Determination of Cobalt in Atmospheric Samples ROBERT G. KEElVAN AND BETTIE $1. FLICK Zndustrial Hygiene Laboratory, C. S. Public Health Service, Bethesda, Md.
A method has been developed for the determination of cobalt in atmospheric dust samples and applied to the analysis of environmental samples from the cemented tungsten carbide industry. The sample is fused with potassium peroxydisulfate after preliminary ignition of an? free carbon present; the cooled melt is dissolved in water, acidified, and diluted to volume; after thorough m i x ing, a suitable aliquot portion is removed, filtered, w-ashed, evaporated almost to dryness, oxidized with concentrated nitric acid, and evaporated to dryness; after solution in water and hydrochloric acid, the cobalt is determined colorimetrically by means of a modification of the nitroso R salt method. The application of this method to atmospheric carbide dust samples has indicated that for samples containing 0.1 mg. of cobalt or more, its accuracy is at least 95%.
A
iY-kLYTICAL methods for the determination of cobalt have been developed for application to such materials as ores, minerals, special steels, alloys and other metallurgical products, enamels, driers, soils, pasture samples, and biological specimens. Many of these methods consist of relatively timeconsuming procedures involving the separation of cobalt from certain common metals which constitute interferences ( I O , 2 2 , B j . I n recent years, considerable attention has been directed toward colorimetric methods of analysis for cobalt. Certain of these methods have proved more suitable for the routine analysis of a large number of samples than either the gravimetric or the volumetric methods in that they are, in general, more rapid, less involved, and more applicable to the lower ranges of cobalt content. However, most of these colorimetric procedures have been developed for specific applications to definite types of samples, and generally they are not too desirable for the analysis of atmospheric dust samples that may contain interfering substances. l-Nitroso-2-naphtho1, in addition to its use as a cobalt precipitant in gravimetric procedures, is a very sensitive and widely used colorimetric reagent for this metal.
It was introduced by Bellucci ( I ) , who stated that with this reagent 1 mg. of cobalt may be detected in 1 to 2 liters of water by the formation of a red precipitate. However, as a quantitative method, the use of 1-nitroso-2-naphthol necessitates long and tedious procedures and is a t times unreliable because of the poor quality of the reagent (4j. Although Black ($1 improved the accuracy of this method by use of the blraching action of sulfite on the yellow color of the reagent, his procedure was applied only t o samples containing less than 7.5 micrograms of cobalt. Applicable also to only the lower range of cobalt is the thiocyanate method (20) in which a blue-colored complex is formed when sodium thiocyanate is added to a cobaltous solution; best results are obtained with a limiting concentration of 16 micrograms of cobalt per milliliter of solution. Iron interferes and must be removed by zinc oxide separation. Nickel, copper, and bismuth also interfere. DeGray and Rittershausen (6) found potassium ferricyanide in ammoniacal solution, which produces a red color with cobalt, to be the most suitable reagent for the routine determination of cobalt in paint driers. This procedure gives an accuracy of only 0.06 mg. of cobalt and many metals interfere. I n a method developed by Cartledge and Xichols (S), the trioxalatocobaltiate ion is formed by oxidation with lead dioxide of a cobaltous solution in the presence of potassium oxalate; the intensity of the dark emerald-green color thus produced may be measured in a spectrophotometer. Copper and manganese must be removed previously and the oxidation must be carried out in a weakly acid solution. When cobalt is oxidized in ammoniacal solution, a pink color is produced by the resulting cobaltammine which is proportional to the amount of cobalt present (24). Iron and manganese, if present, will precipitate and must be removed by filtration. I n order t o avoid interference from nickel, a special procedure is necessary. Organic compounds such as terpyridyl (I@,o-nitrosophenol(5), o-nitrosocresol (7), o-nitrosoresorcinol (19), and dimethylgly-
oxime j211 have been successfully employed as quantitative reagents for the determination of cobalt. However, because of the presence of substances that may interfere in the application of many colorimetric cobalt procedures and the necessity for cobalt determination over a wide range of concentration, it was felt that the nitroso R salt method offered certain advantages of simplicity and adaptability over the other chemical procedures for application to the analysis of atmospheric dust samples. The opportunity to use the nitroso R salt method for the determination of cobalt in atmospheric dust samples presented itself in a recent survey of the cemented tungsten carbide industry by the Industrial Hygiene Division. NATURE OF ATMOSPHERIC DUST SAMPLES
The manufacture of cutting tools and dies in the cemented tungsten carbide industry involves such general operations as powder processing, pressing and rough shaping, sintering, brazing of tips to steel shanks or shrink-fitting of dies into steel cases, and final shaping. The fabricating metals include principally tungsten, titanium, and cobalt and, to a lesser extent, tantalum, columbium, and nickel. The resulting product is evtremely hard and tough, and generally consists of tungsten carbide, KC, or tungsten titanium carbide, WTiCa, sintered in a matrix of metallic cobalt powder. Sickel has sometimes been used in place of cobalt as the metallic binder; tantalum or columbium carbides are used in certain carbide blends for the purpose of providing special properties to the final product. Lampblack serves as the source of carbon in the preparation of carbides of the various metals. The determination of the various metallic constituents in over 800 dust samples collected during the environmental survey of this industry presented the problems of (1) the decomposition of the one or more carbides possibly present, (2) the complete removal of free carbon, (3) the solution of the resulting mixture containing metals-viz., tunisten and cobalt-of directly opposite chemical properties, and (4)the necessity of taking only a portion of each sample (many of these weighing in the neighborhood of 1 mg. and others ranging up to 300 mg.) for the analytical determination of the indicated elements present. The dust samples were collected by means of a direct current electrostatic precipitator; the aluminum collecting electrodes containing the samples were capped and shipped to the laboratory for transfer and analysis (25). PRELIMINARY TREATMENT OF SAMPLES
Determination of Sample Weight. The aluminum collecting electrodes containing the samples were found to have held the dust tightly t o their inner surfaces during shipment. Upon removal of the metal or Bakelite caps, the outer surfaces of the electrodes were wiped clean with alcohol on a gauze pad,
1238
V O L U M E 2 0 , NO. 1 2 , D E C E M B E R 1 9 4 8 dried bl- wiping with a clean pad of dry gauze, and weighed on a n analytical balance after equilibrium betireen the electrode surface and the atmospheric moisture had been established during a 15-minute Lraiting period. The samples were then transferred from the electrodes to Pyrex test tubes of 25-mm. diameter, using a minimum amount, usually 10 to 20 ml., of a 307, solution of ethyl alcohol and a rubber policeman. The inner surface was dried by forcing through it a wad of gauze; the outer surfaces w r e cleaned as described previously and the final weighing was made after another 15-minute eqnilibrium period. The sample weights were obtained from the difference in the two weighings of each tube. By evaporation to dryness in an oven a t 105' C., each sample was concentrated in the bottom of a test tube, after which it was ready for fusion and subsequent solution. Decomposition of the Carbides by Fusion with Potassium Peroxydisulfate. I n addition to the carbides and metallic cobalt, many of the dust samples also contained lampblack or graphite Lvhich was particularly heavy in those collected a t the powder processing operations of tungsten and titanium. I n view of the chemical nature of such a mixture there were certain problems associated with the decomposition of these samples. A number of fusion agents were tried, as it was felt that acid treatment would be too time-consuming and possibly more involved than a direct fusion method. hmong these Lvere potassium bisulfate, potassium hydroxide, and sodium and ammonium nitrates n-ith a little potassium chlorate. None of these reagents produced a clean melt, each effecting only partial decomposition of the sample material as evidenced by the persistence of a considerable amount of grayish black carbide. Finally potassium perosydisulfatc was used : this reagent 'rras found to cause complete decomposition of the carbides and to givc a clean n-hite or yellon-ish-xhite melt in 3 or 4 minutes of heating over a 1Ieker burner. (a blue color was produced in the samples containing a high percentage of cobalt.)
It was found advantageous to heat the samples over the burner before fu4on: this preliminary heating removrd most of the free carbon already preqent in the sample. .I sufficient amount of potassium peroxvdisulfate was then added to each sample, xvhich was hratcd gently a t first and then more qtrnngly, Tvhilc the tube was rotated slowly and a t an angle to allow the molten perouydisulfate to conic in contact wit,h and to decompose the carbide.;. By this treatment cobalt, nickel, and titanium m r e converted to their sulfates, tungsten to potassium tungstate. and tantalum and columbium to oxidea or tantalates and co1umbatc.s. Upon cooline, the melt K R S dissolved in hot water. Csuallv the sample ivas diluted t o a volume of 25 nil. with water and 5 nil. of 6 Ar hvdrochloric acid: the hcavieqt samples were dilllted to 50 or 100 ml. During the course of diluting to a definite volume, tungstic acid and probably some hvdrolyzed titanium salt precipitated. Cobalt, hoxrever, remained dissolved in the 1.2 S acid solution, as p h o n by the folloviing exoeriment. Tire equal portions of a given samplr. one portion filtered frcc of tungqtic acid and the other portion unfiltered. n-ere analyzed by the nitroso R salt method. hlniost identical cobalt values were obtained b r carrying out these detrrminations on a 10.7-mg. dust samnle collected at a powder-pressing operation: the values for the filtered and unfiltered portions were 0.097 and 0.09S mg. of cobalt, respectively. The 0.001-mg. dirparity between the two cobalt values can be ascribed rxincipal1.v to espcrimental errors rather than t o a significant variation in the cobalt content of the filtered and the unfiltrred portions. Before a n aliquot portion v a s removed for each cobalt detcrmination, every sample was shaken thoroughly to distribute the preripitate as evenly as possible throughout the solution. h definite percentage of all constituents vias thus removed, so that the remainder of the sample, with the constituents in the same relative proportions, could be used for subsequent determinations. Those aliquot portions of samples which x e r e more than slightly turbid were filtered and washed three times Ivith distilled water. One of the two modifications of the nitroso R salt method used in this laboratory was then applied. COBALT DETERMINATION BY THE NITROS0 R SALT METHOD
Kitroso R salt was introduced by van Klooster in 1921 as a reagent for the detection of cobalt (14). When R salt, sodium
1239 2-naphthol-3,6-disulfonate is treated with nitrous acid, sodium is formed. This is commonly nitroso 2-naphthol-3,6-disulfonate, known as nitroso R salt. K i t h this substance cobalt forms, in the presence of sodium acetate, a red soluble dye of the composition (CloHjSOsSzSan)aCo.Interfering colors, formed in the presence of iron, nickel, copper, and chromium, are destroyed by boiling with concentrated nitric acid after all the other reagents have been added. Xitroso R salt has been used for the quantitative colorimetric determination of small amounts of cobalt in biological samples (12, 16, 25), soils and pasture samples (11, 12, 171, limonites ( 1 6 ) ,and steels and other metallurgical products (9,271. Certain modifications of this method appeared to be adaptable to the dcterniination of cobalt in the environmental dust samples collected during the course of the cemented tungsten carbide survey. Of spe:ial interest and value was the investigation of 'I-oung, Pinkney, and Dick ( R T ) , who reported that the final cobalt color comparison could be made satisfactorily in the presence of a t least 1000 times as much iron, 100 times as much titanium and tungsten, and not more than 25 times as much nickel. The noninterference of these metals to the extent indicated wa,5 borne out by the present investigation in irhich the authors obtained satisfactory recoveries of known amounts of cobalt added to the environmental dust samples containing all the metals of the final carbide blends. JIoreover, no deviation from the reagent blank (including 0.17, aqueous solution of nitroso R salt) reading of the spectrophotometer was obtained with either sodium tungstate or titanium sulfate when each of these salts was dissolved in \rater and carried through the nitroso R salt procedure for cobalt. Preparation of Standard Cobalt Solution. Cobalt chloride was dissolved in water and converted to the chloropentammino cobaltic chloride 115). The ammino salt n-as then transferred to a porcelain crucible, ignited to black cobaltic oxide, ('0104, and diisolved in dilute sulfuric acid, and the sulfate was crystnllized by evaporation t o drive off excess water. The sulfate XTa; filtered, dissolved in water, and again evaporated to crystallization. These crystals were filtered and dissolved in n-atcr, and to the resulting solution was added a saturated solution of ammoniuni oxalate. The precipitato of the salmon-colored cobalt' oxalate, C O C ~ Owas ~ , filtered on Whatman S o . 50 paper, transferred to a m-ide-mouthed bottle, and dried for 2 or 3 hou solution containing 10 mg. of cobalt per liter disolving 0.0240 gram of this salt in 10 nil. of acid and diluting to 1000 nil. viith water. One hundred milliliters of this solution, to which \rere added 10 nil. of 6 .V hydrochloric acid, were diluted to 1000 ml., and provided a standard solution of 0.001 mg. of cobalt, per ml. Procedure. Two procedures were used for the cobalt d(,termination. One of these (16) n-as applied to the anal of the lighter dust samples who,qe aliquot portions should have contained less than 0.050 nig. of cobalt. Hon-ever, Trith this procedure difficulty n-as encountered in the selection of sufficiently small aliquot portions of certain samples whose cobalt pcrcentages were higher than anticipated. Redilutions and consequent time loss necessitated the adoption of another procedure applicable to a rrider range. The second procedure was that of 1-oung, Pinkney, and Dick ( 2 7 ) . These investigators applied their procedure to the detcrminatiori of cobalt in amounts of 0.010 to 0.500 mg., which \vas a suitable range for most of the environmental dust samples. Since little or no iron or copper was present, the preliminary hydrogen sulfide separation was eliminated. The procedure used in this investigation is as follows:
A suitable aliquot portion was removed from each sample after thorough mixing of the insoluble residue of tungstic acid and the solution, filtered if necessary, and cvashed three times with distilled water. It was evaporated almost to dryness in a 125ml. Phillips beaker, and after 2 ml. of concentrated nitric acid were added, it was evaporated t o dryness. The sample was dissolved in 10 ml. of water and 2 ml. of 6 N hydrochloric acid with boiling. The sample solution was neutralized with 207, sodium hydroxide,
ANALYTICAL CHEMISTRY
1240 Table 1.
Sample KO.
684 739 682 602 797 859 656 877 785 783 934 922 877 936 272 757 793 976 Mean, 18 samples
Recovery of Cobalt and Deviation of Recovery from Expected Value for Varying Amounts of Cobalt Added to Four Equal Aliquot Portions of 18 Dust Samples Cobalt Added, Rlg. Kone 0,010 0.020 0.060 0.100_ Total Cobalt Content of Aliquot Portion as Determined b y Xitroso R Salt Procedure, RIg. a b C d e 0.027 0.036 0.046 0.078 0.127 0.035 0.044 0.054 0.128 0.084 0,044 0.054 0.064 0.144 0.094 0.076 0,083 0.092 0.172 0.123 0.083 0.089 0.102 0.178 0.132 0.115 0.124 0.135 0.208 0.164 0.117 0.127 0,136 0.208 0.165 0.118 0.128 0.138 0.212 0.164 0,149 0.159 0.167 0.242 0.194 0,184 0.192 0,204 0.279 0.232 0.188 0.198 0.208 0.228 0.281 0.211 0.219 0.226 0.256 0.305 0 230 0.242 0.246 0.276 0.325 0.245 0.255 0.265 0,298 0,345 0.274 0.279 0,290 0.318 0.372 0.289 0.298 0.307 0.331 0.381 0,302 0.309 0.324 0.354 0.401 0.353 0,363 0.368 0.400 0.453
0.1688
0.1777
0.1873
0.2162
0.2645
Cobalt Added. Mg. Cobalt Added, hlg. 0,010 0 020 0.050 0.100 O.OIO O.OZo 0.050 o.loo %viation of Recovery from Expected Value, Mg Recovery of Added Cobalt, .\lg - [ ( b - a) [ (C a ) - [ (d - a) [ ( e - a) b - a c - a d a e - a 0 0101 0 0201 0.050l 0.1001 0.009 0.019 0.051 0.100 -0.001 -0.001 +0.001 0.000 0.009 0.049 0,019 0.093 -0.001 -0.001 -0.001 -0.007 0.010 0.020 0.050 0.100 0.000 0,000 0.000 0.000 0,008 0.017 0.048 0.097 -0.002 -0.003 -0.002 - 0.003 0.006 0.019 0.049 0,095 -0,005 -0.004 -0,001 -0.001 0.009 0.020 0.049 0,093 -0.001 -0.001 -0,007 0.000 0.010 0.019 0.048 0.091 0.000 -0,001 -0.002 -0.009 0.010 0.020 0.046 0.094 0.000 0.000 -0.006 -0.004 0.010 0.018 0.045 -0.002 -0,005 0.093 0.000 -0,007 0,008 0.020 0.048 0,095 -0,002 -0,002 0.000 -0.005 0.010 0.020 0.040 0.093 0,000 -0.010 -0,007 0.000 0.008 0.015 0.048 0,094 -0.005 -0 005 -0.006 -0.002 0.012 0 046 0.016 0.093 -0.004 io.002 -0,004 -0.005 0.053 0.010 0.020 f0.003 0.100 0.000 0.000 0.000 0.005 0.016 0.044 0.098 -0.005 -0,004 -0.002 -0,006 0,009 0.018 0.042 0.092 -0.001 -0.002 -0.008 -0.008 0,007 0.022 0.052 0,099 -0,003 +0.002 +0.002 -0.001 0.010 0.015 0.047 0.100 -0.003 0.000 -0 003 0.000
-
-
0.0089
using phenolphthalein as indicator, and 2 ml. of Spekker acid 1150 ml. of phosphoric acid (sp. gr. 1.75) and 150 ml. of sulfuric acid (sp. gr. 1.84) made up to 1000 ml. with water], 10 ml. of 0.1% aqueous nitroso R salt solution, and 10 ml. of 5070 (W/V) sodium acetate trihydrate solution were added. After the solution had been brought to a vigorous boil, 5 ml. of concentrated nitric acid were added and the sample was boiled from 1 to 2 minutes. I t was cooled and diluted to a volume of 100 ml. with distilled water, and the optical density difference xas obtained between it and a reagent blank balanced a t zero on a Coleman Universal spectrophotometer Xodel 11 a t 510 mp. The amount of cobalt was estimated from a standard curve plot of optical density against concentration, prepared from a series of 0- to 0.500-mg. cobalt standards carried through the same procedure. ACCURACY OF METHOD
To test the accuracy of the nitroso R salt procedure for the analysis of the environmental carbide dust samples, a series of cobalt recovery determinations was carried out. I n these determinations five equal aliquot portions Kere taken from each of a number of environmental samples, representative of sampling location and cobalt content. These samples contained tungsten, titanium, cobalt, and such other constituents as are normally present after the preliminary treatment of transferring, fusing, and diluting to a definite volume with water and hydrochloric acid. To four of these portions were added 0.010, 0.020, 0.050, and 0.100 mg. of cobalt, respectively. These, along with the fifth portion containing no added cobalt, were then carried through the analytical procedure described above. The results of this investigation, showing the recoveries of added cobalt with the individual and mean deviations over the range of 0.037 to 0.453 mg. of cobalt, are shown in Table I. The cobalt determinations given in Table I cover essentially the entire working range of those made on the environmental samples of the carbide survey. The four different, known quantities of cobalt xere added to four of the five equal aliquot portions of each dust sample in order to observe what, if any, recovery differences might result on thus elevating two of these aliquotsthose containing 0.050 and 0.100 mg. of added cobalt. The significance of such an approach and the interpretation of the recovery data reported in Table I are brought out in the folloning statistical analysis. Statistical Analysis of Cobalt Recovery Data (by Rosedith Sitgreaves, Assistant Statistician, Industrial Hygiene Division, U. S. Public Health Service). Because of sampling variability and possible errors of measurement, repetitions of the present investigation might well yield different recoveries of the added cobalt. A lower bound on the true amount of added cobalt recovered for each of the four amounts of cobalt added is given in
0.0185
0.0473
0.0957
-0.0011
-
-
-0.0015
-0,0027
-
-0,0043
Table 11. .4nalyses of Cobalt Recovery Data Item Recoverv of added cobalt
- 1.74051 Lowkr bound on 70 of added cobalt recovered a t 95% confidence level
Cobalt Added 0,010mg. 0.020 mg. 0.050 rng. 0.100 mg.
0.0089 0,00040
0.0185 0,00047
0.0473 0.00079
0.0957 0.00073
0,0082
0.0177
0,0458
0,0944
82.0
88.5
91.8
94.4
Table 11. From these lower bounds corresponding lower bound are derived for the per cent of added cobalt recovered.
If p is the true amount of added cobalt recovered when a given - in repeated trials amount of cobalt is added, the quantity __
’sa ’
based on 18 samples is distributed as Student’s t with 17 degrees of freedom. From a table of t (8) it is found that the probability 4 - p
before a trial is made that -will be less than f1.740 is 0.95. SZ Hence it may be said with a confidence coefficient of 9570 that the true amount of added cobalt recovered is no less than 4 - 1.740 Sa. Corresponding lower bounds on the percentage of added co4-p
balt recovered are derived by dividing 100 X ___ by the amount
Sa
of cobalt added. I t may be said, therefore, with a confidence coefficient of 95%, that when 0.010, 0.020, 0.050, and 0.100 mg. of cobalt are added, respectively, t o aliquot portions of dust samples containing varying amounts of cobalt, the added cobalt recovered by the nitroso R salt method described above is no less than 0.0082, 0.0177, 0.0459, and 0.0944 nig., corresponding differences between the amount of cobalt added and the added cobalt recovered being no greater than 0.0018, 0.0023, 0.0041, and 0.0056 mg., respectively. Furthermore, while the absolute value of the difference increased with the increased amounts of cobalt added, the proportion of added cobalt which was not recovered tended to decrease. Indeed it may be said, again with a confidence coefficient of 9570, that the percentage of added cobalt recovered by the nitroso R salt method as described is no less than 82.0, 88.5, 91.8, and 94.47,, respectively, when the amounts of added cobalt increase from 0.010 to 0.100 mg. I n making a statement with a given confidence coefficient it must be remembered that the particular statement made may be either correct or incorrect. However, if in a long sequence of trials (not necessarily involving the same experiment) statements are made consistently Lvith a confidence coefficient of a given magnitude, say, 957& it is expected that the statements made will be correct 9570 of the time. If greater protection against incorrect statements is desired, it is possible to increase the magnitude of the confidence coefficient to, say, 9970. However, the
V O L U M E 20, NO. 1 2 , D E C E M B E R 1 9 4 8 price paid for greater protection is generally an increase in the length of the interval said to cover the true value. Thus, in the present instance, lower bounds on the true amount of added cobalt recovered a t the 99% confidence level are given by 3 2.567 Sz. For each of the amounts of cobalt added these values are less than the lower bounds given above. CONCLUSIONS
A method for the determination of cobalt in atmospheric dust samples has been developed and applied successfully to a large number of environmental samples secured in the cemented tungsten carbide industry. During this investigation there vas developed a means of “opening” the carbide dust, removal of the carbon, and solution of the metallic components as an initial stage of the analysis. The nitroso R salt method has been examined for its accuacy in estimating the cobalt content of atmospheric dust samples. Recoveiies of known amounts of cobalt added to a qeries of atmospheric dust samples reveal that the perccntage of added cobalt recovered tends to increase with increasing aniounts of added cobalt, a lower limit on the percentage recovered approaching 95 as the amount of added cobalt approaches 0.1 ing. Thus, if the amount of cobalt in an atniojpheric duqt sample is of the order of magnitude of 0.1 nig. or greater, it may be concluded that a t least QZm0 of the cobalt is recovered by the method described. LITERATURE CITED
(1) Bellucci, I., Gaze. chim. ital., 49, 294 (1919). (2) Black, I. A., Soil Sci., 51, 387 (1941). (3) Cartledge, G. H., and Nichols, P. M., IND.ENG.CHEM.,ANAL. ED., 13, 20 (1941).
1241 Clarke, W. W., I r o n Age, 150, 45 (Dec. 3, 1942). Cronheim, G., IND.ENG.CHEM.,ANAL.ED., 14, 445 (1942). DeGray, R. J., and Rittershausen, E. P., Ibid., 14, 858 (1942). Ellis, G. H., and Thompson, J. F., Ibid., 17, 254 (1945). Fisher, R. A., “Statistical Methods for Research Workers,” 8th ed., p. 167, New York, G. E. Stechert & Co., 1941. Haywood, F. W., and Wood, A. A. R., J . Soc. Chem. Ind., 62,37 (1943). Ilinski, M., and Knorre, G. van, Ber., 18, 699 (1885). Kidson, E. B., and Askew, H. O., N u Zealand J . Sci. Tech., 21B, 178 (1940). Kidson, E. B., Askew, H. O., and Dixon, J. K., Ibid., 18, 601 (1936). King, .4., “Inorganic Preparations,” p. 103, New York, D. Van Nostrand Co., 1936. Klooster, H. S. van, J . Am. Chem. Soc., 43, 746 (1921). McNaught, K. J., Analyst, 67, 97 (1942). McNaupht, K. J., N e w Zealand J . Sci. Tech., 18, 655 (1937). I b i d . . 26 14 (1938). Moss, M. L., and Mellon, M. G., IXD. ENG.CHEY.,A N ~ LED., . 15, 71 (1943). Overholser, L. G., and Yoe, J. H., I b i d . , 15, 310 (1943). PutschB, H. M., and Malooly, IT’. F., ASAL. CHEW.,19, 236 (1947). Sandell, E. B., “Colorimetric Determination of Traces of Metals.” p. 206, New York, Interscience Publishers, 1944. Scott, W.IT.,“Standard Methods of Chemical Analysis,” 5th ed., N.H. Furman, ed., p. 315, New York, D. Van Nostrand Co., 19.39. Seifert, H. E., Kecnan, 1%. G., and Fairhall, L. T., P u b . HeaZth Rapts., 60, 441 (1945). Snell, F. D., and Snell, C. T . , “Colorimetric Methods of Analysis,” 1701.I. u. 324, New York, D. Van Nostrand Co., 1936. (25) Stare, F. J., and Elvehjem, C. A , J . Bid. Chem., 99, 473 (1933). (26) Treadwell, F. P., and Hall, W. T., “Analytical Chemistry,” 1701. 11, 9th ed., p. 199, New York, John Wiley & Sons, 1945. (27) Young, R. S.,Pinkney, E. T., and Dick, R., IND.ENG.CHEM., -4h-a~.ED., 18, 474 (1946). RECEIVED June 30. 1948.
Semi-Self-Filling Micropycnometers Drainage of Micropipets HERBERT H. ANDERSON Harvard University, Cambridge, M a s s . A new- self-adjusting, or semi-self-filling, Ostwald type of micropycnometer has increased the accuracy fivefold, while capillary action completes the filling; a 1-ml. model has a reproducibility of 1 in 40,000. Satisfactory micropipets from 3.7 to 0.7 ml. in size conformed to drainage values derived from the equation (Robs. - R i ) ( T ) = k; previous investigators have found a similar relationship to apply to larger equipment. Studies on still smaller equipment include those on a self-adjusting, self-filling 0.002589 j=0.000006-ml.micropipet. A brief comparison of these new results and those of certain other investigators is presented.
D
U R I S G sustained inveetigations on inorganic liquids, the author has often found it highly advantageous to be able to manipulate rather moderate volumes of liquids ( 1 ) . I n the present article a reproducibility of 1 part in 10,000 v a s attained in a 0.25-ml. micropycnometer of a new semi-self-filling design which may be considered an improvement on the Ostwald pycnometer ( 1 4 ) .
tomed flask with a ground stopper bearing a capillary hole. Such a model is, in the opinion of the present author, unsuited for liquids such as phosphorus trichloride, and unsuited for accuracy better than 1 to 1000 on ordinary organic liquids. It seems advisable to avoid even a single stopcock, such as is present in a completely linear model ( I C ) , as well as an excessive length of small connecting tubing ( 7 ) .
MICROPYCNOMETERS SELF-ADJUSTING OSTWALD PYCNOMETER
Pycnometers of 1-ml. size show a considerable variation in both design and accuracy; Ostwald-Luther (14, p. 241) Presents a diagram of a 1-ml. pycnometer consisting of a tiny round-bot-
Bulb F (Figure l), of appropriate size, is blown from a 300mm. length of Pyrex capillary tubing of 0.5-mm. inside dianleter and 5-mm. outside diameter; the taper must be smooth, with a
