It has been shown that the discrepancy between tartar emetic and arsenious oxide solutions lies in the ready tendency of the tartar emetic to lose its water of crystallization. If in fine crystalline or powdery form, tartar emetic loses its water of crystallization slowly in the air, more rapidly over sulphuric acid in an ordinary or vacuum desiccator, and somewhat upon standing in closed bottles; at a temperature of 128°-130°, it begins to form an anhydride losing one-half molecule of water; at 160°165° it loses another molecule of water and becomes doubly anhydride. The formation of these anhydrides seems best explained structurally by attaching the antimony to the hydroxyl rather than to the carboxyl end of tartar emetic. If tartar emetic be recrystallized in medium-sized crystals, filtered, washed two or three times with distilled water, drained under suction for from five to ten minutes, and exposed at a temperature not exceeding 25° for a period of from three to four hours to the action of a dry atmosphere, decinormal solutions of the salt thus freshly prepared will correspond exactly to decinormal solutions of arsenious oxide.

THE KENT CHEMICAL LABORATORY

OF YALE UNIVERSITY.

PHOTOMETRIC DETERMINATION OF IRON.

BY J. I. D. HINDS AND MYRTIS LOUISE CULLUM.

Received May 24, 1902.

THE present investigation was undertaken for the purpose of ascertaining whether the photometric method was applicable to colored precipitates.

Solutions of iron of various degrees of dilution were prepared and their strength accurately determined gravimetrically. The iron was precipitated with potassium ferrocyanide. Ferric nitrate, sulphate and chloride were found to give practically the same results. In all the solutions nitric acid was present, though never more than to the extent of I per cent. This was in order to insure that the iron was in the ferric condition.

As in the previous investigations,1 a common candle was used with the simple photometric cylinder. The more accurate instrument suggested by Professor D. D. Jackson2 was not at our command. The work was carried on in the diffused light of the laboratory, but the cylinder was protected by the hand and body against any glare of light from the windows.

Preliminary tests with the solid ferrocyanide showed that constant readings could not be obtained. The reading was found to be a function of the amount of the ferrocyanide added. This is because the ferric ferrocyanide precipitate is soluble in an excess both of the iron and ferrocyanide solutions. We therefore made a 5 per cent. solution of the potassium ferrocyanide and used it from a burette. On adding this drop by drop to a ferric solution, there first appears a blue color, then a coarse precipitate which presently breaks up into a fine state of division and finally dissolves. The greatest opacity and therefore the lowest photometric reading occurs just when the precipitate begins to break up. This point is reached when the amount of ferrocyanide bears to that of the iron present about the proportion of its molecular weight to the atomic weight of iron; or when the ferrocyanide is a little in excess of the amount necessary to convert the iron into Prussian blue. For example, a solution which contained 0.0054 gram of iron required 0.7 cc. of the 5 per cent. ferrocyanide solution, or 0.035 gram

1 This Journal, 18, 661.

2 Ibid., 23, 799.

potassium ferrocyanide. These numbers are in the proportion of 54 to 350 or 56 (atomic weight of iron) to 366 (molecular weight of potassium ferrocyanide 368).

It was found that at the point of maximum opacity, or where the reading had a minimum value, constant conditions and constant readings could be obtained. The following method was then adopted: The solution whose strength was to be determined was placed in a lipped beaker in quantity sufficient to fill the photometer to the required height, and the ferrocyanide added, drop by drop, until the precipitate just appeared. The solution was thoroughly mixed by pouring back and forth, and the reading taken. Another drop of the ferrocyanide was added and another reading taken (which should be lower than the first one). This process was continued until a minimum was obtained and the readings began to rise again. The lowest reading was the one adopted.

With a little practice only four or five readings are necessary to find the minimum, which is the easier to detect because the opacity near the minimum remains nearly the same for two or three readings.

An example and diagram will illustrate the peculiar action under consideration.

It appears that the ascending arm of the curve is nearly a straight line and rises at an angle a little higher than that of the descending arm.

A series of determinations was made and each was tested by numerous readings. In the first column of the table which follows is given the number of the solution; in the second its real strength

in per cent. of iron or grams in 100 cc.; in the third is the photometric reading; in the fourth is the product of the reading and strength; in the fifth is the strength computed from the first equation determined below; in the sixth the difference between the computed and real strength; in the seventh the strength as computed from the second equation; and in the eighth the difference between this and the real strength.

ΙΟ

8.7

0.00359

0.00007

0.00355

0.00003

0.00352 II 0.00313

0.03062 9.7 0.03036 0.00321 0.00008 0.00317 0.00004

Assuming the equation of the hyperbola ry + by a in which r represents the photometric reading and y the strength of the solution, and forming the observation and normal equations and solving for the constants we obtain the equation

Twenty-two sets of observations, including readings from 1.7 cm. to 10.7 cm., were used in the computation of this equation and its efficiency may be judged by the numbers in the first column of differences in the table above.

The readings below 3 cm. and above 9 cm. were found to be rather uncertain. Rejecting these and some others which seemed to be a little inconsistent with the general series and using only the eleven found in the table, we obtain the equation

The values computed from this equation are found in the seventh column, and in the eighth column are the differences between these values and the real strengths of the solutions. These differences run a little more regularly than those in the sixth

column, but practically one is about as good as the other. They all show that the values computed from either equation are true to the fourth decimal-place of per cent., or to parts in 1,000,000. Using the numbers from 2 to 12 for the probable error of one determination, we find, n being the number of equations, q the number of variables, and v the differences,

This method is applicable to all mixed solutions of iron containing no metal which is precipitated with potassium ferrocyanide, such as cobalt, nickel, manganese, etc. In the analysis of phosphates and fertilizers the iron may be determined in the solution made for the determination of phosphorus. It may be necessary to dilute the solution if the iron is present in considerable quantity. The following is an example:

Amount taken: 0.3988 gram phosphate rock, which was dissolved in nitric and hydrochloric acids and made up to 200 cc.

Photometric reading was 6.5 cm., corresponding to 0.00489 per cent. iron. 200 cc. therefore contained 0.00972 gram iron. Of the 0.3988 gram rock, this is 2.45 per cent.

TABLE FOR PHOTOMETRIC DETERMINATION OF IRON.