January 15, 1930
I S D I - S T R I A L A S D ESGISEE'RISG! CHEllISTRY
20
Accuracy of the Gutzeit Method for the Determination of Minute Quantities of Arsenic' J. W. B a r n e s and C. W. M u r r a y IhsCCTICIDE
DIIISIOPI, BCRE4U
O F CHEMISTRY i \ D SOILS. ~ A S H I h G T O \ , D
D
.4TA on the accuracy of the Gutzeit method for the
C
The investigation here reported therefore v a s confined to determination of minute quantities of arsenic, when portions containing less than 40 micrograms of d s 2 0 s , as applied under commercial conditions, are given here. t'ypical of aliquots from samples containing not more than This method is very sensitive to physical conditions and 4 nig. somewhat sensitive to chemical impurities. Accuracy necesThe salient points in the method as used here are: sary under certain circumstances, for instance, in forensic CONCENTRATIONS I N REACTIONCHAMBER-~ldfUriC acid, 2 medicine, mould entail precautions economically unjuqtified in to 2 . 5 m d a l ; potassium iodide, 0.1 m d a l ; stannous chloride, ordinary commercial vork. The technic actually applied 3 drops of 40 per cent stannous chloride in cmcentrated hydroin this investigation is believed to be typical of that used in chloric acid. TSMPERATURE O F BATH-^^' to 20' c. well-equipped laboratories employing the method to deterSTRIPS-standard type (2.5 mm. wide) as found in chemical mine arsenical residue on foodstuffs. supply market. The writers haye been unable to find in the literature any SENSITIZING SOLUTIOX-~ grams mercuric hrsmide p:r 100 cc. definite discussion of the accuracy of this method.? It is 96 per cent ethyl alcohol. TIMEO F SOAKING-At least 3 hours. possible, of course, that such discussion occurs in articles ZIsc-30-mesh (commercial designation). on other subjects, but without mention in the titles, thus There are two parts to the problem: (1) What is the acbeing practically unavailable to the great majority of chemists. Articles in the particular field in which the authors curacy of determining arsenic in a given solution? (2) I s have used this method, that is, arsenical residue on fruits there any loss of arsenic originally o n a sample during the and vegetables, give as correct by simple assertion, or by process of oxidation, etc.? implication, results ranging from 2 t o 6 significant figures, Accuracy of Analysis of Solution a factor of 10,000 in the accuracy of the method. Analysis of the sdution in a method of this type il: subject, It was desirable to duplicate as nearly as possible actual conditions in a representative case of tlie uqe of this method, to error from tiyo sources, the error due to the variations to test the recovery in the process. A number of apples with respect to tlie standard used, and the error of the standwhich had not been exposed to any arsenicals were selected. ard. A composite solution of portions of t'he several oxidized The fruit was peeled and the peels were washed with dilute nitric acid (4 per cent) as a precaution. From the composite samples mas made. The arsenic content of this solution mass of these peels. BO-gram portions (representing approximately 3 applesj were weighed into Kieldahl flasks. ' T o i a i h of these flasks war added a solution of lead arsenate in very dilute 10 nitric acid in quantity calculated to give a con._ venient amount in the final aliquot. The samples were then oxidized with nitric and sulfuric acids, and the solution was made to 6 500 cc. From these solutions aliquots of 25 cc. n-ere dravn. The arsenical originally added =/r theoretically gave each of these aliquots 0.0273 mg. of arsenic trioxide (ds20al,which from 2 c > o - : ' * previous experience n-e had found to be in the region of maximum accuracy. For quantities of arsenic in excess of 40 microq s q o 2 a grams the stains became progressively denser, and t'he length hecame an i&&singly uncertain crciins AS,03 measure of the quantity of arsenic. If tlie aliquot Figure 1 from a sample reads in excess of 40 micrograms a smaller aliquot should be used. If the quantity of arsenic ' was such that 25-cc. aliquots should contain 0.0273 nig. A T ~ O ~ . in the sample is such that an aliquot must be less than 0.01 As minor variations in technic, reagents, etc , occur ironi part of the sample t'o come within t'liis limit, it is advisable day to day, in spite of the greatest care, sets clf these 2 % ~ . to use Smith's method (E. S. Dept. dgr., Bur. Chem., Circ. aliquots were run on several days, 150 aliquots in all iieing 102), whereby tlie arsine is passed into a solution of mercuric run. Figure 1 gives the distribution fiequency of the analyses chloride, which conrerts the AsHB into dsz03 and precipi- of the composite solution. The mean of these analyses is tates calomel (HgCl), the precipitate being Jyeighed and 0.0277 mg. The standard deviation ( u ) is 0.0048 ing. The checked against a titration of the ,2s203. This method allows probable error of the mean of these analyses is *0.00026 the taking of t v o duplicates of 40 per cent of the sample. mg. The probable error of the mean of two observations gives *0.0023 mg. as the accuracy (with respect to the 1 Received August 18, 1929. standard used) of reading duplicates under ordinary condiWhile this paper was in the hands of t h e publisher, a n article entitled tions of this method. "A S t u d y of t h e Accuracy of t h e Gutzeit Method for Arsenic," b y J. R . The formulas according to which these values were obNeller of t h e Washington Agricultural Experiment Station, appeared in the J . Assocn. OPcial A g y . Chem., 12, 332 (1929). tained are:
G ~
z
30
A N A L Y T I C A L EDITION
VOl. 2, No. 1 Recovery of Arsenic
where ZI
=
2v2 = N =
Nl = If Nl = If Nl = If N , =
residual (observation minus arithmetic mean) sum of squares of individual residuals total number of individual analyses used to determine the standard deviation ( u ) number of individuals included in sample for which probable error is sought N, we have the probable error of the mean of the entire series of analyses 1, we have the probable error of a single observation 2, we have the probable error of the mean of duplicates
The standard of reference in this case was a graph representing the analyses of aliquots of various sizes from a solution containing a known concentration of standard Asz03. Each point on the graph represented the mean of 10 determinations, the points being a t intervals of 5 micrograms from totals of 5 t o 40 micrograms. As the conditions under which these standard solutions were analyzed are essentially the same as those just described, the standard graph is subject t o the same probable error. The probable error for the method, then, is a combination of these two effects as expressed in the equation PE where P E
=
.\/(PEl)z
+ (PE%)*
= probable error of method as a whole
( P E l ) = probable error of reading referred t o the standard (0 0023)
(PEl) = probable error of standard (PE for 10 observations being 0.0010)
Substituting in the formula for propagation of error, P E = *0.0025. If, as frequently recommended, the reference is a series of single strips from the standard solution, ( P E 2 ) = 0.0032 (the P E for a single observation), using this value, the error in the method would be P E = *0.0039. (The mean of duplicates refers to single aliquots of a standard solution.) It seemed likely that the probable error thus found is a fixed quantity for the range from 0 to 30 or 35 micrograms. A short series was run as a check. Forty aliquots of 0.010 mg. Asz08 each were drawn from a standard solution. The probable error of duplicates of this series was *0.0023 mg. This seems to indicate that the probable error of reading strips is approximately constant over this range. The relative probable error of the method ranges from about 7 per cent for aliquots containing 30 or 35 micrograms to about 20 per cent for those containing 10 micrograms, and from there down it passes rapidly to a range in which the method is only roughly quantitative and still further to one in which it has only qualitative significance.
The arsenic content of forty samples of the original portions oxidized separately was determined in duplicate. The average of each pair of duplicates was treated as a unit. The mean of this series is 0.0271 mg., and the standard deviation is 0.0034. The probable error of a single observation is =t0.0022 (with respect to the standard graph). As this is practically identical with the P E of duplicates in the other series of determinations, the operations of digestion, etc., introduce no additional errors. Unless the manipulator gives very close attention to the process of oxidation, charring is likely t o occur. Some analysts have expressed a fear that this charring may result in a loss of arsenic. To test this point, aliquots containing 20 mg. of As203 were added to each of six samples. Three of these samples were allowed to char slightly and the other three were more completely carbonized. Within the limits of experimental error of the calomel precipitation method, there were no indications of loss of arsenic from charring. Conclusion
The order of magnitude of the error of the Gutzeit method is indicated by the data here given. Owing to the sensitiveness of this method, however, each laboratory using the method frequently should determine its own probable error. Observation of variations in the Gutzeit strips and application of statistical methods lead to the conclusion that when applied under ordinary commercial conditions, that is, in the absence of extraordinary precautions to control physical conditions, etc., the probable error of the mean of duplicate strips in the Gutzeit method is +0.0039 mg. Thus, for quantities from 35 micrograms down, the error ranges from 11 to 100 per cent, when the reference is a series of single aliquots from the standard solution. Where a graph has been prepared from a large number of aliquots from a standard solution this error may be reduced. Take sets of, say, 20 aliquots a t steps of 10, 20, 30, 40 micrograms. Plot the arithmetic means of these determinations. Use this graph as the reference, or, preferably, calculate the equation of the line and prepare a table of the values of the stains of different lengths. The probable error of the method becomes =t0.0023, which is 7 per cent to 100 per cent for the range discussed. If the total arsenic is more than 4 mg., necessitating an aliquot of less than 0.01 to come under 40 micrograms, the calomel method should be used. There is no loss of arsenic in the process of oxidation, etc., even when charring takes place.
Determination of Alcohol b y Pycnometer' Alex F. F u e r s t 1108 D E A N ST.,BROOKLYX, h-.Y .
HIS method. worked out when the prohibition law went into effect and the number of samples for alcohol determinations in the laboratory of the Pittsburgh Brewing Company increased greatly, is based on the apparent uniform expansion of dilute alcohol up to about 4.5 to 5 per cent and distilled water. Hundreds of alcohol determinations support this assumption. The advantage of this method lies in the avoidance of weighings a t temperatures different from the temperature in the weighing room, with its attendant dis1
Received August 20, 1929.
advantages of dew on the pycnometer and the weighing pan. Another point overlooked in most descriptions is the fact that a difference in the temperature of 0.1 degree affects the result more than a difference in weight of 1 mg. Apparatus
(1) A pycnometer, preferably 50-cc., straight-stemmed, with mark on the stem and without stopper. (2) A thermometer graduated to 0.1 degree. (3) A water bath that can hold several flasks, containing the
