vinkler method for do

8/3/2019 Vinkler Method for DO

1/6

THE ACCURACY OF THE WINKLER METHODFOR DISSOLVED OXYGEN ANALYSIS

James II. CarpenterDcpnrtmcn t of Occnnography, The Johns IIopkins University, Ualtimore, MarylandThe potential errors in the various tcchniqucs for the Winkler m&hod have been cxam-incd and a new tcchniquc dcvclopcd. The accuracy of this technique has been tested bycomparison with standards based on dissolving known quantities of oxygen in oxygen-freewater. An accuracy of 0.1% was obscrvcc~, so the tcchniquc nppcars suitable for oxygenanalysts requiring an accuracy of grcatcr than the 3-50/o obscrvcd with tho commontechniques.

INlnODUClIONThe Winkler method of analysis for oxy-gen dissolved in water is unusual in thatthe result is not based on oxygen as a stan-dard. Studies of the accuracy of themctho,d have been based on comparisonwith the results of another method. Extrac-tion and gasometric measurement has beenthe alternate method in a number of studiessince Winkler proposed the method in 1888.The most recent example of this approach

is the work of Wheatland and Smith (1955),in which the accuracy of the titrimctrictechnique used in the solubility studies ofTruesdale, Downing, and Lowdcn ( 1955)was considered. The results appeared tobc valid, and the solubility values wcrc sug-gested for application to oceanographic(Richards and Corwin 1956) and limno-logical ( Mortimer 1956) research. Solubil-ity mcas,urements by Hots and Benson(I963), as well as others, do not confirmthe Truesdale, Downing, and Lowden re-sults and, therefore, suggest significant cr-rors in their techniques.Discordant results of comparisons oftechniques during International Geophys-ical Year oceanographic expeditions haveled to a question of the accuracy of severaltechniques and to the rejection of the oxy-gen analysts in preparing summary chartsof station data. Differences between nomi-nally analogous procedures have also been

L Contribution number 76 from the ChcsapcakcBay Institute and the Dcpartmcnt of Occanog-raphy, The Johns Hopkins University. This workwas supported by the National Institutes of Healththrough Grant RG-5982.

obscrvcd in an intercomparison experimentconducted by the chemical methods sub-committee of the National Academy of Sci-ences Committee on Oceanography. Re-sults and conclusions from this expcrimcntwill shortly be published as a NASCO re-port (D. E. Carritt, personal communica-tion ) .In view of the discrepancies in the solu-bility measurements and in the techniquesused on oceanographic expeditions, and asa first step in the determination of the solu-bility of oxygen in pure water and sca-water, the accuracy of a technique for theWinkler method is evaluated here so thatthe reliability of the analytical results canbe estimated. The experimentally simplerand logically more straightforward opcra-tion of dissolving known quantities of oxy-gen in oxygen-free water was chosen toproduce a primary standard.The SOLUCCS of error in the manipulationsrequired for the Winkler method for analy-scs of oxygen dissolved in pure water wereconsidered to be the following:1) Air oxidation of iodide.2) Volatilization of iodine.3) Oxygen contributed by the reagentsolution.4) Iodatc contamination of the iodidesolutions.5) Consumption or production of iodineby reagent contaminants.

6) Difference between titration end pointand the equivalence point.These possiblo errors wcrc examined in-dividually and conditions found underwhich all were negligible except oxygen135


2/6

136 JAMES H. CARPENTERcontributed by the reagent solutions. As aresult of these tests, the following procedurewas considered to have minimum errorsand was tested for accuracy.

METHODS AND MATERIALSTechnique for the Winkler method

Reugents-The first two errors can be re-duced by the use of the following reagents:MnCls*4Hz0 600 g/liter 6 ml/liter of samnleNaINaOHH&O4

600 g/liter320 g/liter 6 ml/liter of sample500 ml/liter 3.6 ml/liter of sampleAir oxidation is minimized by a final pHof 2.VoZatiZixation-Loss of iodine during ex-posure of iodine solutions to the atmo-sphere is minimized by the formation of thetriiodide complex with the high iodide con-centrations produced by these reagents.However, a more reliable system was basedon elimination of transfer of the sampleuntil most of the iodine had been reducedwith thiosulfate.

Reagent blanks-The errors produced byiodate contamination and reagent contami-nation were serious but could be reducedby careful selection of chemicals. Whereasiodate contamination is commonly recog-nized, reducing impurities in the reagentswere the major sources of a negative blank.An example of this blank is shown in Table1. The negative values make six or moretitrations necessary for precise standardiza-TABLE 1. lodine titrations for blank andprecision estimate

iadk,:l(mm)

Thiosulfatesolution(6)

Thiosulfatesolution( calculated )(Ex)

Differencel&c)

0.08824 6.127 6.125 +0.0020.09864 6.850 6.851 -0.0010.10432 7.247 7.248 -0.0010.13836 9.624 9.625 -0.0010.15119 10.516 10.520 -0.0040.19976 13.912 13.912 0.0000.20084 13.993 13.988 +0.005Least squares fit:Thiosulfate (g ) = -0.037 + 69.83 iodine (meq)Standard error of estimate 0.003 g

tion of the thiosulfate solution. The data inTable 1 could indicate either an end-pointerror or a negative reagent blank.The association of the negative blankwith the reagents was shown by the follow-ing kinds of experiments. Repeated photo-metric titrations of iodine (from iodate) andthiosulfate were performed using syringemicroburettes to add less than 10 micro-equivalents of either substance to the 200ml of solution containing the reagents, thatis, adding iodine and titrating it with thio-sulfate was followed photometrically sev-eral times in the same solution. The firsttitration was invariably different from t&successive ones in that less thiosulfate wasrequired to reduce the iodine in the firsttitration than subsequently.A second demonstration that the nega-tive blank is associated with the reagentswas the performance of titrations with dif-ferent quantities of reagents present andwith suitable quantities of iodine so thatleast squares regression analysis could beused.0.5-ml reagents: Thiosulfate = -0.011 +68.1 iodine1 *O-ml reagents : Thiosulfate = -0.029 +68.76 iodine2.0~ml reagents: Thiosulfate = -0.059 +68.90 iodine

The dependence of the intercept on thequantity of reagent is clear. The lack ofconsistency in the slope reflects the volu-metric errors with the microsyringe burette,since less than 10 microequivalents ( peq )of iodine were used. The results clearlyshow a loss of iodine associated with theinitial solution rather than a systematic dif-ference between the equivalence point andend point.The source of the reducing impurity wasfound in the sodium iodide. Examinationof sodium iodide from several sources hasshown a variation of two orders of magni-tude in the value of the blank. The blankfrom the sodium iodide may be reduced tonegligible levels by removing dark and dis-colored particles by hand using tweezersand a hand lens.End-point detection-The end-point de-


3/6

ACCURACY OF WINKLER METIIOD 137tection by visual indication with starch andby amperometric and potentiometric mca-surements has been studied by Knowlesand Lowden (1953). Colorimctric dctcrmi-nation with starch and amperomctric mea-surcment were compared in the study ofthe iodine-thiosulfate reaction of Bradburyand Hambly ( 1952). Both these studiesshowed that the starch-indicated end pointis significantly different from the cquiva-lcncc point in titrations of iodine solu-tions at concentrations corresponding to air-saturated water. The authors rccommcndcdthe amperometric technique but remarkedon the variable sensitivity of the elcctrodcs.Since triiodide ion is the most abundantspecies of iodine present in the solution,p,articularly near the end point, determina-tion of triiodide ion was expected to be asuperior method of end-point detection.Triiodidc ion strongly absorbs near-ultra-violet light with maximum absorption at350 rnp. Photometric titration is relativelysimple and free from the erratic behaviorof electrodes. The sensitivities of the scv-cral iodine end-point techniques, as foundby Bradbury and Hambly (1952) andduring this work, are as follows:Visual starch 10 ~eq/litcrColoriinctric starch 2 j..q/li terRinpcronwtric tcclmjque (0.02) j.m~/litcr calculatccl

Ultraviolet aLsorption

free ioclinc0.08 pcq/litcr in 0.01

N ioclicle0.015 ~cq/liter in 0.01

N iodideTitration procedure-A sample wasplaced in a 500-ml Pyrex bottle withrounded stopper and the reagents addedwith syringes. A loo-ml portion of the rc-sulting solution was removed with a pipettecalibrated to contain, Thiosulfate solu-tion was added to the bottle contents froma weight burette to nearly complete re-action (very light straw color). A 2Wmlaliquot of the reaction mixture was trans-

fcrred to the photometric titration appara-tus, which consisted of a Beckman ModelDU spectrophotometer with a modified ccl1compartment to accommodate a 2(&nlFlorcncc flask and a refcrencc air path con-

structcd with four mirrors mounted on amodified cell carriage. The contents of theFlorence flask wcrc stirred with a magneticstirrer and thiosulfate solution added with al-ml syringe microburette until the opticaldensity was approximately 0.2. Incrcmcntsof 0.01 ml thiosulfate solution were addedand the end point found from a plot of opti-cal density vs. added thiosulfate. The prc-cision of the technique is 0.003 g thiosulfatesolution (Table 1).Standardization of thiosulfate-Iodinegenerated from biiodate, iodatc, or dichro-mate is commonly used to standardize thio-sulfate solutions. Potassium dichromatc isnot suitable as a standard for millinormalsolutions, since the air oxidation error isdifficult to avoid at the high acidities rc-quircd. Even with the precautions of gen-eration of iodine under a nitrogen atmo-sphcrc and adjustment of the pI1 to 3 be-fore titration, intercomparison of assayed( bcttcr than 0.1% ) potassium dichromatcand potassium biiodate (G. FredcrickSmith Co., Columbus, Ohio) gave differ-ences of 0.5%.Either potassium iodate or potassium bi-iodate appears to be suitable as an iodinestandard. However, the iodate is easier todry and might be expected to bc more rc-liable. Comparison of potassium biiodate( G. Frederick Smith Co.), which had beendried over magnesium perchloratc and hada stated assay of lOO.Ol%, with potassiumiodatc ( Mallinkrodt ), dried at 18OC andccrtificd as 100.05+-99.95%, gave agrccmcntto 0.04%.

Preparation of dissolved oxygen solutionsThe apparatus shown in Fig. 1 wasused to prcparc solutions of known oxygencontent. Two high-vacuum three-way stop-cocks were joined to form a central cham-ber (4.7 ml) with four entrances, Thel-liter bulb shown at the top permittedvigorous boiling under vacuum, The Teflon

plunger syringe was filled with mercuryand degasscd under high vacuum beforeassembly of the apparatus. Fig. 1 showsthe apparatus during the removal of dis-solved gases from the water.


4/6


5/6

ACCUMCY OF WINKLER METIIOD 1398) The central chamber was opcncd tothe water and the syringe. The oxygen wasgently transferred to the bottle, where solu-tion took place in approximately 10 min.9) The apparatus was unclamped andthe mercury allowed to run back and forthfrom the central chamber thoroughly to mixthe solution, including the portion in thestopcock bore. The walls of the bottle wcrcinspected with a hand lens and all visiblemercury transferred back into the centralchamber by scrubbing with a globule ofmercury.10) The stopcocks wcrc rotated and the

sample bottle rcmovcd from the apparatus.The excess water was allowed to spill, andthe stopper was inserted in the bottle, thusisolating a known fraction of the addedoxygen.11) The fraction of the added oxygencontained in the bottle was determined bythe technique for the Winkler method out-lined above.12) The mercury content of the finaltitrated solution was determined by spcc-trophotometric analysis based on measurc-ment of the 320-rnp absorption peak of mer-curic tctraioclidc and at 350 mp to correctfor the trace turbidity of the solution. A12.5cm absorption cell permits the dctcr-mination of 0.1 +q of mercury to 10%.

RESULTSTable 2 shows the results for six prcpara-tions of oxygen standards. The oxygen wasan Air Reduction Company research grade

gas, with a stated maximum limit of possi-ble impurities of 0,14%. Dr. Bruce Bensonof Amherst College kindly analyzed theoxygen by mass spectrometry. The sum ofTAIILE 2. RL-JSZI~ISf titmtion of oxr~gen standards

0.6016 0.6018 0.0012 0.0012 0.6018 $0.030.5983 0.5990 0.0009 0.0012 0.5987 +0.070.6120 0.6131 0.0002 0.0012 0.6121 +O.Ol0.6162 0.6176 0.0006 0.0012 0.6170 -f-O.130.6188 0.6184 0.0010 0.0012 0.6182 -0.100.6167 0.6168 0.0003 0.0012 0.6259 -0.13

the nitrogen, argon, carbon dioxide, andwater was found to be 0.10 _+ 0.03%.The values for iodine found were basedon standardization of the thiosulfate withiodine generated from biiodatc. AS notedabove, comparison of assayed biiodate withcertified iodate gave agreement to 0.04%.This suggests that the uncertainty in dctcr-mining iodine was not greater than 0.04%,since the precision of the weight photo-metric titration for these quantities of iodinowas 0.01%. The accuracy evaluation at-tempted here is a comparison based on theabove compos,itions of the oxygen and ofthe iodine standards. An observed diffcr-cnce might be ascribed to an error in theindicated compositions, nonstoichiomctricbehavior in the Winkler reaction scquencc,or unknown errors produced in the manipu-lations.It was fortunate that the final solutionfrom the Winkler analysis was prcciscly thecomposition suitable for the simple photo-metric mercury analysis. The sources ofthe mercury were solution of the elementalmercury in the water (solubility 60 jhg/liter),minutc particles adhering to the samplebottle, and possibly the reaction of the mcr-cury with the dissolved oxygen during thesolution process. While the corrections formercury contamination are significant, anuncertainty of not more than 0.02% rcsultcdfrom this. source.The oxygen content of the rcagcnts wasfound from the addition of various amountsof the reagents to water scrubbed withnitrogen to produce a low oxygen conccn-tration and also by addition to water dc-gassed by boiling under vacuum. An un-certainty of not more than 0.02% rcsultcdfrom this source.In view of the several uncertainties, par-ticularly in the gas composition, and thevariance of the results, the small value forthe mean of the differences ( + 0.002%) inTable 2 is clearly accidental. The varianceappears to be produced mainly from thegas volume measurements. The barometricpressure readings had a variance of ap-proximately 0.03%, and the stopcock SYS-tcm requires the use of grease that may ac-


6/6

140 JAMES II. CAnPENTERcumulate and then be removed to produce RlWERENCESvolume changes of approximately 0.05%.The effect of these factors could be rc- ~ADUURY, J, II., AND A. N. HAMuLY. 1952. Aninvestizration of errors in the amncromctricI *duccd by rcdcsign of the system, but an ac- and starch indicator methods for the titrationcuracy of better than 0.1% dots not appear of millinormal solutions of iodine and thio-to serve any useful purpose. The high sulfate. Australian J. Sci. Rcs. Ser. A, 5:(0.02%) precision of the weight photo- ,541~554.KLOTS, C. E., AND B. B. BENSON. 1963. Solu-metric titration does have an advantage for bilitics of nitrogen, oxygen and argon in dis-such purposes as measurements of usolu- tilled water. JT M arine Res., 21: 48-57.bility as a function of temperature and KNOWLES, G., AND G. F. LOWUEN. 1953. Meth-salinity, in which unknown functional dc- ods for dctccting the end point in the titrationpcndencc must be developed from a limited of iodine with thiosulphatc. Analyst, 78:159-164.number of measurements. MO~~TIMER, C. H. 1956. The oxygen content ofair-saturated fresh waters. and aids in calcu-lating pcrccntage saturation. Intern. Assoc.

CONCLUSIONS Thcorct. Appl. Limnol. Commun. No. 6.Since the recovery of oxygen was quan- RICIIARD~, F. A., AND N. COIWIN. 1956. Someoceanographic applications of recent dctermi-titative to the expected accuracy of the prc- nations of the solubility of oxygen in scnpared standards (O.l%), it is rcasonablc to water. Limnol. Oceanog., 1 : 263-267.conclude that the Winkler method as out- TRUESDALE, G. A., A. L. DOWNING, AND G. F.LOWIXN. 1955. The solubility of oxygen inlined here has an accuracy of 0.1%. Com- pure water and sea-water. J.- Appl. - Chcm.parison of this technique With several other London, 5: 53-62.commonly used techniques shows differ- WHEATLAND, A. B., AND L. J. SMM[TII. 1955.enccs of up to 5%, and it is inferred that Gasometric determination of dissolved oxygenerrors of this marrnitudc result from inade- in pure and salinc water as a cheek of titri-mciric methods. J. Appl. Chcm. London, 5:

quacies of the other techniques. An adap- wIN~~L~l~*W 1888 Die Bcrtjmmung t,e~ mtation of the present technique for routinc . .Was&r geliisten Saucrstoffcs. Chcm. Bcr., 21:analysts will be described in another paper. 2843-2855.

vinkler method for do

Documents