Detection Limits of Thin Layer Coulometry with Ionophore Based Ion

Aug 24, 2012 - For the winding of the PVDF tread, we build a computer controlled (CNC) machine schematically shown in Figure 1B. The machine was fully...
0 downloads 12 Views 1MB Size
Download PDF
Article pubs.acs.org/ac

Detection Limits of Thin Layer Coulometry with Ionophore Based Ion-Selective Membranes Alexey Shvarev, Bastien Neel, and Eric Bakker* Department of Inorganic and Analytical Chemistry, University of Geneva, Quai Ernest-Ansermet 30, CH-1211 Geneva, Switzerland S Supporting Information *

ABSTRACT: We report here on a significant improvement in lowering the low detection limit of thin layer coulometric sensors based on liquid ion-selective membranes, using a potassium-selective system as a model example. Various possible processes that may result in an elevated residual current reading after electrolysis were eliminated. Selfdissolution of AgCl on the Ag/AgCl inner element may result in a residual ion flux that could adversely affect the lower detection limit. It was here replaced with an Ag/AgI inner pseudoreference electrode where the self-dissolution equilibrium is largely suppressed. Possible residual currents originating from a direct contact between inner element and ion-selective membranes were eliminated by introducing an inert PVDF separator of 50 μm diameter that was coiled around the inner element by a custom-made instrument. Finally, the influence of electrolyte fluxes from the outer solution across the membrane into the sample was evaluated by altering its lipophilic nature and reducing its concentration. It was found that this last effect is most likely responsible for the observed residual current for the potassium-selective membranes studied here. For the optimized conditions, the calibration curves demonstrated a near zero intercept, thereby paving the way to the coulometric calibration-free sensing of ionic species. A linear calibration curve for the coulometric cell with valinomycin potassium-selective membrane was obtained in the range of 100 nM to 10 μM potassium in the presence of a 10 μM sodium background. In the presence of a higher (100 μM) concentration of sodium, a reliable detection of 1−100 μM of potassium was achieved.

O

of the sequence was a regeneration of the membrane at the OCP. It was shown that the charge difference measured at the first and the second potential pulses represents the total amount of the analyte with the corresponding linear calibration curve with the origin close to zero.5,6 Thus, the possibility of the calibration-free detection was confirmed. The lowest detected concentration was 10 μM for calcium and potassium with the corresponding cation-selective membranes. This paper aims at understanding the chemical origin for the relatively large nonspecific charge that required the abovementioned second pulse to be applied. A reduction of nonspecific processes and residual currents after electrolysis would allow one to develop chemical sensors that would work more reliably without recalibration and that exhibit lower detection limits. Moreover, the efforts described herein aim at improving the general cell design and to make the assembly more optimal and compatible with standard fluidic connectors. Specifically, we postulate that nonspecific signal may originate from four possible processes: (1) The inner element used in previous work was based on a chloridized silver wire, which is known to exhibit an ion product dissolution constant of about Ksp = 5 × 10−10 at room temperature.9 Especially in

n the basis of earlier work on ion-transfer chronocoulometric detection, 1−3 a series of papers recently introduced the concept of controlled potential thin layer coulometry with ionophore-based ion-selective membranes.4−7 This new technique can potentially lead to the development of a family of calibration-free ion-selective sensors where the selectivity is achieved by the ion carrier (ionophore) in the membrane. Indeed, coulometry as being an absolute method based on Faraday’s law does not require the reference standards. The experimental setup4 consisted of a coaxial arrangement of an inner Ag/AgCl wire electrode and the porous polypropylene tube that was used to support the liquid ionselective membrane. The 50 μm gap between the electrode and the membrane formed a sample chamber in contact with a relatively large membrane area (∼250 mm2). The high surface to volume ratio of the sample compartment and its small thickness were necessary to complete the electrolysis in a relatively short time (within several tens of seconds).8 The measurement protocol5,6 consists of the following sequence. First, the open circuit potential (OCP) was determined. Subsequently, a controlled potential pulse of a fixed duration was applied ideally leading to a complete electrolysis of the cationic or anionic analyte. The potential was raised with respect to the OCP. Then, in order to eliminate an undesired buildup of polarization of the membrane, the latter was held at zero current. After that a second potential pulse of the same duration and magnitude was imposed. The final part © 2012 American Chemical Society

Received: July 11, 2012 Accepted: August 24, 2012 Published: August 24, 2012 8038

dx.doi.org/10.1021/ac301940n | Anal. Chem. 2012, 84, 8038−8044

Analytical Chemistry

Article

Figure 1. (A) Scaled up drawing of the 0.4 mm silver wire electrode with the fabricated PVDC monofilament separator and the porous polypropylene membrane, which were used to assemble the coulometric cell; (B) 3D sketch of the CNC machine in the process of electrode separator fabrication; and (C) scaled up cutaway view of the assembled cell.

(KTpClPB), potassium ionophore I (Valinomycin), high molecular weight poly(vinyl chloride) (PVC), tetrahydrofuran (THF), silver wire of 0.5 mm diameter (99.9% trace metals basis), PEEK tubing and 1/8 in. and 1/16 in. HPLC connectors, and all salts were purchased in the highest quality available from Sigma Aldrich. Aqueous solutions were prepared by dissolving the appropriate salts in Milli-Q-purified distilled water. Electrode Fabrication. The electrode manufacturing process started with the 0.5 mm silver wire, which was drawn through a commercially available jewelry tungsten carbide drawplate to gradually reduce the diameter. One of the electrode tips was glued into a 2 cm PEEK tube (o.d. × i.d.; 1.5 mm × 0.75 mm). In order to introduce the separator between the electrode and the membrane, we developed a central electrode with a polymeric thread wound in a tight spiral around it. The tread was a monofilament PVDF fishline (Milo s.r.l., Milan) of the smallest commercially available diameter. Following preliminary trials, these dimensions were selected: 0.4 mm for the inner electrode and 0.05 mm for the thread wound with 0.15 mm pitch. The scaled up electrode assembly is shown in Figure 1A. In this work, we started with 130 mm long electrodes and finally were able to fabricate them up to 800 mm long. For the winding of the PVDF tread, we build a computer controlled (CNC) machine schematically shown in Figure 1B. The machine was fully automated and controlled by a

the absence of a chloride background, self-dissolution of AgCl will result in a non-negligible level of dissociated AgCl in the sample solution that may be at the origin of an undesired residual current. (2) The inner element may be in physical contact with the membrane, since no spacing material is present. Anodic applied potentials may result in the oxidation of membrane components (including tetraphenylborates) or result in the transport of ions between the inner element and outer solution. Either process will result in a parasitic current that is unrelated to the thin layer coulometric measurement. (3) The outer solution composition is relatively elevated with respect to the sample composition and must contain the salt of the analyte ion to ensure a stable membrane potential. As previously described for zero-current potentiometric sensors, a concentration mismatch between the outer and inner solution will result in an electrolyte flux in the direction of the more dilute solution. Consequently, this process may be also at the origin for residual currents after thin layer electrolysis. (4) One of the possible origins previously discussed was the presence of undesired non-Faradaic processes such as capacitive currents.5 This paper aims to eliminate each of these individual contributions in order to find the optimal working conditions and cell design for thin layer ion-selective coulometry.

■

EXPERIMENTAL METHODS Materials and Chemicals. Dodecyl 2-nitrophenyl ether (DDNPE), tetradodecylammonium tetrakis(4-chlorophenyl)borate ETH 500, potassium tetrakis(4-chlorophenyl)borate 8039

dx.doi.org/10.1021/ac301940n | Anal. Chem. 2012, 84, 8038−8044

Analytical Chemistry

Article

Prior to the experiments, the cell was immersed into the solution of 10−4 M KCl and was washed for 1 day by continuously pumping the very same solution at a flow rate of 10 μL min−1. Electrochemical Measurements. The electrochemical experiments were carried out using PGSTAT 101 (Metrohm Autolab, Utrecht, The Netherlands) controlled by a personal computer with Nova 1.8 software. The duration of the electrolysis was 60 s for both raw electrolysis and the background compensation pulse. The solutions were fed into the cell using the ISMATEC peristaltic pump (Glattbrugg, Switzerland) at a flow rate of 50 μL min−1. We used both threeelectrode and two-electrode setups in both of which the coulometric cell served as the working electrode. Depending on the setup, a combination of a gold counter electrode and a double-junction Ag/AgCl/3 M KCl/1 M LiOAc reference electrode (Mettler-Toledo AG, Schwerzenbach, Switzerland) or a single high surface area Ag/AgCl wire electrode were used in the outer aqueous solution. All measurements were performed at room temperature (23 °C). Impedance spectroscopy was carried on with a PGSTAT 302N (Metrohm Autolab, Utrecht, The Netherlands) controlled by a personal computer with Nova 1.8 software.

microcontroller board (Parallax Inc., Rocklin, CA) along with the PC running a LabVIEW application as a high-level interface. Two prototypes were built. The first one was limited to a maximum electrode length of 140 mm making the maximum active length of the membrane below 100 mm. The final (longer) version consisted of two stepper motors rotating the electrode wire stretched between two mini-chucks and a carriage with a filament guide and the bobbin moving along the electrode. The carriage was driven by the third stepper motor and a 1000 mm long precision ball screw (16 mm × 5 mm) typically used for the CNC machine tools. The tension of the filament was controlled by a modified standard RC servo. The servo shaft drove the bobbin and the servo potentiometer (replaced by multirevolution 5k precision wire potentiometer) was connected to a spring loaded dancing arm, which stretched the filament. Increased tension caused the arm movement, and the feedback caused the servo to slowly unload the thread from the bobbin. The tension of the filament was kept constant at 12 ± 2 g. The tips of the PVDF filament were glued to the electrode with epoxy and cyanoacrylate adhesives. After fabrication the electrode was removed and cut to the desired length. Coulometric Cell. In order to ensure the compatibility with standard fluidics, the cell design described previously4 was improved by fitting it with a standard PEEK tube. The polypropylene hollow fiber (Accurel PPQ3/2, Membrana GmbH, Wuppertal, Germany) was cut to the specific length (100−400 mm), and the ends were fitted with two pieces of 60 mm × 0.6 mm silver wire to prevent the membrane from clogging or collapsing in the process of attachment of the connectors. Both ends of the tubing were glued with the epoxy resin into 40 mm long pieces of a PEEK tubing (o.d. × i.d. = 1/ 8 in. × 1/16 in.). After setting of the epoxy, the silver wire was removed and the ends of the PEEK tubing were cut flat with a tubing cutter. The electrode with the separator was carefully inserted into the membrane (Figure 1A,C) forming a sample chamber. The scaled up cutaway view of the assembled cell is shown in Figure 1C. Theoretically the sample layer thickness would correspond to 0.1 mm but in reality, due to the variations in the hollow fiber inner diameter, was smaller. The tips of the membrane and the electrode fitted with the PEEK tube allowed us to use standard HPLC 1/4 in. nuts with 1/8 in. and 1/16 in. ferrules. A Plexiglas T-connector with 1/4 in. thread was used to assembly the T-shaped cell similar to as described in the earlier work.4 The electrode wire was cleaned with methanol and sequentially etched for 1 min in 10% nitric acid and ammonia solutions followed by rinsing in deionized water. An anodic constant current of 10 mA per each 100 mm length was applied for 30 min to do the silver halogenation. For AgI and AgCl coatings, the 10 mM solutions of HCl and KI were used correspondingly. The DDNPE formed a liquid membrane matrix that filled the pores of the polypropylene hollow fiber. The membrane cocktail was composed of 444 mg of DDNPE, 5.5 mg (10 mmol/kg) of valinomycin, and 0.7 mg of KTpClPB (25 mol % relative to the ionophore). Unless otherwise specified, 50 mg (10 wt-%) of ETH 500 were added. The components were dissolved in DDNPE, and ∼100 μL of cocktail was cast on a membrane in the fully assembled cell. Excess of the liquid membrane was wiped with a paper tissue.

■

RESULTS AND DISCUSSION In order to evaluate the coulometric system, a direct comparison of the potassium selective coulometric experiments described recently was performed.6 The composition of the outer solution (1 mM KCl and 10 mM LiCl) and the sample (with 10 mM LiCl background) as well as the potential window were at first identical. The active length of the membrane was 1.5 times shorter (90 mm vs 140 mm) due to our initial inability to fabricate longer electrodes with the separator. In contrast to previous observations,6 the potentiometric response function for potassium was found to be linear within 10−150 μM of KCl with a 57.1 mV slope (Figure 2C, insert). The residual currents at the end of 60 s electrolysis pulses were approximately 36% lower than in the prior experiments. This corresponds to the reduction of the cell length and suggests that the residual currents are proportional to the cell surface. The resulting charges for both pulses along with the difference are plotted on Figure 2 as a function of potassium concentration. At 120 mV applied potential for the first (electrolysis) pulse, the intercept for the linear fit was 19 μC vs 41 μC measured earlier. We also observed clean minima (critical points) on the first pulse curves at low concentrations at all potentials in the window. After charge subtraction of the background (correction) pulse, the 120 mV calibration curve yielded the intercept at 9.2 ± 2.1 μC in agreement with previous observations (7.1 ± 3.8 μC). Taking into account all of the similarities, the “sharper” response in the low concentration range can be attributed to better-defined cell geometry, in other words, more uniform sample layer thickness. It is interesting to note that the elimination of a direct contact between the membrane and inner electrode did essentially not cause a reduction in the charge measured during application of the second pulse. First, this means that the direct contact caused no charge transfer parasitic process, as initially feared. As it can be seen on the graphs (Figure 2), a decreasing potassium concentration below 25 μM causes the concentration−charge curves for the first pulse to level off with subsequent increasing of the charge measured in the second 8040

dx.doi.org/10.1021/ac301940n | Anal. Chem. 2012, 84, 8038−8044

Analytical Chemistry

Article

contamination of the sample with silver and halide ions. In addition, the treatment of the Ag/AgI electrode as an iodideselective electrode would result in high selectivity coefficients with respect to other halides, specifically negligible interference from the chloride present in the sample solution.14 The evaluation of the Ag/AgI inner electrode cell started using conditions similar to the ones described above for the silver chloride cell. However, it became quickly apparent that reaching lower detection limits is feasible only with more dilute outer solutions of KCl (0.1 mM vs 10 mM in previous experiments). The attempt to replace KCl with KI in the outer solution caused further significant deterioration of the cell response. This was likely due to much higher lipophilicity of the latter and substantial coextration and transmembrane transport of KI, thereby causing sample contamination. The results of a direct comparison of two cells with slightly different geometry under similar conditions are shown in Figure 3. In both cases, the solution contained potassium (25 μM for

Figure 3. Observed current decays for electrolysis (1st pulse) and background compensation pulses (2nd pulse) in the cells with Ag/ AgCl (at 120 mV vs OCP) and Ag/AgI (at 100 mV vs OCP) inner electrodes within the 0−20 s time frame. The insert shows the current decays for the last half of the pulse, between 30 and 60 s. In both cases the solution contained potassium (25 μM for KCl and 30 μM for KI) and sodium as a background cation, both with corresponding halide anion (1 mM). The outer solution contained 0.1 mM KCl. The differences of integrated charges between the first and second pulses were 29.8 and 23.8 μC for KCl and KI correspondingly.

Figure 2. (A) Calibration curves with respect to K+ obtained at a 1 min electrolysis time for raw electrolysis pulses, (B) backgroundcompensation pulses, and (C) background-corrected signals recorded for five different excitation potentials 40, 80, 120, 160, and 200 mV vs open circuit potential in the cell with Ag/AgCl inner electrode. The sample solution contains 10−150 μM KCl and 10 mM LiCl as a background. The outer solution contains 0.1 mM KCl. The insert in part C shows the potentiometric cell response in 10−150 μM KCl + 10 mM LiCl sample solution.

KCl and 30 μM for KI) and sodium as a background cation, both with the corresponding halide anion (1 mM) and the outer solution contained 0.1 mM KCl. The applied potential of 120 mV for KCl chloride was slightly higher than 100 mV in the case of KI (both were the closest taken from 40 and 50 mV potential grids, which were used in the corresponding experiments). The differences of integrated charges between the first and second pulses were 29.8 and 23.8 μC for KCl and KI, correspondingly, which was in good agreement with the calculated theoretical charge of 30.1 and 25.1 μC for given potassium concentrations and cell geometries (the “iodide” cell was shorter). Despite the very similar charges there was a striking difference in the magnitudes of the current decay of both first and second pulses as well as residual current and the end of the electrolysis. The latter are shown in the insert. The

pulse. Our attempts to work in the lower concentration range indicated that the calibration curves become inconsistent below 10 μM. Two suspected causes, possible sample contamination and subsequent parasitic processes, were investigated. Lowering of the KCl concentration in the outside solution from 10 mM to 0.1 mM as well as removal of the inert electrolyte ETH 500 from the membrane had little effect on the cell response. In an attempt to utilize a better, yet rather academic, inner electrode, an Ag/AgI pseudoreference system was evaluated. In contrast to AgCl, the silver iodide has a much lower solubility product:9 8.3 × 10−17 vs 1.8 × 10−10. The silver iodide electrode was successfully used as a pseudoreference for the detection with ion-selective electrodes,10−12 and long-term stability of the AgI film was confirmed with an X-ray photoelectron spectroscopy studies.13 Lower solubility product corresponds to lower 8041

dx.doi.org/10.1021/ac301940n | Anal. Chem. 2012, 84, 8038−8044

Analytical Chemistry

Article

integrated charge for the background pulse went down from 20.9 μC vs 0.78 μC, corresponding to a more than 25-fold decrease. Moreover, at given concentrations of potassium the “chloride” cell showed about 50% contribution of the background signal vs 5.2% for the “iodide” cell. Such a drastic improvement in the signal to background ratio lead us to the conclusion that lowering the detection limit for potassium with the Ag/AgI inner electrode was feasible. As mentioned earlier, the main cause of the background current and charge reduction should be attributed to much lower solubility product, thus, much lower activities of silver and iodide in the sample and much smaller sample contamination at the end electrolysis. However, as mentioned above, increasing the sensitivity of the Ag/AgI cell also made it sensitive to the composition of the outer solution. A number of experiments were performed to find the optimal composition. Specifically, we studied the outer solutions containing potassium salts (iodide, sulfate, and chloride) within the concentration range of 10 mM−10 μM. The presence of potassium was necessary for maintaining the equilibrium at the outer membrane interface. The least lipophilic counteranion (chloride, sulfate) was preferred leading to a lower background signal. On one hand, lowering the concentration of the outer electrolyte 100-fold from 10 to 0.1 mM produced an improvement of the detection limit. The results of a systematic study of the electrolysis of 10 μM KI in the presence of 0.1 mM NaI yield a charge close to theoretical in the 0.1−0.3 mM outer solution concentration range. A further increase caused the measured charge to increase with an apparent loss of membrane selectivity. On the other hand, further concentration reduction resulted in a higher solution resistance, and below 0.1 mM the cell response clearly started to deteriorate. No improvement of lowering the background signal was observed, instead the noise increased drastically. The attempts to bring the outer electrode closer to the coulometric cell by coiling it around with various spacers (usually pieces of the PEEK tubing) were unsuccessful. Therefore 0.1 mM KCl was selected as an optimal outer solution. The iodide cell charge vs concentration curves in the range of 1−100 μM KI in the presence of 0.1 mM NaI background is shown in Figure 4. The potentiometric response was checked first to ensure proper functioning of the cell (Figure 4B, the insert). This yielded a linear calibration curve with 62.1 mV/ decade, a slightly super-Nernstian slope, in the range from 1 to 20 μM KI with NaI background. The minor deviation in slope is likely due to an increase in I− activity from 100 to 120 μM since the potential of the inner Ag/AgI element directly depends on the activity of iodide in the sample. The upper and middle graphs of Figure 4 contain the information about the first and the second pulses, and the graph at the bottom shows the charge difference vs concentration of potassium iodide. Because of the low level of the background charges, a different scale was used on the graphs. The potential window was chosen between 40 and 120 mV relative to OCP, the latter theoretically corresponding to 99% electrolysis. The curves for the first pulse clearly show the minima (∼5 μM) shifted toward a lower concentration range in contrast to chloride curves with the minima ∼25−30 μM. The charges measured for the second pulse are at least 2-fold lower than those for the “chloride” cell. The charge difference (Figure 4C) between the first and second pulse yielded linear calibration curves in the case of 100 and 120 mV applied potential relative to OCP within the whole

Figure 4. Experiment in the “iodide” cell: (A) calibration curves with respect to K+ obtained at a 1 min electrolysis time for raw electrolysis pulses, (B) background-compensation pulses, and (C) backgroundcorrected signals for five different excitation potentials 40, 60, 80, 100, and 120 mV vs OCP in the cell with an Ag/AgI inner electrode. The sample solution contains 1−100 μM KI + 0.1 mM NaI. The outer solution contains 0.1 mM KCl. The insert in part B shows a potentiometric calibration curve in 1−20 μM KI + 0.1 mM NaI solution.

concentration range between 1 and 100 μM. Lower potentials become nonlinear as the concentration increases, which clearly indicated the incomplete electrolysis. For the potentials 100 and 120 mV, one obtained 0.91 (±0.06) μC/μM slope with a −0.81 (±0.95) μC intercept and 0.92 (±0.03) μC/μM slope with a −0.24 (±0.50) μC intercept, correspondingly. The experimental values for the slopes are practically equal to theoretical ones and indicate complete electrolysis. In addition, both are in good agreement with the theoretical value of 0.94 μC/μM calculated for this cell. The intercepts were close to zero and practically an order of magnitude lower than those measured for the “chloride cell”. Despite these improvements, both coulometric setups share the same behavior in the low concentration range. The integrated charge measured during the electrolysis pulse 8042

dx.doi.org/10.1021/ac301940n | Anal. Chem. 2012, 84, 8038−8044

Analytical Chemistry

Article

that could have indicated the presence of a charge transfer process at the membrane. On the basis of available data15 for both AgCl and AgI inner electrodes, a simple calculation yields a double layer capacitance of ∼20 μF (10 μF × 2 cm2). In combination with the observed cell resistance of ∼2 kΩ, the corresponding time constant is ∼40 ms. Using “5RC the rule of thumb”, we should expect to charge all of the capacitors to 99% of the total charge within a 0.2 s time interval. This value is more than 2 orders of magnitude smaller compared to the time scale of the electrolysis (60 s) and represents a very fast transition process. In order to evaluate the possible contribution of the nonfaradaic charging current to the total charge, we calculated the amount of charge measured within the first 0.2 s of the applied potential pulse. In both chloride and iodide experiments, the possible fraction of the nonfaradaic process may contribute to 2.1−4.3% of the measured charge, which may be practically relevant. However, with respect to the background compensation pulse this amount lies within 3−6% of the measured background charge leading to the conclusion that the nonfaradaic process may contribute only a small fraction to the background compensation charge.

(Figures 2A and 4A) decreases with the concentration of the primary ion (potassium), the curve passes a minimum, and further reduction of concentration causes a significant amplification of the detected charge due to the increasing contribution of the background signal (Figures 2B and 4B). This charge bump can be easily explained if we take into account that the applied potential is not adjusted with the concentration. The same potential that leads to a fairly efficient electrolysis in high and medium (with respect to the background) concentration range causes a selectivity breakdown at small concentrations of primary ion and subsequent transport of the background (interfering) ion. In order to push the detection limit further, we prepared the membrane without adding the inert lipophilic electrolyte ETH 500 to the ion-selective membrane. A number of preliminary observations showed that in the absence of ETH 500, the residual currents are somewhat smaller. Unfortunately, this came at the price of noisier current signals due to an increasing membrane resistance. The concentration of the background was reduced to 10 μM NaI. The resulting charge vs concentration curves at 120 mV applied potential are plotted in Figure 5. The charge difference despite significant

■

CONCLUSIONS We have developed several improvements for the ion-selective membrane coulometric detection system introduced earlier.4−7 The fabrication of an inert separator between the inner electrode and the membrane proved to be a challenging but doable task. Despite only a marginal improvement of the cell response for cationic species with respect to the background signal, this approach may be the key to the development of anion-selective coulometric sensors. This particular application is currently under investigation in our group. The evaluation of the experimental parameters on lowering the detection limit showed that the inner electrode chemistry is of utmost importance. The replacement of Ag/AgCl with the Ag/AgI pseudoreference system allowed us to reduce the charge detected for the background pulse by a factor of more than 25. In addition, a 100-fold lowering of the ionic strength of the outer solution resulted in an improvement of the lower detection limit for the cell with silver iodide inner electrode. The elimination of the large amount of the inert lipophilic electrolyte ETH 500 had only a minor effect on the reduction of the background charge and residual currents. We demonstrated that with these improvements a linear calibration curve for valinomycin-based potassium-selective membrane can be obtained in the range of 0.1−10 μM of potassium in the presence of 10 μM sodium background. In the presence of a higher (100 μM) concentration of sodium, it was possible to detect potassium in the range of 1−100 μM. For all experiments with the silver iodide inner electrode and new cell design, the intercept of the calibration curve after subtracting the background charge was very close to zero, suggesting a practically feasible calibration-free measurement. The observed impedance spectroscopy data suggest little or no contribution of the nonfaradaic processes to the formation of the background signal. The time frame of the electrolysis is 3 orders of magnitude longer than the expected time for the capacitive currents. It is perhaps possible to envision a “self-regulating” coulometric cell in which the open-circuit potential value corresponds to a complete absence of the primary ion in the presence of an interfering ion in a background. This is perhaps

Figure 5. (A) Calibration curves with respect to K+ obtained for the first electrolysis pulse, (B) background-compensation pulse, and (C) background-corrected signal at 120 mV vs OCP in the cell with the Ag/AgI inner electrode. The membrane does not contain ETH 500. The sample solution contains 0.1−10 μM KI + 10 μM NaI. The outer solution contains 0.1 mM KCl.

contribution of the background signal was found linear within 0.1−10 μM KI range with the slope of 1.18 (±0.11) μC/μM slope with a 1.17 (±0.96) μC intercept. The corresponding theoretical slope was 1.07 μC/μM. The non-Faradic processes such as capacitive currents, in our opinion, should be less likely the cause of the background signal. At first, our impedance measurements within 1−105 Hz (see the Supporting Information, Figure 1S) with the coulometric cells exhibited only a combination of a single semicircle at high frequencies and diffusion impedance in the low-frequency domain. In a typical experiment (sample solution contains 0.1 mM KI + NaI and the outer solution is 0.1 mM KCl), the fitting of the semicircle using the Boukamp model (Figure 1S, Supporting Information) yielded a solution resistance of ∼378 Ω and the cell resistance of 1.78 kΩ with the corresponding capacitance of 0.37 nF. The latter is way below the value one may expect for the double layer capacitance and most likely represent the geometrical capacitance of the cell. We did not observe a second semicircle 8043

dx.doi.org/10.1021/ac301940n | Anal. Chem. 2012, 84, 8038−8044

Analytical Chemistry

Article

most easily achievable with a flow-through cell, such as in flow injection analysis. The proposed electrode fabrication technique will allow one to produce sufficiently long coulometric cells to reach the required residence time. Such experiments are currently in progress in our group.

■

ASSOCIATED CONTENT

S Supporting Information *

Additional information as noted in text. This material is available free of charge via the Internet at http://pubs.acs.org.

■

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS The authors thank the Swiss National Science Foundation and the CSIRO, through the Flagship Cluster “Sensors Systems for Analysis of Aquatic Environments”, for financial support of this research. This work also forms part of a project funded by the Australian Research Council (Grant DP0987851).

■

REFERENCES

(1) Sawada, S.; Taguma, M.; Kimoto, T.; Hotta, H.; Osakai, T. Anal. Chem. 2002, 74, 1177. (2) Yoshizumi, A.; Uehara, A.; Kasuno, M.; Kitatsuhi, Y.; Yoshida, Z.; Kihara, S. J. Electroanal. Chem. 2005, 581, 275. (3) Lee, H. J.; Girault, H. H. Anal. Chem. 1998, 70, 4280. (4) Grygolowicz-Pawlak, E.; Bakker, E. Anal. Chem. 2010, 82, 4537. (5) Grygolowicz-Pawlak, E.; Bakker, E. Electrochem. Commun. 2010, 12, 1195. (6) Grygolowicz-Pawlak, E.; Bakker, E. Electrochim. Acta 2011, 56, 10359. (7) Grygolowicz-Pawlak, E.; Numnuam, A.; Thavarungkul, P.; Kanatharana, P.; Bakker, E. Anal. Chem. 2012, 84, 1327. (8) Bakker, E. Anal. Chem. 2011, 83, 486. (9) O’Neil, M. J.; et al. The Merck Index - An Encyclopedia of Chemicals, Drugs, and Biologicals, 14th ed.; Merck Sharp & Dohme Corp.: Whitehouse Station, NJ, 2006. (10) Bratten, C. D. T.; Cobbold, P. H.; Cooper, J. M. Anal. Chem. 1997, 69, 253. (11) Chiba, K.; Tsunoda, K.; Umezawa, Y.; Haraguchi, H.; Fujiwara, S.; Fuwa, K. Anal. Chem. 1980, 52, 596. (12) Hashimoto, M.; Upadhyay, S.; Kojima, S.; Suzuki, H.; Hayashi, K.; Sunagawa, K. J. Electrochem. Soc. 2006, 153, H155. (13) Strydom, C. A.; van Staden, J. F.; Strydom, H. J. Electroanalysis 1992, 4, 969. (14) Morf, W. E. The Principles of Ion-Selective Electrodes and of Membrane Transport; Elsevier: New York, 1981. (15) Bousse, L. J.; Bergveld, P.; Geeraedts, H. J. M. Sens. Actuators 1986, 9, 179.

8044

dx.doi.org/10.1021/ac301940n | Anal. Chem. 2012, 84, 8038−8044