Analyst, June 1996, Vol. 121 (755-759) 755 Determination of 2-Furaldehyde in Transformer Oil Using Flow Injection With Pulsed Amperometric Detection* John W. Dilleen, Chris M. Lawrence and Jonathan M. Slater Centre for Analytical Science, Birkbeck College, University of London. Gordon House, 29 Gordon Square, London, UK WCIH OPP An on-line flow injection (FI) system suitable for the on-site determination of 2-furaldehyde in transformer oil has been developed. This paper examines the feasibility of combining FI with pulsed amperometric detection (PAD) to produce a method for the rapid on-line analysis of 2-furaldehyde. The determination of 2-furaldehyde by FI with PAD is direct and simple. PAD was applied to the determination of 2-furaldehyde at a platinum electrode in alkaline media.Plots of peak height versus concentration gave linear regions for concentrations in the range 5-40 ppm at a pump rate of 0.4 ml min-1 and detection potential (El) of -750 mV (versus SCE), and in the range 1-8 ppm at a pump rate of 0.8 ml min-l and El of -800 mV. Detection limits of 1 ppm (S/N = 3) were found for El = -750 mV (pump rate = 0.4 ml min-1; injection volume = 67 pI) and 0.5 ppm for El = -800 mV (pump rate = 0.8 ml min-1; injection volume = 100 pl), respectively. 2-Furaldehyde was extracted from transformer oil through a partitioning membrane in a flow through macrodialyser cell. Under optimized conditions the extraction efficiency for 2-furaldehyde from transformer oil, using differential pulse polarography experiments, was found to be 15% at oilheagent flow rates of 0.3 ml min-1.Keywords: Pulsed amperometric detection; platinum electrode; flow injection; 2-furaldehyde (furfural); power transformer; insulating oil; oil condition monitoring; cellulose degradation; partition membrane Introduction Power transformer conductor windings are insulated with paper impregnated with insulating oil. A typical high power trans- former on a national electricity grid contains 10-12 t (1 t = 103 kg) of cellulose based paper, 30-120 mm thick, and 45 t of oil. Due to heat, water and oxygen effects the cellulose degrades over a period of time. This reduces the polymer molecular chain length which in turn reduces its mechanical strength. The life of the insulation usually determines the ultimate life of the transformer, although other factors may cause it to fail early.Analysis of the degradation products of paper insulators in transformers can be used to monitor their degree of depolymer- ization (DP).' 2-Furaldehyde is one such product, so too is 2-acetylfuran and 5-methyl-2-furaldehyde. Other degradation products which may provide useful markers are phenol, m- cresol and xylene which are degradation products of phenol- formaldehyde insulating resins present in the transformer windings. Emsley and Stevens' concluded that 2-furaldehyde and related compounds were suitable markers of transformer * Presented at Sensors and Signals 111, Malahide, Co. Dublin, Ireland, October 26-27, 1995. paper insulation condition. In situ monitoring of these products may provide a simple method of predicting transformer failure.Methods currently employed in the UK for transformer condition monitoring include oil acidity, oil moisture content, dissolved gas in oil analysis and 2-furaldehyde in oil analy- sis.l-4 Burton et al.5 demonstrated methods of analysis based on HPLC of extracts from the oil, and demonstrated an empirical, inverse relationship between log (DP of cellulose) and 2-fur- aldehyde concentration. They found 2-furaldehyde levels between 1-10 pprn in poorly cooled transformers. Paper degradation, and hence 2-furaldehyde levels, increase in transformers prone to overheating. The aldehyde group of 2-furaldehyde should be directly detectable at a solid noble metal electrode by pulsed ampero- metric detection (PAD).6 The method is based on multi-step potential waveforms, which incorporates amperometric detec- tion with successive anodic and cathodic polarizations in a constantly replenished analyte.7-17 The application of alternate anodic and cathodic polarizations cleans and reactivates the electrode surface.9 PAD in conjunction with flow injection (FI) has been applied successfully to the determination of numerous aliphatic compounds including carbohydrates, alcohols, alde- hydes, amines and many organic sulfur species.11-16 FI allows a small volume (typically 100 p1) of analyte to be injected into a carrier stream flowing through a detector.6,10-'6317 This enables manual wet chemistry techniques to be automated in the interests of high speed analysis and drastically reduced cost per sample.In this paper various PAD-FI regimes are considered for determining 2-furaldehyde in aqueous carrier systems.The best regime can be incorporated in a system to determine 2-furaldehyde extracted from transformer oil. Differential pulse polarography (DPP) of 2-furaldehyde and related compounds is well established,l8,19 and was used to determine the effective- ness of the extraction of the target analyte from transformer oil. Experimental Reagents and Solutions All solutions were prepared with deionized water from a Whatman R050 reverse osmosis-deionizer system (Whatman Labsales, Maidstone, UK). Reagents used were sodium per- chlorate (AnalaR grade, Merck, Poole, UK), sodium hydroxide (analytical-reagent grade, Rh6ne-Poulenc, Manchester, UK), 2-furaldehyde (298% purity, Fisons, Loughborough, UK) and 2-acetylfuran (Aldrich, Dorset, UK).Transformer oil standards containing 2-furaldehyde and 2-acetyl furan were supplied by National Power (Surrey, UK) . Solution preparation The carrier solution was 0.1 mol 1-l NaOH-O.1 mol 1-I NaC104.756 Analyst, June 1996, Vol. 121 Equipment and Methods Current-potential curves (14) were obtained by linear scan cyclic voltammetry at a platinum electrode. The detector assembly was a Metrohm 656 electrochemical detector (Met- rohm AG, Herisau, Switzerland) with a thin-layer flow cell. The flow cell comprised a Metrohm platinum working electrode (area 5.3 mm2), glassy carbon auxiliary electrode, and an SCE. The platinum electrode was polished with Metrohm coarse (0363 19) and fine (0363 18) polishing compounds, rinsed with deionized water, sonicated in a water bath for 5 min, and dried before use.The glassy carbon and reference electrodes were used as supplied. The detector was used in conjunction with an FIAstar 5020 Flow Injection Analyser pumping module with flow-through injector (Tecator AB, Hoganas, Sweden). The FI system included two 4-channel peristaltic pumps controlled by a microprocessor. The pumps can be operated independently, e.g., for stopped-flow or intermittent pumping. A macrodialyser cell, SpectraPore MacroDialyser Cell, (Spectrum Medical Industries, Houston, Texas, USA), was used to partition the analyte from the oil to the supporting electrolyte carrier for detection. Samples of 2-furaldehyde in virgin transformer oil were pumped through one chamber of the cell, whilst a supporting electrolyte stream (0.1 mol 1-1 NaOH-O.1 mol 1-l NaC104), was pumped through the adjacent chamber (pump rates = 0.3 ml min-1). The two chambers were separated by a Durapore HVLP04700 (0.45 pm pore diameter) membrane [Millipore (UK), Watford, UK]. The membrane was specially manufactured from polyvinylidene difluoride to facilitate partition between aqueous and non-aqueous solutions. The supporting electrolyte stream could then be introduced into the sample loop for detection using PAD-FI (Fig. 1). Experiments in quiescent conditions were performed using the same working and counter electrodes in conjunction with an Ag/AgCl (saturated KC1) reference electrode. The electrolytic cell was a 25 cm3 beaker. The sample filling was performed by aspiration, via the injection valve tube which was connected to a pump tube.When the valve was activated the sample loop was interposed in the carrier stream. It is necessary to maintain the valve in the inject position for a certain period of time so that the sample loop can be completely emptied; a 10 s period was allowed between samples. In order to establish a stable baseline for the carrier it was necessary to earth the pump tubing using in-line copper tubing. A pulse suppressor was introduced between the pump and the flow through detector. These measures reduced baseline noise but increased injection noise which interfered with the signal. The injection noise was removed from the signal by sufficiently lengthening the tubing from the damper to the detector, and by keeping the injection time to a minimum.An injection time of 10 s was found to be satisfactory. However, for flow rates below 0.6 ml min-1 the sample loop (100 p1) was not fully emptied. Transformer oil Reagent Macrodialysex cell - Injection valve Analogue Fig. 1 Schematic of detection system for transformer oil contaminants. The potential waveform was generated by a Dionex Pulsed Electrochemical Detector (Dionex, Sunnyvale, CA, USA). The I-E curves and PAD peaks were recorded on a Philips PM 827 1 x-y/t chart recorder (Philips, Eindhoven, The Netherlands). For differential pulse polarography (DPP) an EG & G Princeton Applied Research (Princeton, NJ, USA) polaro- graphic analyser and stripping voltammeter, Model 264, was used in conjunction with a Model 303, hanging/dropping mercury electrode stand.The standard platinum auxiliary and dropping mercury electrodes were used but in order to reduce interference due to contamination, a special Ag/AgCl reference electrode, similar to that described by Torrance and Gatford,20 was used. This electrode utilizes a cracked borosilicate glass junction design which is easier to clean and remove con- taminants from than the Vycor frit Ag/AgCl reference electrode that is supplied with the Model 303 electrode stand. All of the analysed samples were first purged with white-spot nitrogen to remove dissolved oxygen for 4 min before each analysis. The sample volume during polarographic analysis was always kept constant (10 ml) so as to remove any variance that may arise from accumulation of the analyte on the mercury drop surface.21 The polarographic glass cup was washed in dilute hydrochloric acid, then rinsed in acetone and deionized water between experiments.All experiments were performed at an ambient temperature of 20 "C. Results and Discussion Voltammetry Quiescent conditions Cyclic voltammetry experiments carried out in quiescent conditions showed that 2-furaldehyde progressively fouled the electrode surface on repeated cycling. Fig. 2 shows cyclic voltammograms recorded at a clean platinum electrode in: (a) background supporting electrolyte solution (0.1 moll-' NaOH- 0.1 moll-' NaC104), (b) 100 ppm 2-furaldehyde in supporting electrolyte, and (c) at a 'fouled' Pt electrode (in supporting electrolyte) which had undergone 10 repeated cycles in 100 ppm 2-furaldehyde in supporting electrolyte and gentle rinsing in de-ionized water.Scan (c) showed a decrease in electrode response, at the detection potentials (around -0.8 V versus Ag/AgCl), and also 0.7 0.2 4.3 4.8 -1.3 E l V versus AglAgC1 Fig. 2 Current versus potential curves obtained by cyclic voltammetry at a platinum electrode under quiescent conditions in: (a, 0) background supporting electrolyte solution (0.1 moll-' NaOH-O.1 mol NaC104), (b, A) 100 ppm 2-furaldehyde in supporting electrolyte, and (c, 0) at a 'fouled' Pt electrode (in supporting electrolyte) which had undergone 10 repeated cycles in 100 ppm 2-furaldehyde in supporting electrolyte and gentle rinsing in deionised water. Scan rate; 100 mV s-l.Analyst, June 1996, Vol.121 757 the appearance of small reduction and oxidation peaks at about 0.45 and 0.15 V versus Ag/AgCl, respectively. Hydrodynamic conditions Cyclic voltammograms of the supporting electrolyte carrier (0.1 moll-1 NaOH-0.1 moll-' NaC104), and 100,200 and 400 ppm concentrations of 2-furaldehyde, respectively, using the FI detector during fluid flow (0.4 ml min-1) and cycling between +1200 and -1200 mV at 100 mV s-1, are shown in Fig. 3. Several well-defined surface reactions of the platinum electrode can be seen. The upper cathodic I-E curve (negative potential scan) shows the reduction peak of surface oxides with a maximum at about -400 mV. The adsorption peak of hydrogen22-24 occurs between -650 and -900 mV. The lower anodic I-E curve (positive potential scan) shows the oxidation peaks for the dissolution of adsorbed hydrogen22-24 in the voltage range between -850 and -500 mV.Between about -300 and +800 mV platinum forms surface oxides, and oxygen evolution starts at about +800 mV.22-24 It was noted that 2-furaldehyde suppresses the cathodic current of the reduction peak of surface oxides and the adsorption peak of hydrogen, and lowers the overpotential for hydrogen evolution during the negative potential scan, between about - 1000 and - 1 100 mV. During the positive potential scan 2-furaldehyde suppressed the anodic current of the oxidation peak of adsorbed hydrogen and enhanced the anodic current of formation of surface oxides on the platinum electrode between about -300 and +800 mV.2-Furaldehyde suppressed the anodic current of oxygen evolution between about +750 and +900 mV. During the cyclic voltammetry experiments the following processes 0ccur:22-2~ (1) Adsorbed 2-furaldehyde is catalytic- ally oxidized during the positive scan at about -300 mV by the formation of PtOH, resulting in an enhanced anodic response. (2) The anodic response for 2-furaldehyde is reduced due to the formation bf the catalytically inactive PtO, giving rise to a plateau betiyeen about - 100 and +600 mV. (3) 2-furaldehyde blocks complete coverage of the electrode with PtO, hence the reduc~on in the cathodic response for the reduction of surface oxides at about -400 mV during the negative potential scan. (4) In the region between about -400 and -900 mV 2-furaldehyde is adsorbed strongly at the oxide free platinum surface, resulting in a decrease in the number of surface sites available for the adsorption of atomic hydrogen.-ve 4 +ve +1200 +600 0 -600 -1200 E f mV versus SCE Fig. 3 Current versus potential curves obtained by cyclic voltammetry at a platinum electrode under fluid flow in 2-furaldehyde in 0.1 rnol 1-I NaOH-O. 1 moll-' NaC104 reference solutions: 0 ppm, H 100 ppm, A 200 ppm, V 400 ppm. Flow rate; 0.4 ml min-1. Scan rate, 100 mV s-1. Extraction of 2-Furaldehyde and 2-Acetylfuran From Transformer Oils 2-Furaldehyde and 2-acetylfuran (a related compound also formed by the decomposition of cellulose) may be determined by DPP since the carbonyl and acetyl groups undergo reduction at a mercury electrode between approximately -1000 and - I800 mV versus Ag/AgCl reference electrode, depending on pH.18719 DPP was performed on 2-furaldehyde in 0.1 mol 1-I NaOH-0.1 moll-' NaC104, aqueous standards and on extracts from 2-furaldehyde-transformer oil samples in the supporting electrolyte stream (0.1 moll-1 NaOH-O.l mol 1-1 NaC104), for comparison.DPP was also performed on similar oil extracts containing both 2-furaldehyde and 2-acetylfuran. Extractant efficiency depends on the relative flows of the transformer oil and reagent through the two compartments of the macrodialyser cell. For these trials a compromise was made between extraction efficiency and sample throughput. The extractant efficiency was examined for the contaminant range in and above that of analytical interest (1-1000 ppm) at flow rates of 0.3 ml min-1.A typical polarogram of the extract from a 1000 ppm 2-furaldehyde oil sample showed the expected DPP profile for 2-furaldehyde in 0.1 mol 1-1 NaOH-O.1 mol 1-l NaC104, with a reduction peak around - 1400 mV. The DPP profile (Fig. 4) for the extract from a 1000 pprn 2-furaldehyde- 2-acetylfuran oil sample was similar but with a small secondary peak at -1570 mV corresponding to the co-extracted acetyl- furan, which was identified from standard additions. DPP was performed on standard solutions of 2-furaldehyde and 2-acet- ylfuran (100 ppm) in 0.1 mol 1 - 1 NaOH-O.l mol 1-1 NaC104 (Fig. 5). It was noted that the reduction of 2-acetylfuran gave much larger current values than the reduction of 2-furaldehyde, the opposite effect to that observed in the extract from the 1000 ppm 2-furaldehyde-2-acetylfuran oil sample.This suggests that 2-furaldehyde is much more readily partitioned from the oil into the aqueous extraction solution. The extraction efficiency for 2-furaldehyde was determined from comparisons between reduction peaks obtained for DPP in 2-furaldehyde standard solutions and oil extract solutions, and was found to be 15% at 0.3 ml min-1. Other workers have reported that the levels of 2-acetylfuran and related compounds are much lower than 2-furaldehyde in problem transformers (by a factor greater than 100 : l).1,25,26 Moreover, the poor extrac- tion efficiency for 2-acetylfuran compared to 2-furaldehyde would further decrease the interference in practical measure- ments, hence 2-furaldehyde alone was tested with PAD-FI.18.0 13.5 5. @ 9.0 5 0 4.5 W 0 -0.8 -1.0 -1.2 -1.4 -1.6 -1.8 I I L EN versus Ag/AgCI Fig. 4 Differential pulse polarogram of sample extracted from transformer oil, containing 1000 ppm of both 2-furaldehyde (A) and 2-acetyl furan (B), into 0.1 mol 1-1 NaOH-O.1 rnol 1-1 NaC104. Scan rate, 5 mV s-1. Pulse height, 50 mV.758 Analyst, June 1996, Vol. 121 Pulsed Amperometric Detection Flow Injection The extent to which 2-furaldehyde suppressed the cathodic current of the adsorption peak of hydrogen was found to be influenced by 2-furaldehyde concentration. PAD was carried out in this region for the indirect determination of 2-furaldehyde at an 'oxide free' platinum electrode. The potential waveform- current detection programmes used are described in Table l(a) and (b), based on the appearance of the cathodic signals for hydrogen.Optimum detection potentials, El, for PAD were between -750 and -800 mV. Values of potentials E2 and E3 for reduction of the electrode surface and oxidative cleaning were - 1200 and +800 V, respectively. The lengthening of the tubing from the damper to the detector increased sample plug diffusion which in turn increased the peak dispersion. The increased peak dispersion produced smaller peak heights but did not change peak areas. In FI dispersion is defined as the ratio of concentrations before and after the dispersion process has taken place in the element of fluid that yields the analytical signal. In a straight tube it is the result of radial and axial mixing, due to convection and diffusion.The rate of equilibration and baseline drift were determined. It was found that severe baseline drift in PAD continued for longer than 20 min. We attribute this to the reported gradual increase in true electrode area which occurs with surface reconstruction under the repeated conditions of alternate anodic and cathodic polarizations in the PAD waveform.14 During the detection process surface reconstruction takes place which is caused by repeated cycles of oxide formation and dissolution. 24 18 5 P 12 B 6 A -1.0 -13 -1.4 -1.6 -1.8 EIV versus Ag/AgCI Fig. 5 Differential pulse polarogram of 100 ppm each of 2-furaldehyde (A) and 2-acetyl furan (B) in 0.1 mol I-' NaOH-O. 1 moll- NaC104. Scan rate, 5 mV s-l. Pulse height, 50 mV.Table 1 Waveform specification for pulsed amperometric detection Potential/mV versus SCE Time/s Function (a) El -800 0 (b) El -750 0.50 Detection 0.59 Reduction of electrode 0.70 Oxidative cleaning EZ -1200 0.51 E3 +800 0.60 Integration period 0.30 to 0.50 s The carrier baseline was run for 2 hours to allow the platinum electrode to attain an equilibrium number of active surface sites, which reduced the baseline drift dramatically. It is important also that the ranges of the potential changes are chosen to span the majority of the voltammetric region for the surface oxide formation; however the potential should not exceed values for onset of significant evolution of oxygen. PAD was carried out on 2-furaldehyde samples of concen- trations between 1 and 80 ppm in carrier, respectively. The results are shown in Fig.6(a) and 6(b); the response was taken to be the peak height. For 2-furaldehyde concentrations in the range 1 to 16 ppm the waveform described in Table l(a) was used, where El = -800 mV. For a higher concentration range, 5-80 ppm, it was necessary to modify the waveform for one with El = -750 mV [Table l(b)]. Plots of 2-furaldehyde concentration against response gave linear regions for concen- trations in the range 1-8 ppm (regression equation; y = 0.861 + 0.293x, Y = 0.994, n = 5) at a pump rate of 0.8 ml min-1, El of -800 mV, and injection volume of 100 p1 [Fig. 7(a)], and in the range 5 4 0 ppm (regression equation; y = 0.642 + O.O67x, Y = 0.998, n = 5) at a pump rate of 0.4 ml min-l, El of -750 mV, and injection volume of 67 p1 [Fig. 7(b)].The limit of detection for 2-furaldehyde was determined by PAD-FI to be 1 ppm for E l = -750 mV (pump rate = 0.4 ml min-1; injection volume = 67 pl) and 0.5 ppm for E l = -800 mV (pump rate = 0.8 ml min-1; injection volume = 100 pl) respectively, using three times the standard deviation of the baseline for each system. The rate at which an analyte accumulates at the electrode surface is transport limited, hence the response which is proportional to surface coverage is often a linear function of concentration. The adsorption of 2-furaldehyde is also affected by changes in El. 2-Furaldehyde appeared to adsorb more strongly at the more negative potential. It appears that other electrode surface reactions occur for 2-furaldehyde concentra- tions above about 50 pprn at El = -750 mV and above about 10 ppm at El = -800 mV, hence the deviation from linearity of the concentration against response plots.This suggests that T 10 min H I I Fig. 6 PAD-FI detection peaks for 2-furaldehyde in 0.1 mol I-' NaOH- 0.1 moll-' NaC104. (a) El = -800 mV, E2 = - 1200 mV, E3 = +800 mV. Sample volume, 100 pl; pump rate, 0.8 ml min-I. (b) El = -750 mV, E2 = -1200 mV, E3 = +800 mV. Sample volume, 67 pl: pump rate, 0.4 rnl min-1.Analyst, June 1996, Vol. 121 759 the response is not entirely controlled by the adsorption isotherm which limits the surface coverage by adsorbed analyte. Conclusion This study shows that the combination of flow through PAD and FI can be used to produce a system suitable for the determina- tion of 2-furaldehyde in alkaline, aqueous media.The work also demonstrates the feasibility of on-line detection of transformer oil contaminants using selective partitioning membranes with a continuous flow extraction system for in-situ monitoring. Over a period of four weeks no noticeable deterioration in extraction efficiency was observed; however over extended periods of time it is likely that if extraction efficiency changes then recalibration of the system would be required. 0 2 4 6 8 10 12 14 16 2-furaidehyde concentration / ppm 0 10 20 30 40 50 60 70 80 2-furaldehyde concentration I ppm Fig. 7 Mean k s response against 2-furaldehyde concentration (a) El = -800 mV. Sample volume, 100 vl; pump rate, 0.8 ml min-1. (b) E l = -750 mV. Sample volume, 67 1.11; pump rate, 0.4 ml min-1.This research was supported by a grant from EA Technology Ltd., Chester, England. References 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 Emsley, A. M., and Stevens, G. C., IEE Proc.-Sci. Meas. Technol., 1994, 141, 324, and references cited therein. Szuta, J., Energopomiar (Poland), 1969, 15, 21. Domun, M., Cornfield, G., and Hadfield, A., CIGRE Symposium, Vienna, 1987, 1020-80, 1. Rybakov, L. M., Rybakova, G. A., Belyakov, I. G., and Akhmetshin, R. S., Prom. Energ. (USSR), 1986, 10,42. Burton, P. J., Carballiera, M., Duval, M., Fuller, C. W., Graham, J., de Pablo, A., Samat, J., and Spicar, E., Proceedings of the CIGRE Conference (Paris), 1988, Paper 15-08. Neuburger, G. G., and Johnson, D. C., Anal. Chem., 1988, 60, Andrew, K.N., Blundell, N. J., Price, D., and Worsfold, P. J., Anal. Chem., 1994, 66, 917A. Gunther, A., and Bilitewski, U., Anal. Chim. Acra., 1995, 300, 117. Johnson, D. C., Polta, J. A., Polta, T. Z., Neuberger, G. G., Johnson, J. Tang, A. P. C., Yeo, I. H., and Baur, J., J . Chem. SOC., Faraday Trans. I , 1986, 82, 1081. Polta, J. A. and Johnson, D. C., Anal. Chem., 1985, 57, 1373. Ngoviwatchai, A., and Johnson, D. C., Anal. Chim. Ada., 1988,215, 1. Lacourse, W. R., Jackson, W. A., and Johnson, D. C., Anal. Chem., 1989,61, 2466. Lacourse, W. R., Johnson, D. C., Rey, M. A,, and Slingsby, R. W., Anal. Chem., 1991, 63, 134. LaCourse, W. R., Mead, D. A. Jr., and Johnson, D. C., Anal. Chem., 1990, 62, 220. Larew, L. A., and Johnson, D. C., Anal. Chem., 1988, 60, 1867. Larew, L. A., Mead, D. A., and Johnson, D. C., Anal. Chim. Acra., 1988, 204, 43. Tait, R. J., Bury, P. C., Finnin, B. C., Reed, B. L., and Bond, A. M., Anal. Chem., 1993, 65, 3252. Zuman, P., Organic Polarographic Analysis, Pergamon Press, Oxford, 1964, and references cited therein. Meites, L., Polurographic Techniques, Interscience, New York, 1955, and references cited therein. Torrance, K., and Gatford, C., Talanra, 1985, 32, 273. Kalvoda, R., Anal. Chim. Acta., 1982, 138, 11. Hughes, S., and Johnson, D. C., Anal. Chim. A m . , 1981, 132, 11. Hughes, S., Meschi, P. L., and Johnson, D. C., Anal. Chim. Acta., 1981, 132, 1. Koryta, J., Dvorak, J., and Kavan, L., Principles of Electrochemistry, Wiley, New York, 2nd edn., 1993, pp. 352-368, and references cited therein. Jakob, F., and Haupert, T., presented at the 16th Technical Conference of the International Electrical Testing Association (NETA), Atlanta, GA, USA, March 1994. Kan, H., Miyamoto, T., Makino, Y., Narnba, S., and Hara, T., presented at the IEEE International Symposium on Electrical Insulation, Pittsburgh, PA USA, June 5-8, 1994. 2288. Paper 51083426 Received December 22, 1995 Accepted March 26, 1996