Extending the Lifetime of the Running Electrolyte in Capillary

Chem. , 2005, 77 (2), pp 607–614. DOI: 10.1021/ .... Unmanned platform for long-range remote analysis of volatile compounds in air samples. Eric T. ...
0 downloads 0 Views 157KB Size
Anal. Chem. 2005, 77, 607-614

Extending the Lifetime of the Running Electrolyte in Capillary Electrophoresis by Using Additional Compartments for External Electrolysis Dosil Pereira de Jesus, Jose´ Geraldo Alves Brito-Neto, Eduardo Mathias Richter, Lu´cio Angnes, Ivano Gebhardt Rolf Gutz, and Claudimir Lucio do Lago*

Departamento de Quı´mica Fundamental, Instituto de Quı´mica, Universidade de Sa˜o Paulo, Av. Prof. Lineu Prestes 748, CEP 05508-900, Sa˜o Paulo-SP, Brazil

The use of two additional reservoirs to accommodate the electrodes of the power source is proposed to improve the stability of the running electrolyte in capillary electrophoresis. The basic idea is to use salt bridges to connect those reservoirs to the ones containing the capillary ends. Although simple, there are several issues that can be considered in the design and implementation of such system in order to prevent undesired transference of material between the electrolysis and the main reservoirs. The use of a sealed electrolysis reservoir without a gas phase, the use of materials that ensure volume stability, and the use of bridges as long as possible are three basic directions. A compromise is involved in the dimensions of the sectional area of the bridge, because a small area diminishes the amount of a species transferred by diffusion but leads to an undesirable increase of the electrical field during the electrophoretic running. Thus, a bridge composed of a main wide-bore tube connected to a small-bore capillary seems to give the best performance for practical use. A simple electrolysis-separated system was adapted to a preexisting capillary electrophoresis system, and its performance was evaluated with a mixture of tartaric, malic, and succinic acids that was separated in sodium benzoate solution (pH 5.5) using the original equipment and the modified one. Due to the water electrolysis and the small buffering capacity of the electrolyte, there was a significant pH change and consequently changes in the effective mobilities of the analytes and loss of resolution after a few runs using the original equipment. Using the electrolysis-separated system, no significant change in the migration time and resolution was observed even after 15 runs. Besides the freedom to prepare running electrolytes with electroactive species or unbuffered solution, high throughput and the use of small reservoirs, such as the ones used in microfluidic devices, are the main advantages of the system. Capillary electrophoresis (CE) is now a well-established analytical technique and has been applied to the determination of a great variety of analytes in an assortment of matrixes. * To whom correspondence should be addressed. E-mail: claudemi@ iq.usp.br. 10.1021/ac0486645 CCC: $30.25 Published on Web 12/03/2004

© 2005 American Chemical Society

Although initially the experiments were done in adapted or homemade equipment, there are several commercial options. The setup depends on the kind of application and mainly the detection system, but one can identify some common features. Although the migration of the species inside the capillary column is the most important phenomenon for this analytical separation technique, one can clearly identify the equipment as an electrolytic cell. Thus, the current required to promote migration and electroosmotic flow (EOF) in the capillary is driven by heterogeneous electron transference at the electrode/ electrolyte interfaces, and this electrolysis is accompanied by unavoidable modification in the composition of the solutions in both semicells. Without physical barriers, the degraded electrolyte will eventually propagate into the capillary, gradually affecting the precision or even dramatically changing the performance of the electrophoretic separation. The effects of this electrolysis on the electrophoresis have been studied.1-3 The main concern is the water electrolysis, which has a direct influence on the pH. The running electrolyte pH is a very important parameter,4 because the mobilities of most species, as well as the EOF, depend on it. The electrolytic formation and consumption of H+ is compensated by using buffer solutions as electrolyte. Although it is a common practice in CE, this approach has at least two limitations: (1) the electrolyte composition is restricted to buffered systems, and (2) other electroactive species may be involved in the electrochemical reactions not being compensated by the buffer. A straightforward solution to this problem is to use reservoirs as large as possible with electrodes as far as possible from the inlets of the capillary to dilute the electrolysis products and prevent their entrance in the capillary. However, this is an inappropriate solution if one takes into account the size of the equipment, the consumption of reagents, the price of some buffers, and the “green chemistry” concepts of minimization of residues. Indeed, the interference of electrolysis becomes even more significant in micro total analysis systems (µ-TAS) or lab-on-a-chip, where reservoirs have only a few microliters or less.5 Thus, detailed studies and (1) Macka, M.; Andersson, P.; Haddad, P. R. Anal. Chem. 1998, 70, 743-749. (2) Carson, S.; Cohen, A. S.; Belenkii, A.; Ruiz-Martinez, M. C.; Berka, J.; Karger, B. L. Anal. Chem. 1993, 65, 3219-3226. (3) Fuller, R.; Sweedler, J. V. Anal. Chem. 1999, 71, 4014-4022. (4) Kenndler, E.; Friedl, W. J. Chromatrogr., A 1992, 608, 161-170.

Analytical Chemistry, Vol. 77, No. 2, January 15, 2005 607

trolysis reservoir), and the end of the capillary is inserted in the other (main reservoir). The geometry and size of the reservoirs and the bridge as well as details of implementation are discussed.

Figure 1. Simplified diagram of a conventional capillary electrophoresis equipment (A) and the modified one (B). ER, electrolysis reservoir; MR, main reservoir; vm, MR volume; ve, ER volume, vg, volume of the gas-phase inside ER, h, difference of liquid levels between ER and MR; a and l, sectional area and length of the bridge between MR and ER, respectively.

new approaches to minimize the effect of electrolysis are necessary. Haddad and co-workers1 carried out detailed experiments about the positioning of the electrode and the capillary inlet, monitoring the pH using indicators. Although some guidelines were drawn about the correct positioning, the authors emphasized the necessity to keep the electrolyte buffered. Oki and co-workers6 have used an ion-selective field effect transistor to measure the pH inside the reservoirs of a µ-TAS device and proposed two methods to reduce pH changes: the use of a salt bridge to isolate the electrolysis region and neutralization by addition of a different pH solution by EOF pumping. Those elaborate approaches allowed pH control in a 2-µL reservoir during experiments in electrophoresis conditions for several minutes. A simpler solution was adopted by Karger and co-workers2 to reduce the residual fluorescence originated by the oxygen produced electrolytically: the electrode and the reservoir were continually washed with fresh running electrolyte. Although a modest flow rate (0.2 mL min-1) was used, a considerable volume is consumed in experiments lasting more than 1 h. The electrolysis is also a problem during electrokinetic injection, when even the analytes can be electrolyzed. To prevent such a disturbance, some authors suggest that the sample vial should be used only once,7 impracticable when a low volume of sample is available and replicates are needed. Fuller and Sweedler3 studied the electrokinetic injection from nanolitervolume conductive vials and concluded that electrolysis makes quantitation problematic. In this paper, a simple approach to minimize the effects of electrolysis on the electrophoretic separation is presented. The basic idea is shown in Figure 1 and consists of the splitting each of the original reservoirs into two compartments with a salt bridge between them. The electrolysis is developed in one vial (elec(5) Reyes, D. R.; Iossifidis, D.; Auroux, P. A.; Manz, A. Anal. Chem. 2002, 74, 2623-2636. (6) Oki, A.; Takamura, Y.; Ito, Y.; Horiike, Y. Electrophoresis 2002, 23, 28602864. (7) Zhang, C.-X.; Thormann, W. Anal. Chem. 1996, 68, 2523-2532.

608 Analytical Chemistry, Vol. 77, No. 2, January 15, 2005

EXPERIMENTAL SECTION Reagents and Solutions. All reagents were of analytical grade, and deionized water (Nanopure-UV, Barnsted, Dubuque, IA) was used to prepare the solutions. The chromate running buffer was prepared as described in the literature.8 A solution containing 5 mmol L-1 chromium trioxide was titrated with 20 mmol L-1 tris(hydroxymethyl)aminoethane (Tris) resulting in a solution with pH 8.5. Afterward, 0.2 mmol L-1 N-cetyl-N,N,Ntrimethylammonium bromide (CTAB) was added as an EOF modifier. Instrumentation. The electrochemical studies were performed using a potentiostat PGSTAT 20 Autolab (Eco Chemie B.V., Ultrecht, Netherlands). UV-visible spectra were obtained with a diode array spectrophotometer HP 8452A (Hewlett-Packard, Palo Alto, CA). Electrophoresis experiments were carried out in modified equipment with contactless conductivity detection, which has been described elsewhere.9,10 Each piece of equipment was interfaced to a microcomputer. RESULTS AND DISCUSSION To demonstrate the vulnerability of the conventional electrochemical cell used in CE, the chromate running buffer was chosen as a model. This is a very common electrolyte used for anion analysis with indirect UV-visible detection,11,12 and it is prone to developing chemical transformation besides water electrolysis. The CE cell was emulated with two vials interconnected by a 1-cm-long silica capillary with a 250-µm i.d. and filled with 1 mL of chromate buffer solution. Platinum wire electrodes were placed in both vials, and a reference electrode (Ag|AgCl|KClsat) was inserted in one of the vials, taken as a working electrode semicell. Electrolysis was conducted galvanostatically, under a constant current of 30 µA. The chosen capillary was much thicker and shorter than usual in CE, to keep the experiment within the (10-V compliance of the galvanostat (no need of high-voltage supply). Two independent experiments were carried out to emulate the anode and cathode reservoirs, and the results are shown in Figure 2. The initial voltage of both experiments (not shown) referenced to Ag|AgCl|KClsat electrode should be the same, because the semicell is exactly the same before electrolysis. However, a few millivolts difference is observed due to the impurities and differences in the conditioning of the electrode surface. After the electrolysis starts, the cathode drifts toward negative potentials, while the anode goes to positive potentials. This drift occurs because the region around the electrode becomes poor in the electroactive species involved in the electron-transfer process and a greater potential is necessary to keep the current. The shapes of the curves suggest that successive processes take place during a few hours of electrolysis. (8) Doble, P.; Macka, M.; Andersson, P.; Haddad, P. R. Anal. Commun. 1997, 34, 351-353. (9) Fracassi da Silva, J. A.; do Lago, C. L. Anal. Chem. 1998, 70, 4339-4343. (10) Fracassi da Silva, J. A.; Guzman, N.; do Lago, C. L. J. Chromatogr., A 2002, 942, 249-258. (11) Jandik, P.; Jones, W. R. J. Chromatogr. 1991, 546, 431-443. (12) Horie, H.; Yamauchi, Y.; Kohata, K. J. Chromatogr. 1998, 817, 139-144.

The following reactions at the cathode should be considered:

CrO42- + 4H2O + 3e- f Cr3+ + 8OHCr3+ + e- f Cr2+ O2(g) + 2H2O + 4e- f 4OH2H2O + 2e- f 2OH- + H2(g) The extent to which each reaction contributes to the electrolysis current depends mainly on the formal reduction potential (including overpotentials), the diffusion coefficients (in the absence of convection), and the concentrations of the species on the electrode. It seems that the first three reactions dominate the experiment, because no gas evolution was observed. After the Cr(VI) concentration at the electrode decays to levels insufficient to sustain the current, the potential drift accelerates until another process sets in simultaneously. This seems to occur after some minutes, with the formation of Cr(II), a species that will chemically react and deplete dissolved oxygen near the electrode, with regeneration of Cr(III). Independently of the relative contribution of the reactions, three of them contribute to the elevation of the pH. Thus, sufficient buffering capacity of Tris is essential to keep the pH nearly constant. It is necessary to realize that the first reaction consumes the chromate, a fundamental species, to allow indirect UV detection. The problem is augmented with reversed EOF that pumps the electrolyte from the cathode reservoir into the capillary. Similar reasoning could be done on the reaction at the anode, where bromine, oxygen, and H+ are formed. A consequence of the pH reduction is the formation of HCrO4-, which precipitates with CTA+ (used for the reversion of the EOF).8 Although it is undesirable, the precipitate does not tend to clog the column, because the EOF goes from the cathode toward the anode. However, it is possible to observe the precipitate inside the reservoir after a few hours of electrolysis. The curves from Figure 2 show that, to keep the desired current, the overall electrolytic cell starts with 0.2 V and drifts to ∼2 V after several minutes. Similar results are observed for other aqueous running electrolytes. Thus, from the several-kilovolt potential difference used in electrophoresis practice, only a few volts are spent to drive the electron-transfer processes at the electrodes. This is why the electrophoresis setup behaves basically as an ohmic conductor with the resistance determined by the column geometry and the resistivity of the running electrolyte. After 4 h of electrolysis, the solutions from both reservoirs were homogenized, 10-times diluted, and their UV-visible spectra were recorded. The spectra of these solutions were similar to the fresh solution one but for the intensities. This reveals that the concentration of CrO42- has changed, which is expected for the cathode solution. However, there was an increase of CrO42concentration in the anode reservoir, which may be understood by considering the migration of this anion from the catholyte to the anolyte compartment. Due to the concentrations of the ionic species in the system, TrisH+ and CrO42- are the main charge carriers. For each CrO42ion reduced to Cr3+, 3 electrons are consumed, and each CrO42-

Figure 2. Chronopotentiograms for the chromate running buffer in constant-current mode (30 µA): (a) anodic and (b) cathodic forms. Working and auxiliary electrodes, platinum. Reference electrode, Ag|AgCl|KClsat.

ion transferred to the anode reservoir carries 2 electrons. Based on the charge balance, it can be demonstrated that the ratio of the variation of CrO42- concentrations at the anode and cathode reservoirs is equal to

∆CA p )∆CC p + 2/3

(1)

where p is the CrO42- transport number. Considering that the mobilities of TrisH+ and CrO42- in the buffer are 8.4 × 10-9 and -81 × 10-9 m2 V-1 s-1, respectively, their transport numbers are approximately 0.17 and 0.83, respectively. Thus, the expected ratio is -0.55, which is in good agreement with the ratio calculated (-0.57) based on the variations of absorbances at 370-380 nm. This shows that there is a significant variation of CrO42- concentration along the capillary and that it is not due only to electrochemical transformation but also to charge transport. The presented example shows that control of pH is only one of the requirements to ensure good reproducibility in electrophoresis, because other transformations may occur besides or instead of pH changes. The idea introduced in the present paper is to confine the electrolysis products in separate compartments, electrolytically connected to the running electrolyte reservoirs in which the capillary ends are immersed. Several implementations of the basic idea were tested, and Figure 1 shows a pictorial representation that summarizes them. It is clear that the bridge should be as long as possible in order to isolate both reservoirs, because the electrolysis products will diffuse to the main reservoir to a low extent in this case. However, several other aspects and parameters should be considered before a new design is proposed. The first aspect to be considered is the composition of the electrolytes. Most of the time, the same running electrolyte can be used to fill the reservoirs and the bridge. Thus, unless otherwise stated, the considerations and experiments described in this paper are based on the use of a single electrolyte. Another important aspect is the possibility of hydrodynamic pumping (siphoning) due to the different liquid levels in the Analytical Chemistry, Vol. 77, No. 2, January 15, 2005

609

reservoirs (h in Figure 1). If the main reservoir has a higher liquid level, the problem is not so dramatic, because its solution will be transferred to the electrolysis reservoir, preventing diffusion from undesired species to the electrophoresis stage. However, if h is positive, the contamination of the running electrolyte in the main reservoir will be accelerated. Of course, if h is large enough, a reservoir may be completely drained or an overflow may occur. The most convenient solution found was to use a closed electrolysis reservoir. In this case, the solution will not be freely pumped from one side to the other. Nevertheless, due to the compressibility of the gases, pumping may occur in a significant extent if an air layer or bubbles remain in the electrolysis reservoir. The geometry of the bridge is another important issue. As previously pointed out, long bridges diminish the extent of contamination by diffusion. Narrow-bore bridges also limit the contamination, because of the small amount of material transported by diffusion. However, in both cases, the electric resistance of the bridge will be increased, which may be deleterious if this resistance becomes comparable to the capillary one. In that case, the effective electric field inside the capillary and the separation performance are diminished. Due to the different situations and ways to implement a separated electrolysis system, it is not possible to define fixed shapes and sizes. A general formulation and typical and extreme cases are presented instead. Diffusion and Migration-Related Issues. Disregarding the bulk motion of the liquid phase due to thermal expansion or pressure gradients, the phenomena that can account for the transfer of material from the electrolysis reservoir to the main reservoir are diffusion and electrophoretic migration through the bridge. Diffusion occurs wherever there is a concentration gradient and migration cannot be disregarded because there is a significant voltage drop across the bridge. In this section, we make some theoretical considerations about the extent of contamination of the main reservoir due to a concentration perturbation developed in the electrolysis reservoir that propagates through the bridge under diffusion- and migration-controlled transport. Only a one-dimensional concentration distribution across the bridge is calculated; thus, time evolution of the system is described by the following equation:

∂2c ∂c ∂c ) D 2 - µE ∂t ∂x ∂x

(2)

where D is the diffusion coefficient of the chemical species under consideration, µ is its electrophoretic mobility, and E is the electric field. The first term on the right-hand side of eq 2 corresponds to diffusion as given by Fick’s law, and the second term is the migration contribution. The electric field is considered uniform throughout the bridge. In the present work, we decided to model a worst-case scenario; thus, we have used a hypothetical species A+ with the same mobility and diffusion coefficient as H+, because it is the fastest species in aqueous medium. However, to simplify the simulation, no chemical equilibrium was considered. The left-hand reservoir (x ) 0) is the electrolysis reservoir, and when the electric field is positive, it drags the species from the electrolysis reservoir toward the main reservoir. Also in accordance with the worst610

Analytical Chemistry, Vol. 77, No. 2, January 15, 2005

Figure 3. Simulation of the steady-state concentration profile along some 10-mm-long bridges between two reservoirs: the left-hand one containing 1 mmol/L A+ solution and the right-hand one containing only the background electrolyte. Curves: (a) diffusion only; (b) i.d. 5 mm bridge (0.04 V/cm); (c) i.d. 5 mm bridge (-0.04 V/cm); (d) i.d. 3 mm bridge (0.11 V/cm); (e) i.d. 3 mm bridge (-0.11 V/cm); (f) i.d. 0.5 mm bridge (4.00 V/cm); (g) i.d. 0.5 mm bridge (-4.00 V/cm).

case assumption, we consider that both reservoirs have uniform concentrations all the time; i.e., there is some kind of highly effective homogenization system in them. Mathematically speaking, Dirichlet boundary conditions on the concentration distribution are enforced. At the left-hand extremity it is considered equal to a given value and at the right-hand one its value is zero. After a long time of electrolysis, the concentration distribution along the bridge reaches a steady state, which depends on the velocity of migration, the diffusion coefficient, and the length of the bridge. The flux of the contaminant species through the bridge from this point on can be calculated from this concentration distribution; thus, some useful information about the long-time behavior of the system can be obtained simply analyzing the steady-state solution of eq 2. Such solution is found by solving the ordinary differential equation obtained when one sets the time derivative term in eq 2 to zero. For the boundary conditions described above, the closed-form solution is

{

c(x) ) c0 1 -

}

1 - exv/D 1 - elv/D

(3)

where c0 is the concentration at the electrolysis reservoir, l is the bridge length, v is the velocity of the ion inside the bridge, and D is its diffusion coefficient. Figure 3 shows the concentration profile along some 10-mmlong bridges. Curve a shows the steady-state profile considering only the diffusion from a 1 mmol/L solution. This diffusion accounts for the transference of A+ in the absence of electric field, i.e., during the time that the equipment rests. Of course, in this case, the longer the tube the smaller the amount of A+ that is transferred. For the simulations of the bridge during electrophoresis experiments, the electric field through bridges of different inner diameters was calculated considering that it was connected in series with a 50-µm-i.d. and 50-cm-long separation capillary under 400 V/cm electric field. Curve b in Figure 3 is the A+ concentration

profile during such an electrophoretic running for a wide-bore (i.d. 5 mm) bridge. The electric field inside the bridge is small (0.04 V/cm), but it becomes an additional stimulus for the transference of A+. Curve c shows the profile of the same species in the cathode bridge. In this case, the electric field acts against the diffusion and diminishes the rate of transference of A+. Curves d and e show the profiles for a 3-mm-i.d. tube (E ) 0.11 V/cm). For the smallest bore tube (0.5 mm) (curves f and g), the electric field (4.00 V/cm) is the most important factor to the transference. In the steady-state condition, the concentration is almost equal to the one at the electrolysis reservoir until very near the right-hand end. The bridge behaves like a simple hole that connects both reservoirs. Of course, this extreme situation occurs for the considered conditions, but real cases may approach this undesirable situation. Another important parameter that can be obtained from the steady-state solution of eq 2 is the asymptotic rate of transference of A+ through the bridge. This is equal to the flux of A+ times the cross-sectional area of the bridge. The very definition of the steady state requires that the flux of A+ be the same anywhere in the bridge. For convenience, the asymptotic rate of transference (r) can be calculated at the right-hand end of the tube, where A+ concentration is equal to zero, as

r ) - Da

(∂c∂x)

) x)l

ac0v 1 - e-lv/D

(4)

where a is the cross-sectional area of the tube. At this point, it is interesting to make more explicit the dependence of the asymptotic rate of material transfer with the bridge inner diameter. To compare different inner diameter bridges under realistic conditions, we decided to consider that the electric field applied to the separation capillary is always the same. It is a trivial matter to deduce that, under this assumption, the velocity of the A+ species inside the bridge is given by v ) µEcdc2/d2, where Ec is the electric field applied to the separation capillary and dc is its inner diameter. The cross-sectional area is given by a ) πd2/4, thus, substituting those expressions into eq 4, one obtains

r)

c0πµEcdc2 1 -lµEcdc2/Dd2 4 1-e

(5)

That is, the d2 factors from the area and the velocity cancel each other out, so that the only remaining dependence on d is in the exponent in the denominator. The conclusion that can be drawn from these considerations is that for small-bore tubes (high velocities), for positive Ec, the exponential in the denominator goes to zero and the rate of mass transfer is almost independent of the inner diameter. However, as the inner diameter grows, the exponential factor in the denominator becomes larger (but always less than 1), and the asymptotic rate of mass transfer also grows. That is, for large-bore tubes, despite the velocity through the bridge being small, the increase in the rate of mass transfer due to the large cross-sectional area outgrows the decrease in the rate due to the small velocity. To evaluate the amount of A+ transferred through the bridge as a function of time, a numerical approach was necessary. The

Figure 4. Simulation of the transferred amounts of A+ for six different bridges: (a) 5-mm i.d. and 100 mm long; (b) 3-mm i.d. and 100 mm long; (c) 0.5-mm i.d. and 100 mm long; (d) 5-mm i.d. and 10 mm long; (e) 3-mm i.d. and 10 mm long; (f) 0.5-mm i.d. and 10 mm long. In all these cases, the electric field in the 50-µm-i.d. and 50cm-long separation capillary was maintained constant (400 V/cm). The inset shows the initial portion of the main graph.

spatial derivatives in eq 2 were approximated by central finite differences and the time evolution of the discretely sampled concentration distribution was calculated using the fourth-order Runge-Kutta method. The initial concentration distribution was considered zero everywhere except at the left-hand extremity of the bridge. Throughout the integration, the material flux at the right-hand extremity of the bridge was calculated and recorded, as well as the integrated amount of material transferred. The simulation program was implemented in ANSI C, and its source code can be obtained from the authors upon request. Figure 4 shows the results of simulation of the transferred amounts of A+ for six different bridges: all the combinations of short (10 mm) and long (100 mm) tubes with three different inner diameters (0.5, 3, and 5 mm). In all these cases, the electric field in the 50-µm-i.d. and 50-cm-long separation capillary was kept the same (400 V/cm). The curves b-f show that 10-7-10-6 mol of A+ are transferred after several hours (more than 10 h). For a 1-mL reservoir, these amounts increase the concentration of A+ by 0.1-1 mmol/L, which is not a negligible fluctuation in several practical cases. Taking into account the 100-mm-long tubes (curves a-c), one can conclude that the wider the tube, the better the resulting bridge. However, considering the 10-mm-long tubes of equivalent diameters (curves d-f), an important feature can be observed. Although the wide-bore tube is submitted to a lower electric field, which diminishes the velocity of the ionic species, the sectional area is large, which implies a large amount of the species is delivered at the end of the tube. Thus, for a short period of usage, wide-bore tubes have better performance (curves d-f). However, for long periods, a wide-bore tube (curve f) may be inappropriate. A general rule for the choice of the sectional area of the bridge states that small area is suitable for the rest condition, while large area tends to be more appropriate for the running condition. On the other hand, the bridge should be as long as possible. Of course, another point is that the larger the area and length of the bridge, the larger will be the volume of electrolyte consumed. Analytical Chemistry, Vol. 77, No. 2, January 15, 2005

611

Pressure-Related Issues. It is possible to design systems with a minimum of gas phase, which practically eliminates the problem of gas compression or expansion and the resulting flow of liquid. However, in some cases, a residual amount of air remains or gas may be produced by electrolysis. To evaluate the extent of liquid transfer between the reservoirs, the gas inside the electrolysis reservoir may be considered ideal and the transferred volume (∆v) may be estimated by

(P -∆p∆p)

∆v ) vg

(6)

where vg is the volume of the gas phase, P is the atmospheric pressure, and ∆p is the difference of pressure given by ∆p ) Fgh, where F is the density of the liquid, g the gravity acceleration, and h is the difference of the liquid levels (Figure 1). Considering the electrolyte solution as water at ambient pressure and temperature, one can note that, for example, at a residual gas phase of 1 mL and a 10-cm level difference between the reservoirs (h), the liquid flow will draw 9.8 µL from one reservoir to the other one. This may be seen as a low volume if the sample vial has ∼1 mL of solution (less than 1% of the total volume). However, for small sample volume (less than 100 µL), the flow may cause a significant contamination or dilution. Additional flow occurs if flexible materials are used as electrolysis reservoir and bridge. Thus, rigid materials should be the first choice. Temperature-Related Issues. Another problem related to the gas phase is the thermal expansion or compression. Considering an ideal gas and for little temperature fluctuations (∆T), the volume of liquid drawn to or from the electrolysis reservoir (∆v) may be estimated by

∆v ) (vg/T)∆T

(7)

where T is the ambient temperature. At normal conditions, for a 1-mL gas phase, a fluctuation of 1 K leads to 3.4 µL of liquid being exchanged between the reservoirs. Again, to keep the transference of liquid below 1%, the volume of liquid in the main reservoir should be at least 1/3 of the gas phase in the electrolysis reservoir for ∆T ) 1 K; the factor rises to 10/3 for ∆T ) 10 K, and so on. Since large temperature fluctuations during sampling or running are not common, thermal expansion of the gas phase seems to be a minor problem if the gas phase is kept at a minimum. Although thermal expansion of liquids occurs in lesser extent than for gases, the volume of solution inside an electrolytic reservoir plus bridge may be quite big, which may be a problem. The volume variation (∆v) is given by

∆v ) ve(eR∆T - 1)

(8)

where R is the volumetric expansion coefficient of the liquid and ve is the volume of the electrolysis reservoir. Considering the volumetric expansion coefficient of water 2.1 × 10-4 K-1 and a 10-mL electrolysis reservoir, 2.1 µL will be transferred to the main reservoir for ∆T ) 1 K. For small temperature fluctuations, a good rule of thumb is to use an electrolyte reservoir with volume smaller than 100 times the main reservoir. 612

Analytical Chemistry, Vol. 77, No. 2, January 15, 2005

Figure 5. Diagram of the electrolysis compartment: (a) 1-mL syringe barrel; (b) 0.5-mm-diameter platinum wire; (c) rubber of syringe plunger; (d) high-voltage cable; (e) Luer-lock fitting; (f) PVC tube (3-mm i.d. and 50 cm long); (h) silicone tube bushing; (g) fusedsilica capillary tip (7 cm long, 250-µm i.d., and 350-µm o.d.).

Simple and Versatile Implementation of a Four-Compartment CE System. The implementation shown in Figure 5 does not meet all the recommendations stated in the previous sections, but it proved to be useful for evaluation of the concepts and versatile enough to be adapted to conventional equipments. Its strong and weak aspects will be timely commented. The system is basically a disposable 1-mL syringe adapted to a 50-cm-long 3-mm-i.d. PVC tube. The rubber of the original syringe plunger was adapted to a 4-mm high-voltage wire that is connected to the high-voltage source. The 0.5-mm-diameter platinum (or other metal) electrode perforates the plunger rubber, which prevents liquid leakage. A short piece of silica capillary (7 cm long, 250-µm i.d., and 350-µm o.d.) was adapted to the extremity of the salt bridge, with the help of a bushing made from short pieces of silicone tubing. The reservoir and bridge are easily filled with electrolyte by suction using the high-voltage cable as the plunger of the syringe. Although the system has a long tube as the bridge, only the tip has a small bore; thus, the restriction to flow is small. Alternatively, the syringe may be disconnected from the tube for a direct draw. Several experiments were carried out with this arrangement. The flexibility of the PVC tube allows one to introduce the silica tip in the main reservoir like a metal electrode, while the electrolysis reservoir lays in a safe place. To prevent volume variations, the plunger must be immobilized. Impressive results obtained with the same CE equipment operated with two and four compartments are presented in Figure 6. A mixture of 200 µmol L-1 tartaric (pKa ) 3.036 and 4.366), malic (pKa ) 3.46 and 5.05), and succinic (pKa ) 4.207 and 5.638) acids was analyzed in 20 mmol L-1 benzoic acid/sodium benzoate (pH 5.5) running electrolyte. This electrolyte was prepared by starting with benzoic acid (pKa ) 4.203) solution and adjusting the pH to 5.5 with NaOH solution. CTAB, 200 µmol L-1, was used as EOF modifier. It is worthwhile to note that the mono- and divalent ionic species related to the three analytes have similar mobilities. However, due to the differences in their pKa values, baseline separation is possible at an appropriate pH. Benzoate was chosen because of its low mobility, which gives a good sensitivity for conductivity detection. However, at pH 5.5, this electrolyte has a low base buffering capacity, because its pH is ∼1 unit above the pKa of the benzoic acid. Figure 6a shows the electropherograms obtained with the conventional equipment without renewal of the running electrolyte. The volume of the solution in the reservoirs is 1 mL. The first run gives baseline separation, but the resolution between the peaks is very low at the ninth run. This may be explained by taking into account that the EOF systematically pumps the electrolyte from the cathode reservoir, which has the pH augmented by water

Figure 6. Electropherograms (the first and the last of a series) and resolution versus the run number for separation of 200 µmol L-1 tartrate (1), malate (2), and succinate (3) in the conventional CE equipment (a, c) and with the four-compartment one (b, d). Resolutions between tartrate and malate and malate and succinate peaks are represented by solid squares and open circles, respectively. Running electrolyte was 20 mmol L-1 benzoic acid/sodium benzoate (pH 5.5) and 0.2 mmol L-1 CTAB. Potential of separation 10 kV; hydrodynamic injection at 10 cm for 20 s; silica capillary with 75-µm i.d. and 48 cm long (40 cm effective); contactless conductivity detection at 600 kHz. The dashed lines in (c) and (d) indicate the minimum resolution for baseline separation (1.5).

electrolysis. After some runs, the effective pH inside the capillary is significantly greater than the initial value (5.5) and all the analytes have similar mobilities, because they are completely dissociated. Figure 6c shows the degradation of resolution at successive runs. The same experiment was done using the four-compartment system, and the results are presented in Figure 6b and d. Even after 15 runs, no resolution degradation was observed. Considering that each complete running takes ∼15 min, this experiment represents 4 h of electrophoresis without renewal of the weakly buffered electrolyte. The bridge used in this experiment is not as simple as those ones presented in the previous section. In fact, it is a composed bridge: a long and wide-bore PVC tube coupled to a short silica capillary. Taking into account the results from the previous simulations, one can note that small-bore tubes have better performance for diffusion-only conditions. However, wide-bore tubes are better suited when migration is also important. Thus, we decided to use a composed bridge, which several simulations suggested to have a good compromise between both features. The use of a capillary tip at the end of the bridge has two additional advantages. First, the capillary tip can have dimensions

similar to those of a platinum electrode, which simplifies the adaptation to a preexisting equipment. Second, during sampling, the small area of a capillary bridge minimizes the extent of sample cross-contamination. A disadvantage of the use of a capillary tip is the increase of the voltage drop on the bridge. For example, the calculated voltage drop on the experimental composed bridges is ∼2.7% of the voltage on the separation capillary. If only the 3-mm-i.d. tube was used, the drop would be 0.13%. Of course, although a significant reduction of the drop would be expected, in both cases only a small fraction of the high-voltage power is wasted, i.e., 29.2 kV of a 30-kV power source would still be available for electrophoresis by using the proposed bridge. CONCLUSIONS Although most CE practitioners pay close attention to the control of the running electrolyte pH, other issues must be carefully considered: electrochemical reactions other than water electrolysis and depletion/enhancement of species by charge transfer. In the example shown (chromate buffer), about half of chromate consumption in the cathode reservoir is due to the electrolysis process; the other half is simply transferred to anode reservoir. The external electrolysis system does not suffer from Analytical Chemistry, Vol. 77, No. 2, January 15, 2005

613

this kind of problem, because the bridge replenishes or drains the species in the main reservoir at the same rates of transfer through the separation capillary. Of course, for extreme cases (long time, high current, and low volumes), this equilibrium may be broken. Several directions should be considered for the design of an electrolysis-separated system, but some of them are somewhat conflictive. For example, a large electrolysis reservoir diminishes the impact of electrolysis and thus extends its lifetime. However, the impact of the temperature fluctuations on the transfer of solution between the reservoirs increases with this volume. On the other hand, some directions seem to be always appropriate, e.g., the elimination of the gas phase in the sealed electrolysis reservoir and the use, when possible, of a composed bridge. Of course there are disadvantages. Increases of the complexity of the setup for electrophoresis, as well as the time expended for its preparation, are undesirable. In fact, for a small set of analyses,

614

Analytical Chemistry, Vol. 77, No. 2, January 15, 2005

perhaps the time consumed does not justify the effort necessary for the modifications. However, there are at least three important reasons to consider an electrolysis-separated system: higher throughput is achievable; freedom to use electrolytes that are unbuffered and contain electroactive species; and very small main reservoirs can be used, e.g., those ones from microfluidic devices. ACKNOWLEDGMENT The authors thank Fundac¸ ˜ao de Amparo a` Pesquisa do Estado de Sa˜o Paulo (FAPESP) and Conselho Nacional de Desenvolvimento Cientı´fico e Tecnolo´gico (CNPq) for the fellowships and Dr. Z. G. Richter for the English revision.

Received for review September 8, 2004. Accepted October 22, 2004. AC0486645