0 . 2 scale division. As each division is equivalent to approximately 0.005 'C.
(see below), this corresponds to a standard deviation of 1 X 10-3' C., which is the same as the value reported by McMullan and Corbett (IO). If only the first six solutions of Table I are considered, the data up to 0.06 molal can be represented by the linear equation m = 1.046 X
Ad
(3 )
with an estimated standard deviation of 0.4 scale division or 2 X C. It is from the slope of Equation 3 (molality per scale division) and the cryoscopic constant reported by Kraus and coworkers (1, 9) for benzene (5.07' C. per molality) that the estimate of 0.005" C. per scale division was obtained for this work. In practice the cryoscopic constant for each new bottle of benzene was calculated from the freezing point depression (expressed in scale divisions) measured for an approximately 0.05 molal solution of triphenylmethane in benzene. A 0.05 molal solution of triphenylmethane in benzene has a freezing point about 50 scale divisions below that of the pure solvent (Equation 3). At this concentration the estimated standard deviation, or error, of 0.4 division corresponds to a relative error of 0.8%. Assuming a negligibleerror in the weights of solvent and solute used for a single determination, the estimated relative error for the determination of the cryoscopic constant from the ratio of molality to freezing point depression should also be 0.8%. This was confirmed by sets of replicate determina-
tions of the cryoscopic constant which were made over a period of 10 months on various lots of benzene. The value of 1.4% so obtained, and based upon 23 degrees of freedom, is not significantly different from the estimate of 0.8% based upon 5 degrees of freedom. The data presented here are for an ideal solute-solvent system and were obtained solely to illustrate the performance characteristics of the equipment. When cryoscopic measurements are to be made on nonideal systems, appropriate extrapolations of the data must be employed to correct for the variation with composition of the apparent molecular weight of the solute. ADVANTAGES OF AUTOMATIC CRYOSCOPY
The following advantages characterize the apparatus described in this report. Once the solute and solvent have been weighed into the cryoscopic cell and the cell has been assembled in the freezing bath, no further operator attention is required. A permanent record of the cooling curve is automatically obtained. The extrapolations required for the determination of the freezing points of solutions are more easily made from a continuous curve than from one plotted manually from discrete temperature readings. ACKNOWLEDGMENT
The author gratefully acknowledges the assistance of Harley W. Middleton, Marie DeVito, and Joyce H. Northrop in the construction and testing of the equipment.
LITERATURE CITED
Batson, F. BI., Kraus, C. A., J . Am. Chem. SOC.56, 2017 (1934).
Beck, 4.,J . Scz. Instr. 33, 16 (1956). Daniels, F., Mathew, J. H., Wdliams, J. W., Bender, R., Alberty, R. S., "Experimental Physical Chemistry," 5th ed., p. 68, McGraw-Hill, New York, 1956. Giguhre, P. A., Secco, E. A., Can. J . Chem. 32, 550 (1954).
Glasgow, A. R., Jr., Ross, G., J .
Research Natl. Bur. Standards 57, 137 (1956). Glasgow, A. R., Jr., Streiff, A. J., Rossini, F. D., Ibid., 35,355 (1945). Herington, E. F. G., Handley, R., J . Chem. SOC.1950, 199. Herington, E. F. G., Handley, R., J . Sci. Znslr. 25, 434 (1948). Kraus, C. A., Vingee, R. A., J. Am. Chem. SOC.56, 511 (1934). Mclfullan, R. K.> Corbett, J. D., J . Chem. Educ. 33, 313 (1956). hfikhkelson, V. Ya., J . Anal. Chem. C.S.S.R., 9, 21 (1954). Pvluller, R. H., Stolten, H. J., ANAL. CHEM.25, 1103 (1953). Sewman, M. S., Kuivila, H. G., Garrett, A. B., J . Am. Chem. SOC. 67, 704 (1945). Richards, L. A., Campbell, R. B., Soil Sci. 65. 429 (1948). Spooner, D. d., J . &ci. Instr. 29, 96 (19.52) \ - - - -
Stull, D: R., IND. ENG.CHEY.,ANAL.
ED. 18, 234 (1946). Stull, D. R., Rev. Sci. Instr. 16, 318 (1945): Kitschonke. C. R.. ANAL. CHEJI. 24, 350 (19E i2). Zeffert, B. M. Hormats, S., Ibid., 21, 1420 (19;49,. Zeffert, B. M.,, Withers oon, R. R., Ibid.', 28, 1701 (19567. Zemany, P. D., Ibid., 24, 348 (1952).
RECEIVEDfor review May 9, 1957. Accepted November 23, 1957. Division of Analytical Chemistry, 131st meeting, ACS, Miami, Fla., ApriI 1957.
Drying and Decomposition of Sodium Carbonate ARTHUR E. NEWKIRK and IFlGENlA ALlFERlS General Electric Research laboratory, Schenectady, N. Y.
b Thermobalance curves are given for the drying and decomposition of sodium carbonate in the temperature range from 25" to 1040" C. using different crucible materials and atmospheres. The reaction of sodium carbonate and silica resulted in a weight loss a t t,emperatures as low as 500" C. It is recommended that sodium carbonate for analytical use b e dried in platinum to prevent errors due to its reaction with silica and silicates.
I
more than a coincidence that the accepted maximum temperature for drying sodium carbonate for use as a T IS
982
*
ANALYTICAL CHEMISTRY
primary standard, and the lower limit for the initiation of the reaction of sodium carbonate and silica are the same -Le., 300" C. This fact is not generally recognized. Directions for drying sodium carbonate usually specify porcelain or platinum containers, but glass weighing bottles are also used. In this Iaboratory, satisfactory results have been obtained by drying in glass a t 250' C. (1). The recommended temperature range from 250" to 300' C. is the result of many studies (6), but the upper temperature limit is surprisingly low in view of the several reliable reports of the thermal stability of
sodium carbonate at temperatures close to its melting point, 851'C. Richards and Hoover (8) found that sodium carbonate heated under carbon dioxide for a long time just below the fusion point lost on fusion only about 0.003% in weight. Duval (2) reported that sodium carbonate, when heated in a thermobalance, is stable in the range from 100" to 840' C. Motzfeldt (7), who made a careful study of this decomposition in a platinum cell, concluded that there is no chance for significant decomposition at any temperatures below 800" C. a t least. He attributes previous discordant results to
reaction of the sodium carbonate with the moisture of the air (or the mater content of the initial sodium bicarbonate) to form sodium hydroxide and carbon dioxide. A recent study by Easterbrook (3) with the aid of a thermobalance showed that sodium carbonate prepared by heating sodium sesquicarbonate reached constant weight at 210" C. in 2 hours; there was no further change in weight after 19 hours a t 270" C. or an additional hour at 340" C. Easterbrook concluded that the slightly high titer of the product (maximum calculated purity, 100.023%) resulted from the formation of oxide or hydroxide produced by reaction with water during the decomposition of the sesquicarbonate. The authors studied the reaction between sodium carbonate and other substances a t elevated temperatures, and heated sodium carbonate and sodium carbonate plus silica in a Chevenard thermobalance using a number of different crucible materials and atmos' pheres. The results and additional data from the literature suggest that another factor causing the discordant results may be reaction of sodium carbonate with glass or porcelain used as a container.
due to water in the sodium carbonate as pointed out by Duval (2). I n runs 8 through 11 the sodium carbonate was dried by heating in weighing bottles of borosilicate glass for 16 hours a t 350" C., and hence showed no initial water loss. The large losses in runs 1 through 4, starting in the neighborhood of 800" C., are chiefly attributed to the reaction
Table 1. Weight Losses in the Reaction of Sodium Carbonate with Porcelain and Silica Calcd. Obsd.
Run
Loss,
LOSS,
Gram
Gram
0.2076 0.2076 0.2076 0.2076
0.1927 .
YO.
1
2 9 10
0.2028 0.2083 0.2087
APPARATUS AND MATERIALS
This work was done with a penrecording Chevenard thermobalance, whose construction and performance have been described (9). Experiments in air were performed with the balance as received, but experiments in controlled atmospheres were performed using a quartz liner tube for the furnace fitted with a loose cap at the bottom and a tube for the introduction of gas at the top. The porcelain crucibles were Coors regular laboratory grade. The alumina crucible was Triangle RR grade supplied by Morgan, England. The gold crucible was spun from pure gold sheet, and the platinum crucibles were of regular laboratory grade. The sodium carbonate (General Chemical Co., anhydrous reagent grade) was stored over Drierite (W. A. Hammond Drierite Go., Yellow Springs, Ohio) except during weighing of the samples. Nitrogen from the laboratory supply was dried by passage through a trap packed with glass wool and cooled in liquid nitrogen. It was humidified by passage through mater in a sinteredglass bubbler a t 25" C. The carbon dioxide (Matheson Co., ''bone dry" grade) was used as received. EXPERIMENTAL RESULTS
The experimental thermobalance curves are reproduced in Figure 1. The explanation for the shape of these curves seems straightforward. The small loss below 100" C. is probably
I
I
of sodium carbonate with the porcelain or alumina crucible. When the heating is carried out for a long period of time, the weight loss is approximately quantitative (Table I). R h e n the sodium carbonate is heated in platinum or gold, the weight loss is much less rapid and is probably due to the decomposition of sodium carbonate to form sodium oxide and carbon dioxide as proposed by Motzfeldt (7) (runs 5, 6, 7 , and 11). As would be expected, the presence of a positive stream of gas to flush out the carbon dioxide resulted in a faster rate of weight loss (run 6 us. run 5). However, when water was present in the gas stream, the observed rate of weight loss was less (run 7 us. run 6). A sample of the sodium carbonate dried a t 350" C. showed no significant change in weight on further heating 12 hours a t 600" C. and 4 hours at 650" C. in a platinum crucible in air. The decomposition of sodium carbonate a t temperatures up t o 1040" C. is completely suppressed by an atmosphere of carbon dioxide (run 8). The reaction of sodium carbonate with coarse silica sand occurs rapidly a t 800" to 850" C. (run 9); the reaction temperature is lowered by grinding the silica (run 10). In this latter run, the first evidence of reaction as shown by weight loss is at about 500" C. The weight loss in these rum was quantitatively equivalent per formula weight to the carbonate present (Table I), there being an excess of silica or silicate in each case. DISCUSSION
Figure 1. Thermobalance runs with sodium carbonate" Curve
Crucible
1 2 3 4 5 6 7 8 9 10 11
Porcelain Porcelain Porcelain Alumina Platinum Platinum Platinum Platinum Platinum Platinum Gold
Atmosphere
Sample
Air NazCOa Air NazCOab Dry Nz' NazCOs Air NazCOa Air NazCOs Dry NzC NazCOs W e t NzC NazCOs COze NazCOa Dry NzC NazCOo Dry NzC NazCOa Dry Nze NazCOa'
Si02 ++ SiOzd
Heating rate 300" C. per hour, except runs 9 and 10. Crucible covered. Gas flow rate 2 5 0 ml. per minute. *Heating rate 300" C. per hour to 520" C., then 50' C. per hour. eMaximum temperature 922" C., but sample cooled and held 1 hour a t 91 5" C. after reaching 922" C.
The above results show that sodium carbonate-silica mixtures will lose weight a t temperatures as low as 500" C. Howarth, Maskill, and Turner (4, 6) have shown that the reaction between these compounds can occur even a t 300" C. Unfortunately, reports of decomposition of sodium carbonate below 800" C. do not specify the material of the vessel: Waldbauer, McCann, and Tuleen (11) report decomposition a t 482" C., and Smith and Croad (10) report decomposition above 300" C. but neither mentions the nature of the container. It is likely that they, and others, were observing a chemical reaction with the container rather than the thermal decomposition of sodium carbonate. The authors suggest that sodium carbonate for analytical use be dried by heating in dry air or carbon dioxide using platinum or other inert material as a container. Under these conditions, any temperature in the range of 250" to a t least 700" C. should be satisfactory. For work of the highest accuracy i t may be necessary to just fuse the VOL. 30, NO. 5, M A Y 1 9 5 0
983
material in an atmosphere of carbon dioxide as proposed by Richards and Hoover (8). LITERATURE CITED
.,
(1) Balis. E. \I7.. Bronk. L. B.. Lieb-
hafsky, H.‘ A., PfeBer, H. -G., ANAL.CHEM.27, 1173 (1955). (2) Duval, C., Anal. Chim. Acta 13, 32 (i955j.
(3) Easterbrook, W. C., Analyst 82, 383
(1957). (4) Howarth, J. T., Maskill, W., Turner, W. E. S., J . SOC.Glass Technol. 17,25T (1933). Howarth, J. T., Turner, W. E. S., Ibid., 14,402T (1930). Kolthoff, I. XI., Stenger, V. A., “Volumetric Analysis,” Vol. 11, p. 80, Interscience, New York, 1947. (7) hlotzfeldt, K., J. Phys. Chem. 59, 139 (1955).
Richards, T. W., Hoover, C. R., J . Am. Chem. SOC.37, 95 (1915). Simons, E. W., Kewkirk, A. E., Aliferis, Ifigenia, ANAL. CHEM. 29,48 (1957). Smith, G. F., Croad, G. F., IND. ENG.CHEM..ANAL. ED. 0. (1937~,_ ,_141 __ ~~
~
Waldbauer, L., hlcCann, D. C., Tuleen, L. F., Ibid., 6 , 336 (1934).
RECEIVED for review August 9, 1957. Accepted December 28, 1957.
Titrimetric Determination of Fluorine Particularly in Aluminum Fluoride LAWRENCE V. HAFF General Chemical Division, Allied Chemical & Dye Corp., Morristown, N. 1. C. P. BUTLER and J.
D. BISSO
General Chemical Division, Allied Chemical & Dye Corp., Porf Chicago, Calif. ,Discordant results for total fluorine were reported from different laboratories analyzing identical samples of aluminum fluoride. Serious losses of fluorine occur in distilling and titrating large amounts of fluorine by conventional methods. Recoveries are especially low and erratic when borosilicate glassware i s employed. Details of a rapid and accurate procedure for pyrohydrolytic assay of aluminum fluoride were worked out. Routine analyses of aluminum fluoride are greatly expedited by the procedure, which, however, requires a standard sample of aluminum fluoride. In assaying the standard sample, appropriate correction must be made for the losses incurred in the distillation and subsequent concentration of the fluorine. The number of replicate samples required to establish the correction and the magnitude of the correction can be greatly reduced by avoiding use of borosilicate apparatus.
A
of conventional volumetric methods for determining fluorine in aluminum fluorine indicates that recoveries of large amounts of fluorine from the distillation procedure are only about 98% complete and further losses are incurred when glassware containing boron is used. Under standardized conditions recoveries are reproducible. Analysis by a pyrohydrolytic method of samples standardized by conventional methods established that pyrohydrolysis offers a rapid, accurate, and convenient procedure for assaying aluSTUDY
984
*
ANALYTICAL CHEMISTRY
in reported assays. It quickly became apparent that recovery of fluorine was never complete by either method. While the relative percentage of fluorine LOSSES I N CONVENTIONAL PROCEDURES lost would not be important in microFor a number of years two procedures analysis, it assumed serious proportions for determining relatively large amounts in assay of a material containing some of fluorides have been used throughout 50% of fluorine. Because some of this company. As applied to aluminum these errors will affect almost all fluoride, both involve fusion of the macrodeterminations of fluorine, some sample with sodium carbonate, followed details of this work are presented. by acidification and steam distillation Standard Samples. No standard fluoride sample was available to the as described by Willard and Winter authors a t the time this investigation (22). The analysis may be concluded as was initiated. Cryolite has been sugby Armstrong’s procedure (1, i?), modified by Ron-ley and Churchill gested as a standard ( T ) , but its purity (20). Practically, this titration is is problematical. Sodium fluoride, used limited to determination of small by Hoskins and Ferris (8), Kimball amounts of fluorine, and the small aliand Tufts (9), Reynolds and Hill quot which must be titrated involves (l7), and hIatuszak and Brown ( l a ) , is of uncertain assay. Kimball and serious magnification of small errors. Alternatively the fluorine in the disTufts (9) reported fluorspar unsatistillate may be converted quantitatively factory as a standard and suggested to fluosilicate, precipitated as potassium analyzed lead chlorofluoride. The presfluosilicate, filtered off, and titrated. ence of large amounts of either calcium or lead in samples intended to simulate This procedure is essentially a macrosodium carbonate fusions of aluminum method and all of the fluoride distilled from a 0.2-gram sample may be titrated fluoride was considered undesirable. conveniently. Unlike the preceding I n this work, the exact weight of titration, this procedure is insensitive to fluorine involved in each analjs’is was the interferences of small amounts of determined by titrating about 0.2 gram phosphates and sulfates and the end of 48% hydrofluoric acid to the phenolpoint is familiar to most analysts. Acphthalein end point with standard 0.1N cordingly it was designated the standard sodium hydroxide solution. The titrations were performed in platinum dishes procedure for the analysis of aluminum and concluded a t the boiling point to fluoride. As rather lengthy evaporaeliminate interference by fluosilicic acid tion and filtration procedures are inin the hydrofluoric acid or dissolved volved, manufacturing locations were silica in the standard base. allowed the option of using the thorium The titrated solution, after cooling, is nitrate titration on all but critical transferred to the still, acidified, and samples. distilled. The distillate is then analyzed An investigation of the procedures was by one or both of the above methods to initiated when discrepancies appeared determine the weight of fluorine reminum fluoride. Recoveries were reproducible and about 99% complete.
7
