Analyst, February, 1968, Vol. 93, $9. 83-92 83 A Solvent-extraction Absorptiometric Method for Determining Nickel in Boiler Feed-water BY A. L. WILSON (Central Electricity Research Laboratories, Cleeve Road, Leathevkead, Surrey) A method has been developed for determining nickel in boiler feed-water. The nickel complex with furil a-dioxime is extracted into chloroform, and is then determined absorptiometrically. The criterion of detection was about 0-3p.g of nickel per litre, and the standard deviation of results varied from about 0.15 to 0-3 pg of nickel per litre in the range of concentrations 0 to 50 pg of nickel per litre; within this range, the calibration graph was linear. No appreciable interference u 7 a s caused by other impurities likely to be present in feed-water, and a batch of ten samples could be analysed in about 24 hours.IMPURITIES entering the feed-water of modern high pressure boilers by corrosion in the condenser and feed-system may initiate corrosion failures in boiler tubes. Such failures result in costly outages, and, accordingly, one of the aims of chemical control in power stations is to control the concentrations of corrosion products in feed-water. Nickel constitutes an important proportion of certain alloys that are used in feed systems. Therefore, a method for determining nickel in boiler feed-water was required for use in power stations of the Central Electricity Generating Board. Because of the minute concentrations of nickel normally present (0 to 10 pg per litre), the method had ta be capable of giving results with a standard deviation no greater than 1 pg of nickel per litre. It was decided that a solvent-extraction absorptiometric method, in which the reagent furil a-dioxime was used, should be suitable.I have previously reported1 the results of a detailed investigation of the effects of several factors involved in the formation and solvent extraction of the nickel complex with this oxime. The work referred to showed the technique to be suitable for a precise and sensitive method. The purpose of this paper is to present full details of the method that has been developed and the tests made of its performance. EXPERIMENTAL REAGENTS, APPARATUS AND TECHNIQUE- All of the experimental details were exactly as previously reported,l and measurements of optical density were always made at 435mp.Distilled water that had been de-ionised with a laboratory-scale mixed-bed column was used for all tests, except where otherwise stated. BASIS OF THE METHOD- The previous results1 had shown suitable conditions for the formation and solvent extraction of the nickel complex. However, the analysis of feed-water requires certain precautions that affect the details of the analytical method. For example, earlier work* on the determination of traces of copper in feed-water had shown the necessity for collecting samples into sufficient hydrochloric acid to give a final acidity of 0.1 N. This acidification was required to prevent losses of copper. As a safeguard, the same acidification technique was adopted when determining nickel. Further, at this stage of the investigation, there was the possibility that some specific treatment might be required to dissolve particulate forms of nickel present in samples; such treatment might have produced samples of varying acidity, Therefore, it was decided to control the pH and ammonia concentration (both parameters are important1) of the solution before the solvent-extraction stage.This was carried out by first adjusting to a pH of about 8-5 with sodium hydroxide solution and dilute hydrochloric acid, and then adding a fixed optimum amount of ammonia solution. The presence of about 0.1 N sodium chloride in the final solution (from the neutralisation of the hydrochloric acid added initially) should aid the extraction of the nickel complex.184 WILSON A SOLVENT-EXTRACTION ABSORPTIOMETRIC METHOD [Analyst, VOl 93 As relatively large amounts of iron would sometimes be present in samples, tartrate was added to the sample before neutralisation to prevent precipitation of the iron.Finally, to prevent interference from i r ~ n ( I I ) , ~ a small amount of hydrogen peroxide was added to the acidified sample to cause oxidation to iron(II1). Hydrogen peroxide, rather than the potassium dichromate used by Taylor,8 was used because the dichromate reduced the optical density due to a given concentration of nickel. The technique finally devised was then as given under Method, and the tests of the effects of several experimental factors not considered previously are reported as follows. EFFECT OF VARIATIONS IN THE AMOUNTS OF REAGENTS- The effects of variations in the amounts of sodium chloride (from the neutralisation of hydrochloric acid), tartrate, furil cc-dioxime and ammonia have already been reported.l Addition of ten times the amount of the phenolphthalein solution specified in the method caused no significant effect, i e ., less than 0.4 pg of nickel per litre. Addition of 0.1 or 0.3 ml of the hydrogen peroxide solution did not significantly affect the results obtained from blank determinations or solutions containing 20 pg of nickel per litre. Finally, during adjustment of the pH of the sample, before the solvent extraction, the addition of 5 drops in excess of the dilute hydrochloric acid (1 + 99) caused no significant effect on the results for blank determinations and for solutions containing 20 pg of nickel or 100 pg of copper per litre.Thus, none of the amounts of the reagents appeared to be critically important. EFFECT OF TEMPERATURE- Solutions containing either 0 or 20pg of nickel per litre were analysed, as described under Method, except that the temperatures of the solutions during the solvent extraction were adjusted to three different values. The temperatures of the chloroform extracts during measurement of their optical densities were within 0.6' C of ambient temperature. Duplicate determinations were made for each condition tested, and the mean results given in Table I show no significant effect of temperature between 21" and 29" C. The efficiency of extraction of the nickel was apparently reduced slightly at the lower temperatures of 12-5" to 14.5' C.TABLE I EFFECT OF TEMPERATURE ON THE EXTRACTION OF NICKEL Mean optical density Mean temperature of 7- sample during extraction, "C Blank hrckel, 20 pg per htre* 12.5 to 14.6 0.024 0.342 21 0.036 0.361 29 0.034 0.352 * These results have been corrected for the blank. The effect of temperature on the optical density of the nickel furil a-dioximate in chloro- form was determined approximately by either warming or cooling a chloroform extract, and then measuring its optical density as the temperature approached ambient. The results indicated a linear relationship between optical density and temperature in the range 18" to 30" C, the optical density at 20' C changing by about 0.3 per cent. per "C. DETERMINATION OF THE CONCENTRATION OF NICKEL IN THE WATER USED FOR BLANK DETER- A technique that seemed suitable for this determination was to analyse 200 and 400-ml portions of the water in exactly the same way, and to equate the difference between the two optical densities to the optical density due to nickel in 200 ml of the water.Tests showed, however, that appreciably greater amounts of chloroform dissolved in the 400-ml than the 200-ml samples. Thus, for a given amount of nickel, the optical density was greater for 400 than 200-ml samples. This effect was overcome by measuring the volume of the extracts before measuring their optical densities. An allowance for the effect of this difference in volumes can then be made as described under Method. For this technique to give accurate results, it is also necessary that the optical density of furil a-dioxime in the chloroform is not affected by the volume of the aqueous phase.Tests in which 200 and 400-ml portions of water were analysed, after addition of 5 and 10 ml MINATIONS-February, 19681 FOR DETERMINING NICKEL IN BOILER FEED-WATER a5 of the furil oc-dioxime reagent solution, showed that the optical density due to the oxime was 0-003 for both 200 and 4Wml samples. Finally, it is also essential that small concentrations of nickel are quantitatively extracted from 400-ml samples. To check this, duplicate 400-ml portions of water, to which either 0 or 0*80pg of nickel was added, were analysed. The mean recovery of the added nickel was 0.84 0.04pg (95 per cent. confidence limits). It was concluded, therefore, that the technique described under Method was satisfactory for determining the nickel content of the water used for blank determinations. DISSOLUTION OF FORMS OF NICKEL INSOLUBLE IN FEED-WATER- Tests were made to determine whether samples were likely to contain insoluble forms of nickel that do not react directly with furil a-dioxime.For this purpose, samples of con- densate and feed-water from one power station were each analysed by two methods. In method A, 200-ml portions of the acidified samples were analysed, as described under Method, except that the initial boiling of the sample was omitted. In method B, the initial boiling was continued until the volume of the sample had been reduced to about 6 ml. This residual volume was diluted with water (containing less than 0-1 pg of nickel per litre) to a final volume of 200 ml, and this solution was then analysed as before.Four portions of each sample were analysed by each method, and blank determinations by both methods were carried out. The results are given in Table 11 and indicate that these particular samples contained in- appreciable amounts of insoluble forms of nickel. Both of these samples had been taken during steady operation of the power station, and, by analogy with iron and copper, it was thought that greater nickel concentrations might be encountered during periods while the turbine and feed-system were being brought into operation. Such greater concentrations might be associated with insoluble forms, and other tests were made to check this possibility. The tests were kindly made by the Scientific Services Department of the South Eastern Region of the Central Electricity Generating Board.Two samples of feed-water were collected during two different start-up operations at the same power station as for the previous samples. The experimental designs were exactly as for the previous tests; methods A and B were used, but the method proposed in this paper (k, samples just brought to the boil and then cooled before analysis) was also used. For method A, samples were analysed within 1 hour of being collected. The results of these tests are also given in Table 11. TABLE I1 EFFECT OF PRE-TREATMENT OF SAMPLES ON APPARENT NICKEL CONTENT Apparent nickel content,* pg per litre, as determined by- - \ Type of sample method A method B proposed method Condensate during steady operation .. . . 1.3 1.6 - Feed-water during steady operation (1) . . 1.6 1.4 - Feed-water during start-up (2) . . 7.6 13.3 - Feed-water during start-up (3) . . 8-8 11.1 10.8 * 96 per cent, confidence limits for these results are about f0-3 pg of nickel per litre. The results indicate that the samples taken during start-up contained small amounts of insoluble forms of nickel that were dissolved by method B. The results obtained by the proposed method were not significantly different from those obtained by method B on the one sample that was tested. Other tests by the South Eastern Region Scientific Services Department showed that the apparent nickel content given by method A increased slowly as the time between sampling and analysis increased.For example, results of 10-0 and 12-45 pg of nickel per litre were obtained after 1 and 3 days, respectively, for feed-water (2). Greatest interest attached to the precise and accurate determination of nickel during steady operation of power stations. Under these conditions, subsequent work has shown that the nickel content of condensates and feed-waters from modern stations is almost invariably less than 6 pg per litre, and is often less than 2 pg per litre. On this basis, it appears that the proposed method should provide a sufficiently rigorous pre-treatment of the sample. The results also showed that the evaporation treatment of method B could be used, if required for any special purposes, without any important loss of precision.86 WILSON : A SOLVENT-EXTRACTION ABSORPTIOMETRIC METHOD [Analyst, Vol.93 METHOD REAGENTS- Except where otherwise stated, analytical-reagent grade chemicals should be used whenever possible. Water-Use water with a small nickel content (preferably less than 0.5pg per litre) for preparing reagents and for blank determinations. Water containing less than 0.2 pg of nickel per litre has been consistently obtained by passing distilled water (from a Manesty still) through a laboratory-scale, mixed-bed, de-ionisation column. Determine the nickel content of the water used for blank determination as described under Procedure. Ethanol-Industrial methylated spirit, 74" O.P. Hydrochloric acid, about 5 r;-Dilute 500 ml of hydrochloric acid (sp.gr. 1.1s) to 1 litre with water.Sodium hydroxide solution, 2.5 N-For all of the results quoted in this paper this reagent was prepared by diluting concentrated volumetric sodium hydroxide solutions (British Drug Houses Ltd) . Preparation of the reagent from sodium hydroxide pellets (analytical-reagent grade) tended to give slightly greater optical densities for blank determinations. Store this reagent in a sealed polythene bottle. Hydrogen peroxide solution (1 + 9)-Dilute 10 ml of hydrogen peroxide (100 vol) to 100ml with water. Prepare this solution freshly each week. Phenolphthalein solution, 0.5 per cent. w/v. Sodiunz potassium tartrate solution, 10 per cen,t. w/v-Dissolve 25 g of sodium potassium tartrate tetrahydrate in water, and dilute with water to 250ml. Prepare this solution freshly each week.Hydrochloric acid (1 + 99). Furil a-dioxime solution, 0.15 per cent. w/v-Dissolve 0.75 g of furil a-dioxime in industrial methylated spirit, and dilute to 500 ml with industrial methylated spirit. Store in a sealed, glass bottle, and keep the bottle in the dark when not in use. This solution has been found to be adequately stable for at least 8 months. Tests showed no significant (95 per cent. confidence limits) difference in the optical densities due to a given amount of nickel when furil a-dioxime from two different suppliers was used. Ammonia solution, 2 ih-Dilute ammonia solution (sp.gr. 0.88) with water so that the normality of the final solution is 2.0 0-1 N. Store the solution in a sealed, glass bottle; its concentration has been found to change by less than 1 per cent. after 2 weeks.Chloroform. Stapzdard nickel solution A-Dissolve 1-000 g of pure nickel metal (rod, foil or wire) by warming with 20 ml of dilute nitric acid (1 + 3). When dissolved, add 10 ml of hydrochloric acid (sp.gr. 1.18) and 100ml of water, cool, and dilute with water to 1 litre in a calibrated flask. 1 ml of solution = 1000 pg of nickel. Determine the normality of the diluted acid. This solution was found to remain stable for at least 6 months. Standard nickel solution B-Dilute 20 ml of standard nickel solution A with 10 ml of hydrochloric acid (sp.gr. 1.18) and water to 1 litre in a calibrated flask. This solution was found to remain stable for at least 6 months. 1 ml of solution = 20 pg of nickel. Standard nickel solution C-Dilute 20ml of standard nickel solution B with water to 1 litre in a calibrated flask.Prepare this solution freshly each day as required. APPARATUS- 1 ml of solution = 0.4 pg of nickel. Care is required to minimise contamination of apparatus when not in use. Pyrex-glass conicaZJEasks, 500-wzZ capacity-Soak the flasks overnight in a cleaning solution of chromic acid, and then wash well with water. Add 250ml of dilute hydrochloric acid (1 + 1) to each flask, and heat the flasks until only 5 to 10 ml of acid remain. Wash the flasks well with water, and store with 50-ml beakers inverted over the necks of the flasks. Pyrex-glass beakers, 50-ml ca#acity-Clean the beakers in the same way as the 500-ml flasks, but with 25 instead of 250ml of the dilute hydrochloric acid.Pyrex-glass separating fzmnels, 500-ml capacity-Soak the funnels overnight in a cleaning solution of chromic acid, and then wash well with water. Before using the funnels for analysis,February, 19681 FOR DETERMINING NICKEL IN BOILER FEED-WATER 87 carry out a blank determination in each funnel, discarding the contents of the funnels after the solvent-extraction stage. Finally, wash each funnel well with water of low nickel content. Do not grease the taps of the separating funnels. Pyrex-glass, stoppered graduated cylinders, 25-ml capacity-Soak the cylinders and their stoppers overnight in a cleaning solution of chromic acid, and then wash well with water. Dry in an oven. Polythene bottles-These bottles were found to be suitable for collecting samples, and any convenient size may be used.Soak the bottles in dilute hydrochloric acid (1 + 1) for 2 to 3 days, and then wash well with water. PROCEDURE- Sample collection-Place sufficient hydrochloric acid (about 5 N) into a polythene bottle to ensure that the final acid concentration, when the sample has been collected, will be 0.1 N (k 0.005 N). Samples acidified in this way were found to be stable for at least several weeks. Make allowance for the acid contained in the sample when measuring the volume required for analysis. AnaZ_vsis of samples-Transfer a volume of acidified sample, equivalent to 200 ml of feed- water, into a 500-ml conical flask. Cover the neck of the flask with an inverted SO-ml beaker, and heat the flask until the contents just begin to boil.Cool the flask, and transfer its contents into a 500-ml separating funnel. Wash the flask with two 5-ml portions of water, and add these washings to the separating funnel. The temperature of the sample should now be between 16' and 30" C. Add to the sample 4 drops of the dilute hydrogen peroxide solution, 2 drops of the phenolphthalein solution and 5 ml of the sodium potassium tartrate solution. While swirling the separating funnel, add the 2.5 N sodium hydroxide solution until the phenolphthalein just turns pink. Then, with swirling, add dilute hydrochloric acid (1 + 99), dropwise, until the pink colour is just discharged. Add 5.0 ml of the furil a-dioxime solution, followed im- mediately by 25 ml of 2 N ammonia solution, and swirl the funnel for a few seconds. Add 15.0 ml of chloroform froin a burette, shake the funnel gently for a few seconds and carefully release the excess of pressure in the funnel.Shake the funnel vigorously (about 200 shakes per minute) for 1 minute, and allow the two phases to separate for at least 5 minutes. Do not alfow direct sunlight to fall on the chloroform extract. Run a small portion of the chloroform phase through a 9-cm Whatman No. 541 filter- paper, discarding the filtrate. Pass the remaining chloroform through the same filter-paper into a clean, dry 4-cm cuvette. Discard the filter-paper. Measure the optical density of the chloroform extract at 435 mp against a 4-cm reference cuvette containing pure chloro- form; the measurement should be made within 2 hours after the extraction.Let the measured optical density be As. After measurement, discard the contents of the sample cuvette, rinse it well with pure chloroform, and allow it to dry before filtering the next extract into the cuvet te. BZnnk determinations-A blank determination should be made with each batch of sample determinations. For this, place 200 ml of water (low and known nickel content, see below) in a 500-ml conical flask, and add sufficient hydrochloric acid (about 5 N) to ensure that the final acid concentration is 0.1 N (k 0.005 N). Analyse this solution exactly as described above. Determination of nickel in the water used for blank determinations-Place 200 ml of the water used for the blank determination in a 500-ml separating funnel and 400ml of the same water in another funnel.Add to each funnel the same volume of about 5 N hydrochloric acid as that added for the blank determination. Treat both funnels exactly as described under Analysis of samples, beginning with the addition of hydrogen peroxide. On completing the solvent extraction, allow at least 30 minutes for the two phases to separate. Pass as much as possible of the two chloroform extracts through 9-cm Whatman No. 541 filter-papers (previously moistened with a few drops of chloroform, so that no excess drops of chloroform remain in the filter-paper or filter funnel) into 25-ml graduated cylinders. Allow all of the chloroform extracts to drain through the filter-papers, insert the stoppers of the cylinders, and measure the volumes of the two extracts. Let the measured volumes for the 200 and 400-ml samples of water be VT and Vp, respectively.Measure the optical densities of the extracts as described above. Let the measured optical densities for the 200 and 400-ml samples of water be AT and AF, respectively. Let the measured optical density be A g .88 for the blank determination is given by the expression- WILSON : A SOLVENT-EXTRACTION ABSORPTIOMETRIC METHOD [A .nabst, Vol. 93 CaZczcZation ofreszclts-The optical density, Ac, due to nickel in the 200 ml of water used The optical density, .AR, due to nickel in the sample is given by the expression- and the concentration of nickel in the sample can then be determined from AK and the calibration graph. PREPARATION OF CALIBRATIOX GRAPH- Into a series of separating funnels, transfer 200, 198, 195, 190, 185, 180 and 175ml of water of low nickel content, and then add 0.00, 2-00, 5-00, 10.00, 15-00, 20*00 and 25.00 ml, respectively, of standard nickel solution C.Add sufficient hydrochloric acid (about 5 N) to each funnel to ensure that the final acid concentration is 0.1 N (2 0-005 N). Treat these solutions exactly as described under Analysis of samples, beginning with the addition of hydrogen peroxide. Repeat this series of determinations until the calibration graph is defined with the desired precision. Subtract the mean optical density for the solution with no added nickel from the mean optical densities of each of the other solutions, and plot the corrected optical densities against the concentration of nickel added to the sample.The calibration graph was linear to at least 50 pg of nickel per litre in samples when 4-cm cuvettes were used, and to at least 230 pg of nickel per litre for l-cm cuvettes. When measurements were made in 4-cm cuvettes with the Spekker absorptiometer and Ilford No. 601 filters, the calibration graph was linear to 30pg of nickel per litre, but the departure from linearity at 50pg of nickel per litre was only 8 per cent, Optical densities measured with No. 601 filters were about 80 per cent. of those obtained by measuring at 435mp. SENSITIVITY AND PRECISION- To determine the precision of the basic solvent-extraction absorptiometric technique, the initial boiling of the sample was omitted. For these tests, on each of 10 days, duplicate analyses were made at concentrations of about 0, 2, 10, 50 and 230 pg of nickel per litre, as described under Preparation of calibration graph; 4-cm cuvettes were used for the first four solutions, and l-cm cuvettes for the last. These analyses were made during a period of 4 weeks, Analytical results are obtained by subtracting the blank value from that of the sample; the precision of these corrected results was, therefore, calculated by allowing for the variability of both the blanks and the samples.For all four levels of nickel there were no statistically significant (95 per cent. confidence limits) “between-days” variations; all of the results for each level of nickel were, therefore, used to calculate the standard deviation of the deter- minations. The “within-days” standard deviation of the blank value was also calculated.These standard deviations are shown in Table 111. The amount of nickel extracted was calculated from the average volume of an extract (13.2 ml) and the molar extinction coefficient1 (1.82 x lo$). TABLE I11 SENSITIVITY AND PRECISION OF NICKEL DETERMINATIONS A R = A S - A B + A C , RESULTS Concentration of nickel added, pg per litre 0.00 1.87 9-32 46.70 233.9 Mean optical density*- corrected for blank value nickel per litre per 10 pg of - --t 0.0340 0.186, 0.176, 0.188, 0.876, 0.187, 1.098 0.046, Amount of nickel extracted, per cent. 98.9 99.7 99.4 99.6 - Standard deviation, pg of nickel per litre 0.1 4 0.13 0.32 2.1 0.11: Degrees of freedom 9 19 19 19 19 * 4-cm cuvettes were used for all of the solutions, except that containing 233.9 pg of nickel t The mean optical density of the blank solutions was 0.014,.$ This is the “within-days” standard deviation only. per litre, for which l-cm cuvettes were used.February, 19681 FOR DETERMINING NICKEL IN BOILER FEED-WATER 89 The precision of anallysing actual samples was determined by a similar experiment with standard solutions containing 0 and 50 pg of nickel per litre and samples of condensate and feed-water from power stations. For these tests, the method given under Analysis of samples was followed exactly, and one batch of analyses was made on each of 5 days. The results were analysed, as before, and are summarised in Table IV. TABLE IV PRECISION OF ANALYSING POWER Mean concentration found, Sample pg of nickel per litre Blank .. .... 49.8 Condensate .. .. 0.0 Feed-water .. .. 10.6 - Nickel, 60.1 pg per litre . . STATION WATERS Standard deviation, Degrees of pg of nickel per litre freedom 0-14* 6 0-28 0 0.20 9 0.22 9 * This is the “within-days” standard deviation only. BIAS- AutaZysis of a feed-water-Twelve portions of a feed-water were analysed, but the equivalent of 1Opg of nickel per litre was added to six of these portions before analysis. Substance Copper(I1) . . .. Zinc(11) .. .. Calcium(I1) .. Iron(II1) .. .. Manganese(I1) . . Cobalt(I1) . . . . Magnesium(I1) . . Chromium(II1) . . Aluminium(II1) . . Molybdenum(V1) . . Vanadium(V) . . Vanadium(1V) . . Titanium(1V) . . Tungsten(V1) . . Silicate . . .. Orthophosphate . . Nitrate . . .. Fluoride . . .. Fulvic acidt . . Detergentst . .Cyclohexylamine . . Tin(I1) . . .. Hydrazine . . .. Blorpholine .. Octadecylamine . . EFFECT Concentration of substance, pg per litre 1000 100 1000 1000 100 100 10,000 1000 100 1000 100 1000 1000 1000 100 10,000 10,000 1000 2500 0000 10,000 1000 100,000 10,000 100,000 10,000 10,000 1000 TABLE V OF OTHER SUBSTANCES Nickel recovered, pg per litre* 0.0 Gg per litre added 2.5 0.3 0.0 2.4 0.0 -0.1 0.1 0.3 0.0 - 0.1 - 0-2 0.0 0-2 0.1 0.1 - 0.2 0.2 - 0.1 0.6 -0.1 - 0.2 0.0 0.2 0.4 - 0.1 - 0.2 - 0.1 9.3 pg per litre 93.4 pg pe; litre added added 11.3 95.4 9-6 94.0 9.2 92.5 10.2 93.8 9.5 95-4 9-3 93.1 9.7 Not tested 9-5 8.4 9-1 6.3 9.0 9.3 9.1 9.3 9.6 86.9 90.5 92-8 63.7 94.9 94.2 Not tested Not tested Not tested 9.4 93.7 9.6 Not tested 9.7 94-4 9.4 9.1 9.3 9.0 9.9 9.0 9.3 9.1 Not tested 93.4 Not tested 92.3 Not tested 91.3 Not tested 92.1 * The ranges of recoveries expected, assuming no interference from the other substances, were (95 per cent.confidence limits)- 0.0 f 0-4 (when 0.0 pg of nickel per litre was added) ; 9.3 f 0.4 (when 9.3 pg of nickel per litre was added) ; 93-4 f 2.4 (when 93.4 pg of nickel per litre was added). t Prepared as described previously.4 1 Omo, Daz, Surf, Dreft, Blue Tide and Quix (1000 pg of each per litre) were used.90 WILSON : A SOLVENT-EXTRACTION ABSORPTIOMETRIC METHOD [ArtdySt, VOl. 93 The difference between the two sets of results was equivalent to 10.1, pg of nickel per litre, with 95 per cent. confidence limits of 3-0.2. Thus, on this particular sample, the recovery of the added nickel was satisfactory. Tests made by other laboratories within the C.E.G.B.have also shown satisfactory recoveries for many different feed-waters. Eflect of other substances-The effects of some other substances were investigated at each of three concentrations of nickel, Le., 0 and about 10 and 90 pg per litre. The analyses were made singly by the proposed method, omitting the initial boiling of the sample; l-cm cuvettes were used for the solutions containing 90 pg of nickel per litre. Solutions containing nickel alone were analysed with each batch of analyses, and the results are shown in Table V. ROBUSTNESS OF THE METHOD- To test the sensitivity of the method to deviations from the recommended conditions, solutions containing known amounts of nickel were analysed by the proposed procedure, except that the amounts of reagents added were varied in the direction thought most likely to cause incomplete extraction of the nickel.The amounts used were: 8 drops of the hydrogen peroxide solution, 4 drops of the phenolphthalein solution, 10 ml of the tartrate solution, 10 drops in excess of the dilute hydrochloric acid (1 + 99), 2.5 ml of the furil a-dioxime solution and 30 ml of the 2 N ammonia solution. All of the solutions were shaken for I minute during the solvent-extraction stage, but both the normal shaking rate (about 240 shakes per minute) and a slower rate (about 120 shakes per minute) were used. One analysis was made for each condition tested, and the results are given in Table VI. TABLE VI EFFECT OF DEVIATIONS FROM THE RECOMMENDED PROCEDURE Shaking rate, Concentration of nickel Concentration of nickel* shakes per minute added, pg per litre found, pg per litre -240 9.3 9.3 93.4 93.0 -120 9.3 9.0 93.4 90.8 * The optical densities of the blank determinations at the two different shaking rates were 0-006 and 0-007.CALIBRATION GRAPH- The results in Table I11 indicate that the calibration graph was linear in the range 0 to 230 pg per litre. The equation of the graph, for 4-cm cuvettes, was: C = 53.5 A where C is the concentration of nickel in pg per litre and A is the optical density due to nickel in the sample. Although 50 pg per litre is regarded as the normal upper limit of the method when 4cm cuvettes are used, the results show that the range of the method can be simply extended to at least 230 pg per litre by using l-cm cuvettes.The results in Table I11 also show that one solvent extraction was sufficient to give essentially complete extraction of the nickel from the sample. PRECISION- The standard deviations of optical density measurements were determined separately by repeated measurements of portions of chloroform solution containing the nickel furil a-dioximate at various concentrations. From these determinations the standard deviations expected for analytical results from measurement errors alone were calculated. These standard deviations ranged from 0.08 to 0.19 pg per litre at concentrations of 2 and 50 pg per litre, respectively. Comparison of these calculated values with those found experimentally (see Tables I11 and IV) shows that the measurement errors were not the sole source of random error.Thus, errors arising during the chemical treatment of the sample were also of im- portance; calculations show that these errors were of about the same magnitude as the measurement errors. The precision of the solvent-extraction technique (Table 111) was satisfactory for my purpose. The results in Table IV show that the precision was not adversely affected by the initial boiling of the sample and that the precision for samples of condensate and feed- water was also satisfactory. DISCUSSION OF THE METHODFebruary, 19681 FOR DETERMINING NICKEL IN BOILER FEED-WATER 91 Since this method was developed, independent determinations of the standard deviations of results have been made in many laboratories of the Central Electricity Generating Board.In one series of tests in five other laboratories values were obtained ranging from 0.11 to 0.59 pg per litre for the standard deviation of results from standard solutions containing 50 pg of nickel per litre. Thus, there appears to be no difficulty in obtaining precise results with this method, as would be expected from the results given under Robustness of the method. The sensitivity of the method can be improved, if desired, by extracting the nickel from larger volumes of the sample. The effect of such a modification has not been examined in detail, but results (see Determination of the concentration of nickel in the water used for blank determinations) indicate that a useful gain in precision might be achieved.CRITERION OF DETECTION- If the criterion of detection5 (95 per cent. confidence) is taken as 2-33 times the “within- batch” standard deviation of blank determinations, it is equivalent to about 0.3 pg of nickel per litre. Again, it should be possible to decrease this criterion of detection by extracting larger volumes of the sample. BIAS- The results given under Bias and from other laboratories of the Central Electricity Generating Board show no signs that the efficiency of extraction of nickel is adversely affected by other impurities present in condensates and feed-waters. The results in Table V show that several of the substances tested can significantly affect analytical results, although none of the effects was important in feed-water analysis, because the necessary concentrations of the interfering substances were greater than would normally occur.The effect produced by the combinations of nitrate, fluoride, fulvic acids and deter- gents is probably caused by partial extraction of fulvic acids, which absorb appreciably at 435 mp. Iron(II1) did not significantly affect results for solutions containing up to 10 pg of nickel per litre. The negative effect of iron at the level of 93 pg of nickel per litre seems likely to be caused by competition between iron and nickel for the furil a-dioxime; even this effect became non-significant for 1OOpg of iron per litre. The effect of chromium is interesting as 1000 pg per litre did not significantly affect the blank, but depressed the result for both 9.3 and 93 pg of nickel per litre by about 30 per cent.As this effect became non- significant at a chromium concentration greater than that expected in feed-water, no further investigation of the nature of the effect was made. The most important interfering substance is copper. The results in Table V show that 10o0 pg of copper per litre was equivalent to about 2 pg of nickel per litre, and that this effect was approximately directly proportional to the concentration of copper. As samples usually contain only a few micrograms of copper per litre, the effect of copper is normally un- important. However, tests of the method in other laboratories of the Central Electricity Generating Board showed that the effect of copper varied a little with different batches of the furil cc-dioxime solid reagent.The greatest effect found was that IOOpg of copper per litre was equivalent to 1.1 pg of nickel per litre. This effect is still unimportant for normal feed-water analysis, but it could cause unexpected errors in solutions containing abnormally large concentrations of copper. For the most precise work with solutions with high ratios of copper to nickel, it seems desirable to check the magnitude of the interference caused by copper for each batch of solid furil a-dioxime. The results in Table I1 indicate that there is little difficulty in converting insoluble forms of nickel (present in the original sample) to forms that react with furil a-dioxime. For many samples, the initial boiling of the sample is likely to be unnecessary, but it is included in the method as a simple precautionary measure. Nevertheless, the detailed tests have been made on relatively few actual samples, and the possibility should be borne in mind that certain samples may contain forms of nickel requiring a more vigorous pre-treatment to dissolve them. It is normal practice in the Central Electricity Generating Board to set the acidified samples aside for 1 day before analysis, to allow partial dissolution of insoluble forms of nickel, before applying the proposed method. This procedure is a useful precaution when analytical results are not required urgently but is unlikely to have any important beneficial effect.92 WILSON SPEED OF ANALYSIS- The method has been shown to be simple to carry out, and a batch of 10 samples may be analysed in about 24 hours. This time would be reduced to about 1Q hours if the initial boiling of the sample were omitted. This paper is published by permission of the Central Electricity Generating Board. I thank Mr. D. Leighton for his co-operation in independently investigating the need for pre- treatment of samples before solvent extraction. I also thank many colleagues in the Central Electricity Generating Board who took part in independent tests of the method, and some of whose results have been quoted in this paper. REFERENCES 1. Wilson, A. L., in Shallis, P. W., Editor, “Proceedings of the SAC Conference, Nottingham, 1965,” 2. - , Analyst, 1962, 87, 884. 3. Taylor, C. G., Ibid., 1956, 81, 369. 4. Wilson, A. L., J . Appl. Chem., 1959, 9, 501. 6. Roos, J. B., Analyst, 1962, 87, 832. W. Heffer & Sons Ltd., Cambridge, 1965, p. 361. Received June 29th, 1967