Analyst, January, 1979, Vol. 104, pp. 55-62 Determination of Chloride in High-purity Waters in the Range 0-20 pg I-' of Chloride Using lon-selective Membrane Electrodes Incorporating Mercury(1) Chloride 55 G. B. Marshall and D. Midgley Ceqztral Electricity Research Laboratories, Kelvin Avenue, Leatherhead, Surrey, KT22 7SE Two types of solid-state mercury(1) chloride electrodes have been used to determine chloride in the concentration range 0-20 pgl-l. At these low concentrations, more chloride will dissolve from the mercury (I) chloride in the electrode than is present in the sample itself. The extent of the dissolu- tion is controlled, however, by the chloride in the sample. In these circum- stances, the electrode potential is linearly related to the concentration of chloride in the sample.With the electrode housed in a flow cell with a thermostatically controlled water jacket, the correlation coefficient between e.m.f. and concentration was always greater than 0.99. The sensitivity (0.18 mV per pg 1-1 of chloride a t 25 "C and 0.4-0.5 mV per pg 1-1 of chloride a t 4 "C) was about ten times greater than that of the silver - silver chloride electrode. Total standard deviations a t 10, 5 and 2 pgl-1 of chloride were 0.4, 0.5 and 0.3 p g 1-1 of chloride, respectively. Keywords : Chloride determination ; ion-selective electrodes ; mercury(I) chloride electvodes ; high-purity waters We earlier developed a solid-state mercury( I) chloride electrode for determining low levels (0.01-1 mg 1-l)" of chloride by a manual methodl and suggested that even lower levels could be determined by using the electrode in a flow cell at a carefully controlled temperature, as has been done with a silver - silver chloride ele~trode.~ The mercury(1) chloride electrode is about ten times more sensitive than a silver chloride electrode because of the lower solubility of the mercury(1) salt.Boiler waters from modern power stations can have chloride contents as low as 0.01 mg 1-1 and even lower levels can be expected in condensed steam. The deter- mination of very low levels of chloride is particularly important for stations with once- through boilers, where the condensate should be monitored before being returned to the boiler. This paper describes the performance of mercury(1) chloride electrodes in the range 0-20 pg 1;l of chloride in flowing solutions, at a controlled temperature during several weeks of operation.Theoretical The potential of an ion-selective electrode is described by a form of the Nernst equation. For a chloride electrode, this is E = E" - k ln[Cl-] where E" is the standard potential of the electrode, [Cl-] is the total concentration of chloride in solution and k = RT/F, R being the gas constant, T the absolute temperature and F Faraday's constant. We have used concentrations instead of activities because in the method the nitric acid added to the sample provides a constant ionic medium and activity coefficients are, therefore, constant. The total concentration consists of the chloride originally present in the sample solution (m) and the chloride dissolved from the electrode itself (s), which are related by the solubility product equilibrium. For mercury(1) chloride * It should be noted that there were misprints in the title and summary of our previous paper.' In each instance the range of chloride concentration should have read 0.01-1 pg ml-1.256 MARSHALL AND MIDGLEY: DETERMINATION OF CHLORIDE IN AnaZyst, VoZ.104 Ks = [Hgz2+fI [Cl-,l2 = 0.5s(m + s ) ~ Ion-selective electrodes are generally considered to be useful only if nz is greater than s, so that there is a linear relationship between E and lnm, but, when nz is less than s, an expression for the potential can be derived as .follows: E = E" - kln(m -t s) = E" - klns -- kln 1 + - ( 3 Expanding In 1 + - and neglecting second.- and higher-order terms, we obtain ( 3 E m E" - 2ilns - km/s At such low values of m, s is virtually constant and hence E , to a very close approximation, is linearly related to m: E m E' - k'.zvt where E' = E" - klns and k' = k/s.Experimental Apparatus Potentials were measured with a Corning 110 digital pH meter reading to 0.1 mV and displayed on a Servoscribe 2s chart recorder. Two types of ion-selective electrodes were used, those made from Radiometer F3012 Universal Selectrodes, as described previously,ll and the Ionel Model SL-01 (Ionel Electrodes, Mount Hope, Ontario, Canada). Both have rnembranes made of a mixture of mercury(I1) sulphide and mercury(1) chloride; in the electrode developed at CERL the mixture is used to impregnate a graphite - PTFE electrode, while in the Ionel electrode the mixture is hot-pressed into a The reference electrode was a mercury - mercury(1) sulphate electrode with a 0.5 mol 1-1 sodium sulphate filling solution (instead of the usual 1 mol 1-1 solution, which would have precipitated at the lowest operating temperature of 4 "C).The reservoir containing the filling solution and the reference element could be raised up to 0.5 m above the remote liquid junction, which was of the ground-glass sleeve type. The chloride electrode and the remote junction were housed in a Perspex flow cell fitted with a water jacket through which thermostatically controlled water was circulated by means of a Techne C-100 thermocirculator operated in conjunction with a Techne, Model 1000, chiller unit. The remote junction was in a separate compartment downstream of the chloride electrode compartment, which contained a magnetic stirrer bar.The flow cell was painted black to exclude light. A Technicon, Model I , proportioning pump delivered the solutions to the flow cell, pumping the sample or standard solutions at 3.9 ml min--l and the acid reagent solution at 0.8 ml min-l. Air was injected at 1 ml min-l to improve mixing. The air-segmented mixture of sample and reagent solutions passed through a glass mixing coil to a T-piece, from which the air and a portion of the solution were extracted at 1.6 ml min-l. The rate of delivery of solution to the flow cell was, therefore, 4.1 ml min-l. The pump tubes were made of Y V C , but the transmission lines were made of PTFE. Before entering the cell, the solution passed through a short length of stainless-steel tubing,, which was connected to the chassis ground terrninal on the pH meter.A stainless-steel wire was immersed in the solution downstream of the electrode pair and also connected to the chassis ground point. Without these connections the signal was noisy. Reagents Water. Town mains water was distilled in .a stainless-steel still (Manesty Machines Ltd., Liverpool), and the distillate passed through a twin-column mixed-bed de-ionisation unitJanuary, 1979 HIGH-PURITY WATERS USING ION-SELECTIVE ELECTRODES 57 (Elga Products Ltd., Model B106/2). This water had a specific conductivity o€ less than 0.1 pS cm-1 at 20 "C as it left the unit and it has been found5 to have a chloride content of about 0.7 pg 1-l.A stock solution (1 000 mg 1-1 of chloride) was prepared by dissolving 1.649 g of sodium chloride in water and making up to 1 1 in a calibrated flask. Intermediate stock solutions (100, 10 and 1 mg 1-l) were prepared by successive dilutions of the above solution. The working chloride solutions were prepared by pipetting appropriate volumes of 1 mg 1-1 stock solution into polyethylene bottles of known mass, of capacity approximately 5 1, and then adding water until the mass of solution in each bottle was 5 kg (using a Mettler P11 balance weighing to 0.1 g). Nitric acid, 0.6 nzol Z-l. The reagent was prepared in 10-1 batches by dilution of 38 ml of concentrated nitric acid (BDH, Aristar grade). At the pumping rates used, this solution gave the same final concentration of acid as that used in the manual rneth0d.l Staizdard chloride solutions.Procedure method,l i.e., solutions were stirred, light was excluded and nitric acid was added. As far as possible, the experimental conditions were kept the same as in the manual Results Both the electrode developed at CERL and the Ionel electrode were tested at 25 "C and at a temperature close to 4 "C. In each instance, five batches of five standard chloride solutions (1, 2 , 5 , 10 and 20 pg 1-l) were analysed in duplicate in random order. Each batch was analysed on a separate day and all five within 7 days. The response to de-ionised water was also recorded. Sensitivity temperatures. Linear responses over the range 0-2Opg1-1 were obtained for both electrodes at both The results of linear regression analysis of the data are shown in Table I.TABLE I CALIBRATION OF MERCURY(I) CHLORIDE ELECTRODES Standard deviation Calibration slope/ of slope/ Temperature/ mV per pg 1-1 mV per pgl-' Correlation Electrode "C of chloride of chloride coefficient CERL . . 25.0 0.1785 0.003 0 0.9995 CERL .. 4.3 0.430 0.001 6 0 9999 Ionel . . . . 25.0 0.175 0.002 2 0.999 7 Ionel . . .. 3.6 0.549 0.0044 0.999 9 The calibration slopes of the two electrodes agreed fairly well at 25 "C. The agreement at the lower temperatures was not so close, but this was caused, at least in part, by the difference in temperature. The linearity of the plots was excellent, as shown by the correlation coefficients. Precision At 25 "C the Ionel electrode gave slightly smaller total standard deviations, but at 4 "C the CERL electrode was slightly better. None of these differences were significant at the P = 0.05 level of the F-distribution. Between-batch standard deviations were zero or non-significant, with two exceptions; one of these was only "possibly significant'' (CERL electrode at 4.3 "C and 5 pg 1-l) and the other appeared because of a freak result for the within-batch standard deviation (Ionel electrode at 25 "C and 1 pg 1-1). The results of precision tests are shown in Tables I1 and 111.58 MARSHALL AND MIDGLEY: DETERMINATION OF CHLORIDE IN Analyst, VoZ.104 TABLE I1 PRECISION OF MEASUREMENTS; WITH THE CERL ELECTRODE Chloride Standard deviation,? Temperature/ concentration/ A e.m.f.*/ A > "C CCg 1-1 mV Within-batch Between-batch Total 26 20 0 0.083 - - (0.47) - - 0.106 (0.69) 0.131 (0.73) 0.167 (0.93) 0.144 0.212 10 1.81 0.077 NSS (0.43) (;;) (0.71) (NOS) (0.79) (;;) 6 2.61 0.126 2 3.21 0.167 1 3.43 0.141 NS O§ (0.93) (0) (0.81) (1.19) (1.14) (NS) 20 0 0.100 - (0.23) - - 10 4.33 0.166 0 0.166 (0.36) 0.224 (0.36) (0) 6 6.49 0.122 O.lSST[ (0.29) (0.44) (0.62) 2 7.74 0.106 NS 0.131 (0.31) 0.166 1 8.16 0.141 (0.33) (NS) (0.36) 3.74 0.204 - (0.24) w;) * Mean e.m.f.normalised with respect t o 20 p g 1-1 solution, e.g., A, = E , .- E,. t Standard deviations are in mV except for the figures in parentheses, which are in concentration 9 De-ionised water. 7 Significant at the P = 0.05 level but not att the P = 0.01 level. units (pg 1-1). NS = not significant at the P = 0.05 levell. 4.3 Criterion of Detection The criterion of detections is given by 1.65;42aB, where U, is the within-batch standard deviation of the blank.The criteria for the Ionel electrode were 1.0 pg 1-1 at 25 "C and 1.7 pg 1-1 at 3.6 "C and for the CERL electrode were 2.6 pg 1-1 at 25 "C and 0.8 pg 1-1 at 4.3 "C. As the blank was not included in the precision trial of the CERL electrode at 4.3 "C, uB was approximated by the within-batch standard deviation of the lowest (1 pg 1-1) standard solution; in view of the criteria of detection obtained at 25 "C and by comparison with those for the Ionel electrode, this is a fair approximation. Accuracy Recovery tests were carried out on a number of power station waters, using the Ionel electrode at 25 "C. The solutions were analysed and then spiked with 5 pg 1-1 of chloride solution.From the standard deviations in Table 111, the recovery was predicted with 95% confidence to be 5.0 & 1.1 pg 1-1 for a single result. The results in Table IV show no definite exceptions to this, but as the condensed steam and feed water samples had concentrations below the criterion of detection, the recovery could only be quantified for the boiler water sample (104%). Response Time The time to reach an equilibrium e.m.f. was dominated by the characteristics of the flow system. About 5 min were required for each fresh solution to reach the sensing electrode. Once the electrode had started to respond to tlhe new solution, 5-10 min were taken to reach equilibrium, depending on the difference in concentration between successive solutions. This time was affected little by temperature or by which electrode was being used (Fig.l), although the Ionel electrode responded faster when immersed in fresh solution in a beaker (about 2 min compared with 5 min for the CElRL electrode). The response time was also affected by stirring.January, 1979 HIGH-PURITY WATERS USING ION-SELECTIVE ELECTRODES TABLE I11 PRECISION OF MEASUREMENTS WITH THE IONEL ELECTRODE Chloride Temperature/ concentration/ 25 20 "C pg 1-1 3.6 10 5 2 1 O§ 20 10 5 2 1 A e.m.f.*/ mV 0 1.77 2.69 3.16 3.31 3.66 0 5.43 8.32 9.82 10.43 10.98 Standard deviation, t Within-batch Between-batch r A 0.095 - 0.063 NS$ 0.063 NS 0.055 NS (0.54) - (0.36) (NS) (0.36) (NS) (0.31) (NS) 0.000 0.1027 (0.44) (0.0) (0.35) - (0.43) (0) (0.37) (NS) (0.00) (0.58) 0.078 0.0 0.190 - 0.236 0 0.205 NS 0.127 NS 0.288 0.410 (0.23) (y) (0.52) '3 (0.75) (0) 59 Total - - 0.072 (0.41) 0.087 (0.49) (0.067) (0.38) 0.102 (0.68) 0.078 (0.44) - 0.235 (0.43) 0.258 (0.47) 0.141 (0.26) 0.288 (0.62) 0.410 (0.75) * Mean e.m.f.normalised with respect to 20 pgl-l solution, e.g., A2 = E, - Ezo. t Standard deviations are in mV except for the figures in parentheses, which are in concentration units ( pg 1-1). $ NS = not significant a t the P = 0.06 level. 5 De-ionised water. 7 Significant at the P = 0.001 level. Effect of Stirring Without stirring in the electrode compartment of the flow cell, both electrodes took more than twice as long to reach equilibrium compared with when the solution was stirred. The e.m.f. shifted by up to 2mV when the stirring was stopped, but the sensitivity of the electrodes was not affected.Effect of Temperature The increased sensitivity at low temperatures is shown by the AmV results in Tables I1 and 111. The precision, however, was not improved by working at low temperatures, possibly because of the greater difficulty of maintaining a steady temperature at sub-ambient levels. Although the flow cell had a thermostatically controlled water jacket and was kept inside an insulated cabinet, at such low concentrations as were tested more precise control of the air temperature would have been desirable. TABLE IV RECOVERY TESTS WITH POWER STATION WATERS Chloride content/pg 1-1 r I Station Unit Sampling point Sample Sample + 5 pgl-l spike A 2 De-aerator <1 6.4 A 2 Extraction pump <1 7.0 A 3 Economiser inlet <1 6.8 A 3 Boiler 9.2 14.4 B 1 Extraction pump <1 6.2 A 3 Extraction pump <1 4.760 MARSHALL AND MIDGLEY: DETERMINATION OF CHLORIDE IN Analyst, VoZ.104 X A 4 20 pg I-' 2 pg ,\ I-' j--y m t B 1 20 pg I--'- t X X X c 2opg I-' Time Fig. 1. Response curves for mercury( I) chloride electrodes : A, CERL electrode a.t 25 "C; B, Ionel electrode at 25 "C : a n d C, Ionel electrode at 4 "C. X = change of solution. The standard potentials of the electrodes increased as the temperature decreased. The e.m.f.s observed a t 20 pg 1-1 of chloride changed by 15 and 21 mV for the CERL and Ionel electrodes, respectively, when the temperature changed from 25 to 4 "C. The main cause of the shift and of the change in sensitivity was the greater concentration of chloride contributed by the dissolution of the mercury(1) chloride of the membrane at the higher temperature.Stability of the Cell Potential and Durability of the Electrodes The CERL electrode operated continuously for a t least 2 months without needing to be re-impregnated. The Ionel electrode has operated for 1 month without needing to be re-polished. When standard solutions were not being passed through the flow cell the apparatus was run on de-ionised water. The day-to-day stability of the e.m.f. can be judged from the results in Table V. TABLE v STABILITY OF E.M.F. OF THE CERL ELECTRODE IN 10 pg 1-1 CHLORIDE SOLC'TION AT 4.3 "c Day I \ 1 2 8 9 10 11 16 E . m. f . /mV -50.6 -51.3 -51.0 -51.3 -50.9 -60.7 -50.9 Reference Electrodes It was noted previously1 that the condition of the reference electrode was critical for obtaining good results. If the head of electrolyte solution above the ground-glass sleeve junction was only a few centimetres, the junction This was also true in flowing solutions.January, 1979 HIGH-PURITY WATERS USING ION-SELECTIVE ELECTRODES 61 needed to be flushed out daily with fresh solution.With the head of electrolyte solution at about 50 cm, the reference electrode required no attention other than re-filling. Discussion Comparison with Silver - Silver Chloride Electrodes Compared with the limiting linear response of the silver - silver chloride electrode a t low concentration^,^ that of the mercury( I) chloride electrode is about ten times more sensitive.The mercury(1) chloride electrode at 25 "C is still four times as sensitive as the silver - silver chloride electrode at 5 "C. The within-batch standard deviations reported by Tomlinson and Torrance3 were smaller, in millivolt terms, than we found, but because of the different sensitivities were larger in concentration units. Silver chloride electrodes respond faster than mercury(1) chloride electrodes, but for most purposes the difference is unimportant. Sekerka et aZ.7 used an Ionel electrode in 1-1000 pg 1-1 of chloride solutions at 25 "C; the within-batch standard deviations of their results at 10 and 20 pg 1-1 were slightly larger (0.64 and 0.68 pg l-l, respectively) than those we obtained. Apparatus The apparatus is similar to that already used with the silver - silver chloride ele~trode.~ Unless the maximum sensitivity is required, however, the apparatus can run at 25 "C or possibly even higher temperatures; in this instance the chiller unit could be dispensed with, giving considerable savings in space and expense.For continuous use, one that has a fairly high head of electrolyte solution or some other means of maintaining a free-flowing junction is recommended. The most troublesome part of the apparatus was the reference electrode. Absolute and Relative Chloride Concentrations Because the de-ionised water used to prepare the standard solutions will never be perfectly free of chloride, absolute measurements at these levels are scarcely possible by direct potentio- metry, whether with silver or mercury(1) chloride electrodes.We have shown, however, that standard solutions can be prepared in a consistent way at concentrations between 1 and 20 pg 1-l. The absolute chloride concentration in the solutions was 0.7-1.0 pg 1-1 higher than nominal, as concentrations of this magnitude were measured in the water we used by Dimmock and Webber.5 Very similar chloride concentrations in other sources of de-ionised water have been found by Florence8 and by Tomlinson and Torrance3 in their work with silver chloride electrodes. No change in signal was observed when we replaced our standard solutions with corresponding solutions prepared with de-ionised water that had been passed repeatedly through a mixed-bed de-ionisation column. If the electrode is used to detect ingress of chloride, e.g., from cooling water leaking into condensed steam, the linear e.m.f. - concentration relationship has the advantage that an increase in chloride concentration is indicated directly by the change in e.m.f.even though the absolute levels are uncertain. If the change in e.m.f. was interpreted by means of the conventional logarithmic relationship, the increase in concentration would also be biased. Conclusions The use of mercury( I) chloride electrodes for determining chloride concentrations can be extended from the levels attainable by manual analysis1 to very low levels (less than 20 pg 1-l) by housing the electrode in a flow cell at a controlled temperature. At a given temperature, the electrode is about ten times more sensitive than the silver chloride electrode used at present. One advantage of this greater sensitivity is that the mercury(1) chloride electrode does not need to be operated at sub-ambient temperatures in order to obtain adequate precision in the concentration range 1-20 pg 1-1. Two kinds of electrode can be used, that developed at CERLl and the Ionel SL-01. The two electrodes are almost equally sensitive and precise, but the Ionel has a faster response time and requires no prepara- tion.62 MARSHALL AND MIDGLEY This work was carried out at the Central Electricity Research Laboratories and is pub- lished by permission of the Central Electricity Generating Board. 1. 2. 3. 4. 6. 6. 7. 8. References Marshall, G. B., and Midgley, D., Analyst, 1978, 103, 438. Marshall, G. B., and Midgley, D., Analyst, 1978, 103, 784. Tomlinson, K., and Torrance, I<., Analyst, 1977, 102, 1. Lechner, J. F., and Sekerka, I., J . Electroanalyt. Chem. Interfacial Electrochem., 1974, 57, 317. Dimmock, N. A., and Webber, H. M., C.E.R.L. Laboratory Note RD/L/N 56/77, Central Electricity Roos, J. B., Analyst, 1962, 87, 832. Sekerka, I., Lechner, J. F., and Harrison, L., ,/. Ass. Off. Analyt. Chem., 1977, 60, 625. Florence, T. M., J . Electvoanalyt. Chem. Interfacial Electrochem., 1971, 31, 77. Research Laboratories, Leatherhead, 1977. Received July 21st, 1978 Accepted August 15th, 1978