I S D USTRIAL A S D EA'GINEERIKG CHEMISTRY
November, 1925
explanation is offered for the failure of these two runs; however, if they are omitted from the calculation of the average yield of runs a t 350" C., the remaining seven show a yield of 89.1 per cent. From these runs i t is apparent that variation of the temperature from 300" to 350" C. does not materially affect the yield or the quality of the hexachloroethane. At 200" C. the yield is considerably decreased.
1183
Conclusions
Hexachloroethane can be produced by direct chlorination of ethylene with chlorine using activated charcoal as a catalyst. . The principal advantages of this method are the high purity of the product, the comparatively low temperature (300 O to 350" C.) a t which the reaction takes place and the high yield (90 per cent) of hexachloroethane obtained.
The Evaluation of Chlorates' A Comparison of Seven Selected Methods Applied to a Single Lot of Potassium Chlorate of High Purity By E. C. Wagner LNIVERSXrY OF PEENSYLVANIA, PHILADELPHIA, P A .
URITY requirements for chlorates are sometimes high. During the war government specifications for primer chlorate called for a purity of 99.85 per cent.2 Satisfactory analysis of such material requires a method the degree of accuracy of which is definitely known. Descriptions of methods in the literature or in the manuals often include no information as to accuracy, or give information which is of doubtful value. Comparisons of seven selected methods applied to the evaluation of a single lot of chlorate of very high purity was therefore undertaken in the hope of revealing their important merits or demerits. d further improvement in Bunsen's evolution method is effected by use of a new receiver.
P
M e t h o d s Investigated
Volumetric Methods (1) Bunsen's evolution method (2) Method of Ditz
(3) Method of Kolb and Davidson (4) Ferrous sulfate excess method Note-Knecht's titanous chloride method [J. SOC. C h e m I n d . , 27, 434 (1908)]was not subjected to trial. Mohr's and Volhard's titrations, for determination of chloride formed b y reduction, are referred to under ( 7 ) .
Gravimetric Methods ( 5 ) Reduction to chloride by ignition with ammonium chloride
(6) Reduction to chloride by evaporation with hydrochloric acid ( 7 ) Reduction t o chloride by sulfurous acid, and precipitation as silver chloride M a t e r i a l Analyzed
The potassium chlorate used for the comparisons had been crystallized four times from hot filtered solution. Each crystallization was analyzed for chloride and bromate, using the methods previously d e ~ c r i b e d . ~The product of the fourth crystallization, dried a t 100" C., was found to contain 0.013 per cent chloride (KC1) and 0.0009 per cent bromate (KBr03). I t s purity may be assumed to be not less than 99.98 per cent. Bunsen's Evolution M e t h o d
The procedure described previously4 was for these trials improved by use of a new receiver, whose design obviates Received February 14, 1925. A t present the requirement is only 99.35 per cent; U. S. Army Ordnance Specification 50-11-11. * THISJOURNAL, 16, 616 (1924). ' For two investigations of Bunsen's method not cited in the previous paper, see Topf, Z . anal. Chem., 26, 295 ( 1 8 8 i ) , and Jander and Beste, C. A , , 18, 1625 (1924). 1
the need to transfer the liquid to another vessel for titration of iodine. There are thus eliminated both the "oxygen error" and loss of iodine by volatilization-two sources of possible error in the usual procedure. APPARArus-The appearance of the assembled apparatus is shown by Figure 1. The construction of the new receiver is more clearly indicated in Figure 2. PROCEDCRE-The analysis was conducted essentially as described previously. Samples were taken as 10-ml. aliquots of a 1 per cent solution, and decomposition was effected by means of 5 ml. of 40 per cent hydrobromic acid, and in'a continuous stream of oxygen-free carbon dioxide. Into the bottle, A , of the receiver were introduced 40 ml. of 10 per cent potassium iodide solution and 100 ml. of water. The trap, C, was charged with 10 ml. of 10 per cent potassium iodide solution. The receiver was placed in a 2-liter beaker containing cold water. When the distillation was ended, the receiver was disconnected and rotated to dissolve solid iodine. The trap was connected a t f with the carbon dioxide generator (Figure l), the cap B was slightly lifted, and the contents of the trap were forced into A by pressure of carbon dioxide. The trap was washed out with several portions of water, introduced through f (the stopper e being removed), using the procedure just described. The cap B was then lifted out, held above A , and the internal surface of the cap and the inlet tube, b, were washed with water. The free iodine was titrated at once with 0.1 N thiosulfate, added from a weighing buret. Blank analyses were conducted in the same way. All the blank analyses made in this apparatus gave a result of zero-i. e., the liquid gave no color on addition of starch indicator.
ANALYTICAL REsuLTs-The last series of trials yielded the following results: 99.91, 100.00, 99.99, 99.96, 99.94 per cent, average 99.96 per cent. The t'ime required for an analysis, exclusive of the preparation of the sample, was somewhat less than an hour. This time can be reduced, without sensible loss in accuracy, by use of a volume buret. Iodometric M e t h o d of D i t z j
The chlorate is brought into contact with potassium bromide and hydrochloric acid in an all-glass apparatus consisting of a 1500-ml. bottle with attached addition-funnel and a trap for potassium iodide solution. After reaction is complete the liquid is diluted, potassium iodide solution added, the contents of the trap washed into the bottle, and the free iodine titrated with thiosulfate. Fifteen results reported by Ditz ranged from 99.81 to 100.17 per cent, and averaged 6 Chem. Zlg., 26, 727 (1901); 2. anal. Chem., 46, 532 (1906); Classen, "Ausgewahlte Methodeu der analytischen Chemie," Vol. 11, 1903 p. 370. Cf. Rupp, Z . anal. Chem. 66, 580 (1917); Moerck, J. A m . Phorrn'l A S S O C . , a, 155 (1913).
.
99.99 per eerrt. The nietliod has b w i rwoinriierrded for standardizat,ioii of tliiosulfate solutions, using putassiinn t,lie priinaiy standard.& 11s of t,liis inet.liod a srnaller (50(hd.) reaction bottle vas iised. The smaller capaeity of the bottle restrict,nl the dilution prcvious to ndrlition of potassium iodide, and tlie grt’ater final cuircentratim of hydriodic acid which resultid imide t,hc 1,lank analysis indis~)ensable. It was ritial. moreover, to time the exposure of the liquid during t,itratioii. and t o stir the blaolcs and the analyses
incan 9!1.!!7 per rent. TVit,li lrra liyrirohro~nicacid the ?esults were low. The time required for this analysis, exclusive of the preparation of t,he mnple, was about 25 minutes. lodometric Method of Koib and Davidson7
Tlic chlorate is caused to react directly with potassium iodide arid hydrochloric acid in an iodine flask previously filled with carbon dioxide. All liquids used should be airfree. TVhen reaction is complete the liquid is diluted (for u~hiclithe authors state that ordinary distilled water may be uscd) and the free iodine titrated. In their test-analyses Kolh and Davidson obtained 99.68 and 99.96 m r cent. meail 99.82 per cent. Kolb and Davidson reDorted a iieelieible blank. The writer’s blanks varied from 0.02 ml. 8.06 ml. of 0.1 N thiosulfate. When ordinary distilled water (not air-free) was used for dilution, the blanks increased to 0.13-0.14 ml., and the main titrations likewise.
o t
I’KOCBUURE-~~O a 500-ml. iodine flask were measured 50 ml. of 25 per cent hydrochloric acid (sp. gr. 1.125), previously saturated with carbon dioxide. Air in the flask was redaced by carbon dioxide, m d 2 grams of potassium iodide (iodate-fr&) were introduced. The chlorate sample (0.1 gram in a I0 mi. alisuot) was added irom a &et. and the flask wss stonoered and allbwed to stand in t h r b k k ’ from 15 minutes to &hour. The liquid was then diluted with 200 ml. of water (250 ml. in the blanks) and the iodine was titrated with 0.1 N thiosulfate, allowing 5 minutes for all titrations.
ANALYTICAL hSULTs-The various conditions used, and the rcsults obtained, are shown below: COXDITIONS
Per cent purity
60 mI, 26 per cent sir-free HCI; 2 grama KI:
100.28
Same; time, 45 minuter
100.40 100.23
Samr; ordinary distilled water lor dilution
100.44 100.38
SO ml. 22 per cent (6 X ) air-free MCI: 2 mama KI; 811 WBLei air-liee; time, 43 minuter
100.47 100.49
Same; ordinary distilled wster for dilution; time, 00 minYteS
100.47
time, 15 minutea
Figure I-Assembled
B ~ n s e nA p p ~ r a f m
about equally during tlie period of exposure. The blank titrations for the procedure adopted ranged from 0.10 to 0.15 ml. of 0.1 A’ thiosulfate. Ditz made no mention of blanks iri reporting liiS work. PROcL‘ounE-htO the reaction bottle were put 2 grams of solid Dotassium bromide, and the chlorate sample, added from a pipet (0.1 gram in a 10-ml. aliquot). The trap was charged with 10 ml. oi 10 per cent potassium iodide solution. The apparatus was assemblcd, and 30 ml. oi concentrated hydrochloric acid wcrc added through the scparatory funnel. After 10 minutes 200 ml. of water C245 ml. in the blanks) and then 10 ml. oi 10 per cent potassium iodide solution were added. The bottle was sliakeir t o dissolve vaporized bromine. The contents of thc trap were blown back into the bottle and the trap was washed out with water. The ground-in cap with its attachments was withdrawn, washed, and the free iodine titrated with 0.1 N thiosulfate. added irom a volume buret. The titration was limited sharply to 5 minutes, and an effort was made to proceed identicaliy in blanks and analyses.
ANALTTICAL ItcsvLTs-The procedure described yielded the rcsu1t.s 100.01 and 99.97 per cent, mean 99.99 per cent. When only 1grain of potassium bromide was used, the results were slightly low: 99234, 99.94, 99.83, 99.i0, 99.92 per cent, svcraging 99.87 per cent. When the potassium bromide and hydrochloric acid were replaced by 40 ml. of 40 per cent hydrobromic acid, the results were 99.93 and 100.00 per cent, a
Pre$euiur and Tetzloff, 2. nnal. Cham.. 45, 530 (1906).
100,57
100.47
Co&wxwi-In practically all trials in which complete decoinposition was assured this method yielded results about 0.4 per cent too high, in spite of carefully conducted blanks and exclusion of air. The failure of the method, due to liberation of excessive iodine, recalls the similar behavior of potassium dichromate in contact with hydriodic acid, in which reaction the iodine separated has been reported to he from 100.11 per cent to 100.7 per cent of the theoreticel.* This irregularity in the dichromate reaction is usually considered to be an “oxygen error,’’ perhaps involving the activity of some elusive catalytic impurity not removable by crystallization. The conditions present in the writer’s trials of the Kolb and Davidson method might be expected to prevent an oxygen error, or to permit its measurement and a correction of the results. No adequate explanation for the persistent positive error has come to mind. I n judging tho method it must be remembered that Kolb and Davidson’s results were not affected by an obvious error of this kind (unless their chlorate was impure), even with blanks omitted. l z . onge?u. Chem., 17, 1883 (19C4); 18, 1098 (1905); Kolb. Cham. E&, 86. 635 (1912). CI. Kolthoff. Phorm. Wsskbl.. 66, 460 (1919); C. A,,
18. 1434 (1918). *Wagner, Z. mors. Chcni.. 19, 427 (1899): McCrosky. f. Am. Chcm. Soc., 40, 1664 (ISIS); Kolthoff. Phorm. Wedbl., 66, 514 (1919); C. A,. 111, 1435 (1919); Vosburgh, J . Am. Chcm. Soc., 44, 2120 (1922); Bray end Millcr, I b i d , , 46, 2210 (1924); Popoff and Whitman, Ibid.. 4P, 2271 (1925).
I N D V S T R I A L AiVD ENGINEERING CHEMISTRY
November, 1925
The iodonietric method of Luther and Rutterg is similar to the preceding method, the reaction being accelerated by tetravalent vanadium and by heating in a closed flask, so that little acid is necessary. This method seems to offer no advantage over others more convenient, and was not subjected to trial. F e r r o u s S u l f a t e Excess Methodlo
The chlorate is heated with a measured excess of standard determined by titration ferrous solution, and excess Fe with permanganate; the heating is best conducted in absence of air. Lunge“ gives the accuracy of the method as 0.2 t o 0.4 per cent. Phelps12 obtained results which ranged from 99.2 to 100.10 per cent, nearly all being low. The method is capable of results much better than these, as many analysts know by experience. The procedure used in the trials here reported involves use of a flask with a Bunsen valve and initially filled with carbon dioxide, and improvement of the end point by use of the Zimmermann-Reinhardt titrating-s01ution.l~ Exclusion of air from the ferrous solution while hot seems to be a wise pre~auti0n.l~ The susceptibility of ferrous sulfate solutions to air oxidation has been studied by several chemist^,'^ who have variously reported that air oxidation of a decinormal solution proceeds at the rate of about 0.06 per cent per hour, but is much retarded by presence of acid, that fifth normal solutions can be boiled in air and stirred while cooling without perceptible change, and that such solutions are relatively stable if allowed to stand until dissolved oxygen is consumed. The 0.1 N ferrous sulfate solution used by the writer was about 2.6 S in sulfuric acid, and had stood for over 2 weeks. Without undue exposure to air during use its normality fell from 0.10337 to 0.10282 in 2 days, and on the following d a y was 0.10258. The concentration of the solution was therefore determined coincidentally with its use, by means of blanks conducted as part of e a c h s e r i e s . The 0.1 N permanganate was about 6 months old, and had been filtered through a s b e s t o s a n d then F i g u r e 2-Receiver f o r B u n s e n A p p a r a t u s through a M a n d l e r filter. It was standardized against Kahlbaum’s sodium oxalate, using the procedure of hlcBride.16 Mean normalit,ies during a 4-day interval were 0.10036, 0,10035, 0.10034, and 0.10032. +
+
* 2. anal. Chem., 46, 521 (1907). Carnot, Compt. rend., 122, 449 (1896); Phelps, Am. J. Sci., [41 17, 201 (1904); Lunge, “Technical Methods of Chemical Analysis,” Vol. I, 1908, p. 497; Classen, “Ausgewahlte Methoden der analytischen Chemie,” Vol. 11, 1903, p. 368; Treadwell-Hall, Vol. 11, 6th ed., 1924, p. 537. 11 “Chem. techn. Untersuchungsmethoden,” Vol. I, 5th. ed., p. 506. 1 2 Loc. c i t . ; see also Gooch, “Methods in Chemical Analysis,” 1912, p . 462. 18 Chem. Ztg., 13, 323 (1889); cf. Barneby, J . A m . C‘hem. Soc., 36, 1429 (1914). 14 Cf. U. S. Army Ordnance Specification 50-11-11,1913, p. 2. 18 Phelps, loc. cit.; Peters and Moody, Am. J . Sci., 141 12, 369 (1899); Banerji, C. A., 18, 1255 (1924). 1 8 J . Am. Ckem. Soc., 34,393 (1912);Bur. Slandards, Circ. 40 (1913). 10
1185
PROCEDURE^. 1 N Ferrous Sulfate Solution. Twenty-eight grams of FeSOa.7H20were dissolved in 200 ml. of warm water containing 50 ml. of 1: 1sulfuric acid. The liquid was filtered, treated with 150 ml. of 1: 1 sulfuric acid, and diluted to 1000 ml. Titrating Solution. Sixty-seven grams of MnS04.4H10 were dissolved in 500 ml. of water, 138 ml. of 85 per cent phosphoric acid and 130 ml. of concentrated sulfuric acid added, and the liquid diluted t o a liter. The sample of chlorate (about 0.08 gram in a 10-ml. aliquot) was pipetted into a heavy Pyrex 500-ml. Erlenmeyer flask. There were added 100 ml. of water (80 ml. in t h e blanks), and air was replaced by passing a stream of carbon dioxide for 5 minutes. Fifty milliliters of 0.1 N ferrous sulfate solution were introduced with a pipet, followed by 40 ml. of water (25 ml. in the blanks), added in such a way as t o wash down the touched-off drops from the pipetting. The stream of carbon dioxide was continued during these operations. A rubber stopper with a Bunsen valve was adjusted, and the liquid was boiled for 5 minutes. The flask was placed in cold water. The cooled liquid was treated with 10 ml. of the titrating solution, and excess F e + + was titrated with 0.1 N permanganate, added from a volume buret. Duplicate blanks were run with each series.
ANALYTICALREsuLTs-The Figure a - ~ i t z A p p a r a t u s final series yielded the results 99.94, 99.97, 99.94, 100.01, 99.94 per cent, average 99.96 per cent. Another series, conducted with less care, gave the results 100.04, 99.95, 99.99, 100.06, 99.99, 100.06 per cent, average 100.02 per cent. When air in the flask was not replaced by carbon dioxide the method was less accurate, five results ranging from 99.93 per cent to 100.15 per cent. The results of the final trials represent the method a t its best. The ferrous sulfate excess method is without question superior to any other method described in this paper for accurate routine analysis. It requires no special apparatus, is rapid, is well adapted to execution of analyses in series, and appears to be as accurate as any volumetric method tried. The time required is about 30 minutes, exclusive of the preparation of the chlorate solution. Reduction to Chloride by I g n i t i o n with A m m o n i u m Chloride“
The chlorate is mixed with excess of ammonium chloride in a tared crucible, and the mixture heated until all ammoium chloride is volatilized; a spongy mass of potassium chloride remains. Allowance must be made for chloride and bromate present as impurities, and for any nonvolatile matter the ammonium chloride may contain. Preliminary trials showed ordinary C. P. ammonium chloride to be unsuitable; even after resublimation it contained weighable nonvolatile matter (0.0008 gram in 4 grams of salt). For the trials reported here C. P. ammonium chloride was dissolved in water, precipitated by alcohol and hydrogen chloride, redissolved, boiled with ammonia, filtered, and hydrogen chloride passed into the cooled solution short of saturation. The precipitated product was free from nonvolatile matter: 10 grams left no weighable residue on volatilization. Two trials of this method by Blangeyl8 yielded the mean result 100.06 per cent. 1’
18
Treadwell-Hall, Vol. 11,6 t h ed., 1924,p. 399. I b r d , Vol. 11, 3rd. ed., 1911, p. 462.
1186
I N D U S T R I A L B S D E,VGI,VEERING C H E X I S T R Y
PROCEDURE-The porcelain crucibles used were of 25 ml. capacity (No. I), each covered with a small watch glass. Larger crucibles were found t o be unsuitable. Each crucible was ignited gently (faint redness), with the cover, until the weight was constant within 0.0001 gram. The sample of about 1 gram of chlorate was transferred to the crucible, 4 grams of purified ammonium chloride were added, and the charge was mixed with a platinum wire. The cover glass was put in place and the crucible heated cautiously with a small moving flame until reaction began; the onset of the decomposition was always rather sudden. When visible reaction was ended, t h e heat was raised so as t o cause rapid volatilization of ammonium chloride. When evolution of white fumes ceased, crucible and cover were cooled and weighed, and the heating was repeated. I n every case t h e second heating brought the weight t o constancy It is apparently permissible in this analysis t o use a temperature somewhat higher than is generally considered safe with potassium chloride obtained by evaporation.
ANALYTICAL REsuLTs-Three trials resulted as follows: calculated from weight of KC1: 99.97, 99.99, 99.98 per cent, average 99.98 per cent; calculated from oxygen lost: 100.02, 99.98, 100.01 per cent, average 100.00 per cent. Some less painstaking trials in 1918, using chlorate very nearly 99.90 per cent pure, gave the results 99.81, 99.88, 99.90 per cent, average 99.87 per cent. This method yields accurate results with certainty, but requires considerable attention, and on the whole is somewhat troublesome. Occasional cracking of a watch glass is an annoyance which could be avoided by use of a silica cover. The ordinary crucible lid must not be used. R e d u c t i o n to Chloride by Evaporation w i t h Hydrochloric Acid"
The chlorate is evaporated with dilute hydrochloric acid in a weighed crucible. The weight of potassium chloride left, or that of the oxygen lost, serves as basis for the calculation. Corrections must be applied for chloride and bromate present as impurities. Trials of the method by BlangeyI8 gave 100.02 per cent as the average of four analyses. PRocEDuRE-Analyses were made in 50-ml. (No 2) porcelain crucibles, ignited gently, with their lids, to constant weight (0.0001 gram). The sample of 1 gram of chlorate was weighed directly into the crucible, and 5 ml. of 1:3hydrochloric acid were added. The crucible was covered with a watch glass, and was warmed gently on a water bath until action ceased. A second 5 ml. of acid was added and heating continued to colorlessness. The watch glass was removed, washed off, and the liquid evaporated t o complete dryness on the water bath. The crucible was covered with its lid, and was kept for an hour a t 105' C. in an electric oven. The covered crucible was then heated with a small flame, first until decrepitation ceased, and then more strongly (not above faint redness), until the weight was constant.
ANALYTICAL REsuLTs-calculated from weight of potassium chloride: 99.99, 100.01, 99.94, 100.00 per cent. average 99.99 per cent; calculated from oxygen lost: 99.98, 99.94, 100.06, 99.96 per cent, average 99.99 per cent. The method is accurate, but rather long. The manipulations are easy, however, and only a moderate amount of actual attention is required. R e d u c t i o n t o Chloride by S u l f u r o u s Acid, a n d P r e c i p i t a t i o n as Silver Chloride
Reduction of chlorate by sulfur dioxide appears to have been first applied quantitatively by Blattner and Brasseur,lg who used it for determination of chlorate in presence of perchlorate. The superiority of sulfurous acid over other reducing agents has apparently not been generally recognized. Only a single manual of analysis,20 among those consulted, specifies its use. The method consists in treatment of the chlorate with aqueous solution of sulfur dioxide, expulsion of excess of 1)
'0
Chcm. Ztg., 24, 793 (1900). Scott, !'Standard Methods of Chemical Analysis
"
Vol. 17, Yo. 11
sulfur dioxide by boiling, and determination of the chloride as silver chloride. Corrections must be applied for chloride and bromate present as impurities; PROCEDURE-A sample of 0.5 t o 0.75 gram of chlorate was transferred to a 250-ml. Erlenmeyer flask, dissolved in 50 ml. of water, and treated with 50 ml. of a fresh saturated solution of sulfur dioxide. The mouth of the flask was covered with a small watch glass, and the liquid was boiled until excess of sulfur dioxide was expelled. The solution was transferred to a 400 ml. beaker, diluted to 150 ml., and 1 ml. of nitric acid added. To the cooled solution a slight excess of 2.5 per cent silver nitrate solution was added slowly from a pipet, with stirring (about 50 ml. for 0.75 gram of chlorate). The liquid was boiled several minutes, and was allowed t o stand several hours in the dark. The liquid was then decanted through a prepared Gooch crucible, the precipitate washed by decantation with water containing a little nitric acid, transferred to the crucible with the aid of a jet of the acidulated water, washed several times more, and finally once with pure water. The crucible was dried for a n hour a t 105' C., and was then put into a No. 2 crucible and heated gradually with a small flame until the silver chloride began to fuse. The first or second heating sufficed always to render the weight constant within 0.0001 gram.
ANALYTICAL REsuLTs-The following values were obtained : 99.96, 99.99, 99.96, 99.97, 99.98, 99.98, 99.96 per cent, average 99.97 per cent. One analysis, evidently because of some error in manipulation, gave 99.87 per cent. This method yields almost faultless results, and its only drawback is the time required. VOLLXETRIC DETERMISATION OF CHLORIDE AFTER RED U C T I O ~BY S02-Trials of these methods were made during the preliminary work in 1918. Mohr's titration appeared to be insufficieiitly accurate: trials with a chlorate of 99.9 per cent purity gave the result's 99.72 per cent and 99.79 per cent. The silver nit'rate was standardized against pure sodium chloride by the hIohr titration (precipitation as silver chloride gave distinctly lower normalities). Polhard's titration, applied to the same specimen of chlorate, gave the results 99.87, 99.99, 99.84, 99.96, 99.81, 99.89 per cent, average 99.90 per cent. I n these trials the entire filtrate mas titrated with thiocyanate to a pink color which persisted after shaking the stoppered flask. The relationship of the silver nitrate and thiocyanate solutions was found under identical conditions. The silver nitrate was standardized gravimetrically. Conclusions
Of the seven methods examined, only that of Kolb and Davidson was found to be inaccurate; its results were consistent but about 0.4 per cent too high. The other methods are suitable for exacting analytical work. For routine use the volumetric methods have the usual advantage. The ferrous sulfate excess method combines accuracy, rapidity, and simplicity, and is especially recommended. The mechod of Bunsen and that of Ditz require only one standard solution, and therefore fewer measurements, but these methods are less rapid and not so well adapted to routine work. Both require special apparatus, which is perhaps less of a drawback in the Bunsen method, because of the wider applicability of the evolution apparatus. The gravimetric analysis of chlorate is most easily effected by evaporation with hydrochloric acid, and most accurately by reduction with sulfur dioxide and precipitation as silver chloride. Ignition with ammonium chloride is accurate but less satisfactory.
Determination of Arsenic in Steel-Correction I n the article by Alan E . Cameron under this title, THIS JOURNAL, 17, 965 (1925), the figures for the comparative determinations a t the bottom of the first column on page 966 refer,to. per cents, not grams.
