Conductivity detection for conventional and miniaturised capillary electrophoresis systems

Since the introduction of capillary electrophoresis (CE), conductivity detection has been an attractive means of detection. No additional chemical properties are required for detection, and no loss in sensitivity is expected when miniaturising the detector to scale with narrow‐bore capillaries or even to the microchip format. Integration of conductivity and CE, however, involves a challenging combination of engineering issues. In conductivity detection the resistance of the solution is most frequently measured in an alternating current (AC) circuit. The influence of capacitors both in series and in parallel with the solution resistance should be minimised during conductivity measurements. For contact conductivity measurements, the positioning and alignment of the detection electrodes is crucial. A contact conductivity detector for CE has been commercially available, but was withdrawn from the market. Microfabrication technology enables integration and precise alignment of electrodes, resulting in the popularity of conductivity detection in microfluidic devices. In contactless conductivity detection, the alignment of the electrodes with respect to the capillary is less crucial. Contactless conductivity detection (CCD) was introduced in capillary CE, and similar electronics have been applied for CCD using planar electrodes in microfluidic devices. A contactless conductivity detector for capillaries has been commercialised recently. In this review, different approaches towards conductivity detection in capillaries and chip‐based CE are discussed. In contrast to previous reviews, the focus of the present review is on the technological developments and challenges in conductivity detection in CE.


Introduction
Most commercial capillary electrophoresis (CE) instruments are equipped with UV-absorption or fluorescence detectors. Though optical detection methods have proven to be valuable techniques, they also have limitations as not all analytes absorb light and/or possess a fluorescent functionality. Some nonfluorescent analytes can be labelled with a fluorescent compound, however, these labelling reactions are not applicable to many simple ions, such as inorganic ions. Additionally, labelling reactions can be time-consuming, may introduce experimental errors, and can be difficult to perform at low analyte concentrations. For non-UV-absorbing solutes, UV-absorbing buffer additives can be used to facilitate indirect absorbance detection. By analogy, addition of fluorescent compounds to the buffer allows indirect fluorescence detection. However, indirect detection is not always desirable due to possible complications and/or restraints on buffer compositions which are imposed to prevent the occurrence of system peaks, unstable baselines, or peakbroadening due to incompatibility of the mobilities of the analyte-and background co-ions. Additionally, detection limits are generally higher as a result of somewhat compromised sensitivity when measuring a small signal on a relatively high and noisy background. The short detection path length in narrow-bore capillaries often results in unfavourable detection limits for absorbance-based detection methods, even when these are applied to UVabsorbing analytes. Therefore, alternative, robust detection methods are required for CE, especially when downscaling to the microchip format.
Electrochemical detection techniques, including amperometry, potentiometry, and conductometry, have the advantages of high sensitivity, high selectivity, and wide linear range [1]. Additionally, electrochemical detection techniques are not mass-sensitive, but concentrationsensitive, implying that downscaling the detector cell size does not result in a loss in sensitivity. Miniaturisation of the detection electrodes could even result in improved sensitivity as a result of the reduced noise. Furthermore, since the detector response is derived directly from an electric parameter of the solution, conversions between different physical parameters such as light intensity and electricity are eliminated, thereby eliminating a potential additional source of noise [2].
In high-performance liquid chromatography (HPLC), optical detection techniques are also the most frequently used detection methods, with only approximately 5% of the detection being performed using amperometric detectors to measure the current associated with the oxidation or reduction of analytes as they are eluted from the column [1]. The suitability of amperometric detection depends on the redox characteristics of the analyte molecules in the environment of the mobile phase. Conductometric detection is almost uniquely applied in ion chromatography (IC). Over the past several years, several groups worked on integration of conductometric, amperometric, and potentiometric methods with a CE separation in capillaries or in microfabricated devices, as reviewed in [2][3][4][5][6][7][8][9].
The present review focuses on the technological development of conductometric detection in CE. Even though often less sensitive than the other two electrochemical detection methods, conductometry has the advantage of being a universal detection technique. A further advantage of conductometric detection is that direct contact of the detector with the solution under measurement is not essential, as conductivity detection can be performed in the contactless mode, exploiting capacitive coupling with the liquid inside the capillary or channel. This is an elegant way to eliminate interferences by the high separation voltage with the detection electronics. An overview of different approaches for contact and contactless conductivity detection will be given and developments in conductivity detection in capillaries and on microfluidic devices will be discussed. Since these developments have often proceeded in specific directions within the work of a certain research group, this review is structured around particular research groups rather than chronologically. For each of the four sections (contact and contactless conductivity detection in capillaries and on microfluidic devices), a table is provided summarising the developments and experimental parameters. This review concludes with an outlook on the expected developments and issues to be solved for conductivity detection in capillary electrophoresis.

Basic principles
In conductivity detection, the solution resistance R (O) is calculated from its conductance G (S), defined as G = 1/R [10]. The value for G may also be determined from the ratio of the specific conductance k (S?cm 21 ) and the cell constant k cell (cm 21 ) and can be given by: In a conductometric cell, it is impossible to measure only R, since the electronic measurement setup will effectively be a network of capacitors and resistors. To correct for differences between different measurement setups, the cell constant is used for determination of G.
Conductivity detection is generally performed in a detection cell where two electrodes are placed at a fixed distance apart. For conductivity measurements, the use of

CE and CEC
an alternating current (AC) signal is preferred over the use of a direct current (DC) signal in order to prevent polarisation of the electrodes, to avoid electrochemical reactions at the electrode surfaces, and to minimise interferences of the DC separation field with the detection electronics. Capacitors play an important role in AC circuits since they can store and release charge when an applied voltage changes over time. Like a resistor, a capacitor impedes the flow of charge during charging and discharging, resulting in a decrease in the magnitude of the current according to Ohm's law. This behaviour is called the capacitive reactance X c of a capacitor [11].
If an AC signal is applied to one of the electrodes, a current will be generated through the solution, and measured at the other electrode. Measurement of this current is used to calculate the impedance Z using Ohm's law. Z consists of a "real" component, called Z real , that reflects the ohmic resistance, and an imaginary component Z imag reflecting the other cell components, such as the capacitance of the electrodes, the leads, etc. The capacitive reactance of a capacitor depends strongly on the applied AC frequency. In bench-top conductivity meters, the solution conductivity is determined in a relatively large volume. This allows the conductometric detector to be designed so that Z imag is very small relative to Z real . If Z imag ( Z real , the frequency-dependence of the capacitive reactances included in Z imag does not influence the measurement, resulting in a relatively frequency-independent measurement of Z, and therefore the conductance of the solution. The detection cells designed for conductivity detection in capillary or microchip CE are small, and the detector cell volume and geometry are designed to minimise analyte zone-broadening. Unfortunately, the optimal detection cell geometry from the analytical point of view is not optimal from the electronic point of view since it results in a relatively large Z imag . This causes the cell constant, or more precisely the apparent cell constant, to be dependent on the applied AC frequency [12]. As illustrated in Eq. (1), G depends strongly on the cell constant, and as such on the applied AC frequency.

Contact conductivity detection
A two-electrode contact conductivity detector, as shown in Fig. 1A, can be represented by its electrical equivalent circuit, given in Fig. 1B. The conductivity detector here is simplified and considered as two capacitors, each with capacitance C D in series with a resistor of resistance R. The capacitance that results from the electrical double layer formed at the electrode surface is considered to be the largest of those in the measurement setup, and (a) Two metal electrodes are placed inside a capillary or microchannel. During conductivity measurements, the AC signal will be applied to the input electrode, whereas the resulting current will be monitored at the output electrode.
(b) Equivalent circuit model for contact conductivity detection. The detector can be represented by a connection in series of two capacitors, C D , and the resistance of liquid, R solution . For more information, see text.
therefore has the largest influence in the conductivity measurements. As explained above, the impedance of a capacitor strongly depends on the applied frequency [13].
For an electrode in solution, the capacitive reactance X D , is given by where f represents the AC frequency applied over the detection electrodes and C D the double layer capacitance. In aqueous solutions, C D is dependent on the electrode area, and can be approximated by where A is the electrode area in cm 2 [13].
In AC circuits, a phase shift exists between the voltages and currents. To simplify the description of the currents and voltages in AC circuits, the algebra of complex numbers was introduced [14]. The impedance, Z, measured during conductivity detection in a two-electrode setup depends on the resistance of the liquid R, and the two double-layer capacitors For proper characterisation of a contact conductivity cell, a frequency-response diagram of the detection cell should be recorded to determine the optimal measurement frequency for measuring the solution resistance R without the influence of the double layer capacitance of the measurement electrodes.

Contactless conductivity detection
As mentioned earlier, conductivity detection in tubes can also be performed in the contactless mode. Instead of having electrodes in contact with the solution, the electrodes are placed on the outside of the capillary wall. At least two electrodes are required: a working electrode to apply the AC signal and a measurement electrode to register the signal after passing through the cell. The distribution of the electromagnetic field in the cell is determined by the permittivity, the permeability, and the conductivity of the solution [15]. When zones of different ionic species in aqueous solution pass the detector, the conductivity of the medium in the detector is the chief parameter which is changed since the differences in permittivity and permeability of aqueous solutions are usually negligible. This means that changes in the signal registered by the measurement electrode are related to changes in the conductivity of the solution in the cell, thereby enabling contactless conductivity detection to be performed.
The two most frequently used configurations for contactless conductivity detection (CCD) in capillaries, namely with and without a grounded shielding electrode between the input and output electrodes, are illustrated in Fig. 2.
The effective sensing volume of a CCD is determined by the gap (d) between the two sensing electrodes, and not by the electrode length l. The conductivity is measured by applying an AC signal to the input electrode, and measuring the resulting current flowing through the detector assembly at the output electrode. In Fig. 3, a schematic representation of the CCD cells is given, together with an equivalent circuit model of the two measurement setups. The similarity between the equivalent circuit models of contact conductivity detection and CCD with ground plane can be noted. CCD with a grounded plane between the two measurement electrodes to shield the input and output electrodes. The AC signal is applied to the input electrode and the current through the cell is monitored at the output electrode. For more information, see text. . Schematic representation and equivalent circuit models for CCD (a) without and (b) with shielding electrode between the input and output electrodes. C air is the capacitance of the air between the input and output electrode, C wall is the capacitance of the capillary wall, and R solution represents the resistance of the solution.
In the equivalent circuit model, C Wall represents the capacitance of the wall separating the electrode and the measured liquid, and C air represents the capacitance of the air between the electrodes outside the capillary. Similar to the model for contact conductivity detection, R solution is the resistance of the solution, which is the parameter of interest. Due to the frequency dependence of the capacitive reactances in the equivalent circuit, the output signal depends on the applied measurement frequency. The optimal measurement frequency should be determined for measuring R solution with as little influence from the capacitors as possible.
In a manner equivalent to the equations given for contact conductivity detection, an equation to describe the detector response in CCD can be written. Configuration A is given by a capacitor C air in parallel with a serial connection of C wall and R solution . Configuration B is represented by a series of C wall and R. For a capacitor C, the capacitive reactance X C is given in Eq. (1). The impedance of the parallel connection in configuration A can be calculated using Eq. (4): where X air is the capacitive reactance of the air between the detection electrodes and X wall the capacitive reactance of the wall between the electrodes and the solution.
Substituting the respective formulae for X air and X wall in Eq. (4), results in the detector response with respect to the measurement frequency: For configuration B, the impedance of the detector can be calculated using Eq. (4) with 1/X air = 0, since the introduction of the grounded shield eliminates the capacitance of the air between the two measurement electrodes. This gives the same result as C air = 0 in Eq. (5). The detector response for configuration B can thus be written as: A frequency-response diagram can be used to graphically present frequency-dependent behaviour of an electric circuit. Three parameters can be distinguished in the total cell impedance in Eq. (4): the frequency-dependent X wall and X air , and the solution resistance R solution , which is independent of the frequency. For accurate determination of the solution resistance, the influence of the capacitive reactance of the wall and the air should be minimised during the impedance measurement. Since R solution is constant, the cell impedance should not change with frequency when R solution is the main contributor to the impedance. In the frequency-response diagram, the frequency range where R solution can be adequately measured is the frequency range where the cell impedance is constant; this is where the impedance curve is parallel with the x-axis.
In Fig. 4, a frequency-response diagram is shown using a dashed line for a contactless conductivity detector without shielding electrode, and using a solid line for a shielded contactless conductivity detector. At low frequency, the measurement is dominated by the influence of X wall , resulting in a rapidly decreasing value for the cell impedance Z for both shielded and the nonshielded CCD. Above a certain frequency, the X wall is too small to contribute to the total cell impedance, and the slope decreases to zero. The impedance measurements only represent an accurate measurement of R solution if the impedance curve is parallel with the x-axis and not influenced by capacitance. From Eq. (5), one can expect that at high frequency X air will influence the impedance measurements in nonshielded CCD. The frequency, above which the impedance measurement is influenced by X air , can be recognised in the frequency-response diagram by a decrease of the cell impedance after the flat section. Above this frequency, the impedance measurement is not representative of the value of R solution .
In nonshielded CCD, R solution can only be measured in the frequency range where neither X wall nor X air influence the cell impedance. Since C wall and C air are of the same order of magnitude, there is only a narrow frequency window where reliable conductivity measurements can be performed. From Fig. 4, one can conclude that in nonshielded CCD only a very narrow frequency range is available where the impedance measurement reflects R solution . The use of a shielding electrode in CCD elim- Figure 4. Typical frequency-response diagrams for CCD with and without the use of a shielding electrode between the input and output electrodes. Dashed line: CCD without shielding electrode; solid line: CCD with shielding electrode. At low frequencies, the cell impedance Z is strongly influenced by the capacitance of the capillary wall, C wall . At medium frequencies, the resistance of the solution, R solution , is measured and the value for Z remains constant. If no shielding electrode is used, the capacitance of the air C air , influences the impedance measurement, resulting in a decrease of the cell impedance for the dashed curve at high frequency.
inates the X air from the impedance measurement, resulting in reliable conductivity measurements at frequencies that are sufficiently high to eliminate the influence of X wall . In Fig. 4, this is confirmed by the shape of the solid curve.
The advantage of the use of a shielding electrode is even more pronounced when the influence of the value of R solution on the position of the flat section in the frequency-response diagram is taken into account, as illustrated in Fig. 5. For nonshielded CCD, three important conclusions can be drawn. (i) The optimal measurement frequency depends on the fluid resistance, and shifts towards a lower frequency with increasing R solution . This implies that changing the background electrolyte in CE requires re-determination of the optimum measurement frequency since the background conductivity is likely to change. (ii) A nonlinear detector response is expected. In Fig. 5 the cell impedance is plotted against the frequency for three values of R solution : R 1 = 10 R 2 = 100 R 3 . On the log-scale, there is no frequency where these three curves have a constant separation, indicating a nonlinear detector response for resistances of solutions in this range. (iii) The optimum frequency for the separation buffer is by definition not the optimum measurement frequency for the analyte zones since the conductivity of the analyte zone must be different from that of the background for the analyte to be detected. For CCD with a shielding electrode, the frequency above which the contribution of the capacitive reactance of the wall can be neglected shifts up with decreasing conductivity, as illustrated in Fig. 5. Once the minimal measurement frequency has been determined for the solution with the lowest expected conductivity, the detector can be used at this frequency for any buffer or analyte zone with higher conductivity. The response in the shielded CCD is expected to scale linearly with R solution since the curves for R 1 , R 2 , and R 3 become parallel with a constant separation on the log scale at higher frequencies. In CCD, one should take into account that not only the detection cell, but also the detection electronics play an important role in the generation of the signal. Amplifiers, for example, have an optimal frequency, and should be selected according to the optimal measurement frequency.

Specific aspects of conductivity detection in CE
In CE, separation is based on the difference in electrophoretic mobility, m, between different analytes. To obtain maximal peak shape during a separation, the mobility of the co-ion in the background electrolyte (BGE) should match the mobility of the analyte to limit electromigrational dispersion and therefore zone-broadening [16]. If m co-ion , m analyte , separation efficiency is reduced by electromigrational dispersion manifested as "fronting" peaks, whereas if m co-ion . m analyte , peak "tailing" will occur. During the separation, the analyte ions replace BGE co-ions on a basis equivalent to their charge according to the Kohlrausch regulating function [17]. Conductivity detection in CE is therefore based on the difference in conductance between the analyte zone and the BGE. A maximal analytical signal will therefore be obtained when there is a maximal difference in conductance between the analyte and electrolyte. However, the equivalent conductance l equiv and the mobility m of an analyte are related via m = l equiv /F, where F is the Faraday constant [18]. This implies an irreconcilable conflict for conductivity detection in CE, since optimal efficiency is obtained with matching m BGE and m analyte whereas optimal sensitivity is obtained with a maximal difference between the two.
In addition to the requirement of matching mobility for analytes and BGE co-ions, a higher ionic strength in the BGE is favourable to take advantage of stacking effects upon injection of the sample. The compromise for conductivity detection in CE was found in the use of BGEs based on amphoteric buffers since systems display low conductivities and can therefore be used at relatively high concentrations, thus limiting the electrodispersion effects [19]. An even more elegant solution to solve the conflict between conductivity detection and CE was presented in the early 1990s [20][21][22][23][24]. Here, conductivity detection was used in combination with a suppressor. An optimal BGE was used for the separation and the background conductivity was suppressed just before detection. This resulted in the analyte being present in a low-conductivity solution at the point of detection, and therefore in an improved signal-to-noise ratio. Suppressed conductivity detection has only been demonstrated in combination with contact conductivity detection.
The idea of conductivity detection in CE originates in the frequent use of conductivity detection in isotachophoresis (ITP). The differences between two-electrode conductivity detection, one-electrode potential detection, and two-electrode potential gradient detection are explained based on their origin in ITP. In contrast with the continuous buffer systems in CE, ITP utilises a nonlinear separation voltage gradient created in a discontinuous buffer system. The sample is injected between the leading (LE) and terminating (TE) electrolytes, where the mobility of the analyte species is intermediate between that of the LE and TE. During separation, the highest mobility analyte forms a zone with approximately the same ionic strength as the LE [25]. The lower mobility of the analyte ensures that the analyte zone does not "overtake" the LE, and since a lower mobility also means a lower specific conductance, the resistance in this zone will be higher than in the LE. All analytes present in the sample will focus into zones of approximately equal ionic strength according to their mobilities. This results in a step-wise decrease in conductivity along the capillary. The constant current produces a step-wise drop in potential along the capillary, and an increase of the potential at the point of detection over time. Detection of different analyte zones can therefore be based on monitoring the resistance/conductivity of the zones. This can be accomplished by active conductivity detection using a two-electrode conductivity detector, or by passive conductivity detection based either on monitoring of the local separation potential using a single sensing electrode, or on monitoring the potential drop over an electrode pair, a technique better known as potential gradient detection (PGD).

Technical developments in conductivity detection in capillary electrophoresis
The objective of this section is to provide an overview of the technological developments in contact conductivity detection in CE rather than focussing on the analytical aspects. For a more analytically oriented review, the reader is referred to Zemann's review [19], or to the original work mentioned therein. This section consists of four subsections discussing the developments in contact and contactless conductivity detection in capillaries and on microfluidic devices, respectively. For each subsection, a table is provided summarising the technological developments and the detection limits achieved, together with references to the corresponding original data in the literature. Tables 1 and 2 deal with contact conductivity and contactless conductivity, respectively, for capillaries. Tables 3 and 4 describe contact conductivity detection and contactless conductivity detection, respectively, for microdevices. Since the organisation of this review is based on the work performed in different research groups, the source of the research is also mentioned in Tables 1-4.

Contact conductivity detection in capillaries
Based on the group's history in ITP, it was a logical step for the group of Everaerts to use the conductivity detector that was available for ITP in their CE separations [26,27]. Similarly, the early research by the group of Kaniansky [28] on capillary ITP-CE with conductivity detection was carried out using a commercially available ITP apparatus. Since most developments in capillary ITP with conductivity detection were performed in commercially available instruments with only little detail of the detector technology, capillary ITP with integrated conductivity detection is not included in this review. In Table 1, an overview of the different approaches for contact conductivity detection in CE capillaries is given.
A two-electrode conductivity detector has been the most popular choice for contact conductivity detection in CE.
For this reason, the theory on contact conductivity detection in Section 2.1 is based on on-column conductivity detection. In on-column contact conductivity detection, the decoupling of the separation voltage and detection electronics should be done electronically, since the sensing electrodes are placed in the separation field. A simplified alternative is end-column conductivity detection, where the measuring electrode is placed at a short distance from the capillary end [29]. Here, the changes in conductivity between the detection electrode at the capillary outlet and a reference electrode placed inside the reservoir are monitored. Since both the measuring and reference electrode are placed inside the same, grounded reservoir, a significantly diminished interference with the separation field is expected. However, a difficulty in this approach is the positioning of the sensing electrode. If the sensing electrode is placed too close to the capillary, interferences with the separation field will occur, but if the electrode is placed too far away from the capillary outlet, peak-broadening will be introduced and the detection sensitivity will decrease. Since few details on the detector technology were provided for the work described above and also for the measurements of the electroosmotic flow in CE [30] and the determination of chlorophenoxy acid herbicides by CE using integrated potential gradient detection [31], this work is excluded from Table 1.
The first group reporting on-column contact conductivity detection in CE was the group of Zare [32]. In their initial paper in 1987, an in-line conductivity cell was constructed by fixing Pt wires in diametrically opposite holes in fused-silica capillaries. The 40 mm diameter holes to accommodate 25 mm diameter Pt wires were drilled using a CO 2 -laser. A transformer was used as a galvanic insulator between the high separation voltage and the detection electronics. The same setup was used for the determination of low-molecular-weight carboxylic acids [33] and for the quantitation of Li 1 in serum [34]. It was also used for determination of the effect of electrolyte and sample concentration on the sensitivity and resolution in CE [35]. Using conductivity detection in 1991, this group presented an improved, end-column column conductivity detector for CE [36]. Again, a laser was used for drilling a hole into the wall of a fused-silica capillary, but in this case the sensing electrode was placed inside the capillary from the capillary ending, and positioned to end just under the FS, fused-silica; FEP, fluorinated ethylene-propylene copolymer; n.a., not available drilled hole. The capillary/electrode ending was sealed using epoxy and the electrode connected to an insulated wire. The drilled hole allowed electrical contact with the grounded electrolyte reservoir and served as fluidic outlet during separation and flushing. In 1998, Zhao et al. [37] presented floating potential electronics to eliminate interference between the separation field and detection electronics. Their detector consisted of insulated Pt wires that were glued to the outside of the fused-silica capillary close to the capillary end. The bare electrode ends were aligned carefully with the capillary outlet and the electrodes were connected to battery-powered detection electronics. An infrared transceiver was used to transfer the data to a personal computer for data acquisition.
In electronics, a signal can be isolated from its source using an optically coupled isolator [38]. A photocouple separator was used for optical decoupling of separation and detection electronics by Mo et al. [39]. The photocouple separator consisted of both an electric signal-to-light converter and a light-to-electric signal converter. Two pieces of fused-silica capillary (70 and 6 cm long, respectively) were connected by a metal wire placed inside the capillaries, leaving a small gap. Two Pt electrodes ([ 30 mm) were then positioned carefully in this gap, perpendicular to the metal wire and capillaries. This assembly was glued together, and after curing of the glue the metal wire was removed, resulting in a conductivity cell at 6 cm from the end of the 76 cm long separation capillary. The conductivity cell was connected to a commercially available conductivity meter that was connected to the recorder via the photocouple separator. The photocouple separator allowed recording of the measurement signal, but prevented direct contact between the detection cell and ground. In this way, the detector was a floating device and interference of the separation field with the conductivity signal was effectively eliminated.
In contrast with all two-electrode systems, Prest et al. [40] introduced a single-electrode conductivity detector for both capillaries and microfluidic devices. The electrode was placed across the separation channel or capillary and connected to a commercially available high-voltage (HV) meter. Using the HV meter, the local potential of the detection electrode was monitored. The functionality of this passive conductivity detector was demonstrated using ITP as separation technique.
The only contact conductivity detector that has been commercially available (and only for several years in the late 1990s) was produced by Thermo Bioanalysis for their Crystal 1000 CE instrument [29]. The conductivity cell consisted of a ConCap I connector containing the fusedsilica capillary, and the ConTip I hosting the conductivity sensor. Inside the ConTip I, a Pt wire electrode ([ 150 mm) was surrounded by polyamide for insulation and a spacer to fit in a hollow stainless steel electrode with an inner diameter of 375 mm. The ConCap I and ConTip I were connected leaving a 24 mm gap between the capillary outlet and the sensor. Applications of the Crystal 1000 with the "ConCap-ConTip" detector can be found in [41][42][43][44][45][46][47][48][49][50][51][52][53][54][55][56] and the wide range of applications of the Crystal 1000 is not covered in Table 1. Unfortunately, the capillary in the Con-Cap I module could not be replaced, which resulted in additional costs for capillary replacement since the ConCap I had to be replaced as well. To circumvent this, Gallagher et al. [45] introduced a splicing method to re-use the detection cell with only little loss in separation efficiency. In this method, the length of the capillary connected to the Con-Cap I was reduced, and a new capillary was connected to the remainders of the original capillary. Splicing could be used to replace broken/clogged capillaries as long as the problem was not in the capillary inside the ConCap I.
As mentioned previously, there is a conflict in the requirements for the BGE composition for optimal separation efficiency and optimal detection sensitivity when using conductivity detection in CE. In IC, the use of a suppressor to lower the background conductivity prior to detection leads to an improvement of detection limit of typically one order of magnitude. Suppressors are based on ion exchange and improve the detection limits and signal-tonoise ratio in conductivity detection in two ways: (i) a decrease in the background conductivity to reduce the noise, and (ii) an increase of the conductance of the sample to decrease the detection limit [57]. The replacement of sodium ions by hydrogen ions to improve the detection of Cl 2 in a background of sodium carbonate using a cation exchanger in the H 1 form is illustrated in Fig. 6. H 1 ions are continuously introduced into the ion-exchanger from the regenerant solution to replace the Na 1 ions, diffusing from the capillary into the regenerant solution based on the concentration gradient. The background conductivity is reduced since the acidic environment will shift the equilibrium of carbonic acid to its nondissociated form so the carbonate ions will be present predominantly as the nonconductive CO 2(aq) . After the exchange of Na 1 for H 1 , the analyte co-ion becomes H 1 and the higher specific conductivity of H 1 increases the conductivity of the analyte zone, and therefore the signal in conductivity detection. Most suppressors are based on acid-base reactions, and in addition to membrane-based systems, packed-column suppressors are also available. Suppressed conductivity detection is the most frequently used detection method in IC. For more information on suppressor technology in IC, the reader is referred to Haddad's review [57].
In 1993, two different research groups reported the use of a suppressor in combination with contact conductivity detection in CE [20,22]. Dasgupta and Bao [22] coupled a Nafion membrane between two pieces of fused-silica capillary. Using a metal wire, a 70 mm wide channel was made inside the Nafion membrane to guarantee fluidic contact between the two capillaries. PVC tubing was placed around the capillary-membrane couplings at each end to keep the assembly in place. In a later version, grafted Teflon tubing was used as the suppressor instead of the Nafion membrane in order to improve flow characteristics. The two detection electrodes were placed inside this poly(vinylchloride) (PVC) tubing and positioned Figure 6. Chemical suppression using a cationexchange membrane in the H 1 form. The background conductivity due to the NaHCO 3 is reduced by exchanging the sodium ions for hydrogen ions. This results in the formation of H 2 CO 3 , which immediately dissociates into H 2 O and CO 2 , thereby reducing the background conductivity. At the same time, the analyte co-ion, Na 1 , is replaced by the more conductive hydrogen ion, resulting in a more conductive analyte/co-ion zone.
carefully to achieve an optimal cell constant for the detection electronics. The ground electrode for the CE separation was placed in the regenerant solution around the suppressor membrane.
Later, a similar suppressed system was used with a bipolar pulse conductivity detector, with a schematic drawing of this system being given in Fig. 7 [23]. In 1970, the use of a bipolar pulse for conductivity measurements was introduced to eliminate the influence of capacitors on AC conductivity measurements [58]. Bipolar pulse conductivity detection involves the application of two successive voltage pulses of equal magnitude and opposite polarity to one of the two detection electrodes. At the second detection electrode, the current through the cell is measured at the end of the second pulse. Because of the pulsing, the capacitors in the cell are charged and discharged to the same extent, which results in no net current from the capacitors and the current through the cell then being due only due to the resistance of the liquid R solution . In 1978, the pulsing system was upgraded to a computer-controlled system [59]. In [23], a bifilar wire punched through a piece of PVC tubing was used as conductivity detector. A bifilar wire consists of two metal wires, separated by an insulating layer. The bipolar pulse was applied to one of the two wires, and the current was measured at the end of the bipolar pulse at the second wire. As illustrated in Fig. 8, the use of a computer-controlled bipolar waveform for conductivity resulted in improved sensitivity and a reduction in the noise, when compared with commercially available electronics [23].
Similar to the suppressed conductivity detector described above, Avdalovic et al. [20] presented a suppressed conductivity detector in 1993. Here, the separation capillary outlet is connected to a chemically modified and functionalized Teflon capillary using sleeves to form a 1 cm long suppressor. At the other end of the suppressor, the Teflon capillary was connected to a piece of fusedsilica capillary containing one of the detection electrodes for end-column detection. This assembly was placed inside a vial containing the electrolyte/regenerant solution and also the grounded electrode for the separation voltage. This electrode also served as the detection reference electrode. Later, a modified version of this detector was used for the development of strategies of selectivity control for small ions [21]. No technical details on the modifications of the detection cell were given.

Figure 7.
Suppressed contact conductivity detection in a fusedsilica capillary. The capillary is connected to a tubular Nafion membrane using PVC tubing. The Nafion ion-exchange membrane is placed inside a reservoir containing both the regenerant solution and the separation ground electrode. A second piece of fused-silica (FS) capillary connects the suppressor to the conductivity cell, consisting of a bare bifilar wire punctured through PVC tubing. A bipolar pulse is applied to one of the wires, and the current is measured at the other wire at the end of the bipolar pulse. Drawing made based on information in [22,23]. Both of the above suppressed contact conductivity detectors were not commercialized. The assembly of the fused-silica separation capillary, the suppressor, and the conductivity cell was a demanding process that involved careful alignment. The use of these detectors was even further complicated by the requirement to dissipate the separation field before detection. In both setups, the ground electrode for the separation was positioned in a vial containing the regenerant solution for the suppressor. This eliminated a significant potential drop over the suppressor, thereby preventing electrophoretic transport through the suppressor towards the detector. It was therefore necessary that the residual EOF should be sufficiently high to carry the analytes through the suppressor towards the detection electrode. This limited the separation conditions to those that provided a high EOF towards the ground electrode.

CCD in capillaries
In a manner similar to the development of contact conductivity detection, the original idea of CCD in CE also originates in ITP. In the late 1970s, Gaš et al. [15] introduced a high-frequency contactless conductivity detector for ITP. Four electrodes were positioned radially around a capillary and a 1 MHz AC signal applied to two of the four electrodes, with the two other electrodes being used for measurement. Five years later, improvements on the electronics were published by Vacík et al. [60]. Devel-opments and applications of CCD in capillary ITP are beyond the scope of this review and are not discussed further. The developments in contactless conductivity detection in CE are summarised in Table 2.
In 1998, two independent research groups introduced CCD in CE capillaries [61,62]. CCD strategies generally use two cylindrical electrodes placed around the separation capillary, as illustrated in Fig. 2. Using capacitive coupling, the conductivity of the liquid inside the capillary can be measured when applying an AC voltage to the input electrode, in accordance with Eq. (5). In Table 2, detection parameters of interest for CE-CCD are indicated. Most of the work has been performed in fusedsilica capillaries with an inner diameter of 50 or 75 mm. Since no influence of the capillary diameter on CCD could be observed, this parameter is not included in Table 2.
Both groups initially used silver paint to make the electrodes. In the initial paper, Zemann et al. [61] already noted the advantage of using cylindrical metal tubes made of hypodermic needles to eliminate the painting process. In the second paper originating from the group of Da Silva and Do Lago, the silver paint was replaced by a cylindrical wire-wrapped copper electrode.
There are two interesting differences between the approaches of the two research groups: the use of a shielding electrode and the measurement frequency. In the initial work from Zemann's group, no shielding electrode was used (configuration Fig. 2A), with the shielding electrode being introduced in their second paper (config-uration Fig. 2B) to reduce capacitive transition between the two electrodes [64]. On the other hand, in the initial work of Da Silva and Do Lago a shielding electrode was used, and in their second paper, this electrode was removed on the basis that no reduction of the signal-tonoise (S/N) ratio could be observed when using a shielding electrode. Additionally, a lower peak-to-peak voltage (V pp ) could be used without increasing the detection limit and without the shielding electrode. The other important difference between the two approaches is the applied AC frequency. The Zemann group initially used 40 kHz and increased this value to 100 kHz when the shielding electrode was introduced. No comments on this increase in measurement frequency were given, but based on Fig. 4B one can conclude that the use of a shielding electrode allows the use of higher frequencies for stable measurements. The group of Do Lago and Da Silva reports optimal sensitivity around 600 kHz. The influences of the measurement frequency and V pp on S/N ratio of the output signal and on the peak area were investigated. As can be expected from Fig. 4A, the detector behaviour at high frequencies was not stable, but an optimal result was achieved at 600 kHz with V pp =2 V.
After demonstration of CCD for CE separations of inorganic cations and anions, the detector used by Zemann's group was utilised for detection after an ion-exchange capillary electrochromatography [65]. The detector was modified [66] and miniaturised [67] for increased flexibility of positioning of the detector on the capillary, leading to successful simultaneous detection of cations and anions. Do Lago and co-workers [68] used their detector for the determination of aliphatic alcohols in micellar electrokinetic chromatography, for the CE separation of mono and disaccharides [69], fatty acids [70], and class IA and IIA metal ions [71]. The detector was also used in combination with CE separations in a project for environmental monitoring [72,73]. The research on CCD in this group resulted in the development of the TraceDec  contactless conductivity detector that became commercially available early 2004 [74].
Jelínek and co-workers [75] presented a CCD for easy capillary replacement in 2000. Instead of cylindrical electrodes, semicircular electrodes were made of aluminum foil. Electrode configurations with both electrodes at the same side of the capillary or on opposite sides were compared with a constant gap in between. No difference in performance was observed, but the first configuration was preferred because of practical advantages in construction and use. A year later, a dual photometric-CCD detector was presented [76]. The unstable behaviour expected at some frequencies (Fig. 4A) in the absence of a shielding electrode was demonstrated below 100 and above 400 kHz. Therefore, measurements were performed at 200 kHz. An identical detector was used for the detection of cyclodextrins and derivatives after CE separation [77]. A CCD was integrated into the capillary cassette of the Agilent 3D CE instrument by Vuorinen et al. [78]. The detector electrodes consisted of 7 mm long coils of copper wire mounted on a printed circuit board. In addition to the separation of five inorganic cations, the setup was used for indirect conductometric detection of four catecholamines.
The confusion in the literature on the use of a shielding electrode in CCD ( Figs. 2A or B) was addressed by the work of Baltussen et al. [79]. Using frequency-response diagrams, the advantage of using a shielding electrode for stable measurements was demonstrated. The elimination of C air from the system resulted in accurate measurements of the solution resistance at frequencies where the effect of the wall capacitance can be neglected. This generally occurs above 100 kHz.
In 2002, Tanyanyiwa [80] changed the CCD standards by using a significantly higher peak-to-peak voltage of 250 V. A frequency-response diagram was used to demonstrate that stable measurements could be expected at frequencies higher than 100 kHz. The signal increase as a result of an increase of V pp from 25 to 250 V is illustrated in Fig. 9. The detection limit for Cl 2 at 190 kHz using 250 V was 0.1 mM, one order of magnitude lower than presented before. In another paper by the Hauser group [81], this detection limit was further improved to 18 nM for Cl 2 using Figure 9. Influence of peak-to-peak voltage (V pp ) in CCD. Electropherograms for the separation of (1) potassium,  [80], with permission. f = 100 kHz and V pp = 450 V. Detection limits for cations were found to be a factor 10 higher than for anions due to baseline instabilities. Detection limits comparable with those reported before were obtained for cations and anions separated in a polyether ether ketone (PEEK) capillary [81], and for the detection of heavy metal cations at 100 kHz and V pp = 300 V in a fused-silica capillary [82]. The setup was also used for the analysis of underivatised amino acids [83].
Based on the developments described in the literature, Kubáň et al. [80] assembled a CCD and adjusted its optimal measurement frequency to match the optimal frequency for the amplifier [84]. This detector was used for simultaneous detection of cations and anions with dualopposite injection. After a slight modification of V pp from 20 V to 10 V, this detector was used for the detection of 22 inorganic cations and anions in less than 3 min [84,85]. As expected, this decrease in V pp resulted in a slightly increased detection limit. This detector was also applied for CCD in a novel, flow-injection analysis CE system [86][87][88]. In collaboration with the research group of Hauser, the V pp was increased to 225 V, resulting in a decrease of the detection limit [89].
In contrast to the approaches given in Table 2, Kaniansky et al. [90] presented the use of a commercially available contactless conductivity detector for ITP detection of CE separations. Polytetrafluoroethylene (PTFE) capillaries with an internal diameter of 300 mm were used and a detection limit of 0.65 mM for Cl 2 was obtained. The detector consisted of four electrodes placed around the capillary. No further details of the detector were provided, and since the detection was also performed in a significantly larger capillary, this research is not included in Table 2.

Contact conductivity detection on microdevices
Since conductivity detection is the detection method of choice in "conventional" ITP, it is not surprising that the first reports of integrated conductivity detection in microfluidic devices were found in this area. Since all developments in microfluidics are fairly recent, research on ITP with integrated conductivity detection on a chip is considered in this review. The developments in contact conductivity detection in microfluidic devices are summarised in Table 3.
Even before the introduction of the miniaturised total analysis system (mTAS) in 1990 [91], Gebauer et al. [92] presented a CE device with an integrated contact conductivity detector. A rectangular, 1 mm wide and 200 mm deep groove was made in a Perspex block, and this channel was sealed with a polymer foil. The device was placed with the foil side on a thermostatted metal plate for cooling purposes. The channel was equipped with two Pt electrodes connected with a voltmeter for potential gradient detection. The device was used for the CE separation of 1 mM nitrate, chloride, nitrite, and sulfate.
The first reports on contact conductivity detection in microfluidic devices after the introduction of the mTAS concept can be found in the proceedings of the mTAS conference in 1998 in Banff [93,94]. Cost-effective microfluidic devices were made by casting a poly(dimethyl) siloxane (PDMS or silicone rubber) chip on an etched copper substrate. In Fielden's contribution, the device was sealed on top of a printed circuit board containing two 100 mm wide, parallel gold electrodes with a 200 mm gap in between the electrodes [93]. These gold electrodes were used for passive single-and for active dual-electrode conductivity detection. In Baldock's contribution, two parallel 100 mm wide gold electrodes were used for conductivity detection [94]. Both devices were used for the ITP separation of dyes, Fielden in combination with a mixture of inorganic cations and Baldock in combination with F 2 . This work, together with the research on single-electrode conductivity detection in capillaries described earlier [40], resulted in an altered design and the use of a wire electrode in a PDMS device [95]. The single-electrode conductivity detector in the microfluidic device was connected to a HV probe for measuring the potential. The ITP-separation in the microchannel using passive one-electrode conductivity detection was compared with an identical separation in a capillary, using active two-electrode conductivity detection. The electropherogram originating from the capillary setup showed steeper steps between different analyte zones and superior reproducibility. The steeper steps in the capillary were a result of the more stable active two-electrode conductivity detection, and the reduced reproducibility was attributed to the manual injection technique used in the microdevices. In their next paper, these issues were solved by the presentation of a bidirectional ITP chip with automated injection and two-electrode conductivity detection [96]. The microdevices were made by an inhouse developed milling technique in poly(methyl methacrylate) (PMMA). The detector consisted of two Pt wires with a diameter of 200 mm, positioned into special slots and secured using UV-curable glue. Isolation of the detection electronics from the high separation voltage was achieved by using an oscillator, employing capacitive coupling to the measurement electrodes [97]. Using this optimised microdevice, selenium(IV) and (VI) were separated from chloride, sulfate, nitrite, fluoride, sulfite, and citrate. The detection limit for both Se(IV) and Se(VI) was  PTFE, polytetrafluoroethylene; PMMA, poly(methylmethacrylate); bp, base pairs; PDMS, polydimethylsiloxane; PET, polyethylene terephtalate; PDG, potential gradient detection 6 mM, and the calibration plot was linear between 10 and 320 mM. An identical experimental setup was used for the determination of ascorbate in photographic developer solutions [98], the determination of Se(III) and Se(V) [99], and for the determination of As(V) in the wastewater streams of a metal processing plant [100].
The same group used polymer electrodes for conductivity detection in injection moulded devices in polystyrene and Zeonor, an amorphous cyclic olefin homopolymer [101]. These polymers were selected due to their favourable characteristics of low shrinkage, high rigidity, optical transparency, and chemical resistance. Three different materials were compared for fabrication of the detection electrodes: 8% carbon black-filled polystyrene and 40% carbon fibre-filled polystyrene for the polystyrene devices and 40% carbon fibre-filled nylon 6/6 for the Zeonor devices. The conductivity of the carbon black-filled polystyrene was found to be insufficient for sensitive conductivity detection. The two other combinations (polystyrene with carbon fibre-filled polystyrene electrodes and Zeonor with carbon fibre-filled nylon electrodes) were both suitable for sensitive conductivity detection in ITP. However, the dissimilarities between the nylon electrodes and the Zeonor devices often led to failing or incomplete bonding, resulting in leakage of liquids between the top and bottom plates. The polystyrene devices with polystyrene-based electrodes were indicated to be favourable because of the similar properties of the materials.
A series of papers on contact conductivity detection on PMMA devices for ITP and ITP-CE by Kaniansky and coworkers was initiated after a study of detection electrode geometries was reported by Grass et al. [102]. The detector consisted of platinum electrodes deposited using physical vapour deposition and patterned using a lift-off process on the cover plate. The channels, 200 mm wide and deep, were made by hot embossing in the bottom plate and sealed to the cover plate. The detection electrodes were initially connected to a commercially available conductivity detector and later to a customdesigned electronic control unit. Cell constants for the different electrode geometries, illustrated in Fig. 10, were calculated based on the electrode area. The narrow electrodes in configuration Fig. 10b were found to have the largest cell constant and were therefore expected to measure the largest resistance for the same solution by all detectors. The relative step heights (RSHs) between leading and terminating electrolyte were measured using each detector configuration. Configuration Fig. 10b measured the largest difference, even though the exact value could not be determined since it was out of range for the electronics used. The different detectors were compared based on their performance in identical ITP runs. The step heights between different zones were significantly higher for Fig. 10b, allowing the detection of small differences in conductivity. The narrow electrodes in Fig. 10b also allowed the detection of narrow zones resulting from analytes present in low concentration. Although this paper demonstrated the advantages of using Fig. 10b, configuration Fig. 10a was used for later research by Grass and by Kaniansky and co-workers [103][104][105][106][107][108][109][110][111][112][113][114][115][116]. In a series of papers, separation conditions for ITP-CE were optimised and used for numerous analyses of "real" samples including wine, beer, plasma, and tomato ketchup. A fluid handling system was developed to prevent syphoning effects in the channel network and resulted in improved reproducibility [107]. Details of the fluid handling system are described in a previous review on hydrodynamically closed separation systems [112].
Guijt et al. [117] presented CE separations in microfluidic devices with integrated contact conductivity detection. Two novel microfabrication methods were used here, namely miniaturised transparent insulating channel (mTIC) and powder blasting. In both cases, the channel wafer was bound to a glass substrate containing the electrodes for conductivity detection. The Pt detection electrodes were deposited into small, etched recesses to ensure a flat and smooth surface of the electrode wafer. After alignment and bonding, the electrodes were present in the serpentine separation channel. The use of mTICs for further research in CE was not recommended due to the extremely fragile nature of the 360 nm thick silicon-rich nitride channel walls. The poor resolution obtained for the separation of Na 1 1 K 1 , and Li 1 in mTICs was blamed on (i) the low separation field, limited by 200 mm wide electrodes present in the separation channel, and (ii) the large sample injection volumes, due to the single-channel injection method. The same injection method and chip layout were used for the fabrication of the powder-blasted devices [118]. These factors, in combination with the high channel wall roughness and related problems in bonding the devices, resulted in low resolution and reproducibility.
Microfluidic devices made from a previously unused material for microfluidic devices, corundum ceramics (Al 2 O 3 ), were used by Deyl and co-workers. In an initial paper, focus was chiefly on the development of a tool for fluid manipulation inside microfluidic devices [119]. In the second paper, the CE separation of SDS-protein complexes in capillaries and microfluidic devices was compared [120]. Even though no gain in analysis time was obtained, a significant improvement of the resolution was observed in the microdevices, possibly as a result of the surface properties of the corundum devices. In both papers, a pair of gold electrodes deposited in parallel and positioned perpendicular to the channel was used for conductivity detection. No further details on the detector geometry or electronics were provided.
Galloway and Soper [121] presented the analysis of polymerase chain reaction (PCR) products by CEC on a microfluidic device with integrated conductivity detection. The embossed PMMA devices contained guide channels for the detection electrodes. Prior to bonding, polished Pt wires were positioned inside these grooves and inspected using a stereomicroscope. The detector design and electronics using a bipolar pulsed waveform were transferred from the conductivity detector used by the same group for the analysis of PCR products by capillary reversed-phase chromatography [122]. The use of a bipolar pulse enabled discrimination between the capacitors and the fluid resistance in the detection cell. A detailed description of the use of a bipolar pulse for conductivity detection can be found in Section 3.1, and in [23,58,59]. Typically, a 5 kHz pulse was used with a pulse width of 5 ms. The separation of the PCR products was performed by reversed-phase ion-pair CEC. The surface of the microchannels was coated with a C 18 -phase to permit effective partitioning of the ion-pair complexes into the stationary phase. An ion-pairing reagent was added to the mobile phase and the PCR products of 100-2000 DNA base pairs were separated and detected.
The most sensitive contact conductivity detection to date was also presented by Galloway et al. [123], using a similar setup as described above, again using a bipolar pulse waveform for conductivity detection. An electropherogram for the CE separation of amino acids and the calibration plot for alanine are given in Fig. 11. A detection limit of 8 nM for alanine was obtained after CE separation, which corresponds to an absolute amount of 3.4 amol. This is in the same order of magnitude as the detection limit for unlabeled amino acids using indirect fluorescence detection in CE microdevices [124].  (4) tryptophan in an unmodified PMMA microchip using bipolar pulse contact conductivity detection. Operation conditions: separation field, 150 V/cm; electrokinetic injection, 3 s; running buffer, 10 mM triethylammonium acetate (pH 7). Detector operated at 5.0 kHz with a bipolar pulse amplitude of 0.5 V. Reprinted from [123], with permission.
Laser ablation was used for the fabrication of devices with integrated conductivity detector in polyethylene terephthalate (PET) by Bai et al. [125]. The detection electrodes consisted of carbon ink deposited into an ablated groove. After pasting the carbon ink into the groove, the separation channel was ablated perpendicularly to the groove, cutting the groove in two pieces. This ensured the two detection electrodes to be perfectly aligned with respect to each other, and to be flush with the channel wall. The carbon paste electrodes were connected to a commercially available conductivity meter and used for detection in a CE separation of 1 mM K 1 , Na 1 , and Li 1 in a poly(allyamine) hydrochloride-coated channel. The same group presented a paper on potential gradient detection (PGD) in CE, where this detection technique was named "passive conductivity detection" [126]. The microfabrication method was identical to the one described above, incorporating a second detection electrode in parallel with the other electrode. In PGD the separation field is used for generation of the measurement signal, thereby simplifying the electronics and eliminating the need for decoupling the separation and detection circuits. The potential difference between the two measurement electrodes, DV D , is measured and is proportional to the electrical field strength, the local solution resistivity and the distance between the two electrodes. When the separation channel is filled with buffer, the values for these parameters are constant, so a stable baseline is obtained. If a sample zone passes the detector the local solution resistivity changes, resulting in a change in DV D . A highimpedance voltmeter is required for the measurements to prevent any current flowing though the detector.
Potential gradient detection in CE was also presented by Feng et al. [127]. Here, a channel was etched perpendicular to and 0.2 mm from the end of the separation channel. This channel ended in a separate reservoir containing the measuring electrode. The potential difference between this electrode and the separation ground (GND) DV D was monitored continuously. Compared with PGD described above, the resistance of the fluid in the side channel is added in series with the resistance of the section of the separation channel where the measurement takes place. Again, DV D is proportional to the electrical field strength and total resistivity between the two electrodes. If an analyte zone migrates to the GND reservoir, a change in DV D is measured as a result of the change in resistivity in the measurement section. The devices were used for the separation of a 2 mM mixture of Na 1 , K 1 , and Li 1 in CE, and of a mixture of the alkaloids strychnine and brucine in nonaqueous CE. Both passive detection techniques were found to be less sensitive than active conductometry.

CCD on microdevices
The developments in contactless conductivity detection in microfluidic devices are summarised in Table 4.
A technologically elegant four-electrode contactless conductivity detector was developed by three groups at the Delft University for Technology, the Netherlands. The detector consisted of four aluminum electrodes that were physically isolated and electrically insulated from the separation channel by a 30 nm thick layer of silicon carbide. The electrodes were deposited in recesses to ensure the detector was flush with the surface to prevent  [129][130][131][132]. For the two-electrode measurements, an AC voltage was applied to one outer electrode, whereas the resulting current was measured at the other outer electrode, a configuration similar to CCD in capillaries. The equivalent circuit of the measurement setup was similar to the one described in Fig. 3a. It consisted of the fluid resistance R solution in series with the capacitance of the double layer and insulating layer and in parallel with the capacitance of the liquid. This should have resulted in a frequency-response diagram comparable with the one shown in Fig. 4a, but in practice, there was no flat area in the frequency-response diagram. This meant that there was no single frequency where reproducible measurements could be preformed with small variations in the background conductivity of the BGE. For each BGE, a frequency-response plot should therefore be made to determine the optimal frequency, and even then the linearity of the detector response with increasing concentration would be in some doubt. During the fourelectrode measurements, an AC signal was applied to one outer electrode and the resulting current (I 0 ) was measured at the other outer electrode. Simultaneously, the potential drop (V 0 ) was measured between the two inner electrodes. The conductivity of the solution was calculated from the real-time division I 0 /V 0 . By measuring simultaneously the resulting voltage and current of the applied AC signal, the influence of the double layer and insulating layer capacitors was greatly reduced. This resulted in stable measurements for frequencies up to 10 kHz, above which the capacitance of the liquid contributed to the total impedance, thus corrupting the measurement. The four-electrode measurement setup is illustrated in Fig. 12. In the paper by Guijt et al. [133], the devices with integrated four-electrode CCD were used for the separation of two peptides.
Another approach for CCD on glass microdevices was presented by Lichtenberg et al. [134]. Here, the two platinum detection electrodes were made perpendicular to and in the same plane as the separation channel. Both detection electrodes were encircled by a grounded electrode for shielding. A 15 mm wide glass wall separated the microchannel and the electrodes. Electronics were designed in-house and optimal sensitivity was determined experimentally to be at an input frequency of 58 kHz and V pp of 15 V. Using field-amplified sample stacking (FASS), peak areas could be increased by up to a factor 4. FASS is achieved when the sample is present in buffer at a lower ionic strength than the BGE. When a high voltage is applied, the analyte species in the sample stack at the front end of the sample plug.
A much simpler device for CE-CCD was presented by Wang et al. [135], using external detection electrodes made of aluminum foil on top of the chip. A hot embossed PMMA device was sealed with a 125 mm thick foil. At 570 mm from the end of the separation channel, two strips of aluminum foil were glued to the top of the device to serve as a conductivity detector. The electrodes were placed in "anti-parallel" position, i.e., the electrode leads going to opposite sides of the device, to reduce interfering capacitance through the air. During the characterisa- Figure 12. Measurement setup for four-electrode CCD. An AC signal is applied to one outer electrode and the resulting current I 0 is measured at the outer electrode. Simultaneously, the potential difference V 0 between the two inner electrodes is measured. The solution conductivity is retrieved from the real-time division I 0 /V 0 . Reprinted from [132], with permission.
tion of the detector, a detection limit for K 1 (1.2 mM) was achieved when using a BGE with lower conductivity than that used for the determination of the other LODs. The device was used for monitoring inorganic ions in explosive residues [136], and for the detection of organophosphonate degradation products [137]. Later, a glass microdevice with identical detection configuration was used for nonaqueous CE of aliphatic amines [137]. Glass microdevices were also used for simultaneous CCD and amperometric detection [138]. Here, the separation channel ended in an in-house constructed reservoir with end-column amperometric detection, whereas the CCD was positioned just before the channel outlet. This microdevice allowed combination of the universal detection character of the CCD with the higher sensitivity and selectivity of the amperometric detector, giving additional value to having just one detector.
An important innovation in conductivity detection in microfluidic devices was the movable CCD described by Wang et al. [139]. A movable detector allows improved insights into the separation process since this process can be monitored at any desired position. The movable detector consisted of 800 mm wide strips of aluminum foil fixed to a PMMA plate using epoxy glue, the distance between the electrodes was also 800 mm. Similar to the detector described above, the electrodes were placed in an antiparallel configuration with a small overlap in the detection area. Both electrodes were wrapped around the PMMA plate to enable connection to the electronics from the top. The detector was clamped to the microfluidic device, and could be positioned at virtually any position by sliding the detector over the microdevice. A similar setup was also used for an enzyme assay of organophosphate nerve agents [140]. The movable CCD also allowed simultaneous detection of cations and anions in a single run [141]. For this purpose, the device layout was slightly modified as illustrated in Fig. 13, to accommodate two injection points at the opposite ends of the separation channel. The sample is simultaneously loaded and injected from the two ends of the separation channel. Once the separation voltage is applied, the cations migrate towards the cathode, and the anions migrate in the opposite direction towards the anode. Careful positioning of the detector was required to prevent overlap between the cation and anion peaks in the middle of the separation channel. A calibration plot was made simultaneously for ammonium chloride and sodium perchlorate for concentrations ranging from 400-2000 mM and 200-1000 mM, respectively.
After the success of CCD with increased peak-to-peak voltages (V pp ) in capillaries, Hauser and co-workers [142] transferred this technique to the microchip format. In the first paper, a glass microdevice was used. Electrodes for conductivity detection were fabricated on the outside of the device and consisted of strips of silver paint (0.5 mm wide and 5 mm long) perpendicular to the channel direction. A grounded electrode was placed in between the two detection electrodes to eliminate the capacitance through the air. Increased detection sensitivity was observed in devices where the electrodes were painted in recesses to reduce the distance between the electrodes ; and (f) ClO 4 2 (all 1 mM). Operation conditions: separation voltage, 1000 V; injection voltage, 500 V; injection time, 1 s; running buffer, 20 mM MES-20 mM His (pH 6.1); sinusoidal waveform with a frequency of 200 kHz and a peak-to-peak voltage of 10 V. The separation channel was 66 mm long, with the detector placed at 27 mm from the anode. Reprinted from [141], with permission. and the channel from 1 mm to 0.2 mm. Later, the silver paint for fabrication of the electrodes was replaced by copper foil, and the device used for the separation and detection of several organic ions [143].
Around the same time that the group of Wang presented their movable CCD, Tanyanyiwa et al. [144] placed the detection electrodes in the chip holder instead of on top or integrated into the microdevice. This procedure allowed the use of devices that only contained the microfluidic network, thereby significantly simplifying the microfabrication process and therefore reducing the costs-per-device. The embossed PMMA devices were covered with 100 mm thick PMMA foil and placed inside the chip holder. The detection electrodes consisted of 2 mm wide strips of copper foil, placed in parallel and leaving a 2.5 mm wide gap in between for the shielding electrode. When the microdevice was placed in the holder, the detection electrodes were positioned perpendicularly to the separation channel at 1 cm from the channel ending. Even though the detection limits achieved in this device were slightly higher (1.5 mM instead of 0.49 mM for K 1 , as indicated in Table 4), the simplification of the microfabrication process makes this paper a significant contribution to the literature.
Later, the performance of the PMMA devices was compared with fused-silica capillaries and glass microdevices for the separation of human immunoglobulin (IgG), all with CCD [145]. The detection limit for IgG achieved on the PMMA devices (34 ng/mL) was higher that the LOD obtained in capillaries (0.15 ng/mL). The difference in sensitivity between capillary and chip CCD was attributed to the less efficient electrode arrangement on the chips. In the capillaries, the detection electrodes are placed around the capillary, whereas on the chip there was only an electrode strip above the channel. The separation of free IgG and the IgG-complex took 6 min in the capillary and 3.5 min on the chip, illustrating the decreased analysis time on microfluidic devices. This was even more pronounced when using glass microdevices, where the analysis could be completed within 1 min.
A very simple microfabrication method was presented by Do Lago and co-workers [146]. Laser-printed traces on overhead transparency films were used for defining the microfluidic structure, and a lamination machine was used for sealing the device. The printed toner lines were 6-7 mm high and served as channel walls. The two parallel detection electrodes were made by attaching two strips of adhesive copper foil to the microdevice. The electronics were based on the work of the same group on CCD in capillaries. Using this device, the separation of 0.1 mM mixture of alkali metal ions was demonstrated.

Summary of major trends
A conductivity detector CE, like any detector, should meet several requirements. The detector should not distort the shape of the passing analyte zones, and it should have a high sensitivity and linear response over several orders of magnitude. In CE, the detector also needs to sense analyte zones separated in a strong electric field in a narrow-bore capillary. Integration of conductivity detection in CE therefore involves a challenging combination of designing and engineering issues. Different approaches presented in the literature to optimising the materials and design, the electronics and the chemistry are summarised below.

Materials
Conductivity detection in CE is presented in capillaries and on microfluidic devices, both in the contact and contactless mode. Capillaries are generally made of fused silica, and only have a limited flexibility with respect to the integration of detection electrodes. For the fabrication of microfluidic devices a wide variety of microfabrication techniques and materials is available. Initially, the microdevices were fabricated in glass or quartz, but a trend towards the use of polymeric devices can be observed. PMMA is the most frequently used polymer, but this probably more reflects its widespread availability than its favourable chemical and physical characteristics.

Design
Contact conductivity detection requires electrodes in contact with the solution, and has been presented "oncolumn" and "end-column". Both modes require labourintensive procedures for positioning of the detection electrodes in capillaries. CCD are easier to manufacture since they typically consist of an assembly of two tubular electrodes that slides around the capillary. In microfluidic devices, the flexibility in design allows the fabrication of intersecting channels to form cross or double-T injectors that function as a sample loop. The sample is generally loaded using electrokinetically driven flow, but dedicated systems for pressure-driven injections have also been presented. Computer-controlled injection systems are essential for the injection of small, reproducible sample zones.
In microfluidic devices, positioning of detection electrodes is more straight-forward than in capillaries. Even though manual insertion of detection electrodes has been presented, integration of electrodes using dedicated microfabrication techniques is preferred for the production of larger batches of microfluidic devices with integrated conductivity detector. For contact conductivity detection, electrodes were positioned perpendicularly to the separation channel, either in parallel configuration or facing each other. The facing configuration has been applied most frequently and has provided a more sensitive means of detection. For CCD, the use of a thin cover plate between the detection electrodes and the separation channel improved the capacitive coupling with the liquid inside the channel, and thereby the sensitivity of the detector. Detection electrodes were initially integrated into the microfluidic devices using metal deposition techniques and later mounted on top of the microfluidic devices. More recently, external, nonintegrated detection electrodes have been presented. The increased flexibility of positioning the electrodes in combination with the simplification of the manufacturing of microfluidic devices can be considered as an important step in the development of affordable, disposable miniaturised analytical systems.

Electronics
The interference of the separation field with the detection electronics can be eliminated electronically or physically. In the electronic approach, the signal can be isolated from ground by using a battery-powered detection system in combination with wireless infrared communication, or from its source by using capacitive coupling, a transformer or an optically coupled isolator. Physical elimination has been presented in two different ways. In end-column contact conductivity detection, the two detection electrodes are positioned in the grounded reservoir, thereby preventing exposure to the high separation voltage. In CCD, no direct contact of the detection electrodes with the solution is required. A less sensitive approach is not to eliminate but to measure the local potential/potential drop resulting from the separation field in the capillary/microchannel using passive conductivity detection techniques.
For active conductivity detection in CE, an AC signal is applied to one detection electrode, and measurement of the resulting current through the cell is made using the other detection electrode. The current through the conductivity cell should reflect the fluid resistance and therefore the conductivity of the liquid inside the detector. However, in AC circuits the capacitive reactance of the different capacitors present in the detection cell contributes to the cell impedance. Since the magnitude of the capacitive reactance depends on the applied frequency, its influence on the cell impedance is frequency-dependent. Capacitive reactances in series and in parallel with the fluid resistance result in unreliable and unpredictable measurements of the fluid resistance at certain frequencies.
In contact conductivity detection, the double layer capacitance of the electrode is considered to be the largest capacitance contributing to the cell impedance. Since the electrode capacitance and the resistance of the liquid are connected in series, they can be added to give the cell impedance. The capacitive reactance decreases with increasing measurement frequency, which makes the capacitive reactance insignificant with respect to the measured fluid resistance above a certain frequency. A more elegant way to eliminate all capacitors in the measurement circuit is the use of a bipolar pulse instead of an AC signal. The bipolar pulse charges and discharges the capacitors, without drawing a current. The only restrictor of current in the cell is the fluid resistance, which is the parameter of interest. The current through the cell therefore reflects the fluid resistance and not the capacitive reactances in the cell.
In CCD, the capacitance of air between the two electrodes is connected in parallel with the resistance of the solution, and the capacitance of the capillary or channel wall is connected in series with the solution resistance. The influence of capacitive reactance of the air is observed at high frequency and can easily be eliminated using a grounded shielding electrode between the two measurement electrodes. The influence of the capacitive reactance of the wall can be corrected for using a fourelectrode CCD. Here, frequency-independent measurements can be performed up to a frequency where liquid capacitance, connected in parallel with the resistance of the liquid, starts playing a role.

Electrolyte composition
Conductivity detection in CE involves an irreconcilable conflict due to the linear relationship between mobility and specific conductivity. For optimal separation efficiency the mobility of the analyte and the background electrolyte co-ion should match, and for optimal detection sensitivity, the difference in conductivity of the analyte and background co-ion should be maximised. Traditionally, this issue has been solved by the use of low conductivity, amphoteric buffers at relatively high concentrations. However, a more elegant solution is the use of a chemical suppressor prior to detection. A suppressor allows the reduction of the background conductivity and an increase in the conductivity of the analyte zone at the same time. Two independent research groups have worked on the development of a suppressed contact conductivity detector for CE but the fabrication process of this detector was too complicated to be commercialised.

Outlook
Conductivity detection in CE has been presented in the contact and contactless mode in capillaries and in microfluidic devices. The recent introduction of a commercially available contactless conductivity detector for capillaries with an outer diameter , 400 mm will result in an increase in the number of applications of contactless conductivity detection in CE. Currently, no preference or trend towards contact or contactless conductivity in microfluidic devices can be observed.
In contact conductivity detection, the influence of capacitors can be minimised by the use of a bipolar pulse instead of a continuous AC signal. Pulsed conductivity detection can be expected to replace the continuous AC signal in contact conductivity detection, both in capillaries and in microfluidic devices. In contactless conductivity detection, the impedance measurements for determination of the resistance of the solution can be corrupted by the capacitive reactance of the air between the two electrodes, or the by the capacitive reactance of the capillary or channel wall. The use of a shielding electrode is an effective measure to eliminate the capacitive reactance of the air between the detection electrodes and is therefore expected to be used predominantly in the future. A four-electrode detection setup can be used for elimination of the capacitance of the wall that is in series with the resistance of the solution. This, however, involves a more complicated detector design and the impedance measurements are influenced by the capacitive reactance of the liquid at high frequencies. Improvements in CCD can be expected through alternative electronic solutions to minimise the influence of capacitive reactances connected both in series and in parallel with the solution resistance.
The intrinsic issue in conductivity detection in CE is the difference in requirements for the background electrolyte for optimum separation efficiency and optimum detection sensitivity. As discussed above, use of a chemical suppressor prior to conductivity detection is an attractive option and could be significantly simplified by using a contactless conductivity detector around the capillary rather than the contact conductivity methods employed thus far. Furthermore, the high precision and flexibility of microfabrication technology should be exploited to produce a microfluidic device with suppressed conductivity detection. Developments in these research areas are expected shortly.