Electrokinetic Chromatography
Electrokinetic Chromatography: Theory, Instrumentation and Applications Edited by U. Pyell # 2006 John Wiley & Sons, Ltd. ISBN: 0-470-87102-4
Electrokinetic Chromatography Theory, Instrumentation and Applications Edited by UTE PYELL University of Marburg, Germany
Copyright # 2006
John Wiley & Sons Ltd, The Atrium, Southern Gate, Chichester, West Sussex PO19 8SQ, England Telephone (+44) 1243 779777
Email (for orders and customer service enquiries):
[email protected] Visit our Home Page on www.wileyeurope.com or www.wiley.com All Rights Reserved. No part of this publication may be reproduced, stored in a retrieval system or transmitted in any form or by any means, electronic, mechanical, photocopying, recording, scanning or otherwise, except under the terms of the Copyright, Designs and Patents Act 1988 or under the terms of a licence issued by the Copyright Licensing Agency Ltd, 90 Tottenham Court Road, London W1T 4LP, UK, without the permission in writing of the Publisher. Requests to the Publisher should be addressed to the Permissions Department, John Wiley & Sons Ltd, The Atrium, Southern Gate, Chichester, West Sussex PO19 8SQ, England, or emailed to
[email protected], or faxed to (+44) 1243 770620. Designations used by companies to distinguish their products are often claimed as trademarks. All brand names and product names used in this book are trade names, service marks, trademarks or registered trademarks of their respective owners. The Publisher is not associated with any product or vendor mentioned in this book. This publication is designed to provide accurate and authoritative information in regard to the subject matter covered. It is sold on the understanding that the Publisher is not engaged in rendering professional services. If professional advice or other expert assistance is required, the services of a competent professional should be sought. Other Wiley Editorial Offices John Wiley & Sons Inc., 111 River Street, Hoboken, NJ 07030, USA Jossey-Bass, 989 Market Street, San Francisco, CA 94103-1741, USA Wiley-VCH Verlag GmbH, Boschstr. 12, D-69469 Weinheim, Germany John Wiley & Sons Australia Ltd, 42 McDougall Street, Milton, Queensland 4064, Australia John Wiley & Sons (Asia) Pte Ltd, 2 Clementi Loop # 02-01, Jin Xing Distripark, Singapore 129809 John Wiley & Sons Canada Ltd, 22 Worcester Road, Etobicoke, Ontario, Canada M9W 1L1 Wiley also publishes its books in a variety of electronic formats. Some content that appears in print may not be available in electronic books. Library of Congress Cataloging-in-Publication Data Electrokinetic chromatography: theory, instrumentation, and applications/[edited by] Ute Pyell. p. cm. Includes bibliographical references and index. ISBN-13: 978-0-470-87102-7 (cloth : acid-free paper) ISBN-10: 0-470-87102-4 (cloth : acid-free paper) 1. Electrokinetics. 2. Chromatographic analysis. I. Pyell, Ute. QC625.E425 2006 5430 .8–dc22 2005031939 British Library Cataloguing in Publication Data A catalogue record for this book is available from the British Library ISBN-13 978-0-470-87102-7 (HB) ISBN-10 0-470-87102-4 (HB) Typeset in 10/12 pt Times by Thomson Press (India) Limited, New Delhi, India Printed and bound in Great Britain by Antony Rowe Ltd, Chippenham, Wiltshire This book is printed on acid-free paper responsibly manufactured from sustainable forestry in which at least two trees are planted for each one used for paper production.
Contents
List of Contributors
vii
Preface
xi
PART I: SEPARATION PRINCIPLES
1
1
Theory of Electrokinetic Chromatography Ute Pyell
3
2
Determination of Critical Micelle Concentrations by Capillary Electrokinetic Techniques Thomas Le Saux, Anne Varenne and Pierre Gareil
33
Selectivity Characterization of Pseudostationary Phases Using the Solvation Parameter Model Colin F. Poole
55
General Aspects of Resolution Optimization with Micellar Pseudostationary Phases Li Jia and Shigeru Terabe
79
3
4
5
Optimization of the Separation Conditions in Electrokinetic Chromatography: Experimental Designs, Modelling and Validation Olga Jime´nez and Maria Luisa Marina
95
6
Microemulsion Electrokinetic Chromatography Alex Marsh, Kevin Altria and Brian Clark
115
7
Polymeric Pseudostationary Phases and Dendrimers Christopher P. Palmer
137
8
Pseudostationary Ion-exchange Phases Philip J. Zakaria and Paul R. Haddad
153
9
Principles of Enantiomer Separations in Electrokinetic Chromatography Bezhan Chankvetadze
179
vi
CONTENTS
10 On-line Sample Enrichment in Electrokinetic Chromatography Joselito P. Quirino
207
PART II: INSTRUMENTATION
233
11 General Aspects of Instrumentation Jan Fischer and Pavel Jandera
235
12 Laser-induced Fluorescence Detection: A Summary Christophe Bayle, Ve´ re´ na Poinsot, Clara Fournier-Noe¨ l and Franc¸ ois Couderc
263
13 Amperometric Detection Frank-Michael Matysik
281
14 Photothermal Detection Werner Faubel, Stefan Heissler and Ute Pyell
289
15 Coupling of Electrokinetic Chromatography to Mass Spectrometry Roelof Mol, Gerhardus J. de Jong and Govert W. Somsen
307
16 Electrokinetic Chromatography on Microfluidic Devices ´ mar Gu´ stafsson and Jo¨ rg P. Kutter O
337
PART III: APPLICATIONS
351
17 Electromigration Separation Techniques in Pharmaceutical Analysis Jean-Luc Veuthey, Laurent Geiser and Serge Rudaz
353
18 Analysis of Body Fluids by Electrokinetic Chromatographic Techniques Ll. Puignou, R. Busquets and M.T. Galceran
373
19 Application of Electrokinetic Chromatography to Food and Beverages Mary Boyce
423
20 Application of Enantioselective Electrokinetic Chromatography Bezhan Chankvetadze
459
21 Environmental Analysis Ana Carolina O. Costa, Gustavo A. Micke, Elisabete A. Pereira, Clo´ vis L. Silva and Marina F.M. Tavares
475
Index
529
List of Contributors
Kevin Altria, GlaxoSmithKline R&D, Building H89, New Frontiers Science Park South, Harlow, Essex, CM19 5AW, UK Christophe Bayle, ERT 1046, Laboratoire des IMRCP, Universite´ Paul Sabatier, 118 Route de Narbonne, 31062 Toulouse cedex, France Mary Boyce, School of Natural Sciences, Edith Cowan University, Joondalup 6027, Western Australia R. Busquets, Department of Analytical Chemistry, Faculty of Chemistry, University of Barcelona, Martı´ i Franque`s 1 – 11, 08028 Barcelona, Spain Bezhan Chankvetadze, Molecular Recognition and Separation Science Laboratory, School of Chemistry, Tbilisi State University, Chavchavadze Ave 1, 380028 Tbilisi, Georgia Brian Clark, School of Pharmacy, Bradford University, Bradford, UK Ana Carolina O. Costa, Institute of Chemistry, University of Sao Paulo, Av. Prof. Lineu Prestes 748, 05508-900 Sao Paulo, Brazil Franc¸ois Couderc, ERT 1046, Laboratoire des IMRCP, Universite´ Paul Sabatier, 118 Route de Narbonne, 31062 Toulouse cedex, France Werner Faubel, Research Centre Karlsruhe, Institute of Technical Chemistry, Water Technology and Geotechnology Division, Postfach 3640, 76021 Karlsruhe, Germany Jan Fischer, University of Pardubice, Faculty of Chemical Technology, Department of Analytical Chemistry, Nam. Cs. Legii 565, 532 10 Pardubice, Czech Republic Clara Fournier-Noe¨l, ERT 1046, Laboratoire des IMRCP, Universite´ Paul Sabatier, 118 Route de Narbonne, 31062 Toulouse cedex, France M.T. Galceran, Department of Analytical Chemistry, Faculty of Chemistry, University of Barcelona, Martı´ i Franque`s l1 – 11, 08028 Barcelona, Spain
viii
LIST OF CONTRIBUTORS
Pierre Gareil, Laboratory of Electrochemistry and Analytical Chemistry, Ecole Supe´ rieure de Chemie de Paris, UMR CNRS 7575, 11 rue Pierre et Marie Curie, 75231 Paris cedex 05, France Laurent Geiser, Laboratory of Pharmaceutical Analytical Chemistry, School of Pharmacy, University of Geneva, 20 Bd d’Yvoy, 1211 Geneva 4, Switzerland ´ mar Gu´ stafsson, MIC, Department of Micro and Nanotechnology, Technical UniO versity of Denmark, 2800 Lyngby, Denmark Paul R. Haddad, Australian Centre for Research on Separation Science, School of Chemistry, University of Tasmania, Private Bag 75, Hobart 7001, Australia Stefan Heissler, Research Centre Karlsruhe, Institute of Technical Chemistry, Water Technology and Geotechnology Division, Postfach 3640, 76021 Karlsruhe, Germany Pavel Jandera, University of Pardubice, Faculty of Chemical Technology, Department of Analytical Chemistry, Nam. Cs. Legii 565, 532 10 Pardubice, Czech Republic Li Jia, Graduate School of Material Science, University of Hyogo, Kamigori, Hyogo, 678-1297 Japan Olga Jime´ nez, Department of Analytical Chemistry, Faculty of Chemistry, University of Alcala´ , 28871 Alcala´ de Henares, Madrid, Spain Gerhardus J. de Jong, Department of Biomedical Analysis, Utrecht University, P O Box 80082, 3508 TB Utrecht, The Netherlands Jo¨ rg P. Kutter, MIC, Department of Micro and Nanotechnology, Technical University of Denmark, 2800 Lyngby, Denmark Maria Luisa Marina, Department of Analytical Chemistry, Faculty of Chemistry, University of Alcala´ , 28871 Alcala´ de Henares, Madrid, Spain Alex Marsh, GlaxoSmithKline R&D, Building H89, New Frontiers Science Park South, Harlow, Essex, CM19 5AW, UK Frank-Michael Matysik, University of Leipzig, Institute for Analytical Chemistry, Linne´ strasse 3, 04103 Leipzig, Germany Gustavo A. Micke, Institute of Chemistry, University of Sao Paulo, Av. Prof. Lineu Prestes 748, 05508-900 Sao Paulo, Brazil Roelof Mol, Department of Biomedical Analysis, Utrecht University, P O Box 80082, 3508 TB Utrecht, The Netherlands Christopher P. Palmer, Department of Chemistry, University of Montana, Missoula, MT 59812, USA
LIST OF CONTRIBUTORS
ix
Elisabete A. Pereira, Institute of Chemistry, University of Sao Paulo, Av. Prof. Lineu Prestes 748, 05508-900 Sao Paulo, Brazil Ve´ re´ na Poinsot, ERT 1046, Laboratoire des IMRCP, Universite´ Paul Sabatier, 118 Route de Narbonne, 31062 Toulose cedex, France Colin F. Poole, Department of Chemistry, Wayne State University, Detroit, MI 48202, USA Ll. Puignou, Department of Analytical Chemistry, Faculty of Chemistry, University of Barcelona, Martı´ i Franque`s 1 – 11, 08028 Barcelona, Spain Ute Pyell, University of Marburg, Department of Chemistry, Hans-Meerwein-Strasse, D35032 Marburg, Germany Joselito P. Quirino, Biologies Analytical Chemistry Department, Scios Inc., 6500 Paseo Padre Parkway, Fremont, CA 94555, USA Serge Rudaz, Laboratory of Pharmaceutical Analytical Chemistry, School of Pharmacy, University of Geneva, 20 Bd d’Yvoy, 1211 Geneva 4, Switzerland Thomas Le Saux, Laboratory of Electrochemistry and Analytical Chemistry, Ecole Supe´ rieure de Chemie de Paris, UMR CNRS 7575, 11 rue Pierre et Marie Curie, 75231 Paris cedex 05, France Clo´ vis L. Silva, Institute of Chemistry, University of Sao Paulo, Av. Prof. Lineu Prestes 748, 05508-900 Sao Paulo, Brazil Govert W. Somsen, Department of Biomedical Analysis, Utrecht University, P O Box 80082, 3508 TB Utrecht, The Netherlands Marina F.M. Tavares, Institute of Chemistry, University of Sao Paulo, Av. Prof. Lineu Prestes 748, 05508-900 Sao Paulo, Brazil Shigeru Terabe, Graduate School of Material Science, University of Hyogo, Kamigori, Hyogo, 678-1297 Japan Anne Varenne, Laboratory of Electrochemistry and Analytical Chemistry, Ecole Supe´ rieure de Chemie de Paris, UMR CNRS 7575, 11 rue Pierre et Marie Curie, 75231 Paris cedex 05, France Jean-Luc Veuthey, Laboratory of Pharmaceutical Analytical Chemistry, School of Pharmacy, University of Geneva, 20 Bd d’Yvoy, 1211 Geneva 4, Switzerland Philip J. Zakaria, Australian Centre for Research on Separation Science, School of Chemistry, University of Tasmania, Private Bag 75, Hobart 7001, Australia
Preface
In their pioneering papers in 1984 and 1985 Terabe et al. [1,2] discovered that separations based on electrokinetic phenomena (electrophoresis and electroosmosis) can be described as chromatographic processes provided that an additive (also called separation carrier or pseudostationary phase) having a velocity different from that of the analytes is present in the separation electrolyte, and this is able to interact with the analytes to be separated. In the first papers on this topic this additive was a micelle-forming ionic surfactant at a concentration above the critical micelle concentration. However, it was also realized very early by Terabe [3] that the presence of micelles is not a prerequisite of electrokinetic chromatography (EKC). Since the presentation of the general separation carrier concept many variants of EKC have been developed that employ ‘polymeric micelles’, microdroplets, other types of colloidal phases, dissolved linear polymers and dendrimers or oligomeric units as separation carriers. In the last two decades EKC has been mainly regarded as a special form of capillary electrophoresis (CE), which has matured into a powerful analytical separation technique that brings speed, reproducibility and automation to the labour intensive methods of classical electrophoresis. However, the IUPAC recommendations concerning the terminology for analytical capillary electromigration techniques published in 2004 [4] clearly regard EKC and CE (also known as capillary zone electrophoresis, CZE) as two equal members of the large family of capillary electromigration techniques. Of course, regarding the unique position of EKC as an interface between electrophoresis and chromatography there inherently remain fuzzy borders between EKC and CE. This book is designed to be a guide to the large, rapidly growing and diverse field of EKC for a broad audience: those new to EKC, those more experienced, those interested in method development including instrumental developments, and those involved with applications research in various fields. The book aims to bring together a thorough theoretical description of methodological aspects, an overview of the current status of the various forms of EKC, and current and emerging applications, as well as looking forward to future developments. This book should be of interest for all those who need a highefficiency separation technique with easily adaptable selectivity and having the additional features of low sample volume requirements, short run times and high versatility. The task of compiling this book required the cooperation of internationally recognized experts with special competence in their respective fields. The volume is composed of 21 chapters organized into three major parts: I Separation Principles, II Instrumentation, and III Applications. Part I includes an introduction to the terminology used to describe the separation process in EKC (Chapter 1), a review of electrokinetic methods to investigate
xii
PREFACE
the micelle-formation process (Chapter 2), an introduction to the fundamentals of the solvation parameter model for the (selectivity) characterization of separation carriers (Chapter 3), a general guide to method development and resolution optimization with micellar pseudostationary phases (Chapter 4), an introduction to concepts for computerbased rapid method optimization (Chapter 5), and reviews of various forms of EKC including microemulsion electrokinetic chromatography (MEEKC), EKC with polymeric pseudostationary phases and dendrimers, EKC with pseudostationary ion-exchange phases and enantioselective EKC (Chapters 6–9). The part is completed by a detailed discussion on techniques employed for on-line sample enrichment in combination with EKC (Chapter 10). In the Part II general aspects of instrumentation are treated including the use of coated capillaries (Chapter 11). This part also presents reviews of different detection methods (laser-induced fluorescence detection, amperometric detection, photothermal detection and mass spectrometric detection) having high potential as sensitive and/or selective detection techniques for EKC (Chapters 12–15). The part is completed by a review of the implementation of EKC on microfluidic devices (Chapter 16). The final part is concerned with applications of EKC in the fields of pharmaceutical analysis, the analysis of body fluids, food analysis, chiral analysis and environmental analysis (Chapters 17–21). With this structure the book intends to illuminate many facets of this relatively young member in the family of separation methods, bringing together expert knowledge from various directions that will not only help the novice to estimate how EKC might help in solving analytical tasks, but will also assist the more experienced user in broadening and deepening their knowledge of this technique, which has already found its way into routine laboratories. I would like to thank all of the contributors and further scientists for their support and cooperation. My personal hope is that bringing together a comprehensive review of the state of the art of this fascinating technique will become a factor in its further dissemination in various fields of application. Ute Pyell
References [1] S. Terabe, K. Otsuka, K. Ichikawa and A. Tsuchiya. Electrokinetic separations with micellar solutions and open-tubular capillaries, Anal. Chem., 56, 111–113 (1984). [2] S. Terabe, K. Otsuka and T. Ando. Electrokinetic chromatography with micellar solution and open-tubular capillary, Anal. Chem., 57, 834–841 (1985). [3] S. Terabe. Electrokinetic chromatography: an interface between electrophoresis and chromatography, Trends Anal. Chem., 8, 129–134 (1989). ˚ . Jo¨ nsson and R.M. Smith. Terminology for analytical capillary electro[4] M.L. Riekkola, J.A migration techniques, Pure Appl. Chem., 76, 443–451 (2004).
Part I Separation Principles
Electrokinetic Chromatography: Theory, Instrumentation and Applications Edited by U. Pyell # 2006 John Wiley & Sons, Ltd. ISBN: 0-470-87102-4
1 Theory of Electrokinetic Chromatography Ute Pyell
1.1 Introduction Electrokinetic chromatography (EKC) is a term that was coined by Terabe and coworkers in 1985 [1,2]. EKC belongs to a family of electromigration separation techniques that employ electrokinetic phenomena (electrophoresis and electroosmosis) for the separation of constituents in a sample. EKC invariably also involves chemical equilibria, e.g. distribution, ion exchange and/or complex formation. According to Terabe and coworkers [1,2] EKC is defined as a capillary electromigration separation technique employing a separation carrier. The separation carrier, also called pseudostationary phase, is a unity (e.g. a microdroplet, a micelle, a dendrimer, or a dissolved polymer) that interacts with the solutes to be separated while its migration velocity is, in general, virtually unaffected by this interaction. The property of the migration velocity of the separation carrier being virtually unaffected by the interaction with dissolved solutes will be taken here to define the difference between a pseudostationary phase and a simple complex-forming agent, which is used in capillary electrophoresis to modify the effective electrophoretic mobility of the solutes to be separated. If the solutes to be separated do not posess an effective electrophoretic mobility without the presence of the separation carrier, the separation carrier must have an electrophoretic mobility. According to the IUPAC recommendations [3], EKC is a separation technique ‘based on a combination of electrophoresis and interactions of the analytes with additives (e.g. surfactants), which form a dispersed phase moving at a different velocity [...than the analytes (editorial note)]. In order to achieve separation either the analytes or this
Electrokinetic Chromatography: Theory, Instrumentation and Applications Edited by U. Pyell # 2006 John Wiley & Sons, Ltd. ISBN: 0-470-87102-4
4
ELECTROKINETIC CHROMATOGRAPHY
secondary phase should be charged’. Micellar EKC (MEKC) is ‘a special case of EKC, in which the secondary phase is a micellar dispersed phase in the capillary’ and microemulsion EKC (MEEKC) ‘is a special case of EKC, where a microemulsion is employed as the dispersed phase.’ In chromatography the observed velocity of a solute zone is the weighted mean of two velocities (velocity of the mobile phase and ‘velocity’ of the stationary phase) resulting from the partitioning of the solute between these two phases. In EKC, as defined above, a noncharged solute will migrate either with the velocity of the electroosmotic flow or with the velocity of the separation carrier. Consequently, the separation of neutral solutes differing in their partitioning coefficients (between the separation carrier and the surrounding phase) is possible and the separation process in EKC can be described in chromatographic terms. In fact, conventional chromatography can be regarded as a special case of EKC, where the observed velocity of the separation carrier is zero. In their first papers on EKC, Terabe and coworkers [1,4] had already emphasized the chromatographic nature of the underlying separation process (re-)defining parameters known from chromatographic theory. Their treatment is the basis of further considerations on rational resolution optimization [5], method development [6] or experimental determination of physicochemical parameters from EKC data [7,8]. One of the peculiarities of EKC is the nonexistence of a stationary phase, hence the solute zone is also transported (in the direction of the detector or in the opposite direction) when incorporated into the pseudostationary phase. Another peculiarity is the possibility of combining electrophoretic and chromatographic phenomena. The instrumentation used in EKC is identical to that employed in capillary electrophoresis (CE) (see Part II ‘Instrumentation’). Separation takes place in a (fused silica) capillary with an inner diameter less than 100 mm and a length mostly varying between 20 and 100 cm. The capillary is filled with the separation electrolyte containing the separation carrier and is immersed at both ends in vessels filled with the same electrolyte. The sample is injected directly into the first segment of the capillary. A high voltage (up to 35 kV) is applied between two electrodes (incorporated into the vessels) producing a very high field strength within the capillary. The migration of analyte zones during the separation process is then caused by electrokinetic effects. In order to avoid instrumental band broadening, detection is mainly done over a short segment of the capillary (e.g. photometric or fluorimetric detection). The resulting trace, detector signal versus time, can be called an electropherogram or chromatogram, which reflects the position of EKC as being between electrophoresis and chromatography. In this chapter chromatographic and electrophoretic terms that are needed to describe and to optimize the separation process will be introduced.
1.2 Electrokinetic Phenomena Two electrokinetic phenomena are of interest in EKC: electrophoresis and electroosmosis. Electrophoresis is the migration of a charged unity (e.g. an ion), surrounded by a medium, due to the presence of an electric field. The charged unity will experience acceleration due to electrostatic forces and friction due to the surrounding medium.
THEORY OF ELECTROKINETIC CHROMATOGRAPHY
5
Under steady-state conditions the two opposite forces balance each other. A final electrophoretic velocity vep is reached which remains constant at constant electric field strength E. vep ¼
eE ¼ mep E 6pZ
ð1:1Þ
where e is the electric permittivity of the surrounding medium, is the electrokinetic potential (zeta potential) at the surface of the charged unity, Z is the viscosity of the surrounding medium, and mep is the electrophoretic mobility of the charged unit in the specific medium. (In a capillary filled with a homogeneous buffer, E is given by voltage U divided by the total length LT of the capillary.) In capillary electrophoresis (CE) solutes can be separated into zones if they differ sufficiently in their effective electrophoretic mobilities meff. In the case of fast equilibria (e.g. protonation equilibria, complexation equilibria) being involved, the efficient electrophoretic mobility is the weighted mean (respecting the degree of dissociation/ protonation or complexation) of the electrophoretic mobilities of all solute species being present in equilibrium. Hence fast equilibria are exploited in CE for resolution optimization, e.g. via the adjustment of the pH (separation of solutes differing in their degree of dissociation/protonation) or the addition of a complex forming agent to the separation buffer (separation of solutes differing in their degree of complexation). The second electrokinetic phenomenon that is important in EKC is electroosmosis. Regarding the typical instrumentation used for EKC, electroosmosis is the bulk flow of liquid inside the capillary at constant velocity due to the effect of the electric field on the counterion layer adjacent to the charged capillary wall. In bare fused silica capillaries the surface is negatively charged under most pH conditions. There will be an excess of positive counterions in the zone forming the boundary layer. This zone of surplus charge adjacent to the capillary wall will be accelerated in the electric field and will also experience friction by the medium next to this layer. In consequence, a steady-state constant velocity, the electroosmotic velocity veo, is reached in the liquid outside the electrical double layer (Helmholtz Smoluchowski equation) [9]. veo ¼
eE ¼ meo E 4pZ
ð1:2Þ
where is the electrokinetic potential (zeta potential) at the surface of the charged wall and meo is the electroosmotic mobility. Equation (1.2) is only valid if the capillary inner diameter is large compared with the thickness of the electric double layer. However, this restriction is fulfilled in practice. It is important to note that veo is independent of the capillary inner diameter and (outside the electric double layer) the velocity of a liquid segment is not a function of the radial position (in contrast to pressure-induced laminar flow). Consequently, neither electrophoresis nor electroosmosis contribute to zone broadening in EKC. The observed (apparent) velocity vs of a solute zone corresponds to the sum of the effective electrophoretic velocity vep of a solute and the electroosmotic velocity veo (vs ¼ vep þ veo ).
6
ELECTROKINETIC CHROMATOGRAPHY
1.3 The Separation Carrier It is the application of a separation carrier that transforms capillary electrophoresis (electrokinetic separation in a homogeneous electric field [10]) into EKC. The term ‘separation carrier’ was coined by Terabe [2,11] generalizing the concept of micellar electrokinetic chromatography (MEKC) [1,3], in which a micellar pseudophase is employed. The term pseudostationary phase, which is more often used in the literature [12], has an identical meaning. The separation carrier is a unit that is added to (or dissolved/dispersed in) the separation electrolyte. In general (see Section 1.1) the separation carrier has an effective electrophoretic mobility and is able to interact with the solutes of interest. In Figure 1.1 the separation mechanism in EKC for a neutral solute and a micelle forming anionic surfactant as separation carrier is depicted. In chromatography, solutes are separated in a system consisting of a stationary and a mobile phase. The velocity vs of a solute zone in the chromatographic bed corresponds to: tmob 1 vmob vmob ¼ tmob þ tstat kþ1
vs ¼
ð1:3Þ
where tmob is the residence time in the mobile phase, tstat is the residence time in the stationary phase, and k is the retention factor (tstat/tmob). In EKC the separation carrier replaces the stationary phase. However, the separation carrier is not immobilized, and hence can have an observed velocity different from zero. Consequently, the observed velocity of a solute zone (neutral solute) is the weighted mean of the velocity of the mobile phase and of the observed velocity of the separation carrier: vs ¼
tmob trsc 1 k vmob þ vsc vmob þ vsc ¼ tmob þ trsc tmob þ trsc kþ1 kþ1
ð1:4Þ
where trsc is the residence time associated with the separation carrier, vsc is the observed velocity of the separation carrier (vsc ¼ vepsc veo), and vepsc is the electrophoretic velocity of the separation carrier.
– – – – – – – – – – – – – – + + + + + + + + + + + + + + -
-
S
-
+
-
-
-
-
-
-
-
S
–
-
-
-
-
+ + + + + + + + + + + + + + – – – – – – – – – – – – – – Figure 1.1 Scheme illustrating the separation mechanism in micellar EKC (anionic surfactant, normal elution mode)
THEORY OF ELECTROKINETIC CHROMATOGRAPHY
7
Table 1.1 Modes in EKC Mode
Separation carrier
Micellar EKC (MEKC)
Micellar pseudophase: anionic, cationic, or nonionic surfactant (or mixture of nonionic surfactant with ionic surfactant) in a concentration above the CMC Microdroplets present in oil-in-water or water-in-oil microemulsions Polymerized micelles or amphiphilic (charged) linear copolymers or cocondensates Charged dendrimers Soluble linear polymers with ion-exchange sites or other moieties with ion-exchange sites (interaction analyte/separation carrier via electrostatic forces) Micellar pseudophase plus dissolved complex ligand or metal complex
Microemulsion EKC (MEEKC) Polymeric EKC Dendrimeric EKC Ion-exchange EKC (IE-EKC)
Secondary-equilibrium modified MEKC, e.g. cyclodextrin-modified MEKC (CD-MEKC) or ligandexchange MEKC (LE-MEKC)
Here the retention factor k is redefined to be the ratio trsc/tmob. Equation (1.4) shows why EKC is considered to be a chromatographic process if the observed velocity vsc of the separation carrier is vitually unaffected by the interaction with the solute. When associated with the separation carrier the solute is transported with the velocity of the free separation carrier. EKC was introduced first in 1984 by Terabe and coworkers [1,2,4], who employed charged micelles formed by anionic or cationic surfactants as separation carriers. In 1991 Watarai [13] introduced microemulsion EKC (MEEKC) employing charged microdroplets as the separation carrier. Later, polymeric micelles, charged dissolved (amphiphilic) polymers and charged dendrimers were used as separation carriers in EKC [14,15]. Table 1.1 gives an overview of developed modes in EKC. Generally, all modes of interaction of the solute with the (pseudo)stationary phase known in chromatography should also be applicable in EKC: hydrophobic interaction, Van-der-Waals interaction, electrostatic interaction, coordinative interaction, etc. In contrast to EKC, a capillary electromigration separation technique employing a true stationary phase is called capillary electrochromatography (CEC) [16]. According to the IUPAC recommendations [3] CEC is ‘a special case of capillary liquid chromatography, where the movement of the mobile phase through a capillary, filled, packed or coated with a stationary phase, is achieved by electroosmotic flow’. EKC was first developed to make it possible to separate noncharged compounds by using an electromigration separation technique without the participation of a stationary phase. Figure 1.2 shows the separation of positional isomers of neutral nitrotoluenes by employing micelles of the anionic surfactant sodium dodecylsulfate (SDS) as the separation carrier. However, EKC also proved to be a versatile tool for mixtures of charged and uncharged compounds as well as for charged compounds with similar effective electrophoretic mobilities. As EKC is invariably performed in a chamber of high electric field strength, electrophoresis will contribute to the separation if the solutes are permanently or partially charged (weak electrolytes). Consequently, measures taken
8
ELECTROKINETIC CHROMATOGRAPHY E
2
7
0.006
5
0.004 4
6
1 0.002 3
8 9
t0
0.000 3
4
5
6
7
8
t/min
Figure 1.2 Separation of neutral nitrotoluenes by micellar EKC. Electrolyte: c(SDS) ¼ 55 mmol L1, c(urea) ¼ 1.8 mol L1; c(Na2B4O7) ¼ 10 mmol L1. Solutes: 1 ¼ 2,4,6,trinitrotoluene; 2 ¼ 2,4-dinitrotoluene; 3 ¼ 2,5-dinitrotoluene; 4 ¼ 2,6-dinitrotoluene; 5 ¼ 3,4-dinitrotoluene; 6 ¼ 2-nitrotoluene; 7 ¼ 2,3-dinitrotoluene; 8 ¼ 4-nitrotoluene; 9 ¼ 3-nitrotoluene; capillary: 75 mm i.d., 50 cm effective length, 56.5 cm total length; voltage: 25 kV; 25 C; pressure injection: 1.5 s; detection: photometric, l ¼ 254 nm. (Reprinted from U. Pyell, U. Bu¨tehorn, J. Chromatogr. A., 716, 81–95 (1995), copyright 1995, with permission from Elsevier)
to modify the effective electrophoretic mobility in CE (e.g. addition of a complex forming agent, variation of pH) can be also applied in EKC to optimize the selectivity of the separation system.
1.4 Separation of Neutral Solutes 1.4.1
Retention Factor
Corresponding to theory in chromatography the retention factor k (older term: capacity factor k0 or ~k0, sometimes referred to as migration factor k0 [17]) in EKC is defined as residence time in the separation carrier devided by residence time in the surrounding liquid phase. If we assume the separation carrier to be a homogeneous (pseudo)phase, the separation process can be understood to be due to distribution between two distinct phases having two different observed mobilities: k ¼ jP ¼
Vsc P Vmob
ð1:5Þ
THEORY OF ELECTROKINETIC CHROMATOGRAPHY
9
(j ¼ phase ratio, Vsc ¼ volume of separation carrier, Vmob ¼ volume of surrounding (mobile) phase, P ¼ partition coefficient). If the molar (or mass) concentration and the partial molar (or specific) volume of the separation carrier are known, the distribution coefficient can be calculated from the retention factor [1]. This makes EKC a valuable tool for the determination of liquid–liquid partition coefficients [18]. Replacing the velocities in Equation (1.4) with the respective distance-over-time and rearranging results in Equation (1.6) [1,4]: ts t0 ð1:6Þ k ¼ t0 ð1 ts =tsc Þ (t0 ¼ migration time of the front of the surrounding (mobile) phase, ts ¼ migration time of the solute zone, tsc ¼ migration time of the front of the separation carrier). It should be noted that micelles (molecular aggregates of surfactants), which are mainly used as the separation carrier, have too small an aggregation number to be regarded as a phase in the usual sense and, on the other hand, also contain too many surfactant molecules to be considered as a chemical species [19]. In contrast to bulk phases, whose properties are invariant with position, the properties of small aggregates are expected to vary with distance from the interface (spatial heterogeneity). However, those solutes that enter the micelle, can diffuse rapidly within the micelle and experience a wide range of microenvironments, so that an averaging effect would prevail. Consequently, micelles are complex solvents and can only be treated in an approximate sense as bulk solvents. In spite of these restrictions, the retention data for a large number of solutes are homogeneous with respect to the construction of solvation parameter models, suggesting a uniform average solvation environment for all solutes [19]. In addition to these considerations, the interaction of the solute with the dispersed separation carrier can also be described by the binding model, where solute–separation carrier ‘binding’ is defined as occurring whenever the solute interacts with the separation carrier unity (e.g. a micelle) [20]. An equilibrium binding constant Kb can be defined for the equilibrium S þ SC Ð [SSC]. On one hand, the binding model is more universal, as it is also possible to describe equilibria where the observed velocity of [SSC] is not identical to the observed velocity of the separation carrier. On the other hand, only oneto-one associates are taken into consideration. In general, the phase model is preferred to the binding model. Equation (1.6) is valid in the case of so-called normal elution mode according to Vindevogel and Sandra [12]. In this elution mode, veo and vsc have identical direction and jveo j > jvsc j. It is important to state that in other elution modes veo and vsc can have opposite directions. Gareil [21] has shown that in the case that the observed velocity of the solute zone being opposite to veo (reversed direction mode according to Vindevogel and Sandra [12]), k has to be determined using Equation (1.7): ts þ t0 k ¼ ð1:7Þ t0 ðts =tsc 1Þ In that case ts and tsc can be determined simultaneously in one run whereas the determination of teo is only possible after the reversal of polarity or the injection of a marker solution at the opposite end of the capillary. Equation (1.4) has to be rewritten,
10
ELECTROKINETIC CHROMATOGRAPHY
provided that only absolute velocities (v ¼ j~ vj) are given: vs ¼
1 k veo þ vsc kþ1 kþ1
ð1:8Þ
In case of observed velocity of the solute zone being opposite to vsc (restricted elution mode according to Vindevogel and Sandra [12] ) Equation (1.9) and (1.10) are valid [21]. 1 k veo vsc kþ1 kþ1 ts t0 k ¼ t0 ðts =tsc þ 1Þ
vs ¼
ð1:9Þ ð1:10Þ
A scheme of the three elution modes compared here is given in Figure 1.3. Figure 1.4 shows the development of solute zones during the separation process for EKC in the normal elution mode. Generally, the effective electrophoretic mobility of the separation carrier is opposite to the electroosmotic mobility of the mobile phase, because a veo vsc vs (a) Normal elution mode
v eo v sc vs (b) Restricted elution mode
v eo vsc vs (c) Reversed direction mode
Figure 1.3 Elution modes in electrokinetic chromatography
THEORY OF ELECTROKINETIC CHROMATOGRAPHY
11
t=0
Injected zone
t >0
SC marker
Solute 3
Solute 2
Solute 1
EOF marker
Figure 1.4 Development of solute zones during the chromatographic run, normal elution mode, neutral solutes, P(Solute 1) < P(Solute 2) < P(Solute 3) [see Equation (1.5)]
separation carrier of opposite charge to the surface of the capillary wall (e.g. a cationic surfactant or a cationic polymer in a negatively charged fused-silica capillary) will be adsorbed onto the surface of the capillary wall reversing the direction of the electroosmotic flow. There are two special cases: (i) vsc equals zero, (ii) veo equals zero. The first case corresponds to conventional chromatography, the second also corresponds to conventional chromatography if we rename the phases. With EKC it is possible easily to vary the volume ratio j (volume of separation carrier/ volume of surrounding (mobile) phase) by increasing the concentration of the separation carrier in the separation buffer. Assuming that the volume of the mobile phase is not significantly reduced by addition of the separation carrier, we would expect a linear increase in the retention factor with concentration of separation carrier [refer to Equation (1.5)]. This expectation has been confirmed by many authors [22]. We would also expect a retention factor of zero at a separation carrier concentration of zero. In the case of a micellar separation carrier it has to be taken into account that the concentration of the separation carier is zero at or below the critical micelle concentration (CMC). Consequently, in micellar EKC (if there is no interaction of the solute with the capillary wall and with surfactant monomers) by plotting the retention factor versus the surfactant concentration, straight lines are obtained that pass through the identical x-axis intercept, which corresponds to the CMC (see Figure 1.5). 1.4.2
Secondary Complex Equilibria
If a compound that forms complexes with the solutes to be separated is added to the separation electrolyte containing a separation carrier, then the apparent retention factor will be decreased due to the coupled equilibria. This effect is used in cyclodextrin-modified
12
ELECTROKINETIC CHROMATOGRAPHY
Figure 1.5 Dependence of the retention factor k for several neutral solutes on the molar concentration of the anionic surfactant SDS in the separation buffer; solutes: & 2-naphthol, toluene, nitrobenzene, ~ phenol, & resorcinol. (Reprinted with permission from M.G. Khaledi, S.C. Smith, J.K. Strasters, Anal. Chem., 63, 1820–1830 (1991), copyright 1991 American Chemical Society)
micellar electrokinetic chromatography for chiral separations and for the separation of hydrophobic compounds [23]. The partitioning coefficient P [see Equation (1.5)] corresponds to the ratio of the concentration of solute in the separation carrier to the concentration of uncomplexed solute in the surrounding phase. The concentration of the uncomplexed solute in the surrounding phase is dependent on the concentration of the complex forming additive and the complex formation equilibrium constant KC. When assuming a 1:1 complex, the degree of complexation b of the solute in the surrounding phase is given by: b ¼
KC cðAÞ 1 þ ½KC cðAÞ
ð1:11Þ
where c(A) is the molar concentration of the free complex forming additive. b remains constant if c(A) remains constant. In general, c(A) can be approximated by the total concentration of the complex forming additive. b can only take values between 0 and 1. It is evident that the situation will become more complicated if several separation carriers or several complex forming additives are used for selectivity optimization. 1.4.3
Resolution
The resolution of two adjacent zones in the normal elution mode is given by [1]: pffiffiffiffi k N a 1 1 t0 =tsc ð1:12Þ Rs ¼ k þ 1 4 a 1 þ ðt0 =tsc Þ k ðRs ¼ resolution, N ¼ plate number, a ¼ selectivity factor, k ¼ mean retention factor ¼ arithmetic mean of the two retention factors for the two solutes investigatedÞ. Comparing Equation (1.12) with the equation for the resolution of two solute zones in conventional chromatography reveals that the dependence of Rs on the mean retention
THEORY OF ELECTROKINETIC CHROMATOGRAPHY
13
1.0 0.9
t0/tsc = 0
0.8 0.7
f (k)
0.6
0.1
0.5 0.2
0.4 0.3
0.4
0.2
0.6
0.1
0.8
0.0 0
10
20 30 Retention factor
40
50
Figure 1.6 Dependence of f(k) on the retention factor in EKC in the normal elution mode [see Equation (1.13)] for several ratios t0/tsc. The ratio t0/tsc ¼ 0 corresponds to conventional chromatography with an immobilized stationary phase (according to [12])
factor is more complex in EKC and that the time ratio t0/tsc has a major impact on the achievable resolution. In normal elution mode, the time span (window) in which a neutral compound can be eluted is restricted to values between t0 and tsc. Consequently, t0/tsc or its reciprocal value tsc/t0 have mainly been used in the literature to characterize the ratio of the observed velocities of the two ‘phases’ in EKC. One widely accepted term for this time ratio is migration (time) window. Plotting the last two factors of Equation (1.12) [ f ðk), see Equation (1.13)] against k reveals that f ðk) reaches a maximum and that this maximum is smaller than 1 in all instances, with 1 as the limiting value if tsc approaches infinity (see Figure 1.6). f ðkÞ ¼
k 1 t0 =tsc k þ 1 1 þ ðt0 =tsc Þ k
ð1:13Þ
In their pioneering paper on micellar EKC Terabe et al. [1] recognized that the lower resolution obtained in the normal elution mode with identical N, a, and k is a disadvantage of EKC as compared with conventional chromatography that can be, however, compensated for by the large plate numbers achievable under routine conditions in EKC (200 000–300 000). It has to be emphasized that the equation to determine the resolution of two solute zones is dependent on the elution mode [see Equations (1.7) to (1.10)] [21]. In the reversed direction mode, Equation (1.14) is valid. pffiffiffiffi N a 1 1 þ t0 =tsc k ð1:14Þ Rs ¼ k þ 1 4 a ðt0 =tsc Þ k 1
14
ELECTROKINETIC CHROMATOGRAPHY 10
t0/tsc = 0.3
0.5
1
8
f (k)
6
4
2
f (k) = 1 0 0
1
2
3
4
5
6
7
8
9
10
Retention factor
Figure 1.7 Dependence of f(k) on the retention factor in EKC in the restricted elution mode [left side of the graph, see Equation (1.15)] and the reversed migration mode [right side of the graph, see Equation (1.14)] for several ratios t0=tsc. The dashed line indicates f(k) ¼ 1 (according to [21])
In the restricted elution mode, Equation (1.15) is valid. pffiffiffiffi N a1 1 þ t0 =tsc k Rs ¼ k þ 1 4 a 1 ðt0 =tsc Þ k
ð1:15Þ
In these modes there is no restricted time span (window) during which a neutral compound can be eluted and f ðk) can exceed unity. Foley p [5]ffiffiffiffiffiffiffiffiffiffi andffi Gareil [21] have shown that in normal elution mode f ðk) is maximum for k ¼ tsc =t0. However, for the reversed direction mode and for the restricted elution mode f ðk) increases dramatically, when k approaches tsc=t0 (see Figure 1.7) [12,21]. Here, if k ¼ tsc =t0 , then the velocity of the solute zone is zero. Consequently, in these modes very high resolution can be obtained even for very low selectivity, but at the expense of migration time of the solutes to be separated. For example, Bushey and Jorgenson [24] succeeded in separating isotopically substituted compounds (dansylated methylamine and dansylated methyl-d3amine) by micellar EKC with migration times of more than 90 minutes. In 1993 Zhang et al. [25] published a paper describing phenomena in EKC based on conventional chromatography theory. They defined three new parameters. One is the phase velocity ratio Pr , which is identical to tsc=t0. They define a negative time as that where the direction of migration is towards the positive electrode (the anode), and a positive time if the direction of migration is towards the negative electrode (the cathode). The second new parameter is the column availability Aco, which corresponds to the last term in Equation (1.13): Aco ¼
Pr 1 Pr þ k
ð1:16Þ
THEORY OF ELECTROKINETIC CHROMATOGRAPHY
15
The third parameter is the virtual column length L0 , which corresponds to the actual length of the capillary to the detector multiplied by the column availability Aco. In the case of conventional chromatography, Aco ¼ 1 and the solute zone is only transported by the mobile phase to the detector. Zhang et al. [25] showed that in normal elution mode, Aco < 1, whereas in restricted elution mode and in reversed direction mode Aco can exceed unity. 1.4.4
Peak Capacity
According to Giddings [26], the peak capacity n corresponds to the maximum number of components resolvable in one chromatographic run. It is obvious that for the analysis of a complex sample with a high number of constituents a high peak capacity is mandatory in order to avoid comigration of solutes. In deriving an equation approximating the number of components resolvable in one chromatographic run, Giddings assumed a constant plate number independent of the solute and the retention factor: pffiffiffiffi N t2 ln ð1:17Þ n ¼ 1þ t1 4 t1 is the migration (or elution) time of the first solute zone, while t2 is the migration time of the last solute zone. In normal elution mode in EKC, t1 is identical to t0 and t2 is identical to tsc, because all neutral solutes have to be eluted within this time span. Consequently, the migration window has a direct impact on the peak capacity in normal elution mode. In other elution modes of EKC there is no fundamental restriction of migration time. It is, however, important to note that the requirement for Equation (1.17), constant plate numbers independent of the retention factor, is not fulfilled in practice. Some authors have therefore preferred to use the separation number SN instead of the peak capacity [27,28]. The separation number is defined as the number of component peaks that can be placed between the peaks of two consecutive homologous standards (e.g. the homologous series of n-alkyl phenyl ketones) with z and z þ 1 carbon chain atoms, separated by a resolution of 1.177. Also the separation number is dependent on the migration window. The separation number takes into account the varying band broadening with increasing migration time. Kolb et al. [28] have therefore suggested calculating the overall peak capacity from the sum of separation numbers within a given z range. 1.4.5 Determination of the Velocities of the Mobile Phase and the Separation Carrier Generally, for the determination of the electroosmotic velocity, which is identical to the velocity of the mobile phase, a sample that contains a neutral compound is injected (the marker of the velocity of the mobile phase), which is not retained by the separation carrier or the capillary wall and which has an effective electrophoretic mobility of zero. Several polar substances have been used to this end: acetone, formamide, and thiourea [29]. The marker must be detected by the detector in use. In the case of a UV detector the baseline disturbance caused by a zone of different refractive index to the separation buffer can be used as a signal [30].
16
ELECTROKINETIC CHROMATOGRAPHY
If a substance (the marker of the velocity of the separation carrier) is available that is exclusively transported by the separation carrier and not transported by the mobile phase (k ! 1) and this substance can be detected by the detector in use, then the velocity of the separation carrier can also be determined with a sample containing a marker. It is important to note that the marker of the velocity of the separation carrier does not have to be a neutral substance, and mainly nonpolar azo dyes (Sudan III, Sudan IV) [1,31], dodecanophenone [32] or polycyclic aromatic hydrocarbons [33] have been employed as marker substances. When using negatively charged micellar phases as separation carriers, Terabe et al. suggested using positively charged compounds with a nonpolar structure unit [34]. One of these substances is timepidium bromide [35]. Another compound employed to this end is quinine hydrochloride [8,36]. However, determinations of the velocity of the separation carrier with a simple marker should be treated with caution. Careful investigations have shown that these data can be misleading, especially if mobile phases are used that contain a considerable volume fraction of an organic solvent [37]. Bushey and Jorgenson [24,38] therefore suggested an iteration procedure to determine the separation carrier migration time. This iteration procedure is now used by many scientists working in the field. For chromatographic separations based on solvophobic interaction the Martin equation holds true: there is a linear relationship between the logarithm of the retention factor and the carbon number of the members of a homologous series. Muijseaar et al. [39] verified that this linear relationship is also valid in EKC with a micellar separation carrier (for the homologous series of n-alkyl benzenes and n-alkyl phenyl ketones as solutes and buffers containing sodium dodecylsulfate, decyltrimethylammonium bromide, or hexadecyltrimethylammonium bromide). According to the iteration procedure suggested by Bushey and Jorgenson, the migration time of the longest chain homologue is taken as approximation of the separation carrier migration time. Then the retention factors of the shorter chain homologues are calculated using this approximated separation carrier migration time. With the retention factors of the shorter chain homologues (plotting logarithm of k versus carbon number) a regression line is calculated. A new value for the retention factor of the longest chain member of the series is obtained by extrapolation of the regression line to the corresponding carbon number. Then a new value for the separation carrier migration time is calculated from the extrapolated value of the retention factor for the longest chain member of the series, and this procedure is continued until the difference between a new value for tsc and the value calculated in the last iteration step is below a threshold value. This general procedure has been used by many groups in estimating the migration time of the separation carrier with different mobile phases and different pseudostationary phases: micelles [40–43], microdroplets [44], dendrimers [45] and polymeric pseudostationary phases [46]. Certainly, the marker method is more convenient than the iterative procedure. Several researchers have shown that in the case of purely aqueous mobile phases, the results for tsc obtained from the migration time of a suitable marker can be equivalent to that obtained by the iterative procedure [30,39]. However, Bailey and Dorsey [47] emphasize that small errors in determining the migration time of the marker can lead to drastic errors in the calculated retention factor. From several potential markers of the velocity of the separation carrier, dodecanophenone has been selected by several authors as the most suitable due to its solubility properties and its high absorbance coefficient [30,47].
THEORY OF ELECTROKINETIC CHROMATOGRAPHY
17
Methods to regulate the migration window have been reviewed recently [37]. The migration window reflects the velocity ratio of the two ‘phases’ involved. While it is difficult to modify the effective electrophoretic mobility of the pseudostationary phase, the electroosmotic mobility generated using a native fused-silica capillary can easily be modified by changing the pH of the separation electrolyte. For neutral solutes, the pH has no impact on the retention factor. Rasmussen and McNair [48] showed for micellar EKC with SDS that the elution order for n-alkyl parabenes can easily be reversed by decreasing the separation buffer pH from 7.0 to 3.37. This corresponds to a change in the elution mode (normal elution to reversed direction mode). Also Otsuka and Terabe [49] investigated the effect of the pH of the separation buffer on the velocities of the mobile and the micellar phase (SDS as surfactant). The pH range investigated was from 7.0 to 3.0. The electroosmotic velocity decreased dramatically with a decrease in pH below 5.5, while the electrophoretic velocity of the micellar phase was almost constant throughout the pH-range investigated. At a pH of 5.0 the absolute velocity of the micellar phase was identical to the absolute electroosmotic velocity. Another approach to modifying the velocity of the mobile phase is to use coated capillaries. Janini et al. [50,51] showed that in micellar EKC with SDS and polyacrylamide-coated capillaries, hydrophobic solutes can be separated with high efficiency in a short run time. With these capillaries the electroosmotic flow is almost completely suppressed [52], so that in this case the separation carrier takes over the role of the mobile phase and the surrounding medium can be regarded as equivalent to the stationary phase of conventional chromatography. Consequently, here the column availability Aco is 1 and there is no restricted migration window. 1.4.6
Retention Indices
In 1994 Muijselaar et al. [39] presented the application of the retention index concept in EKC. In chromatography, retention indices have been used for the identification of solutes because they are considered to express the retention with the best reproducibility and precision. They can also be used for structure–activity relationships and the characterization of stationary (and mobile) phases. Generally, the retention index I of a solute is calculated by the logarithmic interpolation between two neighbouring members of a homologous series according to: I ¼ 100z þ 100
lg kS lg kz lg kzþ1 lg kz
ð1:18Þ
where kz and kzþ1 are the retention factors of the homologues with z and z þ 1 carbon atoms, respectively, and kS ¼ retention factor of the solute. Regarding the determination of k in EKC (normal elution mode) it follows: ts t0 tz t0 lg lg t ts t tz sc sc I ¼ 100 z þ 100 ð1:19Þ tzþ1 t0 tz t0 lg lg tsc tzþ1 tsc tz where ts is the migration time of solute, tz and tzþ1 are migration times of the homologues with z and z þ 1 carbon atoms respectively, t0 is the migration time of the front of mobile
18
ELECTROKINETIC CHROMATOGRAPHY
phase, and tsc is the migration time of the front of the separation carrier. Consequently, for the determination of I the migration velocities of the separation carrier and of the surrounding phase also have to be determined exactly. In contrast to the retention factor, the retention index I (a relative quantity) is independent of the phase ratio and consequently, is also independent of the separation carrier concentration, which follows from theory and has been verified experimentally [39]. The dependence of retention indices on temperature was shown to be very small. A significant decrease in relative standard deviations was obtained by comparing retention indices to retention factors [53]. These features make the retention index the ideal parameter for the identification of peaks [54]. Muijselaar et al. [39] and Ahuja and Foley [55] demonstrated for micellar EKC that the series of alkylbenzenes, of n-alkyl phenyl ketone, and of 1-nitroalkanes show a linear relationship between the logarithm of the retention factor and the carbon number of the homologues. Hence, these series can be applied as retention index standards in MEKC. The determination of I for different solutes and different micellar separation carriers indicated that polar and nonpolar compounds are solubilized in different ways. Hence, the shift in I(I) for the same compound and a different separation carrier can serve as a valuable parameter for characterizing the selelectivity in EKC. I values can be applied to the classification of separation carriers in EKC analogously to using the Rorschneider– McReynolds scale in gas chromatography [56]. Ishihama et al. [57] reported, for microemulsion EKC, a high correlation between I and the logarithmic octanol–water distribution coefficient (log POW) for 53 aromatic sample compounds possessing different functionalities. They preferred I to k, because the reproducibility and the repeatability (batch-to-batch and run-to-run) were drastically improved, when using I as a parameter for correlation studies. 1.4.7
Efficiency
Efficiency is a measure of the band broadening occurring during separation. Terminology developed for chromatography has been transferred to capillary electromigration separation methods. The height equivalent to one theoretical plate, (or plate height) H corresponds to the peak variance (in length units) divided by the migrated distance. The plate number N is the migrated distance divided by the plate height. In EKC with standard experimental parameters plate numbers of 200 000 to 300 000 can be obtained under routine conditions. Yu et al. [58] have shown that plate numbers obtained in micellar EKC with neutral analytes having low to medium retention factors can be estimated by a simple model based on longitudinal diffusion and length of the injected sample plug. The efficiency for these neutral analytes was independent of the sufactant (SDS) concentration (15–100 mmol L1). This model has been refined by attributing the instrumental variance not only to the length of the analyte plug but by estimating it from the peaks of the micellar marker (regression method) [59]. In these investigations it was taken into consideration that the overall diffusion coefficient in a medium containing a separation carrier is the weighted average of the analyte diffusion coefficient in the mobile phase and the diffusion coefficient of the analyte-separation carrier adduct [60]. In the past it has been a subject of debate as to whether the efficiency in EKC is also influenced by nonequilibrium effects (separation carrier mass transfer, transchannel mass transfer) or separation carrier (micellar) polydispersity [61–63]. Recent investigations,
THEORY OF ELECTROKINETIC CHROMATOGRAPHY
19
however, show that lower plate numbers than those that would be expected if only instrumental band broadening and diffusion are responsible mechanisms, can be explained by the radial variation of the effective separation carrier mobility resulting from Joule heat [59,64]. It was experimentally verified that peak asymmetries and efficiency loss found with samples of high analyte concentration (concentration overload) can be explained in terms of nonlinear chromatography [65]. The technique of ‘vacancy injection’ makes it possible to determine the distribution isotherms, which can be described by a Langmuir isotherm [66]. Generally, in capillary electromigration separation techniques, laminar flow due to pressure differences between the two ends of the capillary has to be avoided, as it can dramatically reduce efficiency. In the case of differences in the electroosmotic velocity in different segments of the capillary (e.g. partial filling technique) band broadening can result from intersegmental pressure [67]. A further source of band broadening can be analyte adsorption at the inner capillary wall, which should be suppressed by a suitable composition of the separation electrolyte or dynamic or static coating of the inner capillary wall.
1.5 Separation of Weak Electrolytes 1.5.1
Migration of Acids
The separation mechanism of charged compounds in EKC is based on both chromatographic and electrophoretic principles. In 1985, Otsuka et al. [68] studied the migration behaviour and separation of chlorinated phenols by micellar EKC with SDS as separation carrier. They describe the overall effective electrophoretic velocity of a (partially) ionized solute veps in a micellar medium as the weighted sum of the effective electrophoretic velocity vep of the solute and the electrophoretic velocity of the separation carrier vepsc. veps ¼
1 k vep þ vepsc kþ1 k þ1
ð1:20Þ
The true retention factor k [in contrast to the appararent retention factor kapp calculated according to Equation (1.6)] can only be calculated if the effective electrophoretic velocity vep of the solute zone in the separation electrolyte without separation carrier is known. This quantity is mainly determined by CE experiments. However, it should be noted that several assumptions are made in this case: the influence of the separation carrier on ionic strength, dielectric constant and viscosity are assumed to be negligible and, in the case of micellar or microemulsion EKC interactions of the solute with surfactant monomers, are assumed not to occur [69]. In that case, k can be calculated from: meps mep veps vep ¼ ð1:21Þ k ¼ vepsc veps mepsc meps With this equation, k can be calculated even for a fully ionized solute. Otsuka et al. determined vep for phenolic compounds with a buffer containing 5 mmol L1 SDS, assuming that this concentration is below the CMC. Generally when using this approach, it has to be considered that the CMC of a surfactant in an aqueous separation electrolyte
20
ELECTROKINETIC CHROMATOGRAPHY
can be substantially lower than that in pure water [70]. The phenomenological approach of Otsuka et al. [68] satisfactorily explains the dependence of the retention factor k on the pH of the separation buffer for the chlorinated phenols investigated. It is important to note that the retention factor for the negatively charged species was measurable and different from zero. If the migration time t0sc of a solute in absence of the separation carrier is known, then the true retention factor k can also be calculated from t0sc, from the migration time ts of the solute in presence of the separation carrier, and from the migration time of the separation carrier tsc [71]: ts t0sc ð1:22Þ k ¼ t0sc ð1 ts =tsc Þ Equation (1.22) is analogous to Equation (1.6). The problem here is the determination of t0sc. If the electroosmotic velocity is not independent of the separation carrier concentration, then t0sc cannot be directly determined and has to be calculated from the effective electrophoretic velocity of the solute vep and the electroosmotic velocity in the presence of the separation carrier [47]. In 1991 Khaledi et al. [72] investigated in detail the migration behaviour of acidic solutes in micellar EKC dependent on the pH and the concentration of an anionic surfactant. Their phenomenological approach confirms the observations made by Otsuka et al. [68]. The retention factor k of an acid is the weighted average of the retention factors of its undissociated (HA) and its dissociated (A) forms: aq kHA þ FAaq kA k ¼ FHA
ð1:23Þ
aq where FHA is the mole fraction of the undissociated acid in the aqueous phase, FAaq is the mole fraction of the dissociated acid in the aqueous phase, kAH is the retention factor of the undissociated acid, and kA is the retention factor of the dissociated acid, aq þ FAaq ¼ 1. FHA It is assumed that secondary equilibria with buffer constituents do not occur. The mole fractions of the undissociated and the dissociated acid in the aqueous phase are dependent on the pH and the acid constant Ka:
cðHþ Þ cðHþ Þ þ Ka Ka FAaq ¼ cðHþ Þ þ Ka kHA þ kA ðKa =cðHþ ÞÞ k ¼ 1 þ ðKa =cðHþ ÞÞ
aq ¼ FHA
ð1:24Þ ð1:25Þ ð1:26Þ
According to Equation (1.26), the dependence of the retention factor for acidic solutes on the pH is a sigmoidal relationship with maximum slope for pH ¼ pKa. This also holds true for the partition coefficient P. In EKC not only the retention factor k but also the observed (apparent) effective mobility m can be taken as parameter to describe the migration behaviour of acidic solutes quantitatively: SC mSC þ FASC mSC þ FAaq maq m ¼ FHA A
ð1:27Þ
THEORY OF ELECTROKINETIC CHROMATOGRAPHY
21
where F is the molar fraction of the dissociated or undissociated species associated with the separation carrier or in the aqueous phase, mSC is the observed mobility of the separation carrier, and maq A is the observed mobility of the dissociated acid in the aqueous phase, aq sc þ FAsc þ FHA þ FAaq ¼ 1 FHA
If we define the apparent acid constant Ka,app in a medium containing a separation carrier as: Ka;app ¼
ðcSC ðA Þ þ caq ðA ÞÞ caq ðHþ Þ ðcSC ðHAÞ þ caq ðHAÞÞ
ð1:28Þ
it can be shown that the observed overall effective mobility m of a solute in a medium containing a separation carrier can be described as a function of mHA and mA (the apparent mobilities of the two species in the medium containing a separation carrier), the apparent acid constant Ka,app and the molar proton concentration c(H+): m ¼
mHA þ mA ðKa;app =cðHþ ÞÞ 1 þ ðKa;app =cðHþ ÞÞ
ð1:29Þ
with mHA ¼
mA ¼
SC KHA cðSCÞ mSC SC cðSCÞ 1 þ KHA
ð1:30Þ
SC maq A þ KA cðSCÞ mSC 1 þ KASC cðSCÞ
ð1:31Þ
On the basis of these equations, Smith and Khaledi [73] developed a model to predict the migration behaviour of organic acids in micellar EKC employing an anionic surfactant (SDS) in terms of the acid constant Ka, the separation carrier binding constants KASC and SC , the pH and the molar separation carrier concentration. In the case of micelles as KAH separation carriers, KSC is identical to the partition coefficient P multiplied by the molar volume of the micelles. Quang et al. [74] generalized this phenomenological approach, also taking ion pair interactions into consideration. For an acidic solute this type of interaction would be expected for cationic surfactant micelles as separation carriers. In this case ion pair formation of the anionic species with the cationic surfactant monomers (reducing the effective electrophoretic mobility of the solute in the aqueous phase) can take place. In Figure 1.8 the possible interactions of an acidic or a basic solute in an anionic or in a cationic micellar phase are schematically depicted. We then have to take into consideration the ion pair formation equilibrium A þ Smono ! [ASmono] (Smono ¼ surfactant monomer) with the ion pair equilibrium constant KIP. The concentration of Smono corresponds to the critical micelle concentration. The electrophoretic mobility of the ion pair [ASmono] is zero and following molar fractions have to be considered: aq aq SC FHA þ FASC þ FHA þ FAaq þ FASmono ¼ 1
ð1:32Þ
22
ELECTROKINETIC CHROMATOGRAPHY (a)
µeo
HA KHA,sc -
-
+
Ka
-
-
–
KA,sc
-
A– µep
µepsc
(b)
KHA,sc +
+
A– +
+
KA,sc
–
+ A–
µepsc
(c)
+
Ka
+
+
µeo
HA
µeo
KIP µep
B
KB,sc
BH+
-
-
-
-
Kb -
+
-
BH
+
µepsc
(d)
KBH,sc + +
µepsc
µep
µeo
B
+
+
+
–
KIP
KBH,sc
Kb
+ +
–
KBH,sc BH+
µep
Figure 1.8 Interaction of ionizable solutes with micellar separation carriers: (a) acidic solute (AH) with anionic SC; (b) acidic solute (AH) with cationic SC; (c) basic solute (B) with anionic SC; (d) basic solute (B) with cationic SC (according to [74])
THEORY OF ELECTROKINETIC CHROMATOGRAPHY
23
Table 1.2 Influence of cationic surfactant monomers on electrophoretic mobilities of acidic solutes (electrolyte composition: (a) c(H3BO3) ¼ 10 mmol L1, c(Na2B4O7) ¼ 10 mmol L1, pH ¼ 9.0; (b) c(H3BO3) ¼ 10 mmol L1, c(Na2B4O7) ¼ 10 mmol L1, c(DoTAB) ¼ 5 mmol L1, pH ¼ 9.0. capillary 75 mm i.d., 50 cm effective length, 57 cm total length; voltage 25 kV; temperature 25 C; sample pressure injection: 2 s) Solute 4-Hydroxy-3-methoxybenzyl alcohol 4-Hydroxybenzyl alcohol Ethylvanilline Vanilline 4-Methoxybenzoic acid 4-Hydroxybenzaldehyde Vanillic acid 4-Hydroxybenzoic acid 3,4-Dihydroxybenzoic acid
mep /(103 cm2 V1 s1) CE buffer
mep /(103 cm2 V1 s1) CE buffer þ DoTAB
0.034
0.034
0.036 0.228 0.253 0.265 0.271 0.293 0.311 0.394
0.039 0.223 0.246 0.256 0.269 0.297 0.317 0.393
Where FAaq is the molar fraction of the free dissociated species in the surrounding phase, aq is the molar fraction of the ion-paired dissociated species in the surrounding and FASmono phase. One possibility of studying the presence or absence of ion pair formation of charged solutes with oppositely charged surfactant monomers in the aqueous phase is afforded by capillary electrophoresis experiments with surfactants showing a high CMC. In Table 1.2 effective electrophoretic mobilities of several phenol and benzoic acid derivatives determined by CE in the presence and absence of 5 mmol L1 dodecyltrimethylammonium bromide (DoTAB) are given. The CMC of DoTAB (21–23 C) in these two buffers is 13–14 mmol L1 (determined by conductometric titration [75]). Consequently, in these buffers DoTAB is only present as monomer. The comparison of effective electrophoretic mobilities shows that ion pair formation can be neglected in this case. There is no significant decrease in mep due to the presence of surfactant monomers. It has to be emphasized that with 5 mmol L1 DoTAB there is a reversal of the direction of the electroosmotic flow and consequently a reversal of the migration order. Consequently, DoTAB is also present in form of positively charged hemimicelles formed at the interface liquid–capillary. However, there is no band broadening due to solute–capillary wall (hemimicelle) interactions. It has to be emphasized that the solutes studied here are relatively hydrophilic. Muijselaar et al. [69] verified experimentally that both the mobility model and the retention model describe the migration of monovalent acids in micellar EKC well. However, they observed (hydrophobic) interaction of the (hydrophobic) undissociated form of the acid with surfactant monomers. This interaction is a phenomenon that may have a marked influence on the determination of true retention factors for hydrophobic species. 1.5.2
Migration of Bases
The same concept outlined in the previous section was used by Strasters and Khaledi [76] in 1991 to describe the migration behaviour of cationic solutes in EKC with anionic
24
ELECTROKINETIC CHROMATOGRAPHY
micellar separation carrier. They took into consideration the acid–base equilibrium, ion pair formation between the conjugated acid of the base and the surfactant monomer, and the distribution equilibria of both the base and its conjugated acid between the aqueous phase and the separation carrier (see Figure 1.8c). In the case of an ion pair formation constant KIP approaching infinity the solute will be present in the aqueous phase either as neutral species or as neutral ion pair. Consequently the effective electrophoretic mobility of this solute vep [refer to Equation (1.20)] in the aqueous medium is zero, and the true retention factor can be calculated from [refer to Equation (1.21)]: k ¼
meps mepsc meps
ð1:33Þ
In the absence of ion pair formation k can be calculated from Equation (1.21). Employing these two equations, Strasters and Khaledi calculated the retention factors for several basic solutes dependent on the concentration of SDS in the separation electrolyte. They obtained linear relationships differing in the x-axis intersection. As the x-axis intersection should correspond to the CMC of SDS, negative values clearly indicate the invalidity of the approach. Their results suggest that ion pair formation must not be neglected, especially for the hydrophobic solutes. On the basis of only five experiments, Quang et al. [74] succeeded in correctly modeling the migration behaviour of 17 aromatic amines separated by micellar EKC with an anionic surfactant (SDS) within a parameter range of c(SDS) ¼ 20–85 mmol L1 and pH ¼ 7.0–12.0. In their approach it was assumed that the protonated base is present in the surrounding aqueous phase only as an ion pair formed with the surfactant monomer. It would be interesting to model the migration behaviour of partly protonated bases or partly dissociated acids employing separation carriers that are not present as surfactant monomers (e.g. polymeric separation carriers) in order to verify the phenomenological approach outlined in the previous sections. Those equations concerning the resolution of charged solutes derived by Corstjens et al. [77] assume that the electrophoretic mobility of a charged solute in the surrounding medium is constant and not influenced by the separation carrier.
1.6 Separation of Ions In 1992 Kaneta et al. [78] studied the migration behaviour of inorganic anions in micellar EKC using a cationic surfactant (cetyltrimethylammonium chloride). They employed the overall effective electrophoretic mobility meps [which is not identical to the observed overall effective mobility m of a solute in a medium containing a separation carrier in Equation (1.29)] as the parameter to describe the migration (m ¼ meps þ meo ). In the presence of a cationic surfactant the electrophoretic mobility of an anion is influenced by interaction with the surfactant monomers (ion pair formation) and by interaction with the positively charged micellar pseudophase (‘distribution’). In fact, in this case the micellar pseudophase might be regarded as an ion-exchange separation carrier.
THEORY OF ELECTROKINETIC CHROMATOGRAPHY
25
Below the CMC, the effective electrophoretic mobility mep is given by: mep ¼ FAaq mep;ion
ð1:34Þ
where FAaq is the molar fraction of free ion and mep,ion is the electrophoretic mobility of free ion. Regarding that: KIP ¼
cð½ASmono Þ cðAÞ cðSmono Þ
ð1:35Þ
cðAÞ cðAÞ þ cð½ASmono Þ
ð1:36Þ
and FAaq ¼ it follows: 1 cðSmono Þ KIP 1 ¼ þ mep;ion mep mep;ion
ð1:37Þ
Indeed, assuming that c(Smono) is given by the total surfactant concentration and that plotting 1/mep versus c(Smono) results in a straight line, this verifies the theoretical considerations. At the CMC there is a change in the slope, making it possible to determine the CMC and to determine the electrophoretic mobility of a solute at the CMC using a regression method. Above the CMC c(Smono) remains constant, consequently also the electrophoretic mobility mep,CMC of the anion in the surrounding aqueous phase remains constant independent of the total surfactant concentration. If the electrophoretic mobility of the separation carrier mepsc is known, the retention factor is then given by: meps mep;CMC ð1:38Þ k ¼ mepsc meps Therefore, the method of Kaneta et al. [78] makes it possible to determine true retention factors in a micellar medium for charged and partially charged compounds without neglecting the effects of solute surfactant monomer association.
1.7 Application of Neutral Separation Carriers In 1997 Collet and Gareil [79] reported the MEKC separation of long chain saturated and unsaturated free fatty acids in an alkaline medium with neutral micelles formed by polyoxyethylene-23-dodecyl ether (Brij 35). In this case, the two phases (the separation carrier and the surrounding phase) itself have the same velocity; however, this is not the case for the solutes in the two phases. Neutral separation carriers can only be used in EKC, if the solutes to be separated have an effective electrophoretic mobility in the surrounding phase. If the two solutes to be separated have identical effective electrophoretic mobilities in the surrounding phase, then all considerations made in Section 1.4 can be also applied, provided that the parameter t0 (migration time of the front of the surrounding phase) is replaced by the parameter t0sc, which is the migration time of a solute zone in the absence of the separation carrier.
26
ELECTROKINETIC CHROMATOGRAPHY
1.8 Conclusions As pointed out by Terabe [2] EKC can be regarded as an intermediate between electrophoresis and chromatography. Consequently, the separation process can be described either as a chromatographic or as an electrophoretic process. Both descriptions are valid and fully describe the migration behaviour for neutral solutes or for solutes being present in the mobile phase either in the uncharged form or in the form of an uncharged ion pair. In the case that the solute is also present in a charged form in the mobile phase, both the electrophoretic properties of the solutes and their interaction with the separation carrier have to be taken into consideration.
List of Symbols and Abbreviations Aco c(A) c(H+) CMC E FAaq aq FHA aq FASmono I k k kA kAH ks kz kzþ1 Ka Kc KIP L0 n N P POW Pr Rs t0 t0sc tmob ts tsc trsc
column availability molar concentration of the free complex forming additive molar concentration of protons critical micellar concentration electric field strength mole fraction of the dissociated acid in the aqueous phase mole fraction of the undissociated acid in the aqueous phase molar fraction of the ion-paired dissociated species in the aqueous phase retention index retention factor mean retention factor retention factor of the dissociated acid retention factor of the undissociated acid retention factor of the solute retention factor of the homologue with z carbon atoms retention factor of the homologue with z þ 1 carbon atoms acid constant complex equilibrium constant ion pair equilibrium constant virtual column length peak capacity plate number partition coefficient octanol/water partitioning coefficient phase velocity ratio resolution migration time of the front of the surrounding (mobile) phase migration time of a solute in absence of the separation carrier residence time in the mobile phase migration time of the solute zone migration time of the front of the separation carrier residence time associated with the separation carrier
THEORY OF ELECTROKINETIC CHROMATOGRAPHY
tstat tz tzþ1 veo vep veps vepsc vmob vs vsc Vmob Vsc z a b e Z j m mAaq meo mep mep,CMC mep,ion meps mepsc
27
residence time in the stationary phase migration time of the homologue with z carbon atoms migration time of the homologue with z+1 carbon atoms electroosmotic velocity (effective) electrophoretic velocity overall effective electrophoretic velocity of a (partially) ionized solute electrophoretic velocity of the separation carrier velocity of the mobile phase observed velocity of a solute zone observed velocity of the separation carrier volume of surrounding (mobile) phase volume of separation carrier number of carbon atoms selectivity factor degree of complexation electric permittivity of the surrounding medium viscosity of the surrounding medium phase ratio observed overall effective mobility of a solute in a medium containing a separation carrier observed mobility of the dissociated acid in the aqueous phase electroosmotic mobility (effective) electrophoretic mobility (effective) electrophoretic mobility of a solute at the CMC of the added surfactant electrophoretic mobility of free ion overall effective electrophoretic mobility of a (partially) ionized solute electrophoretic mobility of the separation carrier electrokinetic potential (zeta potential)
References [1] S. Terabe, K. Otsuka and T. Ando. Electrokinetic chromatography with micellar solution and open-tubular capillary, Anal. Chem., 57, 834–841 (1985). [2] S. Terabe. Electrokinetic chromatography: an interface between electrophoresis and chromatography, Trends Anal. Chem., 8, 129–134 (1989). ˚ . Jo¨ nsson and R.M. Smith. Terminology for analytical capillary electro[3] M.L. Riekkola, J.A migration techniques, Pure Appl. Chem., 76, 443–451 (2004). [4] S. Terabe, K. Otsuka, K. Ichikawa and A. Tsuchiya. Electrokinetic separations with micellar solutions and open-tubular capillaries, Anal. Chem., 56, 111–113 (1984). [5] J.P. Foley. Optimization of micellar electrokinetic chromatography, Anal. Chem., 62, 1302– 1308 (1990). [6] U. Pyell and U. Bu¨ tehorn. Optimization strategies in micellar electrokinetic capillary chromatography. One-parameter optimization of the concentration of sodium dodecyl sulfate and the concentration of modifiers (urea and glucose), Chromatographia, 40, 175–184 (1995).
28
ELECTROKINETIC CHROMATOGRAPHY
[7] M.D. Trone, M.S. Leonard and M.G. Khaledi. Congeneric behavior in estimations of octanol– water partition coefficients by micellar electrokinetic chromatography, Anal. Chem., 72, 1228– 1235 (2000). [8] A. Gavenda, P. Bedna´ r, P. Barta´ k, P. Adamovsky´ , J. Sˇ evcˇ´ık, P. Tzoumas and J. Ulrichova´ . Estimation of partition coefficients by MEKC Part 1: Potential pharmaceuticals based on 2-quinolone and 2-indolone, J. Sep. Sci., 24, 723–728 (2001). [9] G. Kortu¨ m. Lehrbuch der Elektrochemie, Verlag Chemie, Weinheim, 1972. [10] J.W. Jorgenson and K.D. Lukacs. Capillary zone electrophoresis, Science, 222, 266–272 (1983). [11] S. Terabe. Micellar electrokinetic chromatography, in Capillary Electrophoresis Technology, N.A. Guzman (Ed.), Chromatographic Science Series 64, Marcel Dekker, New York (1993). [12] J. Vindevogel and P. Sandra. Introduction to Micellar Electrokinetic Chromatography, Hu¨ thig, Heidelberg, 1992. [13] H. Watarai. Microemulsions in separation sciences, J. Chromatogr. A, 780, 93–102 (1997). [14] C.P. Palmer. Polymeric and polymer-supported pseudostationary phases in micellar electrokinetic chromatography: performance and selectivity, Electrophoresis, 21, 4054–4072 (2000). [15] C.P. Palmer and J.P. McCarney, Developments in the use of soluble ionic polymers as pseudostationary phases for electrokinetic chromatography and stationary phases for electrochromatography, J. Chromatogr. A., 1044, 159–176 (2004). [16] J. Knox. Terminology and nomenclature in capillary electroseparation systems, J. Chromatogr. A, 680, 3–13 (1994). [17] J.K. Strasters and M.G. Khaledi. Migration behavior of cationic solutes in micellar electrokinetic capillary chromatography, Anal. Chem., 63, 2503–2508 (1991). [18] A. Berthod and S. Carda-Broch. Determination of liquid–liquid partition coefficients by separation methods, J. Chromatogr. A., 1037, 3–14 (2004). [19] C.F. Poole and S.K. Poole. Interphase model for retention and selectivity in micellar electrokinetic chromatography, J. Chromatogr. A., 792, 89–104 (1997). [20] K.R. Nielsen and J.P. Foley. Micellar electrokinetic chromatography, in Capillary Electrophoresis, Theory and Practice, P. Camilleri (Ed.), CRC Press, Boca Raton, 1998. [21] P. Gareil. Migration time and resolution in micellar electrokinetic chromatography, Chromatographia, 30, 195–200 (1990). [22] H. Harino, M. Tanaka, T. Araki, Y. Yasaka, A. Masuyama, Y. Nakatsuji, I. Ikeda, K. Funazo and S. Terabe. Double-chain surfactants with two sulfonate groups as micelle-forming reagents in micellar elctrokinetic chromatography of naphthalene derivatives, J. Chromatogr. A., 715, 135–141 (1995). [23] S. Terabe, Y. Miyashita, Y. Ishihama and O. Shibata. Cyclodextrin-modified micellar electrokinetic chromatography: separation of hydrophobic and enantiomeric compounds, J. Chromatogr., 636, 47–55 (1993). [24] M.M. Bushey and J.W. Jorgenson. Effects of methanol-modified mobile phase on the separation of isotopically substituted compounds by micellar electrokinetic chromatography, J. Microcolumn Sep., 1, 125–130 (1989). [25] C.X. Zhang, Z.P. Sun and D.K. Lin. Micellar electrokinetic capillary chromatography theory based on conventional chromatography, J. Chromatogr. A, 655, 309–316 (1993). [26] J.C. Giddings. Maximum number of components resolvable by gel filtration and other elution chromatographic methods, Anal. Chem., 39, 1027–1028 (1967). [27] P.G. Muijselaar, M.A. van Straten, H.A. Claessens and C.A. Cramers. Determination of diffusion coefficients and separation numbers in micellar electrokinetic chromatography, J. Chromatogr. A., 766, 187–195 (1997). [28] S. Kolb, T. Welsch and J.P. Kutter. Influence of counter pressure on separation performance in SDS-MEKC, J. High Resol. Chromatogr., 21, 435–439 (1998).
THEORY OF ELECTROKINETIC CHROMATOGRAPHY
29
[29] E.S. Ahuja, E.L. Little and J.P. Foley. Selected organic solvents as electroosmotic velocity markers in micellar electrokinetic capillary chromatography, J. Liquid Chromatogr., 15, 1099– 1113 (1992). [30] E. Fuguet, C. Ra`fols, E. Bosch and M. Rose´ s. Solute–solvent interactions in micellar electrokinetic chromatography: IV. Characterization of electroosmotic flow and micellar markers, Electrophoresis, 23, 56–66 (2002). [31] S. Takeda, S. Wakida, M. Yamane, K. Higashi and S. Terabe. Effect of alkyl chain length of sodium N-acyl sarcosinates on migration behavior in micellar electrokinetic chromatography, J. Chromatogr. A, 744, 135 (1996). [32] E. Van Hove, R. Szu¨ cs and P. Sandra. Alcohol modifiers in MEKC with SDS as surfactant. Study on the influence of the alcohol chain length (C1–C12), J. High Resol. Chromatogr., 19, 674–678 (1996). [33] J. Gorse, A.T. Balchunas, D.F. Swaile and M.J. Sepaniak. Effects of organic mobile phase modifiers in micellar electrokinetic capillary chromatography, J. High. Resol. Chromatogr. CC, 11, 554–559 (1988). [34] S. Terabe, Y. Ishihama, H. Nishi, T. Fukuyama and K. Otsuka. Effect of urea addition in micellar electrokinetic chromatography, J. Chromatogr., 545, 359–368 (1991). [35] S. Terabe, O. Shibata and T. Isemura. Band broadening evaluation by back-and-forth capillary electrophoresis, J High Resol. Chromatogr., 14, 52–55 (1991). [36] P. Tsai, B. Patel and C.S. Lee. Micellar electrokinetic chromatography with electroosmotic step gradient elution, Electrophoresis, 15, 1229–1233 (1994). [37] U. Pyell. Determination and regulation of the migration window in electrokinetic chromatography, J. Chromatogr. A., 1037, 479–490 (2004). [38] M.M. Bushey and J.W. Jorgenson. Separation of dansylated methylamine and dansylated methyl-d3-amine by micellar electrokinetic capillary chromatography with methanol-modified mobile phase, Anal. Chem., 61, 491–493 (1989). [39] P.G.H.M. Muijselaar, H.A. Claessens and C.A. Cramers. Application of the retention index concept in micellar electrokinetic capillary chromatography, Anal. Chem., 66, 635–644 (1994). [40] P.G.H.M. Muijselaar, H.A. Claessens and C.A. Cramers. Parameters controlling the elution window and retention factors in micellar electrokinetic capillary chromatography, J. Chromatogr. A., 696, 273–284 (1995). [41] N. Chen and S. Terabe. A quantitative study on the effect of organic modifiers in micellar electrokinetic chromatography, Electrophoresis, 16, 2100–2103 (1995). [42] N. Chen, S. Terabe and T. Nakagawa. Effect of organic modifier concentrations on electrokinetic migrations in micellar electrokinetic chromatography, Electrophoresis, 16, 1457–1462 (1995). [43] U. Bu¨ tehorn and U. Pyell. Simultaneous use of urea and acetonitrile as organic modifiers for resolution optimization in micellar electrokinetic chromatography, J. Chromatogr. A, 792, 157–163 (1997). [44] S.E. Lucangioli, C.N. Carducci, S.L. Scioscia, A. Carlucci, C. Bregni and E. Kenndler. Comparison of the retention characteristics of different pseudostationary phases for microemulsion and micellar electrokinetic chromatography of betamethasone and derivatives, Electrophoresis, 24, 984–991 (2003). [45] N. Tanaka, T. Fukutome, H. Hosoya, K. Kimata and T. Araki. Polymer-supported pseudostationary phase for electrokinetic chromatography. Electrokinetic chromatography in a full range of methanol–water mixtures with alkylated starburst dendrimers, J. Chromatogr. A, 716, 57–67 (1995). [46] N. Tanaka, K. Nakagawa, K. Hosoya, C.P. Palmer and S. Kunugi. Control of migration time window and selectivity in electrokinetic chromatography with mixed polymeric pseudostationary phases, J. Chromatogr. A., 802, 23–33 (1998).
30
ELECTROKINETIC CHROMATOGRAPHY
[47] D.J. Bailey and J.G. Dorsey. pH effects on micelle–water partitioning determined by micellar electrokinetic chromatography, J. Chromatogr. A., 852, 559–571 (1999). [48] H.T. Rasmussen and H.M. McNair. Influence of pH on elution behavior in micellar electrokinetic capillary chromatography, J. High Resol. Chromatogr., 12, 635–636 (1989). [49] K. Otsuka and S. Terabe. Effects of pH on electrokinetic velocities in micellar electrokinetic chromatography, J. Microcol. Sep., 1, 150–154 (1989). [50] G.M. Janini, G.M. Muschik and H.J. Issaq. Micellar electrokinetic chromatography in zeroelectroosmotic flow environment, J. Chromatogr. B, 683, 29–35 (1996). [51] G.M. Janini, H.J. Issaq and G.M. Muschik. Electrokinetic chromatography without electroosmotic flow, J. Chromatogr. A., 792, 125–141 (1997). [52] P. Jandera, J. Fischer, J. Jebava´ and H. Effenberger. Characterization in micellar highperformance liquid chromatography, in micellar electrokinetic chromatography and in micellar electrokinetic chromatography with reduced flow, J. Chromatogr. A., 914, 233–244 (2001). [53] P.G. Muijselaar. Retention indices in micellar electrokinetic chromatography, J. Chromatogr. A., 780, 117–127 (1997). [54] P. Jandera, J. Fischer and H. Effenberger. Characterization of retention in micellar highperformance liquid chromatography and in micellar electrokinetic chromatography using lipophilicity and polarity indices, J. Chromatogr. A., 807, 57–80 (1998). [55] E.S. Ahuja and J.P. Foley. A retention index for micellar electrokinetic chromatography, Analyst, 119, 353–360 (1994). [56] P.G.H.M. Muijselaar, H.A. Claessens and C.A. Cramers. Characterization of pseudostationary phases in micellar electrokinetic chromatography by applying linear solvation energy relationships and retention indexes, Anal. Chem., 69, 1184–1191 (1997). [57] Y. Ishihama, Y. Oda, K. Uchikawa and N. Asakawa. Evaluation of solute hydrophobicity by microemulsion electrokinetic chromatography, Anal. Chem., 67, 1588–1595 (1995). [58] L. Yu, T.H. Seals and J.M. Davis. Reexamination of dependence of plate number on SDS concentration in micellar electrokinetic chromatography, Anal. Chem., 68, 4270–4280 (1996). [59] J.M. Davis. New assessments of dispersion in micellar electrokinetic chromatography, J. Microcol. Sep., 10, 479–489 (1998). [60] J.M. Davis. Determination of micellar self-diffusion coefficients by micellar electrokinetic chromatography, Analyst, 123, 337–341 (1998). [61] S. Terabe, K. Otsuka and T. Ando. Band broadening in electrokinetic chromatography with micellar solutions and open-tubular capillaries, Anal. Chem., 61, 251–260 (1989). [62] J.M. Davis. Random-walk theory of nonequilibrium plate height in micellar electrokinetic capillary chromatography, Anal. Chem., 61, 2455–2461 (1989). [63] L. Yu and J.M. Davis. Study of high-field dispersion in micellar electrokinetic chromatography, Electrophoresis, 16, 2104–2120 (1995). [64] T.H. Seals, J.M. Davis, M.R. Murphy, K.W. Smith and W.C. Stevens. Influence of organic modifier concentration on plate number in micellar electrokinetic chromatography: .1.2-propanol, Anal. Chem., 70, 4549–4562 (1998). [65] K.W. Smith and J.M. Davis. Quantitative study of concentration overload, peak asymmetry and efficiency loss in micellar electrokinetic chromatography, Anal. Chem., 74, 5969–5981 (2002). [66] S. Katsuta and K. Saitoh. A micellar electrokinetic chromatographic method for determination of solubilization isotherms in surfactant micelles, Anal. Chem., 70, 1389–1393 (1998). [67] D. Michalke and T. Welsch. Study of flow profile distortions and efficiency in counter pressure moderated partial filling micellar electrokinetic chromatography in relation to the relative buffer zone lengths, J. Chromatogr. A., 960, 209–220 (2002). [68] K. Otsuka, S. Terabe and T. Ando. Electrokinetic chromatography with micellar solutions. Retention behavior and separation of chlorinated phenols, J. Chromatogr., 348, 39–47 (1985).
THEORY OF ELECTROKINETIC CHROMATOGRAPHY
31
[69] P.G. Muijselaar, H.A. Claessens and C.A. Cramers. Migration behavior of monovalent weak acids in micellar electrokinetic chromatography. Mobility model versus retention model, J. Chromatogr. A., 765, 295–306 (1997). [70] U. Pyell and U. Bu¨ tehorn. Optimization strategies in micellar electrokinetic chromatography. Optimization of the temperature of the separation capillary, Chromatographia, 40, 69–77 (1995). [71] H. Corstjens, H.A.H. Billiet, J. Frank and K.Ch.A.M. Luyben. Equation for the description of the resolution of charged solutes in micellar electrokinetic capillary chromatography, J. Chromatogr. A., 753, 121–131 (1996). [72] M.G. Khaledi, S.C. Smith and J.K. Strasters. Micellar electrokinetic capillary chromatography of acidic solutes: migration behavior and optimization strategies, Anal. Chem., 63, 1820–1830 (1991). [73] S.C. Smith and M.G. Khaledi. Prediction of the migration behavior of organic acids in micellar electrokinetic chromatography, J. Chromatogr., 632, 177–184 (1993). [74] C. Quang, J.K. Strasters and M.G. Khaledi. Computer-assisted modeling, prediction, and multifactor optimization in micellar electrokinetic chromatography of ionizable solutes, Anal. Chem., 66, 1646–1653 (1994). [75] A. Dworschak. Einsatzmo¨glichkeiten der Mizellaren Elektrokinetischen Chromatographie, PhD thesis, Marburg (Germany) 2000. [76] J.K. Strasters and M.G. Khaledi. Migration behavior of cationic solutes in micellar electrokinetic capillary chromatography, Anal. Chem., 63, 2503–2503 (1991). [77] H. Corstjens, H.A.H. Billiet, J. Frank and K.Ch.A.M. Luyben. Equation for the description of the resolution of charged solutes in micellar electrokinetic capillary chromatography, J. Chromatogr. A., 753, 121–131 (1996). [78] T. Kaneta, S. Tanaka, M. Taga and H. Yoshida. Migration behavior of inorganic ions in micellar electrokinetic capillary chromatography using a cationic surfactant, Anal. Chem., 64, 798–801 (1992). [79] J. Collet and P. Gareil. Micellar electrokinetic chromatography of long chain saturated and unsaturated free fatty acids with neutral micelles. Considerations regarding selectivity and resolution optimization, J. Chromatogr. A., 792, 165–177 (1997).
2 Determination of Critical Micelle Concentrations by Capillary Electrokinetic Techniques Thomas Le Saux, Anne Varenne and Pierre Gareil
2.1 Introduction Surfactants are amphiphilic molecules composed of a hydrophobic hydrocarbonaceous tail comprising at least six–eight carbon atoms, attached to a hydrophilic, more or less bulky, neutral or ionic head group. At low concentration in aqueous solution, surfactant molecules are almost entirely dispersed, although dimers, trimers, etc. can be found [1]. When concentration in solution is over a threshold value, surfactants form aggregates, with the part of solvent-like polarity turned outside, towards the solvent. The first type of aggregate appearing beyond this threshold concentration can often be considered as spherical in shape and is called a micelle [2]. The corresponding threshold is called the critical micelle concentration (CMC). Micellization of surfactants in aqueous media occurs due to the fact that the reduction of the hydrocarbon–water interface is energetically favored. The CMC at which aggregation takes place reflects the fact that the hydrophobic interaction between the hydrocarbonaceous moieties of the surfactant molecules is balanced by hydration and electrostatic repulsive effects of hydrophilic head groups [3]. Thus hydrophobic forces control the formation of micelles, while electrostatic ones limit their maximum size (aggregation number) under determined conditions. CMC is commonly recognized not to be a sharp transition point but a concentration range below which the solution contains negligible amounts of micelles and above which most of the surfactants are found in the
Electrokinetic Chromatography: Theory, Instrumentation and Applications Edited by U. Pyell # 2006 John Wiley & Sons, Ltd. ISBN: 0-470-87102-4
34
ELECTROKINETIC CHROMATOGRAPHY
form of micelles [4,5]. In practice, CMC is the lowest concentration at which a given experimental method detects associates. CMC depends on the surfactant nature (hydrophobicity of the hydrocarbon chain, net charge on the surfactant, nature of the polar head and counterion) [6], medium composition (pH [7], ionic strength [8], counterion in case of ionic surfactants [9], additives [10,11], hydroorganic solvent composition [12]) and temperature [7]. Surfactants have found widespread application for their solubilizing, homogenizing, cleaning or degreasing properties in many industrial processes (e.g. liquid–liquid and cloud point extractions, membrane separations, biotechnologies, formulations) and areas (petroleum, lubricants, painting, phamaceuticals, health care, etc) [13]. Moreover, they play a key role in the principles of several analytical methods, such as polyacrylamide gel electrophoresis, ion-pair chromatography, micellar normal or reversed phase liquid chromatography, and more recently micellar electrokinetic chromatography. For optimizing all of these applications, the knowledge of CMC values is of prime importance. Several tens of methods have been used for the determination of the CMC of aqueous surfactants systems, among them being electric conductivity [14–17], dielectric constant [16] and surface tension measurements [18–20], electrochemical methods such as potentiometry [21] and cyclic voltametry [22], dye solubilization associated with various modes of spectroscopic monitoring (absorbance [5,14,15,23,24], fluorescence [25,26]), static [27,28] and dynamic light scattering [28–30] and even NMR [31] or speed of sound [10]). Insofar as the value attributed to CMC corresponds to a major variation of the physicochemical property measured, reported CMC values often depend somewhat on the technique employed. The choice of a technique may be governed by its availability, easiness of use, and also by the surfactant type (ionic, nonionic) and its spectroscopic properties. In the last decade, several additional methods based on capillary electrophoretic (CE) techniques have been also proposed for determining CMC values of surfactants. These methods can be sorted into three groups according to their overall approach: the current intensity method [32,33], which is no more than a CE instrumental setup of the aforementioned classical conductivity method; the zonal methods [17,34–36], which are based on the migration speed monitoring of a micelle-interacting marker injected as an analyte in a capillary filled with a surfactant solution, and the frontal methods [37,38], relying on partial electrophoretic separations of monomeric and aggregated surfactant forms to produce minimum perturbation of the monomer–aggregate equilibrium, as in a dialysis process, and subsequent quantitation of the monomeric form. Some of these methods have been recently reviewed by C.-E. Lin [39]. The aim of this chapter is to provide the reader with a brief comprehensive presentation of each methodological approach, with special attention being paid to the most critical experimental conditions and to procedures for data treatment. Afterwards, these methods will be compared to the more conventional ones on the basis of CMC values obtained and on method performance and suitability.
DETERMINATION OF CRITICAL MICELLE CONCENTRATIONS
35
2.2 Current Intensity Method This approach, which was first applied by Tickle et al. [32] and later described in greater detail by Cifuentes et al. [33], simply uses CE instrumentation to set up conductivity measurements. According to Ohm’s law, the current intensity inside a capillary, filled with an ionic surfactant solution at a concentration cS lower than the CMC, is related to applied voltage V by: I¼
p dc2 FðmS þ mX ÞcS V 4L
ð2:1Þ
where F is Faraday’s constant, dc and L, the inner diameter and total length of the capillary, respectively, and mS and mX, the mobilities of the surfactant and counterion. For cS higher than the CMC, Ohm’s law becomes: I¼
p dc2 FðmS þ mX Þ CMC þ ðmX þ mmic Þð1 yÞðcS CMCÞ V 4L
ð2:2Þ
where mmic is the mobility of the micelle and y the rate of counterions condensed to the micelle. Plotting current intensity against surfactant concentrations for a given capillary, voltage and temperature, will result in two straight lines corresponding to the monomeric and micellar states of the surfactant. Equations (2.1) and (2.2) explain why the slope of the second straight line (cS > CMC) is normally lower than that of the first one (cS < CMC) and the difference in slopes of these two parts appears to be related directly to counterion condensation and micelle size (with typical y values in the 0.6–0.8 range). Figure 2.1 shows an example of current–concentration plot obtained experimentally. Similar curves are given in the literature [26,37,40,41]. The abscissa of the break displayed is therefore recognized as the CMC. This value can be obtained from classical least-squares linear regression of the data points or from more sophisticated mathematical treatment using second derivatives [16]. For better precision, experimental points too close to the intersection point should be ignored.
Figure 2.1 Plot of the current intensity inside the capillary as a function of the ionic surfactant concentration present in the filling solution. Fused silica capillary: 75 mm i.d. 47 cm length. Filling solution: SDS at varying concentrations in water. Applied voltage: 25 kV. Temperature: 25 C. (Reprinted with permission from [33], copyright 1997 American Chemical Society)
36
ELECTROKINETIC CHROMATOGRAPHY
Capillary diameter and length, and voltage to be applied should be chosen in order to obtain adequate current values, preferably, in the 1–100 mA range [33, 41]. Especially, for low conductivity-surfactant solutions, capillary diameter and voltage should be increased and capillary length decreased. Proper control of Joule heating also requires power to be kept dissipated below 2 W/m inside the capillary [33]. As for the classical conductivity method, the current intensity method applies only to anionic [33] and cationic [40] surfactants and its precision depends on the extent of the difference between the slopes in the premicellar and micellar states, which is related in turn to the system studied. In this connection, it allows CMC determinations in pure water but measurements in high ionic strength background electrolytes (BGE) can become problematic. It also has the great advantage of not needing any marker likely to perturb the aggregation equilibrium. Finally, as compared with classical conductimetry, the CE procedure benefits from having easier temperature control and can be operated unattended. It also requires about one hundred times less surfactant solution volume and amount [33]. CMC values for SDS obtained by the current method in various electrolyte and temperature conditions are given in Lin’s review [39], Table 2.1.
2.3 Zonal Methods Zonal electrophoretic methods for CMC determination all rely on the principles of micellar electrokinetic chromatography (MEKC), as formulated by Terabe et al. [34]. This method was devised to extend the scope of CE to the simultaneous separation of neutral and ionic analytes. Briefly, MEKC consists in implementing CE by using a charged micelle-containing BGE, so that neutral analytes are allowed to partition between the hydrophobic core of the micelles and the electrolyte bulk, just like in a reversed-phase chromatographic system. When an electric field is applied, neutral analytes will gain a nonzero electrophoretic mobility proportional to the micelle mobility, and separate from each other according to their affinity for the micelles. Figure 2.2 schematically depicts a typical chromatogram obtained when MEKC is achieved in a (a) Micelle inj. (b)
Solute column
Water
Solute
Water det. Micelle
Time 0
t0
tR
tmc
Figure 2.2 Schematic representation of (a) the zones separated in a capillary and (b) a typical micellar electrokinetic chromatogram. t0 ¼ teo: retention time of a hydrophilic electroosmotic flow marker. tR: retention times of a neutral analyte. tmic: retention time of a micelle-mobility marker. (Reprinted with permission from [34], copyright 1985 American Chemical Society)
DETERMINATION OF CRITICAL MICELLE CONCENTRATIONS
37
bare silica capillary with an anionic surfactant such as SDS, and when micelles are dragged to the cathode by electroosmotic flow (normal elution mode). The neutral analytes are detected within a range delimited by the elution time of an electroosmotic flow marker, teo and that of a micelle-mobility marker, tmic. For CMC determination applying MEKC principles, the surfactant to be studied is introduced into the BGE of interest at various successive concentrations encompassing the CMC, and the retention time of an appropriate marker interacting with the surfactant micelles is measured. The marker is most often detected from a spectroscopic property that does not belong to the surfactant. UV-transparent surfactants are therefore the best candidates. Four methodological variants are distinguished, mainly according to the mode of data treatment, namely the retention factor, retention time, mobility and zeta-potential methods. 2.3.1
Retention Factor Method
In MEKC, retention factor k is defined as [34]: k ¼ Kmic
Vmic v ðct CMCÞ ¼ Kmic Vaq 1 v ðct CMCÞ
ð2:3Þ
where Kmic is the analyte partition coefficient between micellar pseudophase and aqueous BGE bulk, Vmic and Vaq the volume fractions of both phases, ct the total surfactant concentration, and v the partial molar volume of the surfactant. For ct values of practical utility, v (ct -CMC) is much smaller than 1 and the retention factor can be approximated to: k ¼ Kmic v ðct CMCÞ
ð2:4Þ
For a neutral analyte, k can be experimentally accessed from measurements of its retention time tR and those of electroosmotic flow and micelle markers, teo and tmic respectively, using [34]: k¼
tR teo teo ð1 tR =tmic Þ
ð2:5Þ
According to Equation (2.4), the plot of experimental k values as a function of the total surfactant concentration will give a straight line with the x-axis intercept corresponding to CMC (see Figure 2.3, from the work by Terabe et al. [34]. Similar results were given elsewhere [42, 43]). For CMC determination, this approach requires simultaneous injection of a properly selected micelle-interacting analyte, of an electroosmotic flow marker that does not interact at all with the micelles, and of a micelle-mobility marker. To be valid, Equations (2.4) and (2.5) assume that the surfactant monomers and micelles do not adsorb to the capillary wall and that the micelle-mobility marker does not change the structure of the micelles. In addition, the micelle-interacting analyte should not interact appreciably with the surfactant monomer and should not perturb the micellization equilibrium. As anticipated by Terabe et al. [34], the retention factor method may not be accurate and reliable due to uncertainties in the measurements of teo and tmic. Nevertheless, this method has been used, in particular, for the determination of the CMC values of SDS in various electrolyte solutions (see [39], Table 2.1).
38
ELECTROKINETIC CHROMATOGRAPHY
Figure 2.3 Experimental dependence of retention factor k on total SDS concentration for various neutral analytes: resorcinol (2), phenol (3), p-nitroaniline (4), nitrobenzene (5), toluene (6) and 2-naphthol (7). BGE: 0.1 M borate–0.05 M phosphate buffer, pH 7. Temperature: 35 C. The straight lines represent the linear regression performed. (Reprinted with permission from [34], copyright 1985 American Chemical Society)
2.3.2
Retention Time Method
To obviate some of the aforementioned difficulties, Nakamura et al. [36] proposed direct plotting of the retention time of a micelle-interacting analyte in terms of total surfactant concentration. Retention time is related to retention factor by: tR ¼
teo ð1 þ kÞ 1 þ ðteo =tmic Þk
ð2:6Þ
Below the CMC, retention time should be constant and equal to teo. Above the CMC and taking account of Equation (2.4), retention time should increase nonlinearly upon increasing total surfactant concentration ct. Nevertheless, the experimental data were fitted to straight lines near to the CMC value (Figure 2.4) and CMC was determined at the intersection point. This method does not need calculation of electroosmotic flow or micelle mobility, but they normally should remain constant. Its suitability was demonstrated for several anionic, UV-transparent surfactants in the 1–35 mM CMC range, with RSD value less than 4 % (n ¼ 5) for SDS [36]. It is to note that, similarly, Nagamine et al. [44] plotted (tR teo)/teo (improperly called relative mobility) instead of tR, as a function of ct in case of cationic surfactants and determined CMCs from linear regression.
DETERMINATION OF CRITICAL MICELLE CONCENTRATIONS
39
Migration time (min)
20
15
10
5 00
10
20
Concentration (mM)
Figure 2.4 Experimental dependence of migration time on total SDS concentration for various analytes: toluene (*), naphthalenemethanol (*), naphthol (~). BGE: 20 mM phosphate buffer, pH 7. Temperature: 40 C. (Reprinted from [36] by permission of The Japan Society for Analytical Chemistry)
2.3.3
Mobility Method
This method was initially devised by Jacquier and Desbe`ne [35] to take into account the solvophobic interactions between an anionic surfactant monomer and a neutral micelle-interacting analyte. More generally, the effective mobility meff of the analyte is given by: meff ¼
Ksolv ct m 1 þ Ksolv ct solv
for ct < CMC
½35; 39; 45; 46
ð2:7Þ
and meff ¼
Ksolv CMC m 1 þ Ksolv CMC þ Kmic ðct CMCÞ=n solv
þ
Kmic ðct CMCÞ=n m 1 þ Ksolv CMC þ Kmic ðct CMCÞ=n mic
for ct > CMC
½45; 46
ð2:8Þ
where n is the aggregation number, Ksolv and Kmic are the 1:1 association constants between the analyte and the surfactant monomer, and between the analyte and the micelle, respectively. msolv and mmic are the mobilities of the analyte–monomer complex and of the micelles, respectively. Figure 2.5 shows the experimental variation of the mobility of some neutral analytes as a function of total surfactant concentration. The flat part on the left of the curve indicates that for these analytes, solvophobic analyte–monomer interactions are negligible. Besides, the first break on the mobility curve is representative of the onset of surfactant aggregation, known as critical aggregation concentration (CAC). Accurate CMC determination can be achieved at the intersection point of the nonlinear fittings of the experimental points apart from the micellization regions using Equations (2.7) and (2.8) [46]. Simplified linear fittings were used as well [35], approximating
40
ELECTROKINETIC CHROMATOGRAPHY
Figure 2.5 Experimental variation of the effective mobility of pyridine derivatives as a function of total SDS concentration. BGE: 20 mM phosphate buffer, pH 7. Temperature: 25 C. Reprinted from Lin et al., Capillary electrophoresis study on the micellization and critical micelle concentration of sodium dodecyl sulfate. (Influence of solubilized solutes. J. Chromatogr. A., 924, 83–91, copyright 2001, with permission from Elsevier)
Equations (2.7) and (2.8) to: meff ¼ Ksolv ct msolv þ Kmic
ðct CMCÞ mmic n
ð2:9Þ
The larger the difference between Ksolv and Kmic is, the more precise the CMC determination will be [39]. In the special case where Ksolv Kmic, CMC is derived as the surfactant concentration leading to extrapolated zero electrophoretic mobility [6]. The mobility method does not need the measurement of micelle mobility. It does, however, require electroosmotic mobilities for the calculation of effective mobilities. CMC values obtained by this method for SDS in various BGE conditions and for some cationic surfactants are reported in [39], Table 2.1. This method has also proved useful for the CMC determination of some anionic surfactants in formamide [6]. 2.3.4
Zeta Potential Method
For the sake of completeness, it should be mentioned that zeta potential measurements have also been used to study surfactant aggregation [17]. This parameter, which has been mainly considered in the field of colloidal particles [47], can be related to electrophoretic mobility through the modified Smoluchowski equation: ¼
3 Z meff 2 ef ðkaÞ
ð2:10Þ
where Z is the electrolyte viscosity, e the dielectric constant, k the reverse Debye length, a the radius of the particle and f (ka) the corrective Henry function. In spite of their close
DETERMINATION OF CRITICAL MICELLE CONCENTRATIONS
41
Figure 2.6 Experimental variation of zeta potential in terms of total dodecyltrimethylammonium bromide concentration for various KBr concentrations: (*) 0 mM; (&) 2 mM; (~) 4 mM; () 6 mM. Temperature: 25 C. (Reprinted from [17], by permission of Wiley-VCH Verlag)
connection to electrophoretic mobility, zeta potentials are usually derived from a distinct experimental electrokinetic implementation, referred to as laser-Doppler microelectrophoresis. Briefly, such a commercially available setup involves the migration of the particles to be studied in a short closed capillary channel of cylindrical or rectangular crosssection under alternating high electric field strength (>100 V/cm) of relatively low frequency (typically 1 Hz), to suppress electrode polarization. Light scattering detection from a laser source optical arrangement is also usually provided. A more detailed description of the experimental setup can be found in Reference [47]. Particle mobility meff and size a can be simultaneously extracted by analysing the light scattered with a digital correlator. In effect, the Lorentzian-shaped power spectrum of scattered light undergoes a shift in frequency that is proportional to the particle migration velocity (Doppler effect), while its width at half height is proportional to the particle diffusion coefficient. As illustrated in Figure 2.6, depicts a bell-shaped variation in the concentration range of surfactant aggregation, but the maximum of the curve was reported to be significantly lower (10–20 %) than the CMC values obtained by classical conductimetry [17]. It might rather indicate the onset of counterion condensation. Nevertheless, this method allows clear detection of premicellar association of ionic surfactants in low conductivity media. 2.3.5
Discussion on Operating Conditions
The impact of solubilized solutes on micellization has been emphasized by several groups [39,43,46]. For the best accuracy, the micelle-interacting marker should not therefore be ionizable, for a proper control of ionic strength [39,40] and to avoid anomalous pH effects [48]. It should not interact with the micelles too strongly, otherwise micellization will be favored. It should also have a high molar UV absorptivity (to be used at low concentration), be soluble in water to a certain extent (not to introduce too high a content of organic solvent ( 10 % v/v) with the markers). A lack of solubility in water will probably result in a shift in the micellization equilibrium and
42
ELECTROKINETIC CHROMATOGRAPHY
therefore in systematic errors in CMC values [35,43,46]. Likewise, too high a content of organic solvent in the marker solution will induce disruption of micelles [12,39,49]. Nevertheless, the baseline perturbation generated by the presence of a slight organic solvent content in the injected marker mixture can be used in practice for the measurement of teo. Although questioned by Mrestani et al. [41], the influence of the marker concentration was studied by Lin and Lin [40]. Based on these criteria, 2naphthalenemethanol was the best choice according to Nakamura et al. [36], whereas naphthalene was selected by Jacquier and Desbe`ne in spite of its low water solubility for the case of SDS [35], and anionic markers (cephalosporins) were used successfully for cationic surfactants [40]. Other recommended markers can be found in Lin’s review [39]. For the micelle mobility marker, hydrophobic dyes such as Sudan III [34,42,43], polyaromatic hydrocarbons [50], alkyl phenyl ketones [44,51] or hydrophobic compounds bearing a charge of sign opposite to that of the micelles [52] have been advocated. The nature of the capillary to be used may depend on the type (anionic, cationic) of surfactant. Bare silica capillaries are perfectly adapted to anionic surfactants as electroosmotic flow remains virtually independent of surfactant concentration. As cationic surfactants interact with the inner capillary surface and induce reversal of the electroosmotic flow inside these capillaries [53,54], the choice of the more appropriate capillary has been more controversial in the latter case [39]. Nagamine and Nakamura [44] recommended the use of (3-aminopropyl)triethoxysilane-treated capillaries to produce stable anodic electroosmotic flows, whereas use of bare silica capillaries was not reported to be problematic, once electroosmotic flow reversal has been established [40,45]. Finally, as for the current intensity method, proper temperature control is essential.
2.4 Frontal Methods As recently introduced by Gao et al. for the determination of interaction parameters [55], frontal analysis continuous capillary electrophoresis (FACCE) consists in first equilibrating a capillary with the BGE of interest and then continuously introducing a multicomponent sample dissolved in this electrolyte into the capillary under the influence of an electric field. This electrokinetic process achieves partial separation of the sample components simultaneously with their injection. Resulting electropherograms present successive migration fronts in order of decreasing apparent mobilities, with heights of the corresponding plateaus related to the concentrations of the components. Applied to an interacting system, this process avoids disturbance of the equilibria involved (similarly to dialysis or ultrafiltration), is free from kinetic constraints and allows direct measurement of the concentration of the faster migrating partner from a simple calibration step. As far as CMC determination is concerned, FACCE was first applied to UV-absorbing, anionic surfactants by Le Saux et al. [37]. Figure 2.7(a) shows that, depending on the total surfactant concentration, one or several migration fronts pertaining to surfactant monomer and aggregates can be detected. CMC is thus determined in this case without any marker being needed, as the total surfactant concentration for which the height of the monomer plateau remains almost constant [Figure 2.7(b)]. Importantly, the experimental
DETERMINATION OF CRITICAL MICELLE CONCENTRATIONS (a) mAU
F2
P2
43
F3 P3
P1 F1
600
400 EOF 200 0 1
3
5
min
(b)
Height of plateau P1 (mAU)
450
300
150
0
0
6
12
18
Concentration (mM)
Figure 2.7 Frontal analysis continuous capillary electrophoresis (FACCE) of sodium octylbenzenesulfonate. (a) Electropherograms obtained at various concentrations: from bottom to top: 7, 9, 11 and 13 mM. Bare silica capillary: 50 mm i.d. 35 cm (detection, 26.5 cm). BGE: 20 mM sodium borate buffer, pH 9.2 (ionic strength: 10 mM). Temperature: 30 C. UV absorbance detection at 200 nm. Injection: benzyl alcohol (30 mbar, 2 s), BGE (30 mbar, 2 s), followed by continuous electrokinetic introduction of surfactant sample under þ10 kV. EOF: electroosmotic flow. F1, F2, F3: migration fronts of monomeric, micellar and oligomeric forms, respectively. P1, P2, P3: corresponding concentration plateaus. (b) Height of the concentration plateau P1 of the free surfactant (at 230 nm) as a function of total surfactant concentration. Solid line: least-squares regression straight lines, yielding CMC = 9.5 0.8 mM (95 % confidence level) at intersection point. (Reprinted from Le Saux et al., Determination of aggregation thresholds of UV-absorbing anionic surfactants by frontal analysis continuous capillary electrophoresis, J. Chromatogr. A., 1038, 275–282, copyright 2004, with permission from Elsevier)
setup should be devised so that the discrete plateau of surfactant monomer will be detected first, thus allowing its easy quantitation. For the case of UV-transparent surfactants, a preferably UV-absorbing, micelleinteracting marker M is required [38], as for zonal methods, and FACCE allows direct determination of free marker concentration [M] from the measurement of corresponding
44
ELECTROKINETIC CHROMATOGRAPHY 1.1 A
B
0.8 [M] / [M]t
1 0.96 0.92 0.88 3
5
7
9
0.5
0.2 0
2
4
6 8 10 12 14 Total surfactant concentration (mM)
16
Figure 2.8 Theoretical variation of the free to total marker concentration ratio [M]/[Mt] as a function of the total surfactant concentration, according to Equations (2.11) and (2.12), with (A) Ksolv ¼ 0.5 M1 and Kmic ¼ 150 M1; (B) Ksolv ¼ 5 M1 and Kmic ¼ 1500 M1 and (C) Ksolv ¼ 50 M1 and Kmic ¼ 15 000 M1; CMC ¼ 6 mM; n ¼ 60. (Reprinted from Le Saux et al., Determination of aggregation thresholds of non-UV-absorbing neutral or charged surfactants by frontal- and vacancy-frontal analysis continuous capillary electrophoresis, J. Chromatogr. A., 1041, 219–226, copyright 2004, with permission form Elsevier)
plateau height. When the total surfactant concentration ct is below CMC, the marker can interact only with monomeric surfactant and [M] is given by [38]: ½M ¼
½M t 1 þ Ksolv ct
ð2:11Þ
When ct is greater than CMC, the expression for [M] becomes: ½M ¼
1 þ Ksolv
½M t CMC þ Kmic ðct CMCÞ=n
ð2:12Þ
Figure 2.8 represents the theoretical variations of the fraction of free marker [M]/[M]t for various Ksolv and Kmic values, according to Equations (2.11) and (2.12). The CMC can therefore be determined from the curve discontinuity. It can also be noticed from Figure 2.8 that the most appropriate marker should be of intermediate hydrophobicity, to allow simple linear fitting of the data points. The suitability of this approach was established for SDS in a pH 6.9 phosphate buffer, using a bare silica capillary and a neutral compound, ethylparaben, as a micelle-interacting marker. Under these conditions, the migration front of the free marker, monitored by UV-absorbance, appears before those of the marker–surfactant complexes, as it is solely moved in the capillary by the cathodic electroosmotic flow. The height of the subsequent plateau could hence be easily quantified, yielding a CMC value in very close agreement with literature data [38]. The FACCE method also applies to the CMC determination of UV-transparent cationic
DETERMINATION OF CRITICAL MICELLE CONCENTRATIONS
45
surfactants, as exemplified in Figure 2.9 [38]. In this latter case, an amine-coated capillary and again a neutral marker should be used, to minimize electrostatic interactions with the capillary wall, and electroosmotic flow is reversed. Then, introducing the surfactant/marker mixture by the short end of the capillary (with respect to the position of the detection window) to further suppress wall adsorption effects, and still applying a positive voltage at the conventional capillary inlet, again results in the migration front of the free marker being detected first, the marker-surfactant complexes migrating counterelectroosmotically. This experimental setup is quite symmetrical to that used for anionic surfactants with bare silica capillaries. Finally, the interest in FACCE has also been demonstrated for CMC determination of neutral, UV-transparent surfactants [38]. In this case, an anionic, UV-absorbing amphiphilic marker with a bare silica capillary can be selected. In low ionic strength media, however, electroosmotic flow then drags all the species of interest to the cathodic side of the capillary, and classical implementation of FACCE protocol results in the front of the free anionic marker being detected last. As this situation is not especially favorable for precise quantitation of the corresponding plateau height, due to occurrence of wall adsorption and of noisy, high absorbance levels, a vacancy FACCE protocol should preferably be used. This protocol consists in first filling the capillary with surfactant/ marker mixtures dissolved in the electrolyte of interest, and then continuously injecting this electrolyte alone under an electric field. This implementation results in easier detection of the decreasing migration front of the free anionic marker from a lowabsorbance baseline. Figure 2.10(a) shows the vacancy frontal electropherograms obtained for different concentrations of Brij 35 (polyoxyethylene-23-dodecanol) and a constant concentration of octylbenzenesulfonate, used as a marker. When plotted against the total Brij 35 concentration, the height of the last descending migration front, pertaining to the free form of this anionic marker, exhibits two linear variation ranges [Figure 2.10(b)]. The CMC value determined from the break between these two parts (0.089 mM) was in excellent agreement with the literature value [56].
2.5 Discussion of the Suitability of Electrophoretic and Conventional Methods As compared with the armory of conventional methods, electrokinetic methods based on the use of CE instrumentation appear to be very suitable for the quantitative characterization of surfactant aggregation, both from theoretical and practical viewpoints. Electrokinetic methods are likely more universal than classical ones (especially conductimetry and dye solubilization), with respect to the types of surfactants that can be dealt with, in that they apply to UV- or non-UV-absorbing anionic, cationic and neutral surfactants [36,38,39,45]. The case of mixed surfactants has also been addressed successfully [37]. Although not yet reported [39], they should apply to zwitterionic surfactants as well. For most methods, premicellar aggregation phenomena can be vizualized. Electrokinetic methods offer a large compatibility in medium composition (pH, ionic strength, solvents, additives) [12] and temperature range [57]. The remarkable agreement between CMC values obtained from different CE and conventional methods and their consistency with respect to ionic strength (Table 2.1) lend support to the high
46
ELECTROKINETIC CHROMATOGRAPHY (a) mAU 25 20
F2 EOF / F1
15
F3
10 5 0 EOF / F1
–5 –10 2
4
6
8
10
12
14
min
(b) 18 y = 15.7 n=8
Height of migration front (mAU)
17 16 15
y = – 4.24x + 25.14 R 2 = 0.9989 n=5
14 13 12 11 10 9 8 0
0.5
1 1.5 2 2.5 3 Total TTABr concentration (mM)
3.5
4
Figure 2.9 Frontal analysis continuous capillary electrophoresis (FACCE) of tetradecyltrimethylammonium bromide/ethylparaben mixtures. (a) Electropherograms obtained for constant (0.33 mM) ethylparaben concentration and varying surfactant concentrations in water: from left to right: 0, 0.75, 1.25, 1.50, 2, 3, and 3.50 mM. Amine-coated silica capillary: 50 mm i.d. 62.5 cm (detection, 8.5 cm). BGE: 7.3 mM sodium phosphate buffer, pH 6.4 (ionic strength: 10 mM). Temperature: 25 C. UV absorbance detection at 256 nm. Shortend, continuous, electrokinetic injection of the surfactant/marker mixtures under þ15 kV. EOF: electroosmotic flow. F1, F2, F3: migration fronts of free marker, marker-monomeric surfactant and marker-micelle complexes, respectively. (b) Height of migration front F1 of the free marker as a function of total surfactant concentration. Solid line: least-squares regression straight lines, yielding CMC ¼ 2.25 mM at intersection point. (Reprinted from Le Saux et al., Determination of aggregation thresholds of non-UV-absorbing neutral or charged surfactants by frontal- and vacancy-frontal analysis continuous capillary electrophoresis, J. Chromatogr. A., 1041, 219–226, copyright 2004, with permission from Elsevier)
DETERMINATION OF CRITICAL MICELLE CONCENTRATIONS
47
(a) EOF F2
mAU
F1
15 EOF 10
5
0 0
1
2
3
4
min
(b) 13.6 Height of migration front (mAU)
13.5 y = –9.25x + 14.17 R 2 = 0.996 n=4
13.4 13.3 y = –2.50x + 13.58 R2 = 1 n=4
13.2 13.1 13 12.9 12.8 12.7 0
0.02
0.04
0.06 0.08 0.1 0.12 Total Brij 35 concentration (mM)
0.14
0.16
Figure 2.10 Vacancy frontal analysis continuous capillary electrophoresis (VFACCE) of Brij 35/sodium octylbenzenesulfonate (SOBS) mixtures. (a) Electropherograms obtained for constant (0.064 mM) SOBS concentration and varying Brij 35 concentrations in the BGE: from bottom to top: 0.05, 0.07, 0.09, 0.13, 0.15 mM. Bare silica capillary: 50 mm i.d. 35 cm (detection, 26.5 cm). BGE: 20 mM sodium borate buffer, pH 9.2 (ionic strength: 10 mM). Temperature: 25 C. UV absorbance detection at 193 nm. Initial capillary filling with the Brij 35/SOBS mixtures followed by continuous, electrokinetic injection of BGE under þ10 kV. EOF: electroosmotic flow. F1, F2: migration fronts of free marker (SOBS), and mixed Brij 35/ SOBS micelles, respectively. (b) Height of migration front F1 of the free marker (SOBS) as a function of total Brij 35 concentration. Solid line: least-squares regression straight lines, yielding CMC = 0.089 mM at intersection point. (Reprinted from Le Saux et al., Determination of aggregation thresholds of non-UV-absorbing neutral or charged surfactants by frontal- and vacancy-frontal analysis continuous capillary electrophoresis, J. Chromatogr. A., 1041, 219–226, copyright 2004, with permission from Elsevier)
Current intensity Mobility
Water, 25 C
Extrapolated to zero ionic strength, 25 C Borax trisodium phosphate sodium chloride 2.8 mM sodium phosphate buffer, pH 6.9 (5 mM IS), 25 C 20 mM sodium borate buffer, pH 9.2 (10 mM IS), 25 C 10 mM sodium phosphate buffer, pH 7.0 (20 mM IS), 25 C
Sodium dodecyl sulfate
35 C 30 mM NaCl, 25 C
Retention time (2-naphthalenemethanol)
20 mM phosphate buffer, pH 7, 40 C
Sodium decyl sulfonate
7.8 7.9 6.1 5.3
3.5 6.1 4.8 3.0 3.6
Frontal (ethylparaben) Mobility (naphthalene)
Current intensity Mobility (pyridine) (2,3,5-trichloropyridine) Current intensity Current intensity
8.3
34
30
(naphthalene) (naphthalene)
Retention time (2-naphthalenemethanol)
20 mM phosphate buffer, pH 7, 40 C
CMC (mM)
Sodium decyl sulfate
Method
Conditions
Surfactant
CE techniques
56 46 46 56 33
35
38
12 12
33
36
36
Ref.
Mobility
Surface tension
Mobility (naphthalene)
Absorbance (Sudan III) conductometry Absorbance (Sudan III) conductometry
Method
5
34.3 8.1–8.4
2.9–3.7
5.3
6.9
5, 12
58
12
58
36
26.9 32.9
8.1
36
Ref. 25.0
CMC (mM)
Reference techniques
Table 2.1 Comparison of CMC values obtained form CE techniques and from conventional techniques. Abbreviations used: DoTMAB: dodecyltrimethylammonium bromide; DoTMAC: dodecyltrimethylammonium chloride; TTMAB: tetradecyltrimethylammonium bromide; CTMAB: cetyltrimethylammonium bromide; CTMAC: cetyltrimethylammonium chloride. IS: ionic strength
20 mM sodium phosphate buffer, pH 7.0, 25 C
DoTMAB
Sodium cholate
20 mM phosphate buffer, pH 7, 40 C Water, 25 C KBr 6 mM, 25 C 20 mM phosphate buffer, pH 7.5 (52 mM IS), 25 C 70 mM phosphate buffer, pH 6 (86 mM IS), 25 C
Sodium alkylbenzene-sulfonates: Octyl20 mM sodium borate buffer, pH 9.2, 30 C Dodecyl-
Sodium tetradecyl sulfate Sodium alkenyl sulfates: OctenylNonenylDecenylUndecenyl-
Retention time (2-naphthalenemethanol) Zeta potential Zeta potential Mobility (cephalosporins) Mobility (propazine)
Frontal
Current intensity
20 mM sodium Mobility phosphate buffer, pH 7.0 (pyridine) (41 mM IS) 25 C (2,3,5-trichloropyridine) Retention time 40 C (2-naphthalenemethanol) Current intensity 80 mM sodium borate buffer, pH 9.2 (40 mM IS), 25 C Retention factor 24 mM phosphate buffer, pH 7.0 (50 mM IS), 25 C (catechol) (methylcatechol) 20 mM phosphate buffer, Retention time pH 7, 40 C (2-naphthalenemethanol) 33
3.1
13.4 10.4 12.5 12.4 11 0.1
12.8
2.30 0.09
9.5 0.8
255 119 61 33
17 17 40 40 45
36
37
26
36
43
36
4.4 3.9
4.3 4.0 0.87
46
4.8
Conductivity Conductivity Conductivity
Conductivity Surface tension Conductivity Surface tension Absorbance (Sudan III)
Fluorescence (pyrene)
Absorbance (Sudan III) conductometry
Absorbance (sudan III) conductometry Retention factor, Mobility
15.5 13.9 13.0
17 17 40
36
37
26
36
36 36 59 12
(Continued)
9.8 0.3 7.5–10 unsuitable unsuitable 12.6
252 133 63 31
0.86 0.89
3.5 4.0 2.6–3.2
Water, 25 C 20 mM Britton–Robinson buffer, pH 3, 25 C Water, 25 C 7.3 mM sodium phosphate buffer, pH 6.4 (10 mM IS), 25 C 70 mM sodium phosphate buffer, pH 6 (86 mM IS), 25 C Water, 25 C 20 mM Britton-Robinson buffer, pH 3, 25 C 20 mM Britton-Robinson buffer, pH 3, 25 C 20 mM sodium borate, pH 9.2, 25 C
DoTMAC
Brij 35
Cetylpyridinium
CTMAB CTMAC
TTMAB
Conditions
Surfactant
Table 2.1 (Continued)
0.93 0.75 0.57
Current intensity Mobility (acetophenone) Mobility (acetophenone) 0.0089
1.6 0.1
Mobility (propazine)
Vacancy frontal
2.25
18.8
CMC (mM)
Frontal analysis (ethylparaben)
Method
CE techniques
38
44
33 44
45
38
44
Ref.
Absorbance (Sudan III)
Absorbance (Sudan III)
Conductivity
Absorbance (Sudan III)
Method
0.09
0.54
0.90–0.98 0.67
56
44
5 44
45
56
3.5
1.5
54 44
Ref. 20 17.4
CMC (mM)
Reference techniques
DETERMINATION OF CRITICAL MICELLE CONCENTRATIONS
51
level of accuracy provided by electrokinetic methods. Although performances of CE and conventional methods have been rarely discussed in terms of precision, these are quite similar and mainly depend on the number and the quality of the data points and the surfactant nature [37]. From a practical point of view, CE techniques take advantage of their miniaturized implementation and small volume requirement (a few hundreds of microliters of each varying concentration of surfactant solution) [36], good temperature control [33], and have proven superior to other techniques as regards time needed (especially, as compared to dye solubilization [24]) and ease of implementation, due to instrument automation. Among the CE methods, those involving a marker may lead to inaccurate determinations (lower CMC values insofar as micellization can be favored [33, 35]) if marker nature and concentration are not properly chosen, as emphasized before. This makes the current intensity method particularly attractive for ionic surfactants, especially in low conductivity media. Indeed, detection of the intersection point between the two straight lines becomes imprecise when background conductivity is high [33]. Another drawback of this method is that the slopes of the straight lines may not vary abruptly in the case of premicellar aggregation [39]. When BGE conductivity is markedly higher than that brought by the surfactant in the range of its CMC, the mobility method becomes preferable [39]. Finally, newly developed frontal methods appear to be very effective for fast, precise and accurate determinations of aggregation thresholds of any kind of surfactant, including, for the first time, neutral ones. One fundamental advantage of frontal methods over zonal ones comes from the fact that the surfactant monomer–micelle and surfactant–marker equilibria are perturbed to a minor extent. The choice for the markers can also be enlarged towards compounds of lower hydrophobicity, a favorable situation with respect to marker solubility, detection sensitivity and hence accuracy. On the other hand, frontal methods suffer from the fact that plateau heights can only be measured manually. More specifically, the frontal approach is best suited for UV-absorbing ionic surfactants, for which no marker is needed and zonal methods may present detection difficulties, and for aggregation studies of amphiphilic polymers, which may present slow equilibrium kinetics. Lastly, it is worthy of note that differences between results afforded by the various methods can be increased for polydisperse surfactants, so that complementary insights can be obtained in this latter case.
References [1] L.J. Cline Love, J.G. Harbata and J.G. Dorsey. The micelle–analytical chemistry interface, Anal. Chem., 56, 1132A–1146A (1984). [2] E. Pramauro and E. Pelizzetti. Surfactants in Analytical Chemistry, Elsevier, Amsterdam, 1996. [3] C. Tanford. The Hydrophobic Effect: Formation of Micelles and Biological Membranes, John Wiley & Sons, Inc. New York, 1980. [4] H. Hoffman and W. Ulbricht, in Thermodynamic Data for Biochemistry and Biotechnology, H.J. Hinz (Ed), pp. 297–348, Springer Verlag, New York, 1986.
52
ELECTROKINETIC CHROMATOGRAPHY
[5] P. Mukerjee and K.J. Mysels. Critical Micelle Concentrations of Aqueous Surfactants Systems, National Standards Reference Data Service, National Bureau of Standards (US) Washington, DC, 1971. [6] J.-M. Lin, M. Nakagawa, K. Uchiyama and T. Hobo. Determination of critical micelle concentration of SDS in formamide by capillary electrophoresis, Chromatographia, 50, 739–744 (1999). [7] D. Myers. Surfactant Science and Technology, VCH, New York, 1998. [8] B.C. Paul, S.S. Islam and K. Ismail. Effect of acetate and propionate co-ions on the micellization of sodium dodecyl sulfate in water, J. Phys. Chem., B, 102 (40), 7807–7812 (1998). [9] E.S. Ahuja and J.P. Foley. Influence of dodecyl sulfate counterion on efficiency, selectivity, retention, elution range, and resolution in micellar electrokinetic chromatography, Anal. Chem., 67, 2315–2324 (1995). [10] E. Junquera, G. Tardajos and E. Aicart. Effect of the presence of beta-cyclodextrin on the micellization process of sodium dodecyl sulfate or sodium perfluorooctanoate in water, Langmuir, 9, 1213–1219 (1993). [11] S. Terabe, T. Katsura, Y. Okada, Y. Ishihama and T. Otsuka. Measurement of thermodynamic quantities of micellar solubilization by micellar electrokinetic capillary chromatography with SDS, J. Microcol. Sep., 5, 23–33 (1993). [12] J.C. Jacquier and P.L. Desbe`ne. Determination of critical micelle concentration by CE. Application to organo-saline electrolytes, J. Chromatogr. A, 743, 307–314 (1996). [13] D.O. Shah. Micelles, Microemulsions and Monolayers, Sciences and Technology, Marcel Dekker, New York, 1998. [14] J. Garcia-Anto and J.L. Guinon. Determination of Hyamine 2389 critical micelle concentration by means of conductometric, spectrophotometric and polarographic methods, Colloids Surf., 61, 137–145 (1991). [15] M. Dominguez, A. Fernandez, N. Gonzalez, E. Iglesias and L. Montenegro. Determination of critical micelle concentration of some surfactants by three techniques, J. Chem. Educ., 74, 1227–1230 (1997). [16] M. Perez-Rodriguez, G. Prieto, C. Rega, L.M. Varela, F. Sarmiento and V. Mosquera. A comparative study of the determination of the critical micelle concentration by conductivity and dielectric constant measurements, Langmuir, 14, 4422–4426 (1998). [17] R. Sabate´ , M. Gallardo and J. Estelrich. Electrophoretic properties of dodecyltrimethylammonium bromide micelles in KBr solutions, Electrophoresis, 21, 481–485 (2000). [18] M. Zulauf, U. Furstenberger, M. Grabo, P. Jaggi, M. Regeass and J.P. Rosenbusch, in Methods in Enzymology, S. Fleischer and B. Fleischer (Eds.), Vol. 172, p. 528, Academic Press, San Diego, 1989. [19] Y.D. Smet, L. Deriemaeker, E. Parloo and R. Finsy. On the determination of Ostwald ripening rates from dynamic light scattering measurements, Langmuir, 15, 2327–2332 (1999). [20] S.A.A. Rizvi and S.A. Shamsi. Polymeric alkenoxy amino acid surfactants: I. Highly selective class of molecular micelles for chiral separation of beta-blockers, Electrophoresis, 24, 2514– 2526 (2003). [21] T. Nakashima, T. Anno, H. Kanda, Y. Sato, T. Kuroi, H. Fujii, S. Nagadome and G. Sugihara. Potentiometric study on critical micellization concentrations (CMC) of sodium salts of bile acids and their amino acid derivatives, Colloids Surf. B: Biointerfaces, 24, 103–110 (2003). [22] A.B. Mandal and B.U. Nair. Cyclic voltammetry technique for the determination of the critical micellar concentration of surfactants, self-diffusion coefficient of micelles, and partition coefficient of an electrochemical probe, J. Phys. Chem., 65, 9008–9013 (1991). [23] H. Schott. Solubilization of a water-insoluble dye as a method for determining micellar molecular weights, J. Phys. Chem., 70, 2966–2973 (1996).
DETERMINATION OF CRITICAL MICELLE CONCENTRATIONS
53
[24] C. Maeder, G.M.J. Beaudoin III, E.K. Hsu, V.A. Escobar, S.M. Chambers, W.E. Kurtin and M.M. Bushey. Measurement of bilirubin partition coefficients in bile salt micelle/aqueous buffer solutions by micellar electrokinetic chromatography, Electrophoresis, 21, 706–714 (2000). [25] N.J. Turro and A. Yekto. Luminescent probes for detergent solutions. A simple procedure for determination of the mean aggregation number of micelles, J. Am. Chem. Soc., 100, 5951–5952 (1978). [26] C. Akbay and S.A. Shamsi. Polymeric sulfated surfactants with varied hydrocarbon tail: I. Synthesis, characterization and application in micellar electrokinetic chromatography, Electrophoresis, 25, 622–634 (2004). [27] K.J. Mysels and L.H. Princen. Light scattering by some lauryl sulfate solutions, J. Phys. Chem., 63, 1696–1700 (1959). [28] N.C. Santos, A.C. Silva, M.A. R.B. Castanho, J. Martins-Silva and C. Saldanha. Evaluation of lipopolysaccharide aggregation by light scattering spectroscopy, Chembiochem, 4, 96–100 (2003). [29] H.G. Elias. The study of association and aggregation via light scattering, in Light Scattering from Polymer Solutions, M.B. Huglin (Ed.), pp. 397–457, Academic Press, London, 1972. [30] E.D. Goddard. Polymer–surfactant interaction. Part I: uncharged water-soluble polymers and charged surfactants, Colloids Surf., 19, 255–300 (1986). [31] Y.S. Lee and K.W. Woo. Micellization of aqueous cationic surfactant solutions at the micellar structure transition concentration – based upon the concept of the pseudophase separation, J. Colloid Interface Sci., 169, 34–38 (1995). [32] D.C. Tickle, G.N. Okafo, P. Camilleri, R.F.D. Jones and A.J. Kirby. Glucopyranoside based surfactants as pseudostationary phases for chiral separations in capillary electrophoresis, Anal. Chem., 66, 4121–4126 (1994). [33] A. Cifuentes, J.L. Bernal and J.C. Diez-Masa. Determination of critical micelle concentration values using capillary electrophoresis instrumentation, Anal. Chem., 69, 4271–4274 (1997). [34] S. Terabe, K. Otsuka and T. Ando. Electrokinetic chromatography with micellar solution and open-tubular capillary, Anal. Chem., 57, 834–841 (1985). [35] J.C. Jacquier and P.L. Desbe`ne. Determination of critical micelle concentration by capillary electrophoresis. Theoretical approach and validation, J. Chromatogr. A, 718, 167–175 (1995). [36] H. Nakamura, A. Sano and K. Matsuura. Determination of critical micelle concentration of anionic surfactants by capillary electrophoresis using 2-naphthalenemethanol as a marker for micelle formation, Anal. Sci., 14, 379–382 (1998). [37] Th. Le Saux, A. Varenne and P. Gareil. Determination of aggregation thresholds of UV-absorbing anionic surfactants by frontal analysis continuous capillary electrophoresis, J. Chromatogr. A., 1038, 275–282 (2004). [38] Th. Le Saux, A. Varenne and P. Gareil. Determination of aggregation thresholds of non-UVabsorbing, neutral or charged surfactants by frontal- and vacancy-frontal analysis continuous capillary electrophoresis, J. Chromatogr. A., 1041, 219–226 (2004). [39] C.-E. Lin. Determination of critical micelle concentration of surfactants by capillary electrophoresis, J. Chromatogr. A., 1037, 467–478 (2004). [40] C.E. Lin and K.S. Lin. Determination of critical micelle concentration and interactions between cephalosporins and charged surfactants, J. Chromatogr. A, 868, 313–316 (2000). [41] Y. Mrestani, R. Neubert and H.H. Ru¨ ttinger. Reply to ‘Determination of critical micelle concentration and interactions between cephalosporins and charged surfactants’, J. Chromatogr. A, 868, 317–319 (2000). [42] M.G. Khaledi, S.C. Smith and J.K. Strasters. Micellar electrokinetic capillary chromatography of acidic solutes: migration behavior and optimization strategies, Anal. Chem., 63, 1820–1830 (1991).
54
ELECTROKINETIC CHROMATOGRAPHY
[43] J.K. Strasters and M.G. Khaledi. Migration behavior of cationic solutes in micellar electrokinetic capillary chromatography, Anal. Chem., 63, 2503–2508 (1991). [44] N. Nagamine and H. Nakamura. Measurements of critical micelle concentrations of cationic surfactants by capillary electrophoresis, Anal. Sci., 14, 405–406 (1998). [45] C.E. Lin, T.Z. Wang, T.C. Chiu and C.C. Hsueh. Determination of the critical micelle concentration of cationic surfactants by capillary electrophoresis, J. High Resolut. Chromatogr., 22, 265–270 (1999). [46] C.E. Lin, M.J. Chen, H.C. Huang and H.W. Chen. Capillary electrophoresis study on the micellization and critical micelle concentration of sodium dodecyl sulfate. Influence of solubilized solutes, J. Chromatogr. A, 924, 83–91 (2001). [47] R.J. Hunter. Foundations of Colloid Science, second edition, Oxford University Press, Oxford, 2002. [48] C.E. Lin, W.C. Lin and W.C. Chiou. Migration behaviour and selectivity of dichlorophenols in micellar electrokinetic capillary chromatography. Influence of micelle concentration and buffer pH, J. Chromatogr. A., 722, 333–343 (1996). [49] C.-E. Lin, H.-C. Huang and H.-W. Chen. A capillary electrophoresis study of the influence of cyclodextrin on the CMC of SDS, J. Chromatogr. A, 917, 297–310 (2001). [50] P. Morin and M. Dreux. Factors influencing the separation of ionic and non-ionic chemical natural compounds in plant extracts by capillary electrophoresis, J. Liquid Chromatogr., 16, 3735–3755 (1993). [51] E.S. Ahuja, E.L. Little, K.R. Nielsen and J.P. Foley. Infinite elution range in micellar electrokinetic capillary chromatography using a nonionic anionic mixed micellar system, Anal. Chem., 67, 26–33 (1995). [52] J.R. Mazzeo, E.R. Grover, M.E. Swartz and J.S. Petersen. Novel chiral surfactant for the separation of enantiomers by micellar electrokinetic capillary chromatography, J. Chromatogr., 680, 125–135 (1994). [53] T. Kaneta, S. Tanaka and M. Taga. Effect of cetyltrimethylammonium chloride on electroosmotic and electrophoretic mobilities in capillary zone electrophoresis, J. Chromatogr., 653, 313–319 (1993). [54] D. Crosby and Z.E. Rassi. Micellar electrokinetic capillary chromatography with cationic surfactants, J. Liquid Chromatogr., 16, 2161–2187 (1993). [55] J.Y. Gao, P.L. Dubin and B.B. Muhoberac. Measurement of the binding of proteins to polyelectrolytes by frontal analysis continuous capillary electrophoresis, Anal. Chem., 69, 2945–2951 (1997). [56] H. Nishi and S. Terabe. Micellar electrokinetic chromatography. Perspectives in drug analysis, J. Chromatogr. A, 735, 3–27 (1996). [57] Y. Mrestani, Z. Marestani and R.H.H. Neubert. Characterization of micellar solubilization of antibiotics using micellar electrokinetic chromatography, J. Pharm. Biomed. Anal., 26, 883– 889 (2001). [58] D. Stigter. Physical Chemistry, Enriching Topics for Colloids and Surface Science, H. Van Olphen. and K.J. Mysels, (Eds), Chapter 12, IUPAC Commission I.6, Theorex, La Jolla, CA, 1975. [59] K. Saitoh, C. Kiyohara and N. Suzuki. Mobilities of metal-diketonato complexes in micellar electrokinectic capillary chromatography, J. High Resolut. Chromatogr., 14, 245–248 (1991).
3 Selectivity Characterization of Pseudostationary Phases Using the Solvation Parameter Model Colin F. Poole
3.1 Introduction Separations in electrokinetic chromatography result from differences in the electrophoretic mobility of analytes as well as differences in their distribution between the electrolyte solution and a pseudostationary phase. Since all neutral molecules have the same electrophoretic velocity, it is the presence of the pseudostationary phase that is important in explaining differences in the retention of these compounds. To understand this process it is necessary to have a quantitative model that accounts for the ability of different pseudostationary phases to interact selectively with neutral compounds. In contemporary practice this is achieved by fitting retention data for compounds with varied properties to the solvation parameter model and extracting the contribution of defined intermolecular interactions to the separation process. The solvation parameter model has been widely applied to the characterization of distribution systems in chromatography, and References [1–6] can be consulted for a broader view of these applications. In early studies of the characterization of pseudostationary phases in electrokinetic chromatography the Kamlet–Taft solvatochromic model was often employed [7–10]. This model is similar, but not identical to Abraham’s solvation parameter model. The solute descriptors in the solvation parameter model are all free energy related properties, which is not the case for the solvatochromic model, and for this reason as well as others [1,2,6] the solvation parameter model has largely superseded the solvatochromic model in
Electrokinetic Chromatography: Theory, Instrumentation and Applications Edited by U. Pyell # 2006 John Wiley & Sons, Ltd. ISBN: 0-470-87102-4
56
ELECTROKINETIC CHROMATOGRAPHY
contemporary studies of chromatographic distribution properties. To take full advantage of these early studies, all data using the solvatochromic model has been recalculated by the solvation parameter model and is presented in this article in a common form [11–13].
3.2 Solvation Parameter Model The solvation parameter model is based on a cavity model of solution. Firstly, a cavity of a suitable size to accommodate the solute is formed in the solvent, with the solvent molecules in the same configuration as the bulk solvent. The energy required for this process depends on the forces holding the solvent molecules together and the solute’s size. In the second step, the solvent molecules are reorganized into their equilibrium position around the solute. The free energy for this process is small and can be neglected without much consequence. Finally, the solute is inserted into the cavity and various solute–solvent interactions are set up. For neutral compounds these are identified as dispersion, induction, orientation, and hydrogen bonding. For electrokinetic chromatography, transfer occurs between two condensed phases, and the free energy of transfer is equivalent to the difference in cavity formation and solute–solvent interactions in the electrolyte solution and pseudostationary phase. The contribution of each intermolecular interaction to transfer in the solvation parameter model is represented by the sum of product terms made up of solute factors (descriptors) and complementary solvent factors (system constants). A solute, therefore, has a certain ability to participate in each intermolecular interaction and the contribution of each interaction to the free energy of transfer is the product of solute and solvent properties. The solvation parameter model in a form suitable for characterizing the transfer of neutral compounds from an electrolyte solution to a pseudostationary phase in electrokinetic chromatography is set out in Equation (3.1): log SP ¼ c þ vV þ eE þ sS þ aA þ bB
ð3:1Þ
SP is some free energy related solute property such as a distribution constant or retention factor in electrokinetic chromatography. The remainder of the equations is made up of several product terms consisting of system constants (e, s, a, b, and v) and solute descriptors (earlier symbols for the descriptors are given in parenthesis) E(R2 ), S(pH 2 ), H A (aH 2 ), B (b2 ) and V(VX). Each product term represents a contribution to the solute property from a defined intermolecular interaction. The contribution from cavity formation and dispersion interactions are strongly correlated with solute size and cannot be separated if a volume term, such as the characteristic volume [V in Equation (3.1)], is used as a solute descriptor. Solute transfer between two condensed phases occurs with little change in the contribution from dispersion interactions, and thus, the absence of a specific term in Equation (3.1) to represent dispersion interactions is not a serious problem. For the solvation parameter model to have practical utility it is important that the solute descriptors are accessible for a wide range of compounds by either calculation or experiment. McGowan’s characteristic volume, V, in units of cm3mol1/100, can be calculated by simple summation rules for any compound whose structure is known [14,15]. The characteristic molecular volume is the sum of all atomic volumes less
SELECTIVITY CHARACTERIZATION OF PSEUDOSTATIONARY PHASES
57
Table 3.1 Examples of the calculation of V and E solute descriptors Calculation of McGowan’s characteristic volume, V, for toluene Atomic volumes: C ¼ 16.35, H ¼ 8.71, N ¼ 14.39, O ¼ 12.43, F ¼ 10.48, Si ¼ 26.83, P ¼ 24.87, S ¼ 22.91, Cl ¼ 20.95, B ¼ 18.23, Br ¼ 26.21, I ¼ 34.53. Subtract 6.56 for each bond of any type. Toluene ¼ 7 carbon atoms þ 8 hydrogen atoms – 15 bonds ¼ 114.45 þ 69.68 – 98.40 ¼ 85.73 in cm3mol1. After scaling V ¼ 0.857 in cm3mol1/100. Calculation of the excess molar refraction, E, for toluene using Equation (3.2). The refractive index for toluene (Z) at 20 C (sodium D-line) ¼ 1.496 E ¼ 8.57 (0.292) þ 0.5255 – 2.832 (0.857) ¼ 0.601 in cm3mol1/10.
6.56 cm3mol1 for each bond, no matter whether single, double or triple (see Table 3.1). For complex molecules the number of bonds, NB, is easily calculated from the algorithm NB ¼ N – 1 þ R where N is the total number of atoms, and R is the number of rings. The excess molar refraction, E, characterizes the contribution from the polarizability of solute n- and p-electrons. It is defined as the solute molar refraction less the molar refraction of an imaginary n-alkane with the same characteristic volume and is simply calculated in units of cm3mol1/ 10 from the refractive index of the solute at 20 C for the sodium Dline, Z, as indicated by Equation (3.2) [16]: E ¼ 10V½ðZ2 1Þ=ðZ2 þ 2Þ 2:832V þ 0:526
ð3:2Þ
The use of Equation (3.2) to calculate the excess molar refraction is straightforward for liquids, see Table 3.1, but even for solids refractive index values are easily estimated using available software for molecular property estimations. In addition, excess molar refraction is almost an additive property, and values for solids can be estimated through addition of fragments with known E values [1,6,17]. Initially the solute descriptors for hydrogen-bond acidity and basicity were determined from 1:1 complexation constants measured in an inert solvent [18,19]. These studies also led to scales with a zero origin. A problem remained, however, when these descriptors were used to characterize distribution processes. The effect of solute structure on the distribution process is a consequence of hydrogen bonding of the solute to any surrounding solvent molecules, not just to one. New scales of effective or overall hydrogen bond acidity and basicity, A and B, were determined in conjunction with other solute descriptors using liquid–liquid distribution and chromatographic measurements [15,20]. A minor complication is that some solutes (sulfoxides, anilines, pyridines) show variable hydrogen-bond basicity in distribution systems where the stationary phase absorbs appreciable amounts of water [20]. A new solute descriptor B was defined for these solutes and is generally used in electrokinetic chromatography. Except for the solute types indicated above, the two hydrogen-bond basicity scales are identical. It should also be noted that the scales of hydrogen-bond acidity and basicity are unrelated to proton transfer acidity and basicity expressed by the pKa scale. It would be useful to have descriptors that were related to the propensity of a solute to engage in dipole–dipole and induced dipole–dipole interactions. In the event, it proved impossible to separate out descriptors for the two types of interactions, and it was necessary to construct a solute descriptor for dipolarity/polarizability, S, combining the two interactions [16]. The dipolarity/polarizability solute descriptor is now more
58
ELECTROKINETIC CHROMATOGRAPHY
commonly determined in combination with the hydrogen-bonding descriptors from liquid–liquid distribution constants and chromatographic measurements [1,6,15]. Solute descriptors are available for over 4000 compounds, with the most complete repository being the University College London database incorporated into Absolv (Sirius, UK) [17]. For additional compounds estimation methods are available using fragment constants [1,6,17]. In all other cases it is possible to calculate V and E and determine the other descriptors from experimental distribution constants and chromatographic measurements. This is of little concern for the characterization of pseudostationary phases since the number and variety of solutes with established solute descriptors is more than adequate for the determination of system constants. The system constants in Equation (3.1) are obtained by multiple linear regression analysis for a number of solute property determinations for solutes with known descriptors. The system constants are more than mere regression constants and contain important chemical information about the system. The system constants reflect the difference in solute interactions in the background electrolyte solution and pseudostationary phase in electrokinetic chromatography. The e system constant indicates the difference in the tendency for the electrolyte solution and pseudostationary phases to interact with solute p- and n-electron pairs; the s system constant to the difference in the tendency of the electrolyte solution and pseudostationary phases to interact with solutes through dipole-type interactions; the a system constant indicates the difference in hydrogen-bond basicity between the two phases (because acidic solutes will interact with a basic phase); and the b system constant is a measure of the difference in hydrogenbond acidity between the two phases (because basic solutes will interact with an acidic phase). The v system constant is a measure of the difference in cavity formation in the electrolyte solution and pseudostationary phases together with any residual dispersion interactions that do not cancel when the solute is transferred between phases. 3.2.1
Model Requirements
A reliable model is one that makes chemical sense and is statistically valid. To obtain such a model certain conditions must be met. The dependent variable for electrokinetic chromatography is usually the retention factor (log k) or equilibrium constant (log K) for the distribution of the solute between the electrolyte solution and pseudostationary phase. The retention factor is easily accessible in electrokinetic chromatography and is used predominantly. There are two features of the retention factor that are important with respect to its use in the solvation parameter model. For modeling it is desirable that the log k values cover a reasonable numerical range with a uniform distribution throughout the range. In addition, the general error in the determination of the retention factor increases exponentially as k ) 0 and k ) 1 [21]. To achieve a minimal error and a stable error distribution, most of the k values should fall between 0.15 and 0.75 of the migration window for the separation system. Clustering of a large number of k values into either portion of the undesirable region will result in models with a large prediction error and erroneous or imprecise system constants. The solutes selected to establish reliable system constants should be sufficient in number and variety to properly define all interactions and to establish the statistical validity of the model [12,13,22]. A minimum of seven solutes is sufficient to solve
SELECTIVITY CHARACTERIZATION OF PSEUDOSTATIONARY PHASES
59
Equation (3.1) by multiple linear regression techniques. For statistical evaluation three varied values for each solute descriptor and the intercept is a reasonable minimum, but since individual solutes express several interactions simultaneously, the number of solutes required can be safely reduced from about 18 to 9. This condition presupposes that the error in the solute retention factors is roughly constant and that the descriptor values for each solute are known to the same accuracy. Neither condition is likely to prevail for most cases, and to select the minimum number of solutes for system characterization seems unwise. It is common practice to overdetermine the statistical requirements of Equation (3.1) to obtain an exhaustive fit. With such a fit it should be possible to remove a small set of randomly selected solutes from the model without significant variation in the system constant values. This can usually be achieved using 20–40 varied solutes. The solutes employed for system characterization are not selected at random. They are chosen so as to cover as evenly as possible the descriptor space to be explored. Clustering of descriptor values into narrow ranges should be avoided. This is easily checked by constructing a histogram plot for each descriptor. The presence of a central tendency in the plot is an indication of poor solute selection. The A solute descriptor is a partial exception, reflecting that among any group of varied compounds a significant number are going to be nonhydrogen-bond acids with A ¼ 0. When a varied group of solutes has been selected that adequately populates the descriptor space, it is necessary to check that the descriptor values are not highly cross correlated. Cross correlation results from the unintentional correlation between descriptor values. It results in a loss in capability of the multiple linear regression algorithm to distinguish between the correlated descriptors. There is almost always some correlation between any two sets of independent descriptor values and this is only considered a serious problem for r 0:8. When a high level of correlation exists between two sets of descriptors, it is likely that the model obtained will incorrectly assign values to the complementary system constants or may arbitrarily eliminate one of the system constants from the model entirely. In either case, the models are unlikely to be robust and will have poor predictive properties for new solutes. A useful compilation of solutes for characterizing pseudostationary phases in electrokinetic chromatography is given in Table 3.2. Selection of additional solutes may be appropriate to break cross correlation among subsets of the solutes and to obtain a reasonable range of retention factor values for the chosen separation system. The solute descriptors are defined for compounds in their neutral form. Compounds that will be significantly ionized at the pH selected for measurements should be removed from consideration. The solutes in Table 3.2 have a reasonable absorbance between 210–250 nm to facilitate absorption detection, since this is the common mode of detection in electrokinetic chromatography. Table 3.3 illustrates part of a typical output for the calculation of system constants by multiple linear regression analysis. The overall correlation coefficient, standard error in the estimate, Fischer F-statistic, and the standard deviation in the individual system constants are used to judge the statistical properties of the model. Acceptable models in electrokinetic chromatography will typically have values for the overall correlation coefficient of R > 0:97, a standard error in the estimate SE of 0.05–0.1, and a Fischer F-statistic of 150–1000. The standard error for the system constant should be smaller
60
ELECTROKINETIC CHROMATOGRAPHY
Table 3.2 Solute descriptors for typical compounds used for the characterization of pseudostationary phases in electrokinetic chromatography (UV absorbance detection between 210–250 nm) Descriptor values Compound
E
S
A
B
V
Acetanilide Acetophenone 4-Aminobenzamide Aniline Anisole Anthracene Antipyrine Benzaldehyde Benzamide Benzene Benzenesulfonamide Benzonitrile 2,3-Benzofuran Benzophenone Benzyl alcohol Benzyl benzoate Biphenyl Bromobenzene n-Butylbenzene n-Butyl benzoate n-Butyrophenone Caffeine Catechol 4-Chloroacetanilide 3-Chloroaniline 4-Chloroaniline Chlorobenzene 4-Chlorophenol Corticosterone Cortisone 1,4-Dichlorobenzene 2,3-Dimethylphenol 2,4-Dimethylphenol Estradiol Estriol Ethylbenzene Fluorobenzene Fluorene Geraniol n-Heptanophenone Hydrocortisone Indole Iodobenzene N-Methylbenzamide Methyl benzoate Methyl 3-hydroxybenzoate 1-Methylnaphthalene
0.870 0.818 1.340 0.955 0.708 2.290 1.320 0.820 0.990 0.610 1.130 0.742 0.888 1.447 0.832 1.264 1.360 0.882 0.600 0.668 0.797 1.500 0.970 0.980 1.050 1.060 0.718 0.915 1.860 1.960 0.825 0.850 0.840 1.800 2.000 0.613 0.477 1.588 0.513 0.720 2.030 1.200 1.188 0.950 0.733 0.905 1.344
1.36 1.01 1.94 0.96 0.75 1.34 1.50 1.00 1.50 0.52 1.55 1.11 0.83 1.50 0.95 1.42 0.99 0.73 0.51 0.80 0.95 1.60 1.10 1.50 1.10 1.10 0.65 1.08 3.43 3.50 0.75 0.90 0.80 3.30 3.36 0.51 0.57 1.03 0.63 0.95 3.49 1.12 0.82 1.44 0.85 1.40 0.90
0.46 0 0.80 0.26 0 0 0 0 0.49 0 0.55 0 0 0 0.37 0 0 0 0 0 0 0 0.88 0.64 0.30 0.30 0 0.67 0.40 0.36 0 0.52 0.53 0.88 1.40 0 0 0 0.39 0 0.71 0.44 0 0.35 0 0.66 0
0.69 0.48 0.94 0.50 0.29 0.26 1.48 0.39 0.67 0.14 0.80 0.33 0.15 0.50 0.56 0.51 0.22 0.09 0.15 0.46 0.51 1.33 0.47 0.51 0.36 0.35 0.07 0.20 1.63 1.87 0.02 0.36 0.39 0.95 1.22 0.15 0.10 0.20 0.66 0.50 1.90 0.31 0.12 0.73 0.46 0.45 0.20
1.1137 1.0139 1.0726 0.8162 0.9160 1.4540 1.5502 0.8750 0.9728 0.7164 1.0971 0.8711 0.9055 1.4808 0.9230 1.6804 1.3242 0.8914 1.2800 1.4953 1.2957 1.3632 0.8338 1.2237 0.9390 0.9390 0.8388 0.8975 2.7389 2.7546 0.9612 1.0569 1.0569 2.1988 2.2575 0.9982 0.7341 1.3565 1.4903 1.7184 2.7975 0.9460 0.9746 1.1132 1.0726 1.1313 1.2263
SELECTIVITY CHARACTERIZATION OF PSEUDOSTATIONARY PHASES
61
Table 3.2 (Continued) Descriptor values Compound
E
S
A
B
V
3-Methylphenol Naphthalene 2-Naphthol 2-Nitroaniline 3-Nitroaniline 4-Nitroaniline 2-Nitroanisole Nitrobenzene 4-Nitrobenzyl alcohol 1-Nitrobutane 1-Nitrohexane 1-Nitronaphthalene 4-Nitrotoluene Phenol 4-Phenylbutanol 2-Phenylethanol 4-Phenylphenol a-Pinene n-Propiophenone n-Propylbenzene n-Propyl benzoate Pyrimidine Pyrrole Quinoline Thiourea Thymol Toluene 2-Toluidine 1,2,3-Trihydroxybenzene Valerophenone p-Xylene
0.822 1.340 1.520 1.180 1.200 1.220 0.965 0.871 1.064 0.227 0.203 1.600 0.870 0.805 0.811 0.811 1.560 0.446 0.804 0.604 0.675 0.606 0.613 1.268 0.840 0.822 0.601 0.970 1.165 0.795 0.613
0.88 0.92 1.08 1.37 1.71 1.91 1.34 1.11 1.39 0.95 0.95 1.51 1.11 0.89 0.90 0.91 1.41 0.14 0.95 0.50 0.80 1.00 0.73 0.97 0.82 0.79 0.52 0.90 1.35 0.95 0.52
0.57 0 0.61 0.30 0.40 0.42 0 0 0.44 0 0 0 0 0.60 0.33 0.30 0.59 0 0 0 0 0 0.41 0 0.77 0.52 0 0.23 1.35 0 0
0.34 0.20 0.40 0.36 0.35 0.38 0.38 0.28 0.62 0.29 0.29 0.29 0.28 0.30 0.70 0.64 0.45 0.12 0.51 0.15 0.46 0.65 0.29 0.51 0.87 0.44 0.14 0.59 0.62 0.50 0.16
0.9160 1.0854 1.1441 0.9904 0.9904 0.9904 1.0902 0.8906 1.0902 0.8464 1.1282 1.2596 1.0320 0.7751 1.3387 10.596 1.3829 1.2574 1.1548 1.1391 1.3544 0.6342 0.5774 1.0443 0.5696 1.3387 0.8573 0.9751 0.8925 1.4366 0.9982
than the value of the system constant itself and a t-test used to indicate whether the system constant is statistically different from zero at the probability level selected for the test. In the final evaluation of any model, not only must the system constants be statistically valid but they must make chemical sense as well.
3.3 Micelles as Pseudostationary Phases Surfactants can be conveniently categorized by the charge on the head group as nonionic, anionic, cationic or zwitterionic, or by the type of organic moiety as hydrocarbon, bile salt and fluorocarbon surfactants. A general property of all surfactants is the formation of micelles (molecular aggregates) when their solution concentration exceeds a threshold value, called the critical micelle concentration (CMC). The driving force for micelle
62
ELECTROKINETIC CHROMATOGRAPHY
Table 3.3 An example of part of the output for fitting the solvation parameter model to a micellar electrokinetic chromatographic system by multiple linear regression analysis Descriptors
log k
Solute
V
E
S
A
B
Experimental
Predicted
Phenol Benzyl alcohol Aniline Toluene Ethylbenzene Naphthalene Benzaldehyde Nitrobenzene Chlorobenzene Acetophenone
0.775 0.916 0.816 0.857 0.998 1.085 0.873 0.890 0.838 1.014
0.805 0.803 0.955 0.601 0.613 1.340 0.820 0.871 0.718 0.818
0.89 0.87 0.96 0.52 0.51 0.92 1.00 1.11 0.65 1.00
0.60 0.33 0.26 0 0 0 0 0 0 0
0.31 0.56 0.50 0.14 0.15 0.20 0.39 0.28 0.07 0.49
0.306 0.268 0.386 0.553 0.997 1.185 0.017 0.143 0.618 0.275
0.273 0.252 0.380 0.524 0.937 1.256 0.011 0.225 0.597 0.236
Cross-correlation matrix (r2) V E S A B
1.00 0.19 0.00 0.29 0.00
1.00 0.33 0.22 0.03
1.00 0.03 0.42
Model statistics
1.00 0.17
1.00
R ¼ 0.994 SE ¼ 0.07 F ¼ 570 n ¼ 40
System constants v ¼ 2.99 (0.07) e ¼ 0.46 (0.05) s ¼ 0.44 (0.05) a ¼ 0.30 (0.05) b ¼ 1.88 (0.08) c ¼ 1.82 (0.07)
formation in aqueous solution is the favorable free energy change accompanying segregation of the hydrocarbon-like portion of the surfactant from water by packing hydrocarbon groups into a central core surrounded by the polar head groups. Electrostatic repulsion between head groups in ionic micelles, and the steric repulsion of hydrated head groups in the case of nonionic micelles, oppose this favorable free energy change. 3.3.1
Affect of System Properties on Selectivity
Micelles are dynamic structures whose shape, size and aggregation number fluctuate with changes in the surrounding electrolyte solution [2,23]. The concentration of surfactant, the type and concentration of buffer, pH of the electrolyte solution and temperature are expected to influence retention and selectivity in micellar electrokinetic chromatography through their affect on micelle properties. A systematic study of these variables using the surfactant sodium cholate indicated significant changes in retention but only modest changes in selectivity, Table 3.4 [11]. Numerical differences in the system constants should be established as significant by application of a Student t-test on a case-by-case basis (standard error in the system constants is indicated by the values in parentheses). For the present, only significant general features in the data are discussed. Changes in surfactant concentration effects retention mainly through changes in the phase ratio (incorporated in the c term when the retention factor is the dependent variable) and to a lesser extent through changes in the difference in cohesion between the micelles and electrolyte solution (v system constant). The phase ratio is weakly affected by the choice of buffer and pH but not by the concentration of the buffer over the range studied.
SELECTIVITY CHARACTERIZATION OF PSEUDOSTATIONARY PHASES
63
Table 3.4 Affect of experimental variables on the system constants of the solvation parameter model for sodium cholate micelles in electrokinetic chromatography (a ¼ 0 for all conditions) Experimental variables
System constants
Electrolyte solution v Surfactant concentration Concentration (mM) (mM) Compositiona pH
e
s
b
c
50
20
SP þ SB
75
20
SP þ SB
125
20
SP þ SB
75
20
SP þ SB
75
20
SP þ SB
75
20
SP
75
20
SB
75
10
SP þ SB
75
30
SP þ SB
0.65 (0.09) 0.63 (0.08) 0.48 (0.07) 0.61 (0.07) 0.57 (0.08) 0.59 (0.06) 0.63 (0.07) 0.60 (0.07) 0.60 (0.07)
0.47 (0.10) 0.47 (0.09) 0.46 (0.08) 0.47 (0.08) 0.41 (0.09) 0.48 (0.07) 0.46 (0.08) 0.50 (0.08) 0.50 (0.08)
2.27 (0.13) 2.29 (0.13) 2.14 (0.12) 2.31 (0.12) 2.26 (0.13) 2.26 (0.11) 2.30 (0.13) 2.23 (0.13) 2.29 (0.12)
2.11 (0.13) 1.71 (0.12) 1.34 (0.10) 1.90 (0.11) 1.77 (0.12) 1.79 (0.10) 1.93 (0.11) 1.63 (0.11) 1.65 (0.11)
8
2.59 (0.13) 8 2.45 (0.12) 8 2.39 (0.10) 8.5 2.59 (0.11) 9.0 2.45 (0.12) 8 2.52 (0.10) 8 2.58 (0.11) 8 2.43 (0.11) 8 2.49 (0.11)
a
SP ¼ sodium phosphate and SB ¼ sodium tetraborate
Changes in selectivity are small compared with general retention for the range of experimental conditions explored. In fact selectivity can be considered a reasonably robust parameter that depends on the identity of the surfactant and to only a limited extent on the choice of operating conditions. For those conditions indicated in Table 3.4, the average and standard deviation for the system constants are vAV ¼ 2.50 ( 0.08), eAV ¼ 0.60 ( 0.05), sAV ¼ 0.47 ( 0.03), bAV ¼ 2.26 ( 0.05). The change in system constants with temperature is approximately 0.013 for v, 0.002 for e, 0.004 for s and 0.022 for b for each 1 C change over the temperature range 15 to 35 C. The a system constant is not statistically different from zero for all conditions. Larger effects might be observed at surfactant concentrations closer to the critical micelle concentration where fluctuations in micelle properties are expected to be larger, but these conditions are less likely to be used for method development than those indicated in Table 3.4. The system constants reported by several research groups for the same surfactant are expected to show greater variation than the reproducibility of results from a single group using a robust experimental design. To illustrate this point the system constants for sodium dodecyl sulfate micelles reported by different groups are collated in Table 3.5 [7,9,11,22,24–31]. The data shows no particular trends related to differences in experimental conditions, although in a number of cases the difference in individual system constants are significant at the 95 % confidence level. The central tendency is indicated
64
ELECTROKINETIC CHROMATOGRAPHY
by the average value for each system constant and the dispersion by the standard deviation for all entries in Table 3.5. These values are: vAV ¼ 2.87 ( 0.14), eAV ¼ 0.39 ( 0.11), sAV ¼ 0.40 ( 0.13), aAV ¼ 0.21 ( 0.11) and bAV ¼ 1.80 ( 0.16). A significant amount of the observed variation probably results from a failure to select solutes that homogeneously fill the descriptor space, provide sufficient range for the dependent variable as well as failure to ensure that an exhaustive fit was obtained. Variation in experimental conditions is probably less important than variations in the statistical design of the studies. In the absence of other information the reported dispersion for the system constants in Table 3.4 (single group) and Table 3.5 (different groups) can be taken as an estimate of anticipated variation in compilations of system constants. 3.3.2
Anionic Surfactants
The system constants for a number of common anionic surfactants are summarized in Table 3.6 [11–13,22,27–29,32,33]. In general, the relative ease of cavity formation and dispersion interactions (v system constant) and electron lone pair interactions (e system constant) favor transfer to micelles from the electrolyte solution. Solute interactions of a dipole-type and hydrogen bonding are more favorable in the electrolyte solution, and oppose transfer to the micelles. An exception is the surfactant lithium perfluorooctanesulfonate, which has a positive s system constant, although literature values are in conflict (see Table 3.6). A single feature that stands out in Table 3.6 is the narrow range of v system constant values for all anionic surfactants. This is probably a reflection of the need for a minimum cohesive energy difference for aggregation of surfactant monomers. The solvation properties of the alkane sulfate and sulfonate micelles are not particularly sensitive to changes in the identity of the counterion or alkyl chain length [29,32]. For selectivity optimization there are only minor benefits in replacing sodium dodecyl sulfate with any of the other alkane sulfates and sulfonates in Table 3.6. The perfluorooctanesulfonate and N-alkyl-N-methyltaurine surfactants, however, afford a useful change in selectivity compared with sodium dodecyl sulfate [13,22]. Lithium perfluorooctanesulfonate has different selectivity for electron lone pair interactions (the only negative e system constant) and is the strongest hydrogen-bond acid and weakest hydrogen-bond base of all surfactants in Table 3.6. The N-alkyl-N-methyltaurine surfactants are about as dipolar and stronger hydrogen-bond bases and weaker hydrogen-bond acids than the other alkane sulfate micelles. The bile acids have similar solvation properties as a group with different selectivity to the alkane sulfates. They are more cohesive, stronger hydrogen-bond bases and weaker hydrogen-bond acids [11,12]. The N-alkylsarcosinates have similar solvation properties to the N-alkyl-N-methyltaurine surfactants. The dodecylcarboxylate and dodecylphosphate surfactants are slightly more hydrogen-bond basic than sodium dodecyl sulfate, and could be useful for optimizing the peak position of solutes with different hydrogenbond acidity [27,28,33]. It is clear from the results in Table 3.6 that both the identity of the surfactant head group and relatively small changes in structure in the region of the head group can affect selectivity for dipole-type and hydrogen-bond interactions. These changes in selectivity
SELECTIVITY CHARACTERIZATION OF PSEUDOSTATIONARY PHASES
65
Table 3.5 System constants for sodium dodecyl sulfate micelles determined for different experimental conditions (temperature ¼ 25 C) System constants c
Statistics
Conditions
(i) 50 mM sodium dodecyl sulphate 2.62 0.56 0.67 0.31 1.57 (0.07) (0.09) (0.07) (0.07) (0.08)
1.48 (0.07)
R ¼ 0.996 SE ¼ 0.07 F ¼ 517
2.99 (0.07)
0.46 (0.05)
0.44 (0.05)
0.30 (0.05)
1.88 (0.08)
1.82 (0.07)
R ¼ 0.994 SE ¼ 0.07
2.40 (0.09)
0.51 (0.05)
0.41 (0.07)
0.34 (0.08)
1.76 (0.08)
1.41 (0.06)
F ¼ 569 R ¼ 0.994 SE ¼ 0.10 F ¼ 507
2.98
0.35
0.43
0.02
2.02
1.76
R ¼ 0.993 SE ¼ 0.08 F ¼ 291
2.84 (0.09)
0.38 (0.11)
0.29 (0.13)
0.13 (0.08)
1.69 (0.14)
1.99 (0.12)
R ¼ 0.996 SE ¼ 0.08 F ¼ 445
Buffer ¼ 20 mM sodium phosphate pH ¼ 7 [9] n ¼ 26 Buffer ¼ 20 mM sodium phosphate -sodium tetraborate, pH ¼ 8 [11] n ¼ 40 Buffer ¼ 100 mM sodium tetraborate þ60 mM sodium phosphate, pH ¼ 7 n ¼ 38 [24] Buffer ¼ 20 mM sodium phosphate -sodium tetraborate, pH ¼ 8 [25] n ¼ 27 Buffer ¼ 20 mM Tris, pH ¼ 8, T ¼ 30 C [26] n ¼ 25
(ii) 40 mM sodium dodecyl sulfate 2.81 0.46 0.48 0.16 (0.09) (0.06) (0.07) (0.04)
1.71 (0.08)
1.78 (0.08)
2.72 (0.06)
0.56 (0.08)
0.60 (0.06)
0.27 (0.05)
1.67 (0.07)
1.68 (0.05)
2.95 (0.11)
0.19 (0.07)
0.30 (0.08)
0.17 (0.04)
1.84 (0.09)
1.80
2.86 (0.09)
0.25 (0.06)
0.31 (0.06)
0.15 (0.04)
1.70 (0.07)
1.85
Buffer ¼ 50 mM sodium phosphate pH ¼ 7 [7] n ¼ 59 R ¼ 0.991 Buffer ¼ 20 mM SE ¼ 0.12 sodium phosphate F ¼ 618 pH ¼ 7 [22] n ¼ 63 R ¼ 0.993 Buffer ¼ 10 mM SE ¼ 0.06 sodium phosphate pH ¼ 7 [27] n ¼ 36 R ¼ 0.995 Buffer ¼ 10 mM sodium phosphate pH ¼ 7 [28]
1.72 (0.08)
2.16 (0.06)
v
e
s
a
(iii) 30 mM sodium dodecyl sulfate 2.90 0.42 0.34 0.11 (0.07) (0.06) (0.04) (0.05)
b
R ¼ 0.991 SE ¼ 0.07 F ¼ 565
R ¼ 0.994 SE ¼ 0.06
Buffer ¼ 13 mM sodium phosphate pH ¼ 6.92 [29] n ¼ 36 (Continued)
66
ELECTROKINETIC CHROMATOGRAPHY
Table 3.5 (Continued) System constants v
e
s
a
(iv) 20 mM sodium dodecyl sulfate 2.83 0.47 0.44 0.15 (0.09) (0.06) (0.07) (0.04)
b
c
Statistics
Conditions
1.71 (0.08)
2.17 (0.08)
R ¼ 0.991 SE ¼ 0.07 F ¼ 574
Buffer ¼ 50 mM sodium phosphate pH ¼ 7 [7] n ¼ 59
R ¼ 0.998 SE ¼ 0.06
Buffer ¼ 12.5 mM sodium phosphate –sodium tetraborate, pH ¼ 9.2 [30] n ¼ 18 Buffer 20 mM Tris pH ¼ 7 [31] n ¼ 20
(v)1.0–1.2 % (w/v) sodium dodecyl phosphate 2.74 0.27 0.37 0.23 1.82 1.65 (0.11) (0.08) (0.07) (0.13) (0.16) (0.11)
F ¼ 479 3.17 (0.11)
0.37 (0.06)
0.32 (0.09)
0.31 (0.05)
2.19 (0.13)
2.16 (0.68)
R ¼ 0.998 SE ¼ 0.05 F ¼ 520
are probably determined by the amount of water attracted into the interphase region as a result of the modification of the surfactant structure [22,23,28]. 3.3.3
Cationic Surfactants
System constants for tetradecyltrimethylammonium and hexadecyltrimethylammonium bromide micelles are summarized in Table 3.7 [11–13,22]. They afford complementary selectivity to the anionic surfactants described in Table 3.6. They are about as cohesive and dipolar as the anionic surfactants but are stronger hydrogen-bond bases and among the weakest hydrogen-bond acids of all surfactants studied. 3.3.4
Mixed Surfactant Micelles
The separation characteristics of single surfactant micelles can be modified by forming mixed micelles using surfactants with different solvation properties. Surfactants with a zero net charge, such as nonionic and zwitterionic surfactants, unsuitable for use on their own, when added to an ionic surfactant, form mixed surfactant micelles with an overall net charge and adjustable selectivity. Variation of the mole fraction ratio of nonionic surfactant Brij 35 (polyoxyethylene(23)dodecyl ether) on the selectivity of the mixed surfactant micelles formed with sodium dodecyl sulfate [25,34] or sodium N-dodeconyl-N-methyltaurine [35] produced similar trends, Figure 3.1. The addition of Brij 35 resulted in little change in cohesion of the micelles or the relative importance of dipole-type and electron lone pair interactions. Changes in selectivity resulted mainly from variation in hydrogen bonding interactions, particularly hydrogen-bond basicity at low mole fractions of Brij 35. These trends in selectivity conformed to predictions by the interfacial retention model [24,34,35]. Lithium dodecyl sulfate and lithium perfluorooctanesulfonate mixed micelles have also been studied in some detail, Table 3.8 [36,37]. Changes in the hydrogen-bonding system constants were virtually linear with an increase in the mole fraction of lithium perfluorooctanesulfonate while the other system constants show a quadratic dependence.
SELECTIVITY CHARACTERIZATION OF PSEUDOSTATIONARY PHASES
67
Table 3.6 System constant ratios from the solvation parameter model for anionic surfactants (temperature ¼ 20–25 C except were noted) System constant ratios Surfactant (i) Alkane sulfates and sulfonates Sodium octyl sulfate Sodium decyl sulfate Sodium dodecyl sulfate (36 C) Lithium dodecyl sulfate (35 C) Magnesium dododecyl sulfate (35 C)Mg(DS)2 Copper dododecyl sulfate (35 C) Tris(hydroxymethyl)aminomethane dodecyl sulfate Sodium dodecyl sulfonate (36 C) Sodium tetradecyl sulfate (35 C) Lithium perfluorooctanesulfonate
Abbreviation v SOS SDecS SDS
3.3.5
b/v Reference [29] [29] [11] [28] [22] [32] [32]
Cu(DS)2
3.05 0.11 0.17 0.09 0.63 2.56 0.22 0.26 0.13 0.62
[32] [12]
SDSu STS LPFOSu
2.84 0.12 3.01 0.09 2.36 0.29 2.30 0.23 1.97 0.06 2.88 0.18 3.07 0.23
0.15 0.01 0.63 0.11 0.06 0.60 0.19 0.34 0.26 0.15 0.36 0.23 0.12 0.45 0.23 0.12 0.14 0.82 0.16 0.07 0.84
[28] [33] [13] [12] [22] [27] [11]
2.45 2.67 2.43 2.62
0.26 0.25 0.25 0.26
0.19 0.18 0.14 0.17
0 0 0 0
0.93 0.93 0.85 0.83
[11] [11] [11] [11]
3.02 2.99 3.10 2.92
0.13 0.16 0.14 0.15
0.11 0.14 0.15 0.13
0.16 0.77 0.15 0.82 0.15 0.83 0.17 0.83
[33] [33] [33] [27]
2.99 2.96 3.01 2.96
0.14 0.05 0.08 0.16
0.19 0.13 0.18 0.13
0.05 0.81 0.08 0.60 0.05 0.66 0.04 0.82
[28] [28] [28] [28]
Sodium N-lauroyl-N-methyltaurine SLMT Sodium N-dodecanoyl-NSDMT methyltaurine (ii) Bile acids Sodium cholate SC Sodium deoxycholate SDC Sodium taurocholate STC Sodium taurodeoxycholate STDC (iii) Miscellaneous anionic surfactants Sodium N-lauroylsarcosinate SLN Sodium N-myristoylsarcosinate SMN Sodium N-parmitoylsarcosinate SPN Sodium N-lauroyl-N-methylALE b-alaninate Sodium dodecoxycarbonylvaline SDCV Sodium dodecylcarboxylate (36 C) SDCA Sodium dodecylphosphate (36 C) SDP Sodium lauryl sulfoacetate (36 C) SLSA
0.16 0.12 0.12 0.09 0.22 0.10 0.09
s/v a/v
0.11 0.04 0.66 0.09 0 0.59 0.14 0.08 0.64 0.11 0.05 0.59 0.23 0.12 0.60 0.12 0.07 0.59 0.14 0.09 0.62
LDS
2.85 2.69 2.98 2.86 2.61 3.01 3.02
e/v
Addition of Organic Solvents
Typical water miscible organic solvents (e.g. short chain alcohols, acetonitrile) are known to increase the critical micelle concentration of surfactants and at sufficiently high concentrations inhibit micelle formation [2,23,35,38,39]. Instability of the micelles restricts the range of normal composition variation to 25 % (v/v) organic solvent. Changes in selectivity are consistent with the view that organic solvents moderate the characteristic solvophobic properties of the electrolyte solution and barely interact with the micelles. Some typical results are illustrated in Figure 3.2 [35]. Increasing the volume fraction of
68
ELECTROKINETIC CHROMATOGRAPHY
Table 3.7 System constant ratios from the solvation parameter model for cationic surfactants (temperature ¼ 20–25 C) System constant ratios Surfactant
Abbreviation v
Tetradecyltrimethyl-ammonium TTAB bromide
2.90 2.82 2.63 3.40 2.71
Hexadecyltrimethyl-ammonium CTAB bromide
e/v
s/v
a/v
b/v
0.10 0.13 0.34 0.18 0.41
0.07 0.10 0.23 0.16 0.28
0.29 0.32 0.29 0.17 0.30
0.91 0.95 0.92 0.91 0.90
Reference [13] [12] [22] [11] [22]
4
System constants
v 2 e a s c
0 –2
b –4 0
10
20
30
40
50
Brij 35 (mM)
Figure 3.1 Variation of the system constants for mixed surfactant micelles formed by adding 0–50 mM nonionic surfactant Brij 35 to 50 mM sodium dodecyl sulfate in a 20 mM sodium phosphate and sodium tetraborate pH ¼ 8 buffer (Brij 35: 0–20 mM) or 20 mM sodium phosphate pH ¼ 7 buffer (Brij 35: 25–50 mM). (Plotted from data given in Ref. [34])
Table 3.8 System constants for mixed micelles formed by adding lithium perfluorooctanesulfonate to lithium dodecyl sulfate (40 mM total surfactant, 20 mM lithium phosphate buffer, pH ¼ 7, T ¼ 25 C). (Data from [36]) Mole fraction LPFOSu a
0
0.25 0.50 0.75 1.00a a
System constants v 2.81 (0.09) 2.74 (0.07) 2.64 (0.07) 2.45 (0.08) 2.20 (0.08)
e 0.36 (0.10) 0.27 (0.08) 0.16 (0.08) 0.02 (0.09) 0.25 (0.09)
s 0.43 (0.07) 0.41 (0.06) 0.31 (0.06) 0.16 (0.06) 0 (0.07)
a 0.20 (0.06) 0.37 (0.05) 0.58 (0.05) 0.76 (0.05) 0.92 (0.06)
Different values for the system constants for the single surfactants are reported in [37].
b 1.54 (0.11) 1.20 (0.10) 0.85 (0.09) 0.45 (0.10) 0 (0.11)
c 1.78 (0.08) 1.79 (0.07) 1.85 (0.06) 1.90 (0.07) 1.90 (0.08)
SELECTIVITY CHARACTERIZATION OF PSEUDOSTATIONARY PHASES
69
System constants
4 v
2
e a s c
0
–2
b
–4 0
5
10
15
20
Acetonitrile (%v/v)
Figure 3.2 Variation of the system constants for the mixed surfactant micelles formed by adding 20 mM Brij 35 to 50 mM sodium N-dodeconyl-N-methyltaurine in an electrolyte solution containing 0–20 % (v/v) acetonitrile. (Plotted from data given in Ref. [35])
acetonitrile reduces retention by lowering the difference in cohesion between the micelles and electrolyte solution and by increasing the hydrogen-bond acidity of the micelles relative to the electrolyte solution. Higher mass alcohols, alkyl polyols and neutral polar compounds, in contrast to typical water soluble solvents, are solubilized by the micelles forming mixed micelles with properties similar to mixed surfactant micelles [26,40,41].
3.4 Microemulsions A microemulsion consists of a homogeneous dispersion of charged oil droplets in an aqueous electrolyte [2,42,43]. They are formed from an organic solvent of low water solubility (e.g. octane, heptane, diethyl ether, octanol, etc.), a charged surfactant (e.g. sodium dodecyl sulfate) and a cosurfactant (e.g. butanol, 2-ethoxyethanol, tetrahydrofuran) in an aqueous buffer. The organic solvent and cosurfactant form the core of the droplet solvating a part of the hydrophobic portion of the surfactant with its charged head group located in the electrolyte solution. Stable microemulsions are formed over a narrow composition range for the main components, although a variety of cosurfactants and anionic and cationic surfactants alone or mixed with zwitterionc and nonionic surfactants can be used. The concentration and type of organic solvent chosen to form the oil drop, plays only a minor role in controlling selectivity compared with the concentration and identity of the surfactant and cosurfactant [44,45]. The microemulsion formed by 0.8 % (w/w) n-octane, 6.6 % (w/w) n-butanol, 3.3 % (w/w) sodium dodecyl sulfate and 89.3 % (w/w) of 10–20 mM pH 8.5–9.5 borate or phosphate buffer accounts for most of the practical applications of microemulsions in electrokinetic chromatography. System constants for this microemulsion at several pH values are summarized in Table 3.9 [3,46,47]. An important application of this microemulsion is the estimation of the octanol–water partition coefficient for neutral compounds owing to the strong correlation of the partition properties of the microemulsion with those of wet octanol [3,47].
0.34 (0.10) 0.56 (0.13)
2.24 (0.07) 3.09 (0.18) 2.85 (0.16)
pH ¼ 10
40 mM bis(2-ethylhexyl) sodium sulfosuccinate þ10 % (v/v) methanol 1.8 % (w/w) sodium octylsulphate and hexadecyltrimethylammonium bromide (7:3)
0.37 (0.04)
0.31 (0.06)
2.39 (0.09)
pH ¼ 8
0.41 (0.06)
0.28
e
2.16 (0.10)
3.05
v
pH ¼ 3
1.4 % (w/v) sodium dodecyl sulfate þ 8 % (v/v) n-butanol þ 1.2 % (v/v) n-heptane pH ¼ 7
System
0.57 (0.12)
0.43 (0.13)
0.51 (0.05)
0.52 (0.06)
0.50 (0.06)
0.69
s
0.23 (0.09)
0.02 (0.10)
0
0
0
0.06
a
System constants
3.25 (0.18)
3.02 (0.19)
1.97 (0.07)
2.26 (0.08)
2.01 (0.09)
2.81
b
Table 3.9 System constants for microemulsion and vesicle electrokinetic chromatography
1.52
1.82
0.85 (0.05)
0.87 (0.06)
0.91 (0.07)
1.13
c
R ¼ 0.989 n ¼ 39
R ¼ 0.994 SE ¼ 0.09 F ¼ 792 n ¼ 53 R ¼ 0.984 SE ¼ 0.14 F ¼ 283 n ¼ 42 R ¼ 0.985 SE ¼ 0.16 F ¼ 363 n ¼ 51 R ¼ 0.990 SE ¼ 0.12 F ¼ 471 n ¼ 45 R ¼ 0.987 n ¼ 43
Statistics
[48]
[48]
[47]
[47]
[47]
[46]
Reference
SELECTIVITY CHARACTERIZATION OF PSEUDOSTATIONARY PHASES
71
3.5 Vesicles Vesicles, self-assembling structures that contain a continuous bilayer of monomers enclosing an aqueous core, can be synthesized using surfactants with structures containing two alkyl chains and a polar head group, or by using a mixtures of single chain cationic and anionic surfactants under appropriate conditions. System constants for the vesicles formed by 40 mM bis(2-ethylhexyl)sodium sulfosuccinate and methanol (10 % v/v) in 10 mM phosphate pH 7 buffer and hexadecyltrimethylammonium bromide and sodium octyl sulfate (30:70 mole ratio, total mass 1.8 % w/w) in 50 mM phosphate pH 7 buffer are summarized in Table 3.9. These vesicles were shown to have partition properties suitable for use as correlation models for the estimation of octanol–water partition coefficients [48]. In this respect their selectivity is similar to the sodium dodecyl sulfate, n-butanol and n-heptane microemulsion in Table 3.9, which is a better model for the estimation of octanol–water partition coefficients [3].
3.6 Polymeric Pseudostationary Phases Typical polymeric pseudostationary phases include micelle polymers, polymeric surfactants, and water-soluble anionic siloxanes and dendrimers [49,50]. Micelle polymers [e.g. poly(sodium 10-undecylenate), poly(sodium 10-undecenylsulfate), etc.] are synthesized from polymerizable surfactant monomers at a concentration above their critical micelle concentration [30,51]. These polymers have similar structures to surfactant micelles without the dynamic nature. Polymeric surfactants are polymers with surfactant properties [e.g. acrylate copolymers, polybrene, etc.] [31,52]. Water-soluble anionic siloxane polymers are copolymers of alkylmethylsiloxane and methylsiloxane monomers containing ionic groups [53,54]. Polymeric micelles form stable pseudostationary phases with a critical micelle concentration of virtually zero (aggregation number of 1), and are tolerant of high organic solvent concentrations in the electrolyte solution. The system constants for 34 polymeric pseudostationary phases are summarized in Table 3.10 [30,31,51–55]. The models for polybrene and PDADMA seem unlikely in chemical terms and present poor statistics [51]. They are not discussed further. The polymeric pseudostationary phases have similar cohesion to surfactant micelles, except for the anionic siloxane polymers, which are more cohesive as a group than either the surfactant micelles or acrylate copolymers. Electron lone pair interactions favor transfer to the polymeric pseudostationary phases similar to surfactant micelles. The polymeric pseudostationary phases, except for poly(Na 11-AAU), have weak dipole-type interactions that favor solvation in the electrolyte solution. Some polymer pseudostationary phases have much weaker dipole-type interactions than do typical surfactant micelles, representing a significant difference in selectivity. The polymeric pseudostationary phases exhibit a wide range of hydrogen bonding properties. Some are significantly stronger hydrogen-bond bases than the electrolyte solution, not unlike typical alkylcarboxylate and tetraalkylammonium surfactant micelles. All of the polymeric pseudostationary phases are weaker hydrogen-bond acids than the electrolyte solution, with some of the anionic siloxane polymers among the weakest hydrogen-bond acids observed for pseudostationary phases in electrokinetic chromatography.
72
ELECTROKINETIC CHROMATOGRAPHY
Table 3.10 System constants for polymeric pseudostationary phases used in electrokinetic chromatography System constants System
v
e
s
a
b
c
References
1.18 (0.17) 1.05 (0.13) 0
2.28 (0.11) 1.86 (0.09) 0.09
0
0.81
[51]
3.86 (0.91) 2.91 (0.22) 3.40 (0.20) 3.15 (0.17) 2.84 (0.23) 3.58 (0.26) 3.36 (0.16) 3.39 (0.13) 3.17 (0.16) 3.65 (0.18) 3.76 (0.15) 3.78 (0.21) 2.88 (0.13) 3.39 (0.13) 2.07 (0.2) 1.71 (0.2) 1.76 (0.1) 2.21 (0.3) 2.06 (0.3) 1.34 (0.3)
0.19 (0.05) 0.11 (0.07) 0.04 (0.10) 0.24 (0.09) 0.23 (0.08) 0
0
Poly( sodium undecenylsulfate) pDHCHAt-2-Nae
0.45 (0.08) 0.18 (0.06) 0.35 (0.02) 0.35 (0.07) 0.44 (0.66) 0.60 (0.19) 0.46 (0.17) 0.40 (0.15) 0.32 (0.20) 0.40 (0.22) 0.44 (0.14) 0.53 (0.11) 0.32 (0.09) 0.67 (0.16) 0.69 (0.13) 0.85 (0.18) 0.32 (0.11) 0.53 (0.11) 0.07 (0.1) 0.22 (0.1) 0.38 (0.08) 0.90 (0.2) 1.08 (0.2) 1.05 (0.2)
0.15 (0.13) 0.27 (0.11) 0
PDADMAd
0.18 (0.08) 0.26 (0.06) 0.75 (0.06) 0.75 (0.06) 0
[30]
Polybrenec
1.64 (0.11) 2.11 (0.09) 0
1.39 (0.75) 2.58 (0.28) 3.21 (0.25) 3.19 (0.21) 2.75 (0.28) 3.52 (0.32) 3.22 (0.20) 3.05 (0.16) 2.19 (0.13) 3.70 (0.22) 3.87 (0.19) 3.83 (0.26) 2.45 (0.16) 3.05 (0.16) 1.93 (0.2) 2.36 (0.2) 1.90 (0.1) 2.53 (0.2) 2.29 (0.2) 1.96 (0.2)
2.79
[51]
3.11 (0.20) 2.68 (0.18) 3.15 (0.15) 3.25 (0.20) 2.96 (0.23) 2.86 (0.14) 2.57 (0.12) 2.16 (0.68) 2.84 (0.16) 2.73 (0.14) 2.73 (0.19) 2.69 (0.11) 2.57 (0.12) 2.81 (0.2) 2.37 (0.2) 1.96 (0.1) 1.51 (0.2) 1.65 (0.22) 1.29 (0.3)
[52]
Poly(Na 11-AAU)
a
Poly(Na 10-U)b
pDHCHAt-33-TEAf pLAt-9.2-TEAg pLAt-9-Nah pLAt-13-Nah ptOAm-49-Nai pSAm-28-Naj pOMAt-21k pLMAt-15l pSMAt-13m pSMAt-16m pLMAm-19n pSAm-28o AGENTp OAGENT (10 % C8)q OAGENT (15 % C8)q OAGENT (20 % C8)q OAGENT (25 % C8)q DAGENT (10 % C12)r
0.61 (0.13) 0.65 (0.12) 0.50 (0.10) 0.34 (0.14) 0.39 (0.15) 0.33 (0.10) 0.42 (0.08) 0.37 (0.06) 0.44 (0.11) 0.42 (0.09) 0.65 (0.13) 0.37 (0.13) 0.42 (0.12) 0.76 (0.1) 0.54 (0.1) 0.46 (0.07) 0.49 (0.2) 0.71 (0.1) 0.58 (0.2)
0.02 (0.12) 0.43 (0.08) 0.19 (0.06) 0.31 (0.05) 0.27 (0.08) 0.22 (0.07) 0.50 (0.10) 0.25 (0.06) 0.19 (0.06) 0.45 (0.09) 0.25 (0.08) 0.20 (0.05) 0.11 (0.1) 0.11 (0.1) 0.51 (0.1)
[30] [51]
[52] [52] [31] [52] [52] [52] [31] [31] [31] [31] [31] [31] [53] [53] [53] [53] [53] [53]
SELECTIVITY CHARACTERIZATION OF PSEUDOSTATIONARY PHASES
73
Table 3.10 (Continued) System constants System
v
e
s
a
b
c
References
DAGENT (15 % C12)r
2.46 (0.3) 2.39 (0.2) 1.49 (0.2) 1.22 (0.2) 2.51 (0.3) 1.33 (0.3) 2.49 (0.4) 2.04 (0.2) 2.72 (0.1) 2.25 (0.23)
0.32 (0.2) 0.59 (0.1) 0.16 (0.1) 0.20 (0.2) 0.63 (0.2) 0.57 (0.2) 0.31 (0.2) 0.40 (0.1) 0.46 (0.06) 0.30 (0.16)
0.86 (0.2) 0.78 (0.1) 0.57 (0.2) 0.65 (0.2) 1.14 (0.2) 1.06 (0.2) 0.88 (0.3) 0.17 (0.1) 0.43 (0.08) 0.08 (0.18)
0.21 (0.1) 0.23 (0.07) 0.01 (0.09) 0.05 (0.09) 0.33 (0.1) 0.51 (0.1) 0.21 (0.1) 0.24 (0.08) 0.27 (0.04) —
2.38 (0.3) 2.42 (0.2) 2.00 (0.2) 1.65 (0.2) 2.64 (0.3) 1.95 (0.3) 2.40 (0.3) 2.09 (0.09) 2.46 (0.09) 2.18 (0.29)
1.84 (0.3) 1.98 (0.2) 1.51 (0.2) 1.23 (0.2) 1.75 (0.3) 1.26 (0.3) 1.83 (0.3) 2.50 (0.2) 2.40 (0.1) 1.86
[53]
DAGENT (20 % C12)r SAGENT (10 % C18)s SAGENT (15 % C18)s SAGENT (20 % C18)s C12AGENT-10t C12AGENT-15t C12AGESS-8u C12AGESS-13u Elvacite 2669v a
[53] [53] [53] [53] [54] [54] [54] [54] [55]
Poly(sodium 11-acrylamidoundecanoate) Poly(sodium 11-undecylenate) Poly(N,N,N0 ,N0 -tetramethyl-N-trimethylenehexamethylenediammonium) dibromide d Poly(diallyldimethylammonium) bromide e Poly(sodium dihydrocholesteryl acrylate-co-2-acrylamido-2-methyl-1-propane sulfate) containing 2 % dihydrocholesteryl acrylate monomer f Poly(triethylammonium dihydrocholesteryl acrylate-co-2-acrylamido-2-methyl-1-propane sulfate) containing 33 % dihydrocholesteryl acrylate monomer g Poly(triethylammonium lauryl acrylate-co-2-acrylamido-2-methyl-1-propane sulfate) containing 9.2 % lauryl acrylate monomer h Poly(sodium lauryl acrylate-co-2-acrylamido-2-methyl-1-propane sulfate) containing 9 or 13 % lauryl acrylate monomer i Poly(sodium t-octyl acrylamide-co-2-acrylamido-2-methyl-1-propane sulfate) containing 49 % t-octyl acrylamide monomer j Poly(sodium stearylacrylamide-co-2-acrylamido-2-methyl-1-propane sulfate) containing 28 % stearylacrylamide monomer k Poly(sodium octyl methacrylate-co-2-acrylamido-2-methyl-1-propane sulfate) containing 21 % octyl methacrylate monomer l Poly(sodium lauryl methacrylate-co-2-acrylamido-2-methyl-1-propane sulfate) containing 15 % lauryl methacrylate monomer m Poly(sodium stearyl methacrylate-co-2-acrylamido-2-methyl-1-propane sulfate) containing 13 % or 16 % stearyl methacrylate monomer n Poly(sodium lauryl methacrylamide-co-2-acrylamido-2-methyl-1-propane sulfate) containing 19 % lauryl methacrylamide monomer o Poly(sodium steryl methacrylamide-co-2-acrylamido-2-methyl-1-propane sulfate) containing 28 % steryl methacrylamide monomer p Poly(allyl gycidyl ether N-methyltaurine siloxane) q Poly(octane allyl glycidyl ether N-methyltaurine siloxane) r Poly(dodecane allyl glycidyl ether N-methyltaurine siloxane) s Poly(steryl allyl glycidyl ether N-methyltaurine siloxane) t Poly(allyl gycidyl ether N-methyltaurine siloxane-co-dodecylmethylsiloxane) containing 10 or 15 % of dodecylmethylsiloxane monomer u Poly(allyl gycidyl ether sulfonate siloxane-co-dodecylmethylsiloxane) containing 8 or 13 % of dodecylmethylsiloxane monomer v Poly(methyl methacrylate–ethyl acrylate–methylacrylic acid). None of the solutes used to evaluate this polymer were hydrogen-bond acids. b c
74
ELECTROKINETIC CHROMATOGRAPHY
Often large changes in the selectivity for polymeric pseudostationary phases have been observed for incremental changes in chain length or monomer composition that are not easy to explain. A possible contributing factor is steric hindrance of polar functional groups and changes in hydration of the polymers as a result of alterations in composition of the polymer backbone and substituent chain lengths. Consequently, the best approach for the design of polymeric pseudostationary phases with specific selectivity is unclear at this time. The phase ratios for several polymeric pseudostationary phases (c term in Table 3.10) are unfavorable for optimum retention of typical small molecules separated by electrokinetic chromatography. On the other hand, the range of selectivity observed for polymeric pseudostationary phases is greater than for surfactant micelles, suggesting that further exploration could result in the identification of useful materials if other kinetic and sample loading problems were overcome.
3.7 Selectivity Equivalence of Surfactant Micelles and Other Chromatographic and Biopartitioning Systems The solvation parameter model provides a suitable framework to identify separation techniques with similar properties and surrogate chromatographic models for biopartitioning (and similar difficult-to-study) processes [3,4,15,56–58]. For two processes to emulate each other they would need to have identical system constants. This requirement is quite restrictive and there are no reported successes for systems of interest. Two models will be correlated, however, when the ratio of their system constants (e/v, s/v, a/v and b/v) are (nearly) identical. An outstanding example of this approach for property estimates was the identification of the microemulsion system in Table 3.9 as a suitable model for the high-throughput determination of octanol–water partition coefficients, a widely used measure of lipophilicity in quantitative structure–activity relationships [3,47]. Data base searches of system constant ratios affords the tools for identification of suitable chromatographic models to study biopartitioning processes in a straightforward manner, eliminating the need for inspired guesswork. Data base searches can be made across techniques adding to the flexibility of this approach. For example, data base searches indicated that several microemulsion and micellar electrokinetic chromatography systems would provide suitable models for determining octanol–water partition coefficients but few reversed-phase liquid chromatographic systems would be suitable for this measurement, accounting for the difficulty of identifying reversed-phase liquid chromatographic techniques for this purpose [3,59]. Searching across available data bases also indicates that in general, the selectivity of micellar electrokinetic chromatography and reversed-phase liquid chromatography are largely complementary with limited overlap. Strong similarities exist between the separation properties of immobilized artificial membrane columns and micellar electrokinetic chromatography [60] but not between conventional octadecylsiloxane-bonded silica columns and micellar electrokinetic chromatography.
3.8 Conclusions The solvation parameter model is a useful tool for characterizing selectivity of pseudostationary phases for EKC. A comparison of system constants allows redundant
SELECTIVITY CHARACTERIZATION OF PSEUDOSTATIONARY PHASES
75
pseudostationary phases with similar separation properties to be identified for elimination and a manageable number of pseudostationary phases with suitable selectivity differences identified for selectivity optimization in method development [4,12,22,23]. Selectivity optimization is facilitated by simulation for solutes with known or estimated descriptors for all pseudostationary phases with established system constants. The retention factors are estimated by simple calculations and the predicted selectivity factors ranked in order of usefulness or displayed as a window diagram. For mixed micelles or micelle buffers containing additives retention factors, and therefore selectivity factors, can be simulated as a continuous function of the second component to optimize the composition of the pseudostationary phase for target compound separations [4,25,34–37]. A powerful feature of the solvation parameter model for method development is that it allows simulation of selectivity by gas, liquid and electrokinetic chromatography for target compounds with equal facility, thus allowing the unbiased selection of the preferred separation mode. Beyond method development, the solvation parameter model provides insight into the physical processes responsible for retention and is able to contribute to the development of new pseudostationary phases with desirable properties to extend the selectivity range of available materials.
References [1] M.H. Abraham, A. Ibrahim and A.M. Zissimos. Determination of sets of solute descriptors from chromatographic measurements, J. Chromatogr. A., 1037, 29–47 (2004). [2] C.F. Poole, The Essence of Chromatography, Elsevier, Amsterdam, 2003. [3] S.K. Poole and C.F. Poole. Separation methods for estimating octanol–water partition coefficients, J. Chromatogr. B, 797, 3–19 (2003). [4] C.F. Poole and S.K. Poole. Column selectivity from the perspective of the solvation parameter model, J. Chromatogr. A., 965, 263–299 (2002). [5] C.F. Poole, A.D. Gunatilleka and R. Sethuraman. Contributions of theory to method development in solid-phase extraction, J. Chromatogr. A., 885, 17–39 (2000). [6] M.H. Abraham, C.F. Poole and S.K. Poole. Classification of stationary phases and other materials by gas chromatography. J. Chromatogr. A., 842, 79–114 (1999). [7] S. Yang and M.G. Khaledi. Chemical selectivity in micellar electrokinetic chromatography: characterization of solute–micelle interactions for classification of surfactants, Anal. Chem., 67, 499–510 (1995). [8] S. Yang, J.G. Bumgarner and M.G. Khaledi. Chemical selectivity in micellar electrokinetic chromatography II. Rationalization of elution patterns in different surfactant systems, J. Chromatogr. A., 738, 265–274 (1996). [9] P.G. Muijselaar, H.A. Claessens and C.A. Cramers. Characterization of pseudostationary phases in micellar electrokinetic chromatography by applying linear solvation energy relationships and retention indexes, Anal. Chem., 69, 1184–1191 (1997). [10] Z. Liu, H. Zou, M. Ye, J. Ni and Y. Zhang. Effect of organic modifiers on retention mechanism and selectivity in micellar electrokinetic capillary chromatography studied by linear solvation energy relationships, J. Chromatogr. A., 863, 69–79 (1999). [11] S.K. Poole and C.F. Poole. Characterization of surfactant selectivity in micellar electrokinetic chromatography, Analyst, 122, 267–274 (1997). [12] C.F. Poole, S.K. Poole and M.H. Abraham. Recommendations for the determination of selectivity in micellar electrokinetic chromatography, J. Chromatogr. A., 798, 207–222 (1998).
76
ELECTROKINETIC CHROMATOGRAPHY
[13] M.D. Trone and M.G. Khaledi. Statistical evaluation of linear solvation energy relationship models used to characterize chemical selectivity in micellar electrokinetic chromatography, J. Chromatogr. A., 886, 245–257 (2000). [14] M.H. Abraham and J.C. McGowan. The use of characteristic volumes to measure cavity terms in reversed-phase liquid chromatography, Chromatographia, 23, 243–246 (1987). [15] M.H. Abraham. Scales of solute hydrogen-bonding: Their construction and application to physicochemical and biochemical processes, Chem. Soc. Revs., 22, 73–83 (1993). [16] M.H. Abraham, G.S. Whitting, R.M. Doherty and W.J. Shuely. Hydrogen bonding. Part 13. A new method for the characterization of GLC stationary phases – the Laffort data set, J. Chem. Soc. Perkin Trans 2, 1451–1460 (1990). [17] J.A. Platts, D. Butina, M.H. Abraham and A. Hersey. Estimation of molecular free energy relation descriptors using a group contribution approach, J. Chem. Inf. Comput. Sci., 39, 835– 845 (1999). [18] M.H. Abraham, P.L. Grellier, D.V. Prior, P.P. Duce, J.J. Morris and P.J. Taylor. Hydrogen bonding, Part 7. A scale of solute hydrogen-bond acidity based on log K values for complexation in tetrachloromethane, J. Chem. Soc. Perkin Trans. 2, 699–711 (1989). [19] M.H. Abraham, P.L. Grellier, D.V. Prior, J.J. Morris and P.J. Taylor. Hydrogen bonding, Part 10. A scale of solute hydrogen-bond basicity using log K values for complexation in tetrachloromethane, J. Chem. Soc. Perkin Trans. 2, 521–529 (1990). [20] A.M. Zissimos, M.H. Abraham, M.C. Barker, K.J. Box and K.Y. Tam. Calculation of Abraham descriptors from solvent–water partition coefficients in four different systems; evaluation of different methods of calculation, J. Chem. Soc. Perkin Trans. 2, 470–477 (2002). [21] M.A. Garcia, M.L. Marina and J.C. Diez-Masa. Determination of solute–micelle association constants for a group of benzene derivatives and polycyclic aromatic hydrocarbons with sodium dodecyl sulfate by micellar electrokinetic chromatography, J. Chromatogr. A., 732, 345–359 (1996). [22] E. Fuguet, C. Rafols, E. Bosch, M.H. Abraham and M. Roses. Solute–solvent interactions in micellar electrokinetic chromatography III. Characterization of the selectivity of micellar electrokinetic chromatography systems, J. Chromatogr. A., 942, 237–248 (2002). [23] C.F. Poole and S.K. Poole. Interphase model for retention and selectivity in micellar electrokinetic chromatography, J. Chromatogr. A., 792, 89–104 (1997). [24] B.J. Herbert and J.G. Dorsey. n-Octanol–water partition coefficient estimation by micellar electrokinetic chromatography, Anal. Chem., 67, 744–749 (1995). [25] M. Roses, C. Rafols, E. Bosch, A.M. Martinez and M.H. Abraham. Solute–solvent interactions in micellar electrokinetic chromatography. Characterization of sodium dodecyl sulfate-Brij 35 micellar systems for quantitative structure–activity relationship modelling, J. Chromatogr. A., 845, 217–226 (1999). [26] W.E. Wall, D.J. Allen, K.D. Denson, G.I. Love and J.T. Smith. Exploration of alkyl polyols as ‘class I’ organic modifiers to adjust selectivity in micellar electrokinetic capillary chromatography, Electrophoresis, 20, 2390–2399 (1999). [27] M.D. Trone and M.G. Khaledi. Influence of ester and amide-containing surfactant headgroups on selectivity in micellar electrokinetic chromatography, Electrophoresis, 21, 2390–2396 (2000). [28] M.D. Trone and M G. Khaledi. Characterization of chemical selectivity in micellar electrokinetic chromatography, 4. Effect of surfactant headgroup, Anal. Chem. 71, 1270–1277 (1999). [29] M.F. Vitha and P.W. Carr. A linear solvation energy relationship study of the effects of surfactant chain length on the chemical interactions governing retention and selectivity in micellar electrokinetic capillary chromatography using sodium alkyl sulfate elution buffers, Sep. Sci. Technol., 33, 2075–2100 (1998).
SELECTIVITY CHARACTERIZATION OF PSEUDOSTATIONARY PHASES
77
[30] C. Fujimoto. Application of linear solvation energy relationships to polymeric pseudostationary phases in micellar electrokinetic chromatography, Electrophoresis, 22, 1322–1329 (2001). [31] W. Shi, D.S. Peterson and C.P. Palmer. Effect of pendent chains and backbone functionalities on the chemical selectivity of sulfonated amphiphilic copolymers as pseudo-stationary phases in electrokinetic chromatography, J. Chromatogr. A., 924, 123–135 (2001). [32] M.D. Trone, J.P. Mack, H.P. Goodell and M.G. Khaledi. Characterization of chemical selectivity in micellar electrokinetic chromatography VI. Effects of surfactant counter-ion, J. Chromatogr. A., 888, 229–240 (2000). [33] M.D. Trone and M.G. Khaledi. Characterization of chemical selectivity in micellar electrokinetic chromatography, V. The effect of the surfactant hydrophobic chain, J. Microcol. Sep., 12, 433–441 (2000). [34] S.K. Poole and C.F. Poole. Variation of selectivity with composition for a mixed-micellar buffer in micellar electrokinetic chromatography, J. High Res. Chromatogr. 20, 174–178 (1997). [35] S.K. Poole and C.F. Poole. Influence of composition on the selectivity of a mixed-micellar buffer in micellar electrokinetic chromatography, Anal. Commun., 34, 57–62 (1997). [36] E. Fuguet, C. Rafols, E. Bosch, M. Roses and M.H. Abraham. Solute–solvent interactions in micellar electrokinetic chromatography. Selectivity of lithium dodecyl sulfate–lithium perfluorooctanesulfonate mixed-micellar buffers, J. Chromatogr. A., 907, 257–265 (2001). [37] E. Fuguet, C. Rafols, J.R. Torres-Lapasio, M.C. Garcia-Alvarez-Coque, E. Bosch and M. Roses. Solute–solvent interactions in micellar electrokinetic chromatography, 6. Optimization of the selectivity of lithium dodecyl sulfate–lithium perfluorooctanesulfonate mixed micellar buffers, Anal. Chem., 74, 4447–4455 (2002). [38] K. Sarmini and E. Kendler. Influence of organic solvent on separation selectivity in capillary electrophoresis, J. Chromatogr. A., 792, 3–12 (1997). [39] R.M. Selfar, J.G. Kraak and W. Th. Kok. Mechanism of electrokinetic separations of hydrophobic compounds with sodium dodecyl sulfate in acetonitrile–water mixtures, Anal. Chem., 69, 2772–2778 (1997). [40] S. Katsuta and K. Saitoh. Control of separation selectivity in micellar electrokinetic chromatography by modification of the micellar phase with solubilized organic compounds, J. Chromatogr. A., 780, 165–178 (1997). [41] C.F. Poole. Relationship between liquid–liquid distribution and liquid–micelle distribution systems, J. Chromatogr. A., 807, 307–310 (1998). [42] S.H. Hansen, C. Gabel-Jensen and D.W.M. El-Sherbiny. Microemulsion electrokinetic chromatography – or solvent-modified micellar electrokinetic chromatography?, Trends Anal. Chem., 20, 614–619 (2001). [43] K.D. Altria. Application of microemulsion electrokinetic chromatography to the analysis of a wide range of pharmaceuticals and excipients, J. Chromatogr. A., 844, 371–386 (1999). [44] C. Gabel-Jensen, S.H. Hansen and S. Pedersen-Bjergaard. Separation of neutral compounds by microemulsion chromatography: Fundamental studies on selectivity, Electrophoresis, 22, 1330–1336 (2001). [45] S. Pedersen-Bjergaard, C. Gabel-Jensen and S.H. Hansen. Selectivity in microemulsion electrokinetic chromatography, J. Chromatogr. A., 897, 375–381 (2000). [46] M.H. Abraham, C. Treiner, M. Roses, C. Rafols and Y. Ishihama. Linear free energy relationship analysis of microemulsion electrokinetic chromatographic determination of lipophilicity, J. Chromatogr. A., 752, 243–249 (1996). [47] S.K. Poole, D. Durham and C. Kibbey. Rapid method for estimating the octanol–water partition coefficient (log Pow) by microemulsion electrokinetic chromatography, J. Chromatogr. B, 745, 117–126 (2000).
78
ELECTROKINETIC CHROMATOGRAPHY
[48] W.L. Klotz, M.R. Schure and J.P. Foley. Rapid estimation of octanol–water partition coefficients using synthesized vesicles in electrokinetic chromatography, J. Chromatogr. A, 962, 207–219 (2002). [49] C.P. Palmer. Recent progress in the development, characterization and application of polymeric pseudophases for electrokinetic chromatography, Electrophoresis, 23, 3993–4003 (2002). [50] I. Peric and E. Kenndler. Recent developments in capillary electrokinetic chromatography with replaceable charged pseudostationary phases or additives, Electrophoresis, 24, 2924–2934 (2003). [51] K. Gogova, B. Marchel, B. Gas and E. Kenndler. Electrokinetic chromatography with micelles, polymeric and monomeric additives with similar chemical functionality as pseudostationary phases, J. Chromatogr. A., 916, 79–87 (2001). [52] W. Shi and C.P. Palmer. Effect of pendent group structures on the chemical selectivity and performance of sulfonated copolymers as novel pseudophases in electrokinetic chromatography, Electrophoresis, 23, 1285–1295 (2002). [53] D.S. Peterson and C.P. Palmer. Novel alkyl-modified anionic siloxanes as pseudostationary phases for electrokinetic chromatography: II. Selectivity studied by linear solvation energy relationships, Electrophoresis, 22, 3562–3566 (2001). [54] S. Schutte and C.P. Palmer. Alkyl-modified siloxanes as pseudostationary phases for electrokinetic chromatography, Electrophoresis, 24, 978–983 (2003). [55] M.S. Leonard and M.G. Khaledi. A mixed ionic block copolymer–surfactant pseudostationary phase in micellar electrokinetic chromatography, J. Sep. Sci., 25, 1019–1026 (2002). [56] M.H. Abraham, J.M.R. Gola, R. Kumarsingh, J.E. Cometto-Muniz and W.S. Cain. Connection between chromatographic data and biological data, J. Chromatogr. B, 745, 103–115 (2000). [57] C.F. Poole, A.D. Gunatilleka and S.K. Poole. In search of a chromatographic model for biopartitioning, Adv. Chromatogr., 40, 159–230 (2000). [58] D.M. Cimpean and C.F. Poole. Systematic search for surrogate chromatographic models of biopartitioning processes, Analyst, 127, 724–729 (2002). [59] N.C. Dias, M.I. Nawas and C.F. Poole. Evaluation of a reversed-phase column (Supelcosil LCABZ) under isocratic and gradient elution conditions for estimating octanol–water partition coefficients, Analyst, 128, 427–432 (2003). [60] C. Lepont and C.F. Poole. Retention characteristics of an immobilized artificial membrane column in reversed-phase liquid chromatography, J. Chromatogr. A., 946, 107–124 (2002).
4 General Aspects of Resolution Optimization with Micellar Pseudostationary Phases Li Jia and Shigeru Terabe
4.1 Introduction Electrokinetic chromatography (EKC) is one branch of the capillary electromigration separation techniques. An EKC mode with a micellar solution as pseudostationary phase is called micellar electrokinetic chromatography (MEKC), developed by Terabe et al. in 1984 [1]. MEKC is effective for the separation of small molecules including neutral analytes. The separation mechanism is based on partitioning of the analyte between the micellar phase and the surrounding aqueous phase [2]. The most effective way to obtain high resolution in MEKC is to increase the selectivity factor, as in conventional chromatography. The selectivity factor in MEKC depends on the molecular structure of the micelle and hence on the surfactant used, the pH of solution, and the nature of any additives to the micellar solution. The hydrophilic moieties of surfactant molecules generally affect selectivity more than do the hydrophobic moieties. Mixed micelles consisting of ionic and nonionic surfactants display different selectivity from that of single ionic micelles. Additives such as cyclodextrins, ion-pair reagents, urea and organic solvents can also serve as useful modifiers of the micellar solution for enhancing resolution.
4.2 Resolution Equation MEKC can be performed by dissolving an ionic surfactant in the separation electrolyte solution at a concentration higher than the critical micelle concentration (CMC) with no Electrokinetic Chromatography: Theory, Instrumentation and Applications Edited by U. Pyell # 2006 John Wiley & Sons, Ltd. ISBN: 0-470-87102-4
80
ELECTROKINETIC CHROMATOGRAPHY
instrumental modifications compared to capillary electrophoresis (CE). Under the capillary electrophoretic conditions, the ionic micelle migrates at a different velocity from the bulk solution because the micelle is subjected to electrophoretic migration. Under neutral or alkaline conditions, the electroosmotic velocity is faster than the electrophoretic velocity of the micelle, and hence the micelle also migrates in the same direction as the electroosmotic flow (EOF). When an anionic micelle such as sodium dodecyl sulfate (SDS) is employed, all the neutral analytes migrate toward the cathode due to the strong EOF. The neutral analyte free from the micelle migrates only by the EOF, while the analyte totally incorporated by the micelle migrates at the observed velocity of the micelle (the sum of the electroosmotic velocity and the electrophoretic velocity of the micelle). Other neutral analytes are detected between t0 and tmc, which are the migration times of the EOF marker and the micelle, respectively. The interval between t0 and tmc is called the migration time window. The wider the window, the larger the peak capacity, which is the number of peaks that can be separated during a run. The parameters t0 and tmc can be experimentally determined by injecting acetone, formamide or methanol, which are assumed not to interact with the micelle, and lipophilic dyes Sudan III or Sudan IV, or hydrophobic cations such as timepidium or halofantrine, which are assumed to be totally solubilized into the micelle, respectively. In MEKC the retention factor, k, is defined as: k¼
nmc naq
ð4:1Þ
where nmc is the amount of substance of analyte incorporated into the micelles and naq represents the amount of substance of free molecules in the aqueous phase. For a neutral analyte migrating between t0 and tmc with a migration time (tR), the retention factor for the analyte can be calculated by the following equation derived by Terabe et al. [1]: tR t0 ð4:2Þ k¼ t0 ð1 tR =tmc Þ The fact that neutral analytes must migrate between t0 and tmc is the most significant difference between MEKC and conventional chromatography. If tmc becomes infinite (micellar phase becomes stationary), Equation (4.2) reduces to the analogous equation for conventional chromatography. The resolution Rs in MEKC is given by [3]: pffiffiffiffi N a1 k2 1 t0 =tmc ð4:3Þ Rs ¼ 4 1 þ k2 1 þ ðt0 =tmc Þk1 a where N is the theoretical plate number, k1 and k2 are retention factors of Analytes 1 and 2, respectively, and a is the selectivity factor given by k2/k1. Equation (4.3) differs from that for conventional chromatography by virtue of the last term on the right-hand side, which arises from the limited migration time window. The migration of the micelle inside the capillary causes reduction of the virtual column length [4]. If the micelle migration is completely suppressed or tmc is infinity, the last term equates to unity and the resolution equation is identical to that of conventional chromatography. Equation (4.3) shows that resolution is related to N, a, k and t0 =tmc . Resolution increases in proportion to the square root of N. In MEKC, as in capillary zone
GENERAL ASPECTS OF RESOLUTION OPTIMIZATION
81
electrophoresis (CZE), the higher the applied voltage, the higher the N value and the higher the resolution, unless the temperature increase is too high. Usually a high voltage of 10–30 kV is applied to perform the MEKC separation and 100 000–300 000 theoretical plates are obtained with a 50 cm 0.05 mm i. d. capillary tube within a relatively short time. If N is significantly lower than that, the experimental conditions must be reconsidered. The most probable cause of low N is adsorption of the analyte onto the capillary wall. If this happens, the capillary must be rinsed thoroughly. The retention factor k is important in increasing the resolution in MEKC. The optimum k value to give the highest resolution is determined by the following equation [5]: pffiffiffiffiffiffiffiffiffiffiffiffi ð4:4Þ kopt ¼ tmc =t0 Under neutral or alkaline conditions, tmc/t0 is 3–4; hence, kopt is 1.7–2.0. Large retention factors mean that a high molar fraction of the analyte is incorporated into the micelle, and small values mean the reversed situation or that a high molar fraction of the analyte exists in the aqueous phase. The result given by Equation (4.4) suggests that the retention factors should not be close to either extreme for separation by MEKC. In general, retention factors between 1 and 5 are recommended for optimal resolution. The desired k value in MEKC can be obtained by adjusting the concentration of the surfactant because k can be expressed as: k ¼ KVmc =Vaq ffi KvðCsf CMCÞ
ð4:5Þ
where K is the distribution coefficient of the analyte between the micelle and the aqueous phase, Vmc/Vaq is the phase ratio and Vmc and Vaq are volumes of the micelle and the aqueous phase, respectively, v is the partial specific volume of the micelle, and Csf is the surfactant concentration [3]. Equation (4.5) suggests that k increases linearly with the surfactant concentration. This is a characteristic feature of MEKC since the Csf needed to obtain a given k can be calculated, provided the CMC and k at a certain Csf are known. In contrast to EKC, for reversed-phase liquid chromatography (RPLC), the alteration of the phase ratio is seriously limited. The migration time ratio, t0 =tmc , is directly related to the width of the migration time window. The smaller the value of t0 =tmc , the wider is the migration time window and hence the higher is the Rs value. The value of t0 =tmc is in the range 0.2–0.4 for most ionic micelles under neutral or alkaline conditions. It is necessary to reduce the velocity of the EOF to obtain a smaller value of t0 =tmc . Addition of an organic solvent such as methanol, acetonitrile or 2-propanol, or the use of a nonionic/anionic mixed micelle are useful methods, and changing the pH to acidic is also a possible approach. However, in practice, a longer run time is required. The addition of an organic solvent increases the viscosity of the separation electrolyte and reduces the velocity of EOF and also reduces the velocity of the micelles, resulting in the extension of the migration time window with the side effect of a reduction of the retention factor. The selectivity factor a is the most important and most effective term to increase resolution. The selectivity factor can be manipulated through selection of surfactants and modification of the buffer solution. In RPLC, usually the separation is not manipulated by changing the stationary phase since C18 is the most widely accepted phase and different
82
ELECTROKINETIC CHROMATOGRAPHY
products vary little. In MEKC, however, pseudostationary phases are micelles and the type of surfactant significantly affects the a value. The choice of surfactants and modification of the buffer solution are discussed in more detail below.
4.3 Resolution Optimization The choice of micelle and of modifier added to the aqueous phase is the most effective and important means of enhancing resolution. Another key parameter that affects resolution is temperature. Each experimental factor is discussed separately below. 4.3.1
Choice of the Micelle
4.3.1.1 Ionic Surfactants. For the separation of neutral analytes, the micelle used in MEKC must be ionic, that is, it must consist of an ionic surfactant or a mixture of ionic and nonionic surfactants. Some typical surfactants with their CMC and aggregation numbers are listed in Table 4.1. Surfactants have a hydrophobic and a hydrophilic group within each molecule, and both groups affect selectivity in MEKC. SDS is the most widely used surfactant in MEKC; it possesses a long alkyl chain as the hydrophobic group and an ionic group as the polar group. The micelle formed by SDS is believed to be spherical in shape, with the polar groups being located in the outer zone of the micelle and the alkyl groups constituting a hydrophobic core, as shown in Figure 4.1. When an Table 4.1 Aggregation number (AN) and CMC of some typical surfactants [2] Surfactant Anionic
SDS Sodium tetradecyl sulfate Sodium decanesulfonate Sodium N-lauroyl-N-methyltaurate Sodium polyoxyethylene dodecyl ether sulfate Sodium N-dodecanoyl-L-valinate Sodium cholate Sodium deoxycholate Sodium taurocholate Sodium taurodeoxycholate Potassium perfluoroheptanoate Cationic Tetradecyltrimethylammonium bromide Dodecyltrimethylammonium bromide Cetyltrimethylammonium bromide Cetyltrimethylammonium chloride Non-ionic Polyoxyethylene(23) dodecyl ether (Brij35) Polyoxyethylene(23) sorbitan monolaurate (Tween 20) Zwitterionic 3-[(3-cholamidopropyl)dimethylammonio]1-propanesulfonate N-Dodecyl-N,N-dimethylammonio3-propanesulfonate
CMC (at 25 C; 103 M)
AN
8.1 2.1 (50 C) 40 8.7 2.8 5.7 (40 C) 13–15 4–6 10–15 2–6 28 3.5 15 0.92 1.3 0.1
62 138 40 NA 66 NA 2–4 4–10 5 NA NA 75 56 61 NA NA
0.059
NA
4.2–6.3
10
3.3
NA
GENERAL ASPECTS OF RESOLUTION OPTIMIZATION A
83
B
C
Figure 4.1 Schematic diagram of the structure of an ionic micelle and the interaction between three types of analyte and the ionic micelle. A, on the surface; B, as a cosurfactant; C, within the core. (Reprinted with permission from Ref. [2], copyright 2004 American Chemical Society)
analyte is incorporated into the micelle, three types of interaction are possible as shown in Figure 4.1: A, the analyte is adsorbed onto the surface of the micelle by electrostatic or dipole interaction; B, the analyte behaves as a cosurfactant by participating in the formation of the micelle; C, the analyte is incorporated into the core of the micelle. The effect of the surfactant’s molecular structure on the separation selectivity will differ according to the type of interaction involved. The hydrophilic, or ionic group, is generally more important in determining selectivity than is the hydrophobic group since most analytes interact with the micelle at the surface. Figure 4.2 gives an example of the comparison of selectivity between surfactants possessing different polar groups [6]. The result as shown in Figure 4.2 indicates that the different polar groups of various surfactants will show different selectivity for analytes, even if the surfactants have identical alkyl chain groups. The influence of surfactant type on the separation selectivity can be investigated through linear solvation energy relationships (LSER) [7]. Using the LSER methodology, useful information about the nature of solute interactions with different types of surfactant aggregates can be obtained (see Chapter 3). One result of these studies is that selectivity differences in MEKC between several anionic surfactants are primarily due to hydrogen-bonding interactions rather than the dipolar interactions [7]. Cationic micelles show substantially different selectivity for neutral and for ionic solutes, as compared with anionic micelles because of the different polar group on this surfactant [8]. Most cationic surfactants have an alkyltrimethylammonium group and their counter ions are halides. Cetyltrimethylammonium bromide (CTAB) is the most popular cationic surfactant used in CE, mainly to reverse the direction of EOF due to adsorption of the cationic surfactant on the capillary wall. However, it is not widely used
84
ELECTROKINETIC CHROMATOGRAPHY
Figure 4.2 Comparison of selectivity between surfactants possessing different polar groups: 1, caffeine; 2, acetaminophen; 3, sulpyrin; 4, trimetoquinol; 5, guaiphenesin; 6, naproxen; 7, ethenzamide; 8, phenacetin; 9, isopropylantipyrine; 10, noscapine; 11, chlorpheniramine; 12, tipepidine. Conditions for separation solution: (a) 0.1 M sodium trioxyethylenealkyl (C12–C14) ether acetate (ECT) in 20 mM phosphate–borate buffer (pH 9.0); (b) 0.1 M SDS in the same buffer as for (a). Capillary, 50 mm i.d. 65 cm (effective length 50 cm); applied voltage, 20 kV; detection, UV absorbance at 210 nm. (From Nishi et al., Effect of surfactant on the separation of cold medicine ingredients by micellar electrokinetic chromatography, J. Pharm. Sci., 79, 519–523, copyright 1990 John Wiley & Sons, Inc., reprinted with permission)
as a micelle-forming surfactant in MEKC, probably due to its UV absorption in the short wavelength region and the generation of bromine in the anodic vial during electrophoresis. However, CTAB or the corresponding chloride (CTAC) shows significantly different selectivities compared to anionic surfactants while the migration order of analytes still follows the order of increasing distribution constant, as in the case of anionic micelles. According to studies using the solvation parameter model, CTAB differs significantly from SDS concerning its hydrogen-bond acidity and basicity (see Chapter 3). Therefore, the use of a cationic surfactant instead of SDS is a promising alternative to change the selectivity. Figure 4.3 shows examples of the effect of the type of surfactant (anionic or cationic surfactant) on selectivity [9]. 4.3.1.2 Nonionic Surfactants. Nonionic surfactants themselves do not posses electrophoretic mobility and cannot be used as pseudostationary phases in conventional MEKC. However, nonionic surfactant micelles are useful for the separation of charged compounds, especially for peptides with closely related structures [10]. Since the separation principle is the same as with ionic surfactants, we can classify the technique with nonionic micelles as being an extension of MEKC. This technique was successfully applied to the mass spectrometric detection of peptides separated by MEKC with nonionic surfactants [11]. Nonionic surfactants can also be employed as pseudostationary phases in MEKC with a combination of ionic surfactants. Most nonionic surfactants have polyoxyethylene groups. Mixed micelles of nonionic and ionic surfactants with surfaces covered by polyoxyethylene groups, as shown in Figure 4.4, will have different surface characteristics and hence selectivity from that of the ionic surfactant micelle. Some natural chiral surfactants are uncharged and were used as mixed micelles with SDS.
GENERAL ASPECTS OF RESOLUTION OPTIMIZATION
85
Figure 4.3 Effect of the type of surfactant (anionic or cationic surfactant) on selectivity: 1, water; 2, aniline; 3, nitrobenzene; 4, m-nitroaniline; 5, p-nitroaniline; 6, o-nitroaniline. Conditions: capillary, 50 mm i.d. 65 cm (effective length 50 cm); separation solution: (a) 100 mM SDS in 50 mM phosphate–100 mM borate buffer (pH 7.0); (b) 50 mM CTAB in 100 mM Tris-HCl (pH 7.0); applied voltage, 20 kV; detection, UV absorbance at 210 nm [9]. (Reprinted with permission from Terabe in Capillary Electrophoresis Technology, N. Guzman (Ed.), copyright CRC Press, Boca Raton, Florida)
Figure 4.4 Schematic diagram of the structure of mixed micelle of ionic and nonionic surfactants and the interaction between an analyte and the micelle on the surface. (Reprinted with permission from Ref. [2], copyright 2004 American Chemical Society)
86
ELECTROKINETIC CHROMATOGRAPHY
4.3.1.3 Biological Surfactants–Bile Salts. Bile salts which are a group of natural steroidal surfactants are common biological surfactants used in MEKC as alternative surfactants. Some of the bile salts, which have been employed in MEKC separations, include sodium cholate (SC), sodium taurocholate (STC), sodium deoxycholate (SDC), and sodium taurodeoxycholate (STDC). Nonconjugated bile salts, such as SC and SDC, can be used under neutral or alkaline conditions to keep the carboxyl group ionized, whereas taurine-conjugated forms, such as STC and STDC, can be used in more acidic conditions above pH 3. It should be noted that the deoxycholate solution tends to gelate, when a relatively high concentration buffer solution is employed to prepare the micellar solution. These surfactants are considered to form a helical micelle with a reversed micelle conformation [12]. Bile salts display two characteristic features that differentiate them from long alkyl-chain surfactants: they have a relatively low solubilizing power, and they are capable of chiral discrimination. Highly hydrophobic compounds are not usually well separated by MEKC with long alkyl-chain surfactants, because of excessively large retention factors. However, bile salt micelles have been successfully employed to solve the problem by taking advantage of their low solubilizing effect [12]. Bile salts are chiral and can be used for enantiomeric separations [13]. In this respect, deoxycholate and its taurine conjugate have shown successful results. 4.3.1.4 Micelle Polymers. Micelle polymers and high molecular mass surfactants have been investigated as pseudostationary phase in MEKC [14–16]. These additives form monomolecular micelles (aggregation number of 1), so that an enhanced stability and rigidity of the micelle can be obtained as well as easier control of the micellar size as compared with micelles formed by a conventional low molecular mass surfactant. Since the CMC of the micelle polymer is essentially zero, the net micellar concentration is constant or independent of the concentration and composition of a buffer, pH, additives and temperature, and hence better reproducibility in the migration time can be expected in micelle polymer-MEKC than in low molecular mass surfactant-MEKC. In addition, micelle polymer-MEKC is suitable to be adapted for on-line coupling of MEKC with mass spectrometry (MS) (see Chapter 15). 4.3.1.5 Mixed Micelles. It is known that mixed micelles are formed when more than two different surfactants are dissolved in a solution. Mixed micelle consisting of an ionic and a nonionic surfactant are particularly useful in MEKC. Because such mixed micelles have a larger size and lower surface-charge density than ionic micelles, they provide higher retention factors and a narrower migration time window and also show a different separation selectivity [17]. As described above, most nonionic surfactants have polyoxyethylene groups as hydrophilic moieties, therefore the surface structure of the mixed micelle and hence its separation selectivity will be significantly different from that of an ionic micelle. An example is given in Figure 4.5 [18]. Zwitterionic surfactants are not widely used in MEKC. However, zwitterionic surfactants will be interesting if they are used in mixed micelles, because they should show significantly different selectivities compared to other types of surfactants. 4.3.2
Choice of Buffer Solution
In general, the constituents of the buffer, in which the pseudostationary phase is dissolved do not significantly affect resolution, but the pH of the buffer is an important factor in
GENERAL ASPECTS OF RESOLUTION OPTIMIZATION
87
Figure 4.5 Effect of addition of nonionic surfactant: 7, acetaminophen; 8, caffeine; 9, guaiphenesin; 10, ethenzamide; 11, isopropylantipyrine; 12, trimetoquinol. Conditions: separation solution: (a) 100 mM SDS in 50 mM phosphate–100 mM borate buffer (pH 7.0); (b) 30 mM Tween 60 added to the same solution as in (a); capillary, 75 mm i.d. 57 cm (effective length 50 cm); applied voltage, 18 kV; detection, UV absorbance at 214 nm [18]. (Reprinted from J. Pharm. Biomed. Anal., 10, 705–715, S. Terabe, Selectivity manipulation in micellar electrokinetic chromatography, copyright 1992, with permission from Elsevier)
manipulating resolution of ionizable analytes. An ionized form of the analyte having the same charge as the micelle will be incorporated into the micelle to a lesser extent than its unionized form. The retention factor of an acid is the weighted average of the retention factors of its undissociated (HA) and its dissociated (A) forms. Thus, the distribution coefficient, migration time and resolution can be substantially altered by changing the pH for ionizable analytes [18,20]. The pH of the separation electrolyte is mainly optimized empirically. 4.3.3
Choice of Temperature
The distribution coefficient is dependent on temperature, and an increase in temperature causes reduction of the migration time because of the decrease in the distribution coefficient. The temperature rise also results in increases of the velocity of EOF and micelle by the same extent owing to the lowered viscosity of the separation electrolyte. The dependence of the distribution coefficient on temperature is different among analytes. Therefore, temperature will also affect resolution [21]. Although if the temperature does not have a significant impact on selectivity and resolution it will seriously affect the migration time. Therefore, it is essential for the reproducibility of results to keep temperature precisely constant. 4.3.4
Choice of Additives to the Aqueous Phase
The aqueous phase in MEKC corresponds to the mobile phase in RPLC and therefore the various mobile phase modifiers developed in RPLC are also applicable in MEKC. So far, the following additives, organic solvents, cyclodextrins (CDs), ion-pair reagents, and urea are particularly useful. 4.3.4.1 Organic Solvents. Water-miscible organic solvents such as methanol and acetonitrile can be used in MEKC, just as they are in RPLC. These organic solvents reduce
88
ELECTROKINETIC CHROMATOGRAPHY
the retention factor and alter separation selectivity. A high concentration of organic solvent may break down the micellar structure and therefore it is recommended that the volume fraction of the solvent should not exceed 20 %. The addition of the organic solvent to a micellar solution usually reduces the electroosmotic velocity, thereby extending the migration time window and increasing resolution, probably resulting from the reduced charge on the micelle or the increased size of the micelle due to swelling caused by the organic solvent. A solvent programming operation, where the content of the organic solvent is changed with time, has been developed to improve resolution [22,23]. 4.3.4.2 Cyclodextrins. CDs are now popular in the field of chromatography. Most of the techniques using CD are based on its capability to recognize specific molecules that fit its hydrophobic cavity. From the viewpoint of MEKC, native CDs are electrically neutral and have no electrophoretic mobility, but CD derivatives with ionic group substituents can be used as a pseudostationary phase instead of a micellar separation carrier in EKC. This is especially effective for the separation of a number of neutral compounds, including aromatic isomers and racemic mixtures [24,25]. CD has also been used as an additive to micellar solutions for CD-modified MEKC (CD/MEKC) [26]. When CD is added to a micellar solution, the analyte is distributed among three phases: micelle, CD and water (aqueous phase). From the viewpoint of electrophoretic separation, CD and water migrate at identical velocities, equal to the EOF velocity. However, CD exerts a remarkable effect on the apparent retention factor between the micellar and nonmicellar (i.e. CD and water) phases. Highly hydrophobic compounds tend to be almost totally incorporated into the micelle because of their low solubilities in water. CD is water soluble and capable of including hydrophobic compounds into its hydrophobic cavity. The inclusion complex formation equilibrium constant for an analyte depends on steric parameters. Thus a fraction of the hydrophobic analyte will be included into the CD cavity, even in the presence of micelles. Therefore
Figure 4.6 CD/MEKC separation of eight corticosteroids: a, hydrocortisone; b, hydrocortisone acetate; c, betamethasone; d, cortisone acetate; e, triamcinolone acetonide; f, fluocinolone; g, dexamethasone acetate; h, fluocinonide. Conditions: separation solution. (a) 50 mM SDS in 20 mM phosphate–borate buffer (pH 9.0); (b) 15 mM g-CD and 4 M urea added to the same buffer solution as in (a); capillary, 50 mm i.d. 65 cm (effective length 50 cm); applied voltage, 20 kV; detection, absorbance at 210 nm [27]. (Reprinted with permission from Nishi and Matsuo, J. Liq. Chromatogr., 14, 973–986 (1991), copyright CRC Press, Boca Raton, Florida)
GENERAL ASPECTS OF RESOLUTION OPTIMIZATION
89
the migration time or the apparent retention factor of a hydrophobic analyte that forms an inclusion complex with a CD will decrease with an increase in CD concentration. Separation selectivity among highly hydrophobic analytes depends solely on the difference in their distribution ratio between CD and the micellar phase, because such hydrophobic analytes can be assumed to be insoluble in water. Figure 4.6 shows an example of the separation of highly hydrophobic corticosteroids by CD/MEKC [27]. CD/ MEKC is helpful not only for the separation of highly hydrophobic compounds but also for enantiomeric separations because of the chirality of CD. An example of enantiomeric separation by CD/MEKC is given in Figure 4.7 [28]. 4.3.4.3 Ion-pair Reagents. Ion-pair reagents are often employed in RPLC for the separation of ionic compounds. There is an essential difference in the separation of ionic compounds between MEKC and RPLC although both have close similarities for the
Figure 4.7 CD/MEKC separation of a mixture of five CBI (l-cyano-2-substituted-benz[f]isoindole)-DL-amino acids. Conditions: separation solution, 50 mM SDS, 10 mM g-CD in 100 mM borate buffer (pH 9.0); capillary, 50 mm i.d. 70 cm (effective length 50 cm); applied voltage, 15 kV; detection, laser-induced fluorescence (lex ; ¼ 442 nm, lem ; ¼ 490 nm) (Reprinted with permission from Ref. [28], copyright 1991 American Chemical Society)
90
ELECTROKINETIC CHROMATOGRAPHY
separation of neutral compounds. The difference is due to the electrical charge on the micelle. As the micelle is ionic, ionic analytes having the same charge as the micelle will not be incorporated into the micelle, or if so, then only slightly. On the other hand, ionic analytes having a different charge will strongly interact with the micelle. An ion-pair reagent will strongly modify the interaction between ionic analytes and an ionic micelle mentioned above. For example, a cationic ion-pair reagent such as tetraalkylammonium salts will enhance the interaction between an anionic analyte and an anionic micelle by forming the ion-pair with the analyte or with the micelle. On the other hand, a cationic ion-pair reagent will reduce the interaction between a cationic analyte and an anionic micelle by competing with the analyte for combining with the micelle. The effect of the ion-pair reagent on the migration time or separation selectivity depends on the molecular structure of the ion-pair reagent itself [29]. 4.3.4.4 Urea. Although urea is not often used in RPLC, it is useful for the separation of hydrophobic compounds in MEKC. Urea is very soluble in water up to 7 M. The solution is UV-transparent and the viscosity is not very high. A high concentration of urea is known to increase the solubility of hydrophobic compounds in water as well as to hinder hydrogen bond formation in the aqueous phase. The addition of a high concentration of urea to the SDS solution enabled the MEKC separation of highly hydrophobic compounds, as shown in Figure 4.8 [30]. These steroidal compounds were not successfully resolved with the SDS solution alone because their retention factors were much larger than the optimal value. The high concentration of urea slightly reduced the
Figure 4.8 The effect of urea addition to the SDS solution: a, hydrocortisone; b, hydrocortisone acetate; c, betamethasone; d, cortisone acetate; e, triamcinolone acetonide; f, fluocinolone acetonide; g, dexamethasone acetate; h, fluocinonide. Conditions: separation solution: (a) 50 mM SDS in 20 mM borate–20 mM phosphate buffer (pH 9.0); (b) the same SDS solution as in (a), but containing 6 M urea; capillary, 50 mm i.d. 65 cm (effective length 50 cm); applied voltage, 20 kV; detection, absorbance at 210 nm [30]. (Reprinted from J. Chromatogr., 545, 359–368, Terabe et al., copyright 1991, with permission from Elsevier)
GENERAL ASPECTS OF RESOLUTION OPTIMIZATION
91
EOF velocity and considerably reduced the observed migration velocity of the micelle, resulting in the extension of the migration time window. These observed changes in velocities can not be attributed exclusively to the slight increase in viscosity due to the addition of urea.
Run with a standard MEKC conditions
Yes
Is the separation successful
End or optimize migration times
No Option 2
No
Cationic analyte?
k > 10
Caculate k
k < 0.5
Yes 0.5 < k < 10 Option 1 Use a different surfactant (including a mixed micelle)
Anionic analyte?
No
Increase [SDS]
Yes Option 1
Rs < 0.5
Estimate Rs Rs > 0.5
Optimize k to about 2 and/or use a longer capillary
No
Is the separation successful Yes End
Figure 4.9 Introductory guide to the development of an optimum separation. Standard MEKC conditions: Running solution: 50 mM SDS in 50 mM borate buffer (pH 8.5–9.0); capillary: 50–75 mm i.d. 30–50 cm (from the injection end to the detector); applied voltage: 10–20 kV (keep current below 50 mA); temperature: 25 C or ambient; sample solvent: water or methanol; sample concentration: 0.1–1 mg ml1; injection end: the positive end or anodic end; injection volume: below 2 nl (or less than 1 mm from the end of the capillary); detection: 200–210 nm (depending on the sample). Option 1: use an ion-pair reagent (an ammonium salt) or a cationic surfactant. Option 2: use bile salts instead of SDS, add CD to an SDS solution (CD/MEKC), or add an organic solvent or urea to an SDS solution [18]. (Reprinted from J. Pharm. Biomed. Anal., 10, 705–715, S. Terabe, Selectivity manipulation in micellar electrokinetic chromatography, copyright 1992, with permission from Elsevier)
92
ELECTROKINETIC CHROMATOGRAPHY
4.4 Strategies for Optimizing Resolution The simplest way to reach optimum separation is to follow a guide, as described by Terabe [18]. Such a guide is instructive and helpful for finding preliminary experimental conditions. Figure 4.9 shows a general guide to successful separation by MEKC. A set of standard operating conditions using a 50 mM SDS solution, as given in Figure 4.9, is proposed for the initial stage in developing an analytical method employing MEKC. If a sample consists of a few components, the standard conditions, as shown in Figure 4.9, may be successful. When the resolution is larger than that required, the migration time can be reduced by using a shorter capillary or reducing the concentration of SDS. If the separation is not satisfactory, it is necessary to know the approximate values of the retention factors, according to Equation (4.2). To evaluate the retention factor, t0 and tmc must be measured experimentally (see Chapter 1). The marker or tracer is usually added to the sample solution. The peak of the micelle corresponding to tmc is not always readily observed due to the low solubility of the tracer. In such cases, tmc can be assumed to be three to four times t0. For analytes having retention factors that are too large or too small, either Option 1 or Option 2 are recommended, according to the flow chart. In particular, for hydrophobic compounds, there is a wide range of choices as noted in Option 2. When peaks are partially resolved, further refinement of the conditions will often lead to a successful separation. However, when the separation is unsuccessful with SDS as micellar phase, other surfactants should be explored.
References [1] S. Terabe, K. Otsuka, K. Ichikawa, A. Tsuchiya and T. Ando. Electrokinetic separations with micellar solutions and open-tubular capillaries, Anal. Chem., 56, 111–113 (1984). [2] S. Terabe. Micellar electrokinetic chromatography, Anal. Chem., 76, 240A–246A (2004). [3] S. Terabe, K. Otsuka and T. Ando, Electrokinetic chromatography with micellar solution and open-tubular capillary, Anal. Chem., 57, 834–841 (1985). [4] C.X. Zhang, Z.P. Sun and D.K. Lin. Micellar electrokinetic capillary chromatography theory based on conventional chromatography, J. Chromatogr. A, 655, 309–316 (1993). [5] J.P. Foley. Optimization of micellar electrokinetic chromatography, Anal. Chem., 62, 1302– 1308 (1990). [6] H. Nishi, T. Fukuyama, M. Matsuo and S. Terabe. Effect of surfactant on the separation of cold medicine ingredients by micellar electrokinetic chromatography, J. Pharm. Sci., 79, 519–523 (1990). [7] S. Yang and M.G. Khaledi. Chemical selectivity in micellar electrokinetic chromatography: characterization of solute–micelle interactions for classification of surfactants, Anal. Chem., 67, 499–510 (1995). [8] K. Otsuka, S. Terabe and T. Ando. Electrokinetic chromatography with micellar solutions: Separation of phenylthiohydantoin-amino acids, J. Chromatogr., 332, 219–226 (1985). [9] S. Terabe. Micellar electrokinetic chromatography, in Capillary Electrophoresis Technology, N.A. Guzman (Ed), Marcel Dekker, New York, 1993. [10] N. Matsubara, K. Koezuka and S. Terabe. Separation of eleven angiotensin II analogs by capillary electrophoresis with a nonionic surfactant in acidic media, Electrophoresis, 16, 580– 583 (1995).
GENERAL ASPECTS OF RESOLUTION OPTIMIZATION
93
[11] K. Koezuka, H. Ozaki, N. Matsubara and S. Terabe. Separation and detection of closely related peptides by micellar electrokinetic chromatography coupled with electrospray ionization mass spectrometry using the partial filling technique, J. Chromatogr. B., 689, 3–11 (1997). [12] R.O. Cole, M.J. Sepaniak, W.L. Hinze, J. Gorse and K. Oldiges. Bile salt surfactants in micellar electrokinetic capillary chromatography: application to hydrophobic molecule separations, J. Chromatogr., 557, 113–123 (1991). [13] S. Terabe, M. Shibata and Y. Miyashita. Chiral separation by electronkinetic chromatography with bile salt micelles, J. Chromatogr., 480, 403–411 (1989). [14] H. Ozaki, S. Terabe and A. Ichihara. Micellar electrokinetic chromatography using highmolecular surfactants: use of butyl acrylate–butyl methacrylate–methacrylic acid copolymers sodium salts as pseudo-stationary phases, J. Chromatogr. A., 680, 117–123 (1994). [15] C.P. Palmer and S. Terabe. A novel sulfate polymer as a pseudo-stationary phase for micellar electrokinetic chromatography, J. Microcol. Sep., 8, 115–121 (1996). [16] C.P. Palmer and S. Terabe. Micelle polymers as pseudostationary phases in MEKC: chromatographic performance and chemical selectivity, Anal. Chem., 69, 1852–1860 (1997). [17] H.T. Rasmussen, L.K. Goebel and H. McNair. Micellar electrokinetic chromatography employing sodium alkyl sulfates and Brij 35, J. Chromatogr., 517, 549–555 (1990). [18] S. Terabe. Selectivity manipulation in micellar electrokinetic chromatography, J. Pharm. Biomed. Anal., 10, 705–715 (1992). [19] K. Otsuka, S. Terabe and T. Ando. Electrokinetic chromatography with micellar solutions: separation of phenylthiohydantoin–amino acids, J. Chromatogr., 332, 219–226 (1985). [20] K. Otsuka, S. Terabe and T. Ando. Electrokinetic chromatography with micellar solutions: retention behaviour and separation of chlorinated phenols, J. Chromatogr., 348, 39–47 (1985). [21] S. Terabe, T. Katsura, Y. Okada, Y. Ishihama and K. Otsuka. Measurement of thermodynamic quantities of micellar solubilization by micellar electrokinetic chromatography with sodium dodecyl sulfate, J. Microcol. Sep., 5, 23–33 (1993). [22] M.J. Sepaniak, D.F. Swaile and A.C. Powell. Instrumental developments in micellar electrokinetic capillary chromatography, J. Chromatogr., 480, 185–196 (1989). [23] A.T. Balchunas and M.J. Sepaniak. Gradient elution for micellar electrokinetic capillary chromatography, Anal. Chem., 60, 617–621 (1988). [24] S. Terabe, H. Ozaki, K. Otsuka and T. Ando. Electrokinetic chromatography with 2-Ocarboxymethyl-b-cyclodextrin as a moving ‘stationary’ phase, J. Chromatogr., 332, 211–217 (1985). [25] S. Terabe. Electrokinetic chromatography: an interface between electrophoresis and chromatography, Trends Anal. Chem., 8, 129–134 (1989). [26] S. Terabe, Y. Miyashita, O. Shibata, E.R. Barhart, L.R. Alexander, D.J. Patterson, B.L. Karger, K. Hosoya and N. Tanaka. Separation of highly hydrophobic compounds by cyclodextrinmodified micellar electrokinetic chromatography, J. Chromatogr., 516, 23–31 (1990). [27] H. Nishi and M. Matsuo. Separation of corticosteroids and aromatic hydrocarbons by cyclodextrin-modified micellar electrokinetic chromatography, J. Liq. Chromatogr., 14, 973–986 (1991). [28] T. Ueda, F. Kitamura, R. Mitchel, T. Metcalf, T. Kuwana and A. Nakamoto. Chiral separation of naphthalene-2,3-dicarboxaldehyde-labeled amino acid enantiomers by cyclodextrin-modified micellar electrokinetic chromatography with laser-induced fluorescence detection, Anal. Chem., 63, 2979–2981 (1991). [29] H. Nishi, N. Tsumagari and S. Terabe. Effect of tetraalkylammonium salts on micellar electrokinetic chromatography of ionic substances, Anal. Chem., 61, 2434–2439 (1989). [30] S. Terabe, Y. Ishihama, H. Nishi, T. Fukuyama and K. Otsuka. Effect of urea addition in micellar electrokinetic chromatography, J. Chromatogr., 545, 359–368 (1991).
5 Optimization of the Separation Conditions in Electrokinetic Chromatography: Experimental Designs, Modelling and Validation Olga Jime´nez and Maria Luisa Marina
5.1 Introduction Electrokinetic chromatography (EKC) is a very popular separation technique because of such advantages as high efficiency, short analysis times and small injection volumes required [1–3]. The migration of analytes under the electric field is governed by the electrokinetic properties of the solutes and the partition coefficient differences between a pseudostationary phase and the surrounding aqueous phase. The most popular pseudostationary phases have been micelles (anionic, cationic, and neutral) but other pseudostationary phases have included microemulsions, polymers, vesicles, proteins, resorcarenes, dendrimers and cyclodextrins [2]. The optimization of the conditions of a separation is a difficult task due to the large number of variables affecting the process [4–7]. Among these we can cite pH, nature and concentration of surfactant, nature and concentration of buffer, and nature and concentration of organic modifiers. Furthermore, some instrumental variables, such as the applied voltage and the injection time or volume can also have a great influence on the solute separation. So it is important to follow an adequate strategy to obtain the best conditions for a certain separation. Ideally, an appropriate optimization strategy should
Electrokinetic Chromatography: Theory, Instrumentation and Applications Edited by U. Pyell # 2006 John Wiley & Sons, Ltd. ISBN: 0-470-87102-4
96
ELECTROKINETIC CHROMATOGRAPHY
be able to find good separation conditions in the shortest time with only a few experiments, to assure the baseline separation of all components [8]. To date, optimization strategies can be divided in two great groups: those that employ a univariate optimization, that is, the variables are studied one at a time while keeping the others constant, and those that use a multivariate optimization scheme. Altria et al. [9] and Sentellas and Saurina [5] have reviewed the application of chemometric techniques to the optimization of the separation conditions in CE methods. So, in Table 5.1 the information published has been updated. In this table, the variables (also usually known as factors) and their experimental ranges, the type of model (empirical, theoretical or artificial neural networks (ANNs)) and the response(s), the type of validation, the solutes and the followed strategy are shown. The univariate optimization strategy is only efficient if there is no interaction between the variables. In other situations there is no guarantee of obtaining a global optimum. It must be clear that perhaps the obtained conditions permit a good separation of analytes but these conditions, in most situations, are not the best. This fact has been clearly illustrated by Liu et al. [21]. In their work, a comparison between computer aided simulation optimization and step-by-step (univariate) optimization for variation of a single factor was performed. In both optimization types, solutes were baseline resolved but when computer-aided ANNs simulation optimization was used, the analysis time was approximately reduced fourfold (7.5 minutes vs 30 minutes). Moreover, univariate optimization tends to be time consuming and tedious [4]. Hence the multivariate strategy is highly recommended. This strategy comprises different steps, that is, the use of an experimental design, the modelization step and the model validation, as it is described in the following sections.
5.2 Experimental Designs The objective of an experimental design is to obtain the best conditions for a separation with the minimum number of experiments. As a rule, the factors determining whether an experimental design will be successful or not are [23]: (i) the choice of parameters to optimize; (ii) the interval within which parameters are varied, and (iii) the choice of response used in the optimization. In this section, first, we will try to get an insight into the factors to be taken into account in an optimization scheme. It would be desirable to consider as many factors as possible to find the true optimum [23] but this is difficult due to the large number of them influencing migration in EKC. The factors of interest can be divided into two groups: instrumental or operational parameters, and the separation buffer factors. Instrumental parameters are common to the mode of operation in EKC and are represented by the applied voltage, the temperature, the injection time or injection volume, the capillary length, and the internal diameter, etc. The temperature affects physicochemical parameters like viscosity, pKa, absolute mobilities and critical micellar concentration of various surfactants, and thus separation [4].
Designs
Applied voltage (kV): 24–30 Buffer (mM): 10–20 pH: 8.0–9.6 SDS (mM): 50–80 Temperature ( C): 24–30 SDS (%): 2–5 1-Butanol(%): 5–9 Borate buffer (mM): 10–50 Brij 35 (%): 0–2.5 2-Propanol (%): 0–20 Cassette temperature ( C): 25–40
CM-b-CD (w/v): 0.20–0.40 SDS (mM):10–20 ACN (%): 0–15 pH: 8.2–9.7
Empirical (MLR)/ Prediction error of separation window, five additional plate height, experiments: retention factor, Q 0.2–10.7 %
Fractional factorial (26–2 þ 6)
Nairtazapine and its main active metabolite
Only analytical method
Univariate
Multivariate
Optimization
Anionic, cationic and neutral drugs
Multivariate
Benzodiazepines and Univariate g-hydroxybutyric acid related compounds Ketorolac Multivariate tromethamine and its impurities
One tetrapeptide and nine related substances
Analytes
Only analytical method
Validation
Empirical (MR, ANOVA (model) second order)/ Validity of resolution, analysis prediction time Analytical method
Empirical (MLR)/ resolution, migration time
Model/response function
Doehlert design
Plackett–Burman Methanol (%): 0–15 Full factorial ACN (%): 0–15 Central composite TEA (mM): 0–15 (2,6-DM)-b-CD (mM): 0–20 Buffer (M): 0.05–0.10 Ionic strength: 0.03–0.07 Temperature ( C): 15–40 Applied voltage (kV): 15–30 pH: 2.5–6.0 Type of chiral selector: CM-b-CD, b-CD sulphate, sulfobutyl ether-b-CD, b-CD, g-CD Temperature ( C): 10–25
Factors and levels
Table 5.1 Summary of experimental work on the optimization of conditions in electrokinetic chromatography
(Continued)
[12]
[6]
[11]
[10]
[1]
Reference
b-CD (mM): 5–15 Temperature ( C): 11–30 Applied voltage (kV): 11–25 ACN (%): 0–15
SDS (M): 0.01–0.05 n-Propanol or n-butanol (volume fraction): 0–0.05
Factorial
Multivariate
Multivariate
Univariate
Multivariate
Optimization
()-MethamphetaUnivariate mine, ()-methcathinone, ()ephedrine and ()pseudoephedrine
N-Phenylpyrazoles
Substituted benzenes
Empirical (MLR first and second order polynomials)/Resolution Empirical (MLR)/t0,tm, (ANOVA) Model Prediction errors k, resolution (normalized product) Non-linear regression (theoretical model)/k, resolution (normalized product)
Triangular ORM Cubic ORM
Injection time (s): 3–10 SDS (mM): 0–150 Borate buffer (mM): 10–50 pH: 8–10 SDS (mM): 30–60 pH: 6–8 Cyclodextrin (mM): 10–20
Triazine compounds ANOVA (model) Empirical (MR (herbicides) ANN: prediction second order)/ error for five objective function: additional data chromatographic exponential function Analytical method (CEF) (describes the quality in terms of resolution and duration of the electrophoretic separation) ANN/objective function: CEF Creatinine
Analytes
Central composite
Validation
Borate (mM): 5–50 Phosphate (mM): 5–50 SDS (mM): 5–50 1-Propanol (%): 3–15 pH: 8.5–10
Model/response function
Designs
Factors and levels
Table 5.1 ðContinued)
[15]
[14]
[8]
[2]
[13]
Reference
pH: 8–10.5 SDS (mM): 30–80 Borate (mM): 50–150 CDs: b-, sulfated b-, HP b-, a-, and g-cyclodextrin Organic modifiers: methanol, acetonitrile, propan-2-ol, tetrahydrofuran, and n-butanol
SDS (mM): 25–150 Brij 30 (mM): 10–30 Organic solvent: methanol and ACN H3PO4 (mM): 25–100
Borate (mM):5–25 Orthogonal SDS (mM): 0–20 Uniform ACN (%): 0–20 Buffer pH: 8.0–10.0 Factorial 33 or 22 HP-b-CD (mM): 100–300; 300–500 pH: 4–6 Buffer (mM): 50–200; 200–500 Hexanesulfonate (mM): 0–110 Poly(diallydimethylammonium chloride) (%): 0–0.5 [18]
Prediction error Theoretical model Aromatic carboxylic Multivariate (with the model acids, sulfonates and (Non-linear regression)/observed data set and the opiates prediction data set) mobility Empirical/ normalized resolution product and minimum resolution product Only analytical Lysergic acid Univariate method diethylamide, iso-lysergic acid diethylamide and lysergic acid N,Nmethylpropylamide Xanthones Univariate
(Continued )
[20]
[19]
[17]
[16]
Optical isomers of lac- Multivariate tic acid
Prediction error for Imperatorin and isoim- Multivariate two additional data peratorin
Empirical (MLR)/reso- Only analytical lution method
ANN/resolution
Orthogonal array
Applied voltage (kV): 20–25 pH: 8.0–9.0 Type of surfactant: SDS, NaCh Urea (M): 0.5–1 Borate (mM): 10–20
ANOVA (model)
Validation Pesticides
Analytes
Multivariate
Optimization
ANNs/Two objective Prediction error for Plant hormones Multivariate functions depending five additional data on resolution and resolution and migration time of the last peak Empirical, (MLR Model error Benzene, toluene, Multivariate cubic polynomials)/ ethylbenzene, overall resolution propylbenzene, etc.
Empirical (MLR)/Response function
Model/response function
[22]
[21]
[4]
Reference
ACN: acetonitrile; TEA: triethanolamine; MLR: multiple linear regression; Q: quotient between the effective mobility in the microemulsion and the effective mobility in the corresponding buffer (only for anions); ORM: overlapping resolution mapping; CD: cyclodextrin
pH: 8.0–10.0 Orthogonal Carrier electrolyte (mM): 20–100 Applied voltage (kV): 10–30 Temperature ( C): 15–35 SDS (mM): 20–100 Lithium dodecyl sulfate/lithium perfluorooctanesulfonate ratio
Designs
Factors and levels
Table 5.1 ðContinued Þ
OPTIMIZATION OF SEPARATION CONDITIONS
101
Mikaeli et al. [23] have pointed out that it may be not possible to optimize some parameters that are known to influence the response due to limitations in the experimental instrumentation. For example, in their study the possibility of varying the dimensions of the capillary was limited by the relatively poor sensitivity of UV absorbance detectors. Also, applied voltage is another variable that affects the separation process. High voltages produce a greater number of theoretical plates and at the same time, on reducing analysis times, diffusion is minimized. However, application of high voltages is limited to the generation of an elevated electrical current strength, which produces more heat per unit time through the Joule effect [24,25]. The separation buffer factors depend on the mode of operation. In Table 5.1 we can observe that in micellar electrokinetic chromatography (MEKC) the most frequently studied factors are the concentration of the surfactant, the buffer, and the organic modifiers. Also, in this column, experimental ranges for these factors are included. For weak acids or bases, also, pH can affect the separation process [23]. Other factors that should be taken into account in the optimization of the separation conditions are the surfactant, the buffer and the modifier nature. In cyclodextrin modified micellar electrokinetic chromatography (CD-MEKC) the type and concentration of cyclodextrin are factors that influence the response in an important way. The factors that have been more rarely studied in EKC are ionic strength, the concentration of a secondary organic modifier or the concentration of a secondary surfactant. From the point of view of the optimization scheme, the possibility that so many variables affect the separation process renders an important complexity to the problem. For example, in a typical MEKC separation with eight factors (surfactant, modifier, and buffer concentration, pH, voltage, temperature, capillary length, and injection time) a full factorial design at two levels needs 256 experiments. So, the tendency in recent years has been to apply screening designs, as a previous stage in the optimization scheme, in order to evaluate significant parameters. As screening designs the most frequently used are the Plackett–Burman designs, orthogonal array designs, D-optimal designs and fractional factorial designs. In these experiments the main effects of the variables are studied, so the problem can be simplified with a lesser number of experiments. For example, Zhang et al. [4] studied the separation of nine pesticides in MEKC by means of an orthogonal array design taking into account five relevant factors: type and concentration of surfactant, concentration of buffer, pH and concentration of urea as the modifier. In this case, three or more variables interactions were ignored, since they were much less likely to occur and if they did exist, then they would most likely be of much a smaller magnitude than a main effect. From the results they concluded that the most significant factors were type of surfactant, the urea concentration, and the pH, so the problem was simplified to only three variables. Mikaeli et al. [23] optimized the separation of phenols and the chiral separation of (þ)-1-(9-anthryl)-2-propyl chloroformate-derivatized amino acids by MEKC. They included in their study the following parameters: pH, the concentration of the primary surfactant (sodium dodecyl sulphate, SDS), the concentration of a secondary surfactant (sodium deoxycholate, SDC), the buffer concentration (sodium tetraborate; borax), the temperature in the capillary compartment, the amount of acetonitrile added as an organic modifier, and the separation voltage. In the initial screening designs they used a Plackett– Burman design. All the parameters showed significant effects on the resolution except for
102
ELECTROKINETIC CHROMATOGRAPHY
the concentration of the buffer and the temperature of the capillary compartment. The nonsignificant parameters were thus locked at the value of the respective centre points and the voltage was kept at the lowest value in all further experiments. The significant parameters were thus investigated using a full factorial design in order to elucidate if any interaction effects occurred, and the results showed that the pH of the separation electrolyte was not a significant parameter. The final optimization step was a central composite design, that is, a full factorial design, with axial and centre point experiments. So, with this strategy a total of 45 experiments had been performed, in contrast with a full factorial design for eight factors (28) that required 256 experiments. Similarly, Brunnkvist et al. [1] developed an EKC method for separating the tetrapeptide H-Tyr-(D)Arg-Phe-Phe-NH2 and nine related substances by using heptakis (2,6-di-O-methyl)-b-cyclodextrin (2,6-DM-b-CD) in the separation buffer. The method was developed using experimental design in a four-step procedure, in which eight variables were investigated in a total of 47 experiments. The initial set of experiments was performed to establish basic requirements. This was followed by a Plackett–Burman design for variable selection (eight variables under study), a full factorial design (four variables) to estimate the significance of the selected variables, and, finally, a circumscribed central composite design, including axial points, to investigate carefully the influence of the retained significant variables (two variables, CD concentration and acetonitrile percentage) on the selected responses, resolution and migration time of the last eluting peak. Also, the separation of anionic, cationic and neutral drugs in microemulsion electrokinetic chromatography (MEEKC) was studied with a statistical experimental design. The concentration of SDS, 1-butanol (cosurfactant) and borate buffer and the factors Brij 35 (surfactant), 2-propanol (organic solvent) and cassette temperature were varied simultaneously, while the parameters pH, the concentration of octane (oil), the voltage and the dimension of the fused-silica capillary, were kept constant. The aim was to screen a number of parameters for their contribution in the separation performance of some analytes by means of a fractional factorial design (26-2, resolution IV). From all the results presented in the previous discussion, it is highly recommended that a scheme such as that shown in Figure 5.1 be followed. As an example, in this figure a fractional factorial, a full factorial and a central composite design, for three codified factors, are shown. The objective in using the screening designs (fractional factorial and/or full factorial designs) is to reduce the number of variables to at least two or three, evaluating the main and/or the interaction terms. Then, the nonsignificant variables are usually set at their central points and the significant ones are used in a more sophisticated design, such as the central composite design (usually known as the response surface design) to investigate carefully their influence on the selected response(s). This design includes axial points, in this case, face centred (see Figure 5.1) and a central point (usually, several replicated experiments are performed, typically five). So, the response curvature as a function of the remained variables and the experimental error can be evaluated. However, experimental design techniques are not the panacea: their indiscriminate or inadequate use may lead to erroneous conclusions [5]. Omitting significant factors, selecting inappropriate experimental ranges for study, misinterpreting results, etc., may be the cause of deficient optimization. Chemometric techniques offer excellent possibilities
OPTIMIZATION OF SEPARATION CONDITIONS
Fractional factorial design
1
Full factorial design
1
1 1
–1
Central composite design
1
1
–1 –1
103
–1 –1 1
–1
1 –1 –1 1
–1
Figure 5.1 Experimental designs for three codified factors: fractional factorial, full factorial, and central composite design
in the optimization of CE separations but the analyst has to be aware of the characteristics, possibilities, and limitations of experimental designs [5].
5.3 Modelling Once, through a screening design (or a combination of screening designs) the problem is simplified to a reduced number of variables, an adequate model of the response(s) as a function of these variables must be obtained. Some different strategies have been used in the modelling of the response(s) as a function of the variables of interest; basically, they can be classified into two categories: theoretical models and empirical equations. However, nowadays, soft models, such as artificial neural networks, are acquiring great importance. 5.3.1
Physicochemical Models
Physicochemical models can be divided into equations that describe the separation between solutes and equations that describe the migration behaviour of individual solutes [7]. Table 5.2 summarizes the most representative physicochemical models that have been applied in MEKC. These equations, as can be observed in the table, depend on the surfactant concentration and pH or the surfactant concentration and organic modifier concentration. Terabe et al. [26,27] first introduced in MEKC the equation that gives the resolution (Rs) as a function of the plate number (N), the selectivity factor (a), the mean retention factor (k), the migration time of the electroosmotic flow marker (t0) and the migration time of the micelles (tm). Although the peak plate number depends on the experimental conditions, in most situations it has been fixed to a constant value in order to predict the
104
ELECTROKINETIC CHROMATOGRAPHY
Table 5.2 Most representative physicochemical equations in MEKC Physico-chemical Equations pffiffiffiffi! a 1 1 t0 =tm N k Rs ¼ a 4 1 þ ðt0 =tm Þk 1 þ k KMW;HA þ ½HKþa KMW;A ½M mmc þ ½HKþa maq;A ma;a ¼ 1 þ ½HKþa þ KMW;HA þ ½HKþa KMW;A ½M
Reference
Neutral and ionizable solutes-cationic and anionic micelles
[26–27]
Acidic solute-anionic micelles
[29, 30]
ma;c
KMW;HA þ ½HKaþ KMW;A ½M mmc þ ½HKþa maq;A ¼ 1 þ KIP cmc þ ½HKþa þ KMW;HA þ ½HKþa KMW;A ½M
Area of use
mb;a
KMW;BHþ þ ½HKþa KMW;B ½M mmc þ maq;BHþ ¼ 1 þ KIP cmc þ ½HKþa þ KMW;BHþ þ ½HKþa KMW;B ½M
Acidic solute-cationic micelles
mb;c
KMW;BHþ þ ½HKþa KMW;B ½M mmc þ maq;BHþ ¼ 1 þ ½HKaþ þ KMW;BHþ þ ½HKþa KMW;B ½M
Basic solute-anionic micelles
q P ai km H
meff ¼ i¼0
ðq iÞA
k¼
aq ½ M mm þ mHðq iÞ Ai i
q P ai km H
i¼0
Basic solute-cationic micelles
ðq iÞA
½ M þ 1 i
fK1 1 þ K3 Aaq ½M
2 1 þ K2 Aaq þ K2 K4 Aaq
Basic solute-cationic micelles
[31]
Neutral solute-anionic micelles, alcoholic modifier
[14]
KMW,SH and KMW,S are micellar binding constants of solute in the conjugate acidic and basic forms; Ka is the acid dissociation constant; KIP is the ion-pair formation constant between ionic solute with surfactant monomer; cmc is the critical micelle concentration of the surfactant; mmc is the mobility of the micellar phase; maq,SH and maq,S are the mobilities in the aqueous phase of the solute in the acidic and basic form, respectively; HA is an acidic solute and A the aq dissociated solute; B is a basic solute and BHþ its conjugate acidic form; mHðq iÞ Ai is the electrophoretic mobility of the ith undissociated anion in the aqueous phase; km Hðq iÞ Ai is the association coefficient between the ith undissociated anion and the micelle;
ai is the ith degree of dissociation of the conjugated acid of an anion; k is the retention factor; f is the phase ratio, Aaq is the alcohol concentration; K1 is the equilibrium constant that takes into account the association of the solute with the micelle to form a complex in the micellar pseudophase; K2 takes into account the enhancement of solubility of solute in the separation buffer modified by alcohols; K3 is the formation constant of the complexes among the solute, the alcohol and the micelle in the micellar pseudophase; and K4 the formation constant of the complex between the solute–alcohol complex and the other molecules of alcohol.
resolution between adjacent peaks, or the overall normalized resolution as the way to obtain the quality of the separation under study. Foley [28] differentiated this equation with prespect ffiffiffiffiffiffiffiffiffiffi to k. The maximum resolution was obtained for a mean retention factor, k ¼ tm =t0 , when the other quantities are kept constant (an assumption that cannot be fulfilled in practice). This approach does not focus on the optimization of a given well-defined MEKC system, but it provides a general optimum for MEKC. Although, from this equation a separation buffer composition can be derived, the main disadvantage of this strategy is the possibility of missing the actual optimum of a certain MEKC system.
OPTIMIZATION OF SEPARATION CONDITIONS
105
Khaledi and coworkers [29,30] introduced physicochemical models that describe the migration behaviour of both, acidic and basic solutes, as a function of the separation buffer pH and the surfactant concentration. In Table 5.2 equations relating the mobility of ionizable solutes (mi), acidic solutes, and basic solutes as a function of pH and surfactant concentration (½M ) by using anionic and cationic micelles are shown. Jimidar et al. [31] modelled the effective mobility of anions in MEKC by theoretical and empirical models as a function of pH and the surfactant concentration, and obtained a theoretical model similar to that used by Khaledi and coworkers [29,30]. The results showed that the predictions of the mobilities obtained with linear regression (empirical equations) were slightly more accurate for all the anions, except for carbonate. For this anion the shape of the response surface were clearly sigmoid. More recently, our own research group [14] has modelled the retention factor as a function of the surfactant and the organic modifier (n-propanol or n-butanol) concentrations and the results have been used to classify the retention behaviour of the different test solutes. Also, these results have been compared to those obtained by means of empirical equations. From the comparison of the mean prediction errors obtained by both methods, the statistical test indicated that the physicochemical model was the best to explain the retention behaviour of the solutes under study. Nowadays, chiral separations are gaining much attention and capillary electrophoresis has been demonstrated to be a powerful technique for analytical enantioseparation within the last decade [32]. Also, several physicochemical models for the quantitative treatment of the effect of the experimental variables on the migration behaviour of the analytes in a chiral separation system have been published and have been reviewed by Scriba [32]. These models relate the effective mobility of the enantiomers or the difference in the effective mobility of both enantiomers to the concentration of the chiral selector or with pH and the concentration of the chiral selector. 5.3.2
Empirical Models
In Table 5.1, the most recent applications of these empirical models have been summarized. Generally, the empirical equations used are first or second order polynomials relating, independently, one or more parameters (resolution, retention factor, t0, tm, analysis time, or some type of objective function) with different experimental variables. The chemometric method for obtaining the equation parameters is generally, multivariate linear regression, and when a two-factor system is evaluated, the final results are commonly depicted as response surfaces. One of the main problems that we encountered when we faced an optimization problem was the choice of the appropriate response to model as a function of the variables of interest. For instance, Wan et al. [33] have applied a central composite design to optimize the separation of neurotransmitter amino acids in normal and reversed migration MEKC. The responses used were resolution, efficiency and migration times. In Figure 5.2, the results obtained in normal MEKC for the modelling of resolution between DL-glutamic acid and DL-aspartic acid [Figure 2(a)], the efficiency for DL-aspartic acid peak [Figure 2(b)] and the migration time for DL-aspartic acid [Figure 2(c)] are shown as Pareto charts. The Pareto chart is an option of Statgraphic Plus Software [34] for Windows. The length of each bar is proportional to the absolute value of its associated regression coefficient or
106
ELECTROKINETIC CHROMATOGRAPHY
Figure 5.2 Standardized Pareto charts for the resolution between DL-glutamic acid and DL-aspartic acid (a); the efficiency for DL-aspartic acid peak (b); and the migration time for DL-aspartic acid (c). The three factors considered were the buffer concentration (A), the SDS concentration (B), and the applied voltage (C)
estimated effect. The effects of all parameters, interactions as well as quadratic terms, are standardized. The order in which the bars are displayed corresponds to the order of the size of the effects. The chart includes a vertical line that corresponds to the 95 % limit indicating statistical significance. An effect is, therefore, significant if its corresponding bar crosses this vertical line. From Figure 5.2 we can observe that different results are obtained depending on the response type evaluated. Thus, when the resolution between DL-glutamic acid and DL-aspartic acid is evaluated the applied voltage has no a significant influence on the response. In contrast, when the efficiency for DL-aspartic acid is studied, only the first and second order term, depending on the buffer concentration, influence the response. Furthermore, when the migration time of DL-aspartic acid (the last eluted peak) is considered, the interaction terms between buffer and SDS concentrations and the second order terms of buffer and SDS concentrations have no significant influence on the response. Similarly, Brunnkvist et al. [1] used two response functions, resolution and the migration time of the last eluting peak, to optimize the separation of H-Tyr-(D)ArgPhe-Phe-NH2 and its nine related products by EKC. Their results indicate that the presence of acetonitrile in the BGE does not improve the analysis time, and only the CD (heptakis(2,6-di-O-methyl)-b-cyclodextrin) concentration significantly influences, both, the resolution and the migration time. Also, Bu¨ tehorn and Pyell [35] and Garcı´a-Ruiz et al. [14] modelled t0, tm and k as a function of the concentrations of SDS and the organic modifiers in MEKC. With these predicted values and considering a fixed number for N, the resolution between adjacent
OPTIMIZATION OF SEPARATION CONDITIONS
107
peaks was achieved. Furthermore, the quality of the separation could be evaluated by using the overall normalized resolution so the best conditions for the system considered could be selected [14]. Although one of the most frequently modelled parameters has been the resolution between adjacent peaks [7], it must be taken into account that both the resolution and the separation factor may display discontinuous functions in the case of cross-over and these discontinuities interfere with the use of polynomial equations in the model building process. To overcome this problem, Zomeren et al. [7] suggested allowing the resolutions and separation factors to become negative and calculated them not only for adjacent peaks but for all peak pairs. Apart from the resolution and the migration time of the last eluted peak other researchers have used, as the response, objective functions in order to evaluate the separation quality. Among these, the overall normalization resolution (r) is the most cited in the literature. However, Liu et al. [21] indicated that the electrophoretic window should also be considered as a parameter to evaluate the separation selectivity, together with the overall normalization resolution. Hence they used an objective function, Q, considering the resolution between the first peak and the last peak in the electropherogram, the migration time of the last peak and the resolution between adjacent peaks. From their results, both criteria (r and Q) predicted the same value of pH and concentration of SDS. Also, the simulated concentration of buffer electrolyte and temperature were similar for both criteria. However, the applied voltages were quite divergent. Criterion Q predicted high voltages while criterion r chose a relatively low voltage. Consequently we suggest that as many as possible different responses be used in the optimization scheme, because the results are influenced in different ways by the variables considered. Also, we consider that when empirical equations are used, second order polynomials, as a starting point, should be used because it has been demonstrated that curvature effects in these models are important in the EKC separations [36]. Afterwards, the best model can be chosen by comparing the performance of the models considered, and also, their prediction capability. 5.3.3
Artificial Neural Networks
Nowadays, ANNs are emerging as one of the most promising areas in soft modelling and have been successfully used in the optimization of the separation conditions in MEKC in combination with experimental designs such as central composite designs [13] and orthogonal designs [16,21] (see Table 5.1). Furthermore, the modelling results obtained with ANNs have been compared to those obtained by means of theoretical models [37] and empirical equations [13,38–41], with ANNs as the best choice. Although a detailed description of the theory behind a neural network is out of the scope of this chapter, it can be consulted in the literature [42,43]. It must be pointed out that these soft models do not require formulation of mathematical equations and one of their most attractive advantages is the possibility of modelling nonlinear responses. By contrast, some drawbacks must be cited. As the architecture of the net, that is, the number of the hidden layers, the number of neurones in the hidden layer(s), the transformation of the experimental data, the training algorithm, the momentum, the
108
ELECTROKINETIC CHROMATOGRAPHY
learning rate, the transfer functions for the neurones in the hidden layer(s) and the output layer, must be optimized by trial and error, the optimization scheme is very complex. Also, problems of overfitting must be avoided.
5.4 Validation The purpose of validation is to establish whether or not a method is acceptable for its intended purpose [1], so, in this case, we must distinguish between model validation and analytical method validation. With respect to the model validation, the strategy most usually employed involves the use of two types of data, that is, those necessary to obtain the parameters of the empirical or theoretical equation (for ANNs, the training data set), and those to evaluate the prediction capability of the model. It is important to note that although the goodness of fit of a model is the best as possible, if the prediction capability is poor, the model cannot be considered appropriate anyway. From the literature consulted (see Table 5.1), the analysis of variance (ANOVA) is the most frequently method used to evaluate the significance of the models, and the most widely used descriptive criterion is the correlation coefficient (R2). However, in many cases, neither the significance of the parameters obtained by multiple linear regression (as F ratios, P values or by means of Pareto charts), or the pattern of the residuals is commented on, so the validity of the models may be questionable. Also, it is strongly recommended that central composite designs be used with replicated measurements in the centre to apply the lack of fit test, because this test is designed to determine if the selected model is adequate to describe the observed data. A common error in some works is the reporting of prediction errors for the models taking into account only those data that are used to construct the model (model data set) or, in some instances, both this data set and the prediction data set. It must be urged that the prediction capability of a model should not be evaluated only by the use of the model data set. The prediction data set should be used being the relative mean squared error of prediction (MSEPrel) [7] and the root mean square error of prediction (RMSEP) the most useful predictive criteria. They are calculated as follows: sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi m P 1 ðyk ^yk Þ2 m MSEPrel ¼
RMSEP ¼
k¼1
y
x100 %
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi uP um u ð^yk yk Þ2 tk¼1 m
with y being the average response, yk the measured and ^yk the predicted responses of all the test points (m).
OPTIMIZATION OF SEPARATION CONDITIONS
109
Although there is no agreement about the size of the prediction data set, in our opinion when it is not possible to get at least five points the validation should be achieved by using the cross validation method (leave-one-out method). When the separation conditions have been chosen, the next and final step is the analytical method validation so we can asses the validity of the quantification results. In this case the parameters of interest are the selectivity, the precision, the dynamic range and linearity range, the accuracy, the sensitivity, the detection and quantification limits, and the robustness [45, 46]. In particular, in order to quantify pharmaceutical products, the optimized experimental conditions are usually validated according to the ICH guidelines [47].
5.5 Selection of Optimum Separation Parameters Summarizing the principles outlined in the previous sections, the selection of optimum separation parameters can be viewed as a sequential process, in which each step is influenced by the results obtained in the previous one. The process begins with the selection of the variables (and ranges) influencing the separation process. At this stage the researcher’s knowledge and the bibliographic information can be of great importance. So, in Table 5.1 the variables studied and their variation ranges in the most recent works in EKC have been summarized. Also, the reviews of Altria et al. [9] and Sentellas and Saurina [5] can be consulted. In this selection it would be desirable to include as many variables as possible to assure that global optimum cannot be missed. With respect to response(s) we must be clear what we are looking for. If we are interested only in the base-line separation of the test solutes, an adequate response could be the resolution between adjacent peaks or the overall normalization resolution. The advantage of this latter function is that it is able to indicate the quality of an experiment with only one value, instead of the n 1 values obtained if we consider resolution between adjacent peaks (n being the number of test solutes). By contrast, if we are interested, not only in the base-line separation of the test solutes, but also in the analysis time we could use an objective function such as the Q criterion reported by Liu et al. [21]. At this point, we should construct a screening experimental design in order to reduce the number of variables, not to evaluate the quality of the model. The simplest designs are the fractional factorial designs as that shown in Figure 5.1. By means of multivariate linear regression we can obtain the coefficients of a first order polynomial and evaluate their statistical significance by using the ANOVA analysis or the Pareto charts. With this information we are able to select those variables significantly affecting the response. Then, the other variables can be adjusted for subsequent experiments to a fixed value (usually the mean value for the range studied). Depending on the results obtained, two routes can be followed. If the remaining variables are still many, we can apply a full factorial design at two levels with a first order polynomial to reduce, if possible, to a lesser number. By contrast, if the remaining variables are two or three, we could apply a central composite design with a second order polynomial. At this point, data of the design should be divided into two categories, the model data set and the prediction data set. The experiments conducted with the model data set allow the
110
ELECTROKINETIC CHROMATOGRAPHY
coefficients of the second order polynomial to be obtained, and those conducted with the prediction data set allow evaluation of the prediction capability of the model. In order to evaluate the model developed, the ANOVA analysis can indicate the goodness of fit, the significance of the parameters (hence the possibility of simplification of the model eliminating those statistical nonsignificant terms) and the pattern of the residual can indicate if the requirements of the model are fulfilled. If more than one model is obtained (e.g. because we use different models with different responses), the best model can be chosen comparing R2 values, the number of significant terms in the polynomial and the prediction errors. If the prediction errors are large, then we could repeat the modelling step by using theretical models (if possible) or ANNs, both capable of modelling nonlinear responses. In the last step, the best separation conditions can be chosen graphically or by a gridsearch algorithm (finding the maximum value for the response selected). Modelling also makes it possible to estimate the robustness of a method.
5.6 Concluding Remarks From the discussion in the different sections of this chapter some conclusions can be drawn. The use of experimental designs can help us to find the best conditions for a particular separation by EKC, although special care should be taken to choose the variables of interest and the range of variation. Also, in order to optimize the separation with the minimum number of experiments, it is necessary to introduce a screening step (one or more) to reduce the number of variables to as few as possible. Although the modelling step can be achieved by using theoretical models or by empirical models (empirical equations or ANNs), the best choice seems to be the empirical equations due to their simplicity. Only when we suspect that the response(s) are not linear is the use of ANNs justified. The theoretical models, although permitting an insight in the mechanism of migration, are not very adequate because the number of factors in the models is very restricted. In most cases, only one or two variables are considered. Moreover, they usually need the application of iterative methods to obtain the model constants, so the results depend significantly on the initial estimates. Furthermore, the iterative methods are time consuming and the global optimum can be missed during the optimization process. With respect to the response(s) choice, it must be pointed out that different optimization conditions can be found when different response criteria are used. We consider that objective functions taking into account both the resolution and the analysis time are the most appropriate in EKC with the exception of those chiral separations in which the most important criterion is usually the resolution between enantiomers. Finally, we strongly recommend special care in the validation step. Many reports do not contain enough information about the validation of the model used in the prediction of the best separation conditions or about the analytical method validation, so it is very difficult to establish the usefulness of these methods. The validation step should reflect the goodness of fit of the model, its prediction capability and the characteristics of the analytical method.
OPTIMIZATION OF SEPARATION CONDITIONS
111
Acknowledgements M.L. Marina thanks the Ministry of Science and Technology (Spain) for the research project BQU2003-03638.
References [1] H. Brunnkvist, B. Karlberg, A. Astervik and I. Granelli. Experimental design-based development of a rapid capillary electrophoresis method for determining impurities in the tetrapeptide H-Tyr-(D)Arg-Phe-Phe-NH2, J. Chromatogr. B., 807, 293–300 (2004). [2] E. Pobozy, A. Radomska, R. Koncki and S. Glab. Determination of dialysate creatinine by micellar electrokinetic chromatography, J. Chromatogr. B, 789, 417–424 (2003). [3] S.-W. Sun, S.-S. Lee, L.-Y. Chen and C.K. Chen. Separation of aporphine alkaloids by micellar electrokinetic chromatography, J. Chromatogr. A., 767, 277–284 (1997). [4] Y. Zhang, X. Li, Z. Yuan and Y. Lu. Orthogonal array design experiments for optimizing the separation of nine pesticides by micellar electrokinetic chromatography, Microchem. J., 73, 307–315 (2002). [5] S. Sentellas and J. Saurina. Chemometrics in capillary electrophoresis. Part A: Methods for optimization, J. Sep. Sci., 26, 875–885 (2003). [6] S. Orlandini, S. Fanali, S. Furlanetto, A.M. Marras and S. Pinzauti. Micellar electrokinetic chromatography for the simultaneous determination of ketorolac tromethamine and its impurities. Multivariate optimization and validation, J. Chromatogr. A., 1032, 253–263 (2004). [7] P.V. van Zomeren, M.J. Hilhorst, P.M.J. Coenegracht and G.J. de Jong. Resolution optimization in micellar electrokinetic chromatography using empirical models, J. Chromatogr. A., 867, 247–259 (2000). [8] Y. Zhang, X. Li, Z. Yuan and W. Zhou. Resolution optimization in micellar electrokinetic chromatography using two different overlapping resolution mapping schemes, Chromatographia, 57, 59–65 (2003). [9] K.D. Altria, B.J. Clark, S.D. Filbey, M.A. Kelly and D.R. Rudd. Application of chemometric experimental designs in capillary electrophoresis: a review, Electrophoresis, 16, 2143–2148 (1995). [10] R. Mandrioli, V. Pucci, C. Sabbioni, C. Bartoletti, S. Fanali and M.A. Raggi. Enantioselective determination of the novel antidepressant mirtazapine and its active demethylated metabolite in human plasma by means of capillary electrophoresis, J. Chromatogr. A., 1051, 253–260 (2004). [11] S.C. Bishop, M. Lerch and B.R. McCord. Micellar electrokinetic chromatographic screening method for common sexual assault drugs administered in beverages, Forensic Sci. Int., 141, 7–15 (2004). [12] V. Harang, J. Eriksson, C.E. Sa¨ nger-van de Griend, S.P. Jacobsson and D. Westerlund. Microemulsion electrokinetic chromatography of drugs varying in charge and hydrophobicity: I. Impact of parameters on separation performance evaluated by multiple linear regression models, Electrophoresis, 25, 80–93 (2004). [13] S. Frı´as-Garcı´a, M.J. Sa´ nchez and M.A. Rodrı´guez-Delgado. Optimization of the separation of a group of triazine herbicides by micellar capillary electrophoresis using experimental design and artificial neural networks, Electrophoresis, 25, 1042–1050 (2004). [14] C. Garcı´a-Ruiz, O. Jime´ nez and M.L. Marina. Retention modeling and resolution optimization for a group of N-phenylpyrazole derivatives in micellar electrokinetic chromatography using empirical and physicochemical models, Electrophoresis, 24, 325–335 (2003).
112
ELECTROKINETIC CHROMATOGRAPHY
[15] A.-S. Liau, J.-T. Liu, L.-C. Lin, Y.-C. Chiu, Y.-R. Shu, C.-C. Tsai and C.-H. Lin. Optimization of a simple method for the chiral separation of methamphetamine and related compounds in clandestine tablets and urine samples by b-cyclodextrine modified capillary electrophoresis: a complementary method to GC-MS, Forensic Sci. Int., 134, 17–24 (2003). [16] H. Liu, K. Wang, H. Xu, X. Chen and Z. Hu. Application of experimental design and artificial neural networks to separation and determination of active components in traditional Chinese medicinal preparations by capillary electrophoresis, Chromatographia, 55, 579–583 (2002). [17] L. Saavedra and C. Barbas. Optimization of the separation of lactic acid enantiomers in body fluids by capillary electrophoresis, J. Chromatogr. B, 766, 235–242 (2002). [18] P. Zakaria, M. Macka, J.S. Fritz and P.R. Haddad. Modelling and optimization of the electrokinetic chromatographic separation of mixtures of organic anions and cations using poly(diallydimethylammonium chloride) and hexanesulfonate as mixed pseudostationary phases, Electrophoresis, 23, 2821–2832 (2002). [19] C. Fang, J.-T. Liu and C.-H. Lin. Optimization of the separation of lysergic acid diethylamide in urine by a sweeping technique using micellar electrokinetic chromatography, J. Chromatogr. B, 775, 37–47 (2002). [20] T. Bo, X. Yang, F. Liu, K.A. Li, L. Xiu and H. Liu. Optimized separation of pharmacologically active xanthones from Securidaca inappendiculata by micellar electrokinetic chromatography and microemulsion electrokinetic chromatography, Anal. Chim. Acta, 474, 37–48 (2002). [21] B.-F. Liu, J.-F. Zhang and Y.-T. Lu. Predicting and evaluating separation quality of micellar electrokinetic capillary chromatography by artificial neural networks, Electrophoresis, 23, 1279–1284 (2002). ´ lvarez-Coque, E. Bosch and [22] E. Fu