Conductivity detection for application in capillary...

3/2/2003

Conductivity detection for application in capillary electrophoresis microchips

3/2/2003

3 February 2003, 16:28

Conductivity detection for application in capillary electrophoresis microchips

Proefschriftter verkrijging van de graad van doctoraan de Technische Universiteit Delft,

op gezag van de Rector Magnificus Prof. dr. ir. J.T. Fokkema,voorzitter van het College voor Promoties,

in het openbaar te verdedigen

op maandag 17 maart 2003 om 10.30 uur

door

Frédéric LAUGERE

Ingénieur en électronique, ENSEIRB(Ecole Nationale Supérieure d´Electronique, Informatique et Radiocommunications de Bordeaux)

geboren te Angoulême, Frankrijk

3 February 2003, 16:28

Dit proefschrift is goedgekeurd door de promotoren

Prof. dr. P.J. FrenchProf. dr. ir. M.J. Vellekoop

Toegevoegd promotor

Dr. ir. A. Bossche

Samenstelling promotiecommissie:

Rector Magnificus, voorzitter Technische Universiteit DelftProf. dr. P.J. French, promotor Technische Universiteit DelftProf. dr. ir. M.J. Vellekoop, promotor Technische Universität Wien, OostenrijkDr. ir. A. Bossche, toeg. promotor Technische Universiteit DelftProf. dr. ir. A. van den Berg Technische Universiteit TwenteProf. dr. ir. P.P.L. Regtien Technische Universiteit TwenteProf. dr. ir. J.J. Heijnen Technische Universiteit DelftProf. dr. P.M. Sarro Technische Universiteit Delft

Het onderzoek beschreven in dit proefschrift werd financieel gesteund door Stichting Technologie en Wetenschap, STW, in het kader van het BIOMAS project.

Printed by: Optima Grafische Communicatie, Rotterdam

ISBN 90-6734-186-X

Keywords: miniaturized total analysis system, capillary electrophoresis, conductivitydetection

Copyright © 2003 by F.P.J. Laugere

Printed in the Netherlands

3 February 2003, 16:28

Il semble que le fanatisme, indigné depuis peu des succès de la raison, se débatte sous elle avec plus de rage.

VOLTAIRE (1694-1778)TRAITE SUR LA TOLERANCE (1763)

It should seem that enthusiasm enraged at the recent success of reason, fought under her standard with redoubled fury.

VOLTAIRE (1694-1778)Treatise on tolerance (1763)

3 February 2003, 16:28

CONDUCTIVITY DETECTION FOR APPLICATION IN CAPILLARY ELECTROPHORESIS MICROCHIPS vii

Table of contents

3 February 2003, 18:13

Table of Contents

Introduction 1

1.1 Background and general introduction ..................................................1

1.2 Outline of the thesis .............................................................................3

1.3 References ............................................................................................5

Biochemical detection in CE microchips 9

2.1 Introduction ..........................................................................................9

2.2 Capillary electrophoresis....................................................................11

2.3 Detection in standard CE ...................................................................13

2.3.1 UV / Vis Absorbance detection ...............................................13

2.3.2 Fluorescence detection.............................................................14

2.3.3 Amperometric detection ..........................................................16

2.3.4 Potentiometric detection ..........................................................18

2.3.5 Mass spectometric detection....................................................19

2.3.6 Conductivity detection.............................................................20

2.4 Detection in CE microchips ...............................................................22

2.4.1 CE microchips and Micro Total Analysis Systems .................22

2.4.2 Detection systems ....................................................................24

2.5 References ..........................................................................................26

On-chip conductivity detection 31

3.1 Introduction ........................................................................................31

Table of contents

viii CONDUCTIVITY DETECTION FOR APPLICATION IN CAPILLARY ELECTROPHORESIS MICROCHIPS

3/2/2003

3.2 Contactless detection versus galvanic detection ................................32

3.3 4-electrode versus 2-electrode configuration .....................................34

3.3.1 Electrical model of a conductivity cell ....................................34

Calculation of the cell constant .....................................................34The interface electrode-electrolyte: the double layer....................37The impedance of the insulating material .....................................40

3.3.2 Experiments with glass dipstick...............................................41

Measurements with 2 electrodes ...................................................43Measurements with 4 electrodes ...................................................46

3.3.3 Application to capillary electrophoresis ..................................47

3.3.4 downscaling of the detector .....................................................48

3.4 Design of the sensor ...........................................................................51

3.4.1 Layout of the CE microchip.....................................................51

3.4.2 Dimensions of the electrodes ...................................................52

3.4.3 Properties of the insulating film...............................................52

3.4.4 Choice of insulating material ...................................................55

3.4.5 Experiments on SiC structures.................................................56

3.5 Technology.........................................................................................57

3.5.1 Fabrication of the channel wafer..............................................57

3.5.2 Fabrication of the electrode wafer ...........................................58

3.5.3 Bonding procedure...................................................................60

3.6 Conclusion..........................................................................................60

3.7 References ..........................................................................................62

Electronic interface for on-chip conductivity measurements 65

4.1 Introduction ........................................................................................65

4.2 Electrical model of the detector .........................................................67

CONDUCTIVITY DETECTION FOR APPLICATION IN CAPILLARY ELECTROPHORESIS MICROCHIPS ix

Table of contents

3 February 2003, 18:13

4.3 Control of the current .........................................................................69

4.3.1 Direct current readout ..............................................................69

4.3.2 Regulation of the current .........................................................70

Gain and stability of negative-feedback amplifiers.......................70Current regulation..........................................................................72Influence of parasitic capacitance on the current regulation .........77DC biasing .....................................................................................80Simulation of the current-regulation loop .....................................84

4.4 Differential voltage measurement ......................................................86

4.4.1 The input impedance of the differential amplifier ...................86

Common-mode input impedance of the differential amplifier......86Differential input impedance.........................................................87Electronic setup for bootstrapping.................................................88

4.4.2 common-mode rejection ratio..................................................89

Degradation of the CMRR due to component mismatch ..............91Simulation of the CMRR degradation...........................................96

4.5 Conclusion .........................................................................................98

4.6 References ..........................................................................................99

Full decoupling of detection and separation 101

5.1 Introduction ......................................................................................101

5.2 DC decoupling .................................................................................102

5.2.1 Breakdown of the insulating film ..........................................102

Influence of the separation voltage..............................................103Influence of contact potentials.....................................................104

5.2.2 Protection concepts ................................................................105

Floating readout electronics ........................................................106DC biasing of the readout electronics .........................................107Decoupling with external capacitors ...........................................108

5.2.3 DC decoupling by control on the separation voltage.............108

Table of contents

x CONDUCTIVITY DETECTION FOR APPLICATION IN CAPILLARY ELECTROPHORESIS MICROCHIPS

3/2/2003

Electronic setup...........................................................................109Active control of the DC decoupling voltage .............................110

5.3 AC decoupling..................................................................................111

5.3.1 Leakage of the detection current............................................111

5.3.2 AC decoupling principle ........................................................113

5.4 Measurement results.........................................................................113

5.5 Conclusion........................................................................................116

Conductivity measurements in CE microchips 117

6.1 Introduction ......................................................................................117

6.2 Measurement setup...........................................................................118

6.2.1 Housing of the microchip.......................................................118

6.2.2 Readout electronics ................................................................119

6.2.3 Electronics for decoupling detection and separation .............121

6.3 Electronic baseline suppression .......................................................122

6.3.1 Suppressed conductivity detection.........................................122

Choice of the carrier electrolyte..................................................122Chemical suppressor ...................................................................123

6.3.2 Electronic suppressor .............................................................124

Hardware setup............................................................................125Software setup.............................................................................126

6.4 Comparison between the 2- and the 4-electrode setup.....................127

6.4.1 2-electrode impedance measurement .....................................127

6.4.2 4-electrode impedance measurement .....................................129

6.5 Separation of inorganic ions.............................................................130

6.5.1 Reagents .................................................................................130

6.5.2 Injection procedure ................................................................131

6.5.3 Comparison of the 2- and 4-electrode setup ..........................131

CONDUCTIVITY DETECTION FOR APPLICATION IN CAPILLARY ELECTROPHORESIS MICROCHIPS xi

Table of contents

3 February 2003, 18:13

6.5.4 Measurements with baseline suppression ..............................135

6.5.5 Detection performance...........................................................137

6.6 Separation of organic acids ..............................................................139

6.7 Conclusions ......................................................................................140

6.8 Future work ......................................................................................141

6.9 References ........................................................................................142

Summary 145

Samenvatting 147

Acknowledgements 149

List of Publications 151

About the author 155

xii CONDUCTIVITY DETECTION FOR APPLICATION IN CAPILLARY ELECTROPHORESIS MICROCHIPS

3/2/2003

CONDUCTIVITY DETECTION FOR APPLICATION IN CAPILLARY ELECTROPHORESIS MICROCHIPS 1

3/2/2003

1Introduction

1.1 Background and general introduction

In the development and optimization of miniaturized analytical systems, adelicate combination of science and technology originating frommicroelectronic device fabrication, electrical engineering, and analyticalchemistry is essential. In this multidisciplinary field, microtechnologyexperts combine the demands from analytical chemistry and electronicinstrumentation in the design and fabrication of novel analytical devices[1.1, 1.2]. Chemical analysis systems, such as High-Performance LiquidChromatography (HPLC) or Capillary Electrophoresis (CE), alwaysconsist of the combination of a separation and a detection system.

For separation, CE separation techniques are highly suitable forimplementation on the microchip format. Electrokinetic control of fluidtransport eliminates the need for external components such as pumps andvalves. The separation efficiency is relatively independent of theseparation path length and is, therefore, compatible with miniaturization.

Chapter 1

2 CONDUCTIVITY DETECTION FOR APPLICATION IN CAPILLARY ELECTROPHORESIS MICROCHIPS

3/2/2003

As far as detection is concerned, laser-induced fluorescence (LIF) is, atpresent, the most widely used detection technique in miniaturized analysissystems because of its high sensitivity. The drawbacks of LIF are itslimited compatibility with miniaturization and on-chip integration, and therequirement for labelling of most (bio) chemically relevant compounds.External devices that are relatively large, such as the laser and thephotodetector system, strongly hamper further miniaturization. Thedevelopment of alternative detection methods compatible withminiaturization and full on-chip integration is highly desirable. Sinceelectrode deposition is a well-established process in microfabrication, theimplementation of detection techniques utilizing integrated electrodes hasbecome an attractive approach. Successful coupling of conventional CEwith potentiometry [1.3], amperometry [1.4,1.5] and conductometry [1.6,1.7, 1.8, 1.9, 1.10] has been reported in literature. In addition, bothamperometric and potentiometric detection were also implemented inchip-based CE systems [1.11, 1.12, 1.13]. The primary advantage ofamperometric and potentiometric detection over conductivity detection isthe high selectivity induced by the electrochemical reactions that takeplace at the electrode surface. Only electrochemically active compoundscan be detected using these methods, thereby eliminating interference withother compounds present in the sample. This selectivity, however, can alsoturn into a disadvantage, since it strongly limits the applicability of thedetection system. Additionally, the interference between the electricalseparation field and the detection electrodes and associated electronics is abottleneck for the realization of true on-column detection. Conductometryis a universal detection technique that has been applied for detection in CEeither in the galvanic [1.6, 1.7, 1.14, 1.15] or the contactless mode [1.8,1.9, 1.10, 1.12]. In both cases, a pair of electrodes is placed in theseparation column for impedance measurement. On the one hand,conductivity detection has the advantage to be a universal detectionmethod because separation and detection are based on charged analytes.On the other hand, conductivity detection is an “indirect” detectionmethod whose detection limit is highly limited by fluctuations of thebackground. Furthermore, separation and detection suffer from conflictingrequirements concerning the mobility of the separated ions and of thecarrier electrolyte co-ion [1.16]. Recently, both galvanic [1.17, 1.18, 1.19,1.20, 1.21] and contactless [1.22, 1.23, 1.24, 1.25, where 1.22 and 1.23 areour own work] conductivity detection were implemented on the microchip


Introduction

3 February 2003, 16:39

format. Galvanic detection was combined with isotachophoresis and withcapillary zone electrophoresis. Various analytes were monitored such asmetal ions, amino acids, proteins, and DNA fragments. The fabrication israther easy: the platinum electrodes were either sputtered [1.17, 1.18,1.21] or made from platinum wires sandwiched between two pieces ofpolymer [1.19, 1.20]. Contactless detection is preferred to galvanicdetection for three reasons. First, the electronic circuitry is decoupledfrom the high-voltage applied for separation (no direct DC couplingbetween the electronics and the liquid in the channel). Second, theformation of gas bubbles at the metal electrodes is prevented, and third,electrochemical modification or degradation of the electrode surface isprevented, thereby allowing a wide variety of electrode materials. In 1999,we first came with the design of a contactless detector that is going to bepresented in this thesis [1.22]. However, other contactless configurationswere suggested. Lichtenberg et al. [1.24] presented a microchip where thedetector is constructed with two opposite platinum electrodes placed closeto the microchannel. The detection electrodes are isolated from thechannel by a glass wall of approximately 10 µm. Detection of potassiumand lithium ions down to a concentration of 35 µM was reported. Pumeraet al. presented a very easy-to-construct contactless conductivity detectorconsisting of two planar sensing aluminium film electrodes placed on theoutside of a polymeric microchip [1.25]. A 125 µM thick polymer(PMMA) sheet separates the electrodes and the channel. Detection limitsdown to 2.8 µM and 6.4 µM were reported for potassium and chloride,respectively.

1.2 Outline of the thesis

This thesis deals with the development and investigation of a newconductivity detector for application in capillary electrophoresismicrochips.

In Chapter 2, the detection methods that are most commonly employed incombination with capillary electrophoresis are reviewed. Their integrationand performance in CE microchips is presented. As far as conductivitydetection is concerned, a contactless configuration (metal electrodes

Chapter 1


3/2/2003

protected from the liquid with an insulating layer) is preferred over agalvanic one (metal electrodes in direct contact with the liquid).

Chapter 3 deals with the design of the conductivity detector. It isexplained that contactless conductivity measurements achieved with asingle pair of electrodes (2-electrode configuration) have a reducedsensitivity due to the presence of the insulating layer. In contrast with the2-electrode configuration, the use of four electrodes allows for sensitivedetection with varying carrier-electrolyte conductivity without requiringadjustment of the measurement frequency. Glass microchips with aminiaturized CE channel and an integrated contactless 4-electrodeconductivity detector were fabricated. The detector design and fabricationsteps are given at the end of Chapter 3.

A dedicated electronic interface was developed and is presented inChapter 4. Because of the small dimensions of the detector, the 4-electrode measurement setup is very sensitive to parasitics. Therefore, thetwo main features of the 4-electrode measurement setup had to bereconsidered: i) The control of a constant AC current flow between theouter electrodes. ii) The differential voltage measurement between theinner electrodes with a high input-impedance and a high common-moderejection ratio (CMRR) differential amplifier.

Separation and contactless detection interact in two ways. First, the thininsulating film is subject to breakdown when the separation voltagereaches a threshold value. Second, because the detector is placed at theend of the separation column, not all the detection current flows betweenthe detection outer electrodes, but a part of it leaks to one of the separationelectrodes (electrodes placed in the channel to apply the separationvoltage). A new decoupling setup was developed and is presented inChapter 5.

Chapter 6 deals with the measurements results. First, the measurementsetup is described. In order to improve the detection limit, a setup forelectronic baseline suppression was developed, which is presented in thischapter. The detector design allows for the comparison of the 2- and 4-electrode configurations. On the basis of reproducible separations of amixture of inorganic cations (K+, Na+ and Li+), the merits of the improveddetection concept is demonstrated. Additionally, six organic acids were


Introduction

3 February 2003, 16:39

separated and detected (at concentrations down to 50 µM) , which showsthe potential of the new detection method.

1.3 References

1.1 Dolnik V. et al., “Capillary electrophoresis on microchip”,Electrophoresis, 2000, 21, 41-54.

1.2 Bruin G.J.M., “Recent developments in electrokinetically drivenanalysis on microfabricated devices”, Electrophoresis, 2000, 21,3931-3951.

1.3 Kappes, T. et al., “Potentiometric Detection of Alkali and AlkalineEarth Metal Cations in Capillary Electrophoresis with SimplifiedElectrode Alignment and Enhanced Separation and Sensitivity”,Analytical Chemistry, 1998, 70, 2487-2492.

1.4 Fang, X. M. et al., “Capillary Electrophoresis with ElectrochemicalDetection for Chiral Separation of Optical Isomers”, AnalyticalChemistry, 1998, 70, 4030-4035.

1.5 Gavin, P. F. et al., “Characterization of Electrochemical ArrayDetection for Continuous Channel Electrophoretic Separations inMicrometer and Submicrometer Channels”, Analytical Chemistry,1997, 69, 3838-3845.

1.6 Haber, C. et al., “Quantitative Analysis of Anions at ppb/ppt Levelswith Capillary Electrophoresis and Conductivity Detection:Enhancement of System Linearity and Precision Using an InternalStandard”, Analytical Chemistry, 1998, 70, 2261-2267.

1.7 Huang X. et al., “End-column detection for capillary zoneelectrophoresis”, Analytical Chemistry, 1991, 63, 2193-2196.

1.8 Fracassi da Silva J.A., “An oscillometric detector for capillaryeletcrophoresis”, Analytical Chemistry, 1998, 70, 4339-4343.

1.9 Mayrhofer et al., “Capillary Electrophoresis and ContactlessConductivity Detection of Ions in Narrow Inner DiameterCapillaries”, Analytical Chemistry, 1999, 71, 3828-3833.

Chapter 1


3/2/2003

1.10 Zemann A. et al. “Contactless conductivity detection for capillaryelectrophoresis”, Analytical Chemistry, 1998, 70, 563-567.

1.11 Woolley, A. T. et al., “Capillary Electrophoresis Chips withIntegrated Electrochemical Detection”, Analytical Chemistry, 1998,70, 684-688.

1.12 Rossier, J. S. et al., “Electrochemical Detection in PolymerMicrochannels”, Analytical Chemistry, 1999, 71, 4294-4299.

1.13 Wang J. et al., “Integrated electrophoresis chips/amperometricdetection with sputtered gold working electrodes”, AnalyticalChemistry, 1999, 71, 3901-3904.

1.14 Zhao, H. et al., “Electrically floating conductivity detection systemfor capillary electrophoresis”, Chromatogr. A, 1998, 813, 205-208.

1.15 Mo, J. Y. et al., “An on-column miniature conductivity cell and aphotocouple separator for conductivity detection by capillaryelectrophoresis”, Analytical Communications, 1998, 35, 365-367.

1.16 Mikkers F.E.P. et al., "High-performance zone electrophoresis", J.of Chromatography, 1979, 169, 11-20.

1.17 Graβ B. et al., “A new PMMA-microchip device forisotachophoresis with integrated conductivity detector”, Sensors &Actuators B, 2001, 72, 249-258.

1.18 Masar, M. et al., “Conductivity detection and quantitation ofisotachophoretic analytes on a planar chip with on-line coupledseparation channels”, J. Chromatogr. A, 2001, 916, 101-111.

1.19 Prest, E.P. et al., “Determination of metal cations on miniaturisedplanar polymeric separation devices using isotachophoresis withintegrated conductivity detection”, The Analyst, 2001, 126, 433-437.

1.20 Galloway, M. et al., “Contact Conductivity Detection inPoly(methyl methacylate)-Based Microfluidic Devices for Analysisof Mono- and Polyanionic Molecules”, Analytical Chemistry, 2002,74, 2407-2415.

1.21 Guijt, R. M. et al., “New approaches for fabrication of microfluidiccapillary electrophoresis devices with on-chip conductivitydetection”, Electrophoresis, 2001, 22, 235-241.


Introduction

3 February 2003, 16:39

1.22 Laugere F. et al, “A novel high-resolution liquid-conductivitydetector”, Eurosensors XIII, 13th European Conference on Solid-State Transducers (the Hague), 1999, 211-214.

1.23 Laugere F. et al., “Downscaling aspect of a conductivity detector forapplication in on-chip capillary electrophoresis”, Sensors &Actuators A, 2001, 92, 109-114.

1.24 Lichtenberg, J.; Verpoorte, E.; de Rooij, N.F. “A Microchipelectrophoresis with integrated in-plane electrodes for contactlessconductivity detection“, Electrophoresis, 2002, 23, 3769-3780.

1.25 Pumera, M. et al., “Contactless Conductivity Detector for MicrochipCapillary Electrophoresis”, Analytical Chemistry, 2002, 74, 1968-1971.


3/2/2003


3/2/2003

2Biochemical detection in CE microchips

2.1 Introduction

In this chapter, the basics of capillary electrophoresis (CE) and of thedifferent detection techniques used in classical CE and CE microchips willbe explained. The most common detection techniques will be described,namely UV/Vis absorbance, fluorescence, mass spectometry,amperometry, potentiometry, and conductivity detection. Advancedknowledge about detection in classical CE can be found in books [e.g.2.1,2.2,2.3] and reviews [e.g. 2.4,2.5]. However, no books have yet beendedicated to detection in CE microchips. Nevertheless, scientificpublications are abundant, and many reviews regularly report aboutprogress in the field [1.1,1.2,2.6,2.7]. The major criteria that lead thechoice of a detection scheme are: the properties of the analyte (optical,electrical, electrochemical, and physical), the composition of the samplematrix (do the analytes have the similar properties and/or similar

Chapter 2


3/2/2003

concentrations, etc.), and the required detection limit. Other specificcriteria can be the price, ease of use and implementation, and, as far asmicrochips are concerned, the ease of fabrication and the possibility forminiaturization.

UV / Vis detection is a very popular detection technique in classical CE[2.8,2.9,2.10]. It is used in more than 50% of the CZE (capillary zoneelectrophoresis) applications today and it is found in all the commerciallyavailable CE apparatus. Direct and indirect absorbance measurementsmake absorbance a quasi universal detection technique. The detectionlimit for absorbance measurements in classical CE (direct and indirect) isin the micromolar range. However, absorbance detection is difficult in CEmicrochips because of the reduced optical pathlength (the intensity of themeasured signal is proportional to the detection volume).

Fluorescence detection is very sensitive, and a detection limit down to theattomolar (Alkaline Phosphatase molecules) was reached by using a laseras the light source (laser induced fluorescence, LIF) [2.11]. Therefore,when high performances are required, LIF is the preferred detectionscheme for classical CE, as well as for CE microchips. However, mostcompounds do not fluoresce (amino acids, proteins, and peptides) and apre- or postcolumn derivatization is necessary (adjunction of afluorophore to the analytes). Indirect LIF is another solution for thedetection of non fluorescent analytes, nevertheless poor (when comparedto direct LIF) detection limit have been reported (0.1 µM). This is partlydue to the fact that existing laser sources are unstable, yielding a noisybackground signal.

Electrochemical detection-techniques are very popular in classical CE butthey gained even more interest with the development of CE microchips.Miniaturized optical detectors, namely absorbance and fluorescencedetectors, have limitations that are not encountered with electrochemicaldetectors (see chapter 2.4.2). A large amount of publications are nowdedicated to amperometric detection in CE microchips. The interest inconductivity detection flared up very recently.

Other means of detection (laser-based thermo-optical detection, refractiveindex detection, radioscopic detection) will not be described in thischapter and detailed information can be found in [2.1,2.2,2.3].


Biochemical detection in CE microchips

3 February 2003, 16:39

2.2 Capillary electrophoresis

The purpose of capillary electrophoresis (CE) is to separate analyte ionspresent in a mixture (the sample matrix defines the type of analyte presentin the sample), to identify the different analytes, and to quantify theiramount (related as a concentration). The CE-separation mode describedhere after and illustrated in Fig. 2.1 is called capillary zone electrophoresis(CZE). Other CE-separation modes exist, namely isotachophoresis (ITP),isoelectric focussing (IEF), micellar electrokinetic chromatography(MEKC). Detailed explanations can be found in [2.2,2.3].

carrier electrolyte

injected sample

separated analytesH.V.

A B C

time

CB

A

H.V.A B C

wastesample

detectionwindow

Fig. 2.1 Capillary zone electrophoresis: injection, separation,and detection.

Chapter 2


3/2/2003

In CZE, the separation column is a fused silica capillary of usually 50 µminner diameter and of up to 400 µm outer diameter. The separation columnis filled with a carrier electrolyte (buffer), and the sample is injected in thesample reservoir. A potential-difference of several kV (field strengh up to500 V/cm) is applied between the sample and the waste reservoir. Underthe influence of the applied voltage, the analytes migrate towards thewaste reservoir. They will reach the waste reservoir at different timeintervals, each according to its respective velocity, thus mobility (alsocalled electrophoretic mobility). Separated zones will form, containingeach only one single type of analyte having the same mobility. At the endof the separation column, a detection window measures the amount ofanalytes and peaks are observed. The amplitude and/or area of the peaks islinked to the amount of analytes. The time at which the peak occurs islinked to the identity of the analyte (the analyte is identified through itsmobility).

An important phenomenon accompanying electrophoresis is thegeneration of the electroosmotic flow (EOF), which is the result frominteractions between the wall and the carrier electrolyte. In fused silicacapillaries, silanol groups (Si-O-H) become negatively charged (Si-O-H

↔ Si-O- + H+) when in contact with an acqueous solution. As the wallbecomes negatively charged, counter ions from the carrier electrolyte will

stern layer

diffuse layer

electrolyte

fused silica

Fig. 2.2 double layer formation at the interface between a fusedsilica wall and the electrolyte.



3 February 2003, 16:39

migrate towards the wall vicinity and accumulate there (Fig. 2.2). Thenegative charges from the wall facing the excess of positive charges fromthe electrolyte is called the double-layer [2.12]. The double-layer isformed by the stern-layer (positive charges from the electrolyte that arefixed to the wall by electrostatic forces) and a mobile diffuse layer (therest of the excess of positive charges). When the separation-voltage isapplied, the mobile diffuse layer will move towards the waste reservoir,drawing at the same time the whole carrier electrolyte. The analytes thatare injected will therefore migrate according to their own mobility and,additionaly, according to the EOF (in fact, most of the motion is providedby the EOF). The control of the chemical composition and pH of thecarrier electrolyte and the control of the amplitude of the separationvoltage is crucial for achieving sufficient separation. On the one hand,increasing the separation voltage is necessary in order to reduce the timethe analytes stay in the capillary and therefore to limit diffusion effects.Diffusion is a source of zone dispersion, which may result in anoverlaping of the zones. On the other hand, a too high value of theseparation voltage produces excessive joule heating causing again zonedispersion.

2.3 Detection in standard CE

The major detection techniques in classical CE will be briefly reviewed inthe following paragraphs. More detailed information concerning eachdetection method can be found in books and reviews such as[2.1,2.2,2.3,2.4,2.5].

2.3.1 UV / Vis Absorbance detection

In UV/Vis absorbance detection, the sample is excited by a light source ofintensity Io. The remaining intensity I, after passage of the light throughthe sample, is measured. The ratio of the two previous quantities is

Chapter 2


3/2/2003

denominated as the absorbance A and is linked to the concentration ofdiluted analytes according to the Lambert-law (also called Beer's law):

where d is the pathlength, c is the concentration of the diluted analyte andε its corresponding molar absorption coefficient. The Lambert law, aswritten above, is not directly applicable to circular capillaries. Diffractionand reflection effects strongly modify the pathlength, and therefore acorrection factor must be applied.

A simplified representation of the measurement setup is shown in Fig. 2.3.The detection window is created by removing locally a part of the plasticcoating (often polyimide) protecting the capillary. A narrow light beam (ofintensity Io) is focused onto the capillary, and the photosensor(photomultiplier or photodiode) measures the decrease of intensity due toabsorption in the sample. Common UV light sources are atomic lamp suchas mercury- (254 nm), cadmium- (229, 326 nm), zinc- (214 nm), oriodine-lamp (206, 270 nm). Optical slits and focusing optics are necessarybefore and after the capillary. The monochromator selects the desiredspectral components that reach the photosensor.

In some applications, the measurement of the absorbance versus time doesnot suffice to identify the nature of the analyte. For instance, two analytesthat have the same electrophoretic mobility cannot be identified from anelectropherogram. By measuring at multiple wavelength, a UV-spectrumis obtained of which the profile gives informations on the nature of theanalyte.

2.3.2 Fluorescence detection

Fluorescence detection is not only a very sensitive detection technique butit is, as well, very selective. The detection is applied to compounds thatcontain fluorophores. An excitation source (of intensity Io) is directed

AIIo

----log– ε c d,⋅ ⋅= = (2.1)



3 February 2003, 16:39

toward the sample and, consequently, fluorescence light is emitted. Theintensity of the emitted light F is measured and is linked to theconcentration of the analyte present in the sample.

where ϕ is the fluorescence yield, d is the pathlength, c is theconcentration of the diluted analyte and ε is its corresponding molecularabsorption coefficient.

The wavelength of the excitation source and of the fluorescent emitted-light differ, resulting in a very low measured background signal andconsequently in a low detection limit.

The optical scheme is very similar to the one described for UV/Vis in theFig. 2.3. The use of a laser as the excitation source increases the sensitivitybecause of its high intensity. Conventional lasers are the Ar+, the He-Ne,and the He-Cd. Semiconductor or diode lasers are promising because they

1

2

3

4

5

6 7

Fig. 2.3 On-column UV-Vis absorbance detection-setup. (1) lightsource, (2) slit, (3) capillary, (4) slit, (5) filter, (6) photodetector, (7) read-outelectronics.

F I0 ϕ 1 10 εcd––( ),⋅ ⋅= (2.2)

Chapter 2


3/2/2003

are more robust (longer life time), compact, but also more stable (this isimportant for indirect detection). Laser induced fluorescence (LIF) is, uptill date, the most sensitive detection technique in standard CE, anddetection limits reaching the femtomolar have been reported. A drawbackof using a laser is that a continuous variation of the wavelength is notpossible. Therefore, the sample identification as it is achieved with lamp-based UV/Vis fluorescence (but also absorbance) is not possible.Moreover, working at a fixed wavelength restrains to only few analytes.The use of chemistry and pre- or post-column derivatization with sometype of fluorophores allows the extension of fluorescence detection tomany analytes. The use of a fluorescent carrier electrolyte combined withindirect detection also extends the type of detectable analytes with LIF[2.10,2.14].

2.3.3 Amperometric detection

Amperometric detection makes use of three electrodes: the auxiliary, thereference, and the working electrode (Fig. 2.4). The two first electrodesare used to monitor the potential in the solution, and the working electrodeis the detection electrode. A constant potential is kept between the solutionin the cell and the working electrode. Redox reactions of the analyte at theworking electrode produces the detection signal. The generated current Iis proportional to the amount of redox reactions, thus to the concentrationof the analytes.

where z is the number of electrons transferred in the reaction, F is thefaraday constant, ci is the concentration of analytes that enter the detectioncell, and c0 is the concentration of analytes that exit the detection cell. Thecoulombic efficiency is defined as the percentage of an electroactiveanalyte undergoing an electrode reaction inside the cell, and is linked to ci

I z F ci c0–( ),⋅ ⋅= (2.3)



3 February 2003, 16:39

and c0. The detection sensitivity is strongly linked to the coulombicefficiency.

The reference electrode is in most applications a silver-silver chlorideelectrode (Ag/AgCl). However, the material for the working electrodediffers according to the analyte ions that has to be detected. The employedmaterial are: gold, platinum, palladium, copper, and carbon (carbon ink,carbon paste, carbon fibre).

Amperometric detection is comparable to LIF detection: it is selective(only electroactive analytes are detected) and is has a very low detectionlimit (femtomolar).

The electrochemical reactions taking place at the electrodes may result ina modification of their surface. Fouling effects are prevented by pulseamperometric detection (PAD). The polarisation voltage is applied foronly a short time and is followed by a reverse polarisation in order to cleanthe electrodes.

Classical on-column detection is not possible due to the interactionbetween the separation voltage and the working electrode. Othermeasurement configurations were found in order to apply amperometricdetection to CE [2.15]. Off-column detection using porous conductiveglass [2.16,2.17] or Nafion [2.18] as a decoupler has been developed. Theporous material ensures the electrical contact between the inside of thecapillary and the external grounded electrolyte (Fig. 2.5). The jointbehaves as the cathode for the separation. The ions cannot go through the

R

A WA1 A2

Vr

Vout

Fig. 2.4 Electronic setup for amperometric detection. The potentialin the liquid is controlled by the reference electrode R, the auxiliary electrode A,and the amplifier A1. The current at the working electrode W is measured withthe transimpedance amplifier A2.

Chapter 2


3/2/2003

porous material and they continue their way, driven by electroosmoticpressure, up to the detector. Because the working electrode is still in theseparation column, this configuration has the advantage not to cause zone-dispersion. Another approach, which is also illustrated in Fig. 2.5, consistsin placing the working electrode directly at the end of the capillary[1.7,2.19,2.20]. On the one hand, this solution is easy to implement, but onthe other hand it requires an accurate placement of the electrodes (in orderto prevent the zone-dispersion).

2.3.4 Potentiometric detection

Potentiometric detection is similar to amperometric detection in the sensethat both involve electrochemistry at the surface of a chemically-activeelectrode. In the case of amperometry, a known electrode potential isapplied, and the current resulting from redox activity is measured. Thisimplies that the electrode has a low impedance in order to allow the flowof the redox current. In the case of potentiometry, this is the opposite. Ahigh electrode-impedance keeps the current negligible, and the redoxactivity at the electrode lead to accumulation of charge at the surface. Theresulting potential-drop, denominated as the Nernst potential, is measured.

Detectorseparationcapillary

porousjoint

conductivesolution

R

W

Aseparationcapillary

Fig. 2.5 Off-column configuration with decoupler (right) andend-column amperometric detection (left).



3 February 2003, 16:39

In CE, when a zone passes into the detection cell, a change ∆V of theNernst potential is observed.

where R is the gas constant, T is the absolute temperature, F is theFaraday constant, Z is the charge of the ion i and j, ai is the activity of thesample ion, aj is the activity of the background ion, and kij is theselectivity factor.

First potentiometric measurements in CE were done in an end-columnfashion [2.21]. But it was also shown that on-column detection is possiblebecause of the high electrode-impedance which does not driveelectrophoretic current.

2.3.5 Mass spectometric detection

A mass spectrometer (MS) does not only measure the amount of analytes,but it also achieves separation of analytes according to their mass-to-charge-ratio (m/z). The coupling of MS with CE leads to a very powerfulseparation tool which allows the analysis of complex mixtures. Detectionlimit and sensitivity are comparable to the ones obtained with UV-Visabsorbance detection.

The working principle of a mass spectrometer (as used in CE) is illustratedin the Fig. 2.6. The analytes are accelerated in the RF quadrupole(polarized column). The time they pass along the column (called the timeof flight TOF) depends on their m/z, thus the type of ion. When reachingthe end of the column, they are deflected (by the ion multiplier) on amultipoint ion collector. The curvature of the ion path is also dependent ontheir respective m/z. Finally a current is generated at each ion collector,and this current is proportional to the amount of ions reaching the target.

Work was done in developing a suitable interface between the massspectrometer and the end of the CE-column. Two major coupling methods

V∆ 2.303 RTZF-------

ai aj–( ) kijaj+

ai

------------------------------------ ,log⋅ ⋅= (2.4)

Chapter 2


3/2/2003

are suitable: electrospray ionization ESI [2.22,2.23] and fast atombombardment FAB [2.24].

2.3.6 Conductivity detection

In conductivity detection, the concentration of analyte is determined bythe measurement of the liquid conductivity. Two electrodes are placed inthe capillary. An alternating voltage vin is applied between the electrodesand the resulting current io is measured. The value of the current is

proportional to the liquid conductivity κ (S.m-1).

where K is the cell constant (m-1).

The basis of conductivity measurements were first introduced byEveraerts [2.25,1.16]. In his work, conductivity detection was applied toHPCE (high-performance capillary electrophoresis) in large-scale

separationcapillary interface RF quadrupole

electronmultiplier

multipoint ioncollectormass spectrometer

Fig. 2.6 Experimental setup for CE-MS measurements.

io

vo

K----- κ,×= (2.5)



3 February 2003, 16:39

capillaries (I.D. 0.2 mm, O.D. 0.35 mm). The disadvantage of using baremetal electrodes in direct contact (galvanic-contact) with the liquid is thatelectrode processes can occur, which leads to a poor reproducibility of themeasurements. Gaš improved the detector by using contactless-electrodes(a four electrodes conductivity detector is described) [2.26]. A capacitivecontact with the liquid is created by placing the electrodes outside thecapillary (on the outer-surface of the capillary). The advantage of thecontactless configuration is that the detector is fully decoupled from thehigh-voltage applied for separation and is easy to construct. The first on-column detection in fused silica capillaries (O.D. 0.36 mm, I.D. 50 or 75µm) was reported by Huang and Zare [2.27]. In their work, two facing-holes were drilled by CO2 laser in the capillary’s wall. 25 µm Platinumwire were inserted in the holes to make the electrical contact with theliquid. The electrodes and the read-out electronics were electricallydecoupled from the high separation voltage by using a transformer. Inother works, decoupling capacitors were used instead of a transformer.More recently, contactless conductivity detection in fused-silicacapillaries (making use of 2 electrodes) was reported [1.8,1.10].

If conductivity detection has been extensively used in ionchromatography, it is still not the case for CE. An explanation lies in thefact that conductivity detection combined to CE suffers from conflictingrequirements. It was demonstrated that the sensitivity of the conductivitydetection is linked to the ratio of the background co-ion and analyte ionmobility. The detector is more sensitive when the two mobility-valuestend to differ. However, it was also shown that the difference ofelectrophoretic-mobility values is a source of zone-dispersion. The zonestend to broaden and tailing or fronting effects are observed [2.28,2.29].The zone-dispersion reduces the effective separation efficiency and thepeak height [2.30,2.31].

Conductivity detection combined with chemical-suppression allows toenhance the detection sensitivity in the presence of a highly conductivebuffer [2.32,2.33]. A chemically active membrane is placed before theconductivity detector. This membrane chemically interacts with the buffer

exchanging Na+ ions for Cl- ions (consequently, the value of thebackground conductivity decreases). A detection limit of 10 nM wasreported for inorganic ions [2.33].

Chapter 2


3/2/2003

2.4 Detection in CE microchips

2.4.1 CE microchips and Micro Total Analysis Systems

Capillary electrophoresis microchips is part of the recent and fast-expending field denominated as µTAS (micro total analysis systems)[2.34]. The field of µTAS involves Flow based systems and Microfluidics,Microtechnology, Microreactors, Microchemistry, Separation Systems,and Instrumentation. µTAS aims at miniaturising complete bioanalyticalsystems and at integrating them on a chip platform. Samples may beprepared and analysed on the same miniaturised platform. Many analysiscan be simultaneously and quickly run yielding ultra high throughput. Italso has the advantage that only a little amount of sample and reagent areused in the analysis process.

The first integration of CE on-chip was reported in the early 90’s [2.35].The fabrication of CE microchips is based on the mature semiconductortechnology, which should yield possible mass-production in the future.

The chips are fabricated in glass (pyrex or quartz), using the standardphotolithographic techniques and thin film deposition (metal-, photoresist-, oxide-layer). Channels for sample injection and separation are etched ina first substrate. This substrate is bonded by anodic or fusion bonding to asecond substrate in order to form a sealed channel. Channel dimensionsare 15-40 µm high, and 60-200 µm wide. Glass is usually preferredbecause of its optical transparency, the high EOF that it generates, andalso the possibility to create ultra smooth channel surfaces (by wet etchingtechniques). Recently, polymeric materials including poly(methylmethalcrylate) (PMMA), poly(ethyleneterepthalate) (PET), orpoly(dimethylsiloxane) (PDMS) were also used. They require differentprocessing-techniques like laser ablation, injection molding, hotembossing, and casting. The advantages of polymeric material are a fastand easy fabrication process, and sometimes interesting chemicalbehaviours (e.g. no adsorption of proteins on the wall).



3 February 2003, 16:39

The generally accepted configuration of a CE microchip consists of twocrossed-channels connected to four reservoirs (Fig. 2.7). The two channelsare connected by either a cross injector – the channels are perfectlyorthogonal – or by a double-T injector – there is an offset at theconnecting point. The injection procedure follows two steps: First, thesample migrates through the sample-channel while a voltage is appliedonly between the sample1 and the waste1 reservoir. Second, by permutingthe voltage to the sample2 and the waste2, the separation becomeseffective, while injection stops. This procedure yields a reproducible andwell volume-defined injected zone.

The first CE microchips were fabricated on a relatively large-scale andthey included straight channels of several centimetres-long. With theconstant reduction of the chip sizes, the channel must be folded in order toensure the minimal required length. However, it was shown that abruptturns induce dispersion in the separated zone [2.36]. To overcome thisproblem, the recent designs includes spiral-like channels [2.37] with aradius which is kept much larger than the channel width. Injecting ashorter plug (by reduction of the injector geometry or by stacking-

cross injector

double-T injector

V+ V

_

float.

float.

float. float.

V+

V_

injection

separation

sample1

waste1

sample2

waste2

Fig. 2.7 layout of a CE microchip with four-port injection.

Chapter 2


3/2/2003

injection techniques) is also a solution that allows shorter separation-channel [2.38].

2.4.2 Detection systems

Although a mass spectrometer cannot be miniaturized, mass spectrometricdetection was transposed to the microchip format. Studies have focused ondeveloping an interface between the microchannel and the (bulky) massspectrometer. The interface is an electrospray ionization (ESI) systemwhich requires that a voltage is generated between the end of the channeland the mass spectrometer. This voltage is at the same time the cathodicvoltage necessary for the electrophoretic separation. The voltage isgenerated either by use of a side-channel [2.48], or by the use of a porousglass-membrane [2.49].

Absorbance detection, which is very common in classical CE, is notproper to the CE microchip format. The reduction of the optical pathlengthhas a severe impact on the detection-sensitivity and it makes solutes withpoor absorptivity difficult to detect. Nevertheless, it was shown thatsensitivity could be improved by increasing the pathlength: amultireflection cell allows the light beam to pass in the longitudinaldirection along the flow channel [2.39].

Laser induced fluorescence (LIF) is the most popular detection techniquefor CE microchip because of its ultrahigh sensitivity. The latest setupsapplied in standard CE have, therefore, been transposed to measurementsin microchips. Red diode lasers are preferred because they are more stableand more compact (thus suitable for miniaturisation) than gas phase lasers.A system based on a 635 nm stable red diode laser for detection on amicrochip platform was described [2.40]. However, only a limited amountof analytes can be excited at that wavelength. Recently, a compact andpowerful violet diode emitting at 405 nm with a power of 5 mW waspresented [2.41]. The miniaturisation of the photodetector is also animportant topic. By making channels in silicon, the integration ofphotodiodes is possible and it results in a very compact system [2.42].Very low detection limits were reported (pM of fluorescine molecules)with LIF.



3 February 2003, 16:39

LIF also has disadvantages. Few analytes have the ability to fluoresceunder excitation and pre- or postcolumn derivatization is a necessary step.The derivatization involves the integration of extra-elements such asmixers and extra reservoirs, which makes the system either less compactand/or more complex to fabricate. Furthermore, the fluorophores are oftenbulky and they eliminate at least one charged group. It may result in aquasi-similar electrophoretic mobility for all the analyte ions, andconsequently a longer separation channel is necessary. Therefore, otherdetection schemes have been investigated and electrochemical detectionhas appeared as a promising one. First of all, the measurement potential isnot compromised by miniaturisation, as it is the case with opticaldetection. As far as amperometric and potentiometric detectors areconcerned, it was shown that their miniaturisation contributes to animproved sensitivity and detection limit. The flux and the current densityare increased, and at the same time the capacitive current is decreased.The integration of sensing electrodes in the microchannel only requires afew extra standard steps in the fabrication process. Furthermore, dedicatedread-out electronics can be processed in silicon-based microchips,opening the door to fully integrated smart-sensors.

Potentiometric detection has not been yet implemented in a CE microchip.However, many groups have worked on amperometric detection.Microchips including sensing electrodes were fabricated both in glass andin polymer. Separation and detection were shown in a glass/PDMS-basedmicrochip, for which the channel could be easily unbonded for microchipcleaning or channel changing [2.43]. As for optical detection, thedetection scheme originates from classical CE. The detector wasintegrated most often in an end-column [2.44,1.13] fashion, but also in anoff-column fashion using a decoupler [2.45,2.46]. The decoupler wasrealised either by fabricating a porous membrane (matrix of holes of 10µm diameter) or by using a palladium film as the cathode. Palladium hasthe property to absorb the hydrogen generated at the cathode duringelectrophoresis, which eliminates the risk of gas-bubble generation intothe channel. All the electrode materials used in classical CE weretransposed to the microchip format. However, carbon electrodes are notyet compatible with microfabrication processes and carbon fibres have tobe manually positioned and glued onto the microchip.

Chapter 2


3/2/2003

Till date, few microchips including a conductivity detector have beenpresented. A PMMA microchip for isotachophoresis [1.17] with twoplatinum electrodes was used for separation of organic acids down toconcentrations of 100 µM. CZE separations of organic acids were alsomonitored by conductivity detection in a glass microchip and with twoplatinum electrodes [2.47]. Separations of inorganic ions in glassmicrochips with two contactless electrodes were reported by Lichtenberget al. [1.24]. The works previously mentioned proved that conductivitydetection in the microchip format is viable. However, the reporteddetection limits remain high (50 µΜ for potassium) compared to thevalues reached in standard CE.

In the following chapters, we are going to describe the design, thefabrication, and the experimental setup (electronics and fluidic setup) of aCE microchip that includes a new miniaturised conductivity-detectionscheme: the contactless four-electrode detection. Robustness andperformance in terms of sensitivity and detection limit have beenconsidered.

2.5 References

2.1 Khaledi G.M., “High-performance capillary electrophoresis :theory, techniques, and applications”, Wiley New York, 1998.

2.2 Li S.F.Y., “Capillary electrophoresis; principles, practice andapplications”, Elsevier Amsterdam, 1992.

2.3 Foret F. et al., “Capillary zone electrophoresis”, Veinheim: VCH,1993.

2.4 Swinney K. et al., “Detection in capillary electrophoresis”,Electrophoresis, 2000, 21, 1239-1250.

2.5 Ewing A.G. et al., “Electrochemical detection in microcolumnseparations”, Analytical Chemistry, 1994, 66, 527A-536A.

2.6 Lacher N.A. et al., “Microchip capillary electrophoresis/electrochemistry”, Electrophoresis, 2001, 22, 2526-2536.



3 February 2003, 16:39

2.7 Schwarz M.A. et al., “Recent developments in detection methods formicrofabricated analysis devices”, Lab on a Chip, 2001, 1, 1-6.

2.8 Hjerten S., “High performance electrophoresis elimination ofelectroendosmosis and solute adsorption”, J. of Chromatography,1985, 347, 191-198.

2.9 Walbroehl Y. et al., “On-column UV absorption detector for opentubular capillary zone electrophoresis”, J. of Chromatography,1984, 315, 135-143.

2.10 Hjerten S. et al., “Carrier-free zone electrophoresis, displacementelectrophoresis and isoelectric focussing in a high-performanceelectrophoresis apparatus”, J. of Chromatography, 1987, 403, 47-61.

2.11 Craig D.B. et al. “Detection of attomolar concentrations of alkalinephosphatase by capillary electrophoresis using laser-inducedfluorescence detection”, Analytical Chemistry, 1996, 68, 697-700.

2.12 Delahay P., “Double layer and electrode kinetics”, InterscienceNew-York, 1965.

2.13 Jorgenson J.W. et al., “Zone electrophoresis in open-tubular glasscapillaries”, Analytical chemistry, 1981, 53, 1298-1302.

2.14 Kuhr W.G. et al., “Optimization of sensitivity and separation incapillary zone electrophoresis with indirect fluorescence detection”,Analytical chemistry, 1988, 60, 2642-2646.

2.15 Kappes T. et al., “Recent developments in electrochemical detectionmethods for capillary electrophoresis”, Electroanalysis, 2000, 12(3), 165-170.

2.16 Wallingford R.A. et al., “ Capillary zone electrophoresis withelectrochemical detection”, Analytical Chemistry, 1987, 59, 1762-1766.

2.17 Yik Y.F. et al., “Micellar electrokinetic capillary chromatography ofvitamin B6 with electrochemical detection”, J. of Chromatograph.,1991, 585, 139-144.

2.18 O’Shea T.J. et al., “ Capillary electrophoresis with electrochemicaldetection employing an on-column Nafion joint”, J. ofChromatography, 1992, 593, 305-312.

Chapter 2


3/2/2003

2.19 Matysic F.M., “Improved end-column amperometric detection forcapillary electrophoresis”, J. of Chromatography A, 1996, 742, 229-234.

2.20 Matysic F.M., “End-column electrochemical detection for capillaryelectrophoresis”, Electroanalysis, 2000, 12 (17), 1349-1355.

2.21 Haber C. et al., “Potentiometric detector for capillary zoneelectrophoresis”, Chimia, 1991, 45, 117-121.

2.22 Smith R.D. et al., “Capillary zone electrophoresis-massspectrometry using an electrospray ionization interface”, Analyticalchemistry, 1988, 60, 436-441.

2.23 Lee E.D. et al., “Liquid junction coupling for capillary zoneelectrophoresis/ion spray mass spectometry”, Biomedical andenvironmental mass spectometry, 1989, 18, 844-850.

2.24 Moseley M. et al., “Coupling of capillary zone electrophoresis andcapillary liquid chromatography with coaxial continous-flow fastatom bombardment tandem mass spectrometry”, J. ofchromatography, 1989, 480, 197-209.

2.25 Everaerts F.M. et al., “Isotachoporesis-thoery, instrumentation andapllication”, Elservier Amsterdam, 1976.

2.26 Gaš B. et al., “High-frequency contactless conductivity detection inisotachoporesis”, J. of Chromatography, 1980, 192, 253-257.

2.27 Huang X. et al., “On-column conductivity detector for capillary zoneelectrophoresis”, Analytical Chemistry, 1987, 59, 2747-2749.

2.28 Mikkers F.E.P. et al., “Concentration distributions in free zoneelectrophoresis”, J. of Chromatography, 1979, 169, 1-10.

2.29 Hjerten S., “Zone broadening in electrophoresis with specialreference to high-performance electrophoresis in capillaries: aninterplay between theory and practice”, Electrophoresis, 1990, 11,665-690.

2.30 Gebauer P. et al., “Prediction of zone patterns in capillary zoneelectrophoresis with conductivity detection, concept of the zoneconductivity diagram”, J. of Chromatography A, 1997, 771, 63-71.

2.31 Gaš B. et al., “Optimization of background electrolytes for capillary



3 February 2003, 16:39

electrophoresis, I Mathematical and computational model”, J. ofChromatography A, 2001, 905, 269-279.

2.32 Small H. et al., “Novel ion exchange chromatographic method usingconductimetric detection”, Analytical Chemistry, 1975, 47, 1801-1809.

2.33 Avdalovic N. et al., “Determination of cations and anions bycapillary electrophoresis combined with suppressed conductivitydetection”, Analytical Chemistry, 1993, 65, 1470-1475.

2.34 Freemantle M., “Downsizing chemistry, chemical analysis onmicrochips promise a variety of potential benefits”, Chemical &Engineering News, 1999, 77 (8), 27-36.ssss

2.35 Manz A. et al., “Miniaturized total chemical analysis systems: anovel concept for chemical sensing”, Sensors & Actuators B, 1990,7, 244-248.

2.36 Culbertson C.T. et al., “Dispersion sources for compact geometrieson microchips”, Analytical Chemistry, 1998, 70, 3781-3789.

2.37 Culbertson C.T. et al., “Microchips devices for high-efficiencyseparations”, Analytical Chemistry, 2000, 72, 5814-5819.

2.38 Jacobson S.C. et al., “Elecktrokinetic focussing in microfabricatedchannel structures”, Analytical Chemistry, 1997, 69, 3212-3217.

2.39 Salimi-Moosavi H. et al., “A multireflection cell for enhancedabsorbance detection in microchip-based capillary electrophoresisdevices”, Electrophoresis, 2000, 21, 1291-1299.

2.40 Jiang G. et al., “Red diode laser induced fluorescence detection witha confocal microscope on a microchip for capillary electrophoresis”,Biosensors & Bioelectronics, 2000, 14, 861-869.

2.41 Melanson J.E. et al., “Violet (405 nm) diode laser for laser inducedfluorescence detection in capillary electrophoresis”, The Analyst,2002, 125, 1049-1052.

2.42 Webster J.R. et al., “Monolithic capillary electrophoresis devicewith integrated fluorescence detector”, Analytical Chemistry, 2001,73, 1622-1626.

2.43 Martin R.S. et al., “Dual-electrode electrochemical detection for


3/2/2003

poly(dimethylsiloxane)-fabricated capillary electrophoresismicrochips”, Analytical chemistry, 2000, 72, 3196-3202.

2.44 Henry C.S. et al., “Ceramic microchips for capillary electrophoresis-electrochemistry”, Analytical communications, 1999, 36, 305-307.

2.45 Rossier J.S. et al., “Electrophoresis with electrochemical detectionin a polymer microdevice”, J. of Electroanalytical Chemistry, 2000,492, 15-22.

2.46 Chen D.C. et al., “Palladium film decoupler for amperometricdetection in electrophoresis chips”, Analytical Chemistry, 2001, 73,758-762.

2.47 Schlautmann S. et al., “Powder-blasting technology as an alternativetool for microfabrication of capillary electrophoresis chips withintegrated conductivity sensors”, J. of Micromechanics andMicroengineering, 2001, 11, 386-389.

2.48 Ramsey R.S. et al., “Generating electrospray from microchipdevices using electroosmotic pumping”, Analytical Chemistry,1997, 68, 1174-1178.

2.49 Lazar I.M. et al., “Novel microfabricated device forelectrokinetically induced pressure flow and electrospray ionizationmass spectrometry”, J. of Chromatography A, 2002, 891, 195-201.


3/2/2003

3On-chip conductivity detection

3.1 Introduction

There are limitations when using bare metal electrodes (galvanicdetection) for conductivity detection in capillary electrophoresis (CE).Electrochemical reactions can take place at the electrodes due to theelectrical field which is applied for separation. Everaerts, who appliedconductivity detection to isotachophoresis in large bore capillaries,extensively described the limitations of galvanic conductivity detection.Authors, among whom Gas, Kaniansky, and Zemann, suggested to protectthe metal electrodes with an insulating material having chemical inertnessproperties. The so-called contactless detector is capacitively coupled tothe liquid.

In this chapter, we discuss the possibility to integrate a contactlessconductivity detector in a microchannel. The possibility to perform

Chapter 3


3/2/2003

conductivity measurement, the design and the technology of the detectorare discussed.

It is shown that the capacitive coupling highly deteriorates the detectioncompared to galvanic detection. The conductivity cell is represented by anelectrical network for which the negative contribution of the capacitivecoupling is demonstrated. Four electrodes are used instead of two becausethis theoretically suppresses the influence of the capacitive coupling. Atest setup allowing contactless 2- and 4-electrode detection was made andmeasurements confirmed the theory.

The aspects to be considered for the design of the detector in amicrochannel are the size of the electrodes and the properties of theinsulating film (electrical breakdown strength, chemical inertness,permittivity, and thickness).

Of the various insulating films, we decided to use silicon carbide (SiC).SiC is known to have a high chemical inertness and is a material availablein our cleanroom. Furthermore, thin films (30 nm) with acceptablepermittivity can be obtained by standard PECVD (plasma enhancedchemical vapour deposition). The complete fabrication process isdescribed, including the realization of the CE channel and the patterningof the electrodes.

3.2 Contactless detection versus galvanic detection

Metal electrodes in contact with an electrolyte undergo differentelectrochemical processes.

Redox reactions imply the exchange of electrons between the electrolyteand the electrode. In some cases, electron transfers do not modify thesurface of the electrode. In other cases, electron transfers that causescorrosion, or the growth of a metal or an oxide layer do modify the surfaceof the electrode. Gas generation can be the result of redox reactions.


On-chip conductivity detection

3 February 2003, 16:39

Another type of electrochemical reaction is the adsorption of ions at theelectrode surface. Adsorption assumes electron transfers at the electrodesurface, but without specific interaction between the surface and theelectrolyte ions. In the case of adsorption, the concentration of specificions at the electrode surface is greater than can be accounted for byelectrostatic interactions.

In [3.1], the electrochemical behaviour of different metal electrodes incontact with an electrolyte as well as the resulting influence onconductivity measurements is explained. The electrochemical activity ofthe electrode depends on the concentration of the electrolyte, the pH level,the electrical-signal amplitude and frequency, the electrode material andits treatment, and the electrode history.

Electrochemical reactions originate from and/or depend on the differencein energy level between the electrode and the electrolyte. As a matter offact, applying a potential difference between the electrode and theelectrolyte initiates electrochemical reactions. This is the case forconductivity detection in CE where metal electrodes that are held toground are placed in contact with a liquid in which an electrical potentialis applied (for separation). Conductivity detection in isotachophoresis hasbeen extensively studied by Everaerts [3.2], who considered theelectrochemical modification of the electrodes and the effect on detection.Everaerts showed that electrode surface modifications lead to an unstableresponse over time and that gas generation is a problem because itdecreases the sensing area or blocks the path for the electrophoreticcurrent (in the worse case).

In order to prevent unwanted electrochemistry at the electrodes due to theseparation voltage, different configurations were proposed. In an on-column fashion, the electrochemical activity of the metal electrodes isprevented by cancelling the potential difference between the liquid and theelectrodes. This is done by using an isolation transformer, a battery, ordecoupling capacitors. In an off-column configuration, the sensingelectrodes are placed outside the separation field. In order to avoid zonebroadening, the electrodes are placed inside the capillary and the capillaryis grounded prior to the sensing electrodes by means of a porous glassstructure or by means of an on-column frit structure. In an end-columnconfiguration, the end part of the capillary, where the detector is placed, is

Chapter 3


3/2/2003

enlarged so that the electrical-field strength in that region is lower thanthat in the separation part. This considerably reduces the influence of theseparation voltage on the sensing electrodes, but it also causes unwantedzone broadening.

Protecting the metal electrodes with an insulating material was presentedas a solution which makes the detector more robust, usable in an on-column configuration (that prevents extra zone broadening), andcompatible with all standard reagents. In addition, the fabrication andmaintenance is very straightforward. This was originally pointed out byEveraerts [3.2], and Gas [3.3], who demonstrated the benefits ofcontactless conductivity detection for monitoring isotachophoresisseparation. Kaniansky [3.4], Zemann [3.5], and Fracassi da Silva [3.6]extended the application to capillary zone electrophoresis in small innerdiameter capillaries (50 µm).

The metal electrodes of our detector were also protected by means of aninsulating material. More detailed informations on the choice of insulatingmaterials will be given in Chapter 3.4.

3.3 4-electrode versus 2-electrode configuration

3.3.1 Electrical model of a conductivity cell

3.3.1.1 Calculation of the cell constant

The classical cell configuration consists of two metal electrodes, whichare in contact with the solution (galvanic detection). An alternatingvoltage source vin is connected to one electrode, while the other electrodeis held to ground. Under the presence of the electric field, the cationsmigrate toward the negative electrode (the cathode), and the anions towardthe positive electrode (the anode). The resulting current io is measured and

is related to the conductivity κ of the solution through the value of the



3 February 2003, 16:39

equivalent measured resistance Rliq (Fig. 3.1). A liquid also has adielectric behaviour which dominates the resistive behaviour but only athigh frequencies. A capacitor of value Cliq, placed in parallel with theresistor Rliq, symbolises the dielectric behaviour of the liquid. The value

Cliq of that capacitor is linked to the dielectric constant of the liquid εr

(e.g. εr = 80.2 for pure water at 293 K and 1 bar) so that:

The value of the conductivity κ (S.m-1) and the measured resistance Rliq

(Ω) are linked to each other through the cell constant K (m-1).

The cell constant is calculated according to the dimensions of theconductivity cell, i.e. the electrode area and spacing. In the case of aclassical parallel plate configuration (Fig. 3.2), where A (A = w × L) is the

Rliq Cliq⋅ε0εr

κ---------- ε0 8.85 10

12–×=( ).= (3.1)

+

_

+

iovin

Cliq

Rliq+

_

+

iovin

electrolyte

electrodes

Fig. 3.1 Simple electrical model of the conductivity cell. The liquidresistivity and permittivity is symbolized by a resistor R and a capacitor C,respectively.

κ KRliq

---------= (3.2)

Chapter 3


3/2/2003

area of the electrodes and l the distance between the electrodes, then l/A is

known as the cell constant. This is a very straightforward approximation

of the cell constant value which does not take into account any fringing

effects, but which, nevertheless, gives an acceptable value. In the case of a

planar electrode arrangement (Fig. 3.2), there is no equivalent linear

equation such as for a parallel plate configuration. Nevertheless, an

analytical formula exists which is based on the theory of conformal

transformations [3.7, 3.8]. The dimensions are illustrated in Fig. 3.2.,

where w is the width of the electrodes, L the length of the electrodes, and l

the spacing between the electrodes. The cell constant is calculated

according to the following sequence (note: the following set of equations

is valid for L >> w):

The modulus k and the complementary modulus k’ are related to the

electrode layout:

The solution for the cell constant yields:

k1

12.w

l--------+

------------------=(3.3)

k′ 1 k2

–= (3.4)

K2.K k( )K k ′( )---------------- 1

L---,⋅= (3.5)



3 February 2003, 16:39

where K(k) is the complete elliptic integral of the first kind defined as:

For almost all other electrode configurations, there is no analytical modelavailable. Nevertheless, the value of the cell constant can be computed byusing software that solves differential equations by means of the finiteelements method (i.e. Ansys, or Coventer).

3.3.1.2 The interface electrode-electrolyte: the double layer

At any interface between a metal and an electrolyte there is always adifference between the charge density in the metal and in the electrolyte.Consequently, charges from the solution (ions) accumulate on theelectrode in order to, locally, maintain electrical neutrality. Thisaccumulation of ions forms the electrical double layer [3.9]. In the

K k( ) 1

1 x2

–( ) 1 k2x

2–( )⋅

---------------------------------------------------- x.d0

1∫= (3.6)

Parallel plates configuration

Planar configuration

w

l

L

wl

L

Fig. 3.2 Conductivity cell with parallel-plates configuration andplanar configuration.

Chapter 3


3/2/2003

electrolyte, this excess of charges is located in a region, whose extensionis the greater the lower the ionic concentration of the electrolyte. This pureelectrostatic phenomenon is electrically represented by a double layercapacitance Cdl. Helmholtz, in 1879, considered the double layer asequivalent to a parallel plate capacitor where charges of opposite signwould face at a distance of about one molecular diameter. Gouy, in 1910and 1917, and Chapman, in 1913, enhanced the model by taking intoaccount the fact that the charges from the electrolyte accumulating at thesurface of the electrode also tend to diffuse back into the electrolyte. Sincethen, several improved models (based on the Gouy-Chapman, andHelmholtz theory) have been proposed to explain the origin and the orderof magnitude of the double layer capacitance. An accurate prediction ofthe double layer capacitance value Cdl is rather difficult.

The double layer capacitance is proportional to the electrode area, so that:

where A is the electrode area [m2] (w×L in Fig. 3.2) and Co is the double

layer capacitance per unit area [F.m-2]. Co is a value per unit area, whichmeans that the larger the electrode, the larger Co. It is most often the valueof Co which is found in the literature. Experiments showed that the value

of Co is in the range of 10.10-2 to 40.10-2 F.m-2 [with mercury electrode inref. 3.11, and with gold electrode in 3.12 and 3.13].

The impedance ZCdl of the double layer capacitor is not expressed like fora classical capacitor, but it takes into account a fractional exponent n[3.10].

Cdl Co A× ,= (3.7)

ZCdl1

j 2π f⋅⋅( )nCdl⋅

--------------------------------------,= (3.8)



3 February 2003, 16:39

where j is the imaginary unit, and f the measurement frequency [Hz]. Thefractional exponent has a value usually between 0.80 and 0.95 for noblemetals. The most likely explanation of this non-ideal capacitive behaviouris attributed to the surface roughness of the electrodes [3.10].

The presence of ions in the vicinity of the electrode leads toelectrochemical reactions. Electrons migrate from the electrode to thesolution (or the inverse), oxidizing or reducing the ions from theelectrolyte. This creates an electronic current called the Faradaic current.An impedance called the Faradaic impedance (Zf) symbolizes the

electrochemical and the mass transport processes at the interface. As forthe double layer capacitance, the estimation of a generic Faradaicimpedance magnitude is rather difficult since it depends on theelectrochemistry at the electrode.

The double layer capacitance Cdl, in parallel with the Faradaic impedance

Zf, forms the double layer impedance Zdl (Fig. 3.3). In parallel to the

Faradaic impedance, an adsorption capacitance might be added whenadsorption phenomena occur. When the electrochemistry at the electrodeis known, the electrical model might be enhanced by adding extra resistorsand capacitors. This can lead to complex frequency-dependent models ofthe interface liquid-electrode for which the value of the electrical elementscan be experimentally retrieved by Electrochemical ImpedanceSpectroscopy (EIS) [3.14].

Zf

Cdl

Zdl

Fig. 3.3 Simple electrical model of the double layer impedance Zdl.

Chapter 3


3/2/2003

3.3.1.3 The impedance of the insulating material

The double layer effect described previously for metal electrodes also

occurs for an insulating material in contact with a liquid [3.11, 3.15].

Electron transfer between the insulating material and the electrolyte is

highly limited by the non-conducting property of the insulating material.

Furthermore, the injection of carriers in the insulating material is not a

simple process and it requires a high activation energy. Conduction

through the insulating film is possible in case a sufficiently high electrical

field is applied over the film so that electrical breakdown occurs. Other

possible conduction mechanisms through insulating film, such as cathodic

current, are explained in detail in [3.9].

The double layer effect at an insulating material / electrolyte interface can

in fact be mainly considered as a capacitive effect. For many insulating

materials, such as SiO2, Si3N4, and Al2O3, their surface shows a high

density of A-OH like sites on their surface (i.e. Si-OH). These sites act as

proton donors or acceptors, and they interact mainly with the H+ ions from

the electrolyte. Consequently, the value of the double layer capacitance,

which depends on the charge density at the interface, will depend on the

pH of the electrolyte. This effect has already been used for pH sensing, as

explained in [3.16, 3.17].

The insulating material is represented electrically as a capacitor Cis. Its

value is retrieved by considering it as a parallel plate capacitor whose

dimensions correspond to the electrodes’ area A and to the material

thickness t.

Cis εo εrAt---⋅ ⋅= (3.9)



3 February 2003, 16:39

The capacitance of the insulating material is in series with the double layercapacitance and the total electrode capacitance Celectrode is such that:

Most often, the double layer capacitance Cdl can be neglected whencompared to the insulating film capacitance Cis. Combining Eq. (3.7) andEq. (3.9) yields Cis larger than Cdl rewritten as:

and thus:

Taking, for instance, εr equal to 7.3 (valid for silicon nitride films) and Co

equal to 20.10-2 F.m-2, then Eq. (3.11) is verified. The capacitance of theinsulating film is dominant over the double layer capacitance. The doublelayer behaviour is observable only when the material thickness isdecreased (Eq. (3.12) gives t < 0.32 nm). This proves that in most cases,the double layer capacitance can be neglected.

3.3.2 Experiments with glass dipstick

A test device was fabricated to electrically characterize the contactless 2-and 4-electrode liquid conductivity measurement method (Fig. 3.4). It

consists of four aluminium electrodes (10 × 0.5mm2) deposited on a 10 x

1j C⋅ electrode 2π f⋅⋅---------------------------------------------- 1

j C⋅ is 2π f⋅⋅------------------------------ 1

j 2π f⋅⋅( )nCdl⋅

--------------------------------------.+= (3.10)

εo εrAt--- Co A⋅>⋅ ⋅

(3.11)

tεo εr⋅

Co

--------------.< (3.12)

Chapter 3


3/2/2003

30 mm2 glass substrate. The electrodes are covered with 540 nm siliconnitride (insulating material). The devices are immersed in salt solutions(KCl) with known conductivities. The impedance versus the frequency ismeasured in the range of 10 Hz to 1 MHz for both the 2- and 4-electrodeconfiguration. Fig. 3.5 illustrates the experimental setup allowing both 2-and 4-electrode measurements. In the 2-electrode setup, the two innerelectrodes are floating (disconnected). The signal vin is supplied by the

lock-in amplifier and is connected to one of the outer electrodes. Thecorresponding current io is measured with a transimpedance amplifier

connected to the other outer electrode. The solution impedance value isderived from the ratio vin/io. In the 4-electrode setup, we did not apply a

current steering between the outer electrodes (as in a classical setup) toavoid the effect of current leakage through the capacitance of the leads. Asin the 2-electrode setup, the signal vin is imposed and the corresponding

current io is measured. The two inner electrodes are connected to high

input-impedance buffers (50 MΩ). The outputs of the buffers areconnected to the input (A) and (B) of the lock-in amplifier. The lock-in

(b)

( c)

(a)

10 mm

10

mm

0.5 mm

0.5 mm

0.5 mm

5 mm

Fig. 3.4 Drawings (left) and photograph (right) of the glass 4-electrode dipstick: (a) glass substrate; (b) 600 nm aluminium electrodes; (c) 540nm silicon nitride.



3 February 2003, 16:39

amplifier is used as the differential voltmeter and it measures vo= (A)-(B).

The liquid impedance is retrieved from the realtime division vo/io.

3.3.2.1 Measurements with 2 electrodes

The complete electrical model of the contactless 2-electrode conductivitycell is shown in Fig. 3.6. The double layer impedance has been neglected,as it is assumed that the insulating film capacitance is dominant. Themeasured impedance Zmeas is retrieved from the ratio vin/io.

DEVICE

(sin out)

(A) (B)

lock-in amplifierstandford research SR 830 DSP

(Hi) (Lo)

v =o (A)-(B)

50 M 50 M

1 k1 k

10 k10 kLF 356

1pF

100 k

LF 356

vinMultimeter

Keithley 177

io

Fig. 3.5 Experimental setup for contactless 2- and 4-electrodemeasurements with glass dipstick.

Zmeas2

j 2π f Cis⋅ ⋅ ⋅------------------------------

Rliq

1 j Rliq Cliq 2π f⋅ ⋅ ⋅ ⋅+--------------------------------------------------------+= (3.13)

Chapter 3


3/2/2003

Eq. 3.13 shows that the impedance Zmeas is not only the liquid resistanceRliq. Below a low cut-off frequency flow, the insulating materialcapacitance Cis dominates the measured impedance. The cut-off frequencyflow is equal to:

At frequencies higher than a high cut-off frequency fhigh, it is the liquidcapacitance Cliq which is measured. The high cut-off frequency fhigh isequal to:

+

_

+iovin

a ab

cCliq

Rliq

Cis Cis

Fig. 3.6 Electrical model of the capacitively coupled 2-electrodedetector and measurement setup: (a) metal electrodes; (b) insulating material;(c) conductive liquid; Cis, Cliq, and Rliq are, respectively, the capacitance ofthe insulating film, the capacitance of the liquid and the resistance of theliquid.

flow1

2 π Rliq Cliq 0.5 Cis⋅+( )⋅ ⋅ ⋅---------------------------------------------------------------------.= (3.14)

fhigh1

2 π Rliq Cliq⋅ ⋅ ⋅--------------------------------------.= (3.15)



3 February 2003, 16:39

Only in an intermediate frequency band, the measured impedance is theliquid resistance Rliq. The width of that frequency band depends on theratio fhigh/flow, which is equal to:

The larger Cis/Cliq, the wider the frequency band in which the liquidresistance can be measured.

In the intermediate frequency band (flow < f < fhigh), the measuredimpedance is equal to:

The accuracy of the liquid resistance measurements also depends on theratio Cis/Cliq. The larger Cis/Cliq, the more accurate the measurement ofthe liquid resistance Rliq.

The measurements done with the 2-electrode setup demonstrate theunfavourable influence on the detection performance of the insulating-film capacitance Cis, and of the capacitive coupling Cliq through thesolution (Fig. 3.7). At low frequencies (< 100 Hz), the impedance isdominated by Cis (a value of 600 pF is measured). At high frequencies (>100 kHz), Cliq dominates the measured impedance (the measured value is2 pF). The values of Cis and Cliq restrict the measurement frequency rangeto a small area between 10 Hz and 100 kHz.

The measurements are compared with theoretical values according to Eq.3.13. The insulating film capacitance Cis is retrieved according to Eq. 3.9

(thickness t of 540 nm, area A of 10×0.5 mm2, relative dielectric constantεr of the silicon nitride equal to 7.3). The cell constant is retrieved

fhigh

flow

----------Cliq 0.5 C⋅ is+

Cliq

----------------------------------- 1 0.5Cis

Cliq

---------× .+= = (3.16)

Zint Rliq 1 2Cliq

Cis

---------×+ .×= (3.17)

Chapter 3


3/2/2003

according to the equations given in Chapter 3.3.1 and is equal to 268 m-1

(w = 0.5 mm, l = 7 mm, and L = 10 mm). The liquid capacitance Cliq andthe liquid resistance Rliq are calculated according to Eq. 3.2. Theconductivity of the liquids was measured with a conductivity meter

(Horiba) and was equal to 2 and 16 µS cm-1. The relative dielectricconstant of KCl solution is equal to 80 (at 20ºC). The spacing between theouter electrodes is equal to 6 mm. As shown in Fig. 3.7, the simulatedvalues are in good agreement with the measured values.

3.3.2.2 Measurements with 4 electrodes

A 4-electrode measurement method reduces the errors resulting frompolarization effects at relatively low frequencies (double layer impedance,capacitive coupling). It is a widely used technique for measurements inbiological systems [3.18,3.19,3.20,3.21]. The electrical model of thecontactless 4-electrode detection cell and the measurement setup areshown in Fig. 3.8. The voltage difference between the two inner electrodesis measured with a high-input impedance differential amplifier and thus,no current flows through the capacitors Cis2. Therefore, the voltage dropwhich is measured is the voltage drop over the resistance Rliq2. The

1,0E+04

1,0E+05

1,0E+06

1,0E+07

1,0E+08

10 100 1000 10000 100000 1000000

Frequency [Hz]

Mea

sure

dim

ped

ance

[]

108

106

104

100 10k 1M

2 S/cm

S/cm

Fig. 3.7 Measured impedance versus the frequency with acapacitively coupled 2-electrode setup. The conductivity of the KCl solutions is 2and 16 µS cm-1. The dashed lines correspond to the theoretical calculations.



3 February 2003, 16:39

current io flowing between the outer electrodes is measured; this is alsothe current flowing in the resistor Rliq2. The value of the resistor Rliq2 isretrieved from the ratio vo/io and is not dependent on the insulating filmcapacitance. This is the advantage of using 4 electrodes instead of 2 forcontactless detection. However, the influence of the capacitive behaviourof the liquid cannot be avoided.

Measurements were done with the glass dipstick. The 4-electrodedetection significantly extends the measurement frequency range as canbe seen in Fig. 3.9. Due to the reduced influence of Cis, the detection isextended to the frequency range of 1 kHz to 100 kHz. The contactless 4-electrode probe shows a linear output signal over two decades of KClconcentrations (Fig. 3.10).

3.3.3 Application to capillary electrophoresis

In standard capillary electrophoresis separations, the conductivity of thecarrier electrolyte determines the baseline of the output signal, whereslight changes of this signal have to be detected when a plug with a

+

_

+iovin

a ab

cCliq2

Rliq2

Cis2 Cis2

+ _

vo

Cliq1 Cliq1

Rliq1Rliq1

a a

Cis1Cis1

Fig. 3.8 Electrical model of the capacitively coupled 4-electrodedetector and measurement setup: (a) metal electrodes; (b) insulating material;(c) conductive liquid; Cis, Cliq, and Rliq are, respectively, the capacitance of theinsulating film, the capacitance of the liquid and the resistance of the liquid.

Chapter 3


3/2/2003

different conductivity passes through the detection window. Acapacitively coupled 2-electrode configuration is suitable for suchconditions. However, this system has a major limitation. There is alwaysan optimal measurement frequency which is linked to the conductivity ofthe carrier electrolyte. Supposing that the buffer has a conductivity of 2µS/cm, the previous measurements (Chapter 3.3.2.1) showed that the 2-electrode detector operates optimally at a measurement frequency of 5kHz. Changing the conductivity of the carrier electrolyte requiresadaptation of the measurement frequency in order to achieve optimalsensitivity, linearity, and dynamic range.

The measurement frequency need not to be adjusted when the 4-electrodesetup is used. Thus, it is possible to use various types of conductivebuffers without the need to adjust the measurement parameters.

3.3.4 downscaling of the detector

From the previously described theory and measurements, it is possible topredict the detection performance of a miniaturized detection cell.Considering the 2-electrode setup, its performance in terms or accuracy

105

103

Frequency [Hz]

1k 100k

Mea

sure

dim

ped

ance

[]

10k

0.02mS/cm

0.15mS/cm

1.3mS/cm

Fig. 3.9 Measured impedance versus the frequency with acapacitively coupled 4-electrode setup.



3 February 2003, 16:39

and frequency range relies on the ratio Cis/Cliq as was explained inChapter 3.3.2.1.

The detection cell shown in Fig. 3.4 is considered to be downscaled to thetypical dimensions of a microchannel: the detection cell has the sameaspect ratio, but the cell area is equal to 100 µm × 100 µm instead of 10mm × 10 mm. In the following calculations, all the parameters related to

the miniaturized detection cell will be denominated with a suffix m (e.g.

Cliqm, Cis

m). All the parameters related to the macroscale detection cell

(Fig. 3.4) will be denominated with a suffix M (e.g. CliqM, Cis

M)

Because the aspect ratio does not change, km and k’m in Eq. (3.3) and Eq.

(3.4) remain identical to kM and k’M. However, the length of the electrodeis 100 times smaller. Therefore, Eq. (3.5) yields:

100

1000

10000

100000

100 1000 10000 100000

105

103

1k 100k

Electrolyte resistivity [ cm

Mea

sure

dre

sist

ance

of

the

elec

tro

lyte

[

104

Fig. 3.10 Measured resistance versus the electrolyte resistivity forKCl solutions with a capacitively coupled 4-electrode setup (conductivityranging from 14 µS cm-1 to 2.7 mS cm-1).

Km 100 K

M× ,= (3.18)

Chapter 3


3/2/2003

and thus, according to Eq. (3.2):

According to Eq. (3.1), the liquid capacitance CliqM is 100 times lower:

The thickness of the insulating film cannot be downscaled to a size 100times smaller (because of technological limitations) and thus a 540 nm

thick film is considered. The electrode area Am has been downscaled to a

size 10000 smaller, which decreases the insulating film capacitance Cism

with the same ratio (according to Eq. (3.9)).

Eq. (3.20) and (3.21) yield:

The ratio Cis/Cliq for the miniaturized cell is 100 times lower, whichdecreases the measurement frequency band and the measurementaccuracy. This decrease will seriously reduce the 2-electrode detectionperformance, and thus it is necessary to use a 4-electrode setup.

Rliqm 100 Rliq

M.×= (3.19)

Cliqm 1

100--------- Cliq

M× .= (3.20)

Cism 1

10000--------------- Cis

M×= (3.21)

Cism

Cliqm

------------ 1100---------

CisM

CliqM

-------------× .= (3.22)



3 February 2003, 16:39

In the next section the design aspects for the integration of the contactless4-electrode detector in a miniaturized CE channel will be discussed.

3.4 Design of the sensor

In the previous chapters, we explained why a contactless 4-electrode setupis preferred for conductivity detection in capillary electrophoresismicrochannels. In this chapter, we consider the integration of the detectorin the microchannel. The detector was designed in order to reach thehighest detection performance.

3.4.1 Layout of the CE microchip

For on-chip CE applications, the conductivity detector must be integratedin a microchannel of 70 µm wide and 20 µm deep (Fig. 3.11). The channelis 6 cm long in order to provide enough separation length. At the detectorlocation, the channel is widened to 170 µm in order to place the fourelectrodes for conductivity detection.

Fig. 3.11 CE microchip of 2 cm long and 1 cm wide. The channel is6 cm long, 70 µm wide and 20 µm high. Inlet and outlet reservoirs are 2.3 mmdiameter.

Chapter 3


3/2/2003

3.4.2 Dimensions of the electrodes

The chosen geometry is shown in Fig. 3.12. The spatial resolutionrequirement imposes a maximal electrode length. With our injectionsystem, the peak width is expected to be at least 1 mm long, therefore a100 µm long electrode guarantees a peak width to electrode length ratio of10. Fig. 3.12 shows that there is not a lot of available space in thechannel’s width for the four electrodes. The choice of the dimensions is,therefore, the result of trade-offs:

1) The larger the area of the electrodes, the better because it increases thevalue of the resulting capacitance Cis.

As will be explained in the following paragraphs, the small size of theelectrodes reduces the value of the capacitive coupling Cis to about 10 pF.This has a direct effect on the inner electrode differential-voltagemeasurement (see Chapter 4). Therefore, we preferred to extend the areaof the inner electrodes (33 µm wide) and, consequently, to reduce the areaof the outer electrodes (20 µm wide).

2) The spacing between the inner electrodes (24 µm) must be as large aspossible, in order to increase the cell constant and, therefore, the value ofthe measured resistor Rliq2.

3) A minimal spacing was kept between all electrodes in order to preventthat problems occuring during the fabrication steps (i.e. misalignment)would not cause overlapping of the electrodes. For that reason, the spacingbetween an outer and an inner electrode was kept equal to at least 10 µm.The same spacing is kept between the outer electrodes and the channelside in case the channel would be misaligned on the electrodes.

3.4.3 Properties of the insulating film

When a voltage is applied for separation, the potential in the liquiddecreases linearly along the channel, and also over the insulatedelectrodes. The metal electrodes are referenced to a potential (oftenground level) which is constant along the whole electrode length. The



3 February 2003, 16:39

consequence is that the insulting film over the metal electrode experiencesa potential difference. This is illustrated the Fig. 3.13. For an electrode of100 µm long and a separation field of typically 500 V/cm, the insulatingfilm has to withstand a voltage difference ∆V of ±2.5 Volts. A minimalbreakdown strength is therefore required for the insulating film.

For an electrode capacitance Cis of at least 10 pF and an electrode area Aof 33 µm × 100 µm, the permittivity-to-thickness ratio (εr/t) of theinsulating film should be larger than 0.34/nm.

Possible candidate film materials having chemical compatibility aresilicon nitride SixNy, titanium dioxide TiO2, tantalum pentoxide Ta2O5

[3.22], and silicon carbide SiC [3.23]. The breakdown voltage of thesematerials found in literature is 900, 24, 136, and 220 kV/mm, respectively.The minimal film thickness for prevention of breakdown (for safetyreasons, 5 V instead of 2.5 V is considered) is given in Table2-1, together

10

Separation

2010

33

24

1001

70

all dimensions in m

Fig. 3.12 Geometrical configuration of the four electrodes in the170 µm wide separation channel.

εr

t----

Cis

εo A⋅-------------> 0.34 nm

1–⋅= (3.23)

Chapter 3


3/2/2003

with the permittivity and the permittivity-to-thickness ratio at the given

thickness.

Top view of the detector

!"V

#!"V

Po

ten

tial

inth

ech

ann

el

100m

Velectrode

distance in the channel

Fig. 3.13 Illustration of the voltage drop (DC) over the electrodescaused by the separation voltage applied in the CE channel. The potential dropis linear and the electrode potential is Velectrode.. At both electrode ends, theinsulating film experiences the maximum potential difference of +/- ∆V.

Table 2-1 Data of insulating film materials

Film Material

Electrical breakdown thickness (nm) Permittivity

Permittivity-to-thickness ratio (nm-1)

SixNy 6 7.3 1.2

TiO2 200 40-80 0.2-0.4

Ta2O5 40 25 0.6

SiC 23 9 0.4



3 February 2003, 16:39

3.4.4 Choice of insulating material

Titanium dioxide (TiO2) films can be obtained in two different ways. Thefirst one consists of oxidizing titanium films at a temperature between 300°C and 500 °C. However, the resulting material acts as a semiconductormaterial [3.24]. The second method is chemical vapour deposition (CVD).With this technique, insulating films with a high dielectric permittivity (40to 80) can be obtained [3.25].

Tantalum pentoxide (Ta2O5) is an attractive material because of its highdielectric constant (25) and breakdown voltage (136 kV/mm). Films asthin as 40 nm can be obtained by oxidation of a 20 nm thick tantalum filmat 500 °C in O2 ambient [3.22]. The temperature used for oxidation isabove the limit that aluminium can withstand without being damaged (450ºC), so another metal which can withstand this temperature (e.g. platinum)has to be used.

Silicon nitride (SixNy) film deposition is carried out by plasma enhancedchemical vapour deposition (PECVD). The dielectric permittivity ofsilicon nitride is 7.3 and therefore very thin layers (21 nm) are required toobtain a sufficiently high capacitance.

Silicon carbide (SiC) films are also fabricated by PECVD. The dielectricpermittivity of SiC is higher than for SixNy, which means that with athicker film (30 nm), a sufficient capacitance can be realized.

TiO2 requires a thickness of at least 200 nm to prevent electricalbreakdown. The fabrication of a 19 nm thick PECVD silicon nitride filmwith good electrical properties is difficult (because the deposition rate istoo fast). With a thicker film, the insulator capacitance is too low for ourapplication. Thanks to the lower deposition speed compared to siliconnitride (nitride: 120 nm/min.; carbide 70 nm/min.), SiC can be produceddown to a thickness of 30 nm still having the required electrical properties.With this thickness, similar insulating film capacitance values areobtained as with 200 nm TiO2 having a permittivity of 80. Ta2O5 is anattractive material too, but it cannot be used in combination withaluminium (aluminium layers hardly withstand temperatures higher than

Chapter 3


3/2/2003

400°C). Due to the above given considerations we have chosen to use

aluminium electrodes with SiC as the dielectric material.

3.4.5 Experiments on SiC structures.

Glass dipstick devices with 100 nm silicon carbide were processed. These

devices were used to measure the dielectric permittivity, εr, of the SiC

film. As shown in Fig. 3.14, the measured impedance corresponds to the

insulating film capacitance at frequencies below 1 kHz. In this frequency

area, for the dielectric permittivity a value of 9.4 is calculated from the

measured impedance. Plate capacitors with SiC as the insulating material

were processed in order to measure the breakdown voltage value. Layers

of 27 nm were successfully deposited. These layers are able to withstand

up to 2.5 Volts.

100 10k 1M

Frequency [Hz]

Measured impedance [ ]

104

105

106

107

Fig. 3.14 Measured impedance versus the frequency with acapacitively coupled 2-electrode setup (100 nm SiC as the insulating film). AKCl solution having a conductivity of 20 µS cm-1 was used for the experiment.



3 February 2003, 16:39

3.5 Technology

In this chapter, the fabrication steps of a CE microchip that includes acontactless 4-electrode conductivity detector are summarized. Thetechnology described here was developed at the DIMES-ECTM by AxelBerthold, Wim van der Vlist, and Lina Sarro. A detailed explanation(masks material, reagents, equipment, process parameters, and others) ofthe complete process is found in [3.26]. The complete process includes 9masks and about 100 steps.

The microchip was fabricated out of two glass wafers. Because of the highbreakdown voltage of glass, a glass-to-glass bonded structure is verysuitable for CE applications. The first wafer (channel wafer) contains thechannel and the inlet- and outlet-reservoir holes. The second wafer(electrode wafer) contains the electrodes for conductivity detection.Themain concern with the electrode wafer is that the surface of the wafer mustremain perfectly planar after fabrication. For that purpose, the electrodesare “buried” in the glass. The planarization is essential in view of the finalstep, which is the bonding of the two wafers.

3.5.1 Fabrication of the channel wafer

The channel has a length of 6 cm, a depth of 20 µm and a width of 70 µmat the separation part. At the detection part, the channel widens to 170 µmin order to provide adequate space for the positioning of the fourelectrodes.

Wet-etching steps are involved in the processing of the channel wafer. Theprofile of the channel after wet etching is highly dependent on thecomposition of the glass. The glass does not only contain pure silicondioxide, but also other components (such as, for instance, aluminiumoxide) that etch away of various rates. The choice of the right glassmaterial according to its composition is, therefore, of primary importance.

Corning (#7740), Schott (#8330), and Borofloat® (Borosolicate glass)wafers are possible candidates because of their low content of aluminium

Chapter 3


3/2/2003

oxide. Finally, Borofloat® wafers are preferred because they show ahomogenous composition over the whole wafer.

The reservoirs holes and the channel are created according to thefollowing sequence.

1) A 600 nm thick LPCVD (low-pressure chemical vapour deposition)amorphous silicon layer is deposited on both sides of the glass wafer(3.15a). On the top, the layer is patterned to define the channel (3.15b).

2) Both sides of the wafer are covered with a stack of, sequentially, 400nm of amorphous silicon (LPCVD layer) and 500 nm of silicon carbide(PECVD - plasma enhanced chemical vapour deposition - layer) (3.15c).This stack of layers is able to withstand an etching time of up to 10 hoursin a mixture of phosphoric acid and hydrofluoric acid. The stack of layersplus the 600 nm-thick amorphous silicon layer are patterned to define thereservoir holes (3.15d). The patterning is done on both sides in order toallow a quicker etching of the reservoir holes.

3) The reservoir holes are etched in a mixture of phosphoric acid (H3PO4,70%) and hydrofluoric acid (HF, 5%) at 70 °C (3.15e). The stack of layersdescribed in the previous step is removed so that the channel mask is nowopen for etching (3.15f).

4) The channel is etched with the solution of phosphoric and hydrofluoricacids (same composition as for the reservoir hole etching) (3.15g). The600 nm thick amorphous silicon layer is stripped (3.15g). The channelwafer is finished.

3.5.2 Fabrication of the electrode wafer

The dimensions of the electrodes have been chosen according to theconsiderations given in Chapter 3.4.2. The two outer electrodes are 100µm × 25 µm and the two inner electrodes 100 µm × 35 µm.

In the electrode wafer, a two-step trench (600 nm each) is etched byreactive ion etching (3.16b). In the lower trench the metal interconnects(aluminium sputtered at 20 °C) are deposited (Fig. 3.16d). The uppertrench is used to cover the metal and the bottom of the separation channel



3 February 2003, 16:39

a

b

c

d

e

f

g

h

amorphous siliconsilicon carbide

Fig. 3.15 Fabrication sequence of thechannel wafer.

Chapter 3


3/2/2003

by a 600 nm PECVD silicon nitride film (Fig. 3.16f). The silicon nitride isisotropically etched in an Inductive Coupled Plasma (ICP) etcher. For thecontactless conductivity detection, a very thin dielectric film is required.At the top of the metal electrodes, the silicon nitride is removed andreplaced by a 30 nm thick silicon carbide film (PECVD layer) (Fig.3.16h). The processing of the electrode wafer is finished and it results in aplanar surface that enables leakage-free bonding.

3.5.3 Bonding procedure

For the glass-to-glass anodic bonding, a 160 nm thick PECVD siliconnitride film is deposited onto the “channel wafer” [3.27].

Prior to the bonding, the wafers are cleaned in a 100% HNO3 solution for10 minutes. The cleaning of the wafers is essential to achieve a successfulbonding.

After cleaning, the wafers are manually aligned with an Electronic VisionEV420 and, afterwards, are placed in the bonder (electronic VisionEV501). the wafers are pre-heated for two hours at 400°C and bonded(also at 400°C) at 1000V for 1 hour. The bonding process leads to a sealedelectrophoresis channel that includes four isolated electrodes for liquid-conductivity detection. The channel walls are entirely covered with siliconnitride, which guarantees a uniform electroosmotic flow. A photograph ofthe fabricated CE microchip is shown in Fig. 3.17.

3.6 Conclusion

We have proved the benefits of the 4-electrode configuration forcontactless conductivity measurements in CE microchannels. The 4-electrode configuration allows sensitive and accurate sensing in a widefrequency range, which is not the case with a 2-electrode configuration.

The design of the detector aimed at getting the largest capacitive couplingbetween the metal electrodes and the liquid, but also to prevent the



3 February 2003, 16:39

a

b

c

d

e

f

g

h

aluminium

silicon nitride

silicon carbideFig. 3.16 Fabrication sequence of theelectrode wafer.

Chapter 3


3/2/2003

breakdown of the insulating film. After investigation of differentinsulating materials, we decided to use silicon carbide for our detector.

3.7 References

3.1 Blake-Coleman B.C., “Approaches to physical measurements inbiotechnology”, Academic Press, London, 1993.

3.2 Everaerts F.M. et al., “Phenomena that occur when conductometricdetection is applied”, J. of Chromatography, 1974, 91, 809-818.

3.3 Gas B. et al., “High-frequency contactless conductivity detection inisotachophoresis”, J. of Chromatography, 1980, 192, 253-257.

3.4 Kaniansky D. et al, “Contactless conductivity detection in capillaryzone electrophoresis”, J. of Chromatography, 1999, 844, 349-359.

3.5 Zemann A. et al., “Contactless conductivity detection for capillaryelectrophoresis”, Analytical Chemistry, 1998, 70, 563-567.

3.6 Fracassi da Silva J.A., “An oscillometric detector for capillaryeletcrophoresis”, Analytical Chemistry, 1998, 70, 4339-4343.

600 nm Si Nx y

600 nm Al

30 nm SiC

inlet outlet

channel

conductivitydetector

channel wafer

electrode wafer

160 nmSi N

x y

2 cm

1cm

inlet reservoir outlet reservoir

Fig. 3.17 Photograph of the CE microchip with the contactless 4-electrode conductivity detector (insert).



3 February 2003, 16:39

3.7 Simonyi K., “Foundations of electrical engineering”, PergamonPress, Oxford, 1963.

3.8 Jacobs P. et al., “Design optimisation of a planar electrolyticconductivity sensors”, Medical & Biological Engineering &Computing, 1995, 33, 802-810.

3.9 Bagotzky V.S., “Fundamentals of electrochemistry”, Plenum Press,London 1993.

3.10 Mc Adams E.T. et al., “The linear and non-linear electricalproperties of the electrode-electrolyte interface”, Biosensors &Bioelectronics, 1995, 10, 67-74.

3.11 Bockris J.O’M., “Comprehensive treatise of electrochemistry,Volume 1: The double layer”, Plenum Press, New York, 1980.

3.12 Panzram E. et al., “A capacitance and infrared study of the electricaldouble layer structure at single crystal gold electrodes inacetonitrile”, Phys. Chem., 1995, 99, 827-837.

3.13 Janek R.P. et al., “Impedance spectroscopy of self-assembledmonolayers on Au(111): evidence for complex double layerstructure in aqueous NaClO4 at the potential of zero charge”, J.Phys. Chem. B, 1997, 101, 8550-8558.

3.14 Macdonald J.R., “Impedance spectroscopy - emphasizing solidmaterials and systems” John Wiley & sons, New York, 1987.

3.15 Morrison S.R., “Electrochemistry at semiconductor and oxidizedmetal electrodes”, Plenum Press, New York, 1980.

3.16 Bergveld P., “Development of an ion-sensitive solid state device forneurophysiological measurements”, IEEE Transaction onBiomedical Engineering, 1970, 17, 70-71.

3.17 Bousse L. et al., “Operation of chemically sensitive field-effectsensors as a function of the insulator-electrolyte interface”, IEEETransaction on Electron Device, 1983, 30, 1263-1270.

3.18 Moron Z. et al., “The possibility of employing a calculable four-electrode conductance cell to substitute for the secondary standardsof electrolyte”, IEEE Trans. on Instr. and Meas., 1997, 46 (6), 1268-1273.


3/2/2003

3.19 Steinschaden A. et al, “Miniaturised thin film conductometricbiosensors with high dynamic range and high sensitivity”, Sensorsand Actuators B, 1997, 44, 365-369.

3.20 Kordas N. et al., “A CMOS-compatible monolithic conductivitysensor with integrated electrodes” Sensors and Actuators A, 1994,43, 31-37.

3.21 Volanschi A. et al., “Design of a miniature electrolyte conductivityprobe using ISFETs in a four point configuration” Sensors andActuators B, 1994, 18-19, 404-407.

3.22 Olthuis W. et al., “Planar interdigitated electrolyte conductivitysensors on insulating substrate covered with Ta2O5”, Sensors andActuators B, 1997, 43, 211-216.

3.23 Harris G. (Ed.), “Properties of silicon carbide”, INSPEC, IEE, UK,1995, 231-273.

3.24 Choi Y.K. et al., “Thin Titanium dioxide film electrodes prepared bythermal oxidation”, J. Electrochem. Soc, 1992, 139 (7), 1803-1807.

3.25 Campbell S.A. et al., “Titanium dioxide (TiO2)-based gateinsulator”, IBM J. Res. Dev., 1992, 43 (3), 383-392.

3.26 Berthold A., “Low-temperature wafer-to-wafer bonding formicrochemical systems”, Deltech Uitgevers, Delft, 2001.

3.27 Berthold A. et al., “Glass-to-glass anodic bonding with standard ICtechnology thin films as intermediate layers”, Sensors and ActuatorsA, 2000, 82, 224-228.


3/2/2003

4Electronic interface for on-chip conductivity measurements

4.1 Introduction

Various readout electronic setups for 4-electrode impedancemeasurements have been presented in the litterature. In [4.1], it is shownthat measurements can be done either in a current-controlled mode or in avoltage-controlled mode. In the current-controlled mode, the currentflowing in the detection cell is controlled and the differential voltagebetween the inner electrodes is measured. In the voltage-controlled mode,the differential voltage is controlled and the corresponding detection-cellcurrent is measured. In [4.2], an electrometer with driven shield ispresented. It allows the measurements of low amplitude voltages from ahigh impedance source.

Chapter 4


3/2/2003

All the reported setups concern large scale detection cells. In this chapter,we are showing how the miniaturization of the detection cell influencesthe detection performance. As a matter of fact, the two main features ofthe 4-electrode measurement setup have to be reconsidered:

1) The control of a constant AC current flow between the outerelectrodes

2) The differential voltage measurement between the innerelectrodes with a high input impedance and a high common-moderejection ratio (CMRR) differential amplifier.

The small dimensions of the detector results in a low value of thecapacitive coupling between the metal electrodes and the liquid (pF rangefor Cis1 and Cis2 in Fig. 4.1), which has a determining influence on the

previously mentioned requirements.

It is shown that the control of a constant and known amplitude currentbetween the outer electrodes is not possible when connecting a standardcurrent source. A solution is measuring the current instead of regulating it.For that, a fixed-amplitude voltage source is connected at one outerelectrode and the resulting current amplitude is measured at the otherouter electrode. Another solution, derived from the previous one, consistsof controlling the amplitude of the voltage source in order to regulate thecurrent amplitude. This is done by implementing a stable regulation loop,and then the complete system can be considered as a current source. In thischapter, we consider the design of the regulation loop, its stability, theinfluence of parasitic capacitances, the DC biasing, and its frequencyresponse.

The low value of the capacitive coupling has an effect on the differentialvoltage measurement. Current leakages in parasitic capacitors anddegradation of the common-mode rejection ratio are the mainconsequences. In this chapter, the use of a specific bootstrappingtechnique is introduced as a solution.


Electronic interface for on-chip conductivity measurements

3 February 2003, 18:28

4.2 Electrical model of the detector

The electrical modelling of the contactless 4-electrode detector wasdiscussed in the previous chapter, but will be reconsidered in this chapter.Fig. 4.1 shows the top view and the cross section of the miniaturizedsensors.

The electrodes are 100 µm long and 33 µm wide (inner electrodes) or 20µm wide (outer electrodes). The metal electrodes are covered with a 30nm thick silicon carbide layer. The silicon carbide having a dielectricpermittivity of 9.4, so the corresponding capacitances Cis1 and Cis2 are

equal to 5.5 pF (A=100.10-6×20.10-6, t=30.10-9, Eq. (3.9)) and 9.2 pF

(A=100.10-6×33.10-6, t=30.10-9, Eq. (3.9)), respectively. The liquidresistances (Rliq1 and Rliq2) are rather difficult to calculate. Nevertheless,the range of values for the liquid resistance can be estimated according tothe sequence described in Chapter 3.3.1.1. Considering the couplingbetween the two outer electrodes, Eq. (3.5) yields a value of the cell

constant that is equal to 2.08 m-1 (w=20.10-6, L=100.10-6, and l=110.10-6).MES/His solutions are standard carrier electrolytes in capillaryelectrophoresis. A 20 mM MES/His (pH 6) solution has a conductivity

equal to 317 µS.cm-1. The value of the liquid resistance between the outerelectrodes (2×Rliq1+Rliq2) is retrieved according to Eq. (3.2) and is equalto 655 kΩ. It is then acceptable to say that values ranging between 100 kΩand 10 MΩ can be expected. The capacitive couplings Cliq1 and Cliq2 arelinked to Rliq1 and Rliq2 according to Eq. (3.1).

The capacitive coupling through the bottom glass wafer (Cg1 and Cg2) isconsidered negligible when compared to the capacitive coupling throughthe liquid. This is true, saying the cell constant is the same for bothcouplings and the dielectric permittivity of the glass (εr = 4) is lower thanthe dielectric permittivity of the liquid (εr = 80).

The metal electrodes have been designed slightly larger than the siliconcarbide layer. This was done to ensure proper coverage of the metalelectrodes, independently of any fabrication mask misalignment. Thus thespacing between an inner and an outer metal electrode is not equal to 10

Chapter 4


3/2/2003

µm, but only to 2 µm (Fig. 4.1). The capacitive coupling (Cg3) is

calculated considering a planar capacitor with 2 µm of glass as thedielectric material (εr = 4). The electrode area is defined by the metal-

electrode thickness (600 nm) and length (100 µm). According to Eq. (3.9),the capacitive coupling has a value equal to 1 fF. The influence of such alow capacitance is thus negligible.

The following considerations will demonstrate how the insulating layercapacitance limits the detection. The influence of Cliq1 and Cliq2 will not

be taken into account because it occurs at relatively high frequencies.

b

all dimensions in m

10

0

10 10 2420 33

170

Cliq1

Rliq1c

d

60

0n

m2

0m

600 nm

a

e

Rliq2Rliq1

Cliq1Cliq2

Cis2 Cis2 Cis1Cis1

Cg1 Cg1Cg2

Cg3

Fig. 4.1 Electrical model of the miniaturized contactless 4-electrode conductivity detector. (a) metal electrode, (b) 30 nm silicon carbide,(c) liquid, (d) glass, and (e) 600 nm silicon nitride. Note: the aspect ratio ismodified for better rendering.



3 February 2003, 18:28

4.3 Control of the current

4.3.1 Direct current readout

In a 4-electrode setup, an AC current with a known value is steeredbetween the outer electrodes. It is done by connecting a current sourcebetween the outer electrodes. Connecting an AC current source is notpossible in our case because the detector has a rather high inputimpedance Zinput and most of the current would flow in the parasiticcapacitance Cp1 (Fig. 4.2).

The input impedance Zinput corresponds to ½.Cis1 (supposing Rliq1 andRliq2 are negligible, which is true in a large frequency band). The cablethat connects the current source to the detector has a distributed capacitorbetween the conducting wire and the grounded shield (Cp1 in Fig. 4.2).The value of that capacitor is linked to the cable length: a 1 m long cablehas a distributed capacitance of about 100 pF. By using short connectionsfor the leads, one can expect to reduce the parasitics to about 10 pf.However, Cp1 remains larger than ½.Cis1 and therefore it introduces asystematic error in the current steering.

A solution is connecting a voltage source of fixed amplitude at the firstouter electrode and measuring the resulting current at the second outerelectrode with a transimpedance amplifier (Fig. 4.2). The voltage vin

imposed on the first outer electrode is not affected by the capacitance ofthe cable. The current io, resulting from the applied voltage is measuredwith a transimpedance amplifier which has a low input impedance. Theoutput voltage vcur of the transimpedance amplifier is proportional to thecurrent flowing through the device (vcur = R × io). The high outputimpedance of the detector in combination with the input capacitance of the

Zinput 21

j Cis1 2π f⋅⋅ ⋅--------------------------------- 2 Rliq1 Rliq2+⋅+⋅= (4.1)

Chapter 4


3/2/2003

operational amplifier Cp2 may be a source of instabilities for thetransimpedance amplifier. Oscillations are prevented by adding acapacitor C (1 pF) in parallel with the feedback resistor R [4.5]. The globalconfiguration is insensitive to parasitic capacitors in the frequency rangewe want to work in (100 Hz - 1MHz).

4.3.2 Regulation of the current

As in the direct readout mode, the detector is connected to a voltagesource at one electrode, and to a low-ohmic current readout at the otherelectrode. The current flowing between the outer electrodes can beregulated by making use of a so-called regulation loop or negative-feedback amplifier [4.1]. In fact, the regulated system can be considered asan AC current source.

4.3.2.1 Gain and stability of negative-feedback amplifiers

The general structure of a negative feedback amplifier is shown in Fig.4.3. The open-loop amplifier has a gain A (xo = A.xi). The fraction of theoutput that feeds back to the input is β (xf = β.xo). The feedback signal xf

is subtracted from the source signal xs, to produce the signal xi (xi = xs -

R

C

R +2 Rliq2 liq1%

Cis1 Cis1

Cp1 Cp2vin

vo

io

detector

Fig. 4.2 Electronic setup for direct current readout using atransimpedance amplifier. (Note: the capacitors Cliq1 and Cliq2 are notrepresented).



3 February 2003, 18:28

xf). The gain of the feedback amplifier Af (also called the closed-loop gain)is obtained by combining the above-mentioned equations:

The quantity Aβ is called the loop gain. In many circuits it is rather large(Aβ >> 1). When so, Eq. (4.2) shows that the open-loop gain Af is equal tothe quantity (1/β). This is an interesting result because it shows that thegain is not dependent on the gain of the open-loop amplifier (which ishigh, but not constant), but only on the feedback network (which consistsof passive components having a relatively stable value). A stable high-gain amplifier is then realized. This property will be used to regulate thevalue of the current in the detector.

The gain of the open-loop amplifier A and the feedback gain β are usuallya function of frequency f and then the closed-loop gain Af(jf) is rewrittenas:

where j is the imaginary unit.

Af

xo

xi

----- A1 Aβ+----------------.= = (4.2)

A

&

open loopamplifier

feedback transfer

xf

xox s x i

Fig. 4.3 General structure of a feedback amplifier.

Af jf( ) A jf( )1 A jf( ) β jf( )⋅+--------------------------------------.= (4.3)

Chapter 4


3/2/2003

Considering the frequency f180 at which the loop phase-angle becomes -180°, then A(jf180).β(jf180) will be a negative real number and:

with A(jf180).β(jf180)the modulus of the loop gain.

The condition for a stable feedback is:A(jf180).β(jf180)< 1.

If the modulus of the loop gain at f180 is greater than or equal to 1, then thefeedback is unstable. It results either in oscillations or clipping of thesignal. More detailed information can be found in [4.4].

Bode plots are very useful to determine whether or not the condition onthe loop gain is verified. By plotting the loop gain and its phase shift, onecan immediately tell whether the feedback amplifier is stable. In Fig. 4.4,the gain and the phase of the loop gain are plotted versus the frequency.The curve (3) leads to a stable feedback amplifier sinceA(jf180).β(jf180)< 1. This is not the case for the curve (1) and (2) andhere the amplifier will oscillate.

4.3.2.2 Current regulation

The schematic of the current regulator is given in Fig. 4.5. An intuitiveway to analyse this circuit is as follows:

The detector’s outer electrodes are connected between the output (vo) andthe inverting input (vinv) of the opamp (operational amplifier). Thepotential vinv is considered to be equal to zero. Therefore, the current ioflowing through the resistor R is equal to the ratio vi/R. The same currentis flowing through the detector because the opamp has a very high inputimpedance. The value of the detector current is thus fixed by the ratio vi/R,and not by the detector impedance Z (as it is the case with direct currentreadout, Chapter 4.3.1). In this analysis, we supposed that the potential

Af jf180( )A jf180( )

1 A jf180( ) β jf180( )⋅–------------------------------------------------------,= (4.4)



3 February 2003, 18:28

vinv was equal to zero, which in a real situation is only true in a limited

bandwidth. With the following calculations, we will demonstrate that thecurrent regulation is limited by the opamp and the low value of Cis1.

-180

-90

-270

-360

0

0

Gai

nA

&P

has

eo

fA

&

dB

f180

12

3

Fig. 4.4 Gain andphase of the loop gain versusfrequency. (3) corresponds to astable feedback. (1) and (2)correspond to an unstablefeedback.

R

R +2 Rliq2 liq1%

Cis1 Cis1

vivo

io

detector

vinv

vnoninv

Fig. 4.5 schematics of the feedback amplifier used for regulatingthe current flowing into the device.

Chapter 4


3/2/2003

The potential vinv at the inverting input of the opamp (operational

amplifier) is equal to:

with

where Z is the impedance modelling the detector (between the outerelectrodes), Rliq the total liquid resistance (Rliq = R1iq1 + 2×R1iq2), Cis the

total insulating layer capacitance (Cis = Cis1 / 2), and j the imaginary

operator.

The output voltage vo is defined as:

with A the gain of the operational amplifier and vnoninv the non-inverting

input of the operational amplifier. The operational amplifier gain isconsidered to follow a first-order model, and A(jf) is equal to:

vinv

R vo⋅ Z vi⋅+

Z R+-------------------------------,= (4.5)

Z1

j Cis 2π f⋅⋅ ⋅------------------------------ Rliq,+= (4.6)

vo A vnoninv vinv–( )× A vinv× ,–= = (4.7)

A jf( )Ao

1 j fAo

fu

------⋅ ⋅+

----------------------------,= (4.8)



3 February 2003, 18:28

with Ao the open-loop gain, fu the unity gain frequency. The correspondingvalues are given in every operational amplifier’s data sheet.

Combining Eq. (4.5) and (4.7), we obtain the output voltage vo:

The regulated current is calculated as:

Combining Eq. (4.9) and (4.10), we can rewritte the current io:

One can identify the elements of the feedback amplifier by comparing Eq.(4.2) and (4.11). The corresponding block diagram is given in Fig. 4.6.

voZ– A⋅

Z 1 A+( ) R⋅+----------------------------------- vi.⋅= (4.9)

io

vo vinv–

Z-------------------

vo

vo

A-----+

Z---------------- 1 A+

A Z⋅------------- vo.⋅= = = (4.10)

io1 A+

Z 1 A+( ) R⋅+-----------------------------------– vi⋅

1 A+Z

-------------

11 A+

Z------------- R⋅+

------------------------------– vi.⋅= = (4.11)

R

-vi io

1+AZ

Fig. 4.6 Block diagram of the feedback amplifiers for currentregulation.

Chapter 4


3/2/2003

Looking at Eq. (4.11), if (1+A).R >> Z, then io = vi / R. The condition forregulating the current flowing in the detector is, therefore, to have a highloop gain:

The loop gain is calculated by combining Eq. (4.6), (4.8), and (4.12).

The loop gain versus frequency (Eq. (4.13)) is plotted in Fig. 4.7. TheOPA 604 from Burr-Brown was chosen as the operational amplifierbecause of its large unity gain frequency. It has an open-loop gain Ao of

106, and a unity gain frequency fu of 20 MHz. The liquid resistance Rliq

was taken equal to 100 kΩ, 1 MΩ, and 10 MΩ. The opamp was consideredas a first-order system, which is only true at frequencies below the unitygain frequency fu. Therefore, we will only consider as relevant the curve atfrequencies below 1 MHz (even though the loop gain was plotted forfrequencies up to 100 MHz).

There are two corner frequencies below 1 MHz: f1 and f2. The first cornerfrequency f1 is equal to the ratio fu/Ao. The second corner frequency f2 isequal to the ratio 1/(2π.Rliq.Cis). The gain in the bandwidth is equal to2π.fu.R.Cis (considering that Ao +1 >> 1). A gain higher than one isobtained by using a high unity gain frequency fu and a large value of theresistor R. The bandwidth of the loop gain (and therefore the currentregulation) is extended by using a high open-loop gain Ao. Furthermore, itis clear that the low value of the capacitive coupling Cis (2.75 pF) induceslimitations in the achievable regulation because it reduces the gain in the

1 A+( ) RZ---⋅ 1.» (4.12)

1 A+( ) RZ---⋅

1 Ao+( ) 1 jAo

1 Ao+( ) fu⋅----------------------------- f⋅ ⋅+

j R Cis 2π f⋅ ⋅ ⋅ ⋅ ⋅ ⋅

1 jAo

fu------ f⋅ ⋅+

1 j Rliq Cis 2π f⋅ ⋅ ⋅ ⋅+( )⋅-----------------------------------------------------------------------------------------------------------------------------= (4.13)



3 February 2003, 18:28

bandwidth. The regulated value of the current io (Eq. (4.11)) is also plottedin Fig. 4.7. The input potential vi was taken equal to 1 Volt. It shows thatwhenever the loop gain is greater than unity, then the current is properlyregulated (to a value of 1 µA in that example).

4.3.2.3 Influence of parasitic capacitance on the currentregulation

The current regulation as previously explained did not take into accountthe influence of the parasitics on the loop gain. The parasitics that areconsidered here are the common-mode Ccm (at the non-inverting input)

1 1 0 0 1 0 0 0 0 1 .1 06

1 .1 08

1 .1 07

1 .51 07

2 .1 07

3 .1 07

5 .1 07

7 .1 07

1 .1 06

1 1 0 0 1 0 0 0 0 1 .1 06

1 .1 08

0 .1

0 .5

1

5

1 0

5 050

10

5

1

0.5

1 100 10k 1M

Rliq = 100 k

Rliq = 1 M

Rliq = 10 M

loop gain

frequency [Hz]

f1

f2

Rliq = 100 k

frequency [Hz]

1 100 10k 1M

1

0.7

0.5

0.3

0.2

0.15

0.1

current [ A]io

Rliq = 1 M

Rliq = 10 M

Fig. 4.7 Loop gain (top) and regulated current (bottom) of thefeedback amplifiers shown in Fig. 4.6.

Chapter 4


3/2/2003

and differential-mode Cdm capacitors of the operational amplifier. Thesetwo capacitors are placed in parallel and they form a parasitic capacitanceCpar (Fig. 4.8).

In Fig. 4.8 (top left), the schematics taking into account the parasiticcapacitor Cpar is given. Norton/Thevenin transformations make theschematics equivalent to the one shown in Fig. 4.5 (current regulation

R

RliqCis

vi

vo

detector

Cpar

R

RliqCis

vi vo

detector

CparR

RRliqCis

vi %vo

detectorCpar

11+j.2 .f.R.C( par

Nortonequivalent

Théveninequivalent

11+j.2 .RC .f( par-vi

A+1Z io

R1+j.2 .RC .f( par

Fig. 4.8 (top left) Schematics of the current regulator taking intoaccount the parasitic capacitor (common-mode and differential-modecapacitors of the opamp). (top right) Equivalent schematics after Norton/Thevenin transformation. (bottom) Equivalent block diagram of the feedbacknetwork.



3 February 2003, 18:28

without the parasitic capacitor). The corresponding block diagram of thefeedback amplifier is also shown in Fig. 4.8. It is analogous to the blockdiagram shown in Fig. 4.6, except that the feedback network βcorresponds to the resistor R placed in parallel with the capacitor Cpar.Furthermore, the input potential is not only vi, but vi/(1+j.2π.R.Cpar.f).

The structure of the block diagram is slightly different than that in Fig.4.6, nevertheless, the condition for the current regulation remainsidentical: the loop gain must be greater than unity. When so, the regulatedcurrent is equal to:

The loop gain Aβ is expressed according to Eq. (4.13), and by substitutingR for R/(1+j.R.Cpar.ω)).

Compared to Eq. (4.13), there is an extra corner frequency f3 that is equal

to (R.Cpar.2π)-1.

The data sheet of the OPA 604 indicates a typical common-mode anddifferential-mode capacitor of values equal to 8 and 10 pF, respectively.This makes Cpar equal to 18 pF. The loop gain was calculated, taking intoaccount the parasitic capacitor, and is shown in Fig. 4.9. The loop gain

io1β---

vi

1 j R Cpar ω⋅ ⋅ ⋅+-------------------------------------------⋅

1 j R Cpar ω⋅ ⋅ ⋅+

R-------------------------------------------

vi

1 j R Cpar ω⋅ ⋅ ⋅+-------------------------------------------⋅= =

ivi

R----.= (4.14)

1 Ao+( ) 1 jAo

1 Ao+( ) fu⋅----------------------------- 2π f⋅ ⋅ ⋅+

j R Cis 2π f⋅ ⋅ ⋅ ⋅ ⋅ ⋅

1 jAo

fu------ 2π f⋅ ⋅ ⋅+

1 j Rliq Cis 2π f⋅ ⋅ ⋅ ⋅+( ) 1 j R Cpar 2π f⋅ ⋅ ⋅ ⋅+( )⋅ ⋅-------------------------------------------------------------------------------------------------------------------------------------------------------------------------- (4.15)

Chapter 4


3/2/2003

drops drastically at high frequencies (f > f2 or f3), which reduces the

bandwidth in which the loop gain is greater than unity. Consequently, theregulation is operational in a reduced bandwidth (Fig. 4.9). The risk ofoscillations is also increased by the presence of the third corner frequency:According to Eq. (4.15), the phase at frequencies higher than f2 and f3 is

closed to -180°. The opamp in that frequency range is not an ideal first-order transfer, as it was supposed, and may introduce a phase shift of morethan -90° (see corresponding data sheet). Therefore, the phase of the open-loop gain is likely to be equal to -180° for frequencies in the range of f2and f3. The gain in that frequency range also approaches 1, and in order to

avoid oscillations it must be lower than one (Chapter 4.3.2.1). Simulationswere done with the Microsim simulation package that takes into accountthe real model of the opamp (it is more than a simple first-order transfer).The OPA 604 met the requirement for stability, but other amplifiers likethe AD 820 (from Analog Device) did not.

4.3.2.4 DC biasing

The DC biasing is a problem when considering the integratorconfiguration. The output potential of the opamp is given in Eq. (4.9). Atlow frequencies, the device impedance Z is much larger than the quantity(1+A).R and the closed-loop gain is therefore equal to the opamp’s gain A.Any noise voltage signal can then saturate the opamp, which would reachthe supply level (±15 Volts). This would seriously damage the insulatinglayer. It is possible, by making use of a feedback amplifier structure, toregulate the DC value of the potential vo to the ground level. The

schematic of the feedback amplifier and its corresponding block diagramare shown in Fig. 4.10.

The voltage vi is a function of vin, vf, R6 and R7 and is equal to:

vi

R7

R6 2 R7⋅+-------------------------- vin vf+( )× .= (4.16)



3 February 2003, 18:28

The value of R7 is taken much larger than R6, thus (4.16) becomes:

The gain Ac is equal to the ratio -vo/vi, which is defined according to Eq.(4.9).

1 1 0 0 1 0 0 0 0 1 .1 06

1 .1 08

1 .1 01 2

1 .1 01 0

1 .1 08

1 .1 06

1 1 0 0 1 0 0 0 0 1 .1 06

1 .1 08

1 .1 06

0 .0 0 0 1

0 .0 1

1

1 100 10k 1M

loop gain

frequency [Hz]

1M1 100 10k

1

0.01

0.001

1

0.01

0.0001

1 2f f

3f

Rliq = 100 k

Rliq = 1 M

Rliq = 10 M

current [ A]io

Rliq = 100 k

Rliq = 1 M

Rliq = 10 M

frequency [Hz]

Fig. 4.9 Loop gain (top) and regulated current (bottom) of thefeedback amplifiers shown in Fig. 4.6, taking into account the influence of theparasitic capacitor.

vi12--- vin vf+( )× .= (4.17)

Chapter 4


3/2/2003

The output potential vo is equal to:

+

-

Cis1

R +2×Rliq2 liq1

R

vo

detector

vin

Cis1

+

-

+

-R6

R5

R1

R2

R5

R6

R7

C5

Feedback

network &

open-loopamplifier A

c

vi

vf

io

&

-vin0.5×A

vo

&

vin -vovi

vf

0.5

0.5 Ac c

+

Fig. 4.10 Schematic of the DC regulation loop (top). Equivalentblock diagram of the feedback amplifier (bottom).

vo

0.5 Ac⋅1 0.5 Ac β⋅ ⋅+-----------------------------------– vin .⋅= (4.18)



3 February 2003, 18:28

From Eq. (4.18), we can write that:

Supposing that the condition in Eq. (4.19) is true at low frequencies, andthe transfer β is very large, then vo is equal to zero. This is actually therequirement for regulating vo to the ground potential.

Given the condition in Eq. (4.20) is true in the measurement frequencyrange (higher than 100 Hz), then vo = -0.5.Ac.vin. By definition, Ac is equalto the ratio (vo/vi), thus vi = -0.5×vin. Therefore, the current in the device isregulated as described in Chapter 4.3.2.2, with -0.5×vin as the referencepotential.

The feedback network in Fig. 4.10 combines the two previousrequirements:

The ratio R1.(R1+R2) makes the loop gain 0.5×Ac×β lower than unity forfrequencies higher than 100 Hz. Below the corner frequency f =

(2.π.R5.C5)-1, the loop gain becomes infinite (theoretically) for DCregulation.

The open-loop gain Ac was simulated according to Eq. (4.9). Theinfluence of the parasitic capacitance on the transfer function was taken

v00.5 Ac β⋅ ⋅ 1»

lim1β---– vin,⋅= (4.19)

v00.5 Ac β 1«⋅ ⋅

lim 0.5– Ac vin.⋅ ⋅= (4.20)

βvf

vo

-----R1

R1 R2+------------------

1 j 2 R5 C5 2π f⋅ ⋅ ⋅ ⋅ ⋅+

j R5 C5 2π f⋅ ⋅ ⋅ ⋅---------------------------------------------------------.⋅= = (4.21)

Chapter 4


3/2/2003

into account, as described in Chapter 4.3.2.3. The parameters of the OPA

604 were used. The liquid resistance Rliq was taken only equal to 1 MΩ(in order to simplify the resulting graph). The feedback network was

calculated according to Eq. (4.21), with R1 = 100Ω, R2 = 100 kΩ, R5 = 100

kΩ, and C5 = 50 nF. It clearly shows that the loop gain 0.5×Ac×β is larger

than unity at low frequencies, and the opposite above 100 Hz.

4.3.2.5 Simulation of the current-regulation loop

The complete regulation schematics (Fig. 4.10) was simulated

numerically and with Microsim. The loop gain (Aβ) was calculated as

0 .0 1 1 1 0 0 1 0 0 0 0 1 .1 06

1 .1 08

1 .1 08

0 .0 0 0 0 1

0 .0 1

1 0

1 0 0 0 0

1 .1 07

Rliq = 1 M

frequency [Hz]

1 100 10k 1M10

-2

104

101

1

&

0.5×A

0.5×A &×

10-5

10-8

107

amplitude

c

c

Fig. 4.11 Simulation of the feedback amplifier: (0.5× Ac) is theopen-loop gain, (β) is the feedback network, (0.5×A×β) is the loop gain. Theliquid resistance Rliq is equal to 100 kΩ. The influence of the opamp’sparasitic capacitance Cpar, as described in Chapter 4.3.2.3 was taken intoaccount.



3 February 2003, 18:28

detailed in Chapter 4.3.2.4. According to the block diagram given in Fig.4.10, the output potential vo is calculated as a function of vin.

Ac (equal to -vo/vi) and β were defined in Eq. (4.9) and (4.21),respectively. The values of the components (resistors, capacitors, opamps)were the same as in Chapter 4.3.2.2.

Fig. 4.12 shows that the numerical model matches correctly the microsimmodel. The current is properly regulated in a bandwidth extending

vo

0.5 Ac⋅1 0.5 Ac β⋅ ⋅+-----------------------------------– vin⋅= (4.22)

0 .1 1 0 1 0 0 0 1 0 0 0 0 0 . 1 .1 07

1 .1 09

1 .1 01 2

1 .1 01 1

1 .1 01 0

1 .1 09

1 .1 08

1 .1 07

1 .1 06

frequency [Hz]10 1k 100k 10M

10-7

10-9

Microsimsimulation

Numericalsimulation

10-7

10-9

10-11

45

10-11

current [A]io

current [A]io frequency [Hz]10 1k 100k 10M

Rliq = 100 k

Rliq = 1 M

Rliq = 10 M

Fig. 4.12 Microsim (top) and numerical (bottom) simulation ofthe regulated current flowing through the detector, between the outerelectrodes. The liquid resistance Rliq was taken equal to 100 kΩ, 1 MΩ,and 10 MΩ. (top to bottom curve, respectively).

Chapter 4


3/2/2003

between 45 Hz and 10 kHz. Supposing that the liquid resistance is nothigher than 1 MΩ, then the bandwidth is extended to 100 kHz. There is aslight mismatch between the resonance frequencies of both simulations.Nevertheless, it shows the influence (at medium frequencies) of theparasitics on the current regulation (Chapter 4.3.2.3).

4.4 Differential voltage measurement

There are two major issues for the differential voltage measurement(between the two inner electrodes). First, the differential amplifier musthave a high input impedance so that it does not load the detector (thus, thecurrent which is measured between the outer electrodes is the currentflowing into the measured liquid resistance Rliq2 (Fig. 4.1)). Second, thevalue of the common-mode rejection ratio (CMRR) must remain highenough in order to reject the high common-mode voltage, and that in awide frequency range.

4.4.1 The input impedance of the differential amplifier

4.4.1.1 Common-mode input impedance of the differentialamplifier

Two elements contribute to the common-mode input impedance of thedifferential amplifier (Fig. 4.13). First, the differential amplifier has inputcommon-mode capacitors Ccm which are in the pF range. Second, high-value resistors Rbias (50 ΜΩ) are necessary in order to bias the inputs ofthe differential amplifier.

The output impedance of the detector at the inner electrodes, determinedby the liquid resistances and the electrode capacitances, is in the sameorder as the common-mode input impedance of the differential amplifier.The differential amplifier loads the detector and thus the electrodes supply



3 February 2003, 18:28

a significant current (i1 and i2). The difference between the measuredcurrent io and the current iRliq2 induces a systematic error in themeasurement.

A possible solution to decrease the current difference (io-iRliq2) consists ofbootstrapping the common-mode input impedance of the differentialamplifier with half the input voltage vin. Because of the resultingsymmetry of the setup, the currents i1 and i2 flowing through the two innerelectrodes are equal and in opposite phase. The liquid resistance R2 is nowin parallel with a high input impedance (Fig. 4.14). The error in themeasurement is reduced (io ≅ iRliq2).

4.4.1.2 Differential input impedance

Differential amplifiers have a differential input impedance that includes a

resistor Rd (≅ 1012 Ω) and a capacitor Cd in the pF range (Fig. 4.13). Theresistor Rd has a negligible influence because of its high value. However,the capacitor is an extra error source in the measurement (same influenceas Ccm/2). Inserting a buffer stage (one buffer for each inner electrode)ahead of the differential amplifier is a solution. As for a differential

+

vin

Rliq2

Cis2

Cis2

Cd

Rliq1

Rliq1

Cis1

Cis1

Rbias

Rbias

Ccm

CcmRd

v+

v-

vo

detector

io

differentialamplifier

iRliq2i1

i2

Fig. 4.13 Electrical model of the detector connected to thedifferential amplifier. Ccm and Rbias form the common-mode input impedanceof the amplifier. Cd is the differential input capacitor and Rd is thedifferential input of the differential amplifier.

Chapter 4


3/2/2003

amplifier, the buffers have a Ccm. However, a good isolation between the

two buffers suppresses the presence of the differential input capacitor. Thedifferential voltage vo is linked to the voltage drop over the resistor Rliq2

through the voltage divider formed by Ccm, Rbias, and Cis2.

4.4.1.3 Electronic setup for bootstrapping

The bias resistors are bootstrapped by connecting them to half the inputvoltage. In Fig. 4.15, the resistors R6, which form a voltage divider, supply

the common-mode voltage ½vin. The bias resistors Rbias, because of their

high value (50 MΩ), do not disturb the voltage division.

The input common-mode capacitance of the buffers is bootstrapped bybootstrapping the power supply of the buffers (Fig. 4.15). The connectionsto the power supply is done through a summing amplifier. The summingamplifier has an input stage with a high input impedance so that it does notmodify the voltage division.

An instrumentation amplifier is used as the differential amplifier.

+

vin

Rliq2

Cis2

Cis2

Cd

Rliq1

Rliq1

Cis1

Cis1

2 R% bias

Ccm

Ccm/2Rd

v+

v-

vo

detector

io


iRliq2i1

i2

Fig. 4.14 Electrical model of the detector connected to thedifferential amplifier. The bias resistors Rbias and the common-modecapacitors Ccm have been bootstrapped with half the input voltage vin.



3 February 2003, 18:28

4.4.2 common-mode rejection ratio

The output vo of a differential amplifier is not only a function of itsdifferential gain Ad, but also of its common-mode gain Acm.

+15 V

-15 V

10

k1

0k

10

k1

0k

1 F

1 F

INA 111

+

vin

V+

V-

V+

V-

V+

V-

+15 V

-15 V

vo

Rliq1

Rliq1

Rliq2R6

R6

Rbias

Rbias

R3

R4

R4

Rg

Cis1

Cis1

Cis2

Cis2

instrumentationamplifier

buffer stage

detector

0.5 + 7.5 V%vin

0.5 - 7.5 V%vin

summing amplifier

Fig. 4.15 Complete setup for the differential voltage measurement.

vo Ad vd Acm vcm⋅+⋅= (4.23)

Chapter 4


3/2/2003

with vd and vcm the differential and the common-mode voltage,

respectively.

In Fig. 4.13, the differential voltage vd and common-mode voltage vcm are

equal to:

The CMRR of a differential amplifier defines its ability to attenuate the

common-mode voltage while amplifying the differential voltage. It is

expressed in dB and is defined as:

The CMRR is an important criterion for the choice of the differentialamplifier. An instrumentation amplifier is preferred for its high CMRR.

Considering the Eq. (4.23), a proper differential measurement is achieved

when:

vd

Rliq2

Rliq2 2.Rliq1 21

j Cis1 2π f⋅ ⋅ ⋅---------------------------------⋅+ +

---------------------------------------------------------------------------------- vin,⋅= (4.24)

vcm12--- vin.⋅= (4.25)

CMRR 20.Ad

Acm

---------.log= (4.26)

Ad vd⋅ Acm vcm.⋅» (4.27)



3 February 2003, 18:28

Therefore, the CMRR must satisfy the following relation:

The lower the liquid resistance Rliq2 (and thus Rliq1), the lower thedifferential voltage. However, the common-mode voltage does not dependon the liquid resistance, which makes the requirement on the CMRR (Eq.(4.28)) stronger for low liquid resistances (high liquid conductivities).

Taking Cis1 equal to 5.5 pF, Rliq1 and Rliq2 equal to 100 kΩ, an inputvoltage of 1 V, and a frequency of 2 kHz, Eq. (4.24) and (4.25) yield avalue for the differential voltage and the common-mode voltage equal to3.46 mV and 500 mV, respectively. The INA 111 from Burr-Brown has aCMRR of 110 dB up to 200 Hz (differential gain of 10). Above thatfrequency, the CMRR value drops at a rate of 20 dB/dec. Therefore at 2kHz, the CMRR value is equal to 90 dB, which is enough to reject thecommon-mode voltage (20.log (vcm/vd) = 43.2 dB, thus Eq. (4.28) isverified). The validity of Eq. (4.28) can be extended in the frequencyrange 100 Hz - 1 MHz.

4.4.2.1 Degradation of the CMRR due to componentmismatch

When the differential amplifier is connected to the inner electrode,without the bootstrapping technique described in Chapter 4.4.1.3, anymismatch (between the two inner electrode branches) of the biasresistances, common-mode capacitances, and electrode capacitancesdegrades the CMRR. This is a direct consequence of driving thedifferential amplifier with high output impedance sources. The problemwith a similar configuration was already reported in [4.4, pp. 81]. Thedegradation of the CMRR is demonstrated by considering the equivalentblock diagram in Fig. 4.16.

In Fig. 4.16, the differential amplifier is considered as having a differentialgain Ad, a common-mode gain Acm, and thus the output voltage vo is

CMRR 20vcm

vd

--------.log⋅> (4.28)

Chapter 4


3/2/2003

defined according to Eq. (4.23). The differential voltage vd at the input of

the differential voltage is defined as (v+-v-), and the common-mode

voltage vcm is defined as ½ (v++v-).

As mentioned earlier, the input impedance (Ccm and Rbias in Fig. 4.14) of

the differential amplifier is comparable to the output impedance of the

detector (Cis2), and therefore the potential v+ and v- are not equal to the

potentials vA and vB at the liquid resistance Rliq2. Now, v+ and v- are

redefined as:

+

-

vo

A

B

v+

v-

vA

vB

A d A cmRliq2

Fig. 4.16 Equivalent block diagram for the inner electrodedifferential-voltage measurement.

v0 Ad v+

v-–( )⋅ Acm

v+

v-+

2----------------

⋅+= (4.29)

v+ j Rbias

′Cis2

′ 2π f⋅ ⋅ ⋅ ⋅

1 j Rbias′

Cis2′

Ccm′+( ) 2π f⋅ ⋅ ⋅ ⋅+

-------------------------------------------------------------------------------- vA⋅ A vA⋅= = (4.30)

v- j Rbias

″Cis2

″ 2π f⋅ ⋅ ⋅ ⋅

1 j Rbias″

Cis2″

Ccm″+( ) 2π f⋅ ⋅ ⋅ ⋅+

-------------------------------------------------------------------------------- vB⋅ B vB⋅= = (4.31)



3 February 2003, 18:28

In Eq. (4.30) and (4.31), different bias resistances, common-modecapacitances, and insulating-layer capacitances have been used because itis exactly their mismatch that degrades the total common-mode rejectionratio.

By combining Eq. (4.29), (4.30), (4.31), we can express the outputpotential vo as a function of vA and vB and it is equal to:

The total differential gain Adt and common-mode gain Acm

t define the

total common-mode rejection ratio CMRRt and they verify the followingrelation:

By combining Eq. (4.32) and (4.33), we can express the total differential

gain Adt and common-mode gain Acm

t as a function of the differential gainAd and common-mode gain Acm of the differential amplifier and of the

gain A and B. The total common-mode rejection ratio CMRRt is

calculated as the ratio (Adt / Acm

t) (expressed here in normal units and notin dB):

If we consider the common-mode rejection ratio CMRR of the differential

amplifier equal to the ratio Ad / Acm, then Eq. (4.34) yields the CMRRt as afunction of CMRR, A, and B:

The maximal component mismatch is expected to be only in the order of10% and therefore the difference between (A - B) remains relatively smallwhen compared to (A + B). Combined with the fact that the CMRR

vo Ad A v⋅ A B v⋅ B–( )⋅ Acm

A v⋅ A B v⋅ B+

2---------------------------------

.⋅+= (4.32)

vo Adt

vA vB–( )⋅ Acmt vA vB+

2-----------------

.⋅+= (4.33)

Chapter 4


3/2/2003

remains larger than unity in the working frequency range, thenCMRR.(A+B) >> ½ (A-B). Eq. (4.35) is rewritten as:

and

If there is no component mismatch, then A = B, and Eq. (4.37) yields

CMRRt = CMRR. There is no degradation of the CMRR. Supposing A ≠B and looking at the denominator of Eq. (4.37), then we can say that:

CMRRt Ad

t

Acmt

--------- 12---

Ad A B+( )⋅ 12--- Acm A B–( )⋅ ⋅+

Ad A B–( )⋅ 12--- Acm A B+( )⋅ ⋅+

---------------------------------------------------------------------------.⋅= = (4.34)

CMRRt Ad

t

Acmt

--------- 12---

CMRR A B+( )⋅ 12--- A B–( )⋅+

CMRR A B–( )⋅ 12--- A B+( )⋅+

------------------------------------------------------------------------.⋅= = (4.35)

CMRRt Ad

t

Acmt

--------- 12--- CMRR A B+( )⋅

CMRR A B–( )⋅ 12--- A B+( )⋅+

------------------------------------------------------------------------,⋅≅= (4.36)

CMRRt 1

2--- CMRR

CMRRA B–( )A B+( )

------------------⋅ 12---+

--------------------------------------------------.⋅≅ (4.37)



3 February 2003, 18:28

and therefore:

Combining Eq. (4.39) and (4.37), it appears that CMRRt is always lower

than CMRR and that the mismatch between A and B is the source of the

degradation.

The effect of the component mismatch can be neutralized by driving the

common-mode capacitance Ccm and the bias resistance Rbias with the

common-mode voltage. This is done by introducing the buffers

bootstrapped with half the input voltage. It reduces the effect of

component mismatch and the low output impedance of the buffers

guarantees an optimal common-mode rejection ratio for the connected

instrumentation amplifier.

When looking at Fig. 4.15, one may notice that the buffer stage is the same

as the first stage of an instrumentation amplifier [4.4]. Therefore, the

bootstrapping technique could be done directly on the instrumentation

amplifier, without the need for an extra buffer stage. Such a configuration

has the disadvantage to transfer the common-mode voltage (½vin) on the

output, since the power supply is driven with it. The rejection of the

common-mode voltage is in that case cancelled! The use of the buffer

stage is thus necessary.


------------------⋅ 12---+

12---> (4.38)

1


------------------⋅ 12---+

-------------------------------------------------- 2.< (4.39)

Chapter 4


3/2/2003

4.4.2.2 Simulation of the CMRR degradation

The model of the detector is shown in Fig. 4.13, with Cis1 = 5.5 pF, Cis2 =9.2 pF, Ccm = 2.8 pF, Rbias = 50 MΩ. Rliq1 is equal to Rliq2 and was takenequal to 100 kΩ, 1MΩ, and 10 MΩ.

The total common-mode rejection ratio CMRRt was simulated accordingto Eq. (4.35), and a mismatch of 10% in the value of the bias resistors wastaken into account. This means that one of the bias resistors was taken

equal to 50 MΩ, and the other one equal to 45 MΩ. In Fig. 4.17, CMRRt isrepresented as the degraded common-mode rejection ratio. Whencompared to the common-mode rejection ratio of the INA 111, one can seethat the mismatch in the bias resistances reduces the total common-moderejection ratio. At low frequencies, the 10% mismatch induces adegradation of almost 100 dB for the common-mode rejection ratio. Thisis a considerable value. Such large numbers were also reported in [4.4, pp.81]. Very similar curves are observed for the mismatch of the insulatingcapacitance Cis2.

In Fig. 4.17, the minimum required common-mode rejection ratio is alsorepresented (ratio vcm / vd, with vd and vcm defined according to Eq. (4.24)and (4.25), respectively), with Rliq1 as a parameter. By comparing the setof obtained curves with the common-mode rejection ratio of the INA111,one can confirm the condition for a proper differential readout (Eq.(4.28)). However, by comparing them with the degraded common-moderejection ratio, one can see that the condition for a proper differentialreadout only applies above a limit frequency whose value is linked to theliquid resistance Rliq1. Below the limit frequency, the common-modevoltage is not attenuated enough, and the differential voltage is notsufficiently amplified. Consequently, the output voltage of the differentialamplifier produces a signal with a constant amplitude. This was confirmedwith a SPICE (Simulated program with integrated circuit emphasis)simulation of the complete electronic readout as it is shown in Fig. 4.18.The AD820 (from Analog Device) was taken as the operational amplifierfor the buffer stage. The resistances R3 and R4 were taken equal to 10 kΩ.The instrumentation amplifier was an INA 111, for which the gain was setto 10 (set with an external resistance Rg that is equal to 5.56 kΩ, according



3 February 2003, 18:28

to the data sheet). R5 and R6 were both taken equal to 10 kΩ. All othercomponents have the values mentioned previously in that chapter. Theelectronics for the current regulation was not taken into account in thesimulation. Instead the value of the current iRliq1 flowing in the resistorRliq1 was extracted from the SPICE simulation and taken as the value ofthe current that would be ideally measured by a transimpedance amplifieras described in Chapter 4.3.1.

In Fig. 4.18 (left), the measured liquid impedance (vo / iRliq1) versus thefrequency is shown in the case where there is no mismatch betweencomponents. The bias resistors and the input buffers were and were notbootstrapped. At low frequencies, there is an attenuation of the measuredimpedance due to the voltage divider formed by Cis2 and Rbias. Theattenuation does not affect the linearity of the detector response andchanges of liquid conductivity produce proportional changes of the output

R

=100

k

liq1

R

=1

M

liq1

R

=10

M

liq1

frequency [Hz]

10 1M1k 100k

100

80

Co

mm

on

-mo

de

reje

ctio

nra

tio

[dB

]

60

40

20

CMRR

of theIN

A111

degraded CMRR

Fig. 4.17 Simulation of the CMRR of the INA 111, thedegradation of this CMRR due to component mismatch, and the requiredCMRR for a proper differential-voltage readout (with the liquid resistanceR1iq1 as a parameter).

Chapter 4


3/2/2003

signal. It is also noticeable that when the input buffers and the biasresistors are not bootstrapped, the response remains linear. However, thereis an error (visible at low frequencies) due to the fact that the measuredcurrent iRliq1 is not the current flowing through the resistor Rliq2

(Chapter 4.4.1).

In Fig. 4.18 (left), the same simulation was done, but a 10% mismatch inthe bias resistors was taken into account. The degradation of the common-mode rejection ratio as described previously is visible when thebootstrapping technique is not applied: at low frequencies, the outputsignal is constant. When the bias resistors and the input buffers arebootstrapped, the mismatch does not influence the measurement and thesame curves as in Fig. 4.18 (right) are obtained.

4.5 Conclusion

Due to the low value (pF) of the capacitive coupling between the metalelectrodes and the liquid, parasitic capacitors strongly limit theperformance of the readout electronics.

R = 100 kliq1

R = 1 Mliq1

R = 10 Mliq1

frequency [Hz] frequency [Hz]

10 100 10k 1M1k 100k 10 100 10k 1M1k 100k

100 M

1 M

10 k

100

mea

sure

dre

sist

ance

[]

100 M

1 M

10 k

100

no bootstrapping no bootstrapping

bootstrapping bootstrapping

Fig. 4.18 SPICE simulation of the differential readoutmeasurement with the electronic interface as shown in Fig. 4.15. Thesimulations were done with and without the bootstrapping technique. Nomismatch between the components (left). Mismatch of 10% between the biasresistances (right).



3 February 2003, 18:28

A feedback amplifier is used to control the current flowing in the detectioncell. A stable regulation is possible in a frequency band extending between10 Hz and 100 kHz for liquid resistances up to 10 MΩ.

A high input-impedance differential amplifier is necessary for the readoutof the inner-electrode differential voltage. Here again, parasitic capacitorsat the inner electrodes play a negative role because they drive a significantleakage current. In order to meet the high-impedance requirement, aspecific bootstrapping technique is introduced. In addition, this techniqueprevents the severe degradation of the common-mode rejection ratio dueto component mismatch.

4.6 References

4.1 Tamamushi R. et al., “Instrumental study of electrolytic conductancemeasurements using four-electrode cells”, ElectroanalyticalChemistry and Interfacial Electrochemistry, 1974, 50, 277-284.

4.2 Chroboczek J.A. et al., “Four-probe electrometer system forresistivity measurements”, J. Phys. E: Sci. Instrum., 1995, 18, 568-570.

4.3 Sedra A.S., Smith K.C., “Microelectronic circuits”, OxfordUniveristy Press, New York, chapter 8, 1998.

4.4 Franco S., “Design with operational amplifiers and analog integratedcircuits”, McGraw-Hill Book Co, New York, 1988.

4.5 Wang T., Ehrman B., “Compensate transimpedance amplifiersintuitively”, Burr-Brown application bulletin, March 1993.


3/2/2003


3/2/2003

5Full decoupling of detection and separation

5.1 Introduction

In the previous chapters, we motivated our choice for contactless detectioninstead of galvanic detection. For on-column liquid conductivitydetection, there are various reasons to do so: An insulating materialprotects the metal electrodes from electrochemical destruction and/ormodification, which makes the detector more robust. Gas generation at thesurface of the electrodes is prevented. Furthermore, in principle there is noelectrical DC coupling between the separation voltage source and theconductivity-detector electronics.

In this chapter, we show that separation and contactless detection dointeract in two ways. First, the thin insulating film is subject to breakdown

Chapter 5


3/2/2003

when the separation voltage reaches a threshold value, which leads toelectrochemical reactions and gas formation at the electrode surface.Second, because the detector is placed at the end of the separation column,not all the detection current flows between the detection outer electrodes,

but a part of it leaks to the cathode1.

We developed a full-decoupling technique which is based on the control ofthe potential at the cathode. The control consists of the combination of anAC (AC decoupling) and a DC (DC decoupling) signal. With the newsetup, the detector placed 500 µm ahead of the cathode can withstand aseparation field-strength of up to 300 Volts/cm, without leakage ofdetection current.

5.2 DC decoupling

5.2.1 Breakdown of the insulating film

For on-column conductivity detection, the sensing electrodes are placed inthe separation column at a short distance ahead of the grounded outlet(Fig. 5.1).

The thickness of the insulating film has a direct influence on the detectionin terms of sensitivity, accuracy, and detection limit. As discussed inChapter 3, a thin insulating film is required. In our device, this is a siliconcarbide film of a thickness equal to 30 nm which insulates the metalelectrodes from the liquid. A drawback of having a thin insulating film isthat the breakdown effect is more likely to occur. In this chapter weexplain how the insulating film can be subject to breakdown under theeffect of both the separation voltage and the contact potential at thecathode.

1In this chapter, separations of cations are considered. Therefore the ref-erence electrode for separation is the cathode. In the case anions are sep-arated, the reference electrode would be called the anode.


Full decoupling of detection and separation

3 February 2003, 16:39

5.2.1.1 Influence of the separation voltage

A voltage Vsep1 is applied between the inlet and the outlet forelectrophoretic separation. The potential in the liquid decreases linearlyalong the channel. The potential in the liquid at the detector Vdet is equalto

where d1 and d2 are the distances between the inlet and the detector andbetween the detector and the outlet, respectively (Fig. 5.1). In our 6 cm

Separationvoltage

(DC)Vsep1

Measurementvoltage (AC)vin

Vc2Vdet

Vsep1

Potential in theliquid (Volts)

A B

C

Vc1

d1 d2

Fig. 5.1 Potential distribution along the separation channel. Vc1and Vc2 are the contact potentials at the anode and at the cathode,respectively. Vsep1 is the voltage applied for separation, and Vdet is thepotential in the liquid at the detector.

Vdet

d2

d1 d2+----------------- Vsep1× ,= (5.1)

Chapter 5


3/2/2003

long channel, the detector is placed 500 µm ahead of the outlet. Therefore,applying a potential difference equal to 1800 Volts (equivalent fieldstrength of 300 V/cm) for separation will result in a potential at thedetector equal to 15 Volts (d1 = 59500 µm, d2 = 500 µm).

The metal electrodes underneath the silicon carbide film are held toground and, therefore, the insulating film is subject to a potentialdifference. An insulating film will withstand a limited potential difference(the breakdown voltage), which is determined by its dielectric strengthand by its thickness. According to literature, amorphous silicon carbidefilms have a dielectric strength which is equal to 220 kV/mm. Theelectrical properties of our silicon carbide film were measured inChapter 3. The film has a breakdown voltage that is equal to 92 kV/mm.This means that a 30 nm thick silicon carbide film can withstand amaximum potential difference of 2.8 Volts before breakdown. Applying aseparation voltage higher than 331 Volts will therefore result in thebreakdown of the insulating film (according to Eq. (5.1)).

5.2.1.2 Influence of contact potentials

The cathode and the anode consist of platinum wires that are placed at theinlet and outlet of the separation channel. Denominating Vc1 and Vc2 thecontact potentials which are developed at both electrodes (Fig. 5.1), wecan rewrite Eq. (5.1) as:

The contact potentials are equal to a few hundreds of millivolts. Thecontact potential at the anode does not influence the value of Vdet becauseVsep2 - Vc2 - Vc1 ≈ Vsep2. Eq. (5.2) is rewritten as:

Vdet

d2

d1 d2+----------------- Vsep1 Vc1– Vc2–( )× Vc2.+= (5.2)

Vdet

d2

d1 d2+----------------- Vsep1× Vc2.+≅

(5.3)



3 February 2003, 16:39

Eq. (5.3) shows that the contribution of the separation voltage has adominant contribution on the value Vdet. The contact potential at the

cathode also influences the resulting value of Vdet. A contact potential

equal to a few hundreds of millivolts may cause the potential Vdet to reach

the threshold voltage at which breakdown occurs.

5.2.2 Protection concepts

To prevent breakdown of the insulating film, one must lower or evencancel the potential difference over the insulating film. Three decouplingconfigurations are presented in this paragraph, and their applicability willbe discussed.

detectorVdet

Cis1

Cis2

Cis1

Cis2

+

-0

isolation stage

isolation stage

first stage

io

vo

powersupply

Cpar

Vref

Rpar

Fig. 5.2 setup for floating readout electronics. An isolationstage is used to provide a floating power supply for the first stage. The firststage corresponds to the interface described in Chapter 4. An isolation stageis used to DC decouple the signals before measurement (with a lock-inamplifier in our case).

Chapter 5


3/2/2003

5.2.2.1 Floating readout electronics

A common technique consists in making the power supply of the readoutelectronics floating by means of an isolation stage. Ideally, the referencepotential of the readout electronics Vref follows the potential Vdet and thereis no potential difference over the insulating film. Fig. 5.2 shows the setupwhere isolation stages are used to make the readout electronics floating.

The reference of the floating power supply is coupled to ground through adistributed parasitic capacitor Cpar and leakage resistor Rpar. Thedistributed parasitic capacitor Cpar ranges from pico to nanofarad values,and it can have a negative effect in two ways:

i) It can introduce a large time constant in the floating actuation.When it does, the reference of the power supply does not follow the DCpotential in the liquid instantaneously during switchings of the separationvoltage, and breakdown may occur at that time.

ii) When Cis<Cpar the charge distribution between the twocapacitors results in a significant potential drop over the insulating-film

detectorVdet

Cis1

Cis2

Cis1

Cis2

+

-0

isolation stage

isolation stage

first stage

io

vo

powersupplyVref

Vsep1

R1

R2

Fig. 5.3 Control of the DC level of the readout electronics bymaking use of a resistive divider formed by R1 and R2.



3 February 2003, 16:39

capacitor. The system is not considered to be floating anymore and, again,breakdown may occur.

Another method would consist of using a floating separation-voltagesupply. However, for safety reasons a floating high-voltage supply is notrecommended, and it is necessary that one pole of the high-voltage supplyremains connected to ground.

5.2.2.2 DC biasing of the readout electronics

One can solve the breakdown issue by using a resistive voltage dividerover the separation-voltage supply (Fig. 5.3). The resistors are chosenregarding the dimensions of the channel, so that the resulting potential isequal to Vdet. This potential biases the detector electronics DC level,which cancels the voltage difference over the insulating film. This methodhas some drawbacks: The signals measured with the first-stage electronicsneed to be DC decoupled before being measured, which requires an extraisolation stage. The isolation is likely to introduce noise andnonlinearities. Another drawback is that the power-supply of theelectronics is AC coupled to ground by parasitic capacitors Cpar. Thesecapacitors may lower the performance of the readout electronics asexplained in Chapter 4 (i.e. degradation of the common-mode rejection

detectorVdet

Cis1

Cis2

Cis1

Cis2

first stage

io

vo

Cd

Cd

Cd

Cd

Cd

Cis1

< Cd

Cis2

<

Fig. 5.4 Insertion of coupling capacitor to reduce the voltagedrop over the insulated film.

Chapter 5


3/2/2003

ratio and of the current regulation). For that reason, a grounded readoutelectronics is preferred.

5.2.2.3 Decoupling with external capacitors

The readout electronics can be kept grounded if external couplingcapacitors Cd (DC blocking) are connected in series with the electrodes(Fig. 5.4). These capacitors will be charged until the charge at theelectrode-to-liquid interface is equal to the coupling capacitor charge. Theratio of the capacitances determines the division of the total DC potential.Because of the low insulating-film capacitance Cis (10 pF), one needsexternal coupling capacitors of about 1 pF. This configuration has onemain disadvantage: It reduces the total coupling between the detector andthe readout electronics, which lowers the detection performance andmakes the readout electronics highly sensitive to parasitic capacitors(Chapter 4).

5.2.3 DC decoupling by control on the separation voltage

In the previous paragraphs, we presented various configurations that helpto prevent the breakdown of the insulating film. They all have in commonthat the actuation is done on the readout electronics.

As far as readout is concerned, the parasitics to ground need to becontrolled (Chapter 4). Therefore, it is better to hold the readoutelectronics grounded instead of floating or biased to a fixed potential.

In this chapter, we present a new DC decoupling method that makes itpossible to hold the readout electronics grounded. The decoupling methoddoes not act on the readout electronics, but on the separation voltage. Thereference potential of the separation voltage is brought to a value such thatthe potential in the liquid at the detector is equal to zero. The breakdownof the insulating film is then prevented because it does not experience aDC electrical field.



3 February 2003, 16:39

5.2.3.1 Electronic setup

The electronic setup makes use of two power supplies Vsep1 and Vsep2

(Fig. 5.5). A positive separation voltage Vsep1 is applied at the anode and anegative compensation voltage Vsep2 is applied at the cathode so that thepotential in the liquid at the detector is equal to zero. Such a setup has theadvantage that the readout electronics can remain referenced to groundpotential. Therefore, this compensation method is preferred for ourapplication.

This decoupling technique is also applicable to galvanic detection. In thecase of galvanic detection, the control of the cathodic potential mayprevent electrochemical reactions at the electrodes.

Separationvoltage

(DC)Vsep1

Measurementvoltage (AC)vin

Vc2

Vdet=0

Vsep1

Potential in theliquid (Volts)

A B

C

Vc1

(DC)Vsep2

Vsep2

0 Volt

Fig. 5.5 Electronic setup for prevention of the insulating filmbreakdown. Two voltage supplies are used in order to have the potential inthe liquid at the detector Vdet equal to zero.

Chapter 5


3/2/2003

5.2.3.2 Active control of the DC decoupling voltage

We worked on a new configuration that makes use of a potential probe, ingalvanic contact with the liquid (Fig. 5.6). Gold is used as the material forthe probing electrode because it is chemically inert. 300 nm of gold isevaporated on the top of the aluminium electrodes previously coveredwith a 100 nm titanium layer (adhesion layer). To obtain an accuratereading of the channel potential, we placed the potential probe as close aspossible to the detector’s electrodes (20 µm ahead).

The probe enables us to measure the DC potential in the channel at thedetector area. This information is used to control the amplitude of thevoltage Vsep2. A control loop makes it possible to set the value of Vsep2 sothat the DC voltage over the insulating film is nearly zero, herebylowering the requirements on its isolation properties. The control loop isshown in Fig. 5.7. The probe interface is represented by a high input-impedance amplifier having a gain equal to unity. A high-gain differentialamplifier A is used as a subtractor. The output of the differential amplifiercorresponds to Vsep2. The resistors R1 and R2 symbolize the channelconnection between the cathode, the anode, and the potential probe. Thecontact potential at the potential probe is denominated as Vc3. Theseparation voltage Vsep1 and the contact potential at the referenceelectrode Vc2 are compensated by the regulation loop, thus the regulatedvalue of Vdet does not take into account Vsep1 and Vc2. However, thecontact potential Vc3 at the potential probe does have an influence on Vdet.

potential probedetectionelectrodes

Fig. 5.6 Top view of the contactless four-electrode detector andof the potential probe (galvanic electrode).



3 February 2003, 16:39

It adds an offset in the regulated value of Vdet. This offset can be partlyreduced by setting the reference potential Vref not equal to zero, but to avalue equal to Vc3. The value or the contact potential is not constant andthus the offset cancellation cannot be optimal continuously. Nevertheless,the error on the regulated value of Vdet can be reduced.

5.3 AC decoupling

5.3.1 Leakage of the detection current

Fig. 5.8 shows a drawing of the detector and the cathode, and a suggestedequivalent electrical model. The electrodes are placed 500 µm ahead ofthe cathode and the distance between the two outer electrodes is equal to170 µm. The electronic setup for the contactless four-electrode detectionwas described in Chapter 4. A voltage source vin is applied at one of thetwo outer electrodes, and the resulting current io is measured at the otherouter electrode with a transimpedance amplifier. The voltage differencebetween the inner electrodes vo is measured, and the impedance of the

Vsep1

+

-R1

R2

Vsep2

VrefVc2

Vc3

A

%1

Vdet

Fig. 5.7 Electronic interface for active control of the referencepotential of the separation voltage.

Chapter 5


3/2/2003

liquid (from which conductivity is retrieved) is linked to the ratio (vo/io).

The resistors Rliq1 and Rliq2 represent the resistive couplings through theliquid between the four electrodes. This is the resistance Rliq2 that ismeasured from the ratio (vo/io). The resistive coupling between the

electrodes and the cathode are symbolized by four leakage resistors Rleak,whose value is proportional to the spacing between the electrodes and thecathode. Their respective values certainly differ, but for the sake ofsimplicity they will be considered all identical. Cis1 is the capacitor thatcorresponds to the insulating film (30 nm SiC). Its value is calculated fromthe area A of the electrode, the thickness t and the relative dielectricconstant εr of the insulating film (Eq. (3.9)).

The impedance of the insulating film, Z, is inversely proportional to itscapacitance Cis1 and to the measurement frequency f. At low frequenciesthe impedance of the insulating film increases and becomes larger than theleakage resistance. Consequently, the current generated for conductivitydetection tends to leak towards the reference electrode (through Rleak)instead of flowing between the detection electrodes. The current measuredwith the transimpedance amplifier is, therefore, not the current flowingthrough the resistance Rliq2. Accuracy and sensitivity (and therefore the

detection limit) are consequently lowered.

vin

Rliq1

Cis1

Rleak

17

0m

500 m

vo

io Cis1

Rliq1

Rliq2

vin

Fig. 5.8 Drawing of the detection electrode and of thereference electrode for separation (left). The suggested AC equivalentmodel is shown on the right.



3 February 2003, 16:39

The magnitude of the leakage current is linked to the value of Rleak andCis1. The leakage current is low as long as Rleak remains larger than Cis1.The detector is often placed at the end of the separation channel, close tothe cathode (500 µm in our case), in order to benefit from the completelength of the channel. This will definitely increase the leakage currentbecause Rleak becomes low. A thick insulating film is often chosen inorder to prevent breakdown. This will also increase the leakage currentbecause the insulating layer capacitance decreases.

5.3.2 AC decoupling principle

By connecting an AC voltage source to the reference electrode (for theseparation voltage) with a value equal to half the value of the detectiondriving signal vin, one can significantly reduce the leakage current throughthe cathode. With such a configuration, the cathode draws a currentthrough the two top leakage resistors (see Fig. 5.8), but also supplies thesame current through the two bottom leakage resistors. The leakageresistors can be seen as if in parallel to the resistors Rliq1 and Rliq2. Theleakage resistors have a higher value than Rliq1 and Rliq2 because thedistance between the outer electrodes is smaller than the distance from theelectrodes to the cathode. The induced error is then considerablyminimized.

5.4 Measurement results

A DC voltage source was connected at the reference electrode in order totest the DC decoupling method. An AC voltage source as described inChapter 5.3.2 was added, for AC decoupling, to the DC potential Vsep2.

Fig. 5.10 shows the value of the DC current that flows through theinsulating film while the separation voltage Vsep1 was increased. Above acritical value of Vsep1 (300 Volts for Vsep2 = 0 V), a significant amount ofcurrent was measured, which indicates the breakdown of the insulating

Chapter 5


3/2/2003

film. While applying a DC compensation voltage Vsep2, the critical value

was pushed up to higher values. As shown in Fig. 5.10, a separationpotential of 600 Volts was applicable in combination with a DCcompensation voltage on the reference electrode with a value equal to -1

vin

Rliq1 Rleak

Cis1 0.5 v% invin

Cis1

Rliq1

Rliq2

Rliq1

Cis1

Cis1

Rliq1

Rliq2

1%Rleak

Fig. 5.9 Decoupling principle (right) and equivalent electricalmodel of the AC decoupled detector.

0

20

40

60

80

100

120

140

160

0 200 400 600 800

Separation voltage (V)

DC

cu

rren

t(n

A)

V = 0Vsep2

V = -1Vsep2

Separation voltage V (V)sep1

DC

curr

ent

(nA

)

Fig. 5.10 Measurement of the DC current flowing through theinsulating film while the potential at the reference electrode (for separation)is equal to zero (full line) and -1 Volt (dashed lines).



3 February 2003, 16:39

Volt. Separation potentials of up to 1800 Volts are applicable incombination with a decoupling DC potential of -7 Volts.

The influence of the leakage current through the reference electrode on thedetector response is illustrated in the Fig. 5.11. The measured frequencyresponse (measured impedance versus frequency, see Chapter 6 for moredetails) was obtained while the reference electrode was either floating(ideal response), grounded, or driven with 0.5 vin (this voltage is also

called the common-mode voltage). 2 mM and 20 mM MES/His (pH 6)buffers were used to fill the channel. The reference electrode connected toground drew the leakage current at low frequencies, and a distortion of theresponse was noticed when it was compared to the ideal response. Bydriving the cathode with the common-mode voltage, the measuredresponse followed the ideal response (all separation electrodes floating).

1.00E+00

1.00E+01

1.00E+02

1.00E+03

1.00E+03 1.00E+04 1.00E+05 1.00E+06

1 10 100 10001

10

100

1000

Frequency [kHz]

Mea

sure

dim

ped

ance

[a.u

.]

to ground

floating

0.5 Vin

20 mM

2 mM

Fig. 5.11 Measured frequency response of the detector for variousconfiguration of the reference electrode (floating, grounded, or connected tothe common mode voltage).


3/2/2003

5.5 Conclusion

An electronic setup that allows full decoupling of the contactlessconductivity detector with the separation column was presented. Thedecoupling setup allows one: i) To prevent breakdown of the insulatingfilm and thus to optimize the film properties (a thin film can be used) forimproved detection, and ii) To reduce the AC-current leakage through thecathode allowing placement of the detector close to the channel end. Withthe setup described in this chapter, separations at a field strength of up to300 Volts/cm were possible and no current leakage through the cathodewas detected. In addition, the use of an extra galvanic electrode wasintroduced as a solution for optimal control of the DC voltage at thecathode.


3/2/2003

6Conductivity measurements in CE microchips

6.1 Introduction

A reliable fluidic and electronic setup is a key issue for testing themicrochip and its detector. A dedicated holder that houses the microchipand its associated readout electronics (Chapter 4) was fabricated. Thisholder is compatible with the dispensing rack of an autosampler, whichallows controlled, and thus reproducible, sample injection.

The first series of measurements was aimed at characterising the detectorresponse. The frequency response of the 2- and 4-electrode configurationswere compared, and it was clear that the 4-electrode setup is moreappropriate for detection on the microchip scale. This was confirmed by a

Chapter 6


3/2/2003

second series of measurements, aimed at testing the detector on the basisof separations of inorganic ions (potassium, sodium and lithium).

The detection limit of conductivity detection is relatively poor whencompared to that of other detection systems such as electrochemical andoptical detection. This is partly due to the fact that peaks with a lowamplitude are measured on the basis of a high-amplitude backgroundsignal (carrier electrolyte). A setup for electronic baseline suppression wasdeveloped and is presented in this chapter. This setup, combined with anoptimized injection setup, allowed us to separate and detect six organicacids with concentrations down to 50 µM (each).

6.2 Measurement setup

6.2.1 Housing of the microchip

The microchip is placed in a computer-controlled liquid dispensingsystem operating with a dedicated injection procedure (IBIS TechnologiesBV, Enschede, the Netherlands). The inlet and outlet of the single-channeldevice are connected to 100 µL reservoirs. Platinum wires, serving ashigh-voltage electrodes, are positioned in the reservoirs and connected to ahigh-voltage supply assembled in-house. Injection is performed by meansof a computer-controlled fluid handling program. A drain in the inlet vialenables fast removal of excess sample after injection.

A microchip holder compatible with the dispensing rack of theautosampler was fabricated in-house (Fig. 6.1). The CE microchip wasglued onto a printed circuit board (PCB), which was placed in themicrochip holder. Thin bond wires were used to connect the connectionpads of the microchip to the tracks on the PCB. The low value of theinsulating layer capacitor (pF) requires the readout electronics to bepositioned as close as possible to the detector, in order to minimizeparasitic capacitances. Therefore, the readout electronics circuitry wasimplemented on another PCB and placed in the microchip holder, almostattached to the device. The electrical connections between the electronicsPCB and the microchip PCB were made by 4 mm long pins (no cable wereused).


Conductivity measurements in CE microchips

3 February 2003, 16:39

6.2.2 Readout electronics

The readout electronics implemented on the printed circuit board consistsof the differential amplifier shown in Fig. 3.15 and of the transimpedanceamplifier shown in Fig. 3.2 (direct-current readout mode). The buffer

Dispensing rack

Reservoir holder

inlet vial outlet vial

sample reservoirs

PCB with microchip

Microchipholder

PCB with read-outelectronics

PCB withread-out

electronicsPCB withmicrochip

interconnects

Reservoir holder

Fig. 6.1 Microchip holder compatible with the autosampler’sdispensing rack. The microchip holder houses the microchip and itsassociated readout electronics. A reservoir holder makes a sealed connectionfor dispensing the fluids into the microchip.

Chapter 6


3/2/2003

stage shown in Fig 3.15 has a total gain of 3 (R3 = 10 kΩ, R4 = 10 kΩ, R5 =22.5 MΩ). The instrumentation amplifier (INA111 from Burr BrownCorp.) has a gain equal to 5.2 (an external capacitor of 12 kΩ is connectedaccording to the datasheet). The gain of the transimpedance amplifier isset to 100000, by connecting a feedback resistor R of a value equal to 100kΩ.

The electronics for current control, as described in Chapter 4.3.2.2, isimplemented as a separate element. The connections were made in such away that the current setup is easily switchable from the direct-currentreadout mode to the current-control mode (then, the transimpedanceamplifier is not active anymore).

In the direct current-readout mode, the output of the differential amplifierand that of the transimpedance amplifier are connected to lock-inamplifiers (Model Perkin Elmer 7265) to measure the signal amplitudes(Fig. 6.2). The lock-in amplifier used for reading out the differential-voltage amplitude is also used as the voltage source Vi. In the current-control mode, only the output of the differential amplifier is connected to alock-in amplifier. The lock-in amplifier used for reading out the

Lock-in amplifiersPerkin-Elmer 7265

+

_


transimpedanceamplifier

v

Computer

/

4-e

lect

rod

ed

etec

tor

PCB withelectronics

PCB withmicrochip

i

vo

i o

vo i o

Fig. 6.2 Total electronic setup for 4-electrode conductivitydetection. The configuration shown in here corresponds to the direct-current-readout mode. The electronics for the current-control mode canbe connected as an external element (in that case, the transimpedanceamplifier is not active).



3 February 2003, 16:39

differential-voltage amplitude is then used as the regulating voltage sourcevin (Chapter 4.3.2.4).

The measured signals are sent to a PC through an acquisition card (KPCI3108, Keithley Instruments). The conductivity versus time is displayed bymeans of an interface written and compiled with the software Labview(National Instruments).

6.2.3 Electronics for decoupling detection and separation

The principle for AC and DC decoupling of the separation and detectionwas introduced in Chapter 5.

The electronic setup for combining the DC and AC decoupling of thedetection and separation is shown in Fig. 6.3. The high-voltage supply iscontrolled by computer, with an acquisition card. At the same time, whena voltage is applied for separation, a DC potential (of opposite polarityfrom the separation voltage) is generated by the computer for DCdecoupling. This potential is proportional to the applied separationvoltage, and is equal to about -400 mV for each step of 100 Volts. The DCpotential is summed to half of the detector’s input signal vi for full

-+

-+

10

k

10

k

LF356LF356

10

k

10

k

1 F

high-voltagesupply

vi

Fig. 6.3 Electronic setup for DC and AC decoupling. The DCdecoupling is controlled by computer.

Chapter 6


3/2/2003

decoupling. The output of the summing stage is connected to the referenceelectrode for separation.

When the high-voltage supply is switched from ground level directly tothe separation voltage (up to 1800 Volts), it takes a defined time to reachthe desired value. However, the DC decoupling is done faster because amuch lower value is assigned (5 to 7 Volts). As a consequence, there canbe overcompensation during a short time, which may damage theinsulating film. In order to avoid this, the ramped the separation voltageby steps of 50 Volts, synchronising the separation and the decoupling. Thetotal ramping was done within 500 ms.

6.3 Electronic baseline suppression

Conductivity detection does not have a very low detection limit, comparedto amperometric and optical detection, because small changes of a high-level background signal (baseline) are measured. In standard CEapplications, a chemical suppressor is used in combination withconductivity detection to provide a low measured background signal,thereby enhancing the detection limit. However, the need of a permeablemembrane that is electrochemically reactive with the buffer can be anobstacle to miniaturization. Therefore, we have investigated thepossibility of lowering the level of the background signal electronically.

6.3.1 Suppressed conductivity detection

6.3.1.1 Choice of the carrier electrolyte

Conductivity detection has been extensively used in ion chromatography,but not in CE. An explanation is that conductivity detection combinedwith CE suffers from conflicting requirements. The choice of the carrierelectrolyte is the source of this conflict. To understand the conflict, weconsider the formation of the zones as illustrated in Fig. 6.4 (top). Thezones are formed by substitution of the analyte ions with the buffer co-



3 February 2003, 16:39

ions (the zone remains electrically neutral and the amount of positivecharges is equal to the amount of negative charges). Therefore, theconductivity of the analyte ion must differ from the conductivity of thecarrier electrolyte co-ion in order to get a signal out of the conductivitydetector [6.1],[6.2]. The larger the difference, the larger the amplitude ofthe measured signal (Fig. 6.4 bottom left). Fig. 6.4 (bottom left) alsoshows that large peak amplitudes allow a better adaptation of theelectronics’ dynamic range, thus a lower signal-to-noise ratio (SNR). Onthe other hand, the carrier electrolyte is chosen in such a way that zonedispersion is limited. Zone dispersion (or broadening) is partly a result ofdiffusion, convection, and Joule heating in the separation channel. Zonedispersion is also linked to the difference in velocity (thereforeconductivity) between the analyte ion and the carrier electrolyte co-ion.Zone dispersion occurs when the velocities differ. Usually, The zonedispersion is observed as tailing or fronting peaks [6.3],[6.4] (Fig. 6.4bottom right). Various other peak deformations or the formation ofunwanted system peaks can also be noticed due to mismatches ofvelocities. As a conclusion, finding a suitable carrier electrolyte whenusing conductivity detection involves a trade-off between separationefficiency and detection sensitivity. At the end, the difference inconductivity cannot be too high, and only small changes of a high-levelbackground signal are observed when a zone passes the detector.

6.3.1.2 Chemical suppressor

Conductivity detection combined with chemical suppression improves thedetection limit in the presence of a highly conductive carrier electrolyte[6.5],[6.6]. The setup is shown in Fig. 6.5. Before the suppressor, thecarrier electrolyte is chosen such that it satisfies the requirement foroptimal separation, as explained in Chapter 6.3.1.1: the difference inconductivity between the analyte ions and the buffer co-ion is low. Achemically active membrane is placed before the conductivity detector.This membrane chemically interacts with the carrier electrolyte,exchanging ions with high conductivity for ions with low conductivity

(e.g. Na+ ions for H+ ions). The resulting conductivity of the carrierelectrolyte decreases. Then, it is possible to use a smaller dynamic rangefor the readout electronics because a lower background signal is

Chapter 6


3/2/2003

measured. Using a chemical suppressor, a detection limit of 10 nM wasreported for inorganic ions [6.6].

6.3.2 Electronic suppressor

Chemical suppressors have some limitations. First, only a limited type ofbuffers can be used since they must interact with the suppressor. Second,the integration of the active membrane (often a nafion joint) in themicrochip format has never been reported till date, which is probablylinked to technological limitations. In this section, we describe a methodthat allows us to subtract electronically the value of the baseline beforemeasurements and therefore to extend the usable dynamic range of the

Conductivity

Detector

voltage

supply

analyte 1

Buffer counter-ions and co-ions

analyte 2

outp

ut

signal

req

uir

edd

yn

amic

ran

ge

outp

ut

signal

req

uir

edd

yn

amic

ran

ge

Fig. 6.4 Formation of the zones in a classical capillary-zone-electrophoresis setup (top). Observed changes of conductivity as a function ofthe difference in mobility between the analyte ion and the buffer co-ion.(bottom left) ideal case. Real case taking into account the band broadening(bottom right).



3 February 2003, 16:39

readout electronics. A similar setup has been described for carriersuppression techniques used to measure noise in oscillators [6.7].

6.3.2.1 Hardware setup

When suppressing the baseline, the detector operates in the current-controlled mode. Then, the suppression needs only to be done on onesignal (the differential voltage between the inner electrode).

The baseline suppression technique makes use of a lock-in amplifier (Fig.6.6). The signal proportional to the liquid conductivity is the differentialAC voltage between the two inner electrodes. This signal is conditionedand amplified with the differential amplifier (Chapter 4). Afteramplification, the signal amplitude vo is measured with a lock-in amplifier(Perkin-Elmer 7265) and the value is sent to the computer for on-linedisplay of the electropherogram.

The differential input of the lock-in amplifier is used to suppress thebaseline. On the first input (A), the output signal of the diiferentialamplifier is applied. On the other input (B), a signal of which theamplitude and phase is related to the baseline signal is applied. This signalis self-generated by the lock-in amplifier (through the “sine out” output).When the proper phase and amplitude are set, only the changes ofconductivity due to the peaks are measured. Then, the differential signalcan be measured with a low input range and by taking advantage of the

SuppressorConductivity

Detector

Regenerant

voltage

supply

analytes

highly conductive

buffer-ionsweakly conductive

buffer-ions

Fig. 6.5 Setup for chemical suppression in standard capillaryzone electrophoresis.

Chapter 6


3/2/2003

dynamic range of the lock-in amplifier. It improves the signal-to-noiseratio and thus leads to a lower detection limit.

6.3.2.2 Software setup

The lock-in amplifier is controlled with Labview, and theelectropherogram is acquired according to the following sequence. (1) Atthe beginning of the recording, no baseline suppression is applied. Theinput B of the lock-in amplifier is disconnected. The amplitude and thephase of the output signal of the differential amplifier are measured at theinput A. (2) Shortly before the peaks occur, the measured amplitude andphase are stored. They represent the amplitude and phase of the baselinesignal. (3) A signal with the same amplitude and phase is then generated atthe output “sine wave” and the lock-in amplifier works in the differentialmode (A-B). The baseline suppression now starts. (4) The input range ofthe lock-in amplifier is adjusted to the finer value.

Lock-in amplifierPerkin-Elmer 7265

+

_

Differentialamplifier

voi

Personal computer

A

B sine out (A-B)

Contactless 4-electrodeliquid conductivity detector

Fig. 6.6 Setup for electronic baseline suppression.



3 February 2003, 16:39

6.4 Comparison between the 2- and the 4-electrode setup

The design of the contactless conductivity detector allowed us to comparethe 2- and 4-electrode setup. We compared them on the basis of themeasured frequency response for varying buffer concentrations. Theseparation channel was filled with 20 mM, 10mM, 2 mM MES/His buffer(at pH 6.0), and deionized water. For each buffer concentration, wemeasured the response (measured impedance equal to the ratio of vo/io forthe 4-electrode setup and of the ratio vi/io for the 2-electrode setup, seeFig. 6.2) in the frequency range of 10 Hz - 1 MHz.

6.4.1 2-electrode impedance measurement

The frequency response obtained with the 2-electrode setup (Fig. 6.7)shows similar characteristics for each buffer conductivity, which can bedivided into three frequency bands. At frequencies below 100 Hz, the

10 100 1k 10k 100k

1M

10M

100M

1G

10G

Measu

red

imp

ed

an

ce

[W]

Frequency [Hz]

20 mM

10 mM

2 mM

water

Mea

sure

dim

ped

ance

[]

Frequency [Hz]

Fig. 6.7 Measured impedance versus frequency with a contactless2-electrode setup. The dashed line corresponds to the impedance of an 2.25 pFcapacitor, which is the value theoretically expected.

Chapter 6


3/2/2003

insulating film acting as a capacitor (Cis1, see Fig. 2.8) in series with the

double layer capacitance (Cdl) dominates the measured impedance. The

double-layer capacitance is the only parameter sensitive to changes ofconductivity, resulting in a poor detection sensitivity in that frequencyrange. At frequencies above 100 kHz, stray capacitances Cstray (including

capacitive coupling through the liquid) shortcut the liquid impedance,making the detector insensitive to conductivity changes. At mediumfrequencies (100 Hz - 100 kHz), the measured impedance involves theliquid impedance and partly also the previously mentioned capacitances(Cstray, Cis1, and Cdl). In this frequency band, the detector is sensitive to

changes in conductivity, but the accuracy, linearity, sensitivity, and thedynamic range are determined by the values of Cstray, Cis1, and Cdl.

Concerning the outer electrodes, the theoretical value of the insulating-film capacitor is equal to 5.5 pF, according to Eq. 2.9 (the electrode area is

100 × 20 µm2 and the thickness of the silicon carbide film is 30 nm). At afrequency f of 10 Hz, the measured impedance Zmeas corresponds to twice

the insulating film impedance, thus to:

In Fig. 6.7, the measured impedance at 10 Hz is lower than theoreticallyexpected (212 MΩ instead of 5.78 GΩ).

An explanation is that the surface in contact with the electrode is not onlythe area covered with 30 nm of silicon carbide, but also all the leads. Thelayer of silicon nitride that covers the metal electrode is also not exactly600 nm but a little bit less (about 550 nm). On the one hand, the leads arecovered with 540 nm of silicon nitride, which should limit the coupling tothe liquid. On the other hand, the width of the leads outside the channel isnot only 10 µm, but 52 µm. This increases the coupling to the liquid in thenon-bonded area. The leads for the outer electrodes are each 6500 µm

Zmeas 21

j Cis1 2 π f⋅ ⋅ ⋅ ⋅-------------------------------------⋅ 5.78GΩ.= = (6.1)



3 February 2003, 16:39

long, each. They will contribute to a coupling capacitor Cis1extra of a valueequal to:

With a relative dielectric constant of 7 for the silicon nitride, the extracapacitive coupling is equal to 38 pF. This is a considerable value whencompared the originally expected value (5.5 pF). The total coupling (perelectrode) to the liquid is then equal to 43.5 pF (38 pF + 5.5 pF), and themeasured impedance at 10 Hz is equal to 731 MΩ (Eq. 6.1). This does nottotally explain the low value of the measured impedance at lowfrequencies, however.

6.4.2 4-electrode impedance measurement

As expected from theory, there is a minor influence of Cis1 and Cdl on thefrequency response of the 4-electrode detector (Fig. 6.8). For frequenciesup to 10 kHz, the detector response is linear with a constant and optimalsensitivity, whereas above 10 kHz, stray capacitances affect the detectorresponse in the same way as they affect the 2-electrode setup.

For conductivity detection in capillary electrophoresis, the conductivity ofthe carrier electrolyte fixes the baseline of the output signal. The outputsignal has to change with small variations in conductivity caused byanalyte zones passing the detector. A capacitively coupled 2-electrodeconfiguration is suitable for this application, but, as mentioned before, thedetector response is highly dependent on the measurement frequency,which makes proper adjustment of that frequency a necessity. In addition,changing the conductivity of the carrier electrolyte demands re-adjustmentof the measurement frequency to obtain optimal sensitivity, linearity,accuracy and dynamic range. This adjustment is not required for the 4-electrode configuration, where detection performance (in terms ofsensitivity) is similar for each buffer concentration when performingmeasurements at frequencies below 10 kHz.

Cis1extra 8.85 10 12–⋅ εr6500 10⋅ 6– 52 10⋅ 6–×

550 10 9–×------------------------------------------------------ .⋅ ⋅= (6.2)

Chapter 6


3/2/2003

6.5 Separation of inorganic ions

6.5.1 Reagents

Buffer solutions were prepared daily in MilliQ water (Millipore SAMolsheim, France). 2-Morpholinoethanesulphonic acid (MES) wasobtained from Sigma (St Louis, MO, USA), histidine (His) from Fluka(Buchs, Switzerland). All buffer solutions were filtered and degassed priorto use. Samples were diluted from 1 M stocks prior to analysis. NaCl, KCland LiCl were obtained from J.T. Baker (Deventer, the Netherlands).

The microchip was cleaned once a day (before measuring). First themicrochip was flushed with purified water for 5 minutes. Afterwards, 2 MNaOH was used for reconditioning and cleaning the capillary and theelectrode surface (flushing for 5 minutes). Finally, the microchip wasrinsed with purified water and manually filled with buffer before it wasplaced in the liquid-dispensing system.

100 1k 10k 100k

100k

1M

10M

Measu

red

imp

ed

an

ce

[W]

Frequency [Hz]

20 mM

10 mM

2 mM

water

Mea

sure

dim

ped

ance

[]

Frequency [Hz]

Fig. 6.8 Measured impedance versus frequency with acontactless 4-electrode setup.



3 February 2003, 16:39

6.5.2 Injection procedure

The injection procedure was typically as follows: empty the inlet andoutlet reservoirs, fill the inlet with 70 µL sample, electrokinetic injection,empty the inlet reservoir via the drain port, fill the inlet and outletreservoirs with 90 µL buffer, and, finally, start the separation. Theseparation voltage remained switched on during the entire injectionprocedure, and the platinum electrodes remained in constant contact withthe liquid (even when the vials were emptied, a small amount of liquidremained at the bottom of the vials). At the beginning of the injectionprocedure, the outlet reservoir was emptied in order to enable hydrostaticinjection. However, the electrokinetic injection predominated thehydrostatic injection due to the high flow resistance of the 6 cm longchannel. The entire injection procedure required approximately 30seconds.

6.5.3 Comparison of the 2- and 4-electrode setup

Electrophoretic separations were performed in 20 mM MES/His buffer atpH 6.0. The injected sample consisted of a 1 mM mixture of potassiumchloride, lithium chloride, and sodium chloride in MilliQ water. Theinjection time (the time the sample remained in the sample reservoir forelectrokinetic injection) was 1 second and the applied injection voltagewas 483 Volts. For separation, the voltage applied to the channel wasincreased to 1500 Volts. The measurement frequency was set to 5, 10, and100 kHz. The resulting electropherograms obtained with the 2-electrode

setup are shown in Fig. 6.9. The three peaks corresponding to K+, Na+ and

Li+ are clearly resolved with complete separation in just under 33 seconds.The same measurements were performed with the 4-electrode setup (Fig.

6.10). The peak shapes obtained for K+, Na+ and Li+ are analogous to theones obtained previously using the 2-electrode setup. The distortion whichis observed in the potassium peak is a result of the injection: during theinjection sequence, sample and buffer were drawn in and out of the samplereservoir while a voltage remained applied to the separation channel. Afterelectrokinetic injection, the sample could not be replaced instantaneously

Chapter 6


3/2/2003

with buffer and an excess of sample was injected (observed as thedistortion). To prevent that problem, the voltage applied to the separationchannel could be switched off during dispensing and removal of sampleand buffer in the sample reservoir. An improved peak shape was observed.However, the amount of injected sample decreased considerably, resultingin a poor detection limit. For that reason, we decided to keep the voltageswitched on during the injection sequence.

The obtained electropherograms are in accordance with the frequencyresponses shown in Fig. 6.7 and Fig. 6.8. For comparison, one should keepin mind that liquid impedance is inversely proportional to liquidconductivity. The 2-electrode response shows that the magnitude of thebaseline, which is linked to the conductivity of the buffer, changes withrespect to the measurement frequency (353, 516, and 1200 arbitrary unit at5, 10, and 100 kHz respectively). This is in agreement with the frequencyresponse obtained in Fig. 6.7, where the measured impedance (reciprocal

20 25 30

350352354356358360

5 kHz

10 kHz

100 kHz

Outp

ut

sig

nal[a

.u.]

Time [s]

516520524528532

1190120012101220123012401250

Fig. 6.9 Separation of potassium chloride, sodium chloride,and lithium chloride (1 mM each) in a 20 mM MES/His buffer (pH 6.0).Conductivity was measured with the contactless 2-electrode setup. Theinjection time was 1 second, the injection voltage 483 Volts, the separationvoltage 1500 Volts (250 V/cm), and the measurement frequency was set to 5,10, and 100 kHz.



3 February 2003, 16:39

of the conductivity) decreases with respect to the measurement frequency.

The 4-electrode response shows a behaviour which is consistent with themeasured frequency responses of Fig. 6.8. The baseline value slightlyincreases for frequencies below 10 kHz (2.68 and 2.80 arbitrary units at 5

and 10 kHz, respectively) and, above 10 kHz, it distinctly increases (4.04volts at 100 kHz) due to the effect of parasitic capacitances.

The sensitivity of a conductivity cell, as it is usually considered, is theratio of the magnitude of the output signal and that of the conductivity

(expressed in Volt/Siemens). It takes into account the input signalamplitude and the various amplifications of the measured signal. This isan absolute sensitivity which does not reflect the intrinsic sensitivity of the

detection cell. The sensitivity coefficient S is more suitable forcharacterizing the conductivity cell [6.8]. It is defined as the percent

20 25 30

2.682.722.762.802.842.88O

utp

ut

sig

nal[a

.u.]

Time [s]

2.882.922.963.003.043.08

4.04

4.08

4.12

4.16

5 kHz

10 kHz

100 kHz

Vmin

Vmax

Fig. 6.10 Separation of potassium chloride, sodium chloride andlithium chloride (1 mM each) in a 20 mM MES/His buffer (pH 6.0).Conductivity detection was done with the contactless 4-electrode setup. Theinjection time was 1 second, the injection voltage 483 Volts, the separationvoltage 1500 Volts (250 V/cm), and the measurement frequency was set to 5,10, and 100 kHz.

Chapter 6


3/2/2003

change in output signal due to a one-percent increase in concentration. It isexpressed in percent and is numerically estimated as:

where Vmax and Vmin are the magnitude of the output signal correspondingto the baseline and to the peak height, respectively (see illustration in Fig.6.10 on the potassium peak). From the previous electropherograms, wecalculated the sensitivity coefficient of the 2- and 4-electrode setupresponse for the 3 ions.

In Fig. 6.11, the profile of the sensitivity coefficient versus themeasurement frequency in the frequency range [1 kHz - 200 kHz] isplotted. For each measurement setup, the profile is similar for the three

SVmax Vmin–

Vmin

---------------------------- 100,×= (6.3)

1000 10000 100000

1

10

potassium

sodium

lithium

rela

tive

impedance-c

hanges

[%]

Frequency [Hz]

Fig. 6.11 Sensitivity versus measurement frequency withcontactless 2- and 4-electrode setup. The bottom set of curves (gray lines)concerns the 2-electrode response. The top set of curves (black lines)concerns the 4-electrode response.



3 February 2003, 16:39

ions. The plot for the 2-electrode shows a linear increase of the sensitivity,between 2 and 100 kHz (at logarithmic scale), which is inverselyproportional to the decrease of the “electrode impedance”. The sensitivitycoefficient reaches an optimal value at around 100 kHz with a value equalto 3.84% for potassium. The two previous observations are in agreementwith the measured frequency response shown in Fig. 6.7 (with 20 mMMES/His). Measurements below 2 kHz were not possible because thesignal was below the noise level. As expected from theory, the sensitivitycoefficient with the 4-electrode setup is increased and that in an extendedfrequency range when compared to the 2-electrode setup. In Fig. 6.11, thesensitivity is always higher for the 4-electrode detector in the frequencyrange 1 kHz -70 kHz. At high frequencies, the sensitivity coefficient dropsdue to the influence of the parasitic capacitances. The optimal sensitivitycoefficient (5.12% for potassium) is obtained at a measurement frequencyof 10 kHz. A slight but constant decrease of the sensitivity coefficient isobserved below 10 kHz, which is not consistent with the theory. We do notyet have an explanation for this decrease.

6.5.4 Measurements with baseline suppression

The first experiment consisted of injecting 1mM (each) of potassium,sodium, and lithium ions. 20 mM MES/His (pH 6) was chosen as thecarrier electrolyte. The separation was performed at 600 Volts. Nobaseline suppression was applied. The corresponding electropherogram isshown in Fig. 6.12. The value of the baseline was in the range of 207.5 to208.5 mV, and therefore the input range of the lock-in amplifier had to beset to 500 mV. The peak height is in the order of 4 mV, depending on theion. The sodium peak (second peak) was taken into account to estimate thesignal-to-noise ratio (SNR). The SNR was calculated as the ratio of thepeak height to the noise amplitude (the peak height is the difference insignal amplitude between the analyte zone and the carrier electrolyte). Acorresponding value of 86.8 dB was found.

In the second experiment, the baseline suppression technique was applied.It is visible in Fig. 6.12 that the peak shape remains identical to that in theprevious measurements. Therefore, the baseline suppression does notinduce any distortion or broadening. After suppression, the baseline level

Chapter 6


3/2/2003

decreased to a value in the range of 2 to 4 mV. That is a reduction of thebackground signal by a ratio of at least 50. Such a reduction of thebaseline allowed us to set the input range of the lock-in amplifier to 10mV (instead of 500 mV). The SNR improvement is visible in Fig. 6.12.The SNR is equal to 106.8 dB, which is 10 times higher than withoutbaseline suppression.

In the third experiment, samples containing 1 mM, 100 µM, and 10 µMpotassium, sodium, and lithium ions were injected. In Fig. 6.13, theelectropherogram corresponding to the injection of the 10 µM sample isvisible. At that concentration, peaks would not be visible without thesuppression of the baseline. Taking the detection limit as three times thenoise level, we could detect concentrations down to 3 µM.

A good suppression is only achieved when the two alternating signalshave the same phase. In the setup described here, the phase and amplitudeinformation of the baseline is acquired only once, shortly before the peakspass the detection window. It is then expected that no phase changes occurafterwards. The recording of the phase information showed that when azone passes, a variation of less than one degree is noticed. This is

Output signal [mV] versus Time [s]

with

outsuppre

ssio

nwith

suppre

ssio

n

0.002

0.004

0.006

0.008

90 140 190

0.2055

0.2075

0.2095

0.2115K+

Na+

Li+

90 140 1902

4

8

6

205.5

207.5

209.5

211.5

Fig. 6.12 Separation of 1 mM (each) of potassium, sodium, andlithium ions in 20 mM MES/His (pH 6) carrier electrolyte. The sameseparation was monitored with and without baseline suppression.



3 February 2003, 16:39

acceptable to achieve a good subtraction and as it can be seen in Fig. 6.12,there is no distortion of the peak.

Baseline drift is another source of limitation for baseline suppression. Forlow sample concentrations, the baseline drifts are larger than the peakheight. In that case, the sensitivity is fixed by the changes of the baseline,which fixes the detection limit. This is, for instance observed in Fig. 6.13,where concentrations of 10 µM are detected.

6.5.5 Detection performance

The linearity of the detector was demonstrated by injecting onlypotassium at concentrations ranging from 1.25 mM down to 12.5 µM. 20mM MES/His (pH 6) was used as background electrolyte. The injectionwas done at 483 Volts, and the separation at 1500 Volts. Two measurementcycles were performed for each concentration. Furthermore, themeasurements were done with and without baseline suppression. Themeasurement frequency was set to 10 kHz and the peak area was extractedfrom each electropherogram. The results are shown in Fig. 6.14, where thelinearity of the detector is demonstrated through a correlation coefficient

-4.22

-4.2

-4.18

-4.16

-4.14

-4.12

-4.1

-4.08

100 120 140 160 180

1 mM

100 M10 M

100 140 1804.08

4.12

4.16

4.2K

+

Na+

Li+

Time [s]

Conductivity

[a.u

.]

10 M zoomed

Fig. 6.13 Separation of 1 mM, 100 µM, and 10 µM (each)potassium, sodium, and lithium in 20 mM MES/His (pH 6) carrier electrolyte.

Chapter 6


3/2/2003

for the linear fit equal to 99.96%. The baseline suppression does notinduce non-linearities.

The reproducibility of the detector response has been tested over 21consecutive runs (3 hours in total). For each run, a sample containing 1mM of potassium, sodium, and lithium was injected. 20 mM MES/His(pH 6) was used as background electrolyte. The injection was done at 483Volts, and the separation at 1500 Volts. The measurement frequency wasset to 10 kHz, and the peak area was extracted for all runs. Relativestandard deviations (RSD) of 1.39, 0.92, and 0.72% were obtained forpotassium, sodium, and lithium, respectively.

Electrophoretic separations were performed at concentrations down to 10µM for each ion and at the optimal measurement frequency (10kHz).Detection limits of 5 µM for potassium, 15 µM for lithium, and 10 µM forsodium were obtained (detection limit defined as three times the baselinenoise level). The signal-to-noise ratio (SNR) was the same for bothdetectors when measuring at the optimal frequency (10 kHz and 100 kHzfor the 4-electrode and 2-electrode setup, respectively). However, the SNRdecreased drastically for the 2-electrode setup when measuring atfrequencies different from the optimal one. Below 2 kHz, detection was

Fig. 6.14 Peak area versus sample concentration using contactless4-electrode conductivity detection. The sample consisted of potassium atconcentrations ranging from 12.5 µM up to 12.5 mM. The injection time was483 Volts, the separation voltage 1500 Volts, and the measurement frequencywas set to 10 kHz.



3 February 2003, 16:39

not possible anymore. For the 4-electrode detector, the SNR alsodecreased when we did not measure at the optimal frequency. However,thanks to a higher sensitivity, detection at frequencies as low as 600 Hzwas possible.

The achieved detection limit is slightly higher than the values reported inliterature [6.9] (2.8 and 6.4 µM for potassium and lithium, respectively).However, two major differences between our device and the one describedin [6.9] determine the achievable detection limit. First, the presence ofabrupt turns in the channel design is a source of zone dispersion [6.10].Second, the single-channel geometry does not allow injection of a short,well-defined sample plug. Improvements on peak shape and detectionlimit are expected by the use of a cross injection geometry in combinationwith a straight channel.

6.6 Separation of organic acids

Six organic acids were separated: fumaric, citric, succinic, pyruvic, acetic,and lactic acid [Fig. 6.15]. The buffer was 20 mM MES/His at pH 5.8, towhich 0.2 mM of TTAB was added to reverse the electroosmotic flow. Anegative power supply (assembled in-house) was used for injection andseparation.

The sample was injected electrokinetically by applying a voltage when thesample was present in the sample reservoirs. In contrast with themeasurements of inorganic ions described previously, the high-voltagesource was switched off during liquid-handling steps in the reservoirs. Insuch a way, the injected zone is shorter, which is necessary to separatemore than three ions. The drawback is that less sample is injected,resulting in lower peak heights, thus a higher detection limit. For theelectrokinetic injection, the voltage was ramped from 0 Volt down to -1000 Volts and ramped up to 0 Volt in 800 milliseconds (total time). Alonger injection time overloaded the injected sample and resulted inoverlapping of the peaks.

The separation was done at -1000 Volts (167 V/cm). The directconsequence of the modified injection method is a higher detection limit

Chapter 6


3/2/2003

compared to that of inorganic ions. The electronic baseline suppressionsetup is necessary to detect concentrations lower than the 1 mM. Adetection limit of 50 µM was deduced on the basis of the noise level.

6.7 Conclusions

The benefits of using four electrodes for contactless conductivitydetection in CE microchips has been demonstrated: The influence of theelectrode impedance (double layer capacitance and insulating layercapacitance) on the response is minimized. Consequently, at frequenciesbelow 10 kHz, the detector response (sensitivity, linearity, and accuracy)is almost independent of the measurement frequency and of theconductivity of the carrier electrolyte.

Fig. 6.15 Separation of 1 mM (each) (1) Fumaric, (2) Citric,(3) Succinic, (4) Pyruvic, (5) Acetic, and (6) Lactic acid in a 20 mM MES/His + 200 µM TTAB (pH 5.8) buffer. Applied separation field: 167 V/cm.



3 February 2003, 16:39

The conductivity detection method was evaluated in the CE microchip inwhich a mixture of inorganic cations (potassium, sodium, and lithium)was separated electrophoretically. On-column detection with a separationfield of up to 250 Volts/cm could be achieved. The three peaks could bedetected with a considerable improvement in sensitivity compared to thatof the classical 2-electrode setup. Detection limits ranging from 15 µMdown to 5 µM were obtained (depending on the ion).

Finally, six organic acids were separated, which extends the applicabilityof the contactless four-electrode detector.

6.8 Future work

The separation and detection of six inorganic ions demonstrate thepotential of both the technology and of the detection method. However,the overall performance is expected to be improved by modifying thedesign of the channel. A straight channel will eliminate the bandbroadening that the turns induce in the microchips described in this thesis(Fig. 3.17). Since the detection limit is linked to the peak shape, a decreaseof the detection limit compared to that of the current microchip isexpected. A detection limit below the µM level for inorganic ions such aspotassium, sodium, and lithium would then be reached. The length of thestraight channel is limited by the dimensions of the chips and therefore abetter injector, such as a cross or a double-T injector, is necessary. A newgeneration of microchips that include a straight channel and a double-Tinjector have been fabricated. They are shown in Fig. 6.16. The length ofthe separation channel (from the double-T injector up to the detector) isequal to 2.4 cm. The channel has a width of 70 µm and a depth of 20 µm.The double-T injector allows the injection of a 100 µm long sample plug.The detector has the same dimensions as those mentioned in Chapter 3.4.2(all electrodes are 100 µm long, inner electrodes are 33 µm wide and outerelectrodes are 20 µm). By keeping the same design for the read-outelectronics we can keep the same electronic interface, which is describedin Chapter 4.

Chapter 6


3/2/2003

Measurement results with the new microchips were not available beforethe deadline of submission of this thesis.

6.9 References

6.1 Gebauer P. et al., “Prediction of zone patterns in capillary zoneelectrophoresis with conductivity detection, concept of the zoneconductivity diagram”, J. of Chromatography A, 1997, 771, 63-71.

6.2 Gaš B. et al., “Optimization of background electrolytes for capillary

detector

30 mm

7.5

mm

1 2 3

4

double-T injector

100 m

3 sample1

2 sample2

1 waste1

4 waste2

Fig. 6.16 New generation of microchips. The microchips include astraight channel and a double-T injector. The reservoirs (1, 2, 3, and 4) arereferenced according to Fig. 2.7.



3 February 2003, 16:39

electrophoresis, I Mathematical and computational model”, J. ofChromatography A, 2001, 905, 269-279.

6.3 Mikkers F.E.P. et al., “Concentration distributions in free zoneelectrophoresis”, J. of Chromatography, 1979, 169, 1-10.

6.4 Hjerten S., “Zone broadening in electrophoresis with specialreference to high-performance electrophoresis in capillaries: aninterplay between theory and practice”, Electrophoresis, 1990, 11,665-690.

6.5 Small H. et al., “Novel ion exchange chromatographic method usingconductimetric detection” Analytical Chemistry, 1975, 47, 1801-1809.

6.6 Avdalovic N. et al, “Determination of cations and anions bycapillary electrophoresis combined with suppressed conductivitydetection” Analytical Chemistry, 1993, 65, 1470-1475.

6.7 Walls F.L., “Suppressed carrier based PM and AM noisemeasurement techniques”, Proceedings of the IEEE Internationalfrequency control symposium, 1997, 485-492.

6.8 Jay, F. et al., In IEEE Standard Dictionary of Electrical andElectronics Terms, 3rd ed., The Institute of Electrical andElectronics Engineers, Inc.: New York, 1984.

6.9 Pumera M. et al., “Contactless Conductivity Detector for MicrochipCapillary Electrophoresis“, Analytical Chemistry, 2002, 74, 1968-1971.

6.10 Culbertson, C.T., et al. “Dispersion Sources for CompactGeometries on Microchips“, Analytical Chemistry, 1998, 70, 3781-3789.


3/2/2003


Summary

2 February 2003, 20:22

Summary

In this thesis, conductivity detection for application in capillaryelectrophoresis microchips is discussed.

Because of the liquid environment of operation, it is necessary to protectthe detector´s metal electrodes with an insulating material. This is referredto as contactless detection. We show that using four electrodes instead oftwo is preferred when designing a micro-scale contactless conductivitydetector. It allows sensitive and accurate sensing in a wide frequencyrange and can be applied for various conductive buffers.

The design of the contactless microdetector (for integration in amicrofluidic channel) aimed at getting the largest capacitive couplingbetween the metal electrodes and the liquid, and at preventing breakdownof the insulating film. After investigating different insulating materials, wedecided to use silicon carbide. Silicon carbide has a sufficiently highbreakdown voltage, and layers without pinholes can be deposited with athickness down to 30 nm.

Parasitic capacitors strongly limit the performance of the readoutelectronics, which is a consequence of the limited value (pF-range) of thecapacitive coupling between the metal electrodes and the liquid.Therefore, we designed a dedicated readout interface. A feedbackamplifier is used to control the current flowing in the detection cell. Stableregulation is possible in a frequency band extending between 10 Hz and100 kHz for measuring liquid resistances of up to 10 MΩ. A high input-impedance differential amplifier is necessary for the readout of the inner-electrode differential voltage. Here again, parasitic capacitors at the innerelectrodes play a negative role because they drive a significant leakagecurrent. In order to meet the high-impedance requirement, a specificbootstrapping technique is introduced. In addition, this technique preventssevere degradation of the common-mode rejection ratio due to componentmismatch.

In addition to the electronics for detection readout, an electronic setup thatallows full decoupling of the contactless conductivity detector and the

Summary


2/2/2003

separation column is presented. The reference potential of the separationvoltage supply is controlled in such a way that breakdown of theinsulating film and leakages of the detection current are prevented.

The conductivity detection method was evaluated in a capillaryelectrophoresis microchip in which a mixture of inorganic cations(potassium, sodium, and lithium) was separated. On-column detectionwith a separation field of up to 250 Volts/cm was achieved. The threepeaks could be clearly detected with a considerable improvement insensitivity compared to the classical two-electrode setup. Detection limitsranging from 15 µM down to 5 µM (for potassium ions) were obtained.Finally, the separation of six organic acids, with a concentration down to50 µM, was demonstrated, which proves the applicability of thecontactless four-electrode detector.


Samenvatting

2 February 2003, 20:15

Samenvatting

Dit proefschrift behandeld de toepasbaarheid van geleidingsmeting alsdetectie methode voor capillaire elektro-forese op microchips. Omchemische interactie tussen de vloeistof en elektrodes te voorkomendienen geïsoleerde elektrodes gebruikt te worden.

Aangetoond wordt dat de vierpuntsmeetmethode te prefereren is boveneen tweepuntsmeting wanneer contactloze detectie wordt toegepast. Thevierpuntmethode koppelt hoge gevoeligheid aan een breedfrequentiebereik en is geschikt voor buffer vloeistoffen met verschillendgeleidingsvermogen.

Het ontwerp van de contactloze microdetector was enerzijds gericht opmaximale capacitieve koppeling tussen elektroden en vloeistof, terwijlanderzijds de doorslagspanning van de isolatielaag voldoende hoog moestblijven. Na vergelijking van verschillende isolatiematerialen werdSiliciumcarbide als isolatielaag gekozen omdat het in zeer dunne lagen(tot 30 nm) opgebracht kan worden en toch een voldoend hogedoorslagspanning heeft.

Door de geringe afmetingen van de metalen elektrodes is de capacitievekoppeling met de vloeistof klein (pF-bereik). Parasitaire capaciteitenhebben daardoor een grote verzwakkende invloed op het signaal dat naarde uitleeselektronica gaat. Daarom is een speciale elektronische interfaceontwikkeld waarin een terugkoppelversterker gebruikt wordt om destroom door de detectiecel zo te reguleren dat een stabiel signaal over eengroot frequentiebereik van 10 Hz tot 100 kHz voor vloeistofweerstandentot 10 MΩ.

Een hoogohmige verschilversterker is noodzakelijk om het signaal van debinnenste elektroden uit te lezen omdat hier de parasitaire capaciteitenhier weer een negatieve rol spelen. Om een voldoend hogeingangsimpedantie te krijgen is een speciale “bootstrapping” techniektoegepast. Bovendien wordt hierdoor de “common-mode” rejectievergroot waardoor de invloed van niet geheel gelijke componentenverkleind wordt.

Samenvatting


2/2/2003

Naast de uitleeselektronica is een tweede elektronische schakelingontwikkeld die een goede ontkoppeling van het uitleescircuit en hetseparatie circuit mogelijk maakt. Het referentie signaal van dehoogspanningsbron wordt daarbij zo geregeld dat de gelijkspanningtussen vloeistof en elektroden geminimaliseerd wordt zodat doorslag vande isolatielaag voorkomen wordt.

De meetmethode werd geëvalueerd m.b.v. een CE microchip waarin eenmengsel van anorganische kationen (kalium, natrium en lithium)gescheiden werd. Detectie in het kanaal met een scheidingsveldsterkte tot250 Volt/cm was mogelijk. De drie pieken in het signaal konden duidelijkwaargenomen worden, aanzienlijk duidelijker dan met de klassieke 2-punts methode. Detectielimieten van 15 µM tot 5 µM (voor kalium ionen)werden aangetoond. Tenslotte werd nog een mengsel van zes organischezuren succesvol gescheiden en gedetecteerd voor concentraties tot 50 µM,wat de brede toepasbaarheid van de contactloze vierpuntsmeting aantoont.


Acknowledgements

3 February 2003, 18:04

Acknowledgements

A thesis work is not the work of one person and in order to keep such aproject running, you need support and help. Thus, I would like to expressmy gratitude to all those who gave me the possibility to complete thisthesis:

Michiel Vellekoop, my main supervisor, provided a motivating,enthusiastic, and critical atmosphere during the many discussions we had.I am very glad I started my professional career with him and I amconvinced he teached me a very good working philosophy. Andre Bossche who as my second supervisor gave me full support andguidance to reach the end of my work. Professor Pasqualina Sarro who provided technical and scientific supportin microchip fabrication.Professors Middelhoek, Huising, and French who have made theElectronic Instrumentation Laboratory a reknown research laboratorywhich is also an exciting place to work on the scientific and social point ofview.The STW (project DST.4351), all the associated companies (Akzo Nobel,Unilever, Antec Leiden, DSM, Applikon), and Leo Korstanje for thefinancial support.Jeroen Bastemeijer who played a major role in this thesis work. Heprovided determining comments and inputs during my thesis time as wellas on the preliminary version of this thesis.Gert van der Steen, Rosanne Guijt, Hugo Billiet and Erik Baltussen withwhom, together, we managed to build up a very fruitful scientificcollaboration.Axel Berthold who brilliantly developed the technology for the microchipand Wim van der Vliest who not only took over Axel’s work but alsointroduced me to the field of IC processing techniques.The Biomas group: Professor Gijs van Dedem, Professor Albert van denBerg, Hans Frank, Stefan Schlautmann, Richard Schasfoort, PatrickFowler, Regina Luttge, and also Sabeth Verpoort

Acknowledgements


3/2/2003

All the members of Michiel’s group: Bonnie Gray, Ventzeslav Iordanov,Jeroen Nieuwenhuis, Sang-soek Lee, Bernhard Jakoby, Kari Hjelt, DilailaCriado, Khalid Elbattay, Peter Szczaurski, Estelle Filippozzi, JeremyCicilia. All the members of Andre’s group Jeff Mollinger, Peter Turmezei,Florin Tatar, Vladimir Kutchoukov, and Robert Kazinczi.My roommates Orla O’Halloran, Miodrag Djurica, Serhat Sakarya,Eamon Connolly, and Luis Rocha.I also wish to thank all past and present members of the ElectronicInstrumentation Laboratory for their contribution to the generalatmosphere. For their professional technical and administrative support, Iwish to thank Piet Trimp, Ger de Graaf, Maureen Meekel, HarryKerkvliet, Willem van der Sluys, Inge Edmond, Evelyn Sharabi, TrudieHouweling, and Sabine van den Boer.Charles de Boer, Alex van den Bogaard, Ruud Klerks, Cassan Visser fromDIMES. Gert Schotte and Mario van der Wel from the Workshop. FransBroos and his group for designing the printed circuit boards.Heidi Diedrich, Frank Wiertz, and Gerard Mulder from the KluyverInstitute.Mirjam Niemam for carefully reviewing the english.All my european, australian, african, north and south american, and asianfriends (Hopefully, I did not forget anybody!).Last but certainly not least, my family and Cecilia for their love andcaring.


List of Publications

3 February 2003, 18:30


F. Laugere, W. Lubking, M. Berthold, J. Bastemeijer, M.J. Vellekoop; "Anovel high-resolution liquid-conductivity detector", Eurosensors XIII,13th European Conference on Solid-State Transducers (The Hague, Sept.1999), 211-214.

F. Laugere, W. Lubking, A. Berthold, J. Bastemeijer, M.J. Vellekoop;"Exploring limits for the design of a miniaturized contactless conductivitydetector for on-chip capillary electrophoresis" Eurosensors XIV, 14thEuropean Conference on Solid-State Transducers (Copenhagen, 27-30Aug. 2000), 791-794.

R.M. Guijt, E. Baltussen, G van der Steen, H. Frank, H. Billiet, T.Schalkhammer, F. Laugere, M.J. Vellekoop, A. Berthold, P.M. Sarro, G.van Dedem; "Capillary electrophoresis with on-chip four electrodecapacitively coupled conductivity detection for application inbioanalysis", 2001, Electrophoresis 22, 2537-2541.

J. Bastemeijer, W. Lubking, F. Laugere, M.J. Vellekoop; "Electronicprotection of the conductivity detector in a micro capillary electrophoresismicrochannel", Transducers'O1, 11th International Conference on Solid-State Sensors and Actuators (Munich, June 10-14, 2001), 100-103.

F. Laugere, W. Lubking, J. Bastemeijer, M.J. Vellekoop; "Dedicatedinterface electronics for capacitively-coupled conductivity detection in on-chip capillary electophoresis", Transducers'O1, 11th InternationalConference on Solid-State Sensors and Actuators (Munich, June 10-14,2001), 60-63.

F. Laugere, W. Lubking, A. Berthold, J. Bastemeijer, M.J. Vellekoop;"Downscaling aspects of a conductivity detector for application in on-chipcapillary electrophoresis" Sensors and actuators A, 2001, 92, 109-114.

F. Laugere, A. Berthold, W. Lubking, J. Bastemeijer, R.M. Guijt, E.Baltussen, P.M. Sarro, M.J. Vellekoop; "Experimental verification of animproved method for conductivity detection in on-chip capillaryelectrophoresis systems" Transducers'O1, 11th International Conference



3/2/2003

on Solid-State Sensors and Actuators (Munich, June 10-14, 2001), 1178-1181.

F. Laugere, A. Berthold, R.M. Guijt, E. Baltussen, J. Bastemeijer, P.M.Sarro, M.J. Vellekoop; "Measurement system for biochemical analysisbased on capillary electrophoresis and microscale conductivity detection"Sensor Technology Conference 2001 (Enschede, 14-15 May, 2001), 1-6.

F. Laugere, W. Lubking, J. Bastemeijer, M.J. Vellekoop; "Design of anelectronic interface for capacitively coupled four-electrode conductivitydetection in capillary electrophoresis microchip", Sensors and actuatorsB, 2002, 83, 104-108.

J. Bastemeijer, W. Lubking, F. Laugere, M.J. Vellekoop; "Electronicprotection methods for conductivity detectors in micro capillaryelectrophoresis devices", Sensors and Actuators B, 2002, 83, 98-103.

E. Baltussen, R.M. Guijt, G. van der Steen, F. Laugere, S., G.W.K. vanDedem; "Considerations on contactless conductivity detection in capillaryelectrophoresis" Electrophoresis, 2002, 23 (17), 2888-2893.

F. Laugere, J. Bastemeijer, G. van der Steen, M.J. Vellekoop, P.M. Sarro,A. Bossche; "Electronic baseline-suppression for liquid conductivitydetection in a capillary electrophoresis microchip", IEEE internationalconference on sensors (Orlando, 12-14 June 2002).

F. Laugere, J. Bastemeijer, M.J. Vellekoop, A. Bossche; "Full-decouplingtechnique for on-column liquid-conductivity detection in capillaryelectrophoresis microchip", Eurosensors XVI. 16th European Conferenceon Solid-State Transducers (Prague, 15-18 Sep. 2002).

F. Laugere, G. van der Steen, J. Bastemeijer, R. M. Guijt, P. M. Sarro, M.J.Vellekoop, A. Bossche; "Separation and detection of organic acids in aCE microchip with contactless four-electrode conductivity detection", The6th International Symposium on Micro Total Analysis System (Nara, 3-7Nov. 2002).

A. Berthold, F. Laugere, H. Schellevis, C. de Boer, M. Laros, R.M.Guijt,P.M.Sarro, M.J.Vellekoop; "Fabrication of a glass-implementedmicrocapillary electrophoresis device with integrated contactlessconductivity detection", Electrophoresis, 2002, 23, 3511-3519.



3 February 2003, 18:30

F. Laugere, R.M. Guijt, J. Bastemeijer, G. van der Steen, A. Berthold, E.Baltussen, P. Sarro, G.W.K. van Dedem, M.J. Vellekoop, A. Bossche;"On-chip Contactless Four-Electrode Conductivity Detection forCapillary Electrophoresis Devices", Analytical Chemistry, 2003, 75 (2),306-312.


3/2/2003


About the author

3 February 2003, 18:32

About the author

Frédéric Laugère was born in Angouleme, France, on April 16, 1974. In1992, he started to study electronics at the Technical University ofKourou, French Guyana. He moved to Bordeaux, France, in 1994 tofollow his studies at the Sciences University of Bordeaux. In 1998 hereceived his Master Science degree at the National School of ElectronicsInformatic and Radioelectricity from Bordeaux (ENSEIRB) with aspecialisation in microelectronics. He joined the ElectronicInstrumentation Laboratory as a Ph.D. student in august 1998, and heworked on the development of a conductivity detector for application incapillary electrophoresis.


3/2/2003

Conductivity detection for application in capillary...

Documents

Transcript of Conductivity detection for application in capillary...