertheless, a more thorough inspection of the various mechanisms involved in the conversion of solar energy is still required for a substantial improvement in this technology. The main points which need to be addressed are the stability and lifetime of the cells, which could be investigated by observing the behaviour of devices based on alternative or simplified architectures. In this framework, the development of new characterization techniques are of extreme importance in order to better understand the changes of the various layers composing the system, upon cell working and their effect on the device's lifetime. In particular, it appears extremely useful to monitor the changes experienced by the cells morphological parameters (i.e. the thickness and roughness of the various layers), in order to reveal the occurrence of possible uncontrolled interface phenomena. Here, insitu Energy Dispersive X-ray Reflectivity technique was applied, to study the real-time morphological changes of organic photovoltaic devices in working conditions. In this way, we addressed the crucial requirement, in the development of organic electronics, of device morphological stability.

14 The set-up adopted allowed the electrical properties of the device to be directly correlated to the modification of the electrode morphological parameters (thickness and roughness), which were obtained by in situ EDXR.

2. Experimental

2.1 Sample preparation The two series of bulk heterojunction solar cells used in this article were made from a blend of methano-fullerene [6,6]-phenyl C61 butyric acid methyl ester, denoted as PCBM, respectively with and P3HT or with MDMO-PPV. The cells consisted of an indium thin oxide (ITO) substrate cleaned in an ultrasonic bath of acetone and isopropanol, rinsed in deionized water, dried in an oven and, finally, treated with UV -ozone. The active layers were deposited by spin-casting from an anhydrous chi oro benzene solution, and the devices were completed by deposition of the cathode through a shadow mask with 6 mm diameter openings. The top contact was a 100 nm thick Al layer (nominal thickness). The cells had an active surface of 32 mm2. The annealing was performed in a glove box, under controlled atmosphere «1 ppm O2 and H 20). 2.2 loint X-ray and Photo-current Set-up The experimental apparatus consisted of a non-commercial reflectometer [13] developed at ISM CNR [14] and characterized by a very simple set-up geometry, since neither monochromator nor goniometer are required in the energy dispersive mode, no movement being needed during the measurements. The main elements of the machine are an X-ray tube and an energy sensitive detector, mounted on two benches pivoting around a common central axis. Four adjustable slits are used to define the X-ray optical path. The Bremmsstrahlung of the X-ray tube (3kW power, tungsten anode) is used as a probe, and an EG&G high purity germanium solid-state detector, whose energy resolution is about 1.5-2% in the 15-50 keY energy range, accomplishes the energy scan. The measurements were performed with the device placed inside an X-ray transparent chamber under an N2 gas flux. Sample alignment was checked during the experiment to detect and keep under control possible misalignments of the sample due to heating during illumination. During the EDXR measurements, the photo-current was monitored by using a home-made acquisition software running over Labview. Measurements were carried out in short circuit conditions during illumination with a white light lamp (10 mW/cm2 ).

15

3. Results and Discussion

In the present work the time resolved Energy Dispersive X-ray Reflectometry was used to address the morphological stability of the metallic electrode and of the active layer (with particular attention to their interface), which are considered to be a critical point for improvement of organic PV devices. The study was divided in 3 steps: I.

II.

III.

Definition of the protocol for in situ ED X-ray Reflectometry measurements on organic devices Joint EDXR and photocurrent studies of photo-degradation and stabilization effects at the electrode/active layer interface, in working conditions of the cells. Time resolved morphological investigations of the active layers of efficient organic solar devices (Bulk Heterojunction films)

The use of the time resolved EDXR technique enables the changes in thickness and roughness of the layers of stratified systems to be monitored [13,15,16]. The X-ray Reflectometry is commonly utilized to probe the properties of surfaces and interfaces of layered samples, like films deposited on substrates, multilayers, superlattices etc. This technique is based on the optical properties of X-rays, whose refraction index in a material n = 1 - (t.}/2rc)pro Z,z (A = incident wavelength, p =material density, ro = classical electron radius and Z = atomic number), although very close to 1, is not exactly unitary. As a consequence, the Snell rule still applies and, at the critical angle, it can be written as [17]: cos1'}x = n. Expanding the right hand side to the second order: 1 - 1'}2J2 = 1 pA2roZ212~which corresponds to (1'}c / A) = Z(pro /rc)ll2 = constant, where 1'}c / ~7'Sin1'};y;IA oc qc (critical value of scattering parameter qc=47tsin6/~ Therefore, the variable on which the reflected intensity actually depends is not the deflection angle only but, rather, the scattering parameter q=47tsin1'}/A = (2lhc)E sin,\,} (E=radiation energy, h=Plank's constant, c=velocity of light), as in the case of the X-ray Diffraction. Hence, two ways can be utilized to perform the q-scan, namely either using a monochromatic beam and executing an angUlar scan (Angular Dispersive mode) or using a polychromatic X-ray beam at a fixed angle and carrying out an energy scan (Energy Dispersive mode).

16

Although affected by a lower resolution, the ED [13] technique has some advantages on the laboratory Angular Dispersive (AD) counterpart [18] connected to the immobility of the experimental apparatus during data collection. Indeed, in the grazing geometry required for this kind of measurements, even minimal misallignements of the sample may induce relevant relative errors during the angular scan. In particular, if many scans have to be carried out consecutively, as in the present case, reproducibility problems that may arise because of the mechanical movements of the diffractometer arms are prevented by EDXR. Finally, when laboratory sources (X-ray tubes) are used, the data collection is shorter in ED, since the number of photons concentrated in the monochromatic fluorescence lines (used as primary beam in AD) is much lower than the number of photons distributed along the white Bremmstrahlung component (primary beam in ED).

I.

Definition of the protocol for in situ ED X-ray Reflectometry measurements on organic devices:

Since the device is a multilayered system, as shown in the insert of figure 1, in order to identify the contributions of each layer to the overall X-ray reflection signal, EDXR measurements on samples that correspond to the subsequent stages of cell construction were performed. The reflection patterns of samples that correspond to such stages were compared with the patterns of the complete cell in figure 1. In particular for the measurements of the cells, an experimental procedure was used to maximize the Al signal with respect to the one coming from the other layers, in particular, from ITO. It consisted of tilting slightly the sample under the X-ray beam, by means of a rotating cradle. In this way, it was possible to assign the oscillations visible at lower scattering vectors (curve b in figure 1) to the Al film, the period of the oscillations being related to the film thickness d [19]. These preliminary measurements also allowed the determination of the total reflection edges [15,16]. At higher scattering vectors, the effect of the presence of the other layers is more evident and an interference pattern in the thickness fringes is visible. For this reason, the fit in figure 1 is limited to the lower scattering vector region, where the Al contribution is dominant. Conversely, no Kiessig fringes are visible in curve c (glasslITO sample), probably due to its severe roughness. Of particular interest is the fact that the derivative of the reflection profile fit proves an accurate determination of the total reflection edges of the X-ray

18

II.

Joint EDXR and photocurrent studies of photo-degradation and stabilization effects at the electrode/active layer interface, in working conditions of the cells.

In order to perform the present investigation, the EDXR set up was modified in order to allow a direct comparison of the structural changes with the working efficiency of the device This new set-up permits the simultaneous monitoring of the PV cell photocurrent and the morphological changes (by EDXR) in order to correlate the oxidation process with the decline of the device performances. The results of the in situ EDXR measurements, collected under a controlled N2 atmosphere and upon illumination with a white light lamp are shown in figure 2.

a)

159 ] ' 158

20

'--'

"d

157 l--+----.-~f-----l,--~--,--~---,---

1

20 .

E ~ ;:I U I

0.04

0.05

0.06

0.07

0.08

scattering parameter (AI)

B

.2

~

1.5 1.0

0.5 0.0

r-+-,-~-,~-,--~.-

o

3

6

9

12

Time (hours)

Figure 2. a) Time-resolved EDXR patterns collected on a working organic PV cell. b) morphological data points obtained as the results of data fit of the EDXR patterns and jointly measured photocurrent

The reflectivity profiles do not show any changes in the oscillation period when the sample is kept in the dark (first patterns on the bottom of the graph which are

19

shifted in height for clarity). However, a progressive compression of the Kiessig fringes, during illumination can be noticed, corresponding to an increase of the Al film thickness as a direct consequence of exposure to light. Such an aging effect may be attributed to the formation of an aluminium oxide layer at the Allorganic film interface [15], as a consequence of illumination. Indeed, the aluminium layer in the device is in contact with the oxygen ions (bounded to the polymers of the organic layer), which may be released during illumination, so that photo-induced oxidation may well occur. Moreover, samples have been stored in the dark under ambient conditions before being measured and a certain amount of humidity may have been absorbed by the organic layer. The aluminium oxidation process appears as a thickening of the Al film, because it leads to the formation of a thin layer of aluminium oxides, too thin to be detected by the EDXR technique as a separate layer. Moreover, the interface between the aluminium oxide and the polymer is likely not sharp. As a consequence, in the reflection interaction, it is felt by the X-rays as an increase of the Al bulk, rather than as an independent, well-defined layer. In the following we will discuss how the time-resolved EDXR measurements validate this hypothesis. The resulting d vs. time data points, obtained by the fit of the patterns in figure 2a, are plotted in figure 2b. The stability of the morphology in the dark was verified for a long period. The subsequent effect of illumination is visibly a two step increase in thickness. The fit of the d(t) curves was carried out using two correlated Boltzmann curves: the first curve, d(t)=d,+(d 2-d,) (l-exp-(tl't,)), describes the progressive increase of the film thickness d, from its initial value d, up to a first asymptotical value d 2 in a characteristic time 't, the second one, d(t)=d 3+(d 4-d3) (l-exp-(tJ't2)) describes a further increase in thickness, beginning when the film thickness has reached the intermediate value d3 , until a final asymptotical value d4, in a characteristic time 't2. It can be noticed that the first process is almost concluded at the onset of the second. This characteristic provides further qualitative information that clarifies the nature of such processes, and supports the hypothesis that electrode oxidation has taken place. Indeed, the Al oxidation kinetics is expected to be a two step process, where the formation of sub oxides is the precursor to the growth of a passivating layer of alumina [21]. The cr vs. t data points plotted in figure 2b show that this parameter remains unchanged during the overall process. Since surface phenomena normally produce an increase of the film surface roughness, the fact that in the present case no modification of this parameter is observed over time is a substantial clue

20

that the process is limited to the interface between the aluminium and the organic film, no variation of the surface being expected. During the EDXR measurements, the cell photocurrent was monitored and the result is also reported in figure 2b. It allows the photo-oxidation process of the electrode (observed by EDXR) to be directly correlated with the fading of the cell performances. Moreover, it is worth noticing how the time evolution of the morphological data, due to the elevated temporal sampling, are able to describe the dynamics of this process. Indeed, it is understood that the process takes place in two steps (see curve in figure 2b), while the electrical measurements, normally used to define the aging effects in PV devices (see bottom curve of figure 2b), are unable to provide this information. Therefore, this second study demonstrated a direct connection between the real time morphological changes and the decline in performance of the working device.

III.

Time resolved morphological investigation of the active layers of efficient organic solar devices (Bulk Heterojunctionfilms)

The results obtained in the previous study allowed the association of the observed morphological variation with an oxidation process of the Al electrode at the interface with the active layer. Further results on a similar cell, based on MDMO-PPV blended with PCBM as bulk heterojunction, confirmed the onset of photo-induced oxidation of the Al electrode at the buried interface [16]. The efficiency vs time curves measured for such MDMO-PPV based devices is reported in figure 3.

o

20

40

60 80 100 time (hours)

120

140

Figure 3. Efficiency vs time curves measured for a MDMO-PPV based device

21

The curve is characterized by a first part (first 15 h), well fitted by a fast exponential decay, confirming that also in this there is a fast reduction of the efficiency of the device in the first hours of working. A second part that, almost linear, indicates that a new, subsequent, processes take place. Therefore, two aging process are present: a first fast process is due to a photo-induced oxidation of the Al buried interface, previously discussed. A second slow process may by due to the aging of the active layer and is the focus of the following investigation. In order to investigate the latter process a new set of in-situ EDXR experiment, consisting of collecting a sequence of X-ray reflection patterns on a glass/ ITOIPEDOTIP3HT:PCBM multilayered system, was performed. Each spectrum was acquired over a period of 1 hour 60 minutes. The patterns were collected first in the dark (to verify that samples stored in the dark at room temperature do not show any morphological modification) and then under illumination with a white light lamp. The morphological data points (obtained by analyzing the sequences Of spectra in figure 4a) are fitted according to two Boltzmann growth functions (see figure 4b).

a)

b)

E

.s "0

100

E

.s o

o. O. scattering parameter (A")

o

20

40

60

80

Time (hours)

Figure 4. a) Time-resolved EDXR patterns collected on a P3HTIPCBM film under illumination. b) Morphological data points obtained as the results of data fit of the EDXR patterns.

22

The film response is not immediate (induction time of approximately 20 hours). The amplitudes and time constants obtained by the fits of the data points are: dd = (2.3 ± 0.1) nm and:t= (44 ± 1) hours for the bulk process; dO' = (0.15 ± 0.05) nm and 't = (44 ± 1) hours for the surface one. This structural modification was demonstrated to be due to a thermal effect, since the illuminated sample reaches about 70°C, i.e. a temperature close to the P3HT glass transition temperature. Indeed, EDXR results on an identical sample, measured in the dark but upon heating at T= 70°C for 150 hours, showed that the blend undergoes a structural modification similar to the one observed in figure 3, characterized by a monotonic increase of the film thickness [22].

4. Conclusions

In conclusion, we report an in situ study correlating the vanatlOn of the photocurrent of the device upon working to the electrode-organic film interface morphological evolution monitored by the time-resolved Energy Dispersive Xray Reflectometry technique. The study demonstrated a direct connection between the real time morphological changes and the decline in performance of the working device. This experimental approach gave several important results. First, the morphological changes in the device structure (undergoing a photo induced oxidation of the AI-organic film interface) strongly affect the response of the system. Second, the comparison of the electrical and EDXR data sets shows that, while the photo-current vs time curve reveals only the occurrence of a photo degradation phenomenon, the in-situ EDXR technique is able additionally to describe the dynamics of this effect. In particular, the observed process is limited to the buried electrode interface and is responsible for a rapid decrease in the photocurrent. Moreover, the new approach used in this work allowed the direct detection of a reorganization of the organic molecules in the device active layer. To conclude, the information obtained by the proposed X-ray method uncovers some uncontrolled mechanisms limiting the cell efficiency, stability, and lifetime and may be used for further improvements in organic electronics.

23

References

1. 2. 3. 4. 5. 6. 7. 8. 9.

10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22.

N.S. Sariciftci, L. Smilowitz, AJ. Heeger, F. Wudl, Science 258, 1474 (1992). AJ. Mozer, P. Denk, M.C. Scharber, H. Neugebauer, N.S. Sariciftci, P. Wagner, L. Lutsen, D. Vanderzande, J. Phys. Chem. B 108,5235 (2004). S.E. Shaheen, C.J. Brabec, N.S. Sariciftci, F. Padinger, T. Frornherz, J.C. Humrnelen, Appl. Phys. Lett. 78,841 (2001) c.J. Brabec, Solar Energy Materials & Solar Cells 83,273 (2004). X.N. Yang, J. Loos, S.c. Veenstra, W.J.H. Verhees, M.M. Wienk, J.M. Kroon, M.AJ. Michels, RAJ. Janssen, Nano Letters 5, 579 (2005). F.C. Krebs, H. Spanggaard, Chem. Mater. 17,5235 (2005). R de Bettignies J. Leroy, M. Firon, C. Sentein, Synt. Met. 156,510 (2006). G. Li, V. Shrotriya, J.S. Huang, Y. Yao, T. Moriarty, K. Emery, Y. Yang, Nature Materials 4, 864 (2005). Y. Kim, S. Cook, S.M. Tuladhar, S.A Choulis, J. Nelson, J.R Durrant, D.D.C. Bradley, M. Giles, I. Mcculloch, C.S. Ha, M. Ree, Nature Materials 5, 197 (2006). J. Peet, J. Y. Kim, N.E. Coates, W. L. Ma, D. Moses, A J. Heeger, G.c. Bazan, Nature Mater. 15, 1617 (2007). K. Kim, J. Liu, M.A.G. Namboothiry, D.L. Caroll, Appl. Phys. Lett. 90, 163511 (2007). W. Ma, C. Yang, X. Gong, K. Lee, A J. Heeger, Adv. Funct. Mater. 15 (10), 1617 (2005). V. Rossi Albertini, B. Paci, A Generosi, J. Phys. D-Appl. Phys. 39, 461 (2006). R Felici, F. Cilloco, R Caminiti, C. Sadun, V. Rossi Italian Patent No. RM 93 A 000410, 1993. B. Paci, A Generosi, V. Rossi Albertini, P. Perfetti, R de Bettignies, M. Firon, J. Leroy, C. Sentein, Appl. Phys. Lett. 87, 194110 (2005). B. Paci, A Generosi, V. Rossi Albertini, P. Perfetti, R de Bettignies, M. Firon, J. Leroy, C. Sentein, Appl. Phys. Lett. 89,043507 (2006). R.W. James, The Optical Principles of the Diffraction of X-ray (OX BOW Press, Woodbridge, Connecticut, 1982). R Caminiti, V. Rossi Albertini, Int. Rew. Phys. Chem. 18,263 (1999). S.K. Sinha, E.B. Sirota, S. Garoff, H.B. Stanley, Phys. Rev. B 38, 2297 (1988). L. G. Parrat, Phys. Rev. 95,359 (1954). G. Faraci, S. La Rosa, A R Pennisi, Y. Hwu, G. Margaritondo, Phys. Rev. B 47, 4052 (1993). B. Paci,AGenerosi,V.RossiAlbertini, RGenerosi, P.Perfetti,Rde Bettignies, C.Sentein J. Physical Chemistry C, 112 (26), 9931 (2008).

NOVEL OXIDATION PROCESS ON CHIRAL SILICON NANOWIRES' PAOLA DEPADOVA\ CNR-ISM, via del Fossa del Cavaliere, 00133 Roma, Italy BERNARD AUFRAY, GUY LE LAY CINaM-CNRS, Campus de Luminy, Case 913, 13288 Marseille Cedex 9, France CLAUDIO QUARESIMA, PAOLO PERFETII CNR-ISM, via del Fossa del Cavaliere, 00133 Roma, Italy BRUNO OLIVIERI CNR-ISAC, via del Fossa del Cavaliere, 00133 Roma, Italy In this paper we report a review of the growth and oxidation by self-organization of straight, massively parallel silicon nanowires having a width of 1.6 run. These silicon nanowires, which display a strong symmetry breaking across their widths with two chiral species that self-assemble in large left-handed and right-handed magnetic-like domains, are atomically perfect, and highly-metallic conductors. We show that the oxidation process starts at the Si NW terminations and develops like a burning match. While the spectroscopic signatures on the virgin, metallic part are unaltered, we identify four new oxidation states on the oxidized part, which shows a gap opening, thus revealing the formation of a transverse internal nano-junction.

1.

Introduction

Silicon nanowires (SiNWs) are one of the most stimulating structures In nanoscience due to the central function played by Si in the world of semiconductor industry, where the Si NWs could be the building blocks of many functional nanoscale electronic devices. In this frame, Si NWs-based transistors have been already demonstrated [1]. Likewise, the oxidation of silicon has been an issue of dominant importance in many areas of physics and technology since the interface between silicon and silicon dioxide plays a crucial

• This work is supported by the project "Self-assembled silicon nanowires: Tayloring of their structural, electronic and magnetic properties" of the International Collaboration between CNR and CNRS 2008-2009.

24

25

role in microelectronics devices. With the possibility to undergo towards the manufacturing of nanostructures, the role played by this interface becomes prevalent. The potentiality of self-assembly used to organize spontaneously lowdimensional materials such as quantum dots and nanowires, has been successfully applied recently to the formation of self-aligned straight SiNWs on the silver (110) surface [2, 3]. Room temperature (RT) deposition of a low coverage of Si on the Ag(l1O) surface produced massively parallel, atomically perfect and highly metallic SiNWs sharing a common width of just ~ 16 A, as shown by scanning tunneling microscopy (STM) measurements [2,3]. Surprisingly, these silicon nanowires display a strong symmetry breaking across their widths with two chiral species that self-assemble in large left-handed and right-handed magnetic-like domains. The oxidation process of these silicon nanowires starts at the Si NW terminations and develops like a burning match. High-resolution (HR) photoemission spectroscopy of SiNWs shows the narrowest Si 2p core levels ever measured in solid phase. We determine for the first time, the different life times of the Si 2p 1/2 and Si 2p 312 core holes resulting from a Coster-Kronig transition forbidden in semiconductors, but allowed in metals. In excellent agreement with STM measurements, the HR photoemission spectroscopy shows the presence of all components related to the virgin SiNWs, Si 0, in addition to the oxidation components, Si 1+, Si 2+, Si 3+, and Si 4+ on the features of the Si 2p core-levels. Scanning tunneling spectroscopy (STS) measurements indicate a metallic behavior on the virgin part of the SiNWs and a gap opening of - 0.35 V on the oxidized part, revealing the formation of an internal nano-junction [4].

2. Experimental The STM observations of the SiNWs were carried out at the CINaM-CNRS in Marseille, while the photoemission experiments at the VUV beam line of the Italian synchrotron radiation facility ELETTRA in Trieste. In both laboratories the same procedure has been used for sample preparation, silicon evaporation and Si oxidation. The Ag(l1O) substrate was cleaned in the UHV chamber (base pressure: 8.5 10- 11 mbar) by repeatedly sputtering with Ar+ ions and annealing the substrate at 750 K, while keeping the pressure below 2xlO- 1O mbar during the heating. Si was evaporated at a rate of ~ 0.03 MLimin from a Si source, while the Ag substrate was kept at room temperature (RT). This condition has been shown to give definite LEED patterns, where, in addition to the sharp integer order spots of the unreconstructed Ag( 11 0) surface, thin streaks elongated along the [100]* reciprocal direction, develop through the substrate spots as well

26

as in half-order position along the perpendicular [-110]* direction. The silicon coverage was measured using a quartz microbalance; the error in determining the Si amount was estimated to be less than 10%. The sample temperature was assessed by an infrared pyrometer. The oxidation was obtained upon exposing the so-formed Si NWs at several increasing total doses expressed in Langmuir (1 -6 L = I x 10 Torr per Is) 10, 20, 40, 80, and 300 L of pure molecular oxygen (99.999 %) also at RT. All STM images presented in the following were recorded at RT in constant-current mode at a bias voltage of -1.7 V and -1.8 V for the clean and oxidized SiNWs respectively, and a tunneling current of 1.2

nA. For clean SiNWs the photoemission spectra were recorded after cooling the sample to ~ 150 K. They were acquired using an angle resolved electron energy analyzer with an acceptance of 2°. For core level measurements, the angle between the photon beam and the normal to sample surface was 17°; the collection angle {} = 0° corresponds to normal emission. The photon energy was set to 135.8 eV. For oxidized SiNWs the photoemission spectra were acquired using an electron energy analyzer with an acceptance of 8°, set at an emission angle of 8 = 45°. The angle between the photon beam and the normal to sample surface was also 45°. The photon energy was 132.5 eV. The total energy resolution for all photoemission spectra was better than 50 meV.

3.

Results and Discussion

Figure I displays a 10.2 x 10.2 nm 2 filled-states STM image of - 0.5 ML Si deposited at RT on the Ag(llO) surface. On this image the atomic resolution obtained on the bare Ag( II 0) surface allows to determine directly the x2 periodicity along the edges of the SiNWs parallel to the Ag direction (i.e., 2aAg[_ IlOj). It can be further noticed a misalignment between the right and left side protrusions corresponding to a glide of one aAg[-1I0j lattice parameter. The line profile along the perpendicular [100] direction assigns a lateral size of ~ 1.6 nm for every single SiNW, i.e. an x4 width of 4aAg[lOO]' where aAg[IOOj = 0.409 nm, while the height is about 0.2 nm. It is important to remark that the nanostructures are never terminated along the [100] Ag direction (i.e., perpendicularly to the direction of the SiNWs), but along another definite orientation, which we assign to the [3-34] direction of the Ag(llO) surface. This fact confers to NWs, a surprising asymmetric morphology with respect to the plane perpendicular to the (110) surface.

27

Figure 1. 10.2 x 10.2 mn2 filled-states STM image. The line profiles indicatc the 2x and 'Ix periodicity respectively parallel and perpendicular to the H 10] direction. Adapted from Fig.2 of Ref. 3.

syrmnetrJic transverse shape of the SiNWs is evidenced on the rel)Om:a in 2 of the previous STM image. A dip that determines transverse shape of the SiNWs, is clearly noticeable on the side of each NW. The line profile reveals the presence of a small slanted (l atomic facet oriented at 135° with respect to the (11 0) surface. in all the dip is always on the right hand side of the NW s. we stress that on other extended areas we met the reverse situation where the situated on their left side and the nanosrructures are direction, symmetric to the .....~·v,,\l1 one. the two types of symmetry breaking NWs domains of opposite "spins" to recover the overall '''rrn".,,,.tr'V of the bare surface. The "phase" condensation in these two ~fmHnrte domains implies a cross talk between NWs the same is confirmed by perusal at 2, where one can notice an xl on the bare silver areas, as along the NWs themselves. This to

28 a strain-induced modulation of the substrate extending along the NWs and propagating to a certain distance perpendicularly. This "mattress" effect induces the cross-talk and favors the formation of separate domains because of the along the [-110] direction mentioned above.

2

Figure 2. 3D view of 10.2 x 10.2 nm filled-states STM image: dip asymmetry at right- side. from Fig. 3a of Ref.3.

3 displays high resolution Si 2p core-level spectra measured at t} 0° , .. ~ ...._. emission) of 0.5 ML Si deposited on Ag(llO) surface at RT. The Si is essentially composed of four (spin-orbit splitted) components, and attributed to different Si atoms located at the SiNWs2 • In Si core level collected at t} 0° is completely dominated by the COI1rroonerlt. Its full width at half maximum (FWHM) is only 130 meV. This is line shape ever reported in a solid phase photoemission the narrowest Si eXl,enlmemt, being narrower than the bulk line in a silicon single crystal. This result shows that the Si NWs are atomically precise objects and that the whole ensemble at macroscopic scale is practically free of defects.

29

Figure 3. Normal ({l=OO) emission convoluted Si 2p core levels for SiNWs aftcr the deposition of 0.5 ML ofSi on Ag(llO). Adapted from 4 of Ref. 3.

result arises from the fit of the Si 2p split (D-S) functions, which imposes two 25 meV different values of the Lorentzian FWHMs rl/2 == 40 meVand for the Si 2p3/2 and Si 2p1l2 lines, which have a common Gaussian FWHM of just 95 meV. The evidence of the extra broadening of the Lorentzian FWHM, points to a non-radiative Coster-Kronig LLV U
30

extremities

the

10] direction maintaining an atomic structure pre:terably

11 ..,'''I ...."F·11 along the one-dimension of the wires. This is clearly observed in the

4b at 30 L and at higher oxygen exposures. The oxidation behaves ac(:orlding to a match-burst process, where the extremities of the Si NWs are considered as the head of a match, which reacts with the oxygen atoms. This behavior that the termination-sides of the Si NWs several reactive able to be rapidly saturated by the oxygen atoms. In this way the oxidation propagates along the [-110] direction, like a flame which is the which the clean and the oxidized of the Si NWS. 15L

d

~

Hollow

30L

22runx22nm

~ '<;i

~

Relative Binding Energy (eV)

4. Filled-states STM images ofSi NWs on Ag(llO) at different oxygen doses: (a) 22 x 22 2 at 15 L. (b) 27.3 x 27.3 nm at 30 L. Convoluted Si 2p core levels of of SiNWs grown on Ag(IIO) at: (c) 20 L and (d) 40 L. The colored components , related to the different oxidation states (+ I te +4) grow at the expense of the four initial components (Sf, S2, SJ, S4) of the clean SiNWs. Adapted from Fig. la and lb, and Fig.2 of Ref.4.

These characteristic features discovered through the HR STM observations of the oxidized Si NWs find their corresponding signatures in measurements. 4c and 4d the Si core-level on the Si after 20 and 40 L of exposure. Noticeably, the is strongly modified during the 02 exposure at u ...... "'.""u.•F. that the Si The of the Si NWs on 10) reveals the four UV"UHOLi>, and and the oxidized part at higher binding energy further four new

31

and them to the +1,

located at +0.95; +1.71; +2.42 and +3.85 eV. We +3 and +4 oxidation states of Si.

14.3 run x 14.3 nrn

Figure 5. Scanning tunneling microscopy ofSi NWs exposed to 30 L O 2 and I-V curves. (a) I-Von clean Si NWs. Selected areas on the STM image indicate where the I(V) curves are collected. (b) 2 I-V characteristics of oxidized Si NWs. (c) 14.3 x 14.3 mo filled-states image. Adapted from Fig. 4 of Ref. 4.

32

We obtained another remarkable result by STS measurements. The I(V) spectra reported in Figure 5a and 5b were measured respectively on both clean and oxidized parts of the Si NWs. Figure 5c is the STM where the I(V) curves were collected. On the virgin part, the metallic character is demonstrated by the I(V) spectra with high currents in the nA regime. On the contrary, the I(V) curve acquired on the oxide parts shows a semiconducting behavior revealing a gap of 0.35 V and smaller tunneling currents in the pA regime. We are in presence of a formation of a transverse internal junction between the clean and the oxidized S i NWs parts, along the [-110] direction (i.e. along the nanowires), which opens up interesting perspectives for future functional nanowire devices at the nanoscale. 4.

Conclusions

In conclusion, we have grown at room temperature straight, atomically perfect, and highly metallic SiNWs on the Ag(llO) surface. They display a clear transverse symmetry breaking with two chiral species that, surprisingly, selfassemble in large left-handed and right-handed, "magnetic"-like domains. During the oxidation of the Si NWs, a very peculiar process takes place along the lengths of the wires, similar to a propagating flame front. All oxidation states (+ 1, +2, +3 and +4) are present, which is reflected by four oxidation components, S 1+, S 2+, S 3+ and S 4+ on the Si 2p core level spectra in addition to those related to the still virgin part of the Si NWs. Initially, the oxidation sites are localized at the extremities of the Si NWs. Subsequently, at increasing O2 doses, they move along the [-110] direction: the oxidation process develops like a burning match. Tunneling spectroscopy measurements confIrm the transition from a metallic behavior of the virgin Si NWs to a semiconducting one upon oxidation, with just a small gap because of the extreme thinness. Acknowledgments

The authors thank Peixin Hu for his help in the STM images processing. The fInancial support of the International Collaboration between CNR and CNRS 2008-2009 through the project "Self-assembled silicon nanowires: Tailoring of their structural, electronic and magnetic properties" is greatly acknowledge.

References

1. R. S. Friedman, M. C. McAlpine, D. S. Ricketts, D. Ham, C. M. Lieber, Nature 434, 1085 (2005).

33

2. C. Leandri, G. Le Lay, B. Aufray, C. Girardeaux, C. 1. Avila, M. E. Davila, M. C. Asensio, C. Ottaviani, A. Cricenti, Surf Sci. Lett. 574, L9 (2005). 3. P. De Padova, C. Quaresima, P. Perfetti, B. Olivieri, C. Leandri, B. Aufray, S. Vizzini, G. Le Lay, Nano Lett. 8,271 (2008). 4. P. De Padova, C. Leandri, S. Vizzini, C. Quaresima, P. Perfetti, B. Olivieri, H. Oughaddou B. Aufray, G. and Le Lay, Nano Lett. 8, 2299 (2008).

FROM BECQUEREL TO NANOTECHNOLOGY' G. MARGARITONDO Ecole Polytechnique Federaie de Lausanne CH-1015 Lausanne, Switzerland The IOOth anniversary of Henri Becquerel's death in 2008 is an opportunity to analyze the evolution of scientific dissemination and technology transfer. The facts are shocking: both were much faster and effective at the time of Becquerel. I believe that these dismal failures are primarily rooted in academic and industrial management - and difficult to reverse.

Research conducted for an article l commemorating the lOOth anniversary of the death of Antoine-Henri Becquerel (Fig. 1) led me to discover some facts: at Becquerel's time, scientific dissemination was much faster and effective than Similarly, fundamental discoveries became practical applications more rapidly and efficiently than today. Here I go beyond historical narration to discuss what went wrong with our science management.

Figure I. Antoine-Henri Becquerel as a young student at the Ecole Polytechnque, a mature researcher and an elderly man.

discovery of radioactivity took place2•8 on March 1st, 1896 in Paris - and its very effective dissemination started9 within one The were triggered by a public discussion only 37 before the 111c.~r"'·"',",T __ and had started a couple of weeks before. The framework for these events was the French Academie des Sciences. Each Monday, the Academy held a meeting enabling its members to their latest results, news and speculations. The results were then disseminated all over by an excellent communication network with other scientific institutions - and published in the Academy journal, the "Comptes Rendus" . • This work was supported by the Fonds National Suisse de la Recherche Scientifique and by the EPFL.

34

35

Reciprocally, other scientific institutions communicated through the network and beyond national barriers their latest news, that were presented at the Monday meetings of the Academy. The international communication network and the weekly Academy meetings had a decisive role in Becquerel's work. Presenting the latest results at such meetings was a formidable challenge: in a few days (or hours), the author was forced to digest the data and face a superlatively qualified audience. The event that triggered Becquerel's work happened lo on November 8, 1895 in Wiirzburg: Wilhelm Rontgen discovers x-rays and immediately uses them for radiology. On Monday, January 20, 1896, Henri Poincare announces it to the Science Academy and an exciting discussion follows: what causes the mysterious rays? Poincare notes that in Rontgen's tube the x-rays seem to originate from the same place as fluorescence -- and argues that their emission could be somewhat related to fluorescence stimulated by illumination with visib Ie light. II We know today that this is wrong but at that time the idea was not implausible. Becquerel is very interested in Poincare's hypothesis since he is an expert in fluorescence - and needs a top-level result to definitely establish his independent identity with respect to his grandfather Antoine-Cesar and father Alexandre-Edmond, both outstanding scientists. His recent nomination as professor at the Ecole Poly technique in Paris - his alma mather - had stirred a controversy and opposition of the French Academy of Science president Alfred Cornu. Becquerel is not alone in the race to check Poicare's idea: in France, Charles Henri and Gaston Henri Niewenglowski start their own tests; in London, Sylvanus P. Thompson is experimenting with uranium salts, - the most promising testing ground. Becquerel has some excellent uranium compounds specimens, but he has lent to a colleague the only stable one, S04(UO)K. After getting it back, he rapidly obtains what he (wrongly) considers positive tests of Poincare's hypothesis. He wraps a photographic plate with heavy black paper to prevent accidental exposures to visible light. Then, he places the uranium salt on the wrapped plate. To stimulate phosphorescence and - hopefully - x-ray emission, he exposes everything to sunlight for long periods of time. He then observes in the developed plate a clear image of the salt. After some corroborating tests, Bec~uerel presents the results at the Academy Monday meeting of February 24.1 But he now needs final validation: the next deadline is the meeting of Monday, March 2 nd (1896 is a leap year). He prepares a new test by inserting between the salt and the plate a cross-shaped copper sheet to obtain a well-defined shadow in the image. However, the clouds prevail in the next days: he can only obtain intermitting exposures to sunlight and unreliable exposures to weak ambient illumination. Frustrated, Becquerel interrupts the tests and stores everything in a dark drawer.

36

The clouds continue until Sunday t: Becquerel cannot get a good sunlight exposure in time for the Monday meeting. Thus, he prefers to start a new test by replacing the partially exposed plate with a fresh one. But he does not throw it away: he develops it! Why? He probably hopes that even the erratic sunlight exposures produced a faint image to present at the Monday meeting. The reality is different and astonishing: the image (Fig. 2) is not weak but comparable to the result of a long sunlight exposure! This is disconcerting, and it would be tempting to discard the image as a freak accident. But Becquerel does not cede to temptation and reaches a clear conclusion: illumination is not needed to emit the mysterious invisible radiation that produces the images!

Figure 2. The image that revealed radioactivity: the shape of the uranium salt specimen is clearly visible with a shadow created by a copper cross. The note handwritten by Becquerel himself says: "Uranyl and potassium double sulfate - Black paper - Thin copper cross - Exposed to the sun on the 27 and to diffuse light on the 26 - Developed on March 1"'''.

Within a few hours, he presents9 the new result to the Academy (Fig. 3) and the world. This is a timely announcement: Sylvanus P. Thompson later claims 13 - with no supporting evidence - the independent discover of radioactivity (in his but after Becquerel's announcement he terms, "hyperphosphorescence") abandons his experiments. In the subsequent years, the initially moderate interest in radioactivity is boosted by Marie and Pierre Curie's 1898 discovery of radium. 14 Becquerel collaborates with the Curies, exchanging samples, results and ideas and socializing with them overcoming the barriers that separate a faculty member of the Ecole Polytechnique from Pierre Curie, professor in a minor school, and his woman partner of modest foreign origin. Together, the Curies and Becquerel discover the physiological effects of radioactivity. He gets a bum from the accidental exposure to a radium sample borrowed from the Curies and left in his vest pocket. The Curies voluntarily experiment with their own bodies! The findings are presented in 1901 in a joint article. ls

37

In the same year, the Saint-Louis Hospital in Paris pioneers radiotherapy a stunningly rapid transition from discovery to practical applications. Only twelve years later, after the deaths of Becquerel and Pierre Curie, Marie Curie inaugurates the famous Radium Institute in Paris. » J'insisterai partictllierement sur Ie fait suivant, qui me parait tout Ii fait important et en dehors des phenomenes que I'on pouvait s'attendre aobserver: Les memes lamelles crislallincs, plac\:'es en regard de plaques photographiques, dans les m6mes conditions et au travers des memes ecrans; mais a l'abri de I' excitation des radiations incidentes at main leu ties a l'ohscurite procillisent encor'c les memes impressions photographiques. Void comment j'ai et6 conduit a faire cette observation: Parmi les experiences qui precedent, quelques-tines avaient etc prcparees Ie mercredi 26 at Ie jeudi 27 fevrier ct, comme ces jotlrs-lil, Ie soleil ne s'est montl'c que d'une maniere intermittcntc, j':\Yais conserve les experiences toutes pl'Cparees et rentre les ,chassis 1\ l'ohscurite dans Ie tiroir d'un meuble, en laissant en place les lamelles
Figure 3. The words of Henri Becquerel published in the Comptes Rendus of the French Academy of Sciences (Ref. 9) that announce the discovery of radioactivity,

Nine years before, Becquerel had shared with the Curies the Nobel Prize in physics "in recognition of the extraordinary services he has rendered by his discovery of spontaneous radioactivity". "Becquerel" became the SI unit of radioactivity, streets and institutions were named after him and his image appeared in postage stamps. This was well deserved since radioactivity marked the transition from determinism to the statistical foundations of science leading to the uncertainly principle, quantum mechanics, atomic physics, chemistry and materials science - in fact, to most of modem physics. The above facts show that the dissemination of Becquerel's discovery was much faster than it would have been today. Today, the "safe" method of presenting it would have been publication in a prestigious and "fast" journal like Nature, Science or Physical Review Letters. Dissemination, however, would have started after several weeks or months rather than in a few hours - or even more, considering the problems of peer review. And Becquerel would not have been helped by alternative methods like Web publication and open access that are still quite ineffective. Was fast dissemination possible at Becquerel's time because the scientific community was small? Yes, but the rapidity and effectiveness of our new

38

communication technologies should more than offset this factor. The reasons of our failure must be found elsewhere. I believe that the key factor is the mixing of two different and partially conflicting objectives: dissemination and the quality evaluation of scientists. For this second task, the large number of individuals forces a progressive replacement of the painstaking work of individual assessment by computerassisted "bibliometric" evaluations. This in tum triggers the quantitative ranking of journals and articles. The journals are driven to non-scientific acceptance criteria such as the likelihood that the paper will be cited - whose implementation is slow and a major factor in our dissemination failure. Rapid dissemination was possible in Becquerel's time because the time sequence was reversed: the scientist's quality evaluation occurred first and recognized by membership in scientific academies -- with the opportunity of very fast dissemination of the scientist's subsequent results. The exclusion of newcomers was of course a potential problem. But the system granted generalized fast dissemination via academy members under their responsibility. Journal ranking and bibliometry were thus irrelevant. This system could not work today because the scientist's quality assessment impacts the resource allocation. Due to the pressures on the peer review system, the funding proposals of established scientist are scrutinized less severely than those of newcomers - giving the former an unfair advantage. To offset this inequality, rapid quality evaluations of junior scientists are needed, leading to the temptation of "quantitative" assessment based on bibliometry. This bad phenomenon is worsened by the use of the quality evaluations for promotions and candidate selections. Ideally, the second of the above two objectives should be a secondary goal for the journal leaders - or no goal at all. But it is becoming instead increasingly important, jeopardizing dissemination, the primary objective of. This leads to a "beauty context" in which the personal stature is more important than the real quality of work. Could all this be reversed? Regrettably, I cannot be optimistic. We would need a revolutionary change in our "science society" -- unlikely unless financial problems trigger a collapse of the entire journal system. Similar problems affect technology transfer. At Becquerel's time, new technologies linked to radiology or radioactivity arrived in a vacuum and rapidly led to new products and enterprises. Things are different today: most new technologies must compete with preexisting ones. Its adoption thus implies expensive retooling and reorganization: innovation is automatically welcome only if it results from corporate planning. Our science and technology system is thus less innovative than one century ago. These conclusions from the historical analysis of Becquerel's discovery are disappointing but inescapable. They cannot justify optimism. Real change can only occur with the emergence of a new social and professional conscience among scientists. Is this likely or even possible? Only time can give the answer.

39 References 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. II. 12. 13. 14. 15.

G. Margaritondo, Physics World 21, 26 (2008). A. Allisy, Radiation Protection Dosimetry 68, 3 (1996). L. Badash, Physics Today 21 (1996)

N. S. Kipnis, Phys. Perpect. 2, 63 (2000). E. Segre : "From X rays to Quarks" (Freeman, San Francisco 1980). M. F. L' Annunziata: "Radioactivity: Introduction and History" (Elsevier, Amsterdam 2007). M. Barquins, Bulletin de I'Union des Physiciens 95, 3 (2001). J-L. Basdevant, La Jaune et la Rouge (Journal of the Alumni of the Ecole Poly technique) , 1997. H. Becquerel, Comptes Rendus Academie des Sciences 122, 501 (1896). W. C. Rontgen, Sitzungsberichte der Wiirzburger Physik.-Medic.Gesellschaft. (1898). H. Poincare, letter to W. C. Rontgen (July 1896) and Revue Genf!rale des Sciences Pures et App/iquees 7,52 (1896). H. Becquerel, Comptes Rendus A cademie des Sciences 122, 420 (1896). S. P. Thompson, Phil. Mag. 42 ,103 (1896). P. Curie, Mme. P. Curie and G. Bemont, Comptes Rendus Academie des Sciences 127, 1215 (1898). P. Curie and H. Becquerel, Comptes Rendus Academie des Sciences 132, 1289 (1901).

HIGH-RESOLVED FLUORESCENCE IMAGING OF X-RAY MICRO-RADIOGRAPIDES ON NOVEL LiF DETECTORS F. BONFIGLI a, S. ALMA YIY A a, G. BALDACCHINI a, F. FLORA a, A. LAIa, R.M. MONTEREALI", M.A. YINCENTI a, A. CRICENTI b, C. OLIYA b, A. USTIONE b, A. FAENOYc, T. PIKUZ<, L. REALEd, P. GAUDIO e, S. MARTELLUCCI e, M. RICHETTN

"CR. ENEA, Department of Physical Technology and New Materials, Enrico Fermi 45, 00044 Frascati, Italy bIstituto di Struttura della Materia, CNR, Via Fosso del Cavaliere 100, 00133 Rome, Italy. cKansai Photon Science Institute, Japan Atomic Energy Agency, Kizugawa-cify, Kyoto 619-0215, Japan and Joint Institute for High Temperatures Russian A cademy of Sciences, Moscow, 125412, Russia. dUniversity of L 'Aquila e INFN, Dipartimento di Fisica, Coppito, L 'Aquila, Italy eUniversita di Roma Tor Vergata, Dipartimento Ingegneria dell'Impresa, Via del Politecnico 1, 00133 Rome

Abstract A novel X-ray imaging detector based on optical reading of photoluminescent color centers in lithium fluoride, LiF, is presented. Its main characteristics - i.e. high spatial resolution on a large field of view, wide dynamic range, versatility and simplicity of use - make it a high performance imaging detector for applications in X-ray microscopy, photonic devices, Extreme UltraViolet (EUY) lithography, and materials science, as well as in the characterization of intense X-ray sources, including Free Electron Laser (FEL). The peculiarities of the LiF imaging detector overcome some of the limitations of other commonly used ones, and can be exploited for X-ray microscopy in very simple configurations, such as lensless techniques, even for in vivo investigations of biological samples. Advanced optical microscopy techniques have been used to obtain highly--resolved microradiographies of biological specimens, performed in absorption contrast mode. Its peculiarities seem suitable also for use in phase-contrast experiments. The LiF-based detector versatility allows improvements in order to optimize its response and sensitivity.

INTRODUCTION

The development of X-ray imaging technologies plays a crucial role in understanding the microscopic world. It relies on the improvement of three key objects: X-ray sources, optics and detectors. In X-ray imaging experiments, in order to overcome the spatial resolution limit of the actual detectors, which is of the order of micrometers, sophisticated techniques based on image magnification (using Fresnel Zone Plate (FZP) or zooming apparatus) are currently used. The use ofFZP optics requires monochromatic radiation, with a consequent reduction of the photon flux on the samples. This means long exposure times, with the need of cooling at cryogenic temperature in the case of biological specimens, making impossible dynamic studies and in-vivo observations. Lensless X-ray microscopy has several advantages: it is simple as it does not need optics and the

40

41

spatial resolution is essentially limited by the cooperation of only three factors: the source size, the diffraction effects and the detector resolution. With the available laboratory X-rays sources, in contact mode, i.e. with the detector placed at very short distance from the sample, the resolution is essentially dictated by the detector performances. Moreover, in the case of intense soft Xray sources, like laser plasma sources, in vivo biological samples in their natural environment without any preliminary treatment necessary in electron microscopy can be directly observed. The peculiar characteristics of the LiF-based imaging detector, especially its high spatial resolution, should be particularly suitable for X-ray lensless imaging techniques in absorption and phase contrast modes. LiF is a very promising radiation-sensitive material which is widely used in radiation dosimetry, optoelectronics and integrated optics. It has also been deeply studied as far as basic optical properties of color centers (CCs) are concerned [1]. It is well known that ionizing radiation, such as charged particles (ions and electrons) and energetic photons (X-rays and y-rays), can efficiently generate optically active aggregate CCs in LiF that are stable at room temperature (RT). Significant results have been obtained in the realization of miniaturized active optical devices based on CCs in LiF produced by using low-energy electronbeam lithography, like active waveguides [2], microcavities [3], point light sources [4] and optical memories. In recent years, the use of a LiF-based imaging detector for X-ray microscopy of biological specimens, microstructures and micro-devices investigation, light-emitting patterns transfer and characterization of intense X-ray sources has been proposed and tested [5,6]. Very recently its use has been extended in the hard X-ray energy range, up to about 10 keY [7]. The opportunity to perform images in the EUV, soft and hard X-ray spectral range (photon energies 20 eV - 10 keY), with a very high spatial resolution on a wide field of view, is considered a topical task nowadays. Indeed, the fast development of different types of intense laboratory EUV and soft X-ray sources (laser plasma produced (LPP) sources, X-ray lasers) as well as large scale facilities (synchrotrons, FELs) makes their application very attractive in physics as well as in life sciences. In particular in the (2.2-4.4) nm wavelength interval, known as water-window, water is much less absorbing than organic compounds, such as proteins and carbohydrates, containing carbon. Thus images can be obtained of specimens containing water with a natural contrast [8]. The LiF detector was successfully tested by using several X-ray sources (ELETTRA synchrotron [9], LPP sources [5, 6, 10], capillary discharge lasers [11], table-top soft [12, 13] and hard X-ray sources [7]) for X-ray microscopy of biological samples and for the recording of light-emitting micro and nanopatterns. Several peculiar features of the LiF thin-layer based detector, like very high spatial resolution over a large field of view, wide dynamic range, versatility and simplicity of use, sensitivity of the optical reading technique make it a very promising and attractive imaging plate for X-ray microscopy in the field of

42

material science, characterization of intense X-ray sources and biological investigation, even for in vivo specimens. LlF BASED X-RAY IMAGING DETECTOR Among alkali halides, LiF has several peculiar characteristics. CCs in LiF crystals and films are stable and efficiently emit light in the visible and near infrared spectral ranges, even at RT, under optical excitation. Among CCs produced by ionizing radiation in LiF, we focus our attention on F2 and F3 + defects, which consist of two electrons bound to two and three close anion vacancies, respectively. These centers have almost overlapping absorption bands (M band) centered at around 450 nm [14] and, therefore, can be simultaneously excited with a single pump wavelength in the blue spectral interval. On the other hand, they exhibit two different broad emission bands in the green (F3 +) and red (F2) spectral ranges [14]. Soft X-rays and EUV light are particularly attractive for localized coloration of LiF material. The short wavelength, the neutrality (that avoids charge effect on insulating materials), the low scattering cross section (that limits the lateral spreading of the beam) and the short penetration depth (that limits the photoelectron blurring effects) allow to produce high resolved fluorescence images based on CCs in LiF. Fig. 1 shows a scheme of the LiF utilization as imaging detector in a X-ray contact microscopy experiment. After X-ray exposure of the sample, placed in contact with the LiF surface (Fig.la), the image is stored in the radiationsensitive material and can be read just by illuminating the detector with a blue light and observing the visible photoluminescence signal of F3 + and F2 with an optical microscope operating in fluorescence mode (Fig. Ib». The intensity of the photoluminescence signal is locally proportional to the X-ray transparency of the specimen placed in contact with LiF surface during the X-ray exposure. It is important to stress that the readout process of the LiF-plate, based on the detection of a photoluminescence signal, is particularly simple and efficient. With appropriate laser excitation sources and time-resolved detection, the luminescence sensitivity maybe virtually unlimited: under suitable conditions even single luminescence center present in the sample can be detected. The intrinsic spatial resolution of the LiF-based imaging plate is related to the CC dimensions, that are at atomic scale (- 1 nm), but in practice it is limited by the optical microscope and technique utilized as detection system. By using advanced fluorescence optical microscopes as readout instruments, like Confocal Laser Scanning Microscope (CLSM) and Scanning Near field Optical Microscope (SNOM), optical fluorescence images with sub-micrometric and nanometric spatial resolution can be respectively obtained [15, 16]. After the X-ray irradiation process, the LiF-based detector does not need any development procedure, it is insensitive to ambient light exposure and it stores

43

stable of the sample for very long time (many years or more), unless heated at high temperature (>400°C). ~

Figure 1. Scheme process of liF-based

~

irmdiation process in contact mode configuration (a) and of the readout detector (b)

of Olea europaea pollen have been ......,·F".·.....".'" the sample on the surface of LiF detector and then in the vacuum of Tor chamber of a Nd- Y AG laser plasma source developed at the The LiP-based radiography of the pollen grains has been studied '--''-''~J.YL. model Nikon 80i-Cl and by a SNOM by CNR-ISM 2 shows the X-ray radiography stored in a LiF detector of an Olea europaea pollen grain observed by the CLSM in with fluorescence mode with an objective 60x immersed in oil and (A :::: 458 nm). an laser

Fig. 2. Confocal laser-scanning microscopy fluorescence image of a X-ray radiography of Olea europaea (var. ascalana) pollen grain stored on a liF crystal.

44

The pollen grain X-ray radiography shows strong fluorescence with a fluorescent variations and presents an irregular shape, a not sharp-cut detector of crown. Figure 3 shows the X-ray radiography stored in a LiF the Olea europaea pollen grain observed by a SNOM system on a border of the of fluorescent crown [20J. Figure 3a shows a typical topographical SNOM 3b, the the smooth surface of a LiF crystal. In fluorescence intensity distribution is measured. We observe a lace-like a fluorescent ring, as in the case of the CLSM measurements. In this SNOM image, the ring appears composed by two different distributions. The white line traces the intensity profile that is shown in variation results that this profile is able to follow the weak fluorescence in a width of - 110 nm.

3. (a) SNOM topography of the LiF crystal surface. (b) Corresponding SNOM fluorescence showing the nano-radiography of an Olea europaea pollen grain detail. The pollen in front of the X-ray laser plasma source, absorbed the X-ray radiation, preventing the fOflmation in the back area. The pollen crown is characterized by two different fluorescence intensities: a dark blue crown and a light blue crown are evident. The white line indicates where the fluorescence intensity profile was traced. (c) Fluorescence intensity profile, traced aloug the white line in (b), showing a resolution, measured along an edge, of - 110 nrn.

45 X-rays radiography of a periodic metallic structure stored in a thin LiF crystal has been also obtained at the X-ray laser plasma source in ENEA C.R. Frascati placing in contact with the radiation-sensitive LiF salt a copper mask with 1500 lines per inch. In Fig. 4a a topographical shear-force image of a LiF crystal is presented. In Fig. 4b the corresponding fluorescence intensity distribution is measured. The optical contribution represented in the fluorescence SNOM image arises from the electronic differences between the locally created CCs and the not irradiated blank LiF. In this SNOM image, there is an evidence that some of the fluorescent squares are shifted with respect to the principal ones, indicating that a movement happened to the copper mask during the X-rays exposure procedure. It is worth to note that in Figs 3 and 4, no topographic contribution is present, thus excluding any possible artefact. Figure 4c shows the intensity profile along the white line at the edge of the square of Fig. 4b: an edge width of - 75 nm is obtained in this case. A proposal was presented for exploiting the peculiarities of soft X-ray radiation produced by the future FEL X-ray source SPARX in the field of biological investigation by using single-shot contact microscopy and holography on LiF imaging detector. Coherence, monochromaticity and high brillance of a X-ray free electron laser (X-FEL) as SP ARX will overcome the limitations of the actual soft X-rays sources and will allow to obtain images of biological samples in single-shot experiments both in contact and in holographic configuration with very high spatial resolution. The high and unique brightness of X-FEL (3_8x10 3o Photls/O.l %bw/(mm-mrad)2) allows to reach spatial resolution forbidden with actual sources, especially taking into account the attenuation induced by the monochromatization of a synchrotron radiation source in order to select the coherent part of the beam. Due to the short duration (- 100 fs) of the X-FEL pulse, it could be possible to study living biological specimens by recording images in a very short exposure time, before radiation damage occurs. Due the coherence of X-FEL beam, biological investigation can be performed by single shot holography experiments as a method for a high resolution 3D imaging, also with complex holographic circuits. CONCLUSIONS

A novel X-ray imaging detector based on photoluminescence of CCs in LiF thin layers has been presented. The LiF-based imaging plate has been tested by using several X-ray sources, with emitting energy ranging from 20 eV to 8 keY, for investigation of biological specimens, materials and devices characterization, as well as light-emitting micro and nano-patterns transfer.

46

Fig. 4. (a) 25 IJmx 25 IJm SNOM topography of the LiF crystal surface. (b) COJrrespOlldirlg fluorescence image showing the X-rays radiography of a fluorescent periodic structure by covered by a copper mask. The white line indicates where the X-rays exposnre of a LiF fluorescence intensity was traced. (c) Fluorescence intensity profile, traced along the white line in b, shows an edge width of - 75 nm.

of LiF detector, like high spatial resolution on a a wide range, simplicity of use and U';'''Hill'Y'',"" can be exploited for X-ray microscopy in different COlrmli?:uraw::ms also for lensless and in vivo observation of biological overcome the limitations of the standard detectors and fully the potentlallitles offered by FEL-SPARX peculiar characteristics. rn,.",.r",pn~p"'ITC

of LiP film detector performances and optimisation of its technique are currently under development.

References Physics of Color Centers (W.B. Fowler, New York and A. Mancini, G.C. Righini and S. Pelli Opt. Commun.,

47

(1998),223. [3] A Belarouci, F. Menchini, H. Rigneault, B. Jacquier, RM. Montereali, F. Somma and P. Moretti Opt. Commun., 189, (2001), 281. [4] P. Adam, S. Benrezzak, J.L. Bijeon, P. Royer, S. Guy, B. Jacquier, P. Moretti, RM. Montereali, M. Piccinini, F. Menchini, F. Somma, C. Seas sal and H. Rigneault Opt. Express, 9, (2001), 353. [5] G. Baldacchini, F. Bonfigli, A Faenov, F. Flora, R.M. Montereali, A Pace, T. Pikuz, L.Reale, J. Nanoscience and Nanotechnology 3, 6, (2003), 483. [6] G. Baldacchini, S. Bollanti, F. Bonfigli, F. Flora, P. Di Lazzaro, ALai, T. Marolo, R. M. Montereali, D. Murra, A Faenov and T. Pikuz, E. Nichelatti, G. Tomassetti, A Reale, L. Reale, A Ritucci, T. Limongi, L. Palladino, M. Francucci, S. Martellucci and G. Petrocelli, Review Scientific Instrument 76,1, (2005),113104. [7] S. Almaviva, F. Bonfigli, I. Franzini, ALai, R.M. Montereali, D. Pelliccia, A Cedola, S. Lagomarsino, App\. Phys. Lett. 89, (2006), 054102. [8] R A. Cotton, Microscopy Analysis, 15 (September 1992). [9] R. Larciprete, L. Gregoratti, M. Danailov, RM. Montereali, F. Bonfigli, M. Kiskinova, App\. Phys. Lett. 80, (2002), 3862. [10] G. Baldacchini, F. Bonfigli, F. Flora, RM. Montereali, D. Murra, E. Nichelatti, A Faenov, T. Pikuz, App\. Phys. Lett. 80, (2002), 4810-4812. [11] G. Tomassetti, A. Ritucci, A Reale, L. Arizza, F. Flora, RM. Montereali, A Faenov, T. Pikuz, App\. Phys. Lett. 85, (2004),4163. [12] F. Barkusky, C. Peth, K. Mann, T. Feigel, N. Kaiser, Review of Scientific Instrument 76, (2005), 105102. [13] F. Calegari, G. Valentini, C. Vozzi, E. Benedetti, J. Cabanillas-Gonzalez, A Faenov, S. Gasilov,T. Pikuz, L. Poletto, G. Sansone, P. Villoresi, M. Nisoli, S. De Silvestri, and S. Stagira, Optics Letter 32, 14, (2007), 2593. [14] J. Nahum and D.A. Wiegand, Phys. Rev., 154, (1967), 817. [15] A Ustione, A Cricenti, F. Bonfigli, F. Flora, ALai, T. Marolo, RM. Montereali, G. Baldacchini, A. Faenov, T. Pikuz, L. Reale, App\. Phys. Lett. 88 (2006) 141107. [16] A Ustione, A Cricenti, F. Bonfigli, F. Flora, ALai, T. Marolo, R M. Montereali, G. Baldacchini, A. Faenov, T. Pikuz and L. Reale, Japanese Journal of Applied Physics 45,3b, (2006), 2116. [17] c. Barchesi, A Cricenti, R Generosi, C. Giammichele, M. Luce, and M.Rinaldi, Rev. Sci. Instrum. 68, (1997), 3799. [18] A Cricenti and R Generosi, Rev. Sci. Instrum. 66, (1995), 2843. [19] A Cricenti, R. Generosi, C. Barchesi, M. Luce, and M. Rinaldi, Rev. Sci. Instrum. 69, (1998), 3240. [20] C. Oliva, A Ustione, S. Almaviva, G. Baldacchini, F. Bonfigli, F. Flora, A Lai and R M. Montereali, AYa. Faenov, T. A Pikuz, M. Francucci, P. Gaudio, S. Martellucci, M. Richetta, L. Reale, A Cricenti, J. Microscopy 229, Pt 3, (2008), 490.

GROWTH MECHANISMS OF TIN OXIDE AND ZINC OXIDE NANOSTRUCTURES FROM VAPOUR PHASE LUCIO ZANOTTI, MINGZHENG ZHA, DAVIDE CALESTANI, ROBERTO MOSCA, ANDREA ZAPPETTINI Istituto IMEM-CNR, Parco Area delle Scienze 371A Parma, 43100, Italy Selected morphologies of nanostructured Sn02 and ZnO have been synthesized by using thermal sublimation and controlling the oxidation reaction, growth kinetics, local growth temperature and chemical composition of the source material. This paper focuses on crucial details to optimize the growth of nanowires and nanotetrapods. On the basis of the comprehension of the growth mechanisms, specific procedures are proposed for simple and large-scale production required by device applications.

1. Introduction

Semiconducting metal oxides (MeOx: SnOz, ZnO, Inz03, TiO z, ... ) possess attractive electrical, optical, chemical properties for a large number of device applications (e.g. as gas sensors, as electrodes in solar cells, as catalysts, ... ). Significant activity is underway to synthesize SnOz and ZnO based nanowires, nanobelts, nanorods with high aspect ratios and well controlled crystallinity for enhanced performance in such devices. In the recent years several authors have used vapour phase techniques (VPT) to produce such nanostructures (e.g. see [1-3]). Typically, the processes are performed by generating a vapour precursor that is transported via a carrier gas (such as Ar or N2) to the deposition zone, where single-crystal metal oxide based nanostructures are nucleated and grown. The growth of high aspect ratio nanostructures is accomplished through the use of a temperature gradient which favours the formation of high oversaturation in localized zones of the reactor, where nanocrystals grow usually in a combination of mixed morphologies. This is the case of SnOz and ZnO nanostructures when they are produced by standard vapour transport procedures (Fig. 1). On the other hand, the mentioned device applications of the nanocrystals require that they must be uniform in size and in morphological/physical properties, homogeneously spread on the substrate and, when necessary, grown in confined zone of the substrate.

48

49

Figure 1. SEM images of mixed morphologies of nanostructures grown by normal thermal evaporation process on alumina substrates: a) SnO! nanowires, nanobelts and nanopowders; b) ZnO nanowires, nanocombs, tetrapods and other micro/nanostructures.

For this reason the authors have carried out a "'fJ'v,",''''"' the vapour growth mechanisms of a few nanostructures, i.e. nanowires/-belts (SnOz-NWs), ZnO and ZnO tetrapods which at fJl,,-/Ull"H'c/<, for device applications.

to understand metal oxide fllllnn1JITITf's.l-tllnf'''i

are the most

2. Growth of nanoslrudures from vapour phase The standard usually employed in the preparation of nanostructured MeOx vapour process is represented by a tubular furnace with a tube in which is possible to have vacuum or flow different gases. Source material and substrates are generally located within an alumina in the central part of the tube. The of nanocrystals is a rather complex process, which at traltlSport of the metals in vapour phase, then a chemical reaction which and, at last, nucleation and growth of the occurs in thermodynamic conditions that are very far from equilibrium and with a very high supersaturation level. As a consequence, the growth rate of (up to few mm/hour) and the whole process takes environment, where even concentration of the gas comt:'OIllents is unsteady, as it depends on the reaction rate and the ternp'~rature, which vary along the reactor axis. in order to have a successful growth of selected nanostructures of NWs, ZnO-NWs and ZnO-TPs, we have developed specific procedures for each type of considered morphologies.

50

Tmax

Figure 2. Schematic outline of a typical reactor used in the growth of MeOx nanoslructures from vapour phase.

In the following we describe the experimental approaches to DfC)dulction of these nanostructures, defined and optimized on the basis of the observed growth mechanisms.

3. Growth process optimization for large-scale production 3.1. Tin Oxide nanowires On the basis of experimental evidence (absolute absence of liquid tin on the of NWs), we excluded for our process the claimed vapour-liquid-solid Hlc::cmumiIIl, well-known in the growth of silicon whiskers. On the other hand, the extremely high growth rate of also excludes the mechanism based on simple vapour-solid growth, because diffusion \u ........,,""'" by and concentration gradient) and drift carrier gas from source material cannot provide a sufficient Sn vapour feed for the growth of NWs. This apparent contradiction is overcome by the observation that the first step of the process, tin micro-droplets condense on the whole surface of the substrates and no growth occurs if this crucial step is i:>I\.1.IJIJ''"'U. The growth mechanism can be summarized in the following 2SnO(S) -tSn0 2 (S) +Sn(L....v)source Sn(V) -t Sn(L)snbstr.t. Sn(L)SUbstrate H Sn(y<) Sn(y<) + O 2 -t Sn02n.llowires

(1)

(2) (3)

(4)

51

two steps produce the condensation of Sn in liquid nr("\nl",t", two are related to Sn evaporation from droplet surfaces v.n. .............. reaction to Sn02. In this last step Sn can be vapour phase) faster than in the common source which is far away. Indeed, NWs \.... "',o.u,,,-, (generally less than few micrometers) from the are the droplets and this distance is equivalent to a few path of Sn atoms. In this way, the closeness of the ''''''UUJ'l<. grants a transfer rate that is much higher than the one distances (Fig. 3). This is in agreement with growth rate decreases with the increasing distance 'U'''' ......., and that NW s grow until the total consumption of liquid 'U

,",U

(b)

(c)

Figure 3, Scheme of the proposed growth model for SnOz nanowires: a) Sn vapour is transported along the reactor from source material and condenses on the substrate; b) when Oz is flowed in the reactor tube SnOz crystals are nucleated in contact with Sn micro-droplets; c) SnOz crystals grow in high supersaturation conditions close to the substrate. thus favouring the formation of elongated nanostructures,

described growth conditions can be easily reproduced controlling the main growth parameters (source and substrate tenlperaUlre, flow of the carrier gas, oxygen concentration) [5]. NW s result homogeneously spread on the whole substrate (Fig. 4) and are obtained in a large (a few cm2) zone of the reactor.

52

Figure 4. SEM images of SnOz-NWs obtained by the optimized selective proeess described in the text: a) typical nanowires entanglement; b) low magnification image of an alnmina substrate completely covered by nanowires.

3.2. Zinc oxide nanowires As in the literature, standard vapour phase """",,".'c ZnO nanostructure morphologies 5). Therefore ....".t+",rrYI

a specific growth procedure

Figure 5. SEM images of different ZnO nanostructures which can be obtained by vapour phase growth: a) nanorods; b) nanowires; c) nanoeombs; d) tetrapods

53

with reference to the SnOz-NWs described experience. we observed that the growth of long ZnO-NWs is favoured by the presence liquid Zn. Anyway, in this case, due to the high equilibrium vapour pressure of Zn at the reaction temperature (5OO-600°C), it is difficult to maintain an appropriate amount of liquid Zn condensed on So, a layer has been preliminary deposited on the substrate before the growth. layer thickness has been adjusted in order to supply amount of liquid metal. In our growth system, the best results have with a Zn layer thickness in the range of 5-lOjlm [6]. amount of liquid Zn guarantees high Zn vapour pressure and large ratio in proximity of the nucleation and growth zone of nanowires. Zn metal, as source, remains necessary to provide the proper Zn overpressure an extended region, in other words, it behaves as an "evaporation buffer", while Zn layer on substrate improves timespace-stability of local growth conditions so that the growth nanowires become less sensitive to small fluctuations of growth parameters. In this way ZnO-NWs can be obtained, free of other undesired morphologies, homogeneously spread on the substrates, wherever the Zn layer is present (Fig. 5b). In addition, this procedure may be perfectly compatible with a patterned growth process, since NWs grow only in region the substrate where Zn has been deposited (see for instance 6).

Figure 6. Concept of a patterned growth of ZnO-NWs which can be obtained by metal Zn layer masked deposition. The inset shows the border between the Zn deposited region (right side: nanowires growth) and the "clean" masked region of the alumina substrate (left side).

54

As no trace of metallic Zn has been observed at the rrt"r"","n also ZnO-NWs growth is considered to be vapour-solid mechanism.

a

3.3. Zinc oxide tetrapods the most interesting morphologies ZnO nanostructures it has been found a branched nanostructure with four legs oriented along the axes of a tetrahedron, called "tetrapod". While SnOz and ZnO NWs and other nanostructures grow IllfieClliV substrate where their nucleation occurs, the growth of ZnO tetl:ap()ds COIrnplel1eay takes place in the vapour phase.

detail, it has been observed that these nanostructures and grow while floating in the carrier gas "... ,~..."'~, L~JU"'H'''''' smoke" that moves along the reactor and, once a "uu...,u T"t,.<.....".,-i is reached, they deposit on the substrates and on This dynamic process run out in a very short v .......,,"'Vl! of oxygen in the reactor limits Zn evaporation source material and rapidly reduces Zn concentration in As a consequence, the zone of evaporation and oxidation/nucleation must be physically separated inside by introducing Oz downstream and far away

reactor source

Figure 7. Scheme of Ihe stream-growth of Zno tetrapods, which nucleate and grow while floating in Ihe carrier gas togeIher wiIh Iheir gaseous reagents (Zn vapour and (h). Zn evaporation from sonrce material and oxygen inlet are kept separated in different zones of the reactor and at different temperatures.

55

Setting the proper temperatures for these two zones and using high inert gas flow rates, it is possible to obtain a continuous "stream growth" of ZnO tetrapods, which collects at the end of the reactor, separated from any other different ZnO nanostructure (Fig. 5d). In a subsequent process ZnO-TPs can be deposited at room temperature from an appropriate suspension of organic solvents on different substrates (patent pending). 4. Conclusions We have summarized recent studies which result of basic importance to define the appropriate conditions to grow selected morphologies of tin and zinc oxide nanostructures from vapour phase.

The described processes are a combination of thermal evaporation and controlled oxidation procedures, are carried out at relatively low temperature (500-900°C), don't require the use of any metal catalyst and result in very high yields of "samples per run". In this way we have found key-procedures for low-cost large-scale production of tin/zinc oxide nanostructures required by the applications in the new generation of solar cells, as high-efficiency photocatalysts and as high-performance gas sensors.

References 1. 2.

3. 4. 5. 6.

Z. R. Dai, Z. W. Pan and Z. L. Wang, Adv. Funct. Mater. 13,9(2003). J. G. Lu, P. Chang and Z. Fan, Materials Science and Engineering R 52, 49 (2006). L. Schmidt-Mende and J. L. MacManus-Driscoll, Materials Today 10, vol. 5,40 (2007). Z. L. Wang, Materials Today 7, vol. 6, 26 (2007). L. Zanotti, M. Zha, D. Calestani, E. Comini and G. Sberveglieri, Cryst. Res. Technol. 40,932 (2005). M. Zha, D. Calestani, A. Zappettini, R. Mosca, M. Mazzera, L. Lazzarini and L. Zanotti, Nanotechnology 19, 325603 (2008).

BROKEN-SYMMETRY STATES AT SURFACES: THE (tr)ARPES VIEW L. PERFETTllo

Laboratoire des Solides Irradies, Ecole Poly technique, 91128, Palaiseau Cedex, Prance E-mail: luca. perfetti@polytechnique. edu M. GRIONI

Institut de Physique des Nanostructures, Ecole Poly technique Federale (EPFL) , CH-1015 Lausanne, Switzerland E-mail: [email protected] We discuss recent high-resolution angle-resolved photo emission (ARPES) and time-resolved ARPES experiments on selected surfaces which illustrate the formation of broken-symmetry ground states of very different nature: i) a Mott insulator in the layered dichaicogenide IT-TaSe2, and ii) a spin-orbit split state in the non-magnetic surface alloys BiAg2 and PbAg2.

Keywords: ARPES; time-resolved; surfaces; Mott insulator; spin-orbit splitting

1. Introduction

As illustrated by other articles in this Volume, the physics and chemistry of surfaces and interfaces have been extensively studied under the impulse of applications like the Schottky barrier formation and catalysis. More recently it has been found that new ground states can be realized at surfaces that are not possible or not stable in the material's bulk. They are the result of the reduced dimensionality, the different/ lower symmetry, and the often stronger electron-electron or electron-phonon interactions. Moreover, crucial parameters like the stoichiometry and band filling can often be more easily controlled at surfaces. Hence, surfaces provide useful model systems where specific aspects of many-body physics can be studied. In this paper we illustrate these ideas with two very different examples. In the first we show that stronger correlations at a surface can lead to an insulating state with long-range order over a metallic bulk. In the second, we show that the breaking of inversion symmetry, in the presence of spin-

56

57

orbit interaction, lifts the usual spin degeneracy of non-magnetic solids, and modifies the electronic structure in a way that is potentially interesting for applications. 2. Charge-density-wave and Mott insulating phase Strong many-body interactions can qualitatively modify the structure of solids, leading to broken-symmetry phases with charge, spin, or orbital order. A notable example of broken symmetry is the softening of a phonon branch due to the strong coupling of a lattice mode with the conducting electrons. The resulting lattice distortion comes along with a modulation of the charge density referred to as a charge density wave (CDW). Based on the topology of the electronic structure, the CDW state may lead to an insulating ground state or preserve portions of the original Fermi Surface. Moreover, unexpected ground states may emerge from the interplay of the CDW with electronic correlations. Recently, the essential role of the Coulomb repulsion has been demonstrated in the case of V0 2 and in the transition metal dichalcogeniges. This section describes the Metal-Insulator (M-I) transition observed at the surface of the two sister compounds 1TTaS2 1 and IT-TaSe2. 2 Several spectroscopic techniques are employed in order to identify the nature of the insulating state as well as the stabilization mechanism. Interestingly, the investigation of non-equilibrium states by ultrafast spectroscopy offers a novel and powerful tool to disentangle the electron-electron from the electron-phonon interaction. 1T-TaS2 and 1T-TaSe2 consist of strongly bound X-Ta-X (X=S,Se) layers that are weakly coupled by interplane interaction. Such a layered structure determines a marked anisotropy of the electrical resistivity and of the mechanical properties of the material. X-ray scattering3 and scanning tunneling microscopy (STM)4 show that IT-TaS2 displays a pronounced lattice modulation already at room temperature. The strong coupling between valence electrons and phonons periodically distorts the lattice, modifying the spatial distribution of the charge density. In real space, the local distortion leads to metallic clusters containing 13 Ta atoms each. Extended Huckel calculations 5 suggest that the CDW splits the Ta d conduction band into subbands which contain a total of 13 electrons per unit cell. Two subbands, carrying 6 electrons each, are filled and lie below the Fermi level (E F ). The Fermi surface is formed by a half-filled subband carrying the 13th electron. The opening of a correlation gap in this subband is responsible for the M-I transition observed in bulk 1T-TaS2 (Tc=180 K)l and at the surface of 1T-TaSe2 (Tc=260 K).2

58

VVi::I\revecmr

(1/A)

Fig, 1. ARPES intensity map of IT-TaSz measured in the metallic phase at 300 K and in the Mott insulating state at 70 K (b),

indirectly reflect the dramatic rearra,nglam,ant of the electronic structure at the transition. Photoelectron Qn"f'j"'rr.""rn"T which the function, can a direct view map of 1(a) and (b) display the phase, A broad Ta in the metallic and from the zone boundary and crosses the Fermi at = 0.25 rM. Due to the CDW the whole level band structure should be folded back into the reduced Brillouin zone of the reconstructed structure. However, the "",,,,,..tc,e,, rec:mI'oc:al space is controlled not only also of the scattering potential. 6 still carries most of the weight in the CDW phase, and is transferred to the or shadow bands. As shown by the occupied of the Ta d band into subbands the CDW at 1 0.5 and around The latter is susceptible to electronic localization due to the strong Coulomb interaction between electrons. Below the transition the electronic correlations open an electronic gap, and a lower Hubbard band emerge at 0.2 eV.7 Even though the M-I transition is consistent with the Mott the first order of CDW lattice hinder the description of the transition in terms of such do not take in the sister compound

59

Fig. 2. (a) ARPES spectra of IT-ThSe2 measured at the Fermi wavevector between 300 K and 70 K. (b) UPS spectra of the core levels acquired at 300 K and at 70 K The zero of the energy scale has been set to the position of the first peak.

is commensurate up to 500 K. lUvUlJUI<.U tice is 8 the of occurrence of a M-I transition at at the while the bulk of remains metallic. We ascribe the occurrence of an surface to the combined effect of CDW discan be monitored tortion and Mott localization. The local CDW ""'I'i.+; ..... " of the Ta valence. s Their energy splitting is . .-...,,,,,,,."'/-;"""'0 vanishes in the undistorted As shown the is 40 meV in the insulating (70 than in the metallic one Such an increase of the CDW reduces the electronic bandwidth. Below a critical value at 260 the energy that is necessary for the double of one site becomes than the bandwidth. Therethe metallic state is no stable and electrons localize into the Mott t:t;l,n:.,:tvt: the response of to an ultraThereby, an intense pulse in the infrared "....,,,T .. <> 1 into a highly excited whereas an ultraviDhot(}el!~ctrOlls after a variable time shows a intensity map acquired in at 30 K. oscillations of the electronic start

60

o

2

4

6

Fig. 3. Photoelectron intensity map measured in the MI phase of pump-probe delay and binding energy.

8

as a function

for many 9 The period of these oscillations COlTe!>DC)lldIS to the amplitude mode of the CDW.lO In real space, this excitation results in the coherent motion of the Ta atoms toward the undistorted As a consequence, the release of elastic energy induces a shift of the Hubbard towards Subsequently the nuclear structure breathes until a force the system back to equilibrium. If the Mott insulator is in equilibrium at 30 K, the "n.".. t-"" ",,",U,,1.'V1'I a Lower Hubbard band (LHB) at 0.21 eV and electronic states at Ep. Just after photoexcitation, the energy IJU'>!~"'U in the electronic system is too large for the existence of a Mott insulator. 4(a) shows the photo-induced transfer of from 9 the LHB to the Fermi level. The electronic gap is partially filled by transient electronic states and recovers along with the energy relaxation of the hot electrons. The LHB intensity in figure 4(b) is an measure for the transfer of spectral weight toward the pseudogap. The instantaneous decrease and subsequent recovery of IH can be described by an ex:pO]1le11tll"t1 with a time constant of 680 fs. It follows that the M-I gap vVLL""!-,'O<;O on a timescale much shorter than the pump-probe cross-correlation and monotonically recovers with a subpicosecond timescale. This is consistent with the revival of a M-I groundstate whereas it is not consistent with the dynamics of a Peierls insulator. In conclusion, we show that CDW and electronic correlation are both essential for the correct understanding of 1T-TaX2 • The of atomic "U'""5'U5 with spectroscopic techniques identifies broken

61

(a) -

50fs 250fs 500 fs 900fs -4.5 ps

100 50

0 0.4

(b)

0.2 0.0 Binding

-0.2 -0.4 (eV)

-0.6

_:r:. 1.0

a3

.!::! Iii

E 0

z

t"H:::

680 fs

0.8 0

2 3 Pump-Probe Oelay (ps)

4

Fig. 4. Photoelectron spectra of the photoexcited insulator acquired at several pumpprobe delays. (b) Intensity of the Lower Hubbard Band as a function of pumpprobe delay.

and the Mott insulating phase. Moreover the response of the electronic states to an ultrafast optical perturbation clearly distinguishes the effects of the CDW and of correlations on the band structure.

3. Lifting the spin degeneracy at surfaces In materials it is often taken for granted that states with the same wave vector k and opposite spin have the same energy, but this is true only if both time reversal and space inversion are present. At a surface the latter is broken, and the spin-orbit (SO) interaction 11 This suggests the intriguing possibility of lifts the with electric fields. For a 2D free electron gas = CXR (f. where (f is the Pauli matrices vector. The 'Rashba' coupling CXR is proportional to the surface electric field, which is assumed to be oriented along the surface normal ii. describes the coupling of the spin magnetic moment with the magnetic field appearing in the rest frame of the electron. This field, which lies in the plane of the is pe:rpEmOlICtllar to k so that, for a given energy, the spin polarization describes circles

62

Fig. 5. ARPES intensity maps for the surface alloy (top) and schematic band structure of the Rashba model (bottom) for three different cuts in k space, through r and for A-1 and A-1 (C).

in I\:-soa(:e around band rli~rH>Y·~i

The i.t consists of two on1r\n<,;b, sides of r an offset '" OR. the Rashba energy is The characteristic energy scale of the OIT<.1-n,,, effecdefined the difference between the bottom tive of the

because the electron probes much fields near the atomic cores. VH:;'a..L"Y, both the surface structural inversion asymmeand the atomic "n,n_()"1" 1>,1". interaction contribute to the observed electronic structure. we have observed an even effect in a surface formed by ML of Bi at the Ag(lU) surface. shows energy-wave vector maps which illustrate three

63

leI cuts the band structure. 15 Near r the data are qualitatively consistent with the simple Rashba model (top panel). First-principles, fully relativistic calculations, which correctly reproduce the experimental show that the dispersing features represent the SO split branches of a brid band with Bi 8pz character. A broader angular scan would also reveal a second set of with a Bi Pxy character and a smaller SO with Remarkable in these data are the giant momentum and Rashba energy (E R =200 which are reone and two orders of magnitude larger than for the benchmark case of Au(lll), and orders of magnitude larger than values demonstrated in semiconductor heterostructures. Achieving large band is cru16 cial in view of applications like the proposed spin since L= is the length over which the is reversed.

Fig. 6.

The

SO split band structure near

r

for the PbAg2 and BiAg2 surface alloys

are not unique: we have also observed a SO in the isostructural PbAg 2 surface alloy. Their electronic structures in 6. With one valence electron less in Pb (4) than near r are in Bi (5), all bands shift rigidly upwards in energy. The top of the SO 8Pz branches now lies above and is not accessible ARPES. the SO is reduced, as a consequence of the lower (by one atomic number. the fourfold reduction of ko from 0.13 to 0.03 usti11ed by the change in the atomic SO from eV to (6p(Pb)=0.91. Also, ko is reduced to almost zero in the "1_""'~~"UI".

64

Fig. 7. Constant energy cuts through the ARPES band structure for BiAg2 180 meV and 320 meV below EF.

isostructural SbAg2 surface alloy, where Bi is replaced by Sb neither a simple Rashba scenario, nor a atomic model can fully describe the experimental results. A clue on the possible origin of the giant SO is constant energy contours extracted from the ARPES data, and shown in 7 for The cut taken at 180 meV binding energy (BE), in the between the band maximum and the crossing point, shows two concentric contours corresponding to the inner and outer branches of the Spz band, consistent with 5. The inner contour shrinks to a at BE=320 in correspondence of the crossing of the two branches. Unlike the Rashba model, and the free-electron-like case of 1), the outer contour is not circular, but hexagonal. Similarly the inner contour evolves from a circular to a hexagonal shape as its size increases at energy (not shown). Clearly the bands of the surface alloy are influenced by the lattice potential. More specifically, the anisotropic charge distribution in an in-plane component of the surface electric field presence of an in plane SIA, generates an additional contribution to the 'normal' Rashba effect. Model calculations 17 suggest that this contribution is an essential ingredient of the large SO splitting. 6 shows that large changes in the Fermi level position are induced by the replacing Bi with Pb. A finer control can actually be by tuning the band filling in a mixed (Bi x Pb 1- x )Ag2 alloy.I8 allows to be placed between the top of the band and the band

65

(region I in 8), where the Fermi surface corresponds to the constant energy contour of 7 (b). This situation is intriguing in two respects. Unlike the case of a normal two-dimensional state, the density of states (DOS) is not constant in this region, but (ideally) diverges at the band maximum Emax as (Emax-E)-l. Such a divergence is typical of one dimensional systems, and can be traced back to the fact that the constant energy contours of the SO split bands do not evolve into a point at but into a circle of finite radius. Momentum-averaging scanning tunneling spectroscopy (STS) measurements on the alloy systems have confirmed the strong enhancement of the DOS.19 Theory 2o predicts a strongly renormalized electron-phonon interaction as a consequence of the diverging DOS, when approaches Emax. The intriguing properties of such a 2D Fermi liquid are still unexplored experimentally. The spin polarization expected from the Rashba model is also unique in this region, because the spins 'turn' in the same direction on both Fermi surface contours, yielding a very peculiar spin-chiral state.

ill

i

ill

Wave vector k

Density of States

Fig. 8. Schematics of the Rashba branches (left) and the corresponding density of states (right)

4. Conclusions

New ground states that are not allowed in the bulk can be stabilized at the surface of a solid. The case studies briefly discussed here illustrate this point,

66

and show that ARPES with high momentum and energy resolution is a powerful probe of the new broken-symmetry phases. Time-resolved ARPES experiments provide direct information in the time domain on the dynamics of their electronic structure. The importance of this information, which complements and extends the traditional energy domain view of ARPES, justifies the very rapid ongoing development of this emerging technique. Acknowledgments It is a pleasure to acknowledge our collaborators at Berlin and Lausanne. We are especially grateful to P. Perfetti for the discussions, occasional or frequent but always insightful and enjoyable, we had over the years on various aspects of the physics of surfaces and beyond.

References 1. B. Dardel, M. Grioni, D. Malterre, P. Weibel, Y. Baer, and F. Levy, Phys. Rev. B 46, 7407 (1992). 2. L. Perfetti, A. Georges, S. Florens, S. Biermann, S. Mitrovic, H. Berger, Y. Tomm, H. Hochst, and M. Grioni, Phys. Rev. Lett. 90, 166401 (2003). 3. Akiji Yamamoto, Phys. Rev. B 27, 7823 (1983). 4. Ju-Jin Kim, W. Yamaguchi, T. Hasegawa, and K. Kitazawa, Phys. Rev. Lett. 73, 2103 (1994). 5. K. Rossnagel and N. V. Smith, Phys. Rev. B 73, 073106 (2006). 6. J. Voit, L. Perfetti, F. Zwick, H. Berger, G. Margaritondo, G. Griiner, H. Hochst, and M. Grioni, Science 290, 501 (2000). 7. L. Perfetti, T. A. Gloor, F. Mila, H. Berger, and M. Grioni, Phys. Rev. B 71, 153101 (2005). 8. S. Colonna, F. Ronci, A. Cricenti, L. Perfetti, H. Berger, and M. Grioni, Phys. Rev. Lett. 94, 036405 (2005). 9. L. Perfetti, P. A. Loukakos, M. Lisowski, U. Bovensiepen, H. Berger, S. Biermann, P. S. Cornaglia, A. Georges, and M. Wolf, Phys. Rev. Lett. 97,067402 (2006) . 10. J. Demsar, L. Forro, H. Berger, and D. Mihailovic, Rev. Lett. 66, 041101 (2002). 11. Y. A. Bychkov and E. 1. Rashba, JETP Lett. 39, 78 (1984). 12. S. LaShell, B. A. McDougall, and E. Jensen, Phys. Rev. Lett. 77,3419 (1996). 13. F. Reinert, G. Nicolay, S. Shmidt, D. Ehm, and S. Hiifner, Phys. Rev. B 63, 115415 (2001). 14. Y. M. Koroteev, G. Bihlmayer, J. E. Gayone, E. V. Chulkov, S. Blugel, P. M. Echenique and Ph. Hofmann, Phys. Rev. Lett. 93, 046403 (2004). 15. Ch. R. Ast, J. Henk, A. Ernst, L. Moreschini, M. C. Falub, D. Pacile, P. Bruno, K. Kern, and M. Grioni, Phys. Rev. Lett. 98, 186807 (2007). 16. S. Datta and B. Das, Appl. Phys. Lett. 58, 665 (1990).

67 17. J. Premper, M. Trautmann, J. Henk, and P. Bruno, Phys. Rev. B 76, 073310 (2007). 18. Ch. R. Ast, D. Pacile, L. Moreschini, M. C. Falub, M. Papagno, K. Kern, and M. Grioni, Phys. Rev. B 77, 081407(R) (2008). 19. Ch. R. Ast, G. Wittich, P. Wiihl, R. Vogelgesang, D. Pacile, M. C. Falub, L. Moreschini, M. Papagno, M. Grioni, and K. Kern, Phys. Rev. B 75, 201401(R) (2007). 20. E. Cappelluti, C. Grimaldi, and F. Marsiglio, Phys. Rev. Lett. 98, 167002 (2007).

NANOSTRUCTURING THROUGH LASER MANIPULATION· F. TANTUSSI, N. PORFIDO, F. PRESCIMONE, F. FUSO, E. ARIMONDO, M. ALLEGRINI CNISM, INFMlCNR po/yLab, Dipartimento di Fisica Enrico Fermi, Universita di Pisa Largo B. Pontecorvo 3, 1-56127, Pisa, Italy We have developed a nanofabrication approach that exploits laser manipulation to guide neutral atoms, belonging to a well collimated beam, into regularly spaced positions prior to deposition onto a substrate. The method can be straightforwardly adapted to structured deposition in the low substrate coverage regime. Results demonstrate the achievement of isolated nanostructures with lateral size in the 10 nrn range.

1. Introduction

The relentless search for miniaturization prompts the need for new fabrication techniques in many technological fields. Different strategies and alternative methods have been proposed, developed and, in some cases, introduced into the industrial environment to push resolution down to the few nanometers range, thus overcoming the limitations of conventional techniques. Besides the need for enhancing space resolution, a strong interest is growing for developing bottoms-up methods [I]. Contrary to conventional top-down techniques, where structuring is typically achieved by removing or modifying laterally defined patterns, bottoms-up implies matter manipulation right before, or during, the structure formation. Such new technologies promise radical changes in the fabrication strategies, allowing an unprecedented control on microstructure, stoichiometry and morphology. Advanced applications ranging through, for instance, precise doping of materials for spintronic purposes to hybrid molecular electronics will take advantage of the precise lateral definition and of the virtual absence of any unwanted material damage offered by such techniques. Methods based on scanning probe microscopy (SPM) have been demonstrated able to manipulate matter down to the single atom level [2]; however, SPM techniques are typically cumbersome to be realized and suffer from their inherently serial, hence slow, character.

• This work has been partially supported by EC through FET-JST "Nanocold" and by Fondazione Cassa di Risparrnio di Pisa under the Scientific Project PROS/137.

68

69

In the fundamental areas of laser physics, atomic spectroscopy and metrology, many efforts have been devoted to control the dynamical properties of atoms and molecules in the vapor phase [3], that led to a number of important advancements including, e.g., production of Bose-Einstein condensates and introduction of atom lasers. Atom optics techniques represents a viable route to nanofabrication thanks to the highly accurate matter control that can be achieved. The inherently non-obtrusive character of light and the possibility to develop parallel schemes are additional appealing features ensured by laser manipulation. 2. Laser manipulation tools Application of laser manipulation tools to fabrication brought the development of atomic nanofabrication (ANF [4,5]), also known as atom lithography. Roughly speaking, in ANF the role of matter and light is reversed with respect to conventional optical lithography: a beam of atoms (the matter) is in fact used to produce nanostructures through conditioning by laser radiation (the light). As in electron lithography, the sub-nm de Broglie wavelength of the atom beam prevents diffraction effects to playa remarkable role. Moreover, the process involves low kinetic energy particles, hence detrimental effects like backscattering or sputtering are virtually absent; contrary to charged beam lithography, any issue related with Coulomb repulsion can be neglected as well. Pioneering implementations of ANF used sodium, and, later on, chromium to create arrays of regular nanolines with lateral size in the tens of run range [6,7]. Suitable particle-sensitive resists were then introduced, mostly based on self-assembled monolayers (SAM), whose impression allows transferring the pattern created by laser manipUlation onto the underlying substrate [8]. The main mechanism of ANF is the occurrence of a conservative force, called dipolar force, following the interaction of an atom with a standing electromagnetic field at a wavelength quasi-resonant with an atomic transition. Such a force, classically similar to that felt by electric dipoles immersed in non homogeneous fields, stems from the space modulated light shift for the energy levels of the atom dressed by the external field [4,5]. Assuming a standing wave with an intensity variation along the x-axis, l(x), as produced by retro-reflecting a single laser beam (l-D standing wave), the force F(x) is [4]

nr2 ol(x) F(x) "'" - - - 80ls Ox

(1)

70

where r and Is are the natural rate and the saturation intensity for the considered atomic transition, respectively, and (j is the detuning from resonance. Let us consider an atom travelling along a direction orthogonal to the x-axis: its transverse dynamics will be modified by the dipolar force and the atom will be pushed towards the antinodes, or nodes, of the standing wave, depending on the sign of the detuning. Therefore, an initially homogeneous beam will be spatially segregated into an array of parallel planes spaced exactly half the wavelength. More complex field geometries, e.g., produced by superposing three or more laser beams, lead to differently shaped arrays, consisting for instance of regularly spaced hexagonal or circular dots [4). In absolute terms, the dipolar force of Eq. (1) is typically weak. In fact, detuning (j cannot be set arbitrarily low in order to prevent photon absorption and consequent re-emission, which would produce atom heating and subsequent lack of control in the atom dynamics. For the guiding mechanism to be effective, the transverse velocity of the atoms must be very small, hence highly collimated beams are typically required in ANF, with residual divergence in the mrad range. In order to preserve atom beam density, i.e., to achieve exposure times compatible with practical applications, collimation cannot be attained by simple mechanical means. Laser manipulation tools are used also to this purpose based, e.g., on 2-D optical molasses [3] able to decrease the beam transverse temperature below the mK range. 3. Experimental Our implementation of ANF operates with cesium atoms and exploits a I-D standing wave as the light mask. The setup [9-11] has been designed and built to satisfy all basic requirements of ANF; moreover, contrary to all systems reported in the literature, it includes a modified magneto-optical trap (MOT) working as an atom funnel to produce a longitudinally cooled atom beam [10). Main motivations to such a design choice are: (i) to access the mostly unexplored deposition regime involving arrival of low kinetic energy atoms onto a surface; (ii) to achieve long interaction times in both collimation (optical molasses) and atom guiding (light mask) stages thanks to the small velocity of the beam (around 10 rnIs [10]); (iii) to realize straightforward operation in the low particle density regime. In fact, due to the use of specific laser interaction schemes, the dynamical properties of the atoms are almost deterministically assigned independently of the particle density. Therefore, deposits consisting of few atoms can be attained by reducing the atom flux through externally

71

accessible parameters (e.g., laser power, magnetic field, background vapor density). The system has already proven its capabilities in resist-assisted fabrication of parallel nanotrenches in gold [II], spaced exactly one half the wavelength (Al2 = 426 nm, in our case). In those experiments, the space segregated atom beam was used to impress a particle sensitive film consisting of a self-assembled monolayer of alkylthiol molecules grown on gold. Arrival of cesium atoms, at a dose above 2 atoms per molecule, inhibited protection; a subsequent wet etching process led to nanotrench formation. Size and morphology of the nanotrenches revealed a strong influence of the underlying layers. In particular, the homogeneity of the SAM, ruled in turn by the graininess of the gold layer, posed a limitation in the maximum achievable space resolution (40-50 nm [II]). In order to unravel the intriguing phenomena underlying nanostructure growth in ANF conditions, direct deposition from the structured Cs beam was accomplished on different substrates. Results presented here refer in particular to highly oriented pyrolitic graphite (HOP G) and mono layers of nonanethiol molecules self-assembled onto a thin (100 nm) flame-annealed gold film on mica. Thanks to a tunneling microscope (STM) head (Omicron LS-STM equipped with Dulcinea Nanotec controller [12]), installed in the same ultrahigh vacuum deposition chamber, the properties of the deposited samples could be detected without any need for sample exposure to air. 4. Results and discussion

Many different results have been obtained in a large series of experiments: a few of them will be presented and briefly discussed here. Our experiments were aimed at exploring the regime of moderate to low substrate coverage, meaning that the flux density of the laser-cooled atom beam, time integrated over the whole duration of the exposure, was typically kept below the value corresponding to the growth of a single Cs layer. It must be noted that, due to the number of structural variants possible for Cs on graphite [13], there is no unambiguous definition of the coverage ratio. Results presented here refer to a time integrated flux density corresponding to arrival, on the average, of 0.050.20 Cs atoms per single graphite carbon atom. In practical terms, this roughly corresponds to 100-300 min exposure time at an atom flux around 108 atoms/s over a 25-40 mm2 area. Due to the relatively long duration of the deposition process, special care has been devoted to mechanical stability and immunity against thermal drifts of the deposition setup, concerning in particular the relative alignment between the standing wave and the substrate.

72

Thanks to the long interaction time experienced by the slow atoms crCISSllnl! the wave, the so-called channeling regime is realized. In such atoms are forced to oscillate in the transverse direction; those oscillations confine the particles in an array of channels, spaced exactly one half the whose width can be numerically evaluated. 1 results of numerical simulations of the atom trajectories based on a semiclassical model accounting for spatially modulated light shifts of the atom levels and for the possibility of spontaneous emission following off-resonance absorption. Parameters are in agreement with the experiment; channeling is efficiently the channel width being on the order of 40-60 nm. 4

2

o -2

-4

~

4

~

~

~

0

1

2

345

xll Figure l. Results of the numerical simulation for 150 atom trajectories. The atom beam, assumed to possess a 10 mrad initial divergence, moves along the z-direction (from the bottom to the top of the graph) and interacts with a standing wave directed along the x-direction, assumed to be focused on a waist WL (typically, WL 50-100 /-UTI). Channeling and spatial segregation on a AI2 scale are evident. The peak: intensity of the standing wave was set to 20 mW/cm2 and its detuning was 15= 1 GHz.

Fabrication of a regular array of cesium nanolines should then be p'y,..p"i'",rI· however, the reduced substrate coverage achieved in the experiment prevents formation of continuous lines, leading instead to the growth of isolated nanostructures. This is demonstrated for instance in Fig. 2(a), showing the STM current map of a HOPG substrate exposed to the structured Cs beam. Spots observed in the scan, absent in the pristine substrate and attributed to Cs structures, are mutually aligned along a direction orthogonal to the stanOllng wave vector, with a spacing compatible with the expected).12 value [see the line profile in Fig. 2(b)]. The deposited material is unevenly distributed over the substrate: such a finding is a clear signature of surface processes occurring at the interaction of the deposited atoms with the substrate [14]. In particular,

73 surface diffusion is expected to occur; energy barriers in the diffusive motion Ehrlich-Schwoebel barriers) have been already considered as one of the '''O''P.£1.'''" .., in the morphology of ANF-produced structures we cannot completely rule out the possibility of atom t"Ip.lr1p.tr"tllrlTl intercalation by alkali atoms is a well known process [13J.

(a)

1

X£i.tmJ Figure 2. Tunneling current map of a sample deposited onto HOPG (a); profile analysis along the segment superposed to the map (represented as a double arrow), demonstrating line spacing compatible with »2 426 run (b). The segment direction coincides with the standing wave vector. The estimated coverage ratio of the deposition was -0.05.

STM topography investigations reveal that substrate features affect nanoisland shape and location. The map shown as an ..,"'.....uli-,." demonstrate that Cs deposits are mainly found in the "''''''U''U'.1 naturally occurring fractures between terraces. We found a relationship between height of the steps and adsorbed Cs: islands are not observed close to steps smaller than approximately 0.1 nm, COlrre!lJ)ona:mg to 2-4 graphene sheets. Moreover, nanoislands appear further structured in small droplets [see the magnified scan in 3(b)].

•

II

Figure 3. STM topography of a sample deposited onto HOPG (a) and magnifieation of a Cs nanostructure (b). The double arrow superposed to the left map marks the direction of the standing wave vector. The estimated coverage ratio of the deposition was -0.15.

74

Nanoislands fabricated on HOPG systematically exhibit a 1J\O~,UU'>1 like" with transverse size on the order of 10 nm, or even of a few urn. Their length is variable in the range 10-200 dependent on the presence of substrate features. the axis of the was always almost np,'i"pc·th, with the wave direction; such a behavior, which was not found in carried out without the light mask, a definite demonstration of the atom guiding effect due to the presence of the "utUUllil/:5 wave. size scans demonstrate that the distribution of the de]:losllied material follows the Al2 spacing imposed the """'!!Ull11l'5 vlll
and IV curves acquired on different of the standing wave vector, The

Interaction between alkali atoms and SAM has been process [4,8]. ionic character while calculations that cesium atoms tend to ltPI!JnJI(1CJ[lln.g the substrate because of Coulomb forces. ions can '''_'.......''Hlll\OY towards the substrate, hence the SAM upon arrival onto the underlying gold energy is released to break the thiollAu bond. We have not found any clear evidence for molecular in our morphological Further confurnation was found in local current vs bias measurements STM in selected positions on the surface. An
75

compared with that acquired over a Cs nanoisland, almost spherically shaped with diameter of a few nm. The expected IV behavior for a SAM layer grown over gold was found when analyzing uncoated regions, whereas tunneling through the Cs nanoisland appears inhibited in a relatively large bias range across zero. Such a finding can be explained in terms of Coulomb blockade [I], demonstrating that the Cs nanoisland behaves like a metallic nanocapacitor (estimated capacitance 0.3 aF) able to feel single electron charge processes. 5. Conclusions Thanks to its ability in governing the dynamical properties of neutral atoms, laser manipulation is a viable candidate for the development of novel nanofabrication techniques. Our results confirm that laser manipulation tools can be exploited to produce isolated nanostructures, with a typical minimum size in the ten nm range. Further work, supported by the continuous advancements in laser technologies, will address applications to technologically relevant materials, such as semiconductors or dopant species, in order to set the basis for a new approach in precise doping and hybrid molecular electronics. References 1. For a review, see, for instance: R. Waser Ed., Nanoelectronics and Information Technology (Wiley-VCH, Weinheim, 2005). 2. B. Bushan, Ed., Springer Handbook of Nanotechnology (Springer, New York, 2007). 3. H. Metcalf and P. van der Straten, Laser Cooling and Trapping (Springer, New York, 2001). 4. D. Meschede and H. Metcalf, 1. Phys. D: Appl- Phys. D 36, R17 (2003). 5. M.K. Oberthaler and T. Pfau, 1. Phys.: Condens. Matter 15, R233 (2003). 6. G. Timp, et aI., Phys. Rev. Lett. 69, 1636 (1992). 7. J.J. McClelland, et aI., Science 262,87 (1993). 8. K.K. Berggren, et aI., Science 269, 1255 (1995). 9. F. Tantussi, et aI., Mat. Sci. Eng. C 27, 1418 (2007). 10. A. Camposeo, et aI., Opt. Commun. 200, 231 (2001). 11. C. O'Dwyer, et aI., Nanotechnology 16, 1536 (2005). 12. I. Horcas, et aI., Rev. Sci. Instrurn. 78, 013705 (2007). 13. M. Caragiu and S. Finberg, 1. Phys.: Condens. Matter 17, R995 (2005). 14. J.A. Venables, Surface and Thin Film Processes (Cambridge University Press, Cambridge, 2000). 15. F. Nita and A. Pimpinelli, 1. Appi. Phys. 97 113529 (2005). 16. C. Di Valentin, et aI., 1. Phys. Chern. 109, 1815 (2005). 17. F. Tantussi, et aI., Appi. Surf. Sci, in press (2008).

LOCALIZATION AND DIFFUSIVE PROCESSES IN THE ELECTRONIC TRANSPORT IN QUASI ONE-DIMENSIONAL NANOSTRUCTURES F. FLORES, B. BIEL, P. SUNDQVIST AND F.J:GARCIA-VIDAL Departamento de Fisica Teorica de fa Materia Condensada, Universidad Autonoma de Madrid 28049-Madrid. Spain

Abstract In this paper we present a brief overview of the different electronic transport regimes occnrring in metallic carbon nanotnbes. Three different lengths determine this transport: the nanotube length, the localization length and the electron mean free path. Depending on the ratio between these three lengths, the transport can be ballistic, diffusive or localized. Regarding the diffusive regime, we analyze theoretically the interplay between the scattering of electrons with both optical and acoustical phonons and how this interaction depends on the applied bias and nanotube length. We also discuss the existence of an intermediate regime, mixture of the localized and diffusive ones that emerges when the length of the nanotube is larger than the localization length but smaller than the electron mean free path.

I.

Introduction

It is a great pleasure to participate with this paper in the special issue

honouring Prof. Paolo Perfetti. This is not the place where to summarize his bright scientific career (other people will do it in detail), we will only mention that his contributions in photoemission, metal/semiconductor interfaces and heterojunctions, surface physics and, more recently, in organic interfaces, solar cells and Si-nanowires have been an inspiration to his many friends and colleagues. But, much more than that, Paolo Perfetti has been a perfect friend, always good humoured, with a lot of energy and trying to make life easy to everybody. A combination of these two things, science and friendship, was mostly shown in the Conferences he organized: the one on the "Formation of Semiconductor

76

77

Interfaces" organized by him in collaboration with S.U.Campisano, G.Margaritondo and O.Bisi in Rome (1991), was a great event which we all will remember for the nice time and great scientific atmosphere we enjoyed. Paolo, thanks a lot for your friendship! Although we have many common scientific interests with Paolo Perfetti, mainly regarding organic and inorganic interfaces, we have chosen to discuss in this paper the electronic transport properties of a particular quasi-dimensional system: a metallic single-walled carbon nanotube. We expect this discussion might be of help for understanding the case of other quasi-dimensional system like the Si-nanowires in which recently Paolo was so much involved [1]. Carbon nanotubes are ideal quasi one-dimensional systems [2] where both science and nano-device applications naturally merge [3-6]. Figure 1 shows these single-walled carbon nanotubes (SWCNT), which can be viewed as formed by bending a graphene layer using a chiral vector. SWCNTs with chirality (n,n) are metallic and they are the systems that we discuss in this paper. SWCNTs with a chirality (m,n), as shown in figure 1, are semiconductors and are not going to be considered here. Figure 1b shows the electronic bands of a (lO,lO)-SWCNT: the fundamental property of this system is that there are two bands crossing the Fermi level, in such a way that the system can be considered as a one-dimensional conductor with two channels contributing to its conductance. As we shall discuss below, their transport properties are strongly dependent on small structural variations, defects and also on intrinsic properties as the electron-phonon interaction.

78 (a)

QUIfII veotor - -

(e)

(6,6)

(t().W)SWCST

Figure 1. Schematic figures showing the geometry of several SWCNT: (a) a semiconductor (11,4), (b) a metallic (6,6). (b) Electron bands for a (10,10) metallic-SWCNT. In this paper we analyze the transport properties of these metallic

SWCNTs and show that they present three distinct regimes: (a) ballistic [7]; (b) diffusive [8-11] and (c) localized [12-13]. Three different lengths characterize these regimes: the nanotube length, L, the localization length, La and the electron mean free path, A, associated with the electron-phonon interaction. Figure 2 shows a scheme of the three different regimes. If L« La, A, electrons propagate ballistically between the two electrodes (figure 2a). When A <
79

fc

Figure 2, Schematic pictures representing different transport regimes (a) ballistic (b) diffusive and (c) localized. In section II, we present some relevant experimental evidence published elsewhere [14] which is analyzed in detail in section III. Our theoretical discussion will be reinterpreted in terms of a kind of phase diagram where the nanotube resistance is described as a function of the three lengths, L, LQ and A, and the bias applied between the electrodes of the wire. Although most of the results presented in this paper have been published elsewhere [14,15], our discussion here tries to be a review, analyzing in a more comprehensive way our understanding of the electronic transport in those quasi-one dimensional SWCNTs. II.

Experimental data.

Figure 3 shows the experimental set-up used by Prof. Gomez-Herrero and colleagues to measure the conductance of a (1O,1G)-SWCNT [16]. In this experiment, a nanotube is deposited on an insulator (quartz) and, with an Atomic Foce Microscope (AFM), the nanotube is localized (see

80

thin line in the right inset of figure 3); then, by techniques, a gold electrode is deposited covering partially [ a contact the nanotube. Finally, using a metal-covered (electrode) is established between the tip and the nanotube narlOtllbe resistance is measured by applying a voltage hpjru"'PTl advantage of this set-up is that, by moving the nanotube resistance is measured as a function of the U
lfYll~'ll,rll

set-up used to measure the (from ref.[l4J).

shows the low voltage, V (around 0.1 ''"'o,'''''''u'",'-' as a function of the nanotube length for two '''..AJ'ilJ ..~O case (black squares), nanotube is a commercial one many second case, the nanotubes a standard chemical vapor deposition (CVD) resistance for the commercial sample follows an Anderson localization regime in however, the plots display a linear delJenldelllce indicative of a typical diffusive .-.. ",u ..",. like R:(h/4e 2)(U /L ) (see the discussion 130 nm can be estimated. In a a high bias regime has been also appears when the V>O.2 V, because then electrons can

a

81

phonons whose energy is around 0.2 eV. Figure 4 shows the resistance, R=V/I, and the differential resistance, dV/dI, as measured for V=O.4, 0.7, 1.0 and 1.5 Vas a function of the nanotube length (dots). 140 120.'

(a)

100-

80

• 3

4

5

6

l().Im)

Figure 4. Resistance and differential resistance as a function of L, for V=O.4 (blue), 0.7 (black), 1.0 (red) and 1.5 V (green). Dots: experimental data. Full lines represent the results from our theoretical approach for the four biases that have been experimentally analyze (from ref [15J).

The behaviour of the Rand dV/dI is strongly dependent on V; while for low V, R shows the typical Ohms-law (see also Fig.3), for high V (larger than 0.2 V) the resistance is larger for higher V, but in all the cases it tends to bend downwards showing at long L, an almost linear behaviour. This is reflected in the dV/dI-curve that shows saturation and a decrease for lengths between 1 and 3 microns. We will see in the next section that

82

this is due to the interplay between the optical and acoustical phonons in their way of controlling the electronic transport in the diffusive regime.

III.

Theoretical analysis of the electronic transport. Voltage and length dependence.

(a) Ballistic transport. Let us start our theoretical analysis by briefly reviewing this limit, whereby electrons move from one electrode to the other crossing the nanotube without suffering any elastic or inelastic interaction (see figure 2a). Electrons have a velocity VF, practically constant around E F, and cross the nanotube in a time LlVF; as the density of states, in a onedimensional system, is: dN/dE=L/hvF one concludes that, per one channel and spin, dIldV is given by edN/dE*evp!L=e 2/h (since edV=dE), a very well-known result: including spin, one channel contributes to the conductance with a value of 2e 2/h. In our quasi one-dimensional nanotube with two channels, the ballistic conductance is 4e 2/h and the contact resistance hl4e 2 (6.3 k,Q). Notice that in this case, the chemical potential is constant along the nanotube and then the voltage drop is localized in the nanotube contacts. As we are using a semiclassical picture, this voltage drop occurs at the same interface: a quantum mechanical calculation shows, however, that this voltage drop extends a few layers around the interface [18]. Moreover, notice that in the symmetrical case shown in figure 2, the voltage drop and the contact resistance is symmetrically distributed between the two contacts.

(b) Diffusive transport The ballistic regime is modified by the interaction of electrons with either defects or phonons. We now discuss the diffusive regime set in the nanotube by the electron-phonon interaction. Phonon excitation by the electrons moving along the nanotube is associated with the inelastic processes shown in Figure 2b, whereby one electron injected in the nanotube with energy E creates a phonon of energy hv, jumps to a final state of energy (E- hv), and is scattered predominantly backwards. Then, due to the successive scattering processes the electron suffers, it goes

83

down in energy and describes a random walk which creates a diffusive regime for the electronic transport within the nanotube. In order to understand the different diffusive regimes created in the nanotube, one has to realize that electrons can excite either optical or acoustical phonons; acoustical phonon energies in OO,lO)-SWCNTs are in the range 0-80 meV, while optical phonon energies are around 200 meV. We consiper first the case of low voltage, V<0.2 V, when only acoustical phonons can be excited in the nanotube. This limit defines a very well-known case (see Figure 3 in the case of very clean samples), with the resistance increasing linearly with the nanotube length. In this classical limit, the nanotube resistance behaves as: R=RoI2*0 +LI Aacc )

(V<0.2 V)

(1),

where RoI2=hl4e 2 is the contact resistance we found in the ballistic limit. In equation 0), Aacc is the mean free path associated with the scattering processes between electrons and acoustical phonons [15]. If V is larger than 0.2 V, optical phonons can be excited in the nanotube and its resistance increases. Consider the case of short nanotubes, say, L
(we

findAoP1 :::::50 nm) [15,19), Then, we can repeat the previous argument,

changing the acoustical phonons by the optical ones, and reach the following equation for the nanotube resistance: R=RoI2*(I+LI AoPI )

(V>0.2 V and L
(2)

There appears, however, a new effect associated with the hot phonons created by the moving electrons in the nanotube [15,20]: due to the electron-phonon scattering processes, the population of phonons, nB, in the nanotube increases with voltage, this effect enlarging the electronphonon scattering rate due to the Einstein's factor: (1+ nB) and nB; then

84

AoP! decreases, and the nanotube resistance increases [14,15]. We have analyzed this effect by solving simultaneously Boltzmann's equations for electrons and phonons. Figure 5a shows the distribution of electrons moving along the x-direction for V=1 V and bO.5 J.lm, while figure 5c shows the mean population of phonons in the nanotube as a function of L, for V= 1.5, 1.0, 0.7 and 0.4 V. Notice the strong dependence of the population of phonons on the applied voltage for 1 V and L::::: 0 ,

(nB) : : : 0.6,

while for 0.4 V and L:::::O,

(nB) : : : 0.2.

This explains the

increase in the resistance with the voltage (see Figure 4), because the scattering rate of electrons with optical phonons increases roughly as (1 +2nB), where creation and anhilation of phonons are taken into account [19]. Figure 5c also shows an interesting effect: increasing the nanotube length, the popUlation of optical phonons decreases. In particular, for very long tubes (L» Aacc ), very few optical phonons can be excited; this can be understood by realizing that in this limit electrons do not have enough energy to excite those phonons. This is illustrated in figure 5b, where the distribution of electrons moving along the x-direction is shown for L=4 J.lm; this solution indicates that the chemical potential, f.1 ' (black dot-dashed line) varies linearly between the two electrodes, and that the electron occupation changes quickly (in lengths of about Aacc) around f.1. This shows that, in the limit of very long tubes (L» Aacc ), electrons can only be scattered off by acoustical phonons. Once, we have explained the different diffusive regimes appearing for short ((L«Aacc) and long (L» Aacc ) nanotubes, we can understand the nanotube resistance experimental data shown in Figure 4. Initially, for short lengths, the resistance is controlled by the optical phonons; when the nanotube length increases, the acoustical phonons start to participate in the scattering of electrons and, eventually, they start to dominate the diffusive regime for lengths much larger than A acc ' namely, for lengths around 1-3 J.lm. Differences between cases with different voltages are mainly related to the hot optical phonons excited in the system: hot phonons increase the scattering rate of the electron-optical phonon

85

and makes the resistance larger for larger voltages. Figure 4 shows, in full lines, the nanotube resistance as calculated equations for electrons and phonons under different conditions voltage and length [15].

Figure 5. Density of electrons moving along the x-direction for (a) Jim and (b) L=4 Jim. (c) Population of optical phonons for (blue), 0.7 (black), 1.0 (red) and 1.5 V (green) as a function of the nanotube length. In the inset, the population of optical phonons is shown as afunction ofx,for V=O.4 (blue) and 1.0 V (red) for three different L/s Jim: full line; L=0.5 Jim: dashed line; L=4.0 Jim: dash-dotted line) {from ref (15J).

86

(c) Localized regime Let us now discuss the role played by the nanotube defects in its transport properties. Consider first the case of a high density of defects such that Lo<
(Lo<
(3),

where Lo is the localization length. This exponential law is the fingerprint of a localized regime as the one observed in Figure 3 for commercial samples. Equation (3) describes the nanotube resistance in the limit Lo«L
87

electron-phonon interaction introduces a diffusive process as electrons move incoherently after their inelastic scattering. Putting together those two arguments leads us to the following resistance in the limit La <,,1,<
(La


This is a regime in which electrons, instead of moving between localized states and the electrodes, jump between localized states assisted by the phonons of the system. This effect reduces the effective tunnelling length to"1,, but introduces a diffusive factor: (1+LI A). Equation (4) can be generalized to include the case A »L in the following way: R=RoI2*(1+LI A )*exp[,,1, LI(,,1, +L)Lal (5).

This is, in principle, a valid equation for all values of L, La and A; however, its use is limited to having a precise knowledge of ,,1,. This is the case, as shown above, for values of L either much smaller than ,,1,acc (then, ,,1,= ,,1,opt ) or much larger than ,,1,acc (then, ,,1,= ,,1,acc ); for other cases, A takes an intermediate value depending on the way optical and acoustical phonons contribute to the electron scattering.

IV.

Conclusions

Summarizing our discussion, electronic transport presents three distinct regimes:

In

(n,n)-SWCNTs

(a) Ballistic: L«"1,, La. In this case, electrons move ballistically between the electrodes without suffering any elastic or inelastic process. Here, R=RoI2, because of the nanotube two channels. (b) Diffusive: A
88

for small voltages, acoustical phonons control the electron conduction; for voltages larger than 0.2 V, however, optical phonons are more effective in scattering the electrons if the nanotube length is smaller than,,1,aee • However, even for high voltages, if L is much larger than,,1,aee ' only acoustical phonons are operative. (c) Localized: Lo «A, L. This regime is controlled by the localized states induced in the nanotube by a high density of defects. We have found that in this localized regime are there two cases: if L« A, we find the strong Anderson localization limit, with R=RoI2*exp(LlLo). On the other hand, if A <
Acknowledgments We acknowledge finantial support from the Spanish CICYT under projects MAT2005-01298 and NAN-2004-09183-ClO-07 and the Comunidad de MadridlFeder project under contract 07N/0050/200 1.

89

References 1. P. de Padova et aI, Nanoletters 8, 2299 (2008) ; P. De Padova et aI, Nanoletters, 8, 271 (2008) 2. S.Ijima, Nature 354,56 (1991) 3. S.J.Ajayan and T.W.Ebbensen, Rep.Prog.Phys.60, 1025 (1997) 4. A.BachtoId, P.Hadley,T.Nakanishi and C.Dekker, Science 294,1317 (2001) 5. AJavey et aI, Nature 424, 654 (2003) 6. J.C.Charlier, X.Blase and S.Roche, Rev. Mod. Phys. 97, 677 (2007) 7. C.T.White and T.N.Todorov, Nature 393, 240 (1998) 8. Z.Yao et aI, Phys. R84, 2941 (2000) 9. J. Y.Park et aI, N anoletters 4,517 (2004) 10. A.Javey et aI, Phys. Rev. Lett. 92, 106804 (2004) 11. E.Pop et aI, Phys.Rev.Lett. 95, 155505 (2005) 12. C.Gomez.Navarro et aI, Nat.Mater. 4, 534 (2005) 13. B.Biei et al, Phys. Rev. Lett. 95, 266801 (2005) 14. P.Sundquist et al, NanoLetters 7, 2568 (2007) 15. P.Sundquist, F.J.Garcfa-Vidai and F.Flores, Phys. Rev.B, in press. 16. C. G6mez-Navarro, PJ. dePablo and J. G6mez-Herrero, Advanced Materials, 16,549 (2004) 17. I. Horcas et aI, Rev. Sci. Instrum. 78, 012705/1-8 (2007). 18. P.Pemas, A. Martin-Rodero and F.Flores, Phys. Rev B, 12, 8553 (1990) 19. M.Lazzeri et aI., Phys.Rev.Lett. 95, 236808 (2007) 20. M.Lazzeri and F.Mauri, Phys.Rev.B 73, 165419 (2006) 21. J.Pendry, Advances in Physics

OPTICAL AND ELECTRON ENERGY LOSS SPECTRA OF LIQUID WATER: AN AB-INITIO STUDY

OLIVIA PULCI A , VIVIANA GARBUIO A , AND RODOLFO DEL SOLE A A European Theoretical Spectroscopy Facility (ETSF), CNR-INFM-SMC, Dept. of Physics University of Rome Tor Vergata, Italy Water is the only common substance found naturally in all three common states of matter: solid, liquid, vapor. It is transparent, tasteless, odorless and ubiquitous. It is at the basis of life. Nevertheless, many issues concerning water are still a matter of debate. In this paper, we present ab-initio calculations of the excited state properties of liquid water in the framework of density functional theory and within the many-body Green's function formalism. In particular, the optical absorption spectrum with the inclusion of excitonic effects is calculated by solving the BetheSalpeter equation. The applicability and the accuracy of first-principles methods are also discussed.

1. Introduction

Water is the most common liquid on the Earth's surface. It is essential for living organisms survival and for biological systems in general. Accordingly, it is one of the most extensively studied systems from both a theoretical and experimental point of view. However, despite the single water molecule is a very simple system, many basic aspects of this substance are not fully understood and even the structure of liquid water is presently under strong debate 1,2,3,4,5,6,7,8,9,10,11,12,13,14,15. Many biochemical and industrial processes occur in solution; hence it is crucial to include the role of the solvent in the reactions to gain a deep understanding of these processes. Moreover, the electronic properties of water, as a solvent, are extremely interesting since water can influence many electronic events (electron transfer, photo-excitation) of solute molecules by its dielectric response 16,17 or by actively participating in electronic processes, i.e., in electron-transfer phenomena at biological interfaces 18. The study of the excited state properties of liquid water is therefore fundamental to advance in many research fields. Since the absorption spetrum of water is in the UV region, experimental work is now favoured by the access to the powerful Synchrotron Radiation sources built in recent years. From the theoretical point of view, despite some pioneering studies

90

91

19,20,21,22, the high computational cost needed to address the modeling of a disordered system such as a liquid substance has limited, until recently, the studies of the electronic structures and properties to the gaseous phase 23,24 or to crystalline ice 25,26. At the same time, recent theoretical studies of liquid water 21,27,28,29,30,3,31,32,33,34 have mostly focused on its structure and ground state properties whereas less effort has been dedicated to its electronic structure and optical absorption spectrum. As a consequence, experimental data about excited states are still poorly understood. In the present decade, aided by the rapid increase in computational power, a series of theoretical works on the electronic properties of liquid water have appeared 35,6,36,37,38,39. In this context, we have generalized the application of the Many Body Perturbation Theory (MBPT)40 to liquid systems, and presented a calculation of the optical absorption spectrum of liquid water from first principles, including both self-energy effects and the electron-hole i-nteraction 41 . We showed the occurrence of important excitonic effects, which are crucial for a good description and interpretation of experimental data. We have also calculated the Electron Energy Loss (EEL) spectrum at the Density Functional Theory (DFT) level, which is in good agreement with experiments. In this paper, after a brief introduction to the theoretical and computational schemes used, we presents results for the optical and EEL spectra of liquid water. 2. Green's function theory Green's function theory is particularly suitable for studying excited state properties and hence for interpreting or predicting spectroscopic experimental results. Details of the theory can be found, for example, in Refs. 42,40 2.0.1. Quasi-particle equations In the Lehmann representation, it can be shown that the poles of the Green's function are the electron addition and removal energies, that is, the energy levels of unoccupied and occupied states, respectively, as measured for example in inverse and direct photoemission experiments. For practical calculations, a single-particle-like framework is regained by introducing the concept of quasi-particles (QP) which can be thought of as real particles plus a polarization cloud, due to electron-hole pairs, surrounding them and screening the mutual interaction. The difference between "bare" particles (subject only to the Hartree potential) and quasi-particles

92

can be accounted for by the self-energy operator ~ which is a non-local, non-Hermitian, energy-dependent operator. A Schrodinger-like equation for the QP can be written:

Ho (r)'ljJn (r, w) where Ho(r) = -~V'; has to be found.

+ J dr'~(r, r', w)'ljJn(r', w) = En (w)'ljJn(r, w),

(1)

+ Vext(r) + VH(r);

an adequate expression for ~

2.0.2. GWapproximation It can be shown that the QP equation is equivalent to a Dyson-like equation for the Green's function:

G(1,2)

=

Go(l, 2)

+

J

d(34) Go(l,

3)~(3, 4)G(4, 2).

This is the first equation of a closed set of five equations proposed by Hedin 43,44, the others being: ~(l,

2)

= i J d(34) G(l, 3)r(3, 2, 4)W( 4,1 +);

W(l, 2) = V(l, 2) + J d(34) W(l, 3)P(3, 4)V(4, 2); P(1,2)

= -i

J d(34) G(l, 3)G( 4, 1+)r(3, 4, 2);

r(1, 2, 3) = <5(1,2)<5(1,3) +

J d(4567)

~~t~:;~ G(4, 6)G(7, 5)r(6, 7, 3);

where 1+ stands for (rl, 0"1, tl +<5) and <5 is an infinitesimal positive number. This set of equations also involves the time ordered polarization operator P(l, 2), the dynamical screened Coulomb interaction W(l, 2) and the vertex function r(l, 2, 3). These equations must be solved self-consistently to obtain the exact solution, a procedure that is practically impossible for realistic systems and hence some simplifications have to be found. The simplest approximation consists of starting with a non-interacting system with ~ = 0; in this case G = Go, the vertex correction is neglected and P(1,2) = -iGo(l, 2)Go(2, 1). Hence the self-energy becomes ~(1, 2)

= iGo(l, 2)Wo(2, 1+).

(2)

This is the so-called GW approximation. In principle more iterations should be performed but calculations usually stop at this first step (one-shot GW), and obtain quite accurate results for one-particle excitations.

93

2.0.3. Bethe-Salpeter equation With regards to absorption spectra, it is important to take into account the interactions between holes and electrons by means of the inclusion of vertex corrections. This can be achieved through a second iteration of Hedin's equations, which gives for the vertex the expression

r(123)

=

J(12)J(13)

+ iW(1 +2) J d(67)G(16)G(72)r(673).

(3)

This equation can be transformed into an integral equation for a four-points generalized polarizability by introducing four-point quantities, i.e. 4 P(1234), 4W(1234) = W(12)J(13)J(24) and 4 Po(1234) =

Po(12)J(13)J(24). The Bethe-Salpeter equation for the polarizability can be then derived and gives:

(4) The kernel K is made of two terms: an electron-hole exchange contribution involving the bare potential V, and the electron-hole attraction due to the screened potential W:

K(1234)

=

J(12)J(34)V(13) - J(13)J(24)W(12).

(5)

In practical calculations, an effective two-particle excitonic Hamiltonian is constructed from eq. (4). The eigenfunctions and eigenvalues of this Hamiltonian build up the absorption spectrum. The neutral excited states of the system are now represented as a linear combination of electron-hole couples and the position and shape of the absorption spectrum are deeply modified with respect to the independent quasi-particle spectrum. Details of its derivation can be found in Ref. 40. Bound exciton states within the gap, as well as a distortion of the above-gap absorption spectrum, are the results of the calculations.

3. Computational details The disorder in the liquid system has been modeled by averaging the results over many congurations. We used 20 snapshots of 17 water molecules in a cubic box, with 15 a.u. side. These configurations have been obtained by sampling every 2 ns a 40 ns long classical molecular dynamics (MD) simulation trajectory. A TIP3P water model potential 45 has been used. Equations of motion have been integrated numerically using a time step of 1 fs. The MD run has been done in the NVT ensemble, where thermal equilibrium at 298 K has been achieved applying a Nose-Hoover thermostat

94

Electrostatic interactions were treated using Particle-Mesh-Ewald; all van der Waals interactions between non-bonded atom pairs were included. The electronic states have been first obtained in the DFT framework, within the generalized gradient approximation (GGA), and on top of this calculation, the energy levels have been corrected within the GW approximation to take fully into account exchange and correlation effects; finally the optical absorption spectrum including excitonic effects has been calculated by solving the Bethe-Salpeter equation (BSE). We perform pseudopotential DFT simulations within GGA-PW91. We use 8k points in the Brillouin zone and a kinetic energy cutoff of 50 Ry. GW corrections have been calculated using 19933 plane waves for the exchange part of L:, and 13997 plane waves for its correlation part. The screened W has been calculated within the plasmon pole model, using 600 empty bands. The quasi-particle energies were calculated in first order perturbation theory. Optical absorption spectra at the BSE level have been obtained including all the filled states and 100 empty levels. The resulting 20 excitonic Hamiltonians (one for each MD conguration) have been treated using the Haydock algorithm. 46,47.

4. Optical absorption

The first optical absorption experiments on water date back to the sixties and seventies 48,49,50,51,19. In these works the dielectric constant of water, and hence its optical absorption spectrum, were deduced from KramersKronig analyses of reflectance measurements. Other ultraviolet absorption experiments on liquid water were performed in those years, for example in Refs. 52,53,54,55. More recently, the optical spectrum of water has been measured by low resolution dipole spectroscopy 56 and by inelastic X-ray scattering 57,58 at low momentum transfer, using a synchrotron radiation light source. Examples of these experimental spectra are shown in figure 1, whereas a closer examination of the low energy region, where excitonic effects are more important, is in figure 2. We have calculated 41 the optical absorption spectrum of liquid water according to many-body perturbation theory, through the solution of the Bethe-Salpeter equation. The electronic states have been obtained within DFT for all the 20 MD snapshots and the relative independent particles absorption spectra have been calculated. Then the energy levels have been corrected within the GW approximation. As shown in Ref. 41, the GW corrections are almost identical fot the various MD configurations; hence they have been calculated just for one snapshot and then they have been applied

95

1.5

W"" 1

0.5

10

20

30

40

50

E (eV) Figure 1. Experimental optical absorption spectra of liquid water, from Ref. circles), 50 (red stars), 51 (black diamonds), 58 (green squares).

49

(blue

to all the MD configurations. Finally the optical absorption spectra including excitonic effects have been calculated by solving the Bethe-Salpeter equation for the 20 MD snapshots. The averaged DFT, GW and BSE spectra are shown in figure 3. As expected, the DFT optical spectrum (solid line) shows strong discrepancies with respect to experiment both in the onset and in the lineshape; the overall effect of the GW corrections (dotted line) is to over-shift the DFT spectrum towards higher energies, without improving its shape. The BSE spectrum (dashed line), on the contrary, shows a significant improvement in the agreement with experiment both in the peak positions and in the onset, as well as in the relative intensities of the first two peaks. The first peak is attributed to a bound exciton with a binding energy of 2.4 eV and large oscillator strength. The second peak results from an excitonic enhancement of the oscillator strength of inter band transitions with respect to the single quasiparticle case.

96

...f:>'"

/{

~.

1

.. '

.

.~

0.5

5

6

10

789

11

12

E (eV) Figure 2. Low energy region of the experimental optical absorption spectra of liquid water, from Ref. 49 (blue circles), 50 (red stars), 51 (black diamonds), 58 (green squares).

In Ref. 41, the absorption spectrum of liquid water has also been calculated within TDLDA showing no significant improvement with respect to that obtained within DFT-LDA.

5. Energy Loss Energy Loss Spectroscopy is mainly directed at the observation of plasmons, i.e., the collective excitations of electrons as a response to external perturbations. The loss function is related to the dielectric constant through the relation 1

loss ex Im( - -) C

=

C2 -2--2' c1 c2

+

(6)

The loss function of ice has been measured by electron energy loss experiments, for example in Ref. 59,60

97 3

,, ,, ,, , ,, ,, /\

I

\

2

w'"

I

"'"

\ \

\

I ,

1

....

I

4

5

6

7

8

•

9

10

11

12

E(eV) Figure 3. Optical absorption spectrum of liquid water calculated within DFT (solid black line), GW approximation (dotted red line) and by solving the Bethe-Salpeter equation (dashed blue line); from Ref. 41.

For what concerns liquid water, the loss function has been obtained in Ref. 51 by optical reflectance measurements and, more recently, in Ref. 57, from X-ray scattering experiments: in both studies the main peak is at an energy (the plasmon frequency) of about 22 eV. The loss spectra presented in Ref. 57, for different values of transferred momentum, are shown in figure

4. It is well known that plasmons are well described already at the DFT leve1 62 . Hence, we calculated the energy loss spectrum of liquid water within DFT, for all the 20 MD snapshots61 . The averaged spectrum is shown in figure 4, for a transferred momentum q=O. A good agreement with experiment is reached for what concerns the peak position and lineshape even if, looking at the onset and in general at the low energy region of the spectrum, where excitonic effects are more important, the agreement is less satisfactory.

98

1

I=: 0.8

o ...... ..... c..> I=:

~

4-;

0.6

v:J v:J

o

...... 0.4

0.2

10

20

30

40

50

E (eV) Figure 4. Loss function of liquid for different transferred momenta: from Ref. 57. The solid blue line DFT for a transferred momentum

water obtained with X-ray scattering measurements q=O.19 (circles), q=O.53 (stars), q=O.69 (diamonds)j is the loss function of liquid water calculated within q=Oj from Ref. 61.

6. Final Remarks

In recent years, the electronic and optical properties of liquid water have been the subject of several computational studies. Furthermore, the constant increase in computational power has allowed scientists both to improve the level of theory for the solution of the quantum problem and to infer better statistical averaging to take into account the properties of the liquid disordered phase. In this framework, we have calculated the optical properties of water within MBPT. In this study we achieved a good description of the electronic and optical properties of liquid water. Development of future techniques, algorithms and theoretical approaches aimed at decreasing the computational costs of excited state calculations are desired for future studies on the subject. Time Dependent Density Functional Theory (TDDFT) may be such a technique. At present, it is

99

mostly applied within local (TDLDA) or semilocal (TDGGA) approximations, and it represents a quick and valid alternative to Many-Body Perturbation Theory for small molecules and clusters. However, it heavily fails in describing excitations in extended systems. On the contrary, new TDDFT exchange-correlation kernels based on Many-Body approaches have been successfully used in a variety of systems 63,64,65,66,67. Unfortunately, they do not represent, yet, a computationally convenient alternative to a GW plus BSE calculation since the construction of such extremely non local and frequency dependent kernels causes a remarkable increase in the computational cost. Consequently, efforts should be devoted to speed up the calculation of these kernels, in order to make their usage computationally less demanding with respect to the GW plus BSE method, and competitive with, but more accurate than, the presently widely used local or semilocal approximations to TDDFT.

References 1. P. Wernet, D. Nordlund, U. Bergmann, M. Cavalleri, M. Odelius, H. Ogasawara, L. A. Naslund, T. K. Hirsch, L. OJ arnae , P. Glatzel, L. G. M. Pettersson, and A. Nilsson, Science 304, 995 (2004). 2. J. D. Smith, C. D. Cappa, K. R. Wilson, B. M. Messer, R. C. Cohen, and R. J. Saykally, Science 306, 851 (2004). 3. P. L. Silvestrelli and M. Parrinello, J. Chern. Phys. 111, 3572 (1999). 4. B. Hetenyi, F. De Angelis, P. Giannozzi, and R. Car, J. Chern. Phys. 120, 8632 (2004). 5. A. K. Soper, F. Bruni, and M. A. Ricci, J. Chern. Phys. 106, 247 (1997). 6. A. Hermann, W. G. Schmidt, and P. Schwerdtfeger, Phys. Rev. Lett. 100, 207403 (2008). 7. M. Iannuzzi, J. Chern. Phys. 128, 204506 (2008). 8. M. Cavalleri, M. Odelius, D. Nordlund, A. Nilsson, and L. G. M. Pettersson, Phys. Chern. Chern. Phys. 7, 2854 (2005). 9. D. Prendergast and G. Galli, Phys. Rev. Lett. 96, 215502 (2006). 10. T. Head-Gordon and M. E. Johnson, Proc. Nat. Acad. Sci. USA 103, 7973 (2006). 11. T. Head-Gordon and S. W. Rick, Phys. Chern. Chern. Phys. 9,83 (2007). 12. T. Tokushirna, Y. Harada, O. Takahashi, Y. Senba, H. Ohashi, L. G. M. Pettersson, A. Nilsson, and S. Shin, Chern. Phys. Lett. 460, 387 (2008). 13. M. Leetrnaa, M. P. Ljungberg, H. Ogasawara, M. Odelius, L.-A. Naslund, A. Nilsson, and L. G. M. Pettersson, J. Chern. Phys. 125, 244510 (2006). 14. M. Leetrnaa, K. T. Wikfeldt, M. P. Ljungberg, M. Odelius, J. Swenson, A. Nilsson, and L. G. M. Pettersson, J. Chern. Phys. 129, 084502 (2008). 15. M. Odelius, M. Cavalleri, A. Nilsson, and L. G. M. Pettersson, Phys. Rev. B 73, 024205 (2006). 16. C. J. Cramer and D. G. Truhlar, Chern. Rev. 99, 2161 (1999). 17. J. Tomasi, B. Mennucci, and R. Carnrni, Chern. Rev. 105, 2999 (2005).

100 18. 19. 20. 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. 47. 48. 49. 50.

J. Lin, I. A. Balabin, and D. N. Beratan, Science 310, 1311 (2005). F. Williams, S. P. Varma, and S. Hillenius, J. Chern. Phys. 64, 1549 (1976). R. J. Buenker and S. D. Peyerimhoff, Chern. Phys. Lett. 29, 253 (1974). K. Laasonen, M. Sprik, M. Parrinello, and R. Car, J. Chern. Phys. 99, 9080 (1993). D. Borgis and A. Staib, Chern. Phys. Lett. 238, 187 (1995). N. E. Henriksen, J. Zhang, and D. G. Imre, J. Chern. Phys. 89, 5607 (1988). A. Balkova and R. J. Bartlett, J. Chern. Phys. 99, 7907 (1993). G. Pastori Parravicini and L. Resca, Phys. Rev. B 8, 3009 (1973). L. Resca and R. Resta, phys. stat. sol. b 81, 129 (1977). O. Christiansen, T. M. Nymand, and K. V. Mikkelsen, J. Chern. Phys. 113, 8101 (2000). J. C. Grossman, E. Schwegler, E. W. Draeger, F. Gygi, and G. Galli, J. Chern. Phys. 120,300 (2004). M. E. Tuckerman, K. Laasonen, M. Sprik, and M. Parrinello, J. Phys.: Condens. Matter 6, A93 (1994). H. Sato and F. Hirato, J. Chern. Phys. 111, 5845 (1999). B. Chen, I. Ivanov, M. L. Klein, and M. Parrinello, Phys. Rev. Lett. 91, 215503 (2003). P. Raitieri, A. Laio, and M. Parrinello, Phys. Rev. Lett. 93, 087801 (2004). L. Bernasconi, J. Blumberger, M. Sprik, and R. Vuilleumier, J. Chern. Phys. 121, 11885 (2004). D. Prendergast, J. C. Grossman, and G. Galli, J. Chern. Phys. 123, 014501 (2005). I. Tavernelli, Phys. Rev. B 73, 094204 (2006). D. Nordlund, M. Odelius, H. Bluhm, H. Ogasawara, L. G. M. Petterson, and A. Nilsson, Chern. Phys. Lett. 460, 86 (2008). G. Brancato, N. Rega, and V. Barone, Phys. Rev. Lett. 100, 107401 (2008). J. C. Schwegler, E. Grossman, F. Gygi, and G. Galli, J. Chern. Phys. 121, 5400 (2004). D. Lu, F. Gygi, and G. Galli, Phys. Rev. Lett. 100, 147601 (2008). G. Onida, L. Reining, and A. Rubio, Rev. Mod. Phys. 74, 601 (2002). V. Garbuio, M. Cascella, L. Reining, R. Del Sole, and O. Pulci, Phys. Rev. Lett. 97, 137402 (2006). A. L. Fetter and J. D. Walecka, Quantum Theory of Many-Particle Systems (McGraw-Hill, New York, 1971). L. Hedin, Phys. Rev. 139, A796 (1965). L. Hedin and S. Lundqvist, Solid State Phys. 23, 1 (1969). W. L. Jorgensen, J. Chandrasekhar, J. D. Madura, R. W. Impey, and M. L. Klein, J. Chern. Phys. 79, 926 (1983). S. J. Nose, J. Chern. Phys. 81, 511 (1984). W. G. Hoover, Phys. Rev. 31, A1695 (1985). R. Onaka and T. Takahashi, J. Phys. Soc. Japan 24,548 (1968). L. R. Painter, R. D. Birkhoff, and E. T. Arakawa, J. Chern. Phys. 51, 243 (1969). G. D. Kerr, R. N. Hamm, M. W. Williams, R. D. Birkhoff, and L. R. Painter, Phys. Rev. A 5, 2523 (1972).

101

51. J. M. Heller, Jr., R. N. Harnrn, R. D. Birkhoff, and L. R. Painter, J. Chern. Phys. 60, 3483 (1974). 52. T. I. Quickenden and J. A. Irvin, J. Chern. Phys. 72, 4416 (1980). 53. T. Shibaguchi, H. Onuki, and R. Onaka, J. Phys. Soc. Japan 42, 152 (1977). 54. R. E. Verral and W. A. Senior, J. Chern. Phys. 50, 2746 (1969). 55. D. P. Stevenson, J. Phys. Chern. 69, 2145 (1965). 56. W. F. Chan, G. Cooper, and C. E. Brion, Chern. Phys. 178,387 (1993). 57. H. Hayashi, N. Watanabe, Y. Udagawa, and C.-C. Kao, J. Chern. Phys. 108, 823 (1998). 58. H. Hayashi, N. Watanabe, Y. Udagawa, and C.-C. Kao, Proc. Nat!. Acad. Sci. U.S.A. 97, 6264 (2000). 59. M. Michaud, P. Cloutier, and L. Sanche, Phys. Rev. A 44, 5624 (1991). 60. C. D. Wilson, C. A. Dukes, and R. A. Baragiola, Phys. Rev. B 63, 121101 (2001) . 61. V. Garbuio, M. Cascella, and O. Pulci, J. Phys.: Condens. Matter (in press). 62. V. Olevano and L. Reining, Phys. Rev. Lett. 86, 5962 (2001). 63. F. Sottile, V. Olevano, and L. Reining, Phys. Rev. Lett. 91,056402 (2003). 64. G. Adragna, R. Del Sole, and A. Marini, Phys. Rev. B 68, 165108 (2003). 65. A. Marini, R. Del Sole, and A. Rubio, Phys. Rev. Lett. 91, 256402 (2003). 66. D. Varsano, A. Marini, and A. Rubio, Phys. Rev. Lett. 101, 133002 (2008). 67. O. Pulci, A. Marini, and R. Del Sole, to be published.

ELECTRONIC CONFINEMENT OF SILVER NANOCLUSTERS IN Er3+-ACTIVATED SILICATE AND PHOSPHATE GLASSES L. MINATI\ G. SPERANZA! IFBK-IRSTvia Sommarive 18, 38050 Povo, Italy

A. CHlAPPINf, A. CHIASERA2, M. FERRARI2 2

CNR-IFN, CSMFO Lab. Via alla Cascata, 561C, 38050 Povo-Trento, Italy

S. BERNESCHl3 S. PELLI\ G.c. RIGHINI 3,4 4

3 CNR, Department of Materials and Devices, via dei Taurini 19,00185 Roma, Italy MDF Lab., Nello Carrara Institute ofApplied Physics, !FAC - CNR, Via Madonna del Piano 10, 50019 Sesto Fiorentino (Firenze), Italy

In this work the silver chemical state in Ag-exchanged Er3+ doped silicate and phosphate glasses is analyzed. X-Ray Photoelectron spectroscopy (XPS) shows that thermal annealing leads to a different behavior of silver atoms in these two systems. In silicate glasses silver aggregates in nanopartic1es. In phosphate glasses silver is in an oxidized state and in very small metallic nanoparticles. These results are corroborated by the fact that the surface plasmon absorption band is observed in the absorption spectra of silicate glasses, whereas is absent in the case of the phosphate glasses. These evidences explain also the occurrence or the lack of Er'+ luminescence enhancement in the silicate and phosphate silver-exchanged glasses, respectively.

In the middle 80's Malta et al. observed a radical modification of the Eu3+emission when Ag atoms are inserted in the glassy network [1]. After this pioneering work the mechanism underlying the enhancement of rare-earth photoluminescence (PL) has become one of the hot topics in plasmonics. The main problem concerns the role played by silver in the PL enhancement and the understanding of the mechanism. In the last few years it was demonstrated that the radiation-induced Surface Plasmon Resonance (SPR) occurring in metal clusters is of capital importance for a strikingly diverse range of applications. [2]. Recently it was demonstrated that the enhancement effect strongly depends on the structural organization and on the chemical state of silver [3]. Two processes have been proposed to explain the luminescence enhancement in rare earth doped glasses containing silver: i) local field enhancement induced by the plasmon resonance; ii) energy transfer from the metal clusters to the rare earth ions. Concerning the latter process, Polman et al. [4] suggested that in Er3+ implanted borosilicate glasses luminescence enhancement is due to energy transfer from Ag + centers rather than Ag nanoclusters. In this scenario, XPS analysis of glassy systems may help to clarify the mechanisms underlying the luminescence enhancement, since it provides both chemical and structural information. In this work we have studied Ag ionexchanged silicate and phosphate glasses doped with Er3+ ions, with the aim of identifying the chemical state of silver and correlating it with the Er3+ 102

103

spectroscopic features of the two systems. In XPS the different chemical states are identified by changes of the binding energy (BE), namely the chemical shift, induced by changes of the oxidation state of the atom. The photoemission process is sensitive to changes of the electronic configuration when the long range order typical of bulk metals approaches the nanometric dimensions. When the size of metallic nanoparticles decreases below 5 nm, the core-line BE shifts to higher values [5, 6]. It has been demonstrated that, by using an appropriate calibration, it is possible to correlate the extent of BE shift to the nanocluster dimension [7]. The molar composition of the rare-earth doped silicate glass is: 71.5 Si02, 15 Na20, 10.4 CaO, 1.2 Ah03, 0.4 P 20 5, 0.6 K20, 0.3 Er203, 0.6 Yb 20 3. The phosphate glass has the following molar composition: 65 NaP0 3, 20 Nb 20 5, 15 Ga203, 3 Er203. For each kind of glass two plates, with a thickness of 200 !lm, were cut and optically polished. One of these two plates was Ag-exchanged and the other was kept as reference; the latter ones have been labeled as SAgEr-ref, and PAgEr-ref, respectively. The silicate glass plate was ion-exchanged for 67 hours at 390°C in a molten salt bath of molar concentration 0.5% AgN0 3, 99.5% NaN0 3. Finally, the sample, labeled SAgEr, was annealed at 500°C for 60 min. The phosphate glass plate was ion-exchanged at 280°C for 65 hours using a bath of 0.5% AgN0 3, 49.75% NaN0 3, and 49.75% KN0 3 mol%. After the ionexchange the sample, labeled PAgEr, was annealed for 60 min at 400°C. Absorption measurements have been performed using a ultra-violet-visible-near infrared spectrophotometer (Cary 5000) in dual beam mode. Photoluminescence (PL) measurements in the region of the 411312 - 411512 transition of the Er3+ ion were obtained upon excitation wavelength of 476.5 nm. Great care was taken in these experiments to keep unmodified the experimental conditions, when measuring the PL intensity from the Ag-exchanged glass plate and from the corresponding reference one, in order to make it possible to quantitatively compare the PL intensities [8]. The XP spectra were acquired in a Scienta ESCA 200 instrument equipped with a monochromated Al Ka (1486.6 eV) X-ray source and a 200 mm hemispherical analyzer. The BE scale was calibrated with respect to the Ag Fermi edge. Spectra were acquired at a pass energy of 150 e V, corresponding to an energy resolution of 0.35 eV taking into account the effects of charge compensation. The carbon contamination peak was chosen as an energy reference to calibrate the BE scale. Because in XP spectra of the phosphate samples the Nb3P312 core line is overlapped to the Ag 3d 5/2 peak, all the spectral analyses were performed on the Ag 3d312 component. This is allowed because the lineshape, the energy splitting (6.03 eV) and the intensity ratio between the two spin-orbit components of the Ag 3d are independent of the atom's chemical state [9]. In Figure 1 the Ag 3d 312 component of the SAgEr sample is presented. The BE falls at 374.44 eV, i.e. 0.24 eV higher than that of bulk Ag. Moreover, we observe a line broadening that is associated to the intrinsic disorder derived from the distribution of the particle size. The BE value of the Ag 3d312 core line for the

104

SAgEr sample is very different from the BE value reported in literature for Ag 20 and indicated by the dashed line in Fig. 1 [10]. This observation confirms the formation of Ag nanoparticles. The BE shift of the Ag 3d peak allows us to estimate the cluster dimensions in the SAgEr sample. Using data from literature we calculated a mean nanoparticles diameter of about 2 nm [11]. The Agexchange process leads to the formation of Ag + atoms bonded to the silica host matrix. The annealing at 500°C for 60 min increases the silver mobility in the matrix and induced a silver precipitation in metal nanoparticles. 0.7 - , - - - - - - - - - - - - - - - - - - - ,

Ag 3d 3/2

~ 0.6

.= = of~ 0.5

b

~ 0.4 r.ol

Eo-<

~

0.3

377

376 375 374 373 BINDING ENERGY (eV)

372

Figure 1: Ag 3d 3/2 core line of the SAgEr silicate sample annealed at 500°C for 60 min. The dashed line indicates the BE value of Ag20. This process has been widely studied in literature. A realistic reaction for the reduction of the silver ions was proposed by Wang et al. [12]: (1)

These authors pointed out that this process is thermodynamically favorable and kinetically fast at high temperature due to the tensile strength induced by the greater dimensions of the Ag + respect to the Na + ions. This driving force leads to a diffusion of the silver toward the surface and its precipitation in metallic nanoparticles.

105

Ag 3d 3/2 ~ 1.0

.=

..c=

Ag I I

$'"'

-

~ 0.5

r Jj

Z

f;I;l Eo-<

Z

0.0

378

377

373 375 374 376 BINDING ENERGY (eV)

372

Figure 2: Ag 3d3/z core line of the PAgEr phosphate sample annealed at 400°C for 60 min. The dashed line indicates BE value of the bulk silver. In Figure 2 we report the Ag 3d3/z component of the PAgEr sample, together with the result of the peak fitting with Voigt functions. It is immediately clear that substantial differences exist in respect to the Ag 3d3/z spectrum of the SAgEr sample. In the case of the phosphate glass two different silver chemical states are present. The component at lower BE is assigned to silver bonded to oxygen atoms of the phosphate network. Its BE of 373.93 eV is in perfect agreement with the BE value of AgzO [10]. The second Ag 3d312 component falls at a 375.58 eV, which is blue shifted of about 1.38 eV in respect to that of pure bulk silver, indicated by the dashed line in Fig. 2. This component is assigned to very small metal nanoclusters embedded in the glassy matrix. The higher BE shift in comparison to that measured for the SAgEr sample puts in evidence a stronger effect of quantum confinement, i.e. smaller nanocluster dimensions. Referring to data reported in literature [11] a mean cluster dimension lower than 1 nm was estimated. This is a very special situation in which the XP spectrum displays the early stage of nucleation of silver into nanoclusters. In our opinion there are three possible reasons explaining the different behavior of silver in silicate and phosphate glasses: the lower annealing temperature of the PAgEr sample with respect to the SAgEr (400 °C instead of 500 0C); the lower tensile strength of silver in phosphate with respect to silicate. the higher electron-donor nature of silver silicate bond in comparison with the silver phosphate one. The first point cannot explain by itself the differences between SAgEr and PAgEr samples, even if the kinetics of formation of silver nanoparticles in silicate-glass is high also at temperatures around 250 °C [l3].

106

The last two points deserve more attention. During the ion-exchange process, the silicate glass is deformed by the difference between the Na+ and Ag+ sizes, being the ratio of the respective ionic radii r(Ag+)/r(Na+)= 1.29. The tensile strength induced by this difference leads to the migration of the silver ions to the surface, with successive precipitation in metallic clusters. In phosphate glasses the tensile strength is lower and the silver ions can be easily accommodated in the glass network. This is confirmed by the work of Kern et al. which shows that the introduction of P 20 5 in silica glasses decreases the tensile strength [14]. The last issue concerns the different chemical interactions that take place between silver ions and the two host media. In silica-based glasses the electronegativities of silicon and silver are comparable, leading to a great electron density on the oxygen atoms. On the contrary the electronegativity of phosphorus, being much higher than that of silicon, causes a higher level of oxidation of the silver ions. This is confirmed by the lower value of optical basicity in phosphate glasses than in silicate ones. This parameter, as defined by Duffy et ai, represents the electron-donor ability of the oxygen atoms of the glassy environment [15]. Referring to the work of Takashi et al. [16], that simulates the structure of sodium silicate and phosphate glasses, we can point out another important difference between the silicate and the phosphate networks. Following Takashi, due to the great overlap of P(d)n - O(p)n the P-O bond assumes a double-bond character. In phosphate glasses this makes the negative charges of oxygen atoms more delocalized with respect to those of the Si-O bonds, leading to a stronger covalent interaction between the 0 and the silver ions. These indications are confirmed by Le Flem et aI., who report a fast precipitation of silver into nanoparticles in Ag-enriched zinc phosphate glasses when a low quantity of Si0 2 or Al 20 3 is added [17]. On the basis of these experimental and theoretical evidences we can effectively explain the different behavior of Er3+ luminescence in silicate and phosphate glasses. Figure 3 shows the absorption spectra of the reference and silverexchanged silicate samples. The SAgEr-ref sample is transparent over a wide range, down to 300 nm, and the sharp peaks observed in its absorption spectrum are due to Er3+4f transitions from the 4115/2 ground state to the excited states. The absorption spectrum of the SAgEr sample shows a shoulder at around 420 nm, assigned to the surface plasmon absorption band, which testifies the presence of silver nanoparticles inside the silicate matrix [8]. The inset of Figure 3 clearly shows an enhancement of the 411312 -> 411512 Er3+ fluorescence under 476.5 nm laser excitation, which is a wavelength not in resonance with the Er3+ electronic levels. For what concerns the mechanism of PL enhancement, we do not observe any significant increase of the absorption cross section of Er3+ transitions in SAgEr with respect to the reference sample. This observation indicates that Er3+ photoluminescence enhancement does not originate from an increase of the absorption cross-section in Er3+ ions when they are subject to the strong local electromagnetic field generated by the surface plasmon excitation of silver nanoparticles, but it is promoted by energy transfer

107

from silver nanoparticles to the Er3+ ions. This model is in agreement with the XPS results, which detect the presence of Ag nanoparticles in the SAgEr sample.

0.20 1.4

SAgEr ref --SAgEr

os'"

0.16

-SAgEr ..... ~ SAgEr ref

1.2 1.0

;::1

~ 0.8

(].)

- ; 0.6

U

'0;

ro 0.12

_

=

= .-e

1:1

0 C/)

0.4 0.2

0.0 1700 1650 1600 1550 1500 1450 1400

~

~ 0.08

Wavelength (nm)

\ 0.04

\.

.'

;..\;\.-.......,...;.

300 400 500 600 700 800 900 1000 11 00 1200

Wavelength (nm)

Figure 3: Room temperature absorption spectra of the SAgEr-ref (dotted line) and SAgEr (solid line). The spectra are vertically shifted for clarity. Inset: PL spectra of SAgEr-ref (dotted line) and SAgEr (solid line) upon excitation at 476.5 nm. The situation is different in the case of phosphate glasses. Figure 4 shows that there is no difference between the absorption spectra of PAgEr-ref and PAgEr. The XP analysis indicates that a consistent part of silver is in an oxidized form. The remaining part corresponds to very small nanoclusters. In the case of nanoparticles of very small dimension the surface plasmon absorption band noticeably broadens and the maximum intensity strongly decreases [18], explaining the absence of a prominent plasmon absorption band for the PAgEr sample. For this reason the energy transfer process from silver nanoparticles to the Er3+ ions in the PAgEr sample is not so efficient as in the case of SAgEr sample. This is shown in the inset of Figure 4 where no difference appears in the

108

Er3+ emission spectra, obtained under 476.5 nm laser excitation, for the reference and Ag-exchanged phosphate samples, respectively.

411312 ---+ 411512

0.5

r--.------------;:========;_] 1.4

-PAgEr

•

0.4

--PAger ~ PAgEr ref .~

1.0

'" ~

0.8

;

0.6

Il)

'v;

I:: 0.3

5

0.4

,s

0.2

u

ro

PAgEr reI

12

-B0 CIl

~

<

Wavelength (run)

0.2

0.1 400

600

800

Wavelength (nm) Figure 4: Room temperature absorption spectra of the PAgEr-ref (dotted line) and PAgEr (solid line). The absorption spectra are vertically shifted for clarity. Inset: PL spectra of PAgEr-ref (dotted line) and SAgEr (solid line) upon excitation at 476.5 nm. In conclusion, XPS and optical analysis on Ag ion-exchanged SAgEr and PAgEr glasses doped with Er3+ ions indicate that Ag nanoparticles play an essential role in the process of PL enhancement. The combination of the results obtained from optical and XPS characterizations allowed us the identification of the energy transfer process from silver nanoparticles to Er3+ ions as the principal mechanism of the luminescence enhancement. Acknowledgments. We wish to express Paolo Perfetti our whole, sincere appreciation for his inspiring enthusiasm and dedication. This research was performed in the framework of the COST Action MP0702, PAT (2007-2010) FaStFAL research project, and EFONGA Coordination Action.

109

References [1] O. L. Malta, P. A. Santa-Cruz., G. F. De Sa and F. Azuel, J. Lumin. 33,261 (1985). [2] M. Pelton, J. Aizpurua, and G. Bryant, Laser & Photon. Rev. 2, 136 (2008). [3] G. Speranza, S.N.B. Bhaktha, A. Chiappini, A. Chiasera, M. Ferrari, C. Goyes, Y. Jestin, M. Mattarelli, L. Minati, M. Montagna, G. Nunzi Conti, S. Pelli, G.C. Righini, C. Tosello and KC. Vishunubhatla, J. Opt. A: Pure Appl. Opt. 8, S450 (2006) [4] C. StrohhOfer and A. Polman, Appl. Phys. Lett.81, 1414 (2002). [5] G. K Wertheim, S. B. DiCenzo and D. N. E. Buchahan, Phys. Rev. B 33, 5384 (1986). [6] G. K Wertheim and S. B. DiCenzo, Phys. Rev. B 37, 844 (1988). [7] L. Minati, G. Speranza, L. Calliari,V. Micheli, A Baranov, S. Fanchenko, J. Phys. Chem. A 112, 7856, (2008). [8] H. Portales, M. Matterelli, M. Montagna, A. Chiasera, M. Ferrari, A. Martucci, P. Mazzoldi, S. Pelli and G.c. Righini, J. Non-Cryst. Solids 351, 1738 (2005). [9] D. Briggs and 1. Grant, Surface Analysis by Auger and X-Ray Photoelectron Spectroscopy 1M Publication 2003 p. 47. [10] NIST X-ray Photoelectron Spectroscopy Database', the NIST Standard Reference Database 20, Version 3.4 (Web Version). [11] K Luo, T. P. St. Clair, X. Lai and D. W. Goodman, J. Phys. Chem. B 104, 3050 (2000). [12] P.W. Wang, Appl. Surf. Sci. 120,291 (1997). [13] E. Borsella , G. De Marchi, F. Caccavale, F. Gonella , G. Mattei, P. Mazzoldi, G. Battaglin, A. Quaranta and A. Miotello J. Non-Cryst. Solids 253,261 (1999). [14] W. Kern, G. L. Schnable and A. W. Fisher, RCA Rev. 37,3 (1976). [15] 1.A. Duffy and M.D. Ingram, 1. Amer. Chem. Soc. 93, 6448 (1971). [16] U. Takashi and Y. Ogata, J. Non-Cryst. Solids 181,175 (1995). [Tomokatsu1999] H. Tomokatsu, S. T. Selvan and M Nogami, Appl. Phys. Lett. 74,1513 (1999). [17] I. Belharouak, C. Parent, B. Tanguy, G. Le FIem and M. Couzi, J. NonCryst. Solids 244, 238 (1999) [18] P. Mulvaney, Langmuir 12, 788 (1996).

DYNAMICS AT METAL/SEMICONDUCTOR INTERFACES AND EXOTIC PHENOMENA THROUGH THE LOOKING GLASS

GUYLELAY CINaM-CNRS, Campus de Luminy, Case 913, F-13288 Marseille cedex 9, France, and Universite de Provence, Marseille, France [email protected]

In the early 80's, and for more than a decade, the formation of metal/semiconductor interfaces was a central issue in semiconductor physics, with the aim of solving the intriguing Schottky barrier problem. In this context, where Synchrotron Radiation Photoelectron Spectroscopy emerged at first generation machines as the most powerful investigation tool, Paolo Perfetti was one of the pionneers who initiated the study of the «other side of the problem », i.e., the early stages of the growth and of the development of the electronic properties of inverse semiconductor/metal interfaces, typically Si/polycristalline Au [1]. Few years ago, in Marseille, we renewed this approach using silver single crystal surfaces as substrates, with the novel advanced tools developped in the mid 80's: Scanning Tunneling Microscopy and Spectroscopy as well as High-Resolution SR-PES. «Through The Looking Glass », we discovered a cornucopia of exotic phenomena at Ge/Ag(lll), Ge/Ag(100) and Ge/Ag(llO) interfaces [2-6], as well as at Si/Ag(100) and Si/Ag(110) ones [7-11]. Weird, onedimensional, massively parallel nano-ribbons were discovered

110

111

specific all y at the Si/Ag( 11 0) interface [9-11], and further characterized in great details by HR-SR-PES at the VUV beamline at Elettra, the high brilliance third generation Italian storage ring in Trieste, by P. De Padova et al., [12,13], while theoretical Density Functional Calculations by K. Kara et al., evidenced that these silicon nano-ribbons are actually graphene-like, honeycomb, oneatom thick silicon sheets, i.e., true silicene stripes [14]. In the 90's, another very strange behaviour was discovered at the apparently « simple» and supposedly identical SnlGe,Si(111yhx3 reconstructed surfaces, consisting of one third of a monolayer of tin adatoms positionned in threefold atop sites. Although a single Sn 4d core-level component should have been normally expected, instead, two components were obtained in each case, but with an inverse intensity ratio 1 :2 versus 2 : 1 for Ge( 111)-V3x-V3-Sn versus Si(I11)-V3x-V3-Sn, giving a mirror symmetry for the Sn core-level lines on both surfaces [15]. A tentative explanation of this mystery was suggested: a «phonon assisted charge oscillation» between neighgouring Sn adatoms, that is, a dynamic fluctuation proposed for the first time at a metal/semiconductor interface. Soon after, Carpinelli et al., uncovered a reversible -v3x'h ¢:::> 3x3 phase transition at about 2000K on Ge(I11); they assigned the new 3x3 superstructure to the first evidence of a static surface charge density wave [16,17]. However, the metallic character of the 3x3 phase contradicted this hypothesis, which, in addition, could not explain the preservation of the same two Sn 4d components for both -V3x-V3 and 3x3 phases. For this reason, the present author suggested at the ICFSI-6 conference in Cardiff, UK [18] a novel dynamic phenomenon, which implied a vertical oscillating motion of the Sn adatoms through a kind of Sp2/Sp 3 rehybridization process [19]. This idea of vertical oscillations was conforted by the DFf calculations and the molecular dynamics simulations of Flores' group in Madrid, and then widely recognized [20]. Still, an experimental confirmation had to be given; it was obtained through very delicate STMISTS measurements carried out at room temperature and low temperatures (down to 2.5 K) by Ronci et al.

112

[21] and Colonna et ai. [22] showing, especially, telegraph noise type current time traces, as for the dynamics of oscillating dimers at the Si(100)2xl surface [23]. Furthermore, quantum oscillations were also discovered below - 15 K at the Si(I11)-Y3x-Y3-Sn surface [24]. One could think that this is the end of the story, yet, we will see that this is just the emerged part of the iceberg! Several issues remain a puzzle, especially the assignment of the two Sn 4d components at the Ge(111)-Y3x-Y3-Sn surface and their mirror presence at the Si(111)-Y3x-Y3-Sn counterpart, not to mention their symmetric behaviours upon doping with donor and acceptor species [25-28 ]. [1] Au-Si interface formation,' the other side of the problem, A. Franciosi, D.W. Niles, G. Margaritondo, C. Quaresima, M. Capozi and P. Perfetti, Phys. Rev. B 32 (1985) R6917 [2] Ge/Ag( 111) semiconductor-on-metal growth,' formation of an Ag2 Ge surface alloy, H. Oughaddou, S. Sawaya, J. Goniakowski, B. Aufray, G. Le Lay, J.M. Gay,G. Treglia, J.P. Biberian, N. Barret, C. Guillot, A. Mayne and G. Dujardin, Phys. Rev. B 62 (2000) 16653 [3] Germanium adsoption on Ag( 111),' an AES-LEED and STM study, H. Oughaddou, A. Mayne, B. Aufray, J.P. Biberian, G. Le Lay, B. Ealet, G. Dujardin and A. Kara, J. Nanosci. Nanotechnol., 7 (2007) 1 [4] Ge tetramer structure of the p(2 v2x4 v2 )R(45 0 ) surface reconstruction of Ge/Ag(OOI) " a surface X-ray diffraction and STM study, H. Oughaddou, J.M. Gay, B. Aufray, L. Lapena, G. Le Lay, O. Bunk, G. Falkenberg, J.H. Zeysing and R.L. Johnson, Phys. Rev. B 61 (2000) 5692 [5] Self-organization of Ge tetramers on Ag(OOI) surface,' a 2D realization of unsual substrate mediated interactions, H. Oughaddou, B. Aufray, J.P. Biberian, B. Ealet, G. Le Lay, G. Treglia, A. Kara and T.S. Rahman, Surface Sci. 602 (2008) 506 [6] Self-assembled germanium nano-clusters on silver (110), C. Leandri, H. Oughaddou, J.M. Gay, B. Aufray, G. Le Lay, J.P. Biberian, A. Ranguis, O. Bunk and R. L. Johnson, Surface Sci. 573 (2004) L369 [7] Growth of Si nanostructures on Ag(OOJ), C. Leandri, H. Oughaddou, B. Aufray, J.M. Gay, G. Le Lay, A. Ranguis and Y. Garreau, Surface Sci., 601 (2007) 261 [8] Ordered silicon structures on silver (100) at 230 0 e, c. Leandri, B. Aufray, G. Le Lay, C. Girardeaux, C. Ottaviani and A. Cricenti, J. Phys. IV France, 132 (2006) 311

113

[9] Self-aligned silicon quantum wires on Ag( 11 0), e. Leandri, G. Le Lay, B. Aufray, C. Girardeaux, J. Avila, M.E. Davila, M.e. Asensio, e. Ottaviani and A. Cricenti, Surface Sci. 574 (2005) L9 [10] Silicon quantum wires on Ag( ]10): Fermi surface and quantum well states, M.A. Valbuena, J. Avila, M.E. Davila, C. Leandri, B. Aufray, G. Le Lay and M.e. Asensio, Surface Sci. 254 (2007) 50 [11] Formation of a one-dimensional grating at the molecular scale by selfassembly of straight silicon nanowires, H. Sahaf, L. Masson, e. Leandri, B. Aufray, G. Le Lay and F. Ronci, Appl. Phys. Lett., 90 (2007) 263110 [12] Growth of straight, atomically perfect, highly metallic silicon nanowires with chiral asymmetry, P. De Padova, e. Quaresima, P. Perfetti, B. Olivieri, C. Leandri, B. Aufray, S. Vizzini and G. Le Lay, Nano Lett., 8 (2008) 271 [13] Burning match oxidation process of silicon nanowires screened at the atomic scale, P. De Padova, C ; Leandri, S. Vizzini, C. Quaresima, P. Perfetti, B. Olivieri, H. Oughaddou, B. Aufray and G. Le Lay, Nano Lett., 8 (2008) 2299 [l4] Evidence of epitaxial growth of silicene nano-ribbons, A. Kara, e. Leandri, B. Ealet, H. Oughaddou, B. Aufray and G. Le Lay, submitted [l5] Metal-semiconductor fluctuation in the Sn adatoms in the Sir III )-Sn and Ge(l11)-Sn (V3x-Y3)R30° reconstructions, M. Gothelid, M. Bjorkqvist, T. M. Grehk, G. Le Lay, and U. O. Karlsson, Phys. Rev. B 52, R14352 (1995) [l6] Direct observation of a surface charge density wave, J.M. Carpinelli, H.H. Weitering, E. W. Plummer, R. Stumpf, Nature 381 (1996) 398 [17] Surface charge ordering transition: alpha phase of Sn/Ge( ]]]) J.M. Carpinelli, H.H. Weitering, M. Bartkowiak, R. Stumpf, and E.W. Plummer, Phys. Rev. Lett. 79 (1997) 2859 [I8] 6th International Conference on the Formation of Semiconductor Interfaces, Cardiff, UK, 1997. The ICFSI series, launched in Marseille, France, by Guy Le Lay (Chairman) and Jacques Derrien (Secretary) in 1985. ICFSI-3 was organized in Rome, Italy, in 1991 and chaired by P. Perfetti, while ICFSI-9 was organized in Madrid, Spain, in 2003 and Chaired by F. Flores. [19] Surface charge density waves at Sn/Ge( ]]1)? G. Le Lay, V.Y. Aristov, O. Bostrom, J.M. Layet, M.e. Asensio, J. Avila, Y. Huttel and A. Cricenti, Appl. Surf. Sci. 123 (1998) 440 [20] Dynamical fluctuations as the origin of a surface phase transition in Sn/Ge(l11)? J. Avila, A. Mascaraque, E.G. Michel, M.e. Asensio, G. LeLay, J. Ortega, R. Perez and F. Flores, Phys. Rev. Lett. 82 (1999) 442 [21] Direct observation of Sn adatoms dynamical fluctuations at the Sn/Ge( III ) surface, F. Ronci, S. Colonna, Thorpe S.D., A. Cricenti and G. Le Lay, Phys. Rev. Lett., 95 (2005) 156101 [22] Metallic nature of the a-Sn/Ge( 111) surface down to 2.5 K, S. Colonna, F. Ronci, A. Cricenti and G. Le Lay, Phys. Rev. Lett., under press

114

[23] Dynamics of dimers and adatoms at silicon and germanium surfaces, G. Le Lay, V.Y. Aristov, F. Ronci, S. Colonna and A. Cricenti, in Brilliant light in life and material sciences book series: NATO security through science series b: physics and biophysics (2007) 329 [24] Evidence of Sn adatoms quantum tunneling at the alpha-Sn/Si( 111) surface, F. Ronci, S. Colonna, A. Cricenti and G. Le Lay, Phys. Rev. Lett. 99 (2007) 166103 [25] 1nfluence of charged impurities on the surface phases of Sn/Ge( 111), M.G. Rad, M. Gothelid, G. Le Lay, V.O. Karlsson, T.M. Grehk and A. Sandell, Surface Sci. 477 (2001) [26] Complex behaviors at simple semiconductor and metal/semiconductor surfaces, M.E. Davila, J. Avila, M.e. Asensio and G. Le Lay, Surf. Rev. Lett. 10 (2003) 981 [27] Giant effect of electron and hole donation on Sn/Ge( 111) and Sn/Si( 111) surfaces, M.E. Davila, J. Avila, M.e. Asensio and G. Le Lay, Phys. Rev. B 70 (2004) [28] Perturbation of Ge( 111) and Sir 111 )root 3 alpha-Sn surfaces by adsorption of dopants, M.E. Davila, J. Avila, M.e. Asensio, M. GOthelid, V.O. Karlsson and G. Le Lay, Surface Sci. 600 (2006) 3154

SUPRA MOLECULAR INTERACTION OF CHIRAL MOLECULES AT THE SURFACE G. CONTINI*, N. ZEMA, P. GORI, A. PALMA+, F. RONCI, S. COLONNA, S. TURCHINI, D. CATONE, A. CRICENTI, T. PROSPERI Istituto di Struttura della Materia, CNR, Via Fosso del Cavaliere 100, 00133 Roma, ItaLy +Istituto

per Lo Studio dei Materiali Nanostrutturati, CNR, Via Salaria Km 29.3, 00016 Monterotondo S. (RM), ItaLy

Abstract Two-dimensional supramolecular chemistry on surfaces is strongly governed by directional non covalent forces. The chirality of the system plays an important role, especially in the two-dimensional case due to the confinement in the plane; a strong influence on the self-assembly pattern formation is provided by the absence of certain symmetry elements. For small flexible chiral organic molecules with two heteroatoms a very large self-assembly chiral domain governed by supramolecular interactions mediated by surface potential can be obtained on symmetric metallic surfaces. In the case of the adsorption of D-alaninol (2-amino-l-propanol) on Cu(lOO) surface, molecule-surface interaction may occur through both the amino and the hydroxyl groups or just involving one of them. Adsorbed alaninol molecules have been structurally and electronically characterized as a function of the surface molecular coverage by photoelectron spectroscopy (for core levels and valence region) and scanning tunneling microscopy (STM). The comparison of the experimental results with density functional theory calculations provides further insight into the D-alaninoUCu( 100) adsorption mechanism.

* corresponding author: E-mail: [email protected]

115

116

Introduction

In recent years, an increasing interest has been focused on the surface modification induced by the adsorption of organic molecules on it due to the technological relevance of using molecular adsorbates on surfaces as biosensors, catalysts, in polymer technology, adhesion, and activations of immune cells. A wide class of organic molecules is also capable to add to the substrate special chemical properties due to their chirality. Surface chirality has received increasing attention since 1990s. The study of the heterogeneous catalysis is one of the driving forces for these efforts because of the high potentiality for drug synthesis [1]. Moreover, supramolecular chemistry on crystalline surfaces is largely controlled by lateral interaction, although the substrate plays an important role in mediating them. Only if the adsorbate-substrate interaction allows the molecules to "feel" each other, intermolecular recognition on the surface may take place. The adsorption energy of a single molecule is modulated laterally due to t~e presence of the crystalline surface potential. In order to migrate on the surface, the molecule must overcome the substrate potential and at high coverage the intermolecular interaction becomes prominent and influences the molecule-substrate interactions. The saturation coverage is reached when the amount of repulsion energy within one molecular layer becomes as strong as the adsorption energy of a single adsorbed molecule. Under these conditions, the steric influence due to the adsorbed chiral molecules becomes large and constrains the obtained self-assembled molecular pattern. In some cases, an energetically favored site for single adsorbed molecules may switch to a different binding site when high packing density is reached [2, 3]. In this respect, the interplay between lateral and molecular-substrate interactions determines the two-dimensional self-assembled molecular pattern. The simultaneous presence of chirality and supramolecular effects on the system obtained by the adsorption of the simplest chiral amino alcohol, namely alaninol (2-amino-l-propanol) on Cu(lOO) surface, provides a very interesting system to be studied. The bifunctional nature of alaninol allows the possibility of double interactions with the surface through both the amino and the hydroxyl group, favored also by the fact that the N-O distance in gas-phase alaninol (2.73 A) is comparable with the side length of the surface unit cell of Cu( 100) (2.56

A). It has been shown that alaninol adsorbs on Cu( 100) forming a selfassembled monolayer (SAM) with long-range order [4]. If the D-enantiomer of the molecule is adsorbed, the LEED pattern shows a (4,-111,4) phase of alaninol, leading to a clockwise rotation of 14 degrees of the molecular phase with respect to the [011] direction of the metal surface. This phase is characterized by a surface structural unit that appears to be a tetramer, as evidenced by STM measurements, attributed to four alaninol molecules in view of its dimensions (3.8 Aand 4.4 Apeak to peak distances along two orthogonal directions [4]).

117

has been also characterized by photoelectron (XPS) [5] providing information on the differences of C 0 Is and N Is core-level spectra obtained at low (sub-monolayer) and at full (monolayer) molecular coverages. The main point of interest is the appearance, at monolayer coverage, of a second peak located at lower in the N Is """UH,'>,> energy, the positions of the two peaks being at 399.4 eV and 397.5 eV. This the possibility of two types of interaction of alaninol molecules with the copper surface. Le. through both amino (NH2) and imino (NH) groups, the latter interaction motivating the presence of the new peak in XPS measurements. The modeling of the D-alaninoVCu(100) system by calculations in the framework of Density Functional Theory (DFT) information complementary to the experimental results. In this work, the submonolayer coverage phase has been analyzed considering the adsorption sites of.a single molecule on the Cu(lOO) surface. Some hints on the monolayer coverage phase will also be provided by considering the possible adsorption of a dehydrogenated alaninol molecule in order to provide a model that can help in the interpretation of the photoelectron data obtained by XPS and UPS experunents. CnF'l'trn<1'V""'"

Computational details First-principles calculations have been carried out in the framework using the Generalized Gradient Approximation (GGA) to the Perdew-BurkeErnzerhof (PBE) exchange-correlation functional [6]. Electron-ion interaction is described through ultrasoft pseudopotentials [7]. In the case of Cu, 3d electrons are considered as valence electrons and nonlinear core correction is applied. Wavefunctions are expanded in a plane-wave basis set with a kinetic energy cutoff of 25 The code employed for all calculations is Quantum ESPRESSO [8].

FIG. 1: 3-D representation of the most stable conformer of D-alaninol

118

The Cu(100) surface has been modelled by using a slab geometry, made of three Cu layers subject to periodic boundary conditions (PBC), with the bottom layer constrained during relaxation to its bulk position. Such model was successfully used in previous simulations on Cu(100) [9]. D-alaninol is placed on the top layer. The adopted supercell has a 4x4 surface unit cell containing 16 Cu atoms per layer. This gives a lateral distance between replicas of the molecule of 10.224 A, and describes a low coverage of 8 = 0.0625. A vacuum layer of 10.8 A (corresponding to 7 atomic layers) has been included to avoid spurious interactions with images. The Surface Brillouin Zone (SBZ) has been sampled by a uniform Monkhorst-Pack mesh of 4x4 k-points [10], reduced to 10 k-points by symmetry. The Fermi level has been evaluated using the MethfesselPaxton technique with a smearing parameter of 0.26 eV [11]. Dipole correction has been applied to cancel the artificial field originated by imposing PBC on the electrostatic potential [12]. Supercell parameters have been kept fixed during the optimization and nuclei relaxation has been performed until forces were reduced below 0.005 eVl A. N Is and 0 Is relative core level shifts (CLS's) for the most stable configurations of D-alaninol adsorbed on the Cu( 100) surface have been calculated following a procedure based on DFT total energy differences within a pseudopotential plane waves technique [13]. In this approach, already applied to inorganic and organic systems [14-17], the core-hole binding energies (BE's) are evaluated as differences between ionized and neutral states in the same geometry. The relative BE's are then obtained as differences with respect to a reference configuration. In this way we take into account the electronic relaxation effect in the final state. In this study the calculated CLS's are relative to the most stable local minimum found (see below) which is chosen as a reference configuration.

Geometrical structure

The study of alaninol conformers as isolated molecule and their optimization has been the first step of the geometrical structure determination. The obtained results are in good agreement with previous M~ller-Plesset second order (MP2) calculations for L-alaninol [18]. The above procedure has been applied to D-alaninol and the geometrical structure of the most stable conformer is described in Fig. 1. To study the adsorption process, eleven initial configurations have been considered, with alaninol laying horizontally on the surface (i.e. 0 and N atoms at comparable distances from Cu(100) surface) or vertically (perpendicular to the surface plane), with either the amino or the hydroxyl group close to the surface. The distance between 0 and N in gas-phase alaninol (computed as 2.73 A), which almost matches the side length of Cu(100) surface unit cell (2.56 A,

119

in the [110] direction), might facilitate a double interaction between the molecule and the surface. Horizontal configurations can therefore have the alaninol molecule with 0 and N atoms in the following positions: on top of first Cu atomic layer, on top of second Cu atomic layer (hollow site), in bridge position along one side of the surface unit cell, in bridge position along the diagonal of the unit cell, on top along the diagonal of the unit cell. In the last case, obviously, on top has to be intended as an approximate definition, since the distance between 0 and N is smaller than the diagonal of the Cu( 100) surface unit cell. Vertical configurations taken into account present alternatively either the amino or the hydroxyl group closer to the surface. N or 0 atom is put in a top, bridge or hollow position. Furthermore, we consider two different orientations of the vertical approach: one along the side and the other along the diagonal of the surface unit cell. Molecule's initial configurations in which the methyl group CH3 is the closest to the surface bring to alaninol repulsion (as found, for example, also for alanineINi(lll) [19].) The same behaviour occurs when the molecule is in vertical position with the amino group being the closest to copper. After relaxation, the eleven different initial geometries converged to three local minima. These structures will be referred in the following as Horizontal Top (HT), Horizontal Diagonal Top (HDT) and Vertical with Oxygen on Top (VOT). The adsorption energy per molecule is defined as /).£

= ECu(IOO) + Ealaninol - Ealaninovcu(100) ,

(1)

where ECu(IOO) is the total energy of the relaxed copper surface, Ealaninol is the total energy of the isolated alaninol molecule and EalaninoVCu(100) is the total energy of the adsorbate/substrate system. The values of the adsorption energy (measured in eV/molecule) for the three stable configurations are 0.60, 0.51 and 0.45 respectively for HT, HDT and VOT configurations. These quantities account for the energy gain related to the bonding of the molecule to the metal surface. Each value is in the range found for similar systems and here the adsorption through the alcoholic group gives results ofthe same order of magnitude [19, 20]. The HT geometry gives the most stable adsorption configuration (Fig. 2), where both 0 and N interact with the copper surface. Atomic distances found in this case are Cu-N = 2.14 A and Cu-O = 2.29 A. The Cu-N distance, in particular, is comparable with that of systems in which chemical adsorption takes place (see, for example, glycine adsorbed on Cu(llO) [21]). The Cu atom beneath N relaxes outwards relative to the ideal bulk terminated surface in the z direction. The height of the Cu atom bound to N with respect to the average plane of the other Cu surface atoms is L1ZciN) = 0.18 A. The corresponding variation for 0 is L1zcu(O) = 0.08 A.

120

a)

b)

FIG. 2: (a) side and (b) top views of HT geometry (only the upper layer of the Cu(100) slab is shown for clarity in (b». Oxygen is red, nitrogen turquoise, carbon hydrogen cyan and copper orange. It has been ascertained that ammonia, activated by oxygen, can transform in amine and imine species on Cu(llO) surface as a function of surface temperature and oxygen coverage [22]; on the other hand, an amino molecule, the monomethylamine (CH3NH2), can adsorb on clean Cu(211) at temperature close to RT as amino and imino groups [23]. Following these works, another structure has been considered in which one amino hydrogen atom is removed to yield an imino group (NH) which then binds to the metal. From a computational point of view, it has been verified that the same final structure is obtained after relaxation detaching either one or the other H atom bound to nitrogen in D-alaninol molecule. The final geometry obtained (see Fig. 3) will be referred in the following as HNBOT (Horizontal N Bridge 0 on Top). The results can be summarized in the following way: the Cu-O distance is reduced compared to the most stable configuration HT (bond length of 2.22 A), N binds in a bridge position between two Cu atoms (to which is linked by stronger bonds: CuN=1.99 A) and the detached H atom is located at a fourfold hollow site of Cu(lOO).

121

Ge()mt~try

of HNBOT «a) side view and (b) top view) adsorbed D-alaninol on the top layer of Cu(100) is shown in the top view. Notice the hvrlrn,,,pn detached from the amino group in a Cu(lOO) hollow site.

This in which the imino group is bound to copper is 0.32 eV less stable than the HT configuration and therefore is very _......_.J when a molecule is approaching the copper surface. Npvp.rthp where self-assembled fourfold units of alaninol molecules are observed in a very ordered overlayer reconstruction interactions could favour the dehydrogenation of the NH z group the energy barrier between HNBOT and HT COlltH~ur'itlCm It has been tested if the removed hydrogen prefers to stick to the copper with the H of a neighbouring cell. It results that the former surface or to form is favoured by 0.21 eV, which corresponds to the difference E(HNBOT)noH + 0.5E{Hz) E(HNBOT)H; where is the energy of the system with the detached H excluded from the simulation cell and E(HNBOT)H is the one with H in the simulation cell and close to the copper surface. Electronic .....,,,........ti"'.,

Information about the regions where charge rearrangement occurs as a consequence of molecule-surface interaction can be gained by the spatially resolved electronic charge density difference Ap(r)

p(r)cu(IOO)+aJaninol - p(r)cu(IOO) - p(r)aJauinol ;

122

where p(r)cu(100)+alaninol is the charge density of the adsorbate/substrate and p(r)cu(lOO) and p(r)alaninol are charge densities of each one of the two cornp()ne:nts with atoms in the positions of the adsorbed Positive values of Ap(r) provide information on accumulation of while t'lP,rrM",,'" numbers describe the charge depletion regions. These for the lIT corltl~~ur;atl(m are in Fig. 4 and show how the from the ....E',." ••" around 0, Nand Cu becomes localized between Nand Cu and between o and Cu, that mainly covalent bonds are as also supported the structural analysis of previous section. The situation for the HNBOT is also reported, for comparison, in 4: two accumulation appear between N and its nearest .". ",,,,,,.vv •.u a interaction between molecule and surface. In the subsection we will show how this simple analysis can provide an efficient tool to int,f>rr,rp.t the photoelectron spectra. Core levels Core-level photoelectron spectroscopy is sensitive both to chemical to evaluate and to local environment of an atom. It can therefore be bond variations in reacted systems or in different n"'!'.m .... w'!'. D-alaninoVCu(lOO), it has been found eXI)erlmc~ntally from sub-monolayer to monolayer coverages, N Is core-level separated by approximately the new into two located at lower energy CLS's calculations have been n""'j-n,-rn"rl the previously described methodology.

FIG. 4: \""',I-"""VH of charge), red for positive values for HT and 0.027 e/A3 for HNBOT geometry.

123

In order to understand the chemical origin of this splitting, it has first been speculated on the possibility that it can be due to the coexistence of different ways of anchoring the molecule to the surface in the monolayer phase. Calculations have shown that N Is core-level shifts correlated with different orientations on the surface of the adsorbed D-alaninol molecule are quite small, of the order of 0.1 eV (see Table 1). Therefore, this cannot be the origin of the observed experimental shift. A further possibility is instead related to a more chemical reason like a dehydrogenation process occurring at the amino group, as shown in literature for other systems [22, 23]. To validate this possibility, the HNBOT geometry can be considered as a good candidate for simulations. The N Is core level shift resulting from the comparison between the HT and the HNBOT geometry amounts to 2.26 eV, which is not far from the experimental value of 1.9 eV. This suggests that, in the monolayer regime, the intermolecular interaction may lead to the coexistence of D-alaninol molecules having both amino and imino groups (i.e. different H saturation of N atom) inside the molecular fourfold unit anchored to the Cu(100) surface. Unfortunately, our model, based on a single molecule adsorption simulation, cannot provide information about the driving mechanism, probably strongly affected by intermolecular cooperative forces, which possibly triggers oxidation or chemical changes on Nitrogen atoms upon deposition. Finally it has to be observed that, comparing the same two local minimum configurations, HT and HNBOT, the 0 Is level experiences essentially no shift (4 meV calculated), which is consistent with experimental data of Ref. [5]. TAB. 1. Relative N Is core level shifts calculated with respect to the HT configuration.

X Configuration

VOT

HDT

HNBOT

Exp.ABE (full cov.)

-0.11

-0.03

2.26

1.9

CLS : BE(HT)-BE(X) (eV)

Valence band density of states The electronic structure of a single alaninol molecule adsorbed on the Cu(100) surface has been further analyzed by computing the valence band density of states (DOS) for the HT configuration as well as for the HNBOT configuration. Fig. Sa displays valence band photoelectron data of the clean Cu( 100) surface, of sub-monolayer and of monolayer molecular coverages. To allow a closer comparison with photoelectron measurements, calculated DOS's have taken into account only the eigenvalues at the r point. In Fig. 5b-c we show the DOS projected (pDOS) over the atomic orbitals of some selected atoms of the adsorbate/substrate system for the HT and for the HNBOT configuration, respectively.

124

A

between experimental and calculated spectra should be into account that some source of discrepancy between them may come from the use of DFf eigenvalues to describe DOS and band structure these limitations in mind, it is anyway possible to extract useful information from the comparison, in particular regarding details of adsorbatesubstrate interaction.

non-intEmacting Gu 3d states CuS 3d states CuW 3d states N 2pstates

021' stales

nOfHnt:era.cTll1!Q Gu states Cu63dstates Cu1 0- 11 3d states N 2p states 021' stales

FIG. 5: valence band for clean Cu(lOO), submonolayer, and mOinOjlaYI~r D-alaninol coverage. (b) pDOS on selected atomic orbitals for the HT (c) on selected atomic orbitals for the HNBOT configuration. The Cu curve has been obtained as an average over Cu atoms not involved in bondings with a1anino!. For a better PDOSs have been multiplied by a factor of 2 for N 2p and o states. The zero of the energy scale is at the Fermi level, but theoretical curves have been shifted toward binding energy by 0.8 eV to align Cu main 3d

125

The experimental valence band spectrum in Fig. Sa obtained for selfassembled monolayer shows, after comparison with the clean copper spectrum, the appearance of several new structures due to the interaction of D-alaninol with the surface copper atoms. The main structures appear at -1.9 eV, -2.3 eV, 4.S eV, marked in Fig. Sa as "a", "b", "c" respectively. On the other hand, some of these structures are not visible in the valence spectrum recorded in the submonolayer regime, witnessing a different chemistry at the surface; the main features observed are the "b" and "c" peaks. In order to compare theoretical and experimental results for sub-monolayer to monolayer coverages, Fig. Sb reports the theoretical pDOS's projected on selected atoms of the system: Cu atoms non interacting with D-alaninol, Cu atoms bound to 0 (Cu6) or to N (CulO in HT, CulO and Cull in HNBOT; see Figs. 2 and 3 for Cu atoms numbering), nitrogen and oxygen atoms of the molecule. Note that for a better comparison, the theoretical energy scale has been shifted by 0.8 eV toward higher binding energy in order to align Cu main 3d peak. We discuss first the part of the spectra around Cu main 3d peak. At submonolayer coverage, the two features "b" and "c" are also present in the theoretical DOS of the HT configuration and are attributed to copper-oxygen interaction. In particular, the "b" shoulder can be associated to 3d states of the Cu6 atom bound to 0 found at -2.3 eV and "c" to the 0 2p level at -4.S eV. It is worth noting that these two structures remain unaltered in the HNBOT configuration, although a small N 2p contribution is also present, in accordance with the experimental results obtained for the 0 Is core line BE, which does not change significantly as a function of coverage. In the saturated self-assembled monolayer regime, a peak at -1.9 eV, the "a" structure, appears in the experimental spectrum. The HT pDOS does not present any signal around this energy, whereas in the HNBOT configuration a clear structure appears both in the 3d pDOS of the two copper atoms bound to N (CulO and Cull) and in the N 2p pDOS.

126

of the alaninoIlCu(100) system in HNBOT correslDOllldillljl; to the -3.2 eV (a) and -1.0 eV correspOrtUUljl; to -4.0 and -1.8 eV in the shifted energy scale).

\"U'H'i'\'l.Ua'"Ull

the N state, further information can be obtained from calculations. The theoretical DOS of isolated D-alaninol ascribes the HOMO-l level to N When the interaction between D-alaninol and the as in the HNBOT the DOS copper surface is on the N state two of structures, above and below eu 3d states: the first is located at -4.0 eV and contributes to the "c" the other is a double at -1.8 and -1.2 eV and to the Ha" feature. These states can be identified as and as follows from the NewnsAnderson model to adsorbates on metal surfaces with narrow d bands. A can be deduced from the formation of these two states, which to the mixed the fact that appear both in the of these states is shown in wavefunction at the r to the P.o,,,,... ,,,,1 eV and -1.8 eV in the shifted energy clear borldulg 'un....''''''... ,,,/'; characters are In the energy range from -5 to -11 several other structures are found both in the distribution and in the theoretical DOS. For a better of these 7 of Nand 0 atoms obtained for HT should be of the self'UU'HVJ,aY'vl situation. It is found that these structures can be associated to non mteralctlng alaninol orbitals because of the lack of copper states in this energy These states are out over the entire molecule with the main contribution '-U"U"l": from carbon states.

127

u;-

'2:::I

9

N2pslales - " 0 2p states - - C 2p states

8

-

C 2p + N 2p + 0 2p

, I

6

I I

.05

I

1

~

(/)4

o0

3 2

O~~~~-L~~L-~~-L~~~L-~~~~~~~~~

-13

-12

-11

-10

-9

-8

-7

-6

-5

-4

-3

-2

-I

0

Energy (eV)

FIG. 7: pDOS of 2p carbon, oxygen and nitrogen summed for HT and HNBOT configurations,

Scanning tunneling microscopy and spectroscopy In order to go deeper into the investigation of the adsorption of Dalaninol/Cu(lOO), STM measurements on D-alaninol SAM, obtained as described in ref. [4] have been taken at room temperature using a tungsten tip in the constant current mode (0.1 V, 0.2 nA for the image shown in Fig. 8b).

128

b)

FIG. 8: curves averaged over points of the types A and B (m
'V,"'''U'S

129

seen in the STM image as a single bright protrusion and the main contribution to the image comes from the part of the molecule that is closer to the observer. This is related to the fact that, because of the chemisorption regime that is attained and because of the metallicity of the whole system, molecular orbitals are smeared out over the continuum of copper metallic states and proximity effects result to be dominant in the image formation. The simulated image also shows that it is reasonable to explain the experimentally observed tetramer with the presence of four adsorbed alaninol molecules. Similarly, in the HNBOT configuration the simulated STM image (Fig. 8d, obtained for a bias of 0.1 V) displays essentially a bright spot, but accompanied by a quite dimmer side spot, related to the fact that in this case the molecule is more tightly bound to the substrate and two parts with comparable heights over the surface are singled out.

Conclusions

Supramolecular chemistry and chirality in two-dimensional systems have been investigated in the case of the adsorption of D-alaninol on Cu( 100) !n:!ffa€e. This molecule shows chiral self-assembling at room temperature. Selfassembled pattern formation seems strongly governed by supramolecular directional non-covalent forces controlled by the chiral charge distribution in the molecule. The chiral properties of the molecules play an important role in twodimensional geometry due to the space confinement imposed by the substrate surface. D-alaninol, a small flexible chiral organic molecule, presents two possible ways to bind to the surface, due to the presence of two functional groups, interacting either through both the amino and the hydroxylic group or just through the hydroxylic group. The comparison between DFT -based calculation and experimental results obtained by photoelectron spectroscopy and STM provides good understanding of the mechanisms of the adsorption of D-alaninol on Cu( 100) and its chemical interaction while considering different molecular amount on the surface. The interaction with the copper surface takes place through the intact alaninol molecule, with on top adsorption for both nitrogen and oxygen in sub-monolayer regime, or through the dehydrogenated amino group of the molecule, allowing chemisorption onto the surface through imino, rather than an amino group. The coexistence of these two types of molecular structures on the surface provides an interpretation of photoelectron experimental results for core levels and valence band. The modification of the bonding at the amino group seems to be responsible of the chemical interaction between Cu and alaninol driven by the growth of a self assembled monolayer at the surface.

130

REFERENCES [1] R Raval, CATTECH 5, 12 (2001). [2] S.M. Barlow, R. Raval, Curf. Opin. Colloid & Interface Science, 13, 65 (2008). [3] S.M. Barlow, S. Louafi, D Le Roux, J Williams, C Muryn, S Haq, R Raval, Langmuir 20, 7171 (2004). [4] S. lITera, G. Contini, N. Zema, S. Turchini, J. Fujii, S. Sanna, and T. Prosperi, J. Phys. Chern. BIll, 7478 (2007). [5] S. lITera, G. Contini, N. Zema, S. Turchini, S. Sanna, P. Moras, C. Crotti, and T. Prosperi, Surf. Sci. 601, 2562 (2007). [6] J.P. Perdew, K. Burke, and M. Ernzerhof, Phys. Rev. Lett. 77, 3865 (1996); 78, 1396 (1997); 80, 891 (1998). [7] D. Vanderbilt, Phys. Rev. B 41,7892 (1990). [8] S. Baroni, S. de Gironcoli, A Dal Corso, P. Giannozzi, http://www.pwscf.org! [9] RB. Rankin and D.S. Sholl, J. Phys. Chern. B 109,16764 (2005). [10] H.J. Monkhorst and J.D. Pack, Phys. Rev. B 13,5188 (1976). [11] M. Methfessel and AT. Paxton, Phys. Rev. B 40, 3616 (1989). [12] L. Bengtsson, Phys. Rev. B 59, 12301 (1999). [13] E. Pehlke and M. Scheffler, Phys. Rev. Lett. 71, 2338 (1993). [14] A Pasquarello, M. S.Hybersten, and R Car, Phys. SCf. T66, 118 (1996). [15] Y. Travaly, D. Vanderbilt, and X. Gonze, Phys. Rev. B 61, 7716 (2001). [16] S. Meloni, A Palma, A Kahn, J. Schwartz, and R Car, 1. Am. Chern. Soc. 125, 7808 (2003). [17] M. Carbone, A Palma, R Caminiti, Phys. Rev. B 75,245332 (2007). [18] R Fausto, C. Cacela, and M.L. Duarte, J. Mol. Struct. 550-551, 365 (2000). [19] L.M. Ghiringhelli, P. Schravendijk, and L. Delle Site, Phys. Rev. B 74, 035437 (2006). [20] S. D'Agostino, L. Chiodo, F. Della Sala, R Cingolani, and R Rinaldi, Phys. Rev. B 75, 195444 (2007). [21] M. Nyberg, J. Hasselstrom, O. Karis, N. Wassdahl, M. Weinelt, and A Nilsson J. Chern. Phys. 112,5420 (2000). [22] B. Afsin, P.R Davies, A Pashusky, M.W. Roberts, D. Vincent, Surf. Sci. 284, 109 (1993). [23] P.R Davies and J.M. Keel, Surf. Sci. 469, 204 (2000). [24] A Marini, G. Onida, and R Del Sole, Phys. Rev. Lett. 88,016403 (2002). [25] D. Catone, S. Turchini, G. Contini, N. Zema, S. lITera, T. Prosperi, M. Stener, D. Di Tommaso, and P. Dec1eva, J. Chern. Phys. 127, 144312 (2007). [26] B. Hammer and J.K. Norskov, in Chemisorption and Reactivity of Supported Clusters and Thin Films, edited by R M. Lambert and G. Pacchioni (Kluwer Academic, Dordrecht, 1997), pp. 285351; R Di Felice, A Selloni, and

131

E. Molinari, J. Phys. Chern. B 107,1151 (2003). Phys. Rev. B 75,115407 (2007). [27] J. Tersoff and D. R. Hamann, Phys. Rev. B 31, 805 (1985).

MICRORADIOLOGY IMAGING OF THE BIODISTRIBUTION OF POLYETHYLENE GLYCOL (PEG) MODIFIED GOLD NANOPARTICLES IN CANCER BEARING MICE CHIA-CHI CHIEN

Department of Engineering and System Science, National Tsing Hua University, Hsinchu 300 and Institute of Physics, Academia Sinica, Nankang, Taipei 115, Taiwan CHI-JEN LIU, HSIANG-SHIN CHEN, CHANG-HAl WANG, SHIN-TAl CHEN, WEI-HUA LENG

Institute of Physics, Academia Sinica, Nankang, Taipei 115, Taiwan Y.HWU

Institute of Physics, Academia Sinica, Nankang, Taipei 115, Department of Engineering and System Science, National Tsing Hua University, Hsinchu 300, Institute of Optoelectronic Sciences, National Taiwan Ocean University, Keelung 202 and National Synchrotron Radiation Research Center, Hsinchu 30076, Taiwan. We explored a very interesting gold nanoparticle system -- pegylated gold in colloidal solution -- and analyzed its uptake by cancer bearing mice. Large amounts of such nanoparticles are uptaken by the tumors as verified by quantitative inductive coupled plasma - optical emission spectroscopy (ICP-OES) and real time microradiology. With high time and lateral resolution phase contrast microradiology, we conclusively demonstrated at the animal level and by direct visual observation that the particles strongly accumulate in tumor regions - specifically more than in normal muscle tissue. This accumulation increases with the time after injection -- whereas for most non-tumor regions it saturates or decreases. The potential impact of this result is discussed with special emphasis on passive targeted drug delivery and the delineation and early imaging of small tumors.

1. Introduction

The objective of radiation therapy is to reduce the dose and the damage to healthy tissues and organs: nanotechnology continuously demonstrates that it has much to offer to facilitate the progress towards this objective [1, 2]. Different approaches were previously developed to improve the selectivity: active targeting, passive targeting and active/passive combinations. Active targeting exploits tumor-specific bio-molecules such as antibodies or functional peptides exhibiting strong affinity with the targeted cells [3-5]. Passive targeting

132

133

approaches are based instead on the specific properties of tumor tissues such as vascular leakage or abnormal vessel architecture. This can facilitate the uptake of specific drug carriers - a property that Maeda et al. called the "EPR (enhanced permeation and retention) effect" [6-9]. Tumors larger than 1-2 mm need new blood vessels to supply nutrients and oxygen, with anomalous density and morphology and ineffective physiological functions (e.g., lymphatic clearance)[lO, 11]. These characteristics lead to the EPR effect for suitably small (nano) particles; in turn, such particles could be exploited for targeted drug delivery [12-15]. Gold nanoparticles are top candidates for both approaches because of their excellent biocompatibility and wide range of surface modification possibilities. For example, they will enhance the cell and tumor tissue response to high-energy photon irradiation (x-rays and gamma rays) [16, 17] - and selectively carry drungs to target areas. In animal studies, differential accumulation at tumor sites with 5-8 fold increase in concentration was recently discovered for pure gold nanoparticles. - leading to substantial dose enhancement without cancer drugs or labelers [18]. In a series of recent articles [18-20], we announced a new approach for the synthesis of gold nanoparticles coated with PEG ("PEGylated"). The synthesis was activated by irradiation with intense synchrotron x-rays (photon energy 8-15 keY) yielding highly concentrated, stable colloids without using any reducing agents. The PEG-Au nanoparticles were stable and biocompatible in both cellular and animal tests. Their potential for cancer therapy was by irradiating CT-26 cells with and without PEG-Au nanoparticles and studying their viability and their cellular damage by synchrotron radiation Fourier transform IR spectromicroscopy. In the present work, we first demonstrate that such nanoparticle system can be used very effectively as a contrast agent for microangiography. We also show that after injection in cancer-bearing mice they strongly accumulate in the cancer areas - and this tendency increases with time up to -12 hours after injection. Even though nanoparticles showing the EPR effect are being tested for cancer therapy in clinical trials, their specific accumulation in tumors was only demonstrated by macroscopic visual observation [4, 21] and post-mortem chemical analysis [22, 23]. Our present study presents conclusive microscopic evidence that PEG-coated Au nanoparticles do strongly accumulate in mice tumors, much more than in normal muscle tissues. This result was obtained with the parallel use of different experimental techniques including in particular phase contrast micro-radiology with a synchrotron x-ray source [24-26] - plus ICPOES, transmission electron microscopy (TEM), transmission x-ray microscopy

134

(TXM) and pathology imaging. Specifically, ICP-OES enabled us to quantitatively characterize the phenomenon. The microscopy and microradiology techniques enabled us to follow the evolution in real time of the preferential nanoparticie uptake: to the best of our knowledge, this is the first direct dynamic observation of passive targeting with nanoparticies.

2. Experimental

2.1. Preparation and characterizations of the pegylated gold nanoparticie solution Pegylated gold nanoparticies were synthesized by the synchrotron x-ray irradiation method reported elsewhere [18, 19, 27]. In short, a mixed water solution of gold precursors (2 mM HAuC14-3H20 (Aldrich, MO, US) with NaOH (0.1 M, Showa Inc., Japan) and polyethylene glycol (PEG) (MW 6000, Showa Inc., Japan) were placed into polypropylene conical tubes (15 mI, Fa1con®, Becton Dickinson, NJ) for x-ray irradiation. The irradiation time was set to 5 min to guarantee the complete reduction of the gold precursor. The exposures were performed at the "white light" x-ray microscopy beamline BLOIA of the National Synchrotron Radiation Research Center (NSRRC) , Hsinchu, Taiwan [28]. The energy distribution of the x-ray photons was centered at -12 keY, with a broad width of 8-15 keY [29]. The particie morphology, structure and size were analyzed and reported in Ref. 30.

2.2. Animal models EMT -6 syngeneic mammary carcinoma cell lines were cultured under standard conditions. The BALB/c ByJNarl tumor models were generated by inoculating lxl06 cells in 10 III PBS into the thigh. The mice were used for the study 1 week after inoculation, when the tumor had grown to 50-90 mm3 (estimated as half the product of the square of the smaller diameter multiplied by the larger diameter).

2.3. High resolution microradiograpy The real time microradiology observations were performed at the same beamIine as the above exposures. The time of the each frame was 3 ms and the field of view (FOV) was 3 mm. Preliminary observations were performed with no nanoparticies and using two types of the iodine solutions, a hydrophilic solution (Telebrix®, Guerbet company, France) and a hydrophobic solution

135

(Lipiodol® ultra-fluide, Guerbet company, France). 400-600111 of the selected solutions were injected through the tail vein for imaging. Whole body images of the mouse were taken and also kinetic images of the local region of the organ. Then, nanoparticle effects were tested by taking images before and after the injection of 100 III (-75 mg/rnl) of gold nanosols via the tail vein by a syringe pump. All tests on live animals were generally anaesthetized and monitored following standard procedures (Laboratory Animal).

2.4. Biodistribution and blood circulation The accumulated gold amount was tested by ICP-OES after injection of 200 III of gold nanosol; some tests performed with lower nanosol volumes (100 Ill, -75 mg/ml) confirming the results. The tumor-bearing mice were sacrificed 30 or 90 min or 4 hr after the colloidal injection,. A total of 3 mice were used for each after-injection time. After sacrifice, all of the most important organs or tissues (blood, lung, tumor, muscle, brain, heart, liver, spleen and kidney) were collected for gold analysis by ICP-OES. 2.S. TEM sample preparation and observation

To reveal the Au nanoparticle distribution, TEM samples were prepared as follows: the organs were immediately fixed with glutaraldyhyde at 4 C for 24 hrs. After removing the glutaraldyhyde by O.lM PBS, the samples were further fixed and stained with 1% osmium tetraoxide in buffer and dehydrated by a series of alcohol treatment, embedded in resin, and sliced to 90-100 nm in thickness using a Leica Ultracut R ultramicrotome. After being double stained with uranyl acetate and lead citrate, the specimens were observed in a Hitachi H7500 TEM operating at 100 keY.

2.6. Histology analysis To examine the pathological characteristics of PEG-gold loaded organs/tissues, different organs - tumor, spleen, liver, lung, kidney and muscle were immediately fixed in 10% formalin and dehydrated by a series of immersions in 50, 70, 90 and 100% ethanol. They were then embedded in paraffin wax and sectioned to 2-5 11m slices with a Leica RM2235 microtome. After histological H-E staining, the slices were observed by a confocal laser scanning microscope (Leica TCS-ST, Germany).

136

3. Results and Discussion:

3.1. In vivo distribution of PEG-gold nanoparticles The Iep-OES results indicated that the nanoparticles accumulated in tumors and reveal specific time-dependent patterns. The concentration in all tumors monotonically increased with the time after ,...... '''f''tuYn as shown in 1. 2 shows the fraction of the injected gold uptake:n each organ with to all the organs we examined.

50

_Omin 30 min 90 min

E

!:! 40 OJ

Qi

4 Hr

0. Q) <J)

30

0 "0

20

10

0

blood

lung

tumor muscle brain

heart

liver spleen kidney

Targeted organs Figure 1 - ICP-OES (Inductive Coupled Plasma Optical Emission Spectroscopy) analysis shows the biodistribution of PEGylated gold nanopartic1es in tumor bearing mice for different times after their injectiou. The results are shown for tumor regions and for a series of non-tumor areas.

These results reveal that much more gold accumulated in tumor areas than in muscle, brain kidney or heart tissues. After 4 hrs, the gold in tumors reached a concentration even higher than in the blood. The gold concentration in the blood was found to decrease with the time after the exposure, consistent with other pharmacokinetic studies [1, 31]. The specimen analysis revealed a concentration than in other organs. For the liver and spleen substantial was observed that increased with time. The comparison between the tumor and muscle is revealing as far as the EPR effect is concerned. In Fig. 2, we show the fraction to tumors and muscle and the tumor/muscle uptake ratio for two

137

different times after the injection. After 30 there was no detectable became visible at 90 minutes but then it in muscle tissues. This to an increase in the tumor/muscle ratio up to 34.1. b) 'l

I 15

j Targeted organs

organs

Figure 2 ICP-OES results similar to those of Fig. 1: in this case, the bars show the fraction of accumulated nanoparticies for each organ with respect to aHlhe organs we examined. (b) Enlarged version of the data of (a) for tumors and non-tumor muscle tissue. The inset shows the tumor/muscle increase. ratio emphasizing its

3.2. Real-time analysis of the EPR effect by microradiology To prepare the real-time

of PEG-Au naIloP,artlc uptake by phase

contrast microradiology, we first performeda serious of experiments using

traditional absorption contrast agents tovisualize microangiography, tumor and the blood circulation. Different contrast agents were found distributed in the mouse with very different characteristics highlighted different organs. The radiographin 3a was taken without any contrast agentand showed only bone structure and thoracic cavity.While analyzing

the distribution of 600111 ofLipiodolcontrast agent in the organs

it was found that this solution was not diluted by the body fluid and could show the vessel structure in the mouse body On the the Telebrixcontrast was easily diluted by the In this case, all organs showed dark-bright imagecontrast without sharp edges,except the bladder The circulation time of the iodine solution in the mouse

was 2 min. These tests indicate that the hydrophilic Telebrixcontrast could be used to image drugcirculation in real time.

Fig. 3 - (a) the mouse without any contrast agent, it showed clear bone structure, (b) the mouse was Injected with Liplodol® contrast agent, It showed dear vessel structure, (c) the mouse was injected with Telebrix* contrast agent, the contrast agent diffused In the mouse.

The Lipiodol® led to the clear visualization of the vessel structure of heart, liver and kidney. The resolution of the corresponding microradiographs was quite good: in some cases, the vessels smaller than 20 \tm could be dearly identified. This is actually not limited by the instrumental resolution but by the natural size of these features (vessels). Therefore, synchrotron microangiography is much better than standard clinical imaging modalities whose resolution is limited to hundreds of pm. The magnified vessel images of various organs are shown in Fig. 4. The vessel structure is particularly clear in the kidney and liver areas, Fig. 4a and 4d. The most complex structure was observed for the lung radiographs with clear identification of vessels, capillaries and alveoli.. Images taken at the tumor site (mouse thigh) showed the small vessels winding around the tumor (Fig. 4c) - associated to microangiogenesis. Contrary to Lipiodol®, the Telebrix® contrast agent could freely circulate throughout the whole body, but it did not lead to a clear detection of the vessel structure (Fig. 5) because of its high solubility and diffusivity. Other types of contrast agents, such as emulsion and microencapsulated agents all had limitations similar to Lipiodol or Telebrix. Therefore, the corresponding images did not reflect the true circulation of because of the high viscosity or the contrast agent diffused too quickly to outline the desired detailed structure. None of such agents was actually able to deposit differentially at the tumor sites and therefore has limited value as an "active contrast agent".

139

The gold nanoparticles solution could be used as an effective C01IlPiennerltru'v contrast agent. A compromise can be reached between the low concentration rec!uu'ed to see the circulation and the high concentration required for strong contrast, enabling one to highlight very small vessels and to detect crc1an;gmgellesls from cancer.

Figure 4 — (a) Clear kidney vessel with the Lipiodol* contrast agent. (b) Lung with clear vessel structurc. (c) Tumor on the leg with 1I0nciear capillaries. (d) Liver.

The amount of gold nanopractice in the micro vessel was found to increase with time indicating that the nanoparticles do not freely circulate in small vessels but tend to accumulate there. This phenclme:non was never previously observed with any clinical imaging technique. Our images show that the nanoparticles accumulate not only in the microvessles but also in the tumors themselves by leakage. Both mechanisms — retention by microverssels and leakage into the tumors - contribute to the EPR effect — and point to a possible use of PEG-Au as a contrast agent to locate cancer tumors,

140

Figure 5 - (a) Obsure kidney shape with the Telebrix* contrast agent. (b) Lung showed clear hone structure but vessel. (c) tumor site was no vessel structure. (d) Liver structure without clear vessels.

Figure 6 shows a time sequence of microradiographs taken at a specific tumor size and monitoring the nanoparticle accumulation in real time with coherent microradiology. Specifically, Figs. 6a and 6b show the tail region at two times after injection separated by 0.33 sec and reveal (dark areas) the nanoparticles passing through blood vessels. Figures 6c-6f show a region cOiltainiI1lg a tumor and blood vessels. Approximately 10 sec after the injection, the ll1tu~uratH)nS of the main vein vessel, microvasculature and tumor sites is clearly delineated by the darkening effect of the accumulated nariopartl After 2 min, both the tumor outlines and the intra-tumor tissue structure were observed with increasing contrast. The strongest contrast appeared 15 min after injection: the tumors became clearly visible with no image processing. However, compared to the 50 sec images the boundaries of large vessels were less visible. This indicates that between 50 and 100 sec there were no longer nanoparticles in the large vessels whereas the accumulation in the tumor and in the nearby microvascularisation

141 increased. This indicates that the EPR effect for different organs has different and sometimes complex time evolution, not revealed by mere visual or postmortem inspection.

Figure 6 - Microradiography snapshpots extracted from real-time video sequences taken during and after the injection of PEGylated Au nanoparticles. (a) 3sec and (b) 4 sec: the tail site showing the passage of the injected particles (dark); (c)-(f) images of a tumor-containing region taken 10 min, 15 min, 10 min and 15 min after injection. The dark wire is a tumor-locating reference. The white arrows indicate the tumors.

3.3. Microscopic-scale TEM analysis of the nanoparticle distribution This technique enabled us to tackle sophisticated questions like: are the uptakeo gold particles located inside the cells or within the intercellular matrix? TEM micrographs of various mice organs and tumor are shown in Fig. 7. The individual gold particle size was 30-40 ran, much larger than the originally

42

administrated 6 nm. By anlogy with other experiments, we believe that this is due to colloidal flocculation and eventual aggregation of small particles.

Figure 7 - TEM micrographs of different regions with accumulated PEGylated Au particles 24 hrs after injection: (a) tumor; (b) liver; (c) spleen; (d) kidney; (e) lung; (f) heart.

The distribution of gold in liver is shown in Fig. 7b. It was found that gold particles agjgregalted in the endosome were almost confined within the cytoplasm of the liver cells. Here, the individual particles size was 50-iGO nm and was again the result of agglomeration.

143

3.4, Histology imaging Figure 8 shows microscopy images of tumor, spleen, liver and lung, kidney and muscle after H-E staining. In Fig, 8a, we see a cross-section of a small blood vessel in a tumor containing region. The vessel is filled with gold nanoparticles (the dark regions). Some nanoparticles are also observed in the neighboring inter-cellular matrix. For the spleen, Fig. 8b, the accumulated nanoparticles were mostly observed within the red pulp (with splenic cords and sinuses in between them) - whereas they are not found in the lymph nodes

Figure 8 - Histology images (40X) of PEGylated Au particles in different organs 24 his after injection: (a) tumor; (b) spleen; (c) liver; (d) lung; (e) kidney; (f) muscle.

144

Figure Sc shows the nanoparticle distribution within the network-like lobules of liver. The images indicate the formation of nanoparticle aggregates. The hepatocytes within the lobule captured the aggregates that accumulated within both the eosinophilic cytoplasm and the interface region. Flake-like large gold aggregates also appeared within the non-tumor micro-vasculature of lungs, as shown in Fig. Sd. The aggregates were mostly close to vessel walls. For kidneys (Fig. Se), less gold particles than in lungs were observed and they accumulated within the inner microvasculature and the renal cortex. Some particles were present in the muscle tissue, but not in the skeletal muscle fibers, eosinophilic cytoplasm, and the small peripheral nuclei - as illustrated in Fig. Sf. In essence, the histology imaging results confirmed that, large amounts of PEGylated gold nanoparticles accumulated in tumors, lungs and major RES organs, consistent with the ICP-OES and TEM data. The histology images detected no pathological changes related either to inflammation or to active immunocytes congregating in these organs. 4. Conclusion Our new synthesis approach for high density, stable colloidal solutions of pegylated gold nanoparticles appears quite effective in enhancing the preferential uptake by tumors consistent with the EPR effects. Our combined quantitative ICP-OES, microscopy and microradiology tests demonstrated that 5 nm PEGylated gold particles are strongly concentrated in tumors in mice. This concentration is larger than in normal muscle tissues, comparable to that of blood and only smaller with respect to RES organs and lungs. This direct evidence could facilitate the possible use of PEGylated gold nanosols - in particular those prepared by x-ray irradiation - for targeted drug delivery, radiotherapy and radiological diagnosis. Acknowledgments This work was supported by the Biomedical NanoImaging Core Facility, National Science Council (Taiwan), by the Academia Sinica (Taiwan). References 1. 2.

3.

M. Ferrari, Nat. Rev. Cancer 5, 161 (2005). I. H. EI-Sayed, X. H. Huang, M. A. EI-Sayed, Cancer Letters 239, 129 (2006). T. M. Allen, Nat. Rev. Cancer 2, 750 (2002).

145 4. 5. 6. 7. 8. 9. 10. 11. 12.

13.

14. 15. 16. 17. 18.

19.

20.

21. 22. 23. 24. 25.

G. F. Paciotti, L. Myer, D. Weinreich, D. Goia, N. Pavel, R. E. Mclaughlin, L. Tamarkin, Drug. Deliv. 11, 169 (2004). Z. Liu, W. Cai, L. He, N. Nakayama, K. Chen, X. Sun, X. Chen, H. Dai, Nature Nanotech. 2, 47 (2007). H. Maeda and Y. Matsumura CRC Crit. Rev. Ther. Drug Carrier Syst. 6, 193 (1989). R. Haag Angew. Chern. Int. Ed. 43, 278 (2004). K. Iwai, H. Maeda, T. Konno, Cancer Research 44, 2115 (1984). Y. Matsumura, H. Maeda, Cancer Research 46, 6387 (1986). J. Folkman, J. Natl. Cancer Inst. 82,4 (1990). L. W. Seymour, Y. Miyamoto, H. Maeda, M. Brereton, J. Strohalm, K. Ulbrich, R. Duncan, Eur. J. Cancer 31,766 (1995). T. Konno, H. Maeda, K. Iwai, S. Tashiro, S. Maki, T. Morinaga, M. Mochinaga, T. Hiraoka, I. Yokoyama, Eur. J. Cancer Clin. Oncol. 19, 1053 (1983). T. Konno, H. Maeda, K. Iwai, S. Maki, S. Tashiro, M. Uchida and Y. Miyauchi , Cancer 54, 2367 (1984). H. Maeda, Adv. Drug Deliv. Rev. 6, 181 (1991). H. Maeda, K. Greish, J. Fang, Adv. Polym Sci. 193, 103 (2006). D. M. Herold, I. J. Das, C. C. Stobbe, R. V. Iyer, J. D. Chapman, Int. J. Radiat. BioI. 76, 1357 (2000). J. F. Hainfeld, D. N. Slatkin, H. M. Smilowitz, Phys. Med. BioI. 49, N309 (2000). C. -H. Wang, T. -E. Hua, C. -C. Chien, Y. -L. Yu, T. -Yo Yang, C. -J. Liu, W. -H. Leng, Y. Hwu, Y. -CO Yang, C. -CO Kim, J. H. Je, C. H. Chen, H. M. Lin, G. Margaritondo, Mat. Chern. Phys. 106, 323 (2007). C. -H. Wang, C. -C. Chien, Y. -L. Yu, C. -J. Liu, C. -F. Lee, C. -H. Chen, Y. Hwu, C. -So Yang, J. H. Je and G. Margaritondo, J. Synchrotron Radiat. 14, 477 (2007). C. -H. Wang, C. -J. Liu, C. -L. Wang, T. -E. Hua, J. M. Obliosca, K. H. Lee, Y. Hwu, C. -So Yang, R. -So Liu, H. M. Lin, J. H. Je, G. Margaritondo, J. Phys. D. 41, 1953 (2008). G. F. Paciotti, D. G.I. Kingston and L. Tamarkin, Drug. Develop. Res. 67, 47 (2006). J. F. Hainfeld, D. N. Slatkin, H. M. Smilowitz, Phys. Med. BioI. 49, N309 (2004). J. F. Hainfeld, D. N. Slatkin, T. M. Focella, H. M. Smilowitz, Br. J. Radiol. 79,248 (2006). G. Margaritondo, Elements of Synchrotron Light for Biology, Chemistry, and Medical Research (Oxford University Press, New York, USA 2002). Y. Hwu, W. L. Tsai, H. M. Chang, H. I. Yeh, P.c. Hsu, Y. C. Yang, Y. T. Su, H. L.Tsai, G.M. Chow, P. C. Ho, S.c. Li, H. O. Moser, P. Yang, S. K.

146

26.

27. 28. 29.

30.

31.

Seol, C. C. Kim, J. H. Je, E. Stefanekova, A. Groso, G. Margaritondo, Biophys. J. 87, 4180 (2004). Y. Hwu, W. L. Tsai, J. H. Je, S. K Seol, B. Kim, A. Groso, G. Margaritondo, K H. Lee, 1. K Seong, J. K, Phys. Med. BioI. 49, 501 (2004). Y. C. Yang, C. H. Wang, Y. K Hwu, J. H. Je, Mater. Chern. Phys. 100,72 (2006). G. Margaritondo, Y. Hwu, J. H. Je, Riv. Nuovo Cimento 27,7(2004). P. C. Hsu, C. H. Wang, T. Y. Yang, Y. Hwu, C. S. Lin, C. H. Chen, L. W. Chang, S. K Seol, J. H. Je, G. Margaritondo, J. Vac. Sci. Technol. A 25, 615 (2007). C. J. Liu, C. H. Wang, C. C. Chien, T. Y. Yang, S. T. Chen, W. H. Leng, C. F. Lee, K H. Lee, Y. Hwu, Y. C. Lee, C. L. Cheng, C. S. Yang, Y. J. Chen, J. H. Je, G. Margaritondo, Nanotechnology 19, 295014 (2008). V. P. Zharov, E. N. Galitovskaya, C. Johnson, T. Kelly, Lasers in Surgery and Medicine 37,219 (2005).

147

Albi, nostrorurn serrnonurn candide iudex quid nunc te dicarn facere in regione Pedana? Scribere quod Cassi Parrnensis opuscula vincat an taciturn silvas inter reptare salubres, curantern quicquid dignurn sapiente bonoque est? (Q. Horatius F. - Epistolae I, IV)

THE FRASCATI EXPERIMENTS G. FARACI Dipartimento di Fisisca Universita di Catania

In July1965 after my Laurea in Physics, at the University of Catania, my advisor Prof. Italo Federico Quercia proposed to me a fellowship of the Centro Siciliano di Fisica Nucleare. I had to join his group working in Frascati in experiments aiming to detect temperature effects in the annihilation process of positrons in metals and semiconductors. Prof. IFQ (as we usually called him) was at that time the director of the Frascati National Laboratories and one of the directors of the Institutes of Physics of the University of Catania. I was twenty one, and the idea of a research activity in Frascati was very challenging, so I immediately accepted although my salary would have been only 90 kLire (Euros 45) per month! Fortunately, when in Frascati, I usually slept in the guest-house and had my meals at the excellent cafeteria at the symbolic price of 100 Lire. However, my institute was paying also my travel and living expenses for an amount of about 4000 Lire per day. In Frascati I was therefore considered a person full of money, and actually often we were dining in town. I remember some excellent dinners in the familiar "Osteria da Rocco" where at the ceiling many "prosciutti"(pork legs) were hanged. Usually, we tasted there the most exquisite "amatriciana" of all area of the Castelli Romani.

148

At that time, my boss, Prof. Quercia, was quite famous in Italy being the father of the electro synchrotron Adone with Giorgio Salvini, Bruno Touscheck, Carlo Bernardini, Marcello Conversi, Mario Puglisi, etc. However, the local research group of IFQ which should have worked in Frascati in the new experiment, was really represented by only one person, Maurizio Spadoni, who had just obtained his Laurea in Rome, with a thesis on low temperature experiments at 4.2 K with Giorgio Careri. He was the expert on low temperature, whereas I studied positron annihilation in silicon during my thesis work. A third young physicist Elio Turrisi often joined us from Catania. Of course, this orchestra was under the direction of IFQ. In Frascati our experimental apparatus was hosted in the hangar of the superconductivity group of Prof. G. Sacerdoti, where I met Nicola Sacchetti, Giovanni Sanna, Giorgio Pasotti, etc. I spent two years there looking for De Haas van Alphen oscillations in the density of states of bismuth, at the liquid helium temperature, as a function of the magnetic field, using the positron annihilation technique. During this period 1966-67, I do not remember exactly the month, a new student arrived for developing his thesis in the Sacerdoti' s group, Paolo Perfetti. A very friendly atmosphere was established between us and I remember frequent conversations with him especially in the morning when he was crossing our experimental area near the main entrance of the laboratory. Here, we mounted an angular spectrometer with arms moving around the center of the system. These arms with sodium iodide detectors were able to measure the 0.511 Me V annihilation photons. Their position was moving within few mrad, the width of the angular correlation curve. At the center was put the Bi single-crystal hollow cylinder having the Na22 positron source inside. The crystal within a low temperature cryostat was hosted between the poles of a 15 kgauss magnet. The location of this equipment was on the left side of the main entrance. All the people of the Sacerdoti's group coming at work, including Paolo, were spending few words for greeting us every morning.

149

But in an occasion everybody avoided to use the main door, and nobody came around. It happened when during the transfer of our radioactive Na-22 positron source, in the form of tiny spheres of amberlite, one of these spheres jumping continuously on the floor was lost. It was useless to look for it with several Geiger counters, we never found it. IFQ was furious with us and he participated to the unsuccessful search. Paolo too did not cross our area for several days. Fortunately the activity of the bead was quite low with a decay time of 2.6 years ........ . Twenty years later attending an EPS congress in Berlin (1985), my interests switched to cluster physics, a new topic at that time with few adepts. I was thinking to some experiments to be performed with synchrotron light. I contacted Paolo for a visit to his beamline, in Frascati.

6

8

Fig. 1. XAFS oscillations of Xe clusters vs. temperature, taken at the French beamline ESRF - Grenoble. (G. Faraci et al. Phys. Rev. B 74, 235436, 2006)

150

Fig. 2. Xe nanocrystal parameters vs. temperature, obtained by XRD at the Troika beamline ESRF-Grenoble. The diameter of the clusters is in angstrom. (G. Faraci et al. Eu. Phys. J. B 51,209,2006)

In that period Paolo was organizing a workshop on Synchrotron Radiation experiments where I met Settimio Mobilio, Piero Chiaradia and other physicists involved in different researches using the light of Adone. At the PULS my first experiment was performed on Xe clusters implanted in a Si matrix, where we detected solid crystalline Xe nanocrystals.

151

The beamline was the XAFS station and a stimulating collaboration with Settimio Mobilio and his group brought to nice results on cluster physics observed by EXAFS technique firstly in Frascati and later in Paris-Lure and in Grenoble-ESRF (figs. 1,2). In 1987 during a meeting in Trieste for the development of the Elettra machine I met Giorgio Margaritondo to whom I required some beam time in SRC-Madison where he was the director for research. Giorgio encouraged my ideas trying to apply the photoemission spectroscopy to clusters. A nice first trip to Stoughton was organized for preliminary experiments. I was impressed by the very good resolution of the photoemission beamline but also by the excellent fish of the Red Lobster restaurant. Several others trips to Madison followed, with interesting results (fig. 3,4), and some exhibition of Giorgio in wild-chicken-dances in a german beer-house. In 1988 I convinced one of my PhD students Tony Terrasi to spend about a year in Madison.

152

-~ 11

'2 ;::J

Allgraphite

• Experimental Data - - Flttlng Curve

~ (JJ c (!)

:g c

AI 2p doublet

:1 ~

~

0-

·6

·2

0

2

Binding Energy (eV) Fig. 3. Photoemission spectrum taken at the beamline Mark II of SRCMadison, Wisconsin. ( G. Faraci et al. J. Appl. Phys. 78, 4091, 1995)

Later an other of my students, Salvo La Rosa, enthusiastically decided to work in Madison and in Lausanne, with Giorgio. Salvo will obtain a permanent job entering the staff of Elettra at the spectromicroscopy beamline.

153

Cs 4<1

t ML

10

0.1 0,0 ao~--~~~--~--~~~--~--~--~

-0.1

20

0.1

0.0 -0.1

0.1

0.0 .0.1

6

5

t=:=::::::t:==:.::J.:.~~--I.

___--J 0.1

~~~~~Y!.! ' - - _......._ - ' - _ - - ' - _ - - - " '_ _L.-_.L.-_......_...J

-80

·78

0.0 -0.1

-76

Binding Energy ( eV ) Fig. 4. Photoemission spectrum taken with the Scienta ESCA 300 at the IPA-EPFL- Lausanne. (G. Faraci et al. Phys. Rev. B 53, 3987,1996)

154

Although I performed several experiments in SRC-Wisconsin (fig. 3), in Lausanne (fig. 4) and also in Elettra-Trieste, in no one I had the opportunity to collaborate with Paolo. Anyway we encountered several times for meetings or congresses. In particular, I remember that held in Lisbon in 1990 .....

AFM AND SNOM TECHNIQUES AT ISM: AN OVERVIEW M. GIRASOLE, G. LONGO, G. POMPEO and A. CRICENTI CNR - Istituto di Struttura della Materia, V. F osso del Cavaliere 100, 00133 Roma - Italy.

Atomic force microscopy (AFM) and Scanning Near-field Optical Microscopy (SNOM) are well established techniques for high resolution morphological and optical studies and, indeed, they have been applied to study surfaces for microelectronics, electrochemistry and other solid state physic's research field. Moreover, AFM and SNOM are very attractive for biological studies too. This is because their characteristics allow to observe, in air and liquid, conducting and non conducting sample with resolution well below the diffractive limit of the conventional optical microscopes and can add a description of the optical properties of the surface at super-resolution. In this paper we present the AFM and SNOM techniques with particular emphasis on some applications performed in the field of biophysics, material science and space sciences.

1.

INTRODUCTION

Scanning Probe Microscopies (SPM)l are a class of versatile techniques, the most popular being Scanning Tunneling Microscopy (STM)2, Atomic Force Microscopy (AFM)3.5 and Scanning Near-field Optical Microscopy (SNOM)6.8 whose invention have drastically changed and modernized the concept of microscopy. Historically, the SPM techniques have been introduced in the eighties when the development of the STM deserved a Nobel price to the inventors Gerd Binning and Heinrich Roher9 . In the following years, the concept underlying the development of STM have been generalized to different experimental context and new techniques, such as the AFM and, afterward, the SNOM, were introduced. The main idea behind all these techniques is the use of a sharp probe, scanned by sub-nanometric steps while maintained at very small distance from the sample surface, to locally investigate topography and different properties of the specimen. Indeed the SPMs, initially developed as pure imaging techniques, allow the collection of magnified images of the sample with a resolution well below the limit of the conventional optical microscopy (roughly speaking half the wavelength of the light used) and, at the same time, can produce maps of

155

156

physical quantities not always linked to the bare sample's morphology. The last decades of use and application of SPMs have proved that the detection of these properties and the determination of their relation with the sam~le morphology are, indeed, a precious method to solve many scientific problems 0-12. The additional physical quantities that can be collected through SPMs can differ, depending strictly on the nature of the probe employed. For instance, the AFM probe is an extremely sharp, pyramidal, silicon (or silicon nitride) tip mounted at the apex of a very flexible cantilever which is the best method to detect, through the deflection of the cantilever, the magnitude of the forces arising between the tip and the sample's surface. In the magnetic force microscopy (MFM) 13, the choice of a magnetic probe enables the high spatial resolution mapping of magnetic domains. Similarly, in the STM, the probe is a conductive (often tungsten) tip brought very close to a conductive sample in such a way that the quantistic tunneling current between the two can be collected and monitored. In the SNOM, on the other hand, the probe is a tapered optical fiber used to illuminate and/or to collect light from the sample in the near field regime thus enabling the investigation of the local optical properties well below the diffraction limit. Obviously, the choice of a specific probe allows the investigation of definite properties of the sample and determines the intrinsic sensitivity of the related microscopy as well as the spatial (lateral) resolution that can be achieved in an experiment. Namely, the d 6 dependence of the force vs. distance typical of the AFM, as well as the d4 of the SNOM, makes such techniques particularly appropriate for molecular scale studies, while the exponential dependence from the distance of the tunneling current makes STM by a long way the most sensitive SPM technique enabling sophisticate studies down to the atomic scale. Regarding SNOM, the coupling of a suitable light source, typically a laser, with the microscope allows to study the local optical properties (usually reflectance, or transmittance) of the samples. The use of tunable sources or a dye lasers allows, in principle, to perform local spectroscopy measurements, thus delivering high resolution chemical information. Moreover, treating the sample with suitable fluorescent dyes, discloses the possibility to investigate also the functional behavior of biological samples. In technical terms, the probe size and shape is the most important factor affecting the resolutions that can be achieved in a SNOM experiment. Indeed, on one hand, the spatial resolution is obviously dependent by the probe size for it determines key factors such as contact area and probe-sample convolution and, on the other hand, the optical resolution ultimately depends on the size of the tip aperture (typically some tens of nm). To improve the optical and/or the topographic resolution, novel SNOM setups have been developed, such as the tapping mode operation that allows to extend most of the AFM knowledge and technology to the SNOM world 14 . Concerning the AFM, the capability to detect the normal and lateral forces occurring between the tip and the sample can be exploited to collect high

157

resolution three dimensional topographies while probing the local chemical properties (qualitative maps of surface friction) of the sample. All these abilities have been demonstrated, over the years, unique tools in many research fields ranging from the tribology (where AFM can bridge the gap between the Surface Force Apparatus and the investigation of the interaction occurring at the nanoscale - nanotribology)15to the cellular or molecular biology. In particular the applications in these latter fields have been demonstrated remarkably interesting since the pioneer days of the AFM. Indeed, the technique allows high resolution, quantitative, non destructive imaging on the macromolecular scale coupled with the possibility to perform studies in different environments including liquids and physiological buffers l6 Such peculiarities make AFM a unique tool to perform high resolution studies of the time evolution of living systems or of artificial compartments (for instance, liposomes) as well as to investigate the interaction of biomolecules in physiological environments 17. More recently, another class of applications in which the AFM has demonstrated it's importance, providing auantitative information on the nanoscale, is the so-called force spectroscopy 1 . These applications are based on the collection of the force curves that are plots of tip-sample force vs. tip-sample distance. In practice, the force curves are obtained by vertically scanning the tip over the sample while collecting the force-induced deflection of the cantilever. The possibility to use a functionalized sample, for instance a suitable substrate upon which a polymer or a protein has been immobilized, allows to probe even the single molecule nanomechanical properties 19. In this context, indeed, a sophisticated analysis of the force curves allows the direct measurement of quantities such as bonding energies 20 , interaction forces between molecules 21 , elastic and viscoelastic properties of proteins, nucleic acids and polymers22, measure of the adhesive forces in air and liquid (see fig. 3) and all these properties can be directly tested on the nanoscale through repeated reversible or irreversible stretches of the samples. The force spectroscopy is a rapidly developing field whose results, valuable per se, can be of extreme interest for the investigation of fundamental problems (for instance the protein folding) as well as for practical purposes (test of polymers or of the interaction of a contaminant with a surface) and are very promising for the study of the material properties and the development of new biology-inspired materials.

2.

SPM ANALYSIS

The AFM and SNOM microscopes used in the present work are described elsewhere 23 ,24 For all the reported experiments, the AFM contact-mode measurements were performed in air (at room temperature and about 30% relative humidity) with the microscope working in the regime of repulsive interaction with forces

158

smaller than 1 nm from zero cantilever deflection. Gold coated cantilever (from Veeco) with a statistical apical radius of 5-10 nm were used and an elastic constant of 0.03-0.06 N/m. Particular care was used to avoid the occurrence of plastic deformation of the samples and to reduce as much as possible the extent of the unavoidable elastic deformations. The SNOM investigation have been performed operating the microscope in shear mode or in tapping mode. In both modes of operation particular care was dedicated to ensure a low-force scanning of the sample. The measurements were performed in air, at room temperature and controlled (30%) relative humidity. The SNOM probe was an optical fibre with an apex tapered by means of a pipette-puller to produce tips with different shapes and apical radii between 50 and 100 nm. The reported images have been treated by only a background subtraction and, when required, a polynomial plane alignment. No digital filter were applied.

3.

RESULTS

3.1

AFM and Force Curves In the everyday laboratory practice, the force curves are acquired by collecting the cantilever deflection while the tip is pushed toward the sample surface (approach curve), and subsequently withdrawn (retraction curve), through the elongation/retraction of a piezoelectric scanner. A dedicated software has been developed in order to collect force curve data. The operation requires a direct driving of the Z-scanner which can scan a selected approach/retraction length dividing the movement in a maximum of 65536 (2' )microsteps. At every microstep the tip can holds it's position for a definite time in order to decrease the noise and to collect the data in quasistatic conditions. Since the cantilever exhibits a typical elastic response, the deflection can be converted into force value by exploiting the relation: F=-k.,ffLix where Lix is the deflection and k.,ff the elastic constant of the probe-sample system. This procedure, however, allows to collect a force vs scanner displacement curve rather than a force versus tip-sample distance.

Deflection

[~~~~~~

Distance Sample FIG 1. Cantilever deflection and associated tip-sample distance

159

Thus, before the information about the sample properties can be obtained, the force vs. displacement curve must be analyzed and converted into a real force vs. distance curve by taking into account the cantilever properties and the effects of the force experienced by the tip during the data collection. The differences between the two kind of curves are reported below. Steel

~

Steel- Force vs Distance

Force V'i Displac.ement

1,5 c - - ' r - - - , - - - - , - - - - - , - - - - - ,

I~FOrCe(n~1

z

s

zc:

LL

LL

'; 0,5 ....... '.

~ 0,5 (;

;.

:=o

,[-

v -0,5 0'---~5'-0- 1 O-0-1...:'5"-0--:2~OO::------'2c'c50:--~300 J

Displacement (nm)

-0,5

~:-----="=---c-'-::---:-:'c' I~

50

100

150

Distance (nm)

FIG 2. A typical conversion from force vs displacement to force vs distance curve. Data are collected on a stainless steel surface. Exploiting the linear behaviour in the contact line (non-indentation hypothesis), this surface can be used to calibrate the force axis, that is, to determine the photo-detector sensitivity.

In the following, we will show examples of force versus displacement curves. All the experiments were performed with a liquid cell, in order to study the behaviour of the surfaces in liquid environment. In figure 3 we report the force curves collected on a freshly cleaved mica surface in air. This is a typical test sample for this kind of investigations, and the curves shown in figure 3(a) are useful to describe the differences between approach and withdrawal curves. The main differences regard the contact line (1) and the position of the jump-off contact point (2). The contact line, is expected to be identical in approach and retraction conditions. The difference shown in fig 3(a) is due to several factors, most important of which is the scan speed and a non linear response of the Z piezo transducer. The second difference between the two curves is the jump to contact in the approach curve compared to the jump off contact in the retraction curve. In particular, the jump off contact occurs at much larger pulling (negative) forces respect to the jump to contact. This behaviour is a physical consequence of the meniscus interaction (mediated by the wetting layer) between the tip and a thin layer of water adsorbed onto the surface 18 .

160 Mica in Air 1.6 0.8

! 8

-0,8

@l>;

-1,6

.. •

.f -2.4

o

-3.2

..

Mg..1\! SL, A1 0 OM

-4

Displacement (nm)

FIG. 3 (a): Approach and withdrawal curves on a mica surface in air at 30 % relative humidity. (b): The structure of a vertical slice through freshly cleaved muscovite mica. Mica in liquid: withdrawal curves

Mica in liquid: approach curves 0,8 0.6

0.4

Z

0.2

GI

0

.5 I!

If

·0,2

-0.4 -0,6

2

'''\i

I~~A~

.=.Z

0

GI

~

of

-1

-2

r ······ ..·· ········ .. ·~·I+············ .. ··;···············~

-0.86'::-0---=8'::-0----::'0"=0----::12"=0----:-'-:-----'

Displacement (nm)

Displacement (nm)

FIG. 4. Force curves acquired on mica surface in liquid at various pH. The curves are artificially shifted to evidence the differences. 4(a) Approach curves: note the disappearance, at increasing pH, of the attractive regime. 4(b) Withdrawal curves: the pH modulate the jump off contact amplitude

The use of a liquid cell allows to perform more sophisticated experiments: it allows to remove the meniscus force and study the relevance of other phenomena (primarily double layer and Van der Waals forces). For instance, the force curves collected on the very same mica sample in liquid at different pH in the range 5 to 8 (and low ionic strength: 30-50 JlM) are reported in figure 4. Comparing the withdrawal curves in air and liquid (Figure 4b and the red curve of Figure 3a), a strong reduction of the jump off distance is observed. Moreover, the chosen pH range is large enough to evidence a trend in the forces behaviour: at increasing pH the gradual disappearance of the attractive regime (approach curves) and a strong reduction of the amplitude of the jump off contact (withdrawal curves) is observed. The disappearance of the attractive

161

regime, in particular, is due to the balance of the attractive Van der Waals forces by the double-layer forces produced by the resident charges (ions and surface dissociation) in the liquid. These introductory experiments should have clarified that the force curves are useful and particularly sensitive tools to study the physics of surfaces and the variation that a sample can experience, even in chemical terms and over time, when exposed to specific environmental conditions. 3.2

SNOM: instruments development and mode of operation. The microscopy group at ISM has a long term experience in the development of new SPM instrumentation for a variety of applications. Such an activity produced a continuous improvement in the performances and range of application of the microscopes and have resulted also in several patents. One, in particular25 , regards a novel method to oscillate and detect the motion of an optical fiber that has been implemented in a SNOM, resulting in a tapping mode operation for an aperture microscope employing a non bent optical fiber. Running a SNOM microscope in tapping mode, rather than with the usual shear mode, has several practical and conceptual advantages. For instance allows to extend the huge knowledge and expertise developed for the tapping mode AFM to the SNOM world and allows to better control the physical parameters of the data collection and to obtain quantitative physical information on the sample. There are other technical advantages when choosing a tapping mode operation. Usually, the collection of the optical information with a SNOM requires the use of a lock-in amplifier to largely improve the quality of the signal and to get rid of stray fields and environmental noise. As a reference for the lock-in, is common practice to modulate the excitation laser (see figure Sa) using a chopper or an electronic switching of the laser itself. The use of a tapping mode oscillation allows to access a different modulation method: the vertical oscillation of the tip can, indeed, produce an intrinsic modulation of the signal (large signal for tip close to the surface, small signal for tip far from the sample) that can be used to operate the lock-in (see figure Sb).

162

FIG 5. The different SNOM setup employed. They differ by the method to modulate the illumination in order to lock the reflected signal: left= laser modulation; modulation Such a tip-modulation results in an and almost completely t>ac:Kg,rouna--tn:e collection that is not available in shear mode. In that case, indeed, the net vertical component of the tip's lateral oscillation is too little to introduce the required modulation in the A comparison between the two different chopping methods modulation and tip-modulation) is shown in figure 6, where the very same area of a is imaged by an aperture SNOM using a «lOOnm uncoated) fiber for the collection of the local reflectivity signal. The sample is constituted by an oxide (WOx) deposited on a zrOz substrate through trivial wet-chemistry method and a morphological transition associated subsequently annealed at 800°C in order to to a sharp enhancement of its catalytic activity (for more details about the see Colonna et al.

FIG. 6_ Comparison between(5j,tm x 5 j,tm) images collected with (a,b,c) laser- and (d,e,f) tip-modulation set-ups. Panels (a) and (d) are topog,raphies while (b, c) and (e, f) are local retlectivities acquired with different collection parameters. Images collected in tip-modulation clearly show more structural details and a better resolutions.

163

The data of figure 6 clearly evidences that better imaging are obtained when the tip-modulation is used. The data of figure 6b, indeed, contains more structural details which are the consequence of a lesser background and of a local enhancement of the actual optical signal produced by the tip at the apex of its oscillation. It is worth noting that the setup just described is similar to the setup typically used in the apertureless SNOM, and yet the standard interpretation of the apertureless field-enhancement (plasmon resonance, field confinement etc.) cannot be directly invoked in the present case. As consequence, the reported results are more likely to be the consequence of a predominant "cleaning" and/or refinement of the local signal which is expected to be further improved through tip-engineering. AFM and SNOM applied to the study of chromosomes 3.3 Deoxyribonucleic acids spend most of their time in eukaryotic cells as extremely long and hierarchically organized filaments, while they condense in the macroscopic structure of chromosomes during the metaphase of the cell cycle. This is probably the best time to study, in the condensed phase, how the genomic information is passed by, and to study the stability of the genetic structure in transgenic organism, i.e. when external genes (or whole operons) are introduced in a given specie. A very challenging, although debated, field of study regards the production of transgenic species, typically pigs or sheep, in which certain genetic material is introduced into the pristine genome in order to lower the probability of organ reject in a "X specie-to-human" transplant. The research strategy used in the case of sheep cells consists in producing in vitro a micro-chromosome using the centro mer of the animal lines to which the selected trans-genes are linked. Suitable telomere material may be added as well, in order to facilitate the acceptance of the micro-chromosome into the cell line. Beyond the complex production of the micro-chromosome, several problems can be encountered in the evaluation of the chromosome stability after its introduction into the host cell line. Indeed, after internalization, the chromosome can be non-replicated or even expelled at the following cell duplication. Alternatively, it can be replicated but some genetic material, often relatively large amounts of non-coding sequences, can be added to the telomere extremity of the chromosome. In this latter case, the chromosome size increases generation after generation and, if such increase goes on unregulated, sooner or later the micro-chromosome becomes useless or is expelled by the cell. The standard molecular biology-based strategy to follow the process is based on the use of staining or fluorescent dyes. For instance, the efficiency of translation can be checked by inserting in the micro-chromosome the gene coding for a GFP (green fluorescent protein), while the presence of the chromosome after a given

164

number of cell

can be verified by the fluorescent of some micro-chromosome. Concerning this latter issue, COJmpllcatHJnS come from the size of the micro-chromosome that can be small at the limit of the optical resolution and by the effect of the fluorescent L~V'~LU"'" that can non-specific cellular and/or genetic material. The consequence is that both the presence and the putative size variation of the micro-chromosome cannot be verified. In this framework SPM techniques can a valuable role.

FIG. 1. Panel (a): AFM reconstruction of a sheep chromosome metaphase. Panel (b) higher resolution image taken on the top-right border of the metaphase that shows in detail the micro-chromosome. Panel (c): fluorescence high resolution image enhancing the fluorescence labeled centro mer.

In Figure 7, an AFM mosaic reconstruction of a sheep somatic metaphase is shown. The mosaic is necessary for the typical spread of metaphase chromosomes which is usually much larger then the maximum scan size of our AFM (40 !lm x 40 !lm). The screening of the metaphase allows to rec:ogm2:e every original chromosome and to identify a good candidate for the microchromosome 7b). Of course, the AFM can only provide a pure morphological characterization while the parallel use of SNOM, and in particular of its ability to couple morphology with optical signals is required to unambiguously identify the inserted genetic material (Figure 7c) by detecting, for instance, the specific fluorescent labelling of its centromer. The shape of the candidate can be observed in detail by AFM and its size can be measured and compared to known chromosomes. In particular, we can compare the micro-chromosome size with, for instance, the size of chromosome 1 (or with any chromosome whose base-pair content is known).

165

FIG. 8: of the of the micro-chromosome and with the size of a chromosome of known base-pair content. The major error source is the in chromosomes of different sizes. an evaluation of the 100 nm tip-convolution effect, the measured dimensional ratio is 14,1. This can be used to estimate the micro-chromosome base-pair content.

This provides an independent estimate of the contained in the micro-chromosome. Such method can, obviously, be employed to report the occurrence of chromosome size increase after a certain number of cell division, i.e. the final chromosome ---'-----J 3.4 AFM and SNOM applied to the study of extraterrestrial Samples The association between the most abundant population of meteorites, the so chondrites" (DC), and their parent airless bodies is one of the caned main in the quest to understand the evolution of the solar system. Such association is mainly inferred from the analysis of spectra in the visible and near-infrared range, where the large majority of the asteroids show reddened reflected curves. A spectrum is usually defined "reddened" when it shows reflectivity at the wavelengths of red and infrared. The phenomenon of the reddening consists in the fact that the reflectivity an asteroid collected when the asteroid is still in space and when it has landed (or crashed) on Earth are different. In particular, the spectrum in space is reddened. In the analysis of lunar soils, the reddening of the spectra was attributed to the presence of nanometer-size metallic particles, and an interpretation of the

166

reddening was attempted by connecting it to the vaporization of Fe-bearing silicates. In this context, we have identified an alternative process for surface alteration of aidess bodies that can be invoked to solve this n"l~Tl>r" shock-induced phase transformation of Fe-Ni alloys. An AFM meteorites. The mechanical ,",Vll"llHII5 the that would characterization of oe evidenced the presence of a number of metal or metalsilicate inclusions in the meteorites. Moreover the presence of nanostructured of the metallic inclusions. material has been observed inside or on the LAaUJI,",'~;" of that are in 9.

FIG. 9: An olivine inclusion analyzed by SEM (panel a, ba(:ksl~att:ering) b, derivative The presence of nanoparticles on the of the oliviue can be Self evident nanoparticle structure can also be observed in observed in the AFM martensitic contained in the meteorite (panel c)

A allowed to relate the presence of nall0J)aJ"tlCl martensitic in the meteorite. Martensite is a metastable characterized by a crystallographic Bee structure that can be obtained and an austenitic phase (crystal structure austenite to martensite transition is. thus, accompanied by a transition and by the production of nanoparticles. This occurs, for mstaIllce,

16?

forge, the well known treatment used for hardening swords and, in general, metal-containing material. In the forge process, the phase transition is driven and triggered by mechanical shocks that, in the case of meteorites may be associated to the shocks experienced as consequence of hits and bombardment with cosmic powders and other micrometeorites. The data of figure 10 show that, remarkably, the reflectivity spectra collected by micro-spectrophotometry on small sections of the meteorites appear to perfectly match the martensitic content (i.e. the nanoparticle amount).

FIG. if. On the left panel, the meteorite sections (from A to E) where the reflectlYtty spectra (reported on the right panel) have been collected. Red spectra in A,B,C,D and E sections have, respectively a 0.05, 1.00, 0.85, 0.45, and 0.05% Martensite content. The green spectra have been collected as reference spectra on sections with a Martensite content below 0.05%. With the purpose to shed more light on the mechanism of nanophase induction associated with the FCC to BCC crystal structure transition and to the corresponding reddening of the reflectivity spectra, terrestrial steels have been investigated as well. In particular AISI 286 and AISI 316 steels have been mechanically treated by air blast shot peening with 100 \tm silica bullets for different times and their nanoscale morphology and reflectivity spectra have been characterized. The results, reported in figure 11 show the induction of nanopartieles and the simultaneous reddening of the steel reflectivity spectra in a way that agrees completely with the results observed on the meteorites and definitely support the hypothesis of the mechanical-induced phase transition. More information on the formation mechanism can be gained by using SNOM as a refinement technique that, retaining the capability to characterize nanophases spread into rocks or inclusions, can simultaneously collect the local reflectivity signal and this can be used to understand the characteristics and the composition of the observed nanophase. In particular one target of the study could be to distinguish, using a multi-wavelength approach, metal (optically featureless), olivine (tipically, one minima at 950 nm) and pyroxene (tipically,

168

two minima at 900 and 1800 nm and one maxima around 1250 data are available """\-a'I<"'.Y of SNOM as a nanoscale

AISI316 1,30

1.10

500

1000

of AISI 286 (top) and AISI 316 (bottom) steels r",""",f'li"",'I" and after (right panels) 180 second of treatment are 2/lm x The treatment the induction of a whose extent is proportional to the duration of the treatment. the reddenmg of the spectra can be observed and it increases at exposures.

4.

CONCLUSIONS In conclusion, some application of AFM and SNOM have been described in order to show some potentialities of such techniques in a of different The given examples are taken from the recent scientific fields of of the SPM group at the CNR-ISM where, in the of the techniques, some of the first microscopes operating in Italy have been __ ... ""..__ and built. AFM and SNOM are, nowadays, well established whose is continuously increasing thanks to their peculiarities as well as to the work of a community of researcher and to the encouragement and active ""r,nArl ,.,rr"",1",r! by the most visionary and ambitious scientists. Paolo Perfetti's work and career deserve a mention for him in the core of this vv.UHJCH.UJUJ

Acknowledgement We wish to acknowledge Prof. George Baran, Dr. Pier Francesco Fabrizia Somma, Prof. Adriana Maras for the support provided for the preparation and the helpful discussion.

169

REFERENCES

1. Meyer E, Hug HJ, Bennewitz R. Scanning Probe Microscopy: The Lab on a Tip. Heidelberg: Springer-Verlag; 2003. 2. Binnig G, Rohrer H. In touch with atoms. Review of Modern Physics. 1999;71 :S324-S330. 3. Hansma HG, Pietrasanta L. Atomic force microscopy and other scanning probe microscopies. Current Opinion in Chemical Biology. 1998;2:579-584. 4. Giessibl FJ. Advances in atomic force microscopy. REVIEWS OF MODERN PHYSICS. 2003;75:949-981. 5. Jandt KD. Atomic force microscopy of biomaterials surfaces and interfaces. Surface Science. 2001;491:303-332. 6. Dunn C. Near-Field Scanning Optical Microscopy. Chemical Reviews. 1999;99:2891-2927. 7. Wu S-f. Review of near-field optical microscopy Front Phys China. 2006;3:263-274. 8. Lange Fd, Cambi A, Huijbens R, et al. Cell biology beyond the diffraction limit: near-field scanning optical microscopy. Journal of Cell Science. 2001;114:4153-4160. 9. Binning G, Rohrer H, Gerber C, Weibel E. Surface Studies by Scanning Tunneling Microscopy. Physical Review Letters. 1982;49:57-6l. 10. Simpson GJ, Sedin DL, Rowlen KL. Surface roughness by contact versus tapping mode AFM. Langmuir. 1999;15:1429-1434. 11. Magonov SN, Elings V, Whangbo M-H. Phase imaging and stiffness in tapping-mode atomic force microscopy. Surface Science. 1997;375 L385-L391. 12. Girasole M, Cricenti A, Generosi R, et al. Different Membrane Modifications Revealed by Atomic ForcelLateral Force Microscopy After Doping of Human Pancreatic Cells With Cd, Zn, or Pb. Microscopy Research and techniques. 2007;70:912-917. 13. Rugar D, Mamin HJ, Guethner P, et al. Magnetic force microscopy: General principles and application to longitudinal recording media. Journal of Applied Physics. 199068:1169. 14. Girasole M, Longo G, Cricenti A. An Alternative Tapping Scanning Near-Field Optical Microscope Setup Enabling the Study of Biological Systems in Liquid Environment. Japanese journal of Applied Physics. 2006;45 23332336. 15. Tocha E, Schonherr H, Vancso GJ. Quantitative Nanotribology by AF¥: A Novel Universal Calibration Platform. Langmuir. 2006;22:2340 -2350. 16. Bustamante C, Rivettit C, Keller DJ. Scanning force microscopy under aqueous solutions. Current Opinion in Structural Biology 1997;7:709-716.

170

17. Dufrene YF, Lee GU. Advances in the characterization of supported lipid films with the atomic force microscope Biophysica Biochimica Acta. 2000;1509: 14-4l. 18. Butt H-J, Cappella B, Kappl M. Force measurements with the atomic force microscope: Technique, interpretation and applications Surface Science Reports. 2005;59: 1-152. 19. Duwez A-S, Cuenot S, Jerome C, et al. Mechanochemistry: targeted delivery of single molecules Nature Nanotechnology. 2006;1:122-125. Noy A, Vezenov DV, and M. Lieber C. Chemical Force Microscopy. 20. Annu Rev Mater Sci 1997;27:381-421. 21. Allison DP, Hinterdorfer P, Han W. Biomolecular force measurements and the atomic force microscope. Current Opinion in Biotechnology. 2002;13:47-51. 22. Zhang w, Zhang X. Single molecule mechanochemistry of macromolecules. Prog Polym Sci. 2003;28. 23. Cricenti A, Generosi R, Barchesi C, Luce M, Rinaldi M. A multipurpose scanning near-field optical microscope: Reflectivity and photo current on semiconductor and biological samples. Review of scientific instruments. 1998;69:3240-3244. 24. Girasole M, Pompeo G, Cricenti A, et al. Roughness of the plasma membrane as an independent morphological parameter to study RBCs: A quantitative atomic force microscopy investigation. Biochimica et Biophysica Acta - Biomembranes. 2007;1768:1268-1276. 25. Girasole M, Cricenti A, Generosi R, Longo G, Luce M. Dispositivo per l'eccitazione e la rivelazione dell'oscillazione di una fibra ottica. Pat. N.RM2004A000626. Italy; 2004. 26. Colonna S, Pompeo G, Girasole M, Gazzoli D, Pettiti I, Valigi M. Thermally-induced morphological transition in WOx deposited on a Zr02(lOO) substrate. Surface Science. 2007;601:1389-1393.

a-Sn/Ge(111) AND a-Sn/Si(111) SURFACES STUDIED BY STM

MEASUREMENTS AND AB-INITIO CALCULATIONS

S. COLONNA, A. CRICENTI, P. GORI and F. RONCI CNR - Istituto di Struttura della Materia, V Fosso del Cavaliere 100, 00133 Roma - Italy. O. PULCI NAST Centre and ETSF, INFM-CNR Dipartimento di Fisica, Universita di Roma Tor Vergata", Via della Ricerca Scientica 1, 1-00133 Roma, Italy. G. LE LAY CINaM-CNRS, Campus de Luminy, Case 913, F-13288 Marseille Cedex 09, France. UFR Sciences de la Matiere, Universite de Provence, Marseille, France.

Since its invention by Gerd Binning and Heinrich Roher at the IBM laboratories [1], Scanning Tunneling Microscopy (STM) has become one of the most popular investigation tool for surface science physics. STM is a powerful experimental technique that allows an unprecedented imaging resolution and an electronic structure investigation with a high local sensitivity. The development of this revolutionary microscopy technique earned its inventors the Nobel prize in 1986. The history of STM can be traced back to the first experiments of tunneling spectroscopy [2] and the attempts to control the electrons tunneling in vacuum with two electrodes brought few A apart [3]. These experiments, however, were oriented toward tunneling spectroscopy rather than microscopy. Only in a previous attempt a microscope system similar to STM was proposed working in the field emission limit [4]; this instrument was, indeed, acknowledged by the Nobel committee of the STM. The operating principle of STM is equivalent to the stylus profilometry, but instead of bringing the tip in mechanical contact with the sample surface a small gap is maintained and a beam of electrons

171

172

tunneling through the vacuum barrier scans the surface. STM images are obtained by approaching a sharpened metallic probe with apex of atomic dimensions at distances smaller than about 1 nm from a conductive sample, applying a proper voltage bias between the tip and the sample and recording the flowing tunneling current while scanning the chosen sample area. Both the in-plane (x-y) and out-of-plane (z) tip movement is performed by piezoelectric actuators that are able to move the tip with very high resolution and precision. Because the solutions of the Schrodinger's equation inside a rectangular barrier, modeling the vacuum region between the sample and the tip, have the form: (1) where k 2 = 2m (VB - E)/h 2 and VB is the is the potential barrier, this technique has a very high sensitivity in the z direction mainly due to the exponential dependence of the tunneling current on the distance. Hence, it is possible to image samples with very low corrugation, like e.g. metallic surfaces, at an atomic resolution level. Two operation modes are typically used in STM measurements: constant current mode and constant height mode. In the former mode, a feedback circuit operates the z-piezo actuator in order to keep the tunneling current to a constant value by retracting or approaching the tip while it is scanned on the sample surface; the movement of the z-piezo actuator is recorded in order to obtain the STM image. On the other hand, in the latter mode, the tip height is kept constant during the scanning process and the tunneling current is recorded. The first surface physics problems addressed by STM were the resolution of the Au(ll 0) and the (7x7) Si(1ll) surfaces reconstruction [5,6]. The success in imaging the (7x7) reconstruction attracted interest even outside the field of surface physics like, for example, the first theoretical approaches to the STM image interpretation by Tersoff and Hamann [7] and Baratoff [8]. Following the Bardeen perturbative approach [9] the tunneling current, in first order perturbation theory, is given by:

173

2 1= 2;LI',v{t(EI')[1- f(Ev)] - f(Ev)[l- f(EI')]}IMl'vI 8(Ev + V - EI')'

(2)

where feE) is the Fermi distribution function, V is the tip-sample bias and M/1v the matrix element which describes the electron tunneling in the perturbative approach [9]: (3)

where l/l/1 and l/lv are the wave functions of the two electrodes, the tip and sample in the STM case, In order to calculate the tunneling current, and hence the STM image or spectrum, it is necessary to have explicitly the wave function of both sample and the tip. According to the Tersoff-Hamann model the tip is locally approximated to a sphere of known radius and the electronic states in the tip are assumed with spherical symmetry (s waves). In this approximation and for very small tip-sample voltages the tunneling current reduces to [8]:

where per, EF ) is the electronic density of states of the sample surface around EF . At larger voltages, in the usual experimental conditions, Eg. (4) can be generalized by: (5)

neglecting the energy dependence of the tunneling matrix and the effect of the voltage on the surface wave functions. A more general expression of the tunneling current has been proposed by Selloni et al. [10]:

I oc

fEF+V JEF

per, E)TeE, V)dE

(6)

where the voltage dependent tunneling transmission coefficient TeE, V) takes into account the effects of the voltage on the surface wave

174

functions and the tunneling matrix. Thus, this simplified scheme shows that STM images are electronic density of states (DOS) per, E) maps rather than sample topographs like AFM, making STM a unique technique for studying the local electronic properties at conducting surfaces. Indeed, the very large potentiality of this technique becomes evident when it is used not just as a simple imaging tool, but rather in association with spectroscopy measurements (STS) in which a linear voltage ramp is applied to the gap while detecting the tunneling current with the feedback circuit switched off. Such STS I(V) curves contain information about the local electronic density of states (LDOS) and, after a proper normalization procedure, give a direct measurement, to a first approximation, of the band structure around the Fermi energy at atomic level. In fact from (6): dI(V) dV

oc p (E)T(E r, , V)

(7)

The coefficient T(E, V) can be assumed constant on a small voltage range or can be approximated by the total tunneling conductance [11]: T(E, V) ==

I(E,v)

v

(8)

STM and the related scanning probe techniques have evolved in the last years reaching a mature status opening the possibility to probe different properties of the sample surface. Good examples of the wide applications of STM measurements are spin resolved STM [12], high speed STM imaging for real time studies [13] (a beautiful example of STM dynamic study can be found in ref. [14] where the in situ growth of InAs quantum dots is monitored by real time STM measurements in a MBE chamber) and phonon detection on nanostructures [15]. STM has demonstrated to be a powerful investigation tool in studying systems where an accurate description of the surface density of electronic states is necessary. A good example of this kind of systems are the surfaces obtained by evaporating one third of a monolayer of group IV ad-atoms on Ge(111) and Si(111) substrates (the so called a

175

phase). Such surfaces are interesting systems for their triangular surface crystal lattice supposed to be good model systems for studying interesting electron correlation effects. Furthermore, theoretical calculations have predicted magnetic ordered insulating ground states for these surfaces at low temperature and even the possibility of surface superconducting phenomena [16]. In this paper we report a theoretical and experimental study of the aSn/Ge(lII) and the a-Sn/Si( Ill).

Sn/Ge(l11) and Sn/Si(111) systems: the a phase The a-phase is characterized by metal adatoms regularly located on one out of three T4 sites of the bulk terminated semiconductor (111) surface, resulting in a Ch x --J3)R30° reconstruction observed by STM at room temperature. In particular, the a phase of the isoelectronic Sn/Ge( Ill), Pb/Ge( Ill) and Pb/Si( Ill) interfaces have been extensively studied for their (--J3 x --J3)R30°+-+3 x 3 reversible phase transition observed by STM below a critical temperature [17-19]. An intriguing aspect of this class of surfaces is that this phase transition is observed at all the interfaces except the Sn/Si(lll) one, whose STM images maintain the (--J3 x --J3)R30° periodicity at temperature as low as 2.3K [20], in spite of the fact that they are isoelectronic and share the same atomic structure. On the other hand, the Sn core level photoelectron spectroscopy signature is equivalent for the a-Sn/Ge(1II) and the a-Sn/Si(1II) surfaces regardless of temperature indicating that the two systems should be equivalent [21,22]. Such puzzling evidences are still under discussion after more than 10 years since the first observation of the (--J3 x --J3)R30°+-+3 x 3 reversible phase transition at the a-Sn/Ge(lll) surface [18], in spite of the relative simplicity of such systems. Experimental The experiments reported in this paper were carried out using an Ultra High Vacuum (UHV) apparatus operating in our Institute since the

176

beginning of 2001. This experimental station is currently composed of two UHV connected vacuum chambers (base pressure 1.10- 10 mbar). The first chamber houses a Low Temperature STM (Omicron LT-STM; temperature range: 2.5 K..;- room temperature). Recently, the system was upgraded by adding at the STM stage a) a superconducting coil capable of generating a 400 Gauss magnetic field perpendicular to the sample surface and b) 4 spring-loaded electrical sample contacts that allow to insitu drive experimental devices, measure signals and/or apply additional potentials. Furthermore, the STM preamplifier was upgraded as well, enabling the detection of very small tunneling current (i.e. below 1pA) and thus allowing the measurement of samples with low conductivity even at very low temperatures. The second chamber, used for sample preparation, is equipped with LEEDIAuger optics, two evaporation sources (a Knudsen effusion cell and an electron bombardment cell), quartz crystal thickness monitor, ion gun and sample manipulator with direct and indirect sample heating. The effusion cell for Sn evaporation was thoroughly outgassed before use in order to maintain the pressure in the 10- 10 range during the metal deposition. The evaporation rate was measured by using the quartz crystal thickness monitor. Germanium and silicon substrates were cut from Ge(lll) and Si(lll) n-type wafers (resistivity 4 mOcm for Si and 0.3 Ocm for Ge). The sample preparation started from clean Ge(lll) and Si(lll) surfaces. The Ge(lll) c(2 x 8) surface was obtained, after sample degassing at 500°C overnight keeping the pressure in the 10- 10 mbar range, with 3-5 cycles of Ar+ sputtering (E = 500 eV, 1= 6 !-lA, t = 10 min., Tsample = 500°C) and annealing (T = 600-700 °C, t = 5 min.). The Si(lll) (7x7) surface was prepared by annealing at 900°C and flashing the sample at 1250 °C for a total time of about 60 s. The surface reconstructions were confirmed both by LEED and STM before Sn evaporation. A nominal 113 ML Sn deposition was performed at RT, followed by sample annealing (200°C and 650°C for Ge and Si respectively). Again, the formation of the a-phase (-Y3 x -Y3)R30° was checked by LEED and STM. Electrochemically etched tungsten tips were used after a cleaning procedure by field emission discharge against a metal electrode or electron bombardment. Careful attention was

177

devoted to thermal and piezo drifts by letting the instrument stabilize overnight every time the temperature was changed. a-Sn/Ge(lJJ) The Sn/Ge(111) system has been studied for its (-V3X\j3)R30° ~ 3x3 reversible phase transition, a well known, but not yet completely understood surface transition, initially explained as a Charge Density Wave (CDW) formation below a critical temperature [17]. However, due to the absence of nesting at the Fermi surface and to the presence of a double signature in the Sn-4d core-level spectra [21], this model has been questioned and, eventually, alternative models describing the -V3 x -V3 to 3 x3 transition as an order-disorder transition have been put forward. Among them, the "dynamical fluctuation" model [21] suggests that the 3 x3 reconstruction is the ground state and the -V3 x-V3 appearance, observed at high temperature, results from the thermally activated rapid oscillations of the Sn atoms. Below the critical temperature the ad-atoms fluctuations are frozen in the 3x3 periodicity which presents an hexagonal surface lattice with three Sn atoms per unit cell with two inequivalent Sn adatoms placed at different heights with respect to the underlying Ge substrate giving rise to a buckled surface. In Fig. 1 constant current STM images of the a-Sn/Ge(111) surface are shown. The images collected at room temperature are characterized by a hexagonal pattern in which the bright spots are Sn adatoms which appear all equivalent (except surface defects, usually vacancies or Ge substitutional defects). Decreasing the temperature below Tc == 220K, the a-Sn/Ge(lll) images exhibit complementary modulation of the surface with two brighter spots and one darker resulting in a honeycomb motif if imaged by empty electronic states; on the contrary an hexagonal pattern appears (with one brighter spot and two darker one) when filled electronic states are probed. The exact structure of the a-Sn/Ge(111) 3 x3 reconstruction is still a matter of debate. In fact, two possible configurations of this surface have been proposed, one with two Sn adatoms in a higher position with respect to the third one (two atoms in up position and one in down position, 2UID for brevity hereafter) and the opposite configuration (1 U2D). Different surface-sensitive structural

178

techniques applied to the a-Sn/Ge( Ill) surface have produced conflicting results [23-26]. Recently, non-contact AFM investigations of the a-Sn/Ge(lll) surface have also been carried out [27]. Even if this technique should in principle provide a true topographic description of the surface, it is shown how honeycomb or hexagonal images can result for the 3 x3 phase depending on tip-sample distance. The STM technique provides a picture of the surface electronic density of states and, hence, is not capable to tell whether the surface configuration is 1U20 or 2U 1o. Indeed, imaging the empty or the filled electronic states results in a honeycomb (an apparent 2UI0) image or a complementary hexagonal (an apparent 1U20) one, respectively, with the average image flat similar to the C../3x--.j3)R30° phase observed at room temperature. In order to assess the issue of the real (3 x 3) surface structure we employed a combination of first-principle calculations and bias dependent STM imaging. First-principles calculations have been carried out using a repeated slab geometry consisting of six Ge layers of 9 atoms each, saturated by H atoms on the bottom layer and with Sn adatoms on top. Calculations of electron eigenenergies have been performed within a OFT+GW approach [28] to allow a close comparison of theoretical and experimental results. Norm-conserving LOA pseudopotentials have been employed (ultrasoft gradient-corrected pseudopotentials have also been tested, giving comparable results). A kinetic energy cutoff of 12 Ry, increased to 20 Ry for convergence checking, has been used in the expansion of wave functions in a plane-wave basis set. The 3 x3 surface Brillouin zone (SBZ) has been sampled by a uniform mesh of 15 x 15 k-points, both for OFT and GW calculations. A plasmon pole model has been applied to parametrize the dielectric function.

79

RT

180

simulated 5.0 x 5.0 nm2 STM (b) configurations. Panel c displays the cOI-respOIldiIlg of experimental constant current (0.2 nA, T=

In 2a and 2b calculated STM images for D reconstructions as a function of the bias voltage V are ..."' ....r""t'~rl configurations the simulated images obtained at than 0.2 V show a honeycomb pattern in both empty and filled states as from a careful inspection of the calculated structure. increasing the bias voltage, in the 1U2D case the filled states images gradually revert to the expected passing through an apparent -V3 x -V3 reconstruction at about while the empty states images preserve the honeycomb natteUrl. calculated images between 0.3 V and 1.0 V show well-known complementary honeycomb and hexagonal patterns (for empty and respectively) reported in many papers. Interestingly enough, a

181

further unexpected transition from honeycomb to hexagonal is observed in the empty states images at higher bias voltage. As a matter of fact, both the filled and empty states simulated images at 2.0 and 2.3 V show a hexagonal pattern. The 2UID simulated STM images reported in Figure 2b show an opposite behavior: increasing the bias voltage above 0.2 V, the honeycomb to hexagonal transition occurs in the empty states images, crossing the unbuckled appearance at about 0.28 V, while the honeycomb pattern is maintained in the filled states images. Further increasing the bias voltage, a new hexagonal to honeycomb transition is observed in the empty states series, resulting in a honeycomb pattern for both empty and filled states images. Summarizing the results obtained by analyzing the simulated STM images, we found that: a) at low bias voltage (i.e. lower than about 0.2 V), for both the lU2D and the 2UID models, STM images with honeycomb pattern are predicted in both empty and filled states; b) at intermediate bias voltage the simulated STM images of the I U2D system confirm previous results obtained at ± 0.55V [29] showing honeycomb and hexagonal patterns for empty and filled states images, while the 2U 1D results give the opposite condition; c) at high bias voltage, calculations predict that the STM images, for both empty and filled states, should provide a picture of the true surface reconstruction (i.e. hexagonal for the I U2D and honeycomb for the 2UID). Thus, theoretical calculations suggest that a series of bias dependent STM images is a convenient method to discriminate between IU2D and 2UID configurations. We therefore acquired a series of STM images at 78 K on the same sample area ranging from ±0.1 V to ±2.3V, reported in Figure 2c. The evolution of such experimental STM images is strikingly similar to the one reported in the simulated STM images series for the 1U2D configuration (Figure 2a). The first predicted transition from honeycomb to hexagonal at low bias voltage in the filled states series is clearly visible in the first three images collected in the 0.1 -:- 0.3 V range. In particular, the 0.2 V image shows an apparent 1/3 x 1/3 reconstruction, as predicted by the simulated STM image at 0.27 V. Furthermore, the second expected transition from honeycomb to hexagonal at high bias voltage in the empty states series was verified as well in the experimental

182

empty states STM images. From the combined theoretical and experimental study of the u-Sn/Ge(lll) surface it is possible to asses that the 3 x3 reconstruction is characterized by a 1U2D surface structure. This result demonstrates that the combination of image simulation and STM measurements probing different electronic states is an effective method for surface structure determination. The dynamical behavior of the surface atoms was investigated in our laboratory by monitoring the tunneling current as a function of time on top Sn adatoms. A similar method was successfully applied for the study of the flip-flop motion of the asymmetric dimers at the Si(lOO) 2x 1 surface [31,32]. These tunneling current vs. time traces ("current traces" hereafter) were acquired simultaneously to the acquisition of constant current images by interrupting the tip scan on a 80x80 grid over the chosen area. At every single grid point the STM feedback loop was switched off and the tunneling current was recorded during 12 ms with a sampling rate of 33 kHz. In this way, it was possible to obtain 6400 current traces on every STM image. Considering that all measurements were performed over a lOx 10 nm 2 area, the resulting distance between adjacent current traces is 0.125 nm, much smaller than the distance between Sn adatoms sitting on T4 sites. With this method, we can observe steps in the current trace if an adatom underneath the tip moves in the z direction between two stable levels. We will refer to this instability as "flipflop" throughout the paper to stress the fact that it occurs between two well defined z levels. Furthermore, it is possible to exactly locate on the images the position of the flipflopping adatoms. The driving force responsible for the u-Sn/Ge(lll) reversible (--./3 x--./3)R30° ~ 3x3 transition is an interesting issue which has been the object of extensive investigation. This surface transition can be understood in the framework of the dynamical fluctuation model. According to this model the surface ground state is the 3 x3 reconstruction having inequivalent adatoms; at temperatures higher than the critical temperature the fluctuation of the Sn atoms can induce the switching of the domain configuration. In fact, we have seen that the crystallographic cell contains three atoms, one at higher position with respect to the other two.

183

b

120K

80K 80K

o

2

468

Time (ms)

10

10

20

30

# of steps

FIG. 3: a) Typical current traces collected at various temperatures at the a-Sn/Ge(111) surface. b) The relative histograms obtained by summing the histograms of all the stepped current traces detected on the sampled area

Hence, it is possible to arrange three energy equivalent configurations of the surface domain, each configuration representing a minimum of the surface potential energy separated by an energy barrier from the other two minima. In Fig. 3 current traces collected at different temperatures on the a-Sn/Ge(lll) are reported (see the experimental explanation in the introduction for technical detail). It is worthwhile to stress that these are only an example of a large set of current traces collected at every temperature. Clearly, these measurements of the tunneling current as a function of time are characterized by an evident telegraph noise superimposed on the set-up value of the current. These steps demonstrate that the adatoms undergo fluctuation movements. These flip-flop movements are the result of thermal activation, as evidenced by the fact that the number of steps increases as the temperature is increased. This issue is better monitored in Fig. 3(b) where for each temperature an

184

histogram reporting the frequency of the observed number of steps is shown. Due to the thermal energy there is a finite probability that the surface domain can overcome the energy barrier and switch to a different configuration. Increasing the temperature, the thermal fluctuation intensity increases resulting in a higher transition probability. At temperatures well above the critical temperature the adatoms strongly fluctuate between the up and down position locally retaining the I U2D configuration. The observed fluctuations in the tunneling current are interpreted as a confirmation of the dynamical fluctuations model. Thus, the average fluctuation frequency 1 is expected to follow an Arrheniuslike trend as a function of the sample temperature:

1 = 10 exp[E~ kT

-

1

(1)

To better analyse the dynamical behaviour of this system we traced an Arrhenius plot of the mean value of the histograms with the relative standard deviation value (Fig. 4), from which we can derive an estimate of the energy barrier E~ from the slope of the linear fit.

240 200

160

T (K) 120

7.5

100

80

7.0 "-,6.5 ..5 6.0 5.5 5.0 0.004

0.006

0.008 liT (K')

0.010

0.012

FIG. 4: Arrhenius plot of the observed flip-flop frequency f at the uSn/Ge(1ll) surface.

185

We obtain Et = 13 ± 7 meV, which corresponds to a transition temperature Tt ~ 150 ± 80 K. The typical value reported for Tt is 220K [17,18], but 70K for a defect-free surface [35], which corresponds to Et = 19 meV and 6 meV, respectively. The Et and Tt values we find (despite the large error bars, most probably due to big variations of the defect density values throughout the experiments) are in the expected range and thus give a further confirmation to the validity of the reported data. To summarize the results of the STM investigation of the a-Sn/Ge( Ill) surface it is possible to assert that the observed transition at this surface can be described in a more general framework of order-disorder phase transition and that the stable phase is the 3x3 reconstruction. As demonstrated by band calculations and valence band photoemission measurements, the a-Sn/Ge(lll) surface posses a narrow conduction band resulting from the unsaturated Sn dangling bonds. As a consequence theoretical calculations have predicted that the strong electron correlation effects should lead to a Mott transition at the surface resulting in a magnetic ordered insulating surface [16]. These interesting calculations have stimulated a wide experimental activity aimed at the direct observation of a possible correlation driven surface transition at low temperature (below the (v'3 x v'3)R30°+-+3 x3 critical temperature). In particular at the Pb/Ge( 111) surface a glassy phase was observed below 76K and more recently at the Sn/Ge(l11) surface a new transition from the 3x3 to a low temperature v'3 (LT-v'3 phase hereafter). However, because the 3x3 reconstruction stabilization is the result of a subtle equilibrium between the elastic strain and the electron energy gain, extreme caution is necessary in the determinations of these surfaces. In fact, as demonstrated in ref. [19], STM imaging can influence the surface structure due to tip-surface interaction phenomena. In Fig. 5 STM images of the Sn/Ge(lII) surface acquired at 80K and 10K, (tunneling parameters: ±IV; InA) are presented. The empty and filled electronic states images collected at 80K show the well known complementary decoration of the surface, exhibiting a honeycomb and a hexagonal pattern respectively, typical of the 3x3 reconstruction. Decreasing the temperature to 10K the filled electronic states image shows an apparent flat surface, i.e. all the adatoms appear equivalent, in

186

agreement with previously reported studies [36]. The observation of this L T-'1'3 reconstruction below 25K was presented as the experimental evidence of a low temperature phase transition to a Mott insulating phase. Surprisingly, looking at the 10K empty electronic states STM image in Fig. 5, a sharp 3 x3 periodicity is still visible. This inconsistency between the empty and filled electronic states images calls for a more accurate investigation in order to understand the real nature of the uSn/Ge(1II) surface at low temperature. In particular, a close examination of the possible interaction effects between the microscope tip and the investigated surface becomes necessary. In order to understand the influence of the tip during the image acquisition, we report in Fig. 6 filled states STM images collected at different tunneling currents (gap voltage -I V, sample temperature 5K) along with the- related 2D-Fourier transforms. The STM image collected at InA exhibits a clear '1'3 periodicity confirmed by the related Fourier transform too; decreasing the tunneling current to 0.2 nA, the 3 x3 periodicity becomes again visible in both the topographic image and the Fourier transform; further decreasing the current down to 0.05 nA no changes are observed in the STM image, indicating that below 0.2nA the expected topography is recovered. The observation of a 3x3 periodicity and a metallic characteristic at the surface contradict the low temperature MIT proposed in Ref. [36]. In fact, the '1'3 periodicity is observed only in filled states images and under specific measurement conditions, namely high tunneling current, pointing to a tip surface interaction effect rather than an intrinsic behaviour of the sample surface.

187

0,-

188

189

collected at low temperature. As demonstrated in previous works, a flickering STM image is the fingerprint of fluctuating adatoms [32,37]. It is possible to argue that below 20K a tip-surface coupling mechanism exists which activates the Sn adatoms flip-flop fluctuation. An analogous trend is operative at the Si(l 00)(2x 1) surface, where it was experimentally demonstrated that the flip-flop frequency of the Si atoms in the dimers, at low temperature, depends on the tunneling current[38]. Taking into account inelastic tunneling phenomena the dimers fluctuation was calculated to non-linearly depend on the sample temperature and tunneling current [33,34]. It is possible to argue that a similar inelastic tunneling process could be active at the a-Sn/Ge(lll) surface, giving rise to an apparent flat surface at specific tunneling conditions which could explain the appearance of the LT -~3 phase. It is worthwhile noting that the low temperature fluctuation of the Si dimers at the SiC 100) surface can also be excited by the electron beam in Low Energy Electron Diffraction (LEED) measurements as reported in Ref. [39]. Assuming a similar behaviour of the Si(lOO) and a-Sn/Ge(lll) surfaces it is possible to explain why the first STM observation of the LT-~3 was supported by LEED patterns showing a ~3x~3 periodicity. As a last remark it is interesting to note that the hypothesis of a low temperature transition at the So/Gee 111) surface was confirmed by valence band photoemission spectroscopy (PES) measurement [36]. The PES symmetrized spectra showed a gap opening below the critical temperature. This behaviour was explained as the fingerprint of a Mott transition. On the contrary, our tunneling spectroscopy measurements exhibit a metallic surface at a temperature as low as 2.5K. This seeming inconsistency can be explained by taking into account that PES measurements in Ref. [36] were performed along a fixed direction of the k-space. Whereas, STS measurement are k-integrated, thus more representative of the surface metallic state. Based on the reported experimental results the presence of a low temperature transition at the aSn/Ge(lll) appears untenable.

190

uuurs'''''' obtained at at at voltage: -1.0 feedback current: 1.0 nA.

(1-

191

In this case, the interconversion energy barrier Et between the three different 3 x3 configurations should be much smaller than in the Sn/Ge(111) system and, as a consequence, a transition to a 3 x3 reconstruction should be visible by further decreasing the temperature. In this view, we tried to reach the lowest possible temperature to search for a transition to a 3x3 reconstruction in the STM images. We reached a minimum temperature of 2.3 K (see Fig. 7b) at which the a-Sn/Si(lll) still retains the ...J3 x...J3 reconstruction. To better comprehend this surprising behavior, we performed a series of measurements varying the temperature from 80 K to 2.3 K avoiding the surface regions adjacent to defects.

Ti Ille

(illS I

FIG. 8: Typical current traces detected at 4.6 K at the a-Sn/Si(lll) surface. A flat trace is reported for comparison purpose.

Stepped current traces were clearly observed only for temperatures below 32 K. In Fig. 9 a small representative set of current traces collected at the a-Sn/Si(lll) surface at different temperatures is reported (it is worthwhile noting that each current trace is representative of the whole measurement set at each temperature). A remarkable decrease of the average fluctuation frequency 1 is observed from 32K (~800 Hz) to 15K (~150 Hz), as expected for a thermally activated process. Unexpectedly, by further decreasing the temperature down to 2.3 K, no significant frequency change is observed. We calculated the average fluctuation frequency 1 for all the measurements in which stepped current traces were observed. The results are reported, together with the

192

ones obtained for Sn/Ge( Ill), in Figure 10 as an Arrhenius plot (In 1 vs. liT, panel a) and as a function of temperature (1 vs. T, panel b), where the error bars represent the standard deviation. Making the assumption that the value of the pre-exponential factor fo was the same in both cases, we obtain that the frequency reduction observed at the Sn/Si(lll) surface between 32K and 15K would result in an interconversion energy barrier value of E~ = 3.0 ± 0.5 meV (solid red line). Figure lOb clearly shows that, given the detectable fluctuation frequency range, stepped current traces are observable, in the case of the Sn/Si(lII) system, at lower temperature and in a much narrower temperature range with respect to Sn/Ge( 111).

,

.'

6

7

K

9

In

II

12

Time (ms)

FIG. 9: Typical current traces collected at various temperatures at the uSn/Si(lII) surface.

According to this model, the Sn/Si(lII) surface should freeze to a 3 x3 configuration at temperatures lower than about 15K. On the contrary, we found no transition and the persistence of almost constant frequency fluctuations, suggesting a residual form of instability which prevents the

193

stabilization of the 3 x3. A possible explanation of this instability is suggested by the a constant jump rate as a function of temperature. In fact, as demonstrated in different physical systems a constant jump rate as a function of temperature between two stable states is the fingerprint of quantum tunneling phenomena [41-45]. It is possible to envisage that in the case of the a-SnlSi( 111) surface due to the very low Et value it is possible that a single surface domain could switch the configuration through a quantum tunneling process. Thus decreasing the temperature below thermodynamic critical temperature the expected surface transition is not permitted because the quantum phenomena give rise to residual instability of the Sn adatoms.

100.0 20.0

5.0

10.0

T(K)

2.5

_-----1--------1--1" -- -------------11000

----- .. -.... o

----

50

- -- - 'SnlGe(III)!(E1=13 meV) --SnlSi(1ll)f(E1=3 meV)

Cfl

'-'

"'-,

100 0.00

I I I I I I I I I I I I I I I I

j 0.05

0.l0

0.l5

150

lOOT (K)

• •

SnlGe(1ll) expo data SnlSi( III) expo data

Ii 0.20

200

f1

j 0.25

0.30

0.35

0040

0045

liT (K-l)

FIG. 10: Arrhenius plot of the observed flip-flop frequency at the aSnlSi(111) surface and, for comparison purpose, at the a-SnlGe(111) surface.

194

Conclusions In summary, the study of the a-Sn/Ge(111) and a-Sn/Si(111) systems demonstrates that STM is a very effective investigation tool for surface physics mainly for its ability to provide spectroscopic information with a high spatial resolution. From the above measurements we demonstrate that the two systems are equivalent, the only difference being the height of the energy barrier between the different domain configurations, which results so small in the case of Si that possible tunneling phenomena of the Sn adatoms prevent the 3x3 stabilization. Moreover, the flip-flop detection at these surfaces has demonstrated that the STM can be a useful detection method for slow surface dynamic phenomena. This kind of experiment is very promising for future dynamic studies if ad hoc experimental set-up are taken into account. REFERENCES [1] G. Binning, H. Roher, Ch. Gerber and E. Weibel, Phys Rev Lett. 49 (1982) 57 [2] I. Giaver, Rev. Mod. Phys. 46 (1974) 245 [3] W.A. Thompson and S.F. Hanrahan, Rev. Sci. Instrum. 47 (1976) 1303 [4] R. Young, l. Ward and F. Scire, Rev. Sci. Instrum. 43 (1972) 999 [5] G. Binning, H. Roher, Ch. Gerber and E. Weibel, Surf. Sci. 131 (1983) L379 [6] G. Binning, H. Roher, Ch. Gerber and E. Weibel, Phys Rev Lett. 50 (1983) 120 [7] l. Tersoffand D.R. Hamann Phys. Rev. B 31 (1985) 805 [8] A. BaratoffPhysica B 127 (1984) 143 [9] l. Bardeen, Phys. Rev. Lett. 6 (1961) 57 [10] A. Selloni, P. Camevali, E. Tosatti and C.D. Chen Phys. Rev. B 31 (1985) 2602 [11] l.A. Stroscio, R.M. Feenstra and A.P. Fein Phys. Rev. Lett. 57 (1986) 2579 [12] M. Bode, Rep. Prog. Phys. 66 (2003) 523 [13] R. Curtis, T. Mitsui and E. Ganz, Rev. Sci. Instrum. 68 (1997) 2790 [14] S. Tsukamoto, T. Honma, G.R. Bell, A. Ishii and Y. Arakawa Small 2 (2006) 386

195

[15] M Ouyang, J. Huang, C. Li Cheung, C.M. Lieber Science 291 (2001) 97 [16] G.Profeta and E. Tosatti, Phys. Rev. Lett. 98 (2007) 086401 [17] J. M. Carpinelli, H. H. Weitering, M. Bartkowiak, R. Stumpf, and E. W. Plummer, Phys. Rev. Lett. 79 (1997) 2859 [18] J. M. Carpinelli, H. H. Weitering, E. W. Plummer, and R. Stumpf, Nature 381 (1996) 398 [19] I. Brihuega, O. Custance, R. Perez, and J. M. Gomez-Rodriguez, Phys. Rev. Lett. 94(2005)046101 [20] F. Ronci, S.Co\onna, A. Cricenti, G. Le Lay, Phys. Rev. Lett. 99 (2007) 166103 [21] J. Avila, A. Mascaraque, E.G. Michel, M.C. Asensio, G. Le Lay, J. Ortega, R. Perez and F. Flores Phys Rev. Lett. 82 (1999) 442 [22] L. Ottaviano, G. Profeta, S. Santucci and L. Petaccia, Surf. Rev. Lett. 9 (2002) 675 [23] J.S. Okasinski, c.Y. Kim, D.A. Walko and MJ. Bedzyk Phys. Rev. B 69 (2004) 041401(R) [24] O. Bunk, J. H. Zeysing, G. Falkenberg, and R. L. Johnson Phys. Rev. Lett. 83(1999)2226 [25] M.E. Davila, J. Avila, M.C. Asensio and G. Le Lay Phys. Rev. B 70 (2004) 241308(R) [26] T. Lee, S. Warren, B.C.C. Cowie and J. Zegenhagen Phys. Rev. Lett. 96(2006)046103 [27] I. Yi, R. Nishi, Y. Sugimoto and S. Morita, Appl. Surf. Sci. 253 (2007) 3072 [28] S. Baroni, S. de Gironcoli, A. Dal Corso and P. Giannozzi, http://www.pwscf.org; F. Aryasetiawan and O. Gunnarsson, Rep. Prog. Phys. 61 (1998) 237. [29] L. Jurczyszyn, J. Ortega, R. Perez and F. Flores, Surf. Sci. 482-485 (2001) 1350 [30] A.V. Melechko, J. Braun, H.H. Weitering and E.W. Plummer Phys. Rev. B 61 (2000) 2235 [31] T. Mitsui and K. Takayanagi, Phys. Rev. B 62 (2000) R16251 [32] K. Hata, Y. Sainoo, and H. Shigekawa, Phys. Rev. Lett. 86 (2001) 3084 [33] H. Kawai, Y. Yoshimoto, H. Shima, Y. Nakamura and M. Tsukada J.Phys. Soc. Jpn. 71 (2002)2192 [34] H. Kawai and O. Narikiyo J. Phys. Soc. Jpn. 73 (2004) 417

196

[35J A. V. Melechko, J. Braun, H. H. Weitering, E. W. Plummer, Phys. Rev. Lett. 83(1999)999 [36J R. Cortes, A. Tejeda, J. Lobo, C. Didiot, B. Kierren, D. Malterre4 E. G. Michel, and A. Mascaraque, Phys. Rev. Lett. 96 (2006) 126103 [37J T. Sato, M. Iwatsuki, and T. Tochihara, J. Electron. Microsc. 48 (1999) l. [38J S. Yoshida, T. Kimura, o. Takeuchi, K. Hata, H. Oigawa, T. Nagamura, H. Sakama and H. Shigekawa Phys. Rev. B 70 (2004) 235411 [39J T. Shirasawa, S. Mizuno, and H. Tochihara Phys. Rev. Lett. 94 (2005) 195502 [40J M. E. Davila, J. Avila, M. C. Asensio, and G. Le Lay, Surf. Rev. Lett. 10 (2003) 981 [41J E. Merzbacher, Quantum Mechanics (1. Wiley, New York, 1970). [42J L. J. Lauhon and W. Ho, Phys. Rev. Lett. 85 (2000) 4566 [43J J. Heinrich, C. P. Lutz, J. A. Gupta, and D. M. Eiglerm, Science 298 (2002) 1381 [44J P. Ohresser, H. Bulou, S.S. Dhesi, C. Boeglin, B. Lazarovits, E. Gaudry, I. Chado, J. Faerber, and F. Scheurer, Phys. Rev. Lett. 95 (2005) 195901 [45J C. Z. Zheng, C. K. Yeung, M. M. T. Loy, and X. Xiao, Phys. Rev. Lett. 97,(2006) 166101

DEVELOPMENT OF SCANNING PROBE MICROSCOPY ELECTRONIC CONTROL AT ISM

M. LUCE, M. RINALDI, R. GENEROSI, A. CRICENTI Istituta di Struttura della Materia, CNR, via del Fossa del cavaliere 100, 00133 Rama, ITALY Abstract In this paper we present several implementations of control systems and

software for Scanning Probe Microscopy (SPM) that have been developed at ISM in the last 20 years. Our systems for controlling SPM employ proprietary DOS or Windows based software both for the data acquisition as well as for display and image processing. Introduction

Scanning probe microscopy (SPM) consists of a well established group of techniques that allow the imaging of surface on a nanometres scale. In addition of investigation of topography, information regarding the electronic states of the surface can be recovered by means of the acquisition of correlated electric or optical signals. The minimal set of tasks to be handled by an SPM control system are: rough approach between probe and sample, control of scanning probe on an atomic scale, and data acquisition and graphical reproduction of the scanned surface. Since the invention of scanning tunnelling microscopy! (STM), many other related experimental techniques have been developed: atomic force microscop/ (AFM) and near field optical microscopy3 (SNOM) are two of the most common. Several commercial SPMs are available; however they usually cannot be easily adapted to new approaches. Additions to the electronic hardware from the manufactures; software modification are even more difficult. As a consequence, researches working on SPMs are often involved in the development and testing of their own experimental apparatus. Discussion of several implementations of control systems and software for SPMs are widely described in scientific letterature 4 -7 • Many of the earliest articles describe system which were based on large and

197

198

powerful computer architecture; real time digital data processing as well as the possibility to handle huge memory blocks were available, until recently, only by using complex and expensive computers. The range of practical applications for digital technology in laboratory process control and real time data manipulating has expanded rapidly as the processing speed of microprocessors has increased. Most of the commercial systems for controlling SPMs employ proprietary DOS or Windows based software both for the data acquisition as well as for display and image processing. In particular. The Windows based software has many built-in advantages over DOS based software since it allows extended memory for running multiple applications as the same time. Thus making easy the interconnection among different tasks. Experimetal set-up

A low-cost data acquisition system for SPM based on a PC486 controller was first developed in the mid of 90' s for controlling an STM microscope. It consisted of a '486 based PC which controls STM scanner, DACs for voltage output, a digital proportional integrator (DP!), for controlling STM current, and three digital phase lock-in amplifier for data reading. With such configuration it is possible to perform costant current, spectroscopy and barrier height experiments simultaneously using both IEEE 488 and Keithley 110 ports. Fig.1 shows the block diagram of the data acquisition system 8. Four IEEE 488 register based DACs, supply ramp for the scanner, tip-sample voltage and a voltage reference for the differential amplifier which set the operating current. The electric diagram of the DAC was conceived to produce fast and automatic incremental ramps, so it has been provided with two external connector (Trig-In and Trig-Out). After the PC's IEEE 488 programming message, DAC start to stepping-up the output voltage each time an external trigger pulse occurs at its Trig-In. Such trigger is provided by the computer to the Trig-In of DAC-X, using a bit of 110 port. Once the DAC-X reaches its ramp end, i.e., one line of 512 points has been acquired, it provides, through its output Trig-Out, a trigger to DAC-Y Trig-In, that increases of one step. Afterwards the position of DAC-X is reset, and a new line can be acquired.

199

....lJlH"UlLllS

current is by an I/V converter on OPA128 operational amplifier, output (fixed at the operating This difference is than digitized and int.PfYf'",t,,·rI counting and accumulated in a digital to analog converter provides to

amplifier has been designed for our v .... ..,,""',,'" a frequency of several KHz, depending on input signal is digitized by a voltage to A phase adjustable square wave at signal by gating one up-down counter COllm~ct~:a output of the counter is then digitally logic unit. Constant can extem:al rotary switch, Sensibility of the lock-in HJV''''-U'

200 ~ 10

J.l V. Gain of the input preamplifier is adjustable by a programmable operational amplifier in the range 1-1000. Time constant can be set from 1 to 10 period of the sine wave supplied as modulating signal. The data acquired from lock-in or DPI are fed to the computer by using one I/O digital port. Four 16 bit data ports (having own handshake lines) can be addressed on the PI096 card. Each one controls a different operating mode of the STM. The next development was a control unit capable of controlling also an AFM and SNOM microscope: figure 2 is a schematic diagram of our SPM control and data acquisition system. The proximity signal measured by the probe (tunnelling current in STM, cantilever deflection in AFM and SNOM)9 is compared against an analog setpoint reference (REF) to obtain an error signal. This error signal is fed in a digital proportional integrator module (DPI) whose digital output is back converted into an analog value (D/A - Z) and finally amplified by a high voltage amplifier necessary to drive the z micropositioner of the piezo scanner. In Fig.l, the feedback loop is represented by a bold line. The experimental hardware is linked to a digital input-output board which is installed in an ISA slot of the personal computer bus. We used the model PI096 by Keithley which is organized as 12X8 bit wide ports. Some bits of one of the ports perform the enable and disable of several switches in the analog section of the hardware. These switches are used for controlling the feedback gain as well as for enabling feedback and modulation of the probe. Four digital to analog converters (D/A) are used for computer control of X scan, Y scan, probe voltage, and reference voltage. All the converters (DAC703 Burr-Brown bipolar ±10V, 16 bit resolution) are connected to the I/O port by means of a 2X8 bit memory latch (SN74LS373).

201

two ports are A pulse provided by a programmed into converter (high off) is about 60flS. z(x,y) is lTlt''''M''\"rpt"r! v(z) applied to the "'.,vu,'"'' while the current digital output DPI module is latching

same

'7_1"1,1<"7£>

as is

202

asynchronously latched by a trigger provided by an analog circuit acting as a window comparator (WC). Comparing the proximity signal and the reference (REF), this comparator supplies an output which is at a high level when both signals habe the same value (0). In such manner, topography values stored in the latch register (LATCH) are always related to the correct distance between the probe and the sample surface. The data acquisition handshake protocol assures that the scanning speed (X and Y driver voltage waveforms) is dependent on the degree of surface roughness: the rougher surface, the slower the scan. An advantage of such asynchronous data acquisition is the fact that the error signal (i.e., the difference between the proximity signal and the reference voltage) is negligible. In a synchronous data acquisition system, the data points are acquired at preset time points independent of the value of the error signal; the result is that the noise level in an image can be very high, depending on the electronic feedback performance, and even large vanatIOn (spikes) are taken as good points. It happens very often, particularly on modulation experiments (like tapping mode, noncontact AFM, and shear force SNOM) where the feedback is driven by a slow electronic apparatus like a lock-in amplifier, that the error signal has more information than the feedback voltage applied to the z axis of our system: in fact, a data point is stored only when the proximity signal equals the reference signal, i.e., the applied force is really constant for all data points. When a new point is stored in the register, a set-reset flip-flop (SR) is also triggered; this flip-flop asserts the data-valid (DAV) line of the handshake ring. Sensing this line, the computer proceeds to acquire the data into its memory, therefore moving the scanning probe to the next position of the rastered image. When a new position has been reached, the computer, asserting the handshake return line named data acknowledge (DAKN), resets the flip-flop and starts a new data polling of line DAV to acquire the next data point. The data acquisition sequence is described in Fig. 3. During the scanning of the topography, a samplelhold (SH) circuit used in conjunction with a 16 bit analog to digital converter (ADC) allow the acquisition of further data information. In STM mode of operation, i/v curves are measured at constant tip-

203

in NSOM are rl;~,an+ll" the ADC synchronously with the acquisition

scanner it is a samples with

to multiple

l ...... '!H,.~'
204

must be acquired and overlapped. A reliably performing system for large motion of the sample must be added to the microscopy setup. In our system, three dc motors (3 AXIS STAGE) fed by pulses, controlling x, y, and z axes, allow the rough movements of the sample in the range (8X8X1 mm). The data acquisition software has a dedicated procedure allowing the complete control of the three axes. Holding the correlation among the subsequent images, the investigation of large areas of a biological sample is therefore allowed. Controlling the three axis stage by computer is also extremely important for AFM studies of the growth of biological material since it allows the cantilever to be positioned over an area of interest. The area can be imaged, and the probe retracted to a appoint far above the sample; this is important to avoid any interaction between lever and the sample during material growth that might influence the process. The area of interest can then be visualized at any time depending on the particular experiment. Since the movement of the sample is computer controlled, the above procedure can be repeated several times without manual control, thus following in real time any variation process. The easiest way to develop applications for the Windows environments is the Microsoft Visual Basic programming system. The programmer create the application by drawing objects such as buttons, slider, option boxes, menu-bars in a generic form which will be used as the user interface. An object contained in the form reacts to each occurring event by executing a defined code procedure. The application can be organized in multiple windows, each one performing different tasks. It is very useful for developing data acquisition control panels as well as experimental setup panels. Our data acquisition system is based on a digital I/O interface board mapped in the memory of PC. The board uses 16 consecutive I/O addresses within the PC's I/O address space. The controlling software must be able to perform low level instructions to assign values to and read values from the physical addresses of the memory map of the Pc. Even though Visual Basic does not allow such instructions, with DLL (dynamic link library) the performances of the languages can be upgraded. In order to control the Keithley PIO 96 board, which is the central node of our hardware setup, a DLL named

205

InpOut. DLL has been used. InpOut. DLL is a complete «lNP» and «OUT» replacement for Visual Basic written in Assembler. To use it, it is needed to place the following declarations in the global or general module. Declare Function Inp Lib «InpOut.DLL» (Byval Port%) As Integer Declare Sub Out Lib «InpOut.DLL» (Byval Port%, Byval Value%) The procedures can be used just like their QuickBasic counterparts. OutPort%, Value----or X%=Inp(Port%) Initially, the front end of our software for controlling SPM appears such as a simple menu bar. The user can have access either to the data acquisition setup panels or to the images display and analysis program. The can operate in both the environments with the same graphical interface. In reality the images elaborating program and data acquisition program are two completely different applications. Visual Basic main program works as a command server for the client application PV-Wave. Since PV-Wave is a command based environment, it works by executing the commands typed at its prompt and followed by (carriage return). In the Windows environment, dynamic data exchange (DDE) would easily provides to data exchange between different Windows application. However the version of PV-Wave personal edition we are using do not support DDE. Nevertheless, because both applications run in the Windows operating environment, some communication is always possible. We established a server to client communication sending command to PV-Wave from Visual Basic as they were virtual keystrokes. Sending keystrokes to another application is very simple for Visual Basic. Only two commands are needed: APPACTIV ATE and SENDKEYS. To send keystrokes to another application by means SENDKEYS, the client application must be activated with the APPACTIV ATE instruction. The arguments for APP ACTIV ATE is the name of the application window where to send keystrokes:

206

APP ACTIV ATE «PV -WAVE main window» SENDKEYS «surface {enter}» To send control characters they have to be surrounded by brackets ( {}). In such a way, the connection among the global experiment graphical interface, which containing also the data acquisition program, and the separate image enhancement program was completed. PV-Wave is a Windows application with its editor and a command prompt that we used to implement the software for image processing and displaying. It is a powerful, interactive command language with several built-in functions and procedures that simplify the writing of the program code. It has all the program control options of a structured programming language in a easy-to-leam form. Every command or call to a routine must be typed to the command prompt and it is possible to generate pop-menu and widgets as well as buttons and slides. By choosing an element of the menu or pushing a button, an automatic call to a subroutine is generated. In this way it would be necessary only to type the command that creates these objects at the beginning of your work session. Our image enhancements program takes its input data from the data acquisition system which provides a square matrix containing information of the scanned surface, the range, and the maximum corrugation of the surface. The two dimensional image is displayed by mapping the values of the square matrix on a well defined palette of colours. This palette is composed by almost 256 colours (some colours are reserved by the PV-Wave) and it is possible to choose among 16 pre-defined palette or create a user defined palette. If the palette is made by a single-colour scale, as a grey scale, the intensity value of each pixel will be proportional to the data value: if the data are related to the topography, the darker is the pixel the lower is the topographic height at that point. Another important feature for displaying images is the possibility to select a subimage in order to cut a noised part of it or to focus the attention on a particular zone. In this case the selected sub image will have a bigger resolution because a smaller range of values are mapped in the same range of colours. In order to magnify or to narrow an image, a zoom function is available too.

207

Very often the image appears as it was laying on a sloping plane. The background plane removal filter calculates this plane, called background plane, by using the least square method and subtracts it to the image. Before the removal of the background plane, the two dimensional image appears with a colour saturated zone. Another important filter is the step-removing filter. It is utilized for removing the lines appearing sometimes in the images due to the electric and/or mechanical noise. In general, every removed line is replaced by a new one obtained as an arithmetic average of the two closest lines. During the scanning of the sample, a part of the tip could break off or a part of a soft sample could have been compressed by the tip. In both of these cases a step would appear in the image. The plane alignment filter can remove this step by subtracting from each row of data matrix its average and adding the average of the whole image. Smoothing and sharpening filters are also provided, the first primarily for diminishing spurious effects that may be present in a digital image as a result of a poor sampling system or transmission channel, and the second for highlighting edges. These filters are made with methods both in spatial or in the frequency domain; our smoothing filters (mean and median filter) utilize neighbourhood techniques while the sharpening filter uses Roberts gradient technique. In the frequency domain, it is possible to use a low pass filter for the smoothing and an high pass filter for the sharpening. The equalization filter achieves enhancement by modifying the colour histogram of a given image in a specified manner.

References lG. Binning, H. Rohrer, Ch. Gerber, and E. Weibel, Appl. Phys. Lett. 40, 178 (1982); Phys. Rev. Lett. 49, 57 (1982); Physica B 109/110,2075 (1982). 2G. Binning, C.F. Quate, and Ch. Gerber, Phys. Rev. Lett. 56, 930 (1986). 3U. Diirig, D. W. Pohl, and F. Rohnere, J. Appl. Phys. 59, 3318 (1986). 4T. L. Porter, Rev. Sci. Instrum. 64, 3530 (1983). 5E. I. Altman, D. P. Di LelIa, J. The, K. Lee, and R. J. Colton, Rev. Sci. Instrum. 64, 1239 (1993).

208

6A. J. Hoeven, E. J. Van Loenen, P. I. G. M. Van Hooft, and K. Oostveen, Rev. Sci. Instrum. 61, 1668 (1990). 7R. Piner and R. Reifenberger, Rev. Sci. Instrum. 60, 3123 (1989). 8c. Barchesi, A. Cricenti, R. Generosi, C. Giammichele, M. Luce, and M. Rinaldi, Rev. Sci. Instrum. 68, 3799 (1997). 9H. Muramatsu, N. Chiba, T. Ataka, H. Monobe and M. Fujihira, Ultramicroscopy, 57, 141 (1995).

NANOSTRUCTURES INDUCED BY THE ADSORPTION OF FULLERENES: STRUCTURAL AND ELECTRONIC PROPERTIES

ROBERTO FELleI t European Synchrotron Radiation Facility, 6 rue J. Horowitz Grenoble, F-38043, France

MADDALENA PEDIO TASC National Laboratory,CNR-INFM Triese, Italy The adsorption of C60 onto metal surfaces can lead to the formation of complicated reconstruction. In this contribution we show how the combination of inverse photoemission spectroscopy and surface x-ray diffraction experiments can give the details of the adsorption process providing the information on the bonding geometry and on the electronic states. The combination of these two techniques show that it is not possible to describe the process in terms of electronic exchange but the molecule plus substrate has to be considered as a whole for a full comprehension of these systems.

1. Introduction

The investigation of the correlation between electronic, molecular and structural properties of the interface between large organic molecules (LM)s with surface substrates is fundamental for the understanding of their interaction mechanism and attracts considerable interest because of their possible technological applications. Page numbers are included at the top of the page for your guidance. The final pagination of the volume will be done by the Publisher. Large molecules (LMs), upon adsorption on substrate surfaces, can form strong chemical bonds with to the topmost layer atoms and, because of their typical size in the nanometers scale, induce nano-restructuring of the surface. Complex molecules can be used as basic building blocks for molecular nano-devices based on self-organization where nano-size objects with specific shape, composition, and functional properties can be built on surfaces. Exploiting LMs for the fabrication of surface-supported functional nanostructures requires the

t

email: [email protected]

209

210

understanding and characterization of the structural and electronic properties of the molecular-surface interface, in particular in those cases where the interface is strongly modified by the interaction. Substrates surfaces may change drastically when foreign atoms or molecules are adsorbed on them. In the case of large molecular adsorbates the details are sometimes difficult to access because they extend over a large number of substrate unit cells and of sites. To control these systems we must answer the following questions: i) which is the part of the molecule interacting with the substrate playing the predominant role in the bond formation? ii) is the substrate structure modified and which is the new structure? Fullerenes are considered as model systems of 3 dimensional large molecules. They are widely used to create complex molecular structures on surfaces for their stability, electronic properties (doping), functionalization, capability of forming reconstructed surfaces. [1]. The paper is organized as follow: in section 1 the Inverse Photoemission Spectroscopy (lPES) and the Surface X-ray Diffraction (SXRD) techniques, focusing on fullerene films characterization, are shortly reviewed. Section 2 summaries the main results achieved on single C60 ordered monolayers.

2. Inverse Photoemission and comparison with other techniques When applied to C60 films IPES allows to study their electronic structure providing access to molecular empty states. In IPES, the electrons, that are generated by an electron gun with a preset kinetic energy, fall into unoccupied levels while emitting a photon with an energy equal to the difference between the initial kinetic energy and the energy of the unoccupied level. Thus, in principle, the final state of the system is obviously not the electronic ground state but an (excited, negatively charged) ionized state. Experimental details can be found in refs [2,3] and references therein. In combination with UPS, IPES is used to probe the correct values for the transport levels of the occupied and unoccupied levels directly by using ultraviolet radiation (for occupied valence levels, UPS) or electrons (for unoccupied levels, IPS) [see for example ref. 4]. IPES probing depth can be estimated from the mean free path of the typical incident electron energy (Kinetic Energy=4-30 eV) as smaller as 0.7 nm [5,6], that is the C60 diameter. Another useful method is the combination of near edge x-ray absorption fine structure (NEXAFS). Both IPES and NEXAFS are used to probe unoccupied states. In NEXAFS, the incident x-rays with an energy close to the absorption edge of one of the elements in the investigated material excite a core electron into an unoccupied state. Thus, information about unoccupied states can be derived, even though the interpretation may be complicated by the influence of the core hole.

211

In recent literature of molecules adsorbed on surfaces many studies are conducted by using NEXAFS [7] that provides information on the unoccupied states and molecular average orientation [8]. Here we discuss briefly the different processes at the base of these two techniques: IPES is based on the process consisting of a transition from a free-electron-like state above the vacuum level to a lower energy unoccupied state below the Fermi energy [3]. No core state is involved in the process and the final system has N+ 1 electrons. The final state is relatively non local and the spectrum closely approximate the density of states of the ground state, with possibly an uniform upward shift in energy levels [see for example ref. 9]. In NEXAFS the X-ray absorption fine structure of the x-ray absorption K-edges can be interpreted as due by dipole transition from the localized core level of the element of interest into unoccupied electronic states with dominant p character (Joint Density of States) [Stohr]. The presence of a core hole in NEXAFS generates an attractive potential; alternatively the creation of the core hole can be treated in terms of equivalent core or Z+1 approximation whose effect is to pull the ground state levels down in energy, as shown by the calculated NEXAFS spectra for C60 by density functional theory of ref. 10. In order to get standard information on molecular empty states we measured multilayer samples of C60 and C70 , deposited on different substrates at room temperature. The fullerene intermolecular interaction in the films is governed by Van der Walls interactions. The multilayer desorption takes place after thermal treatment at about 630 K for C60 and about 530 K for C70 • The IPES of multilayer polycrystalline samples of C60 and C70 , deposited on Si and Cu polycrystalline substrates, were measured together with the C K-edge NEXAFS. Fullerene C60 and C70 NEXAFS C K-edge spectra are shown in Fig.la) and compared with theoretical simulation performed by Density Functional Theory [11] taking the molecule with a core hole. C70 NEXAFS measurements is taken from ref. 12. Fullerene C60 and C70 IPES spectra are shown in Fig.l b) and compared with the analogous calculations performed by Nyberg [10] but for the ground state of the molecule, in absence of a core hole. The first antibonding states of fullerenes are characterized by Jt molecular states. The IPES spectrum of the C60 multilayer shows three extremely sharp features and two wider ones in agreement with previously reported data. In C K-edge NEXAFS spectra these molecular states show a different energy separation. The overall good agreement between the IPES spectra and the DFT ground state simulations confirms that information obtained by IPES can be related to the molecular ground state properties [13]. IPES C60 spectrum, presents the characteristic features of the localized antibonding Jt molecular states at 1.4 eV, 2.6 eV, 3.7 eV, that are assigned to LUMO (t lu ), LUMO+l (tl g) and LUMO+2 (hg+t2u) respectively, and have almost pure antibonding Jt character [14,15]. In NEXAFS the first 3 states (below 10 eV above the edge) have Jt* character, with energy transition at 284.5,

212

285.85 and 286.4 eV for LUMO, LUMO+l and LUMO+2 respectively. The Jt antibonding states in IPES C70 spectrum are located at l.2 eV (LUMO), l.95 eV (LUMO+l) and 2.85 eV (LUMO+2) , while NEXAFS at 284.5, 285.3 and 286.3 eV for LUMO, LUMO+l and LUMO+2 respectively. The features at higher energies present fraction or total a character. LUM)

C K-edge NEXAFS

IPES

Ground state " , , / \ ,I o

Full core hole

Ground state

284

286 Photon Erler

288 e

290

o

2 4 6 Energy relative to Fermi level (eV)

8

Figure 1. Left: Multilayers spectra ofC60 and C70. (a) NEXAFS at the C K-edge ofC60 (bottom) and C70 (up) are compared with the full core DFT calculations (lines) of ref [11]; (Right: IPES of C60 (bottom) and C70 (upper curve) are compared with ground state calculations of Nyberg [10].

As discussed in refs [10 and 16] the differences in the empty states IPES and NEXAFS spectra of C60 and C70 are mainly related to the LUMO+ 1 state. This feature is strongly perturbed by intermolecular interaction for C60 solid phase. The LUMO and LUMO+2 transition are lined up, while the LUMO+l state is shifted toward lower energies for the C70 respect to C60 in both techniques. The morphologic variation between the two molecules is related to the elongation of C70 along the "equatorial" plane that makes it more similar to a rugby ball. The breaking of symmetry perturbs more the Jt orbitals than the a orbital because the Jt are more sensitive to bond angle that is reduced for the C atoms at the equator of C70 • The features of C70 IPES of ref. [14, 16] show strong cross section effects, considerably varying the relative intensities with the probe energy from 19.25 to 32.25 eV. In our case the detected photon energy is 9.5 eV and the LUMO+l in case ofC7o results as a shoulder at about 0.75 eV from LUMO. Nevertheless the agreement with the assignment of the feature to the LUMO+ 1 state is confirmed

213

by the calculation shown in the figure for the C70 ground state. Moreover the less symmetric shape of C70 can introduce orientational disorder in the grown film, leading to less defined spectral features with respect to C60 spectra. Physisorbed single layer of C60 on surfaces leads to IPES features related to all the the Jt* molecular states lying at energies substantially unperturbed with respect to the multilayer spectrum [17] as reported for C60 interacting weakly with GeS surface. The comparison between IPES and NEXAFS shows that the density of states is modified substantially upon core-hole creation confirming that IPES spectra are a good measure of the quasi-particle density of state ground state. A discussion between IPES and NEXAFS measurements of C60 deposited on different materials is reported for example in refs [18 and 19].

3. Surface x-ray diffraction In order to study the adsorption of organic molecules at the surface of metal surfaces the standard techniques based on the use of x-rays are: X-ray reflectivity (XRR) and Surface X-Ray Diffraction (SXRD). These techniques are fully complementary and each of them provides a piece of information for the determination of the thickness and roughness of the deposited layer together with its atomic structure. XRR is based on the x-ray optical properties of each material. In the case of x-rays, the refractive index of a material is given by [20]:

1 2 i n = 1- - ' NZr, A + - ' NAil 2n 0 4n r'a

(1)

where A is the x-ray wavelength, N is the numerical density of the material, Z is the atomic number of the material, ro is the classical electron radius and f-la is the absorption coefficient. The imaginary part is typically one or two order of magnitudes smaller than the real component and it can usually be neglected in the calculation. The refractive index is always slightly smaller than 1 implying that for a given angle x-rays will be totally reflected when their wavelength A is larger than a critical Ac value and viceversa for a given A the x-rays will be reflected for an incidence angle smaller than a critical value, or cumulating the two statements when the vertical exchanged momentum q.l is smaller than a critical value q .l,C'

214

After having defined the refractive index the reflectivity as a function of q.L can be calculated using all the classical optics relations [21]. The reflectivity vs q.L and the refractive index vs z are connected by a mathematical relation, implying that by a measurement of R(q.L) it is possible to determine n(z). For q.L » q .L,C' it can be demonstrated that the reflectivity is given by the Fourier transform the of the derivative of the refractive index as a function of the distance from the surface, z [22]. A surface or an interface presents some kind of disorder, which is a generally described as a roughness. Its main effect is to decrease the reflected intensity. Mathematically this is represented by an exponential term depending on the average roughness dimension of the surface [23]. In the case of interface roughness this term can be directly included in the algorithm describing a sequence of distinct layers [24]. While for XRR we have not done any assumption to the sample's lattice properties, SXRD can only be applied in the case of ordered samples. This technique provides information on the structure of both the terminating layers of the ordered substrate lattice or of an ordered new structure which is formed at the surface. In the first case, the basis of SXRD is the measurement of the diffracted intensity raising from the termination of the periodic bulk structure which is located along rods in the reciprocal space having their origin at the Bragg points of the bulk structure and which are perpendicular to the terminating surface [25]. These rods are usually referred as Crystal Truncation Rods (CTRs). In the second case the diffracted intensity is along continuous rods, still perpendicular to the surface with an in plane periodicity defined by the surface unit cell. If the surface unit cell is due to a reconstruction of the bulk cells and then it has in-plane dimensions which are commensurate with the bulk unit cell, the associated rods take the name of Fractional Order Rods (FORs) [26]. A measurement of the intensity of the CTRs as a function of the continuous variable I, which is the reciprocal space coordinate perpendicular to the scattering surface, provides information on the expansion or contraction of the top layers of the substrate and on the occupancy (surface roughness) of the atoms on the topmost substrate layers while the measurement of the intensity of the FORs as a function of I gives access to the position of all the atoms belonging to the surface unit cell. To be noticed that in the FORs there is no information on the substrate structure being the contribution of the substrate atoms to the structure factors of the FORs always zero. To gather sensitivity to a particular atomic species, all the above techniques can also be coupled to energy scans through absorption edges. In the simple case

215

the change in the atomic scattering power at the resonance will give he possibility of determining the contribution of a particular atomic specie to the scattered intensity. In addition an exact determination of the dependence of the scattered intensity versus the energy shows intensity oscillation which can be interpreted using the same theoretical approach of EXAFS giving the possibility of determining also the chemical and structural local environment of the resonating element [27,28 and 29].

4.

C60 monolayer

The adsorption of organic molecules on metal surfaces can lead to complex nano-structuration of the supporting substrate. Thin films of C60 adsorbed on crystalline substrates usually tend to form hexagonal or quasi-hexagonal ordered structures with a large variety of the substrate-adsorbate interactions, ranging from covalent to ionic character. An important issue is the investigation of the consequences of the C60 adsorption on the underlying substrate structure, that in some cases, i.e. for Au(llO) [30], showed dramatic changes from the reconstruction of the clean surface. Surface X-Ray Diffraction (SXRD) measurements make this information accessible. We present results on the following systems: IML of C60 Pt(lll), and (2xl) Au(llO) surfaces. 4.1 C60/Pt(1l1)

The Pt(lll) surface has a close packed fcc structure and presents the higher work function among the ordered metal surfaces. This implies a very compact and stable metallic layer. Figure 4 shows a set of IPES spectra for low coverages of C60 deposited on Pt(lll) at RT and the single monolayer evolution after annealing at different temperatures. The bottom curves, related to the multilayer and the clean Pt(lll), are shown for comparison. The top most spectrum is related to the chemisorbed (6x5) C6o /Au(llO) that will be discussed in the next section. The C60 deposited on Pt(lll) at room temperature weakly interacts with the substrate atoms and molecular features measured in IPES spectra [2] are similar in intensity and energy separation to the isolated molecule or to C60 multilayer. After annealing the fullerene bond with pte III) substrate become strongly covalent. The deposition of one monolayer of C60 on the Pt(lll) surface at T < 300°C leads to an ordered double domain hexagonal reconstruction, that is present also for coverages exceeding I ML [2]. X-ray diffraction presents the advantage of

216

being able to observe the buried interface even when it is covered by a thick layer of adsorbed molecules giving the opportunity of determining the substrate structure during the different growth phases. The clean Pt(111) surface does not show any reconstruction, however an analysis of its CTRs shows that the surface top layer is slightly relaxed with respect to the nominal bulk termination and their fitting allows to determine an expansion of 4.5 ± 0.5 pm (+2.0%) with respect to the ideal bulk termination, slightly larger than the value of 2.5±0.l pm determined by quantitative LEED analysis [31]. About 10 MLs of C60 were deposited at room temperature. Then the system was slowly annealed while monitoring with the x-rays the appearance of the ..J13 x ..J13R13.9° reconstruction. At about 420 K sharps peaks corresponding to this reconstruction appear. This temperature is much lower than the multilayer desorption of the fullerenes and, in fact, x-ray reflectivity scans show the presence of a well ordered C60 multilayer. In Fig.2 we show an azimuthal plot of the intensity of the fist order reconstruction. The two P6 domains separated of 27.8° with a practical equal intensity are clearly observable. 20000 r - - - - , - - - - - - - - , - - - - - - - , - - - ,

15000

10000

, I' "

5000

:~ J "

0 50

100

150

200

Figure 2. Azimuthal scan of the x-ray scattered intensity for an exchanged vector modulus x ..J13R13 .9° reconstruction. The peaks belong to the corresponding to the first order of the two hexagonal domains separated of27.8°.

..J13

217

Figure 3. In (a) we show the Patterson map of the .J}3 x.J}3R13.9° reconstruction. Because the pattern is dominated by the signal due to a vacancy in the surface unit cell this map has an inverted sign usual. The negative peaks (dashed lines) correspond to the positions of atoms participating in the reconstruction. The dashed rhombus is the Pt(lll) surface unit cell. In (b) we show a top view of the best model fit and in (c) a detail on the C atoms forming bonds with the substrate.

Further annealing at 650 K removes the multilayer and leave only one bonded to the surface. The analysis of the CTRs provides the following infonnation: a) the vertical expansion of the topmost substrate layer reduces to 1.1 +1- 0.4 pm, b) the increase in surface roughness is compatible with the presence of one vacancy per reconstruction unit cell in the substrate surface. From the x-ray diffraction analysis this vacancy helps enonnously the structural solution of the system. This situation is very similar to the well known case of the presence one heavy atom per unit cell [32]. In this case if the origin of the unit cell is taken at the position of the heavy atom all the reflection phases are very close to zero and the Patterson map is very close to the real electron density map. This same approach has been used for solving the structure of gold deposited on a vicinal Si surface [33]. In our case the vacancy can be considered as a heavy atom with a negative number of electrons and the Patterson must then be interpreted by looking at the negative peaks (Fig.3). The dashed rhombus in the figure shows the surface unit cell of the Pt(lll) substrate while the large continuous line rhombus is the surface unit cell of the .Jl3 x .Jl3R13.9° reconstruction. The dashed contour lines the l1f'·""tlVf' values of the map and the deepest minima are localized around the vacancies and at the Pt surface atom positions. A fitting of the data, assuming only 13 structural parameters, leads to a very simple solution of this complicated system. The surface Pt atoms displace slightly from their bulk tenninated positions while the C60 lies on top of the vacancies pointing to the substrate surface with one of its hexagonal surfaces. The fullerene enters as much as possible in the vacancy trying to maximize the number of bonds which are

218

formed with the Pt atoms and about 12 Carbon atoms form directs bonds with the Pt substrate [34]. Similar results have been also observed in the case of the C60 adsorbed onto the Au(lll) surface. Also in this case the dominant effect is the formation of vacancies at the substrate surface which stabilize the position of the fullerene molecules. The C60 lies with one of its hexagonal faces on top of the vacancies at a height slightly larger than in the case of the Pt substrate. IPES spectra of the hexagonal phases of a single monolayer are substantially different from the C6o-Au(110) where was estimated by electron spectroscopies and an ionic character dominates interaction a charge transfer of 3+1 [see for discussion ref. 2 and references therein] electron. When the surface is annealed a 2-domain structure is formed rather dramatic changes in the spectrum IPES spectrum occur: the LUMO-derived feature appears centered at 0.5 eV and the spectral weight near EF is greatly

Energy above EF (eV) Figure 4. Normal incidence Inverse Photoemission for different C60 overiayers on Pt(lll). The spectra ofC,o Multilayer (bottom) (shifted 0.12 eV) and (6x5)C60IAu(llO) (top) are plotted for comparison.

219

increased, indicating that there is a redistribution of the empty states. This fact has been interpreted as a strong interaction between the substrate and the C60 overlayer when the covalent bond takes place. This result nicely confirms the formation of 12 Pt-C bonds, suggested by the SXRD analysis. The ordered layer formation is a kinetically controlled process: The double hexagonal pattern was observed previously for 1 ML adsorbed at 100 K and annealed above 770 K [35]. When the sample is heated above 800 K or if less than a complete monolayer is adsorbed at T?:670 K, a third domain becomes visible in the LEED pattern. For the two-domain structure on Pt(III), vibrational studies have indicated no evidence for charge transfer. Actually in case of Pt substrate a further annealing of the system leads to a 3-domain structure and to fragmentation of the fullerene. The corresponding LEED pattern shows graphitic rings. Higher T and longer annealing produce changes in the IPES spectra. At this stage it has been suggested that the fragmentation products consist in pentagon-hexagon groups, still arranged in a pseudo-order related to graphitic-like features in diffraction, as discussed in ref. 2.

4.2 C60IAu(1JO)

The (110) surfaces of metals are characterized by the strong surface diffusion anisotropy energies. Typically this surface is quite unstable in all metals and shows reconstructions ranging from the lx2 to the lx5 unit cell. Clean Au(110) presents the lx2 reconstruction. Upon deposition of 1 ML of C60 it is possible to observe by STM a nice hexagonal pattern of the fullerene layer showing alternating brighter and darker lines. The uncovered regions of Au modify their reconstruction from 1x2 to a 1x5 surface cell. This 1x5 reconstruction is in perfect register with the brighter and darker lines in the STM images leading the authors to conclude that this kind of reconstruction was extending under the fullerene layer [30 and refs therein]. On clean 2xl-Au(110) deposition of few monolayers of C60 followed by thermal annealing leads to the formation of large terraces showing a (6x5) reconstruction. The evolution of the IPES spectra of C60 deposited on Au(11 0) for different molecular depositions at room temperature (RT) [36] shows, for coverages :s 1 ML, that all the molecular structures are broader and their energy separation is different with respect to the multilayer spectrum. The first structure (LUMO) is practically not detectable and a clearly Fermi level emission is present. As soon as we reach the second layer, the LUMO feature appears clearly, though the

220 peaks are still broader than in the multilayer system. For 1 ML deposition at about 450°C we observed, in the LEED pattern, a (6x5) superstructure. The energy position of the features in the IPES spectra are the same as in the IML C60 grown at R T, while their intensities are much more pronounced. The LUMO is almost absent and the Fermi level emission is still present. By comparing our 1 ML IPES data with the ordered system C6o /GeS(001) [17], where C60 is physisorbed on the surface, it results that the configuration of the empty states is considerably different. Fig. 5 (top) most spectrums show the 6x5 C601Au(l10) IPES spectrum. During the formation of the first monolayer, when the system is annealed (either during C60 deposition or after) there is a strong deformation of all the molecular C60 localized states due to a charge transfer from the substrate to the C60 indicating a chemisorbed phase. Comparison with NEXAFS results will be discussed. SXRD data confirmed a strong redistribution of the substrate Au atoms induced by the formation of the 6x5 reconstruction after thermal annealing. When analyzed with x-ray diffraction the structure of the interface appeared to be much more complicated. The substrate reconstruction was not a "simple" lx5 but a complex 6x5 reconstruction with no obvious symmetries. This makes the data analysis an almost impossible task. For solving this system we have applied direct methods which are able to provide hints on the electronic density maps of the surface unit cell [30].

221

from the model proposed by the direct methods, we have been able to refine the analysis and obtain a full 3D solution of the system. The peculiarity of this solution is the proof that fullerene adsorption induces a strong mass transport involving several layers of the substrate. The gold surface

-4 -3 -2 -1

0

0123456 [eV]

Figure 5. Top Comparison between experimental and calculated in-plane data corresponding to the CoolAu(1IO)-p(6x5) surface reconstruction. The measured values an their associated uncertainties are proportional to the radii of the two empty semicircles. The filled semicircles are proportional to the calculated values using the final structure Right: Lateral views of the CooIAu(IIO)-p(6x5) surface final structure. The observed corrugation has a height of one atomic level

222

modifies itself in order to form a kind of calyx structure where the fullerene molecules fit. Using this proposed structure we have been able to calculate the electronic properties of the interface which can then be compared with the photoemission data. (Fig.5 ) The larger contact area between the C60 and the metal, resulting from the substrate rearrangement, allows for the formation of strong directional C-Au bonds. 5. Conclusions We have shown that the combination ofIPES and SXRD techniques are able to highlight all the details of the adsorption of fullerenes on noble metal surface elucidating the complex structure which is formed at the interface. Even though the fullerenes are considered quite inert molecules they are able to case massive atoms displacement in the metals surfaces in order to maximize the number of bondings which can be formed. References See for example P. Rudolf, in: H. Kuzmany, J. Fink, M. Mehring, S. Roth, (Eds.), Fullerenes and Fullerene Nanostructures, World, Scientific, Singapore, 1996, pp. 263-275; Rosei, F. et al. Properties of large organic molecules on metal surfaces. Prog. Surf Sci. 71,95-146 (2003). 2. M. Pedio, K. Hevesi, N. Zema, M. Capozi, P. Perfetti, R. Gouttebaron, J.J. Pireaux, R. Caudano, P. Rudolf, Surf Sci. 437,249-260 (1999) 3. N. V. Smith Rep. Prog. Phys. 51, 1227 (1988) and references therein; V. Dose Surf. Sci. Rep. 5, 337 (1985); F. J. Himpsel, Surf Sci. Rep. 12, I (1990). 4. S Krause, M B Casu, A SchOll and E Umbach New J. Phys. 10 085001 (2008). 5. G.K. Wertheim, D.N.E. Buchanan, E.E. Chaban and lE. Rowe, Solid State Comm. 83, 785 (1992) 6. A. Goldoni, C. Cepek, S. Modesti, Synthetic Metals 77,189 (1996) 7. l StOhr, "NEXAFS Spectroscopy", Springer Series in Surface Sciences Vol. 25 Springer-Verlag, Berlin, 1992 8. M. Pedio, B. P .. Doyle, N. Mahne, A. Giglia, F. Borgatti, S. Nannarone, S K Henze, R. Temirov, S. Tautz , L. Casalis, R. Hudej, M F Danisman, B Nickel 2007 Appl. Surf Sci. 254, 103 (2007) 9. M. B. Jost, N. Troullier, D. M. Poirier, J. L. Martins, J. H. Weaver, L. P. F. Chibante, R. E. Smalley, Phys. Rev. B 441966 (1991). 10. M. Nyberg, Y. Luo, L. Triguero, L. G. M. Pettersson, H. Agren, Phys. Rev. B 60,7956 (1999) 1.

223 11. B. Wastberg et al. Phys. Rev. B 50, 13031 (1994); A. J. Maxwell, P.A. Brtihwiler, A. Nilsson, N. Martensson, P.Rudolf, Phys. Rev. B 49, 10717 (1994) 12. A Goldoni, C. Cepek, R. Larciprete, L. Sangaletti, S. Pagliata, L. Floreano, R Gotter, A. Verdini, A. Morgante, Y. Luo, N. Nyberg 1. of Chern Phys 116, 7685 (2002) 13. J.Schnadt, J.Schiessling P.A.Brtihwiler Chern. Phys. 312,39 (2005)39 14. M. B. Jost, N. Troullier, D. M. Poirier, J. L. Martins, J. H. Weaver, L. P. F. Chibante, R E. Smalley, Phys. Rev. B 44 1966 (1991) 15. S. Satpathys, V. P. Antropov, O. K. Andersen, O. Jepsen, O. Gunnarsson, A. I. Lichtenstein, Phys. Rev. B 46, 1773 (1992) 16. M.B. Jost, P. J. Benning" D.M. Poirier, J.H. Weaver, L.P.F. Chibante, RE.Smalley, Chemical Physics Letters 184 423 (1991); PJ. Benning, D. M. Poirier, T. R Ohno, Y. Chen, M. B. Jost, F. Stepniak, G. H. Kroll, J. H. Weaver, J. Fure, R. E. Smiley, Phys Rev B 456899 (1992) 17. J.-M. Themlin, S. Bouzidi, F. Coletti, J.-M. Debever, G. Gesterblum, L.-M. Yu, J.-J. Pireaux, P. A. Thiry, Phys. Rev. B 46,15602 (1992) 18. K.-D. Tsuei, J.-Y. Yuh, C.-T. Tzeng, R.-Yu Chu, S.-c. Chung, K.-L. Tsang, Phys. Rev. B 56,15412 (1997) 19. R. Schwedhelm, L. Kipp, A. Dallmeyer, and M. Skibowski, Phys. Rev. B 58, 13176 (1998) 20. B.L. Henke, et aI., Atomic Data and Nuclear Data Tables 54 no.2, 181-342 (July 1993) 21. M. Born and E.Wolf, Principles of Optics, Cambridge University Press, Cambridge, UK, 1999 22. R. Jacobsson, in Progress in Optics, edited by E. Wolf, North-Holland, Amsterdam, 1966 23. L. Nevot and P. Croce, Phys. Appl. 15,761 (1980) 24. J Penfold and R K Thomas, 1. Phys.: Condens. Matter 2,1369-1412 (1990) 25. I K Robinson, Phys. Rev. B 333830-3836 (1986) 26. R Feidenhans'l, Surface Science Reports, 10 105-188 (1989) 27. H Stragier et ai, Phys. Rev. Lett. 69,3064 (1992) 28. M G Proietti et ai, Phys. Rev. B 59,5479-5492 (1999) 29. M Benfatto and R Felici, Phys. Rev. B 64115410 (2001) 30. M Pedio, R Felici, X Torrelles, P Rudolf, M Capozi, J Rius, and S Ferrer, Phys. Rev. Lett. 85, 1040 (2000) 31. N Materer, et al. Surf Sci. 325, 207-222 (1995) 32. H B Dyer, Acta Crystal/., 4 42 (1951) 33. I K Robinson et aI., Phys. Rev. Lett. 88096104 (2002) 34. R Felici, M Pedio, F Borgatti, S Iannotta, M Capozi, G Ciullo, A Stierle , Nat. Mat. 4 688-692 (2005) 35. C Cepek, A Goldoni, and S Modesti, Phys. Rev. B 53 7466 (1996) 36. M. Pedio, M L Grilli, C Ottaviani, M Capozi, C Quaresima, P Perfetti, P A Thiry, R Caudano, P Rudolf,. EI. Spec. and ReI. Phen. 76405 (1995)