Analyst, January, 1979, Vol. 104, pp. 63-72 63 Assessment of Glass Electrodes for Determining pH in Boiler Feed Water D. Midgley and K. Torrance Central Electricity Research Laboratories, Kelvirt A venue, Leatherhead, Surrey, K T22 7SE Six types of commercial glass electrodes have been tested in the laboratory for their suitability for measuring pH in ammonia-dosed boiler feedwater of moderately low specific conductivity (about 5 pS cm-l). The electrodes were chosen to represent the range of pH-sensitive glasses available. All of the electrodes showed a near-theoretical sensitivity, had stable standard potentials and responded sufficiently quickly. In the dilute ammonia solutions, however, the electrodes indicated pH values that could differ by as much as 0.3 pH unit when the solution was flowing slowly through the measuring cell.When the solution was stirred the maximum bias was 0.05 pH unit. For most industrial purposes, the differences in performance between the various types of electrode are unimportant and glass electrodes are less of a problem than reference electrodes for pH measurements in this type of water. Keywords : pH determination ; glass electrodes ; boiler feed water Experience of continuous pH measurements in power station waters shows that there is considerable dissatisfaction with some aspects of this long established technique. Much of the trouble can be attributed to reference electrodes, which have been the subject of a previous study.l This work constitutes a similar appraisal of glass electrodes. The conditions in which the electrodes were tested were intended to simulate those prevailing in boiler feed water.The pH of de-ionised water was adjusted to about 8.5 by adding ammonia to give a concentration of 0.5 mg 1-1. The resultant solution had a specific conductivity of about 5 pS cm-l. Four general-purpose, screened, industrial electrodes were chosen for testing, together with two types of low-resistance electrode suitable for meas- urements at low temperatures or in high-purity water. Of the latter type, one (type F) had a glass with a low specific resistance and limited working ranges of temperature and pH, while the other (type D) had the same type of glass as the general-purpose electrode (type C) from the same company. The electrodes and their properties are summarised in Table I.Experimental Apparatus Two electrodes of each of the types listed in Table I were tested. They were placed in a 126 mm diameter water-jacketed Perspex flow cell such that electrodes of the same type were opposite each other. This arrangement minimised the risk of local conditions being TABLE I ELECTRODES TESTED Resistance/ Electrode Mi2 EIL 1072-110t . . .. 100 Coming 476022t . . .. Schott 9201s . . .. 100 Schott 92025 . . .. 20 Metrohm EA-107T . . 600 Metrohm EA-lO7Tq . . 30 - Working temperature range* /"C -6 to 100 -6 to 100 0 to 80 10 to 100 -30 to 80 -15 to 40 Working pH range* 0-14 0-14 0-14 0-14 0-14 0-1 1 Code A B C D E F * Manufacturers' figures. t Electronic Instruments Ltd., Chertsey, Surrey. $ Corning - EEL, Evans Electroselenium Ltd., Hdstead, Essex.3 H. V. Skan Ltd., Shirley, Solihull.64 MIDGLEY AND TORRANCE : ASSESSMENT OF GLASS ELECTRODES Analyst, Vol. 104 favourable to particular types of electrodes. The solution in the cell could be stirred, as required, by means of a magnetic stirrer bar. The temperature in the cell was maintained at 30.0 j, 0.1 "C by means of a Churchill thsermocirculator. A Pye 305 calomel reference electrode with 3.0 mol 1-1 potassium chloride solution as the reference electrolyte was situated in the central position of the flow cell, 45 mnn from all the glass electrodes. The flow cell was kept in a cabinet maintained at 30 "C by a 300-W heater controlled by a mercury relay thermometer. The flow cell and the cabinet have been described in more detail e1sewhere.l Potentials were measured with a Corning 110 digital pH meter connected through its recorder output terminals to the Central Datat Acquisition and Processing System (CDAPS) at CERL, which automatically gave a paper-tape record of the e.m.f.s and the time of measurement.The recorder span control of the pH meter was adjusted so that the meter acted as a unity-gain amplifier. The electrodes were switched in turn through the pH meter by a modified signal multiplexer with reed relays. Each relay was latched shut for 4 s before the e.m.f. was recorded. Reagents Buffers. Standard NBS phosphate (pH 6.853 at 30 "C) and disodium tetraborate (borax) (pH 9.139 at 30 "C) buffers were prepared fresh each time the electrodes were standardised. Stock ammonia solution.A 20-ml volume of 35% ammonia solution, sp. gr. 0.88, (BDH, Aristar) was added to 2.47 kg of de-ionised water in the glass reservoir of an automatic pipette fitted with a soda-lime guard tube on the air inlet. This solution had an ammonia concentration of approximately 2.5 g 1-l. Each week a 96-1 batch of approximately 0.5 mg 1-1 ammonia solution was prepared in two interconnected 50-1 aspirators by dilution of the stock solution with de-ionised water that was fed directly into the aspirators from the outlet of a mixed-bed de-ionisation unit. Working ammonia solzttio?~. Procedure The working solution was pumped from the aspirators to a header tank above the cabinet. The solution flowed from the header tank, through a glass capillary into the flow cell, at a rate of about 9 ml min-l and was then run to waste.The solution was not stirred while passing through the flow cell, except as detailed below. The header tank and the aspirators were protected from atmospheric carbon dioxide by means of soda-lime guard tubes. The aspirators held enough solution for continuous running for 1 week. The electrodes were monitored by CDAPS every 2 h for changes in their potentials. This enabled us to obtain an estimate of the short-term standard deviations of the electrode potentials [step (i)], to test the effects of stirring [step (ii)], to measure the response times [steps (iii) and ( i v ) ] and to check the calibration of the electrodes [steps (v) and (vi)]. (i) The potentials were recorded once a minute for at least 10 min. (ii) The stirrer was started and the potentials were again recorded at intervals of 1 min for at least 10 min.(iii) With the stirrer in operation, the flow of solution was stopped. The potentials were recorded at 1-min intervals. (in) After at least 10min a 1-ml portion of stock ammonia solution was injected into the solution in the cell by means of a syringe. 'The potentials were recorded until they were steady. ( v ) The electrodes were removed from the flow cell, rinsed with de-ionised water and placed in phosphate buffer solution. The temperature of the rinse water and the buffer had previously been brought to 30 "C. The potentials were recorded every minute for a t least 6 min after a steady value had been reached. (vi) The electrodes were rinsed with de-ionised water and placed in borax buffer solution.The potentials were recorded every minute for at least 6 min after steady values had been attained. (vii) The electrodes were rinsed with de-ionised water and returned to the flow cell. (viii) A new batch of working solution was connected to the pump and CDlAPS set to record at intervals of 2 h again. Step (i) was repeated on two other occasions during each week. For the next 12 weeks, only step (viii) was carried out, except in weeks 16, 20 and 25, when the full procedure was followed. After a further 11 months, during which time the electrodes were stored in de-ionised water, exclept for a few tests in dilute ammonia solution and one in acidic solution, the calibration was checked as in steps (v) and (vi). Once a week for the first 13 weeks the procedure below was carried out.January, 1979 FOR DETERMINING PH I N BOILER FEED WATER 65 Results In most instances the results are shown for only one electrode from each pair as electrodes of the same type gave very similar results.Exceptions to this rule are noted. Stability of the Standard Potential The e.m.f.s measured when the electrodes were immersed in the phosphate buffer solution were used to check the variation of the standard potential with time. Electrodes of the same type behaved similarly, but there were considerable differences between the rates of change of standard potential for electrodes of different types (Table 11). The standard potentials changed steadily from week to week, except for electrode B1, the standard poten- tial of which was constant for the first 7 weeks.The potentials of the type A electrodes changed in the opposite direction to the others. TABLE I1 RATES OF CHANGE OF STANDARD POTENTIAL Electrode Rate of change*/mV week-1 Al, A2 -0.2, -0.3 B1. B2 0.6t, 0.2 c1, c 2 0.5, 0.3 D1, D2 0.2, 0.2 E l , E2 0.1, 0.1 F1, F2 0.2, 0.1, * Average over 12 weeks. t The potential was constant for 7 weeks before it started to increase; the figure given is for the next 5 weeks. Because reference electrodes may not be truly constant, the observed changes in standard potential cannot be assigned exclusively to the behaviour of the glass electrodes. The differences between the changes in standard potential are real, however, as all the potentials were measured against the same reference electrode.Stability of the Slope Factor The slope factor was calculated from the difference between the e.m.f.s when the electrodes were immersed in the two buffer solutions. A selection of the results, expressed as a per- centage of the theoretical value at the same temperature, is shown in Table 111. TABLE I11 SLOPE FACTOR AS A FRACTION (yo) OF THE THEORETICAL VALUE* Electrode Week 1 3 5 7 9 11 13 16 20 25 73 A1 99.421. 99.641 99.271. 99.34 98.911. 99.051. 97.821. 98.691. 98.621. 98.251. 97.70t B1 99.85 99.85 99.20: 99.64 99.131. 99.051. 99.131. 98.761. 98.47t. 98.401. 98.071. c1 99.71: 99.711 99.85 99.201. 99.56: 99.20t 98.761. 98.05t 99.49: 98.911. 98.91f D1 100.29 99.85 100.29 99.491 99.271. 99.277 99.277 99.131. 98.83t 98.16t 98.981. El 99.71; 99.85 99.71: 100.36 99.341.99.201. 99.277 98.691. 98.767 98.761. 98.501. F1 99.93 99.93 99.85 99.93 99.341. 99.341. 99.341. 98.917 99.05t 98.83t 98.531. * Not significantly different from the theoretical value at the 95% confidence level, unless t Significantly different from the theoretical value a t the 99% confidence level. : Significantly different from the theoretical value a t the 95% confidence level. otherwise noted.66 MIDGLEY AND TORRANCE: ASSESSMENT OF GLASS ELECTRODES Analyst, VoZ. 104 All the electrodes showed a tendency for the slope to decrease with time, but the variations from week to week were small. The slope factors of all the electrodes except those of type A were essentially theoretical at the start of the test, but from the eighth week onwards all the electrodes deviated significantly from the theoretical value ; none, however, would have been rejected for normal use.Precision The precision of the e.m.f. measurement was estimated by measuring the e.m.f .s at intervals of 1 min, assuming that there were no changes in pH or temperature during the period of the test (6-10 min). Results for tests in different solutions, all performed on the same morning, are shown in Table IV. The standard deviations varied erratically from week to week, but there was no over-all tendency for the standard deviations to increase or decrease during the 25-week test. Table V shows an extract of the results for the weekly variations for the electrodes in the borax buffer solutions. TABLE IV STANDARD DEVIATIONS OF THE ELECTRODE POTENTIALS IN DIFFERENT SOLUTIONS Standard deviation/mV for a single result' in- Phosphate Electrode buffer t A1 0.43 A2 0.26 B1 0.38 c1 0.29 c 2 0.31 D1 0.37 D2 0.13 El 0.28 E2 0.47 F1 0.15 F2 0.10 Borax buffer t 0.33 0.44 0.30 0.37 0.35 0.42 0.18 0.2 1 0.43 0.15 0.20 Unstirred ammonia solution: 0.58 0.19 0.19 0.27 0.20 0.29 0.15 0.27 0.44 0.14 0.14 Stirred ammonia solution: 0.56 0.32 0.37 0.27 0.59 0.41 0.36 0.50 0.59 0.43 0.42 * Results taken from the ninth week of testing.t Five degrees of freedom. Nine degrees of freedom. The standard deviations observed in the phosphate buffer, borax buffer and unstirred working solution were generally similar on any one occasion. The standard deviations observed in the stirred working solution were generally 2-3 times larger than in the other solutions, the biggest proportional increases being found with the electrodes with the smallest standard deviations.STANDARD DEVIATIONS* (mV) IN BORAX BUFFER SOLUTIONS Week 1 3 5 7 9 11 13 16 20 25 A1 0.21 0.43 0.50 0.67 0.33 0.11 1.01 0.96 1.13 0.36 B1 0.20 0.26 0.46 0.26 0.30 0.28 0.12 0.23 0.25 0.26 (: 1 0.20 0.21 0.34 0.28 0.37 0.32 0.44 0.29 0.32 0.28 D1 0.15 0.14 0.39 0.55 0.42 0.08 0.08 0.08 0.08 0.19 E l 0.33 0.26 0.33 0.34 0.21 0.39 0.61 0.25 0.37 0.42 F1 0.15 0.14 0.12 0.34 0.15 0.14 0.22 0.18 0.17 0.18 * Standard deviation for a single result with five degrees of freedom.JnnNary, 1979 FOR DETERMINING PH IN BOILER FEED WATER 67 Electrodes with a low electrical resistance, k., types D and F, had lower standard devi- ations than the others, type D electrodes being slightly better than type F.The electrodes with relatively high impedances had, on average, very similar standard deviations, but the type A electrodes were less consistent than the others. Response Time The equilibrium response times varied in the range 1-7 min for almost all the electrodes. I t was assumed that the mixing characteristics of the cell were the same in each test, as the temperature, the arrangement of the electrodes, the volume of solution in the cell and the stirring rate were kept constant. Typical response curves are shown in Fig. 1. y: E LLi 7th week - s' 20th week - 0 4 8 0 4 8 Time after injectiodmin Fig. 1. Time response curves for pH electrodes: OI A l ; Al B1; A, C1; a. D1; ., El, 0, F1.There was a tendency for theiresponse times to increase as the tests progressed over the 25-week period, but the changes were irregular. The increases in response times were most pronounced for the C and D electrodes and least pronounced for the A and F electrodes. Electrode B2 had consistently the best response time before a faulty relay made measure- ments unreliable. Electrode E2 was markedly slower than the others, although El was among the fastest. The electrodes can be grouped in terms of their equilibrium response times, which include a mixing time of less than 1 min, as follows. Initially the B electrodes were faster, but the F and E electrodes were more consistent over the full period of the tests. Typical response times were 1-3 min. Electrode A1 gave response times of 3 4 min through- out the trial, but the others were fairly rapid at first (2-3 min), becoming slower (5 min) over the second half of the trial.Group 3: the response of electrode E2 was consistently the slowest (5-7 min) and did not change significantly during the trial. Group 1: B1, B2, F1, F2, El. Group 2: A l , A2, D1, D2, C1, C2. Bias The electrodes indicated different pH values for the approximately 0.5 mg 1-1 ammonia solution, which flowed through the flow cell at 9 ml min-l but was otherwise unstirred. Types A, C and D indicated lower pH values than average and types B, E and F higher than average values. The same pattern of electrode behaviour was observed with all batches of68 MIDGLEY AND TORRANCE ASSESSMENT OF GLASS ELECTRODES Analyst, vd.104 solution, regardless of any differences in pH between the batches, but the absolute value of the mean bias between electrodes varied cclnsiderably from batch to batch. The mean biases between electrodes were calculated with respect to one electrode of each type for each weekly batch of solution. These mean biases were significantly different from zero at the 5% level in almost all instances. Typical examples, measured with respect to electrode D2, are shown in Table VI for 6 consecutive weeks. TABLE VI MEAN BIAS (PH UNITS) OF ELECTRODES AGAINST ELECTRODE D.2 Week Electrode A1 B1 c 1 D1 El F1 Number of readings 7 8 -0.006* 0.124 0.017 0.036 0.093 0.029 68 9 -0.076 0.096 0.020 0.036 0.073 0.019 66 10 - a m 6 01.122 0.043 0.062 0.128 0.075 81 11 o.ooo* 0.221 0.045 0.068 0.183 0.098 56 12 o.ooo* 0.132 0.017 0.040 0.127 0.030* 70 -----l 13 0.042 0.200 0.063 0.105 0.208 0.105 71 * Not significantly different from zero a t the 5% level.When the dilute ammonia solution was stirred, the bias between the pH values indicated by the different electrodes was reduced; high and low pH values in the unstirred solution always showed a decrease and an increase, respectively, when stirring started. Results from a typical batch are given in Table VII. Over all the batches, the standard deviations for the determination of the mean pH of the results from the 12 electrodes varied from 1.4 x 10-3 to 6.1 x to 7.2 x Because of the good agreement between electrodes in stirred solutions, the mean of the values they indicated was taken as the best estimate of the true pH of the dilute ammonia pH unit in the stirred solutions, compared with from 1.6 x pH unit in the unstirred solutions.TABLE VII pH VALUES GIVEN BY DIFFERENT ELElCTRODES I N DILUTE AMMONIA SOLUTION Electrode A1 A2 B1 B2 c 1 c 2 D1 D2 E l E2 F1 F2 Mean Standard deviation** I Instirred solution* 8.656 8.603 8.727 8.694 8.648 8.633 8.668 8.562 8.712 8.728 8.658 8.729 8.667 0.055 Stirred solution* 8.690 8.650 8.678 8.684 8.694 8.644 8.690 8.655 8.684 8.648 8.686 8.648 8.671 0.020 Average bias? between pH indicated in unstirred solution and true pH -0.076: - 0.0695 N.S.7 N.S.7 - 0.115: -0.129; -0.103; -0.1875 0,103; 0.052j 0.058j 0.07611 * Results for the ninth weekly run: each p€I is the mean of 10 readings. t Average of 10 different batches; the true pH for each batch is the mean of all pH values in Significantly different from zero at the 59; level.$ Significantly different from zero a t the lT, level. 7 Non-significant. (1 Significantly different from zero a t the 0.1% level. ** Standard deviation for a reading by a single electrode (11 degrees of freedom). the stirred solution.January, 1979 FOR DETERMINING PH IN BOILER FEED WATER 69 solution. The errors of the pH electrodes in unstirred solutions were determined by com- paring the true pH with the individual pH values indicated immediately before stirring started. The means (over ten batches) of such errors are also shown in Table VII. Except for the type B electrodes, the errors in unstirred solutions were significantly different from zero.Fig. 2 shows, for one electrode of each type, the error in the unstirred solution plotted against the true pH. For the types C and D these errors increase with the pH itself (linear correlation coefficients of 0.94), but in the other instances no relationship is evident, except possibly for the type I;. 8.1 8.3 8.5 a. 7 Mean pH in stirred solution Fig. 2. Deviation of pH in unstirred solution from the mean pH in stirred solution : solid lines, correlation coefficient > 0.94; broken lines, correlation coefficient G0.60. 0, A l ; x , B1; A, C1; 0, D2; 0, El; A, F2. The experiments with stirred and unstirred solutions were repeated with neutral and acidic solutions. In de-ionised water the same pattern of electrode behaviour was observed as in dilute ammonia solution, i.e., low-reading electrodes indicated an apparent increase in pH when stirring started and conversely for high-reading electrodes. The standard devi- ations of measurements with a single electrode (9 degrees of freedom) were 0.048 and 0.012 pH unit in unstirred and stirred solution , respectively. In unstirred 10-3 moll-1 hydrochloric acid solution the electrodes showed far less bias between one another than in de-ionised water or dilute ammonia solution. In contrast to the effects of stirring the neutral and alkaline solutions, all of the indicated pH values changed in the same direction (lower) when the solution was stirred (cj. , Tables VII and VIII) , although these changes were relatively small. TABLE VIII pH VALUES GIVEN BY DIFFERENT ELECTRODES IN mol 1-1 HYDROCHLORIC ACID Electrode Unstirred solution Stirred solution A2 3.087 3.077 B1 3.106 3.087 c1 3.128 3.113 D1 3.112 3.096 E l 3.122 3.108 F1 3.111 3.101 Mean* 3.118 3.102 Standard deviation? 0.015 0.013 * Mean of results for 10 electrodes (Al, B2 excluded).t Standard deviation for a reading by a single electrode (nine degrees of freedom).70 MIDGLEY AND TORRANCE : ASSESSMENT OF GLASS ELECTRODES AnaZyst, VoZ. 104 Effect of Temperature The temperature dependence of the e.m.f. of a glass electrode cannot usefully be discussed without simultaneously considering the nature of the reference electrode, the pH of the solution in which it is immersed and the teimperature dependence of that pH. When an electrode is used to give direct meter readings of pH, the temperature compensation circuit of the pH meter will also influence the apparent temperature dependence of the electrode.It was considered that a study of the effect OE temperature on different glass electrodes used with an arbitrarily chosen reference electrode in a small number of particular solutions would not be a useful guide to the practical aspects of pH measurement at different tempera- tures. Attention was concentrated therefore on the recovery of electrodes from a rapid change in temperature, such as may occur in a process stream. This is a matter of general applicability and is independent of the other factors mentioned above. The electrodes were immersed in phosphate or borax buffer at 30 "C and allowed to reach a steady e.m.f. They were then placed (in blatches of six) in a second solution of the same buffer at 20 or 40 "C for 1 h, before being returned to the original solution at 30 "C.The reference electrode remained in the solution at 30 "C throughout. The e.m.f.s were recorded on the CDAPS system as the electrodes returned to their original temperature. All the electrodes regained their original potentials (A0.5 mV) within 30-60 min and, although the initial displacement of the e.1n.f. from the value at 30 "C varied between electrodes in the order A > C w D > E m F m B, for most purposes the differences between the electrodes were of no importance. Disciussion Stability The stability of both the standard potential and the slope factor, as determined by measure- ments with buffer solutions, were very good for all types of electrodes over a period of 17 months.The changes observed would not have produced a significant error (0.05 unit) in pH over a period of at least 4 weeks between standardisations. Changes in performance in the dilute ammonia solutions as the electrodes aged were hard to assess because of the large variations in precision and bias between weekly batches. The response times of the electrodes increased during the period of the tests, but the changes occurred slowly and were unimportant for the first 8-12 weeks of operation. The precision of measurements in buffer and dilute ammonia solutions did not change with time over a 6-month period. Performance The slope factors of all the electrodes were initially very close to the theoretical value and declined to 99% of the theoretical value over' a period of about 12 weeks.The precision of e.m.f. measurements was approximately the same in buffer and dilute ammonia solutions; a typical electrode had a. standard deviation of 0.5 mV (0.008 pH unit) for a single measurement, which is adequate for almost all industrial purposes, and the electrodes with low-resistance membranes (types D and F) had standard deviations as low as 0.15 mV. The response times of all of the electrodes were adequate for most industrial purposes even after operation for 6 months. The recovery of the electrodes from large changes in temperature (10 "C) was fairly satisfactory (30-60 min), but was better for types B, E and F than the others. The recovery time of a pH electrode is only one aspect of the effect of temperature changes on pH measure- ment and should not be the only consideration for choosing between electrodes; the nature of the reference electrode and the temperaturle compensation circuitry in the pH meter must be considered simultaneously if errors caused by temperature variations are to be mini- mised.In unstirred dilute ammonia solutions, the bias between electrodes could be as large as 0.3 pH unit (Table VI), but in stirred solutions the bias between individual electrodes was rarely above 0.05 pH unit and usually there was no significant bias from the mean of the pH values indicated by all the electrodes. 'The bias developed differently in alkaline andJanuary, 1979 FOR DETERMINING PH I N BOILER FEED WATER 71 acidic solutions (cf, Tables VII and VIII) and only a small part of it could be attributed to changes in standard potential, as the largest change in Table I1 is equivalent to only 0.015 pH unit per week. Pairs of similar electrodes showed similar deviations from the true pH, independently of their positions in the flow cell.It was inferred, therefore, that any local variations in flow or temperature within the flow cell were of little importance. These results show the importance of adequate stirring for measurements in dilute ammonia solutions; the type B electrodes were the least affected by this problem and the worst affected were type C and D electrodes. The significance of solution agitation for on-line industrial pH measurements arises in relation to the rate and turbulence of the flow of sample.Bias is further discussed below in terms of the glass composition. Glass Composition All were predominantly lithium oxide - silica glasses modified by small amounts of other oxides, of which the most important were those of tantalum or niobium. The electrodes with a negative bias in dilute ammonia solutions had the higher tantalum contents, but the bias could not be correlated directly with the tantalum content, presumably because of the modifying influence of the other oxides present. Electrodes that gave a negative bias also had slower responses than the others, but the stability of the standard potential and the slope factor could not be correlated with other properties. The short-term standard deviation depended on the total resistance of an electrode rather than on the glass composition, at least under the conditions tested in this work.Two mechanisms may be postulated to explain the effect of stirring on the bias, the positive and negative deviations from the true pH in alkaline solution and the difference of behaviour in alkaline and acidic solutions. Firstly, water hydrates the glass and extracts alkali metal hydroxides from the structure. The hydroxide may accumulate in pores near the surface of the glass and cause the electrode to indicate a higher pH than that of the bulk of the solution. Stirring will displace the accumulated hydroxide and the indicated pH will decrease towards that of the bulk solution. Secondly, hydroxide ion in solution can attack the silicate lattice, as follows: At the end of the tests, the electrode glasses were analysed.This process consumes hydroxide in the surface layers of solution adjacent to the glass and causes the electrode to indicate a lower pH than exists in the bulk of the solution. Stirring will supply more hydroxide and the indicated pH will increase towards that of the bulk solution. In acidic solution the first reaction should predominate, because the rate of the second reaction depends on hydroxide concentration. As expected from this hypothesis, the indicated pH decreased when the acidic solution was stirred and this occurred with all of the electrodes. Even in the unstirred dilute ammonia solutions the type B, E and I-; electrodes gave high pH readings, showing that the first reaction was still dominant for these electrodes. The types A, C and D electrodes, however, gave low pH readings, indi- cating that the second had become more important. The bias between electrodes in slowly flowing solutions can, therefore, be explained qualitatively as being caused by the relative susceptibilities of the electrodes to both mechanisms under a particular set of solution conditions. Conclusions All the electrodes tested had nearly theoretical responses and for most industrial purposes any difference in performance between them would be of no importance. If, however, the sample was unstirred, poorly buffered and either static or flowing only very gently, errors could reach a level of 0.2 pH unit. Under the conditions of the tests, electrodes of types A, B, E and F were more satisfactory than the others in this respect. Low-resistance electrodes had no significant advantages over general-purpose electrodes for the application to power station waters, although they were capable of slightly higher precision.72 MIDGLEY AND TORRANCE The good performance obtained from a variety of glass electrodes, in contrast to the poor performance of some reference electrodes,l confirms power station experience that reference electrodes are the less satisfactory half of industrial pH cells. This work was carried out at the Central Electricity Research Laboratories and is pub- lished by permission of the Central Electricity Generating Board. Reference 1. Midgley, D., and Torrance, K., Analyst, 1976, 101, 833. Received July 25th, 1978 Accepted August 14th, 1978