ANALYST, JANUARY 1992, VOL. 117 51 Comparison of the Influence and Contribution of the Response Times of Coated Open-tubular Solid-state Bromide- and Chloride-selective Electrodes on the Analytical Throughput (Dispersion) in Flow Injection Systems Jacobus F. van Staden Department of Chemistry, University of Pretoria, Pretoria 0002, South Africa An evaluation of the influence and contribution of the response times of coated open-tubular solid-state bromide- and chloride-selective electrodes on the analytical throughput (dispersion) in flow injection (FI) systems is presented using concentration ranges of 10-5000 mg dm-3 for bromide and chloride, respectively, in the study. For both electrodes the adsorption rate is independent of concentration and does not contribute significantly to any decrease in the sampling rate, although the adsorption rate of the chloride-selective electrode in the FI system seems to be slightly faster than the bromide system.For both electrodes the desorption mechanism process is mainly responsible for the sampling rate obtained in an optimized FI-ion-selective electrode system. An electrode memory of the bromide-selective electrode is the main reason for the sampling rate of this system being lower than that obtained for a similar FI system with a chloride-selective electrode. Keywords: Flow injection; tubular solid-state bromide- and chloride-selective electrodes; flo w-through system; dispersion; response times It was shown in a previous paper' that the shape and quality of the analytical signal obtained from a coated open-tubular inorganic-based solid-state silver-silver chloride ion-selective electrode incorporated into a flow injection (FI) system are directly influenced by both the electrode and the FI manifold.It was also shown that the electrode characteristics and response of a chloride-selective electrode in an FI flow- through system are basically controlled by two mutually dependent components: ( i ) a contribution from the type of electrode used, which depends on the chemical properties of the specific electrode used at the moment in time, e.g., silver-silver chloride, i.e., the electrode aspect; and (ii) a contribution from the partial dispersion originating from flow dynamics resulting in a concentration-time profile obtained from both convection and diffusion transport, i.e., the FI manifold aspect.The study was extended in this paper to compare the influence and contribution of the response times of coated open-tubular inorganic-based solid-state bromide- and chloride-selective electrodes on the analytical throughput (dispersion) in FI systems using a slightly different approach in connection with the data acquisition system. Experimental Reagents and Solutions Analytical-reagent grade chemicals were used throughout unless specified otherwise. Doubly distilled, de-ionized water was used to prepare all solutions. The water was tested beforehand for traces of chloride. All solutions were de- gassed before measurements were made by using a vacuum pump system. The main solutions were prepared as follows.Ionic strength adjustment reagent Dissolve 202.22g of potassium nitrate in 2dm3 of distilled water in a calibrated flask to obtain a 1 mol dm-3 solution of potassium nitrate. Standard bromide solutions Dissolve 29.7860 g of dried potassium bromide in 2 dm3 of distilled water to give a stock solution with a bromide concentration of 10 000 mg dm-3. Prepare working standard bromide solutions by dilution of appropriate aliquots of the stock solution to cover the range 5-5000 mg dm-3. Standard chloride solutions Dissolve 32.9680g of dried sodium chloride in 2dm3 of distilled water to give a stock solution with a chloride concentration of 10 000 mg dm-3. Prepare working standard chloride solutions by dilution of appropriate aliquots of the stock solution to cover the range 5-5000 mg dm-3.Instrumentation Construction of the coated tubular flo w-through electrodes The basic design, preparation and conditioning of the coated tubular flow-through solid-state bromide-selective' and chloride-selective3 membrane electrodes were similar to those described previously.2.3 Five different coated tubular flow- through solid-state bromide- and chloride-selective electrodes were prepared using the same procedures as previously described2.3 in order to validate the data. Flow injection system, data acquisition and assimilation A manifold system with a sampler unit, sampling valve and peristaltic pump similar to the system described previously~-3 was used in this study. The tubular flow-through solid-state bromide- and chloride-selective electrodes were incorporated separately, as needed, into the conduits of the FI system for the specific studies.Experiments were performed on five different coated tubular flow-through solid-state bromide- and chloride-selective membrane electrodes, respectively, in order to validate the data properly. Bromide and chloride standard solutions were injected into the corresponding FI-ion-selective electrode (ISE) systems in order to obtain representative runs. The experiments were also conducted in a series of five consecutive runs for each electrode type to ensure that the results obtained fell within the required accuracy and precision. A very good reproducibility between the different prepared ISEs was obtained and also between the series of five consecutive runs.The potentials were measured at room temperature with an Orion Model 901 microprocessor Ionalyzer. A binary coded decimal (BCD) output signal, equivalent to the displayed value in front of the Orion microprocessor Ionalyzer, was directed via a 40-pin 8-4-2-1 BCD connector at the rear of the instrument with appropriate interfacing to an XT IBM-com- patible microcomputer, 640 kbyte RAM, equipped with an 80287 mathematics co-processor, EGA colour graphics card with 256 kbyte graphics memory and high-resolution monitor.52 -40.0 0) ANALYST, JANUARY 1992, VOL. 117 I 'Oo0 .- 5000, I I I The acquired data were manipulated using ASYST software (ASYST Software Technologies, Version 1.04, Macmillan Software, New York). The background noise received with the raw data from the electrodes was removed by applying the DS (data set) Cat and Smooth unary operation of the file processor of the ASYST software.With the DS Cat and Smooth operation, consecutive data sets in long data files are catenated and then smoothed by convolution with a filter computed from a low-pass Blackman window frequency response. Selection of a cut-off frequency (cycles per point) is available in a prompt list to allow flexibility in the amount of smoothing. These background-corrected data were plotted on a Hewlett-Packard Model 7475A-compatible serial plotter. The constructed flow-through tubular indicator electrodes were used in conjunction with an Orion Model 90-02 double-junction reference electrode with 10% m/v potassium nitrate as the outer chamber filling solution.Results and Discussion The relevant parameters that are normally selected in the quantitative evaluation of the shape and quality of the analytical signal obtained from coated open-tubular inorganic- based solid-state ion-selective electrodes interfaced to an FI system are peak intensity and residence time. These two parameters resulted in a concentration-time profile and the observed peak dimensions revealed valuable information on the electrode characteristics of a specific type of electrode used. This ultimately formed a very important part in the selection of optimum conditions as far as precision, accuracy and sample throughput are concerned in analytical measure- ments for a specific application. The electrode characteristics and response, in a flow-through system, are controlled by the chemical properties of the specific electrode used and the partial dispersion arising from the flow dynamics of the system.The basic background related to the contribution of the electrode and FI manifold to the analytical signal has been described previously. 1 The BCD acquired raw data for representative runs, transferred as raw data to the ASYST system, contained less background noise than the detector output as previously directed via an analogue-to-digital ( N D ) converter to the ASYST system.' The main reason for the noisier background previously experienced' was probably the accumulation of noise in the detector output and the channelling of the analogue data from the detector output (normally used for a recorder) via a screened cable to the daughter board of the LabMaster card. The integration time function in the A/D converter of the Orion model circuit consists of a conversion cycle of the integrator output which contains four phases.The first phase (zero period) is initiated by the line synchronization and is terminated by the counter exactly 100ms later. The second phase (input integration) is initiated by the counter 100 ms later, i.e, 200 ms after synchronization. The third phase (reference integration) is initiated by the counters at 300 ms after line synchronization and is terminated by a zero crossing, 0 to approximately 260 ms later, depending on the value of the integrator voltage at the beginning of phase three. The fourth phase (idle) is initiated at the end of the update pulse.The counters are then reset to zero, turning on the zero switch to hold the integrator at zero while it waits for line synchronization and the start of a new conversion cycle. The purpose of the idle period is to provide synchronization with the a.c. line frequency. Two features combine to provide rejection of noise at the a.c. line frequency. Firstly, line synchronization ensures that each conversion begins at the same point in the line waveform. Secondly, the input integration period is chosen to be an exact multiple of both the 50 and 60Hz line frequencies, so that in either situation the integrator receives a whole number of line cycles of any power frequency ripple superimposed on the input from the pre- amplifier. This study, however, still showed some amount of background noise with the raw data.By applying the DS Cat and Smooth unary operation of the file processor of the ASYST software, the background noise was removed. The background-corrected graphs, as they appeared on the moni- tor screen, are represented in this paper. A typical representative DS Cat and Smooth operated background-corrected graph for the determination of bromide with the FI system and coated tubular bromide-selective electrode is presented in Fig. l ( a ) . Standard bromide solutions in the range 10-5000 mg dm--7 were injected, each standard in duplicate. The total run was performed in 2300 s and the signal range varied between -60 and 140 relative mV. An experimental calibration plot of analytical signal versus pBr = -log[Br] of a typical representative DS Cat and Smooth operated background-corrected run, illustrating the dynamic linear response of the bromide-selective electrode, is pre- sented in Fig.l(b). The linear range depends on the pre- treatment of the electrode and care should be taken during the preparation of each new coated tubular electrode. When incorporating an electrode into the conduits of an FI system, the sample volume is not only crucial to the design of efficient FI systems but also plays a major part in the sensitivity and linearity of a calibration graph. The DS Cat and Smooth unary operation of the file processor of the ASYST software might also have a very small effect on the represented background- corrected graph. The calculated Nernstian response of the tested electrodes at steady state, using the same procedure as described previously,z is 57.7 k- 1 mV per decade with a correlation coefficient of 0.9996, which confirmed previous results.' The dynamic linear response range in Fig.l(b) was less than those at steady state, but was consistent with those previously obtained.' The results from Fig. l ( a ) revealed two 120.0 80.0 40.0 0.0 250 750 1250 1750 2250 .- - 60 - 40 - 20 0 20 40 60 80 100 I I I I I 0 - 1 -2 -3 -4 PBr Fig. 1 ( a ) Typical representative DS Cat and Smooth operated background-corrected graph of raw data for a run with different standard bromide solutions from the tubular bromide-selective electrode-FI system as directed via a BCD output and as transferred to thc ASYST system and as it appeared on the monitor screen.From left to right, 5000-10 mg dm-3 standard bromide solutions; each standard injccted in duplicate. Total run performed in 2300 s and the signal range varied between -60 and 140 relative mV. ( 6 ) Experimen- tal calibration plot of analytical signal versus pBr of a typical representative DS Cat and Smooth operated background-corrected runANALYST, JANUARY 1992, VOL. 117 important points: the tendency of a slower return to the baseline of the higher bromide standards compared with the lower standards and a slight shift in the baseline over the entire run. When these results are compared with a typical represen- tative DS Cat and Smooth operated background-corrected graph for the determination of chloride with a similar FI system and a coated tubular chloride-selective electrode as outlined in Fig.2(a) (chloride solutions in the range 10- 5000 mg dm-?, running time 2300 s, variation of the signal range between SO and 205 relative mV), the first observation is that the return to baseline of the higher chloride standards seemed faster and the baseline seemed to be constant over the entire run. An experimental calibration plot of analytical signal i'emfs pC1 = -log [Cll of a typical representative DS Cat and Smooth operated background-corrected run, illustrating the dynamic linear response of the chloride-selective electrode, is presented in Fig. 2(h). The calculated Nernstian response of the tested electrodes at steady state, using the same procedure as previously described,3 is 57.7 k 1 mV per decade with a correlation coefficient of 0.9998, which confirmed previous results.-3 Again the dynamic linear response range in Fig.2(6) was less than that at steady state, but was consistent with those previously obtained.3 The typical respresentative DS Cat and Smooth operated background-corrected graph for the deter- mination of chloride obtained riia the BCD output signal [Fig. 2(u)] is a better representation of the chloride-selective electrode-FI system than that described previously. 1 The main reason for this conclusion is that the amount of background noise received previously' together with the raw data was of such a nature that it definitely influenced the DS Cat and Smooth unary operation of the file processor of the ASYST software. 195.0 165.0 135.0 105.0 75.0 li I loo0 [15000, I 1 I 250 750 1250 1750 2250 .- 4- a Time/s - L .- m o zn v) .- 50 100 150 200 I I I I 0 -1 -2 -3 -4 PCI Fig. 2 ( L I ) Typical representative DS Cat and Smooth operated background-coi-rcctcd graph of raw data for a run with different standard chloride solutions from the tubular chloride-sclcctivc clcc- trodc-FI systcm as directed viu a BCD output and as transferred to the ASYST system and as it appeared on the monitor screen. From lcft to right, 5000-10 mg dm-.3 standard chloride solutions: each standard injcctcd in duplicatc. Total run pcrformcd in 2300s and thc signal range varied between 50 and 205 relative rnV. ( h ) Experimental calibration plot of analytical signal wI:su.s pCI of a typical rcprcscnta- tivc DS Cat and Smooth operated background-corrected run 53 The two typical representative graphs [Figs.l(a) and 2(a)J did not give a detailed observation of the individual peaks. In both graphs all the peaks did not, for example, reach the baseline before the start of the next peak, a condition that is more prominent within the higher concentration range. In order to compare the different peaks over the whole concen- tration range a more in-depth study of the performance of individual peaks of the different bromide concentrations was necessary. Separate representative runs of the different bromide concentrations were therefore conducted in order t o obtain a baseline-to-baseline performance of each individual peak. This was carried out as follows. Each bromide standard solution was injected in duplicate into the ISE-FI system allowing sufficient time for both peaks in a run to reach the baseline.This was also conducted in a series of five consecu- tive runs to ensure that the results obtained fall within thc required accuracy and precision. By accumulating data in such a way, it was possible to evaluate the results in a more logical way, which led to better conclusions. The beauty of the ASYST system is the manipulation of individual peaks from which valuable information can be revealed. Figs. 3 and 4 show a selection of individual pcaks from the above-named runs of individual bromide (Fig. 3) and chloride (Fig. 4) standards where the position of the individual peaks was manipulated in such a way that overlapping of peaks was possible.The main aim was to try to obtain an overlapping of individual peaks from different concentrations at a fixed point, thereby enlarging the selective individual bromide and chloride peaks to the same graphical intensity on the y-axis and to study the response behaviour of the bromide- and Time - Fig. 3 Enlargements of thrcc sclcctcd individual bromide pcaks to the samc graphical intensity on the y-axis: 5000 mg dm- 3 of Br- bctwccn -SO and 130 relative mV, 500 mg dm-3 of Br- bctwccn 0 and 130 relative mV and SO rng dm-3 of Br- bctwccn 59 and 135 relative mV. A cut-off frequency of 1000 s was used on the x-axis 5000, 50 5b0 500C ,-500 50 Time - Fig. 4 Enlargements o f thrcc sclcctcd individual chloridc peaks to the samc graphical intensity on the y-axis: SO00 mg dm-3 of CI- between 58 and 203 rclativc mV.500 mg dm-3 of Cl- bctwccn 1 1 0 and 205 relative mV and 50 mg dm- 3 of CI- bctwccn 164 and 204 rclativc mV. A cut-off frequency of 1000 s was used on the x-axis54 ANALYST, JANUARY 1992, VOL. 117 chloride-selective electrodes for different bromide and chloride concentrations, respectively, from this common point. The rising part of each curve was used for simultaneous overlapping. Three selected individual bromide peaks were enlarged to the following graphical intensity on the y-axis: 5000mgdm-3 Br- between -50 and 130 relative mV, 500mgdm-3 Br- between 0 and 130 relative mV and 50 mg dm-3 Br- between 59 and 135 relative mV (Fig. 3). The three selected individual chloride peaks were enlarged to the following graphical intensity on the y-axis: 5000 mg dm-3 Cl- between 58 and 203 relative mV, 500 mg dm-3 CI- between 110 and 205 relative mV and 50 mg dm-3 CI- between 164 and 204 relative mV (Fig.4). In order to comply with the main goal of baseline-to-baseline performance, no enlargement on the 124.0 93.2 62.0 30.8 -0.400 620 628 636 644 652 122.0 1 I E . - 126.0 0, m .- 107.0 87.5 68.5 620 628 636 644 652 128.0 113.0 97.0 81.4 65.8 624 632 640 648 656 Time/s Fig. 5 Enlargements of the leading edges of four selected individual bromide peaks to the same graphical intensity on the y-axis and the same cut-off frequency of 40 s on the x-axis. ( a ) 1000 mg dm-3 of Br- between -16 and 140 relative mV; ( 6 ) 500 mg dm-3 of Br- between 0 and 135 relative mV; (c) 100mgdm-3 of Br- between 40 and 135 relative mV; and (d) 50 mg dm-2 of Br- between 58 and 136 relative mV x-axis was made and a relatively large cut-off frequency of 1000 s was used.It is clear from the above information that the influence of the ‘wash-in’ and ‘wash-out’ of a bromide analyte sample plug in the flow detection conduits of the coated tubular bromide- selective electrode can be divided into four sections: two for the wash-in part and two for the wash-out part. As previously outlined,’ the output form for a certain part of a peak is controlled by the electrode mechanism, whereas the flow dynamics of the manifold system dominate another output region of the peak. Also, the distortion of the Gaussian peak shape at the lower front and rear part of the peak was attributed to dominance of the electrode mechanism, which is in agreement with the adsorption-desorption mechanism that occurs at low concentrations.1 198 174 1 50 126 1 02 647 655 663 671 679 200 180 160 140 > E 120 Q, .- 644 652 660 668 676 *.’ - Q, 191 179 167 154 620 628 636 644 652 1681 , . , , , ,c 660 668 676 684 692 Time/s Fig. 6 Enlargements of the leading edges of four selected individual chloride peaks to the same graphical intensity on the y-axis and the same cut-off frequency of 40 s on the x-axis. ( a ) 1000 mg dm-2 of CI- between 90 and 210 relative mV; ( b ) 500mgdm-3 of C1- between 109.8 and 209.8 relative mV; (c) 100 mg dm-3 ofC1- between 148 and 210 relative mV; and ( d ) 50 mg dm-3 of CI- between 164 and 205 relative mVANALYST. JANUARY 1992.VOL. 117 55 Although the plots in Figs. 3 and 4 clearly differentiate between the leading (wash-in) and tailing (wash-out) edges of the peaks, the large cut-off frequency of 1000 s on the x-axis only gave a global, but still valuable, insight into these two parts. A closer examination of the baseline-to-baseline performance of the ‘wash-in’ part of the bromide and chloride samples was necessary, however, in order to analyse the leading edges of the peaks. In Figs. 5 and 6, enlargements of the x-axis were made by using a cut-off frequency of 40 s. The results in Fig. 5 clearly divide the performance of the bromide-selective electrode-FI system of the leading edge (wash-in part) of the bromide peaks into two sections. A time delay was observed in the first section near the baseline, where the front part of the sample plug starts to move into the conduit of the tubular bromide-selective electrode.As the bromide concentration in a very small fraction of the front edge of the sample plug is very low, the time delay was probably mainly due to the chemical properties of the electrode at the moment in time and more specifically to the adsorption mechanism process at the initial lower portion of the peaks. The flow dynamics of the manifold system dominate the second section of the leading edge, where the bulk of the sampling zone starts to move through. This became more obvious (Fig. 5 ) when the front part of the sampling zone (where the bromide is concentrated) in the FI conduits took over for a period of time, after which the peak jumped to a maximum output. The adsorption rate seems to be indepen- dent of the concentration evaluated. I t seems that the increase in the dominance of the adsorption of the electrode mechan- ism becomes more obvious with a decrease in bromide concentration as observed in Fig.l ( a ) , where the peak intensities of the range of peaks are different. This is, however, not the situation as seen from Fig. 3, where the three peaks reflecting different concentrations of 50, 500 and 5000mgdm-3 of Br- were enlarged to the same peak intensity, a fact which is also true for the concentration range 10-5001) mg dm-3 of Br-. This is confirmed by the enlarged plots given in Fig. 5. It is, therefore, clear that the leading edge of the bromide peaks with different bromide concentrations in this FI-ISE study did not influence the analytical throughput of a specific optimized FI-system. The influence of the flow dynamics and adsorption rate of the bromide-selective elec- trode-FI system (Figs.3 and 5 ) was also compared with the flow dynamics and adsorption rate of a similar chloride- selective electrode-FI system (Figs. 4 and 6). The contribu- tion from the flow dynamics of the manifold system was identical for both electrode systems, but the adsorption rate of the chloride-selective electrode-FI system seems to be slightly faster than the bromide system. The results in Figs. l(a) and 3 clearly illustrate that for the top part of the peak, where the centre portion of the sample plug with the higher bromide concentration is moving through the electrode conduit, the influence of dispersion flow dynamics becomes predominant.I t also follows (Fig. 3) that the top portion near the peak maximum of the peak shape is due to the bulk of bromide analyte moving through the electrode conduit. This is confirmed by the results in Fig. 7, where enlargements of both the x- and y-axis were made. This is followed by peak tailing until the baseline is reached. The results in Figs. 3 and 7 also clearly distinguish between two sections in the performance of the bromide-selective elec- trode-F1 system of the wash-out part of the bromide peaks. The first portion of the rear part of the peak tailing (Figs. 3 and 7) is attributed to the tendency of the rear of the sample zone to move to the tube centre.This output region of the peak tailing nearest to the peak maximum is mainly due to the wash-out effect originating from the flow dynamics of the manifold system, where the rear part of the sample zone moves through the electrode conduits. Confirmation of this is obvious from Fig. 7. Although this tailing of the peaks has an effect on sampling rate, the contribution is reduced for bromide concentrations between 20 and SO0 mg dm-3. The influence, however, increases for bromide concentrations above 500mgdm-3 as observed from Figs. 3 and 7. In the second section, as the peaks approach the baseline, the effect of the rear part of the sample zone decreases owing to the flow dynamics of the manifold system and the desorption mechan- ism starts to dominate. The major contribution to any delay in analytical sampling rate of a specific optimized FI-ISE system for different bromide concentrations comes, however, from the part nearest to the baseline, where the peaks of 10- 250 mg dm-3 of Br- approach baseline conditions.From 250 to 5000mgdm-3 of Br-, the influence of the desorption process on sampling rate becomes more marked (Fig. 7). Figs. 3 and 7 clearly illustrate that the desorption mechanism process is mainly responsible for thc sampling rate obtained in an optimized FI-ISE system with a coated tubular bromide- selective electrode. When the rear part of the peaks of the coated open-tubular 124.0 93.2 62.0 30.8 -0.400 658 694 730 766 802 122.0 94.5 67.5 40.5 > 5 13.5 F > .- +d - \ - m & 126.0 v) .- 107.0 87.5 68.5 49.5 688 724 760 796 832 658 694 730 766 802 735 771 807 663 699 Time/s Fig.7 Enlargements of the tailing cdges of four sclcctcd individual bromide pcaks to the same graphical intensity o n the y-axis as in Fig. 5 . but with a cut-off freqiicncy of 180 s on thex-axis. ( a ) 1000; ( h ) 500; (c.) 100; and ( d ) 50 mg dm--’ of Br-56 ANALYST, JANUARY 1992, VOL. 117 198 174 150 126 102 688 724 760 796 832 200 180 1 60 140 > 5 120 > .- .I- - 683 719 755 791 827 2 . - m & 204 v) 191 .- 1 79 167 154 663 699 735 771 807 I V I I I I I 703 739 775 811 847 Time/s Fig. 8 Enlargcmcnts of the tailing cdges of four selected individual chloride peaks to the same graphical intcnsity on the y-axis as in Fig. 6 , but with a cut-off frequency of 180 son thcx-axis. ( a ) 1000; (6) S00; (c) 100; and (d) 50 mg dm-3 of Cl- solid-state bromide-selective electrode-FI system (Figs.3 and 7) was compared with the corresponding results obtained for the chloride-selective electrode-FI system (Figs. 4 and 8), the following observations were made. In the first section of the rear part of the output region of the peak tailing for both the bromide- (Figs. 3 and 7) and chloride-selective (Figs. 4 and 8) electrodes, where the sample zones in both instances moved to the tube centre, the contributions to peak shapes were mainly due to the wash-out effect originating from the flow dynamics of the manifold system. It seems from Figs. 3 and 4 that the flow dynamics of the manifold systems contribute more to the first section of the tailing of both the bromide and chloride peaks, which is actually not the whole explanation.Although there is a differentiation between the performance of the wash-out part of the bromide and chloride peaks, when compared in Figs. 3 and 4, the large cut-off frequency of 1000 s only gives a global view of the situation. When an enlargement I ; Time - Fig. 9 Enlargcmcnts of pairs of bromide and chloride pcaks to the same graphical intensity on thc y-axis as in Figs. 5-8. The pairs of pcaks wcrc manipulated in such a way that overlapping of twin pcaks was possible. A cut-off frcqucncy of 1000 s was used on thc x-axis. A , 1000 mg dm-3 of Br-; B, 1000 mg dm-3 of CI-; C, S O 0 mg dm-3 of Br-; D, 50Omgdm--i of C1-; E, 100mgdm-2 of Br-; and F, 100 mg dm-3 of CI- (Figs. 7 and 8) of the x-axis was achieved by using a cut-off frequency of 180 s, a better view of the situation was obtained.I t was clear from these results (Figs. 7 and 8), and more obviously from the bromide peaks (Fig. 7), that the desorption rate is already starting to contribute in the first section of the tailing edge. The contribution of the desorption rate increases as the peaks move nearer to the baseline and eventually dominates in the region of the baseline. A further conclusion to be drawn is that the desorption rate of the bromide- selective electrode system is slower than that of the chloride- selective electrode system and that this contribution also comes into effect in this rear part of the peak shape. This conclusion is confirmed from the representations in Figs. 4 and 8. I t is clear from the results that the chloride-FI system responds faster than the bromide-FI system for the tailing edge of the peak shape. The results, however, show that the major contribution to the analytical throughput of both electrode systems in the tailing edge comes from the desorp- tion process. The desorption rate of the chloride-selective electrode is, however, faster than that of the bromide- selective electrode. This is confirmed in Fig. 9, where three peaks reflecting different concentrations of 100, 500 and 1000 mg dm-3 of Br- and CI- were enlarged to the same peak intensity and curves were drawn wherc the corresponding peaks overlapped. I t can therefore be concluded that the electrode memory of the bromide-selective electrode is mainly responsible for the sampling rate of the bromide-selective-FI system being lower than that obtained for a similar FI system with a chloride- selective electrode. The Foundation for Research Development (FRD) , Pretoria, and the University of Pretoria are thanked for financial support. J. C. Lindeque is also thanked for assistance in performing some of the experiments. References 1 2 3 van Stadcn, J . F.. Anulysr, 1990, 115. 581. van Staden, J . F., Analysr. 1987. 112, 595. van Staden, J . F., Anal. Cliim. A m . 1986, 179, 407 Paper O105092.1 Receiried Nmyernber 13, I990 Accepted June 18. I991