ANALYST, JANUARY 1992, VOL. 117 47 Flow Injection Amperometric Determination of Cyanide on a Modified Silver Electrode SneZana D. Nikolic and Emil 6. Milosavljevic Faculty of Chemistry, University of Belgrade, P.O. Box 550, I1001 Belgrade, Yugoslavia James L. Hendrix and John H. Nelson Departments of Chemistry and Chemical and Metallurgical Engineering, Macka y School of Mines, University of Nevada, Reno, NV 89557, USA A composite coating based on a mixture of phosphatidylcholine, cholesterol and stearic acid was used for fabricating a silver-based amperometric sensor. The sensor was used in a flow injection (FI) configuration for the selective and sensitive determination of cyanide. The method is based on the permselectivity of the lipid membrane, which rejects from the surface of the working silver electrode the undesired, potentially interfering species, while allowing the transport of the analyte as hydrogen cyanide. Logarithmic calibration graphs were linear up t o the maximum concentration of CN- investigated (0.100 mmol dm-3).The precision of the technique was better than a relative standard deviation of 3.5% at the 0.5 Vmol dm-3 level and better than 2% at the 0.1 mmol dm-3 level, with a throughput of 30 samples h-1. Under optimum conditions, the detection limit was 0.1 pmol dm-3 (0.26 ng CN-). The effects of working potential, type of electrode coating, acidity of the reagent stream and interferents on the FI signals were studied. Keywords: Flow injection; amperometric detection; modified silver electrode; cyanide determination Amperometric detection for flowing streams based on a silver working electrode is inherently sensitive.However, it is also non-selective, as any species that forms either an insoluble silver salt or stable complex with Ag+ would necessarily interfere. Hence, in order to solve a particular analytical problem, such detection has to be coupled with a suitable separation step. Rocklin and Johnson' developed a method for the determination of cyanide, sulfide, iodide and bromide utilizing ion chromatography and electrochemical detection via a silver working electrode. More recently, the same group2 developed an automated system for the determination of cyanide in water samples. Two in-series separation steps were employed. The analyte (as HCN) is separated from most interferents using gas diffusion and then from the remaining interferents, such as sulfide, by ion-exchange chromatography and is detected by an amperometric detector with a silver working electrode.Flow injection (FI) methods based on the same detection principle have also been successfully used for cyanide3-5 and sulfide determination.6 However, either continuous distillation3 or gas diffusion44 pre-separation steps had to be employed. On the other hand, lipid membranes, based on stearic or palmitic acid or phospholipids, have previously been used for modifying carbon electrode~7-~~) to be employed under static conditions. As was pointed out recently by Wang and Lu," a major obstacle to the development of lipid-based flow detectors had been the poor mechanical stability of the lipid layer.However, they developed a highly stable composite phospholipid-cholesterol electrode coating for the glassy carbon electrode for amperometric monitoring of hydrophobic substances in flowing streams. Under the FI regime, they employed the developed amperometric sensor for the determination of hydrophobic drugs (e.g., promethazine and trimipramine) while nearly eliminating interferences from ascorbic acid and tyrosine. The same group" described an interesting and innovative approach for simultaneous analyses using FI. They used a sensor array of several glassy carbon amperometric electrodes, each coated with a different permselective film, one of which was a phospholipid. Successful multicomponent analyses of neurologically significant catechol compounds were performed using pattern recognition.The technique is based on the partial but different selectivity of each sensor. This paper describes a novel approach for the selective and sensitive FI determination of CN-. An in situ separation step is incorporated in a relatively simple FI manifold by modifying the silver electrode with a phosphatidylcholine-cholesterol- stearic acid coating. The lipid composite membrane acts as a barrier for ionic species while retaining permittivity for the hydrophobic HCN. This work also demonstrates the possibil- ity of modifying electrodes that participate directly in the amperometric detection (anodic dissolution of silver in the presence of the analyte), as opposed to those which are only a medium for the electron transfer (glassy carbon or platinum electrodes).Experimental Reagents and Materials All chemicals were of analytical-reagent grade. The aqueous reagent and standard solutions were stored in glass containers. De-ionized water was used throughout. A stock solution of 0.1 mol dm-3 CN- was made from potassium cyanide (Merck, Darmstadt, Germany) and checked using the Liebig method. 13 Standard cyanide solutions, which in most experi- ments were made to be 1 mmol dm-3 in NaOH, were prepared by diluting aliquots of the stock solution to the appropriate volume. i2-cx-Phosphatidylcholine type XI-E (Sigma, St. Louis, MO, USA) and stearic _acid (Merck) were used as received. Cholesterol (Zorka, Sabac, Yugoslavia) was re- crystallized from chloroform. Instrumentation and Apparatus The FI manifold is illustrated in Fig.1. Two peristaltic pumps were used, Mini S-840 (Ismatec, Zurich, Switzerland) and a Model HPB 5400 (Iskra, Kranj, Yugoslavia). The injection valve was a Model 5020 (Rheodyne, Cotati, CA, USA) equipped with a 100 1-11 sample loop. Interconnecting poly- (tetrafluoroethylene)(PTFE) tubing had an i.d. of 0.5 mm. [Caution: to prevent HCN evolution, the waste lines should be connected to the NaOH solution.] A thin-layer flow-through amperometric cell was part of an LC-17A package (BAS, West Lafayette, IN, USA) and was equipped with a Model MF-1008 silver working electrode (BAS) and a Model MW-2021 Ag-AgCI reference electrode (BAS). A 0.13 mm48 ANALYST, JANUARY 1992, VOL. 117 I RE I II P I I W Fig. 1 FI manifold used for amperometric determination of cyanide: C, carrier (water); R, reagent (pH 8 borate buffer); P, peristaltic pump; I .injection valve; MC, mixing coil (30 cm X 0.5 mm i.d.); EC, electrochemical flow-through cell; PO, potentiostat; RE, recorder; and W, waste. Flow rates are given in ml min-l thick Model MF-1047 PTFE gasket (BAS) was used to separate the working electrode from the cell body. Before coating the electrode, the silver surface was polished with polishing alumina (BAS) and thoroughly rinsed with de-ionized water. The composite coating was prepared by adding the desired amounts of cholesterol (usually 12 mg) and stearic acid (12 mg) to 1 ml of chloroform solution containing 10 mg of L-a-phosphatidylcholine. Each silver disc of the dual-disc electrode block was covered with 5 1.11 (1 drop) of the chloroform solution.A 10-20 PI Wiretrol I1 micropipette (Drummond Scientific, Broomall, PA, USA) was used for the delivery of chloroform solution. The potential was applied to the flow-through amperometric cell and currents were measured with a Model MA 5450 polarograph (Iskra); the resulting FI signals were recorded on a Servograph Model 61 strip-chart recorder (Radiometer, Copenhagen, Denmark) equipped with a REA 110 unit. Results and Discussion The FI amperometric determination of cyanide on a modified silver electrode was performed with the manifold illustrated in Fig. 1. The alkaline (pH 11) CN- standard or sample, after injection (I), is washed by the water carrier (C) to a mixing point with a reagent (R) (pH 8 borate buffer).The mixing coil (MC), positioned downstream, ensures thorough trans- formation of CN- to HCN (pKHCN = 9.21).14 The hydro- phobic hydrogen cyanide formed on-line in the FI manifold diffuses through the lipid membrane and is detected at a silver working electrode. The anodic current measured is propor- tional to the CN- concentration in the injected standard or sample. The effects of several parameters on the performance of the FI system were studied. The effect of the applied potential at the working silver electrode was investigated in the range -0.20 to +0.30 V versus the Ag-AgC1 reference electrode. Taking into account the signal-to-noise ratio achieved, and the interference effects, the optimum potential was found to be 0.10 V.It had been established earlier' that the optimum potential for CN- determination at the bare silver electrode is 0.0 V. Of the four acceptor solutions tested (0.1 mol dm-3 HCl, pH 4.5 acetate buffer, 0.1 mol dm-3 KN03 and pH 8 borate buffer), the best results were obtained with the borate buffer. Also, the highest FI signals were achieved with the shortest mixing coil utilized. Hence, for all subsequent experiments, a potential of 0.10 V versus Ag-AgCI reference electrode, a borate reagent stream and a 30 cm X 0.5 mm i.d. mixing coil were used. As noted earlier, Wang and Lull developed a highly stable phosphatidylcholine-cholesterol coating for the glassy carbon electrode. These experiments indicated that better selectivity towards CN- is achieved when stearic acid is added as a third component to the lipid membrane.The addition of stearic acid Scan --.-c Fig. 2 Response of the amperometric detector with ( a ) bare and ( 6 ) lipid-modified silver electrodes t o three repetitive injections of: A, cyanide (0.100 mmol dm-'); B, thiosulfate (0.100 mmol dm-3)); and C, thiosulfate (1 .OO mmol dm-3) increases the concentration of negatively charged sites at the surface of the coating, which then, in turn, cause better discrimination against potentially interfering anions. The permselectivity of the three-component membrane is illus- trated in Fig. 2, which depicts typical FI peaks for CN- and S2032- at the bare and lipid-modified silver electrodes. As can be seen, unlike the complete elimination of the thiosulfate signals at 0.1 mmol dm-3 levels, large peaks are still observed for CN- at the coated electrode. It was suggested" that the ratio between the currents at the film-coated electrode over that at the bare electrode, im/ih, could be used as a measure of the permeability and thus selectivity.These values for S2032- and CN- at 0.1 mmol dm-3 concentrations were found to be 0 and 0.10, respectively. Even for a thiosulfate concentration ten times higher this ratio was only 0.004. Hence the membrane exhibits effective discrimination against S2032- and tolerable attenuation of the signal for CN- ion. The corresponding im/ib values for a two-component (phos- phatidylcholine-cholesterol) membrane at 0.1 mmol dm-3 concentrations of CN- and S2032- were found to be 0.11 and 0.07, respectively.A plausible rationalization that clearly explains the ob- served experimental selectivity relationship could be based on the Donnan equilibrium effects. A composite membrane based on a mixture of phosphatidylcholine, cholesterol and stearic acid can be envisioned as a cation exchanger. The membrane has fixed carboxylic acid sites. At pH values greater than about 6.8 these sites are completely ionized and therefore it is possible that the membrane would exclude co-ions (in this instance anions) to an extent determined by the Donnan equilibrium. This phenomenon could be responsible for hindering the passage of the ionized species (S2032-, for example) to the lipid-modified silver electrode, while allowing the transport of the non-ionized analyte (HCN) through the membrane to the working electrode.The ability of the FI manifold developed here to determine CN- selectively in the presence of other anions will be illustrated later. Wang and Lull demonstrated that composite coatings fabricated from cholesterol and phosphatidylcholine were stable when applied to the glassy carbon electrode. In order to show that addition of the stearic acid to the membrane and its application to the silver working electrode also produce a stable coating under dynamic FI conditions, the following experiment was performed. A 0.1 mmol dm-3 CN- standard was sequentially injected for a 2 h period. No change in the FI signals was observed, which indicates the integrity of the lipid membrane under dynamic flow conditions. The same experi- ment has an additional significance: it shows that it is possibleANALYST, JANUARY 1992, VOL.117 I I I I 49 Table 1 Comparison of FI results for the determination of CN- in the presence of other anions (all samples contained 50.0 pmol dm-3 CN- and 1 .00 mmol dm-3 of the anion investigated) t l 1 al d.-0.15 pA 1 rnin scan - Fig. 3 Ra id scan response to two repetitive injections of a 0.100 mmol dm-?cyanide standard at ( a ) bare and ( b ) lipid-modified silver electrodes. Valve switchings are indicated by the arrows 2.8 2.4 h p 2.0 2 2 1.6 1.2 L 2 0 -I to modify, with lipid coatings, electrodes that participate directly in the amperometric detection process (dissolution of silver electrode) and not only those which function solely as a medium for electron transfer (glassy carbon or platinum electrodes). A comparison of the responses of the bare and modified silver electrodes to the dynamic changes in the concentration of the analyte is shown in Fig. 3.As can be seen, the time elapsed from the injection of the CN- standard to the appearance of the FI signal is almost the same with both the bare and the lipid-modified silver electrodes [compare the traces in Fig. 3(a) and (b)]. This indicates that the kinetics of the diffusion of HCN across the lipid layer are fast on the FI time-scale. Also, it can be seen that there is no significant change in the peak profiles. This indicates an effective wash out of the analyte from the lipid membrane, which is a prerequisite for successful FI operation. The linearity studies were conducted by injecting in triplicate eleven CN- standards with concentrations between 0.500 pmol dm-3 and 0.100 mmol dm-3. A linear relationship was found between the logarithm of the anodic peak current (nA) and logarithm of the concentration (pmol dm-3).A typical calibration graph, illustrated in Fig. 4, had a slope of 0.55 k 0.02 with a correlation coefficient of 0.9978 (all the statistics were calculated for a 95% confidence level). The detection limit, calculated according to the recommended procedure,lS was found to be 0.1 pmol dm-3, which corre- sponds to 0.26 ng of CN- (the sample loop volume was 100 yl). Under these conditions, the relative standard deviations for the determination of cyanide were 1.7 and 3.1 % (n = 9) for 0.1 mmol dm-3 and 0.5 ymol dm-3, respectively.As can be seen from Fig. 4, a similar linear relationship was found also for a bare silver electrode. Hence it can be inferred that the nature Ion added Acetate B ro mat e Bromide Chloride Carbonate Citrate Fluoride Iodide Nitrate Nitrite Oxalate Phosphate Sulfate Sulfide Sulfite Thiocyanate Thiosulfate Added as CH3COONa KBr03 KBr NaCl Na2C03 Sodium citrate KF KI KNO3 KNO? K?C?OJ KZHPOJ Na7S04 Na2S Na2S03 KSCN Na2S203 Difference (%)* 0 0 0 0 0 -1.8 1.9 121-1 0 0 0 0 0 --$ 0 0 0 * Compared with a pure CN- (50.0 vmol dm-3) standard; as -t Does not interferc at a 2-fold cxccss. -$ Out of scale; interferes significantly at all concentrations. determined by a t-test at a 95% confidence level. Table 2 Analysis of a synthetic water sample Concentration/pmol dm-3 Taken Found* 1 .00 1 .oo 5 0.01 5.00 4.96 k 0.01 10.0 10.03 k 0.02 25 .o 25.22 * 0.04 50.0 51.6 k0.l * Mean k standard deviation ( n = 5 ) .~ ~ Table 3 Comparison of CN- determination by FI and argentimctric titrationI3 in samples of industrial salt wastes obtained after thermic treatment of steel Concentration ( % ) 4: Sample No. I t 20.1 k 0 . 1 20.3 1 2 0.04 2 3 A rge n t i me t ri c F1 (4.7 * 0.1) x 10-4 (1.4 k 0.6) x 10-4 (5.05 k 0.04) x lo-' ( 1 .30 -t 0.01) x 1 0 - 4 * Mean t standard deviation ( n = 5 ) . t Before thermic treatment. of the calibration relationship is not influenced by the lipid coating . The effects of possible interferents on the cyanide determi- nation are summarized in Table 1. As can be seen, the presence of the following ions in a 20-fold excess relative to CN- do not interfere: acetate, bromate, bromide, chloride, carbonate, citrate, fluoride, nitrate, nitrite, oxalate, phos- phate, sulfate, sulfite, thiocyanate and thiosulfate.I t is noteworthy that Br-, C1-, SCN- and S203z- do not cause a positive error in CN- determination at the lipid-modified silver electrode as they would at a bare electrode. Iodide ion interferes at a 20-fold excess, more than doubling the anodic peak current. However, it does not cause a change in the CN- peak when present in a 2-fold excess. The Donnan equilibrium effects could also be used to explain why higher levels of I- (1.00 mmol dm-3) interfered whereas the same levels of Br-, for example, did not. The Donnan equilibrium-type exclusion is species dependent, so it is plausible that at higher I - concentrations some of the anion is transported across the composite membrane and is detected at a silver working electrode.However, the selectivity of the membrane is still operational as the i,,,/i,, value for this species is less than 0.03.50 ANALYST, JANUARY 1992, VOL. 117 Sulfide ion was found to be the only serious interferent. This is not surprising, as H2S formed on-line in the manifold would be expected to diffuse through the lipid coating. However, if SZ- is present in the sample, it can be removed by precipita- tion with lead(i1) acetate; lead(i1) does not form strong complexes with the CN- ligand. I n order to illustrate the potential of the proposed FI method, synthetic water samples (Table 2) and industrial salt wastes obtained after thermic treatment of steel (Table 3) were analysed.This work demonstrates that composite permselective lipid coatings, by rejecting the potentially interfering species from the surface of the electrode while allowing the transport of the analyte, offer a substantial enhancement of amperometric FI methods. The authors acknowledge the financial support of the US Bureau of Mines under the Mining and Mineral Resources Institute Generic Center programme (Grant number G1125132-3205, Mineral Industry Waste Treatment and Recovery Generic Center) and the Serbian Republic Research Fund. References 1 Rocklin, R. D., and Johnson. E. L.. Anal. Chem.. 1983,55.4. 2 3 4 5 6 7 8 9 10 11 12 13 14 15 Liu, Y . , Rocklin. R. D., Joyce. R. J.. and Doyle, M. J., Anal. Cliem., 1990, 62. 766. Pihlar, B., and Kosta, L., Anal. Chim. Acta, 1980, 114, 275. Utley, D., Analyst, 1990, 115, 1239. Pollema, C. H., Hendrix, J. L., Milosavljevid, E. B . , Solujid, L., and Nelson, J. H.. J. Photochem. Pliotobiol. A: Cliem., submitted for publication. Milosavljevid, E. B., SolujiC, L., Hendrix, J . L., and Nelson, J. H., Anal. Chem., 1988.60, 2791. Garcia. 0. J.. Quintela, P. A., and Kaifer. A. E., Anal. Chem.. 1989, 61, 979. Chastel, O., Kauffmann, J. M., Patriarche, G. J., and Christian, G. D., Anal. Chem.. 1989. 61. 170. Tanaka. K., and Tamamushi. R., J . Electrounal. Cliem., 1987. 236, 305. Uchida, I . , Ishiho. A., Matsue, T., and Itaya, K.. J . Electyo- anal. Cliem.. 1989, 266. 455. Wang. J., and Lu, Z . , Anal. Cliem.. 1990, 62. 826. Wang. J.. Rayson. G. D.. Lu, Z . , and Wu, H . , Anal. Chem.. 1990, 62, 1924. Vogel. A. I . , A Text Book of Quantitative Inorganic Analysis, Longman, London. 1961, 3rd edn.. pp. 271-272. Izatt. R. M., Christenscn, J. J., Pack, R. T., and Bench, R., Inorg. Cliem.. 1962. 1, 828. Analytical Methods Committee, Analyst, 1987. 112, 199. Paper I l02344F Received May 20, 1991 Accepted August 29, I991