DECEMBER 1979 The Analyst Vol. 104 No. 1245 Absorptiometric Determination of Low Oxygen Concentrations in Power-station Waters Part I. Manual Method G. I. Goodfellow and H. M. Webber Central Electvicity Research Laboratories, Kelvin Avenue, Leatherhead, Suvrey, KT22 7SE A simple, fast method for determining low concentrations of oxygen in power-station waters has been developed, based on the reaction of dissolved oxygen with the leuco-base of methylene blue to give a soluble blue oxidation product the absorbance of which is a function of the oxygen concentration. A special glass cell has been devised, which acts sequentially as a sample- collection vessel, a reaction vessel and a spectrophotometric cuvette. The cell design permits the easy addition of the leuco-base and also the air- saturated water used for calibration.A novel technique of “zero-time extrapolation” for the determination of the reagent/cuvette blank circumvents the difficulty of making this measurement with oxygen-free water. The calibration graph is linear up to 50 pgl-l, but satisfactory measure- tnents may be made up to 100 pg 1-l. The criterion of detection is approxi- mately 1.0 pg 1-l with standard deviations ranging between 0.4 and 1 . 7 pg l-l, depending on the concentration. The analysis time is 5-10 min for a single determination. Iron(I1) and copper(I1) ions are the only ions likely to be present in boiler waters that cause serious interference and these must be removed before analysis by passing the water sample through a cation-e:.;changc column.Keywords ; Oxygen determination ; absovptiometvy ; watev analysis ; manual annly s is For many years the manual method recommended in the CEGB for determining oxygen in feed-water has been that described by Potter and Whitell the detailed procedure being given in BS 2690.2 This method is accurate if interfering ions are removed from the sample and is extremely precise, a standard deviation of 0.7 pg 1-1 being attainable. However, the technique is difficult and time consuming and even an experienced operator can analyse.no more than 10-12 samples per day. For these reasons the less sensitive and less precise indigo carmine spectrophotometric technique (BS 26902) or the North Hants Engineering Co. Ltd. portable] electrometric instrument (now supplied by Automatic Samplers Ltd.) has been preferred.A more convenient manual method was therefore desirable and the work of a number of Russian workers (Rotshtein and Shemyakin13 Sutotskii and Gramatchikov* and Devdariani et nZa5), who used a spectrophotometric procedure involving oxidation of the leuco-form of methylene blue in glycerol, seemed applicable. This reagent appeared to have a number of advantages over the indigo carmine reagent : (i) better stability; (ii) greater sensitivity; (iii) the formation of a single coloured species, the intensity of which is directly pro- portional to the concentration of oxygen. A detailed investigation of the experimental parameters was undertaken to develop a technique suitable for routine use in power stations, Basis of the Method Methylene blue is reduced in an alkaline glucose solution to the colourless leuco-form.The leuco-base is insoluble in water but is readily soluble in glycerol; this solvent has the 11051106 GOODFELLOW AND WEBBER: ABSORPTIOMETRIC DETERMINATION OF Analyst, vol. 104 additional advantage that because of its high viscosity oxygen diffusion is very slow and therefore the bulk of the reduced reagent soluition is relatively stable, even when no pre- cautions are taken to prevent contact with air. On adding the reagent to a sample containing oxygen a coloured oxidation product of the leuco-base is produced, the intensity of which is proportional to the concentration of oxygen as long as there is sufficient excess of the leuco-base. Method Apparatus Spectrophotometric cuvettes.The special cuvette shown in Fig. 1 can be obtained from either Chandos Intercontinental (New Mills, Nr. Stockport, Cheshire) or Hellma (England) Ltd. (Westcliff-on-Sea, Essex). The taps shou.ld be evenly but lightly greased with com- mercially available rubber grease. Silicone grease should not be used as this may be troublesome to remove should it get on to the internal optical faces. inlet Outlet Fig. 1 . Spectropho tometric cuvette. Sflectrophotometer. Most commercial spectrophotometers can be used for this deter- mination although minor modifications may be required to enable the cuvette to be con- tained in the sample compartment. Spectrophotometers with digital read-out are recommended as with null-point and analogue display instruments difficulties have been experienced in obtaining accurate readings during the initial stages of the reaction.However, many of these latter types of instrument have a recorder outlet to which a digital voltmeter can be connected. Although the voltage output may not be linear with respect to absorbance, a conversion graph may easily be obtained. A diagram of a suitable 5- or 10-m1 burette assembly with 0.1-ml graduations and a reservoir is given in Fig. 2. Alternatively, a self-filling micro-burette may be obtained from Baird and Tatlock Ltd., catalogue number 241/0420. The delivery tip of this burette must be modified to that shown in Fig. 2. Micro-syringes of 25- and 100-pl capacities fitted with 70-mm hypo- dermic needles. Make the column with approximately 20 ml of cation- exchange resin (Amberlite IR 120, analytical grade, has been found satisfactory) packed in a glass tube of approximately 180 x 15 mm.This volume of resin is sufficient to remove divalent cation impurities at concentrations of 100 pg 1-1 from approximately 5000 1 of water. (Ammonia present at the levels norma.lly in feedwater, i.e., 0.1-1.0 mg l-l, should not affect the efficiency of removal of divalent cations.) Burette. Micro-syringes. Cation-exchange resin column. Those supplied by Scientific Glass Engineering (U. K.) Ltd. are suitable. Reagents Water. in reagent preparations. Distilled water passed through a mixed-bed de-ionisation unit is suitable for useDecember, 1979 LOW OXYGEN CONCENTRATIONS I N POWER-STATION WATERS. PART I 1107 Leuco-methylene blue.Solution A: dissolve 0.123 g of methylene blue (technical dye grade) and 0.65g of glucose (analytical-reagent grade) in 35 ml of water and dilute the solution to 2000 ml with glycerol (analytical-reagent grade). This solution is stable for several weeks. Solution B : dissolve 100 g of potassium hydroxide (analytical-reagent grade) in water and dilute to 200 ml with water. To 100 ml of solution A add x ml* of solution B. Mix by shaking and transfer the solution into the burette. This reagent may be used for at least 3 weeks, although the sensitivity of the determination decreases slowly with the ageing of the reagent. The reagent in contact with air becomes oxidised (e.g., at the top of the burette) but this does not affect the bulk of the reagent although it does affect the ease with which the meniscus may be seen. The reagent in the burette should be stored out of sunlight and preferably in the dark, at 15-20 “C.Allow the solution to stand overnight for air bubbles to separate. 1 I Reagent reservoir 5-or 10-ml burette “Push-fit” 1 insert 35 mm PTFE tube (0.d. % 2.0 mm) Fig. 2. Reagent burette modified with PTFE tip. Air-saturated water. Pass a stream of air, from a sintered-glass frit held just below the surface, through a continuously stirred volume of water until equilibrium is reached; with volumes greater than 1 1 this may take several hours. The oxygen content of this water (the temperature of which must be known) may be obtained from Table I, together with the following correction for the barometric pressure : SP s = - 760 where P = the observed pressure in millimetres, S, = solubility at pressure P and S = solubility at 760 mm at the observed temperature, Volumes of “low-oxygen” water, sufficient for the preparation of the calibration graph, may be prepared conveniently by the following procedure.Fill a glass container (size 5-10 l), which can be either an aspirator or a bottle fitted with an inlet/ outlet syphon, with water from which much of the oxygen has been removed; this may be obtained either from a suitable power-station source taken in the same manner as a sample or by boiling de-ionised water vigorously for about 20 min and then cooling under an inert- gas atmosphere. The oxygen remaining in the water can be reduced to less than 1 pg 1-‘ by passing a stream of inert gas, e.g., white-spot nitrogen, through the water for 90-120 min at a flow-rate of 500-1000 ml min-I.“Lozel-oxygen,” water. * N.B. I t is not possible to state precisely the volume of solution B as it is dependent, among other factors, on the ambient temperature (see, for example, Table III), but it will be between 0.5 and 2.0 ml. The actual volume, x ml, must be determined by a preliminary trial as that volume, using a ca. 100 p g 1-1 oxygen solution, which gives a reaction time of 5-8 min to achieve maximum absorbance. For this test it is necessary only for the larger air bubbles to separate from the solution (by standing for ca. 30 min) as the oxygen present in any residual bubbles will be insufficient to affect the result of the trial.1108 GOODFELLOW AND WEBBER : ABSORPTIOME'TRIC DETERMINATION OF Analyst, I/'ol.104 TABLE I SOLUBILITY OF OXYGEN IN WATER^ Temperaturel'C Solubility* 0 14.63 1 14.23 2 13.84 3 13.46 4 13.11 5 12.77 6 12.45 7 12.13 8 11.84 9 11.55 10 11.28 11 11.02 12 10.77 13 10.53 14 10.29 15 10.07 Temperature/"C Solubility* 16 9.86 17 9.65 18 9.46 19 9.27 20 9.08 21 8.91 22 8.74 23 8.57 24 8.42 25 8.26 26 8.12 27 7.97 28 7.84 29 7.70 30 7.57 * Solubility of oxygen in water (mg 1-l) in equilibrium with air at 760 mm. Procedure Sample collection Attach the cation-exchange resin column to the sample point using thick-walled neoprene tubing and a butt joint. (Other tubing with low permeability to oxygen may be satis- factory.) Before initial use, flush the column with up to 100 1 of sample, at a flow-rate of approximately 100 ml min-l, ensuring that no air bubbles are trapped in the system.Leave the column attached to the sample point with the outlet closed. On subsequent samplings flush the column with 1-2 1 of sample immediately before collecting an aliquot for analysis. With both inlet and outlet taps of the cuvette open, connect the inlet port of the cuvette to the outlet of the resin column using a butt joint with tubing similar to that used above. (N.B. The temperature of the sample should be less than 30 "C and within &3 "C of the temperature at which the calibration graph was prepared.) At a flow-rate between 30 and 100 ml min-l all.ow at least 150 ml of sample to flow through the cuvette, ensuring that no air bubbles are trapped in the system. First close the inlet tap and then immediately close the outlet tap on the cuvette, disconnect it from the resin column and store it immersed in a suitable container filled with water low in oxygen (this may conveniently be collected from the overflow of the cuvette).(N.B. Failure to close the taps in the order given will cause breakage of the cuvette if the sample supply is under any significant pressure.) The sample should be analysed as soon as possible after collection. Analytical procedure The "reagent / cuvette blank" absorbance for each individual measurement is obtained by taking absorbance measurements over the initial stages of the reaction and from these calculating, by linear extrapolation, the "zero-time" absorbance. To obtain accurate values of this absorbance the mixing and insertion of the cuvette into the spectrophotometcr must be completed as quickly and efficiently as possible.I t has been found that efficient mixing is obt,ained in ca. 5 s by holding the cuvette with the first and second (or second and third) fingers on the inlet and outlet tubes and the thumb under the cuvette and vigorously oscillating the wrist through 180". The first absorbance reading should be made within 20-25 s of cornmencing shaking; a further two readings within the first minute should normally be adequate. Immediately before the reagent addition allow 0.1 ml of reagent to flow to waste from the tip of the burette. Open tap C on the cuvette and insert the PTFE tip of the burette so that it protrudes about 2 mm below the bore of the tap. Add 0.3 ml of reagent carefully so as to avoid mixing and reacting the sample with the reagent.This mixing is minimisedDecember, 1979 LOW OXYGEN CONCENTRATIONS IN POWER-STATION WATERS. PART I 1109 by tilting the cuvette slightly so that the leuco-base forms a shallow pool in one of the lower corners of the cuvette. Start the stop-watch and simultaneously begin to shake the cuvette for 5 s, ensuring complete mixing of the sample and reagent. Quickly dry the optical faces of the cuvette and place it in the spectrophotometer set at 647 nm. Take absorbance measurements at known intervals over the first minute from the start of the mixing procedure and then note the absorbance at longer intervals until maximum absorbance, A e , is attained.Remove the tip of the burette carefully and close tap C. Calculation of Results Determine the absorbance of the solution at zero time, A,, by extrapolating the readings made over the first minute of the reaction. The absorbance, As, due to oxygen in the sample is given by From A , and the calibration graph calculate the oxygen concentration in the sample. Preparation of the Calibration Graph The calibration graph is prepared from results obtained using a single batch of low-oxygen water. Known additions of air-saturated water using the micro-syringe are made via the inlet port of the cuvette to aliquots of the low-oxygen water contained in the cuvette and the resulting solutions treated as samples. This gives a range of added oxygen concentrations from 0 to approximately 70 pg 1-1 (depending on the temperature of the water, the atmospheric pressure and the volume of the cuvette). The determinations should be repeated enough times to define the calibration graph with the required precision.Subtract the average absorbance for the solution containing no added oxygen from the average absorbances for each of the other solutions and plot the corrected absorbances against concentrations of oxygen calculated from the following equation : Additions of 0, 20, 40, 60, 80 and 1OOpl of air-saturated water are recommended. Oxygen concentration (pg 1-11 = ES, v2 where V, pl = volume of air-saturated water added; V , ml = volume of cuvette; and S , mg 1-1 = concentration of oxygen dissolved in air-saturated water. The calibration graph is linear up to at least 50 pg 1-1.The sensitivity of the method depends upon the age of the leuco-methylene blue reagent and it is recommended that the calibration graph should be prepared afresh whenever tests give results outside the expected limits. Once the linearity of response of the method has been confirmed by the calibration procedure described above it is often more convenient to use an alternative calibration procedure. For this purpose samples of low-oxygen water should be obtained to which the reagent is added and the absorbance measured in the usual way. Immediately maximum absorbance has been attained a known volume of air- saturated water is introduced into the cuvette and the increased maximum absorbance measured. The increase between the two values of maximum absorbance is directly pro- portional to the concentration of oxygen added.Results Reagent Formulation Initial tests showed that the formulation of the leuco-methylene blue used by Rotshtein and Shemyakin3 was unsatisfactory owing to the formation of a precipitate in the glycerol solution, while that used by Devdariani et a1.5 gave a poor sensitivity under the conditions used owing to the reduction of the methylene blue by the excess of glucose present. The compositions of the Russian reagents, together with that finally used in the proposed method, are given in Table 11.11 10 GOODFEL1,OW AND WEBBER ABSORPTIOMETRIC DETERMINATION OF TABLE I1 REAGEKT COMPOSITIONS Rotshtein and Devdariani Sht:myakin3 et al.5 Methylene blue . . . . . . .. 0.246 g 0.3 g Water . . . . . . . . . . 70 ml 70 ml Reagent A (per litre) (in glycerol)- Glucose . . * . . . . . . . . 2.6 g 1.2 g Reagent B (per 100 ml) (in water)- Potassium hydroxide . . . . . . 50 g 40 g Leuco-base- Reagent B:reagent A . . . . . . ;1: 19 1 : 19 Analyst, VoZ. I04 CERL 0.062 Q 0.33 g 17.5 ml 50 g 1:50 to 1 : 200 I t will be seen that there is a variable ratio of reagent B to reagent A in the CERL prepara- The reason for this is discussed in detail below under Reagent comeosition a i d volzime tion. added to a sarYzple. Development of the Technique It was recognised at the outset that the technique would have to be simple, easy to cali- brate and precise if it were to be a satisfactory alternative to the Potter and White1 method. Four basic requirements were identified : the design of a leak-proof sampling vessel, capable of being used also as a spectro- photometric cuvette, which was simple to manufacture and easy to use ; a means of reagent addition without atmospheric contamination ; a technique for measuring a meaningful reagent/cuvette blank ; a suitable method of calibration.(i) (ii) (jii) (IV) The last requirement was necessary because the technique proposed by Rotshtein and Shemyakin3 of using standard solutions of fully oxidised methylene blue for calibration had been shown by Devdariani et to cause large errors because the absorption spectrum of such standards was not the same as that given by methylene blue formed in solutions by oxidation of the leuco-base. Spectrophotometific cztvette One of the primary requirements for a successful manual technique is that the sample should be manipulated as little as possible to minimise the risks of atmospheric contamina- tion.Ideally a vessel was needed which would serve for sample collection, for colour development and for spectrophotometric measurement. In addition, it not only had to be leak-proof to atmospheric oxygen, but had to allow the addition of reagent without intro- ducing extraneous oxygen. After trials with different designs of cell and different techniques for reagent addition the cuvette shown in Fig. 1 was adopted. I t was found that the reagent could be added without contamination by inserting through the bore of the stopcock a 2 mrn o.d. PTFE tube attached to the modified tip of a burette. The reagent inlet limb was designed so that ingress of air into the sample was minimised both during and after the time of reagent addition.Initially PTFE stopcocks were used as these required no lubrication. However, it was found that these often deformed easily and this resulted in ingress of air. Lightly greased borosilicate glass stopcocks were, therefore, used in later batches of cuvettes. Because of the necessary restriction in size of the cuvette for its accommodation in a spectrophotometer and also to allow access of the PTFE burette probe, specially manufactured glass micro- stopcocks of 2.5-mm bore were used. Teclznique of menswement The conventional method of spectrophotometric measurement is to measure the equilibrium absorbance of the sample to which reagents have been added, and to deduct from this reading the absorbance due to the reagent and any self-colour in the sample.This correctedDecember, 1979 LOW OXYGEK CONCENTRATIONS IN POWER-STATION WATERS. PART I 11 11 absorbance is then proportional to the concentration of determinand in the sample. The reagent blank is normally determined by measuring the absorbance of a solution containing no significant concentration of the determinand, to which reagents have been added. In the oxygen determination it is extremely difficult for routine analysis to ensure that this will be the case. An alternative technique for obtaining the absorbance of the reagent blank was therefore tested and found satisfactory. when the glycerol solution of reagent is initially added to the sample, because of the differences in viscosity an insignificant amount of mixing and reaction occurs, and by adjusting the reagent composition and concentration the time required for the reaction to give maximum absorbance can be controlled. By using a relatively slow reaction time to obtain maximum absorbance, the increase in absorbance of the solution may be measured accurately against time after the mixing of sample and reagent.The resultant line plot can then be extrapolated to “zero time” and the intercept with the absorbance axis gives the reagent blank absorbance plus the absorbance from any self-colour in the sample; this technique also eliminates the cuvette blank. The difference between the “zero-time” absorbance and maximum absorbance is then proportional to the oxygen content of the sample.This technique is based on two factors : (i) (ii) Reagent cona$osition and volume added to a sawfile The leuco-methylene blue reagent solution is prepared by mixing a solution of methylene blue and glucose in glycerol with a small volume of an aqueous potassium hydroxide solution and allowing the mixed solution to stand in normal daylight until colourless (15-30 min). The rate of reduction of methylene blue is dependent on the concentrations of both the glucose and potassium hydroxide. However, in alkaline solution the leuco-methylene blue is not excessively stable and decomposes slowly to an inactive form. The relative con- centrations of the three constituents have, therefore, to be carefully controlled to ensure : ( a ) reasonably fast reduction of the methylene blue; ( b ) adequate stability of the leuco-methylene blue reagent; (c) a colour development time commensurate with obtaining an accurate estimate of the reagent/cuvette blank ; ( d ) adequate stability of the final colour, when the reagent is added to aqueous solutions containing dissolved oxygen.Initially the leuco-base reagent used was a mixture of 1 ml of reagent B to 40 ml of reagent A (see Table 11) and this reagent was added in the ratio of 1 : 40 to water in the filled cuvette. Under these conditions maximum colour development occurred within 2-3 min of mixing the reagent with the sample, but the colour faded at >lyo min-l thereafter, and was there- fore not sufficiently stable for the precision required or for the mode of measurement used.The volume of reagent B added to reagent A was therefore reduced. Lsing the same ratio of the various leuco-methylene blue reagent solutions to water, it was found that as the concentration of alkali was reduced, the reaction time between the reagent and oxygen increased and also the final product became more stable. I t can be considered that there are two competing reactions: (i) (ii) oxidation of the leuco-methylene blue by dissolved oxygen, and reduction of the methylene blue formed in (i) by the glucose and hydroxide present. Decreasing the hydroxide concentration retarded the second reaction and the value of the maximum absorbance produced by a given concentration of oxygen was thereby increased. The most suitable ratio of reagent B to reagent A was found to lie between 1 : 50 and 1 : 200, the actual value depending upon conditions, such as ambient temperature.The most suitable ratio therefore needs to be experimentally optimised. The addition of 0.3 ml of the alkaline leuco-base to the cuvette (volume 12 ml) filled with sample gave maximum colour development times in the range 2-8 min. The time depended on the concentration of oxygen, the shorter time arising from lower concentration. The maximum absorbance with freshly prepared reagent remained constant for at 1 .ast 8 min. As the reagent aged over a number of days, when stored in the burette in the normal laboratory atmosphere, it was found that the reaction time decreased by about 50% and the absorbance sensitivity decreased by about 20% over a period of 8 weeks. I t was also found that once the maxi-1112 GOODFELLOW AND WEBBER: ABSORPTIOMETRIC DETERMINATIOK OF AnaZyst, VOZ.104 mum absorbance for any one measurement had been obtained it remained constant for only 2-3 min before the colour started to fade. When leuco-methylene blue solution was freshly prepared by mixing reagent A with reagent B the sensitivity and stability of the reaction were essentially the same whether or not these two separate reagent solutions were them- selves freshly prepared or up to 2 months old. When a solution of potassium carbonate equivalent to the 50% potassium hydroxide solution was used to prepare the leuco-base solution (simulating the maximum absorption of carbon dioxide that could occur in the potassium hydroxide solution during storage) the sensitivity was not affected, although the reagent took 2-3 h to decolorise, and the subsequent reaction time to maximum colour was increased to such an extent (50 min) as to be impracticable.This was similar to the effects experienced when less potassium hydroxide was used, and indicates that both the reduction of methylene blue and the reaction between oxygen and the leuco-base are markedly pH dependent. Effect of Temperature Sample temperature This effect was tested by taking samples of low-oxygen water at known temperatures to which the leuco-methylene blue reagent was added. After the reaction had reached equilibrium, air-saturated water was introduced into the cuvette and the increase in absorbance recorded.At both 18.5 and 33 "C the increase in absorbance when 50 pl of air- saturated water was added was 0.194 (triplicate determinations, standard deviation of the means 0.006). However, at the higher temperature the maximum absorbance was obtained after only 2.5 rnin (90% at 1.35 min) compared. with 4.5 min (90% at 2.5 min) at 18.5 "C. At 33 "C, the absorbance started to decrease within 4 rnin of the air-saturated water being added. Once a water sample has been obtained and the taps have been closedl, temperature changes will cause differential expansion or contraction of the water and glass, and this may lead to air or air-saturated water being drawn into the body of the cuvette from the keyways of the stopcocks. To minimise this effect it is recommended that cuvettes filled with sample are stored under excess of sample-water collected in a suitable vessel during the cuvette flushing-out stage of the sampling procedure.Variations in temperature can affect the integrity of the cuvette. Reagent temperature It has been shown above that the leuco-methylene blue reagent is unstable and that the rate of deterioration increases with increasing temperature. This instability is related to the concentration of potassium hydroxide in the reagent, as the results in Table I11 show. Not only is there a decrease in sensitivity with increases in storage temperature, age of reagent and amount of potassium hydroxide, but there is a concomitant decrease in the time for equilibrium absorbance to be reached, TABLE I11 EFFECT OF TEMPERATURE ON REAGENT STABILITY Reagent (prepared on day 1) Sensitivity, absorbance per 100 pl of air-saturated A I 7 waterltime to equilibrium (min) Composition, Storage condition prior ,- A \ A/B* to test Day 2 Day 3 Day 4 18-21 "C 10.49 1 1 8 0.48817 0.51118 l O O / l .O Ambient temperature, lOO/l.6 28 "C for 24 h 0.46 81 5 1 001 1 .o 28 "C for 48 h 0.47017 1001 1.5 28 OC for 48 h 0.43014 100/0.8 28 "C for 72 h 0.47215 loo] 1 .o 28 "C for 72 h 0.46015 l00/1.6 28 "C for 72 h 0.41313 * A is the glycerol solution of methylene blue and glucose; B is 50% m l V potassium hydroxide solution.December, 1979 LOW OXYGEN CONCENTRATIONS IN POWER-STATION WATERS. PART I 1113 Tests with Different Batches of Methylene Blue Samples of methylene blue were obtained from three different sources: Aldrich Chemical Co., BDH and Eastman-Kodak.Dilute aqueous solutions (2 mg 1-l) of these were prepared, the visible absorption spectra of which indicated that all three samples were of essentially the same purity; this was confirmed when leuco-reagents prepared from each source gave similar absorbances when added to solutions containing excess of oxygen. Calibration graphs prepared from each of these reagents gave similar sensitivities and again indicated no significant differences in the composition of the three samples. The detailed results are given in Table IV. TABLE IV COMPARISON OF THREE BATCHES OF METHYLENE BLUE Absorbance a t peak wavelength (ca. 650 nm) t Aldrich BDH Eastman-Kodak (U.S.P.) (technical dye) (certified) I h 7 Aqueous solution (2 mg 1-I) .. . . 0.41 1 0.421 0.415 pg 0,1-1* 15.7 31.4 47.1 62.7 Excess 0,t 0.101 (0.002,) 0.094 (0.003,) 0.091 (0.002,) 0.189 (0.010) 0.187 (0.005,) 0.169 (0.002) 0.267 (0.014) 0.257 (0.001) 0.252 (0.002,) 0.355 (0.002,) 0.340 (0.004,) 0.330 (0.007) 0.899 (0.019) 0.918 (0.029) 0.913 (0.014) * Each figure for the absorbance is the mean of three results. t Each figure for the absorbance is the mean of five results. The figures in parentheses are the standard deviations of the mean results. Storage of Samples If the cuvette containing a sample is leak-tight it should be possible to store the sample indefinitely. However, the water remaining in the exterior limbs of the taps may become contaminated with oxygen and this may in turn contaminate the sample when the reagent is added.Eight cuvettes were filled with low-oxygen water and set aside for 2 h, four immersed in the overflow from the low-oxygen water supply and the remainder stored in air. In a second test the five cuvettes which showed little or no contamination during the first test were stored, filled with low oxygen water, for 2 h in air; the remaining three cuvettes were stored under water. The results of these tests are shown in Table V and indicate that two of the cuvettes (2 and 8) were definitely not leak-tight, two (1 and 4) gave slightly variable results while the remainder were unaffected by the mode of storage. As a safeguard it is recom- mended the cuvettes should be tested for leak tightness in air as above, but for routine practice samples should be stored under low-oxygen water until they are analysed.Addition- ally, samples should be analysed as soon as possible after collection. Wash- ou t of Cuvet tes When a new sample was collected this displaced the contents of the cuvette. The time and volume of water required to do this were determined by flushing a cuvette (volume 13.5 ml) con- taining a methylene blue solution, whose absorbance was 0.350, with water a t a flow-rate of 400 ml min-l and measuring the absorbances at 15-s intervals. After 15 s the absorbance was 0.005 and no further reduction occurred with two further periods of flushing. A more stringent test was carried out in which an empty cuvette was flushed with low- oxygen water for various time periods until a minimum absorbance value was obtained when the reagent was added.At a flow-rate of 200 ml min-1 the absorbance after 20 s was 0.020; after 40 s this was reduced to 0.008 and no further reduction occurred with flushing times up to 2 min. Further work showed that the flow-rate could be as low as 30 ml min-l provided that at least 150 ml of sample were flushed through the cell. The cuvettes were normally stored with the previous sample plus reagent in them.1114 GOODFELLOW AND WEBBER : ABSORPTIOMETERIC DETERMINATION O F Analyst, vol. 104 TABLE V STORAGE OF SAMPLES Storage for Absorbance, corrected for Equivalent oxygen Cuvettc No. 2 h original 0, concentration concentration/pg 1-1 1 Air 0.007 1.4 2 0.050 10.0 3 0.002 0.4 4 0.01 7 3.4 5 6 7 8 Water Air 2 Water 4 ,9 0.001 0.000 0.000 0.01 6 0.004 0.000 0.000 0.000 0.000 0.033 0.000 0.01 3 0.2 0 0 3.2 0.8 0 0 0 0 8.2 0 2.6 Performance of the Method range up to 30 ,ug 1-l) are summarised in Table VI.gave the equation The results from five determinations on each of five concentrations of oxygen (in the A regression analysis on these results y = 0.001, $. 0.004,x where y = absorbance units and x = pg 1-1 of oxygen, and the correlation coefficient was >0.99. TABLE VI PRECISION OF RESULTS added/ pg 1-' Mean absorbance* 1-1 Concentration of oxygen Standard deviation/ 0 7.28 14.55 21.83 29.10 0.003 0.0368 0.0'7 5 0.109 0.139, 0.40 1.02 1.33 1 .fi9 0.91 * Mean of five results corrected, where appropriate, for the absorbancc of the no-added- oxygen solution. To test the accuracy of the method a continuous supply of water with a known concentra- tion of oxygen was produced by passing a controlled mixture of nitrogen and oxygen through oxygen-free water.The water was supplied continuously to an EIL, Model 9430, oxygen monitor, and samples were taken periodically for analysis by both the Potter and White methodl and the methylene blue procedure. Because of the time taken to determine the oxygen concentration using the Potter and White method the results of the Potter and White determinations were used to show the accuracy of the results obtained with the EIL monitor. Results from the monitor were then compared with those obtained using the leuco-meihylene blue technique. The mean of five Potter and White determinations was 16.7 pg 1-1 (standard deviation 0.4 pg 1-l) compared with a mean of 16.8 (0.2) pg 1-1 obtained from readings taken simul- taneously on the EIL monitor.The results obtained using methylene blue are given in Table VII, together with the corresponding readings from the EIL monitor; they show no statistically significant difference.December, 1979 LOW OXYGEN CONCENTRATIOKS IN POWER-STATION WATERS. PART I 1115 TABLE VII DETERMINATION OF OXYGEN BY METHYLENE BLUE AND BY EIL 9430 MONITOR Methylene blue/ EIL monitor/ 17.5 16.8 18.6 17.3 17.7 17.0 17.9 17.0 16.7 17.0 17.0 16.8 16.5 16.6 16.8 16.8 Pf3 1-' Clg 1-' Difference/ P.g I-' 0.7 1.3 0.7 0.9 - 0.3 0.2 -0.1 0 Mean: 0.4 Interferences The effects of a number of substances likely to be present in feed-water samples were tested and the results are detailed in Table VIII.The test solutions were prepared by dissolving the substance in air-saturated de-ionised water and adjusting the concentrations so that suitable amounts of the substance and dissolved oxygen could be introduced into the cuvette, containing low oxygen de-ionised water, with a hypodermic syringe. TABLE VIII EFFECT OF OTHER SUBSTAXES Conccn tration/ Substance pu.g 1-l N,H, . . . . 90 T\;H,OH . . 1000 Fez+ . . .. 40 80 200 Fe3+ . . . . 200 cu2+ . . . . 100 Na+ . . . . 100 Ca2+ . . . . A l p + . . . . Zn2+ . . . . 100 c1- . . . . 150 NO,-- . . . . 1000 :::: ] Effect / p g 1-1 of oxygen I A > [O,] = 10 pg I-' [O,] = 30 pg1-I <0.5 t 0 . 5 - <0.5 <0.5 - [O,] = 60 - 13 - 26 - 58 - - - - - - - <0.5 - 25 32 - <0.5 <0.5 - <0.5 (0.5 <0.5 <0.5 To attempt to overcome the interference effects of copper(I1) and iron(I1) ions, complexing agents were introduced into the reagent.Both and 2 x lW4 M solutions of the di- sodium salt of EDTA increased the equilibrium response time to over 1 h, and although a 2 x 1 0 - 5 ~ solution gave a satisfactory response time its concentration was too low to eliminate the interference of either of the metal ions. Glycine, at a concentration of 0.25 g 1-1 of reagent, reduced the interference from copper(I1) ions to an insignificant amount but had no effect on the interference due to iron(I1) ions. As it appeared unlikely that normal complexing reagents would successfully eliminate the interference from copper(I1) and iron(I1) ions it was decided that a better alternative would be to remove them by ion exchange, as in the Potter and White technique.Removal of Cations by Ion Exchange Davies et aL7 showed that ions interfering with the Winkler determination could be removed successfully on fresh ion-exchange resin columns, but no work has been published on the effect of reducing cations [e.g., iron(I1) or hydrazinium ions] already absorbed on the resin on oxygen present in a solution passing through the resin. This situation would arise if a column were used for the removal of cations from many samples (or a continuously flowing sample) before regeneration.1116 GOODFELLOW AND WEBBER: ABSORPTIOMETRIC DETERMINATION O F AW&.St, Vd. 104 A cation-exchange column was saturated with iron(I1) ions, and was then placed in the flow-line from a supply of low-oxygen water.A second cation-exchange column in the hydrogen form was connected in series with the iron(I1) ion loaded column and sampling points were installed before and after the two columns. The second column ensured that any iron(I1) ions leaking from the first column would be removed prior to analysis. A small concentration of oxygen was chosen for these tests so that relatively small decreases in oxygen due to the effect of the reducing species would be analytically significant and this was achieved by passing low-oxygen water through a sufficient length of silicone-rubber tubing so that oxygen from the air diffused into the water to give a concentration of ca. 5pgl-I. The water was then passed through the ion-exchange columns, the flows from both sampling points being kept constant at about 50 ml min-l (i.e., a linear flow-rate of 30 cm min-l through the column).The experiment was repeated with the iron(I1) ion column replaced by a column saturated with hyclrazinium ions. The results given in Table IX show that no significant decrease in the oxygen content of the water occurred when it was passed through either the iron(I1) ion or hydrazinium ion column at the temperature of the experiment, 25, "C. TABLE I:X EFFECT OF REDUCING SPECIES ABSORBED ON CATION-EXCHANGE RESINS ON THE CONCENTRATION OF OXYGEN I N WATER PASSED THROUGH THE RESIN Oxygen concentration found*/pg 1-1 -7 -_ 7-- Reducing species Before ion exchanger After ion exchanger 17C2f . . .. 4.9 4.6 N,H,+ . . .. 5.2 5.2 * Each result is the mean of three determinations. The standard deviation in each instance was ca. 0.2 pg 1-'. Discussion Development of the Analytical Method The analytical method described is based on the work of Devdariani et aZ.,5 who showed that under controlled conditions the leuco-base of methylene blue is oxidised by oxygen dissolved in water and that the intensity of the blue colour produced is proportional to the oxygen concentration. The major obstacles to the development of a suitable manual technique were expected to be (i) the design of a suitable spectrophotometric cuvette, which could also be used as both the sampling and reaction vessel, (ii) the method of addition of the reagent and (iii) the calibration procedure. Once a technique had been developed for manufacturing a suitable cuvette it was found that the reagent could be added simply from a semi-micro burette that had been modified so that the tip of the delivery tube could be inserted directly into the cuvette.Two specialist spectrophotometric cuvette manufacturers can now make cuvettes to the CEGR specifications. The cuvettes fit into the cell compartments of Pye-Unicam SP600, SP6-500, SP1700 and SP1750 instrument:;; it is possible that with some other makes of spectrophotometer it may be necessary to make a simple modification to the cell cover to accommodate the cuvette. The development of a calibration technique involved two stages: (i) the preparation of solution containing known amounts of oxygen and (ii) the determination of the reagent / cuvette blank.Stage (i) involved the addition of air-saturated water to the cuvette con- taining low-oxygen water and it was found that this could be done conveniently and precisely by using a hypodermic syringe with the needle inserted through the bore of an open tap in the cuvette. I t was also found that air-saturated water could be prepared satisfactorily by passing a stream of compressed air through de-ionised water. Provided that sudden temperature changes of the water were avoided the possibility of super-saturation did not occur with this technique. To determine the reagent blank it is normally necessary to provide a sample containing zero concentration of the determinand, but to pi-epare a sample of water known to containDecember, 1979 LOW OXYGEN CONCENTRATIOSS I N POWER-STATION WATERS.PART I 1117 no dissolved oxygen on a routine basis is virtually impossible as the inert gases in the gas- stripping techniques used cannot be guaranteed to be 1 0 0 ~ o pure. The difficulty of measuring the reagent blank was overcome by a novel procedure. The reagent is an alkaline solution of leuco-methylene blue in !%yo glycerol and because of its high viscosity the reagent could be added to the sample in the cuvette as a narrow stream from the PTFE probe of the burette and collected at the bottom of the cuvette without any significant mixing occurring. Prior to the mixing of the two phases the integrated absorbance at the wave- length of measurement would be due to that of the reagent plus any self-colour of the sample.This is, of course, a hypothetical condition as it is not possible to distribute the leuco-reagent uniformly in the cuvette for spectrophotometric measurement, without initiating the reaction with dissolved oxygen in the water. With an appropriate choice of reagent composition the reaction rate between the leuco-metliylene blue and oxygen dissolved in the sample is made to be relatively slow and essentially linear for the initial reaction period. By moni- toring the increase in absorbance against time after mixing the sample and reagent, the reagent/cuvette blank can be estimated by extrapolating the absorbance curve to zero time, i.e., when the mixing commenced. In practice it was found that the linear part of the reaction curve extended for the first 40-60 s of the reaction so that two measurements within this period were sufficient to calculate by simple proportionation the absorbance at zero time resulting from the reagentlcuvette blank.This method eliminates the effect of any transmission losses of light passing through the optical faces of the cuvette. Range of the Method For power-station applications measurements of dissolved oxygen are normally required in the range up to 50 pg 1-l. I n this range the leuco-methylene blue method gives a linear calibration graph. At higher concentrations the graph becomes progressively non-linear and, with the reagent concentration used, tlie practical limit of the method is approximately 10Opgl-l. I t is possible that the range may be extended by increasing the amount of reagent added to the sample but this will affect the reaction rates.Maximum absorbance will be reached more rapidly and it is likely that this absorbance will not be stable so that accurate measurements will not be possible. However, as an indication of high oxygen concentrations such a modification to the method may prove useful for investigational purposes. Performance of the Method Tlie results given in Table VI show that at an oxygen concentration below 1 p g 1-1 the standard deviation was 0.40 pg 1-l. At higher concentrations, in the range 0-30 pg l-l, it varied between 0.91 and 1.69 pg 1-l and these higher values were thought to result from the increasing errors associated with the estimation of the zero-time absorbance when the absorbance increased rapidly over the initial stages of the reaction.However, when con- sidered as a proportion of the final concentration the errors were negligible. I3otli iron(I1) and copper(I1) ions caused serious interference with the determination and, as these ions are present in virtually all feed systems, it was considered necessary to remove them from samples prior to analysis. Cation exchange has been used successfully for many years in conjunction with the Potter and White method although doubts had been raised wlietlier concentrations of reducing agents, namely iron(I1) ions and hydrazine, might build up on the resin and reduce the oxygen dissolved in the sample passing through the resin column. The results obtained indicate that even when the resin is saturated with these reducing species no reduction of the oxygen in the sample occurs a t ambient temperature at the low concentrations normally found in power-station feed water and condensate.l'lic most probable cause of bias will be contamination of the sample by the atmosphere and tliis can occur a t any stage of tlie determination. l;or example, an air bubble of diameter 0.4 mm, which could easily escape visual detection, will increase the oxygen content of a samplc. in a cuvette containing 12 ml of water by 1 pg 1-l. However, the results obtained during the (levclopment of this method indicate that any bias was less than 0.5 pg 1-l. The results of comparative determinations made both with the EIL, Model 9430, oxygen monitor and by tlie Potter and Wliite method at a dissolved oxygen concentration of ca.17 pg I--' indicate that there was no significant bias between the methods.1118 GOODFELLOW AND WEBBER The leuco-methylene blue reagent deteriorates slowly with age and this causes a slow, but significant, decrease in sensitivity with a consequent change in the calibration graph. The sensitivity of the reagent should be checked during each batch of determinations by the analysis of a control standard. However, this deterioration in the reagent occurs over a number of weeks even when no special precautions, except the exclusion of direct sunlight, are taken with the storage of the reagent. Because the zero-time procedure is used to determine the reagent/cuvette blank absorbance, samples cannot be analysed in batches as with most types of conventional spectrophoto- metry. Each sample has to be treated individually and the rate of colour development measured until peak absorbance is obtained. The time required for this depends on the oxygen content of the sample and the concentration of reagent added. The concentration of reagent used in the recommended procedure was chosen to give a sufficiently fast over-all reaction rate to attain peak absorbance in a reasonable time whilst enabling the reagent/ cuvette blank to be determined with sufficient accuracy by extrapolation of the absorbance veysuus time graph to zero time. Using the given procedure the time required for the measurement of one sample is approxi- mately 10 min. This does not include the time taken to collect the sample or transfer it to the laboratory. Extension of the Technique Smaller experimental cuvettes (capacity 5 ml) have been made and used successfully for determining oxygen in water from corrosion rigs at CERL, where only limited volumes of sample are available. The technique has also been used for continuous measurement; this work is published separately in Part II.8 The authors record their appreciation to Mr. P. Madden, who advised on and made the experimental spectroplwtometric cuvettes, and to Mr. M. Owen of the CEGB South West Region, Scientific Services Dept., who provided the facilities for the comparative analyses and who made the Potter and White determinations. The work was carried out at the Central Electricity Research Laboratories and is published by permission of the CEGB. References 1 . 2. 3. 4. 5. 6. 7. 8 . NOTE-l-leference 8 is to Part 11 of this series. Potter, E. C., and White, J . F., J . A p p l . Chem., 1957, 7, 459. “Methods of Testing Water Used in Industry,” I’art 2, Methods 2 and 3, BS 2690 : 1965, British Standards Institution, London. liotshtein, V. P., and Shemyakin, V. N., Teplodnwgetika, 1962, No. 2, 54. Sutotskii, G. P., and Gramatchikov, M. V., Teploc’nergetika, 1966, 13(10), 86. Devdariani, 1. V., Partskhaladze, K. G., Petruzashvili, L. G., and Shmal’tsel’, G. N., Teplodner- Department of the Environment, “Analysis of Raw, Potable and Waste Waters,” HM Stationery Davics, l . , Redfearn, M. N., and Remer, I). E. Y . , Analyst, 1956, 81, 113. Goodfcllow, G. l., Libaert, L). F., and W’ebber, H. M., Analyst, 1979, 104, 1119. getzka, 1970, 17(10), 76. Office, Jdondon, 1972, p. 106. Received April 27th, 1979 Accepted M a y 30th, 1979