of the components is solid, and to the instances where there is an increase instead of a decrease of solubility.

The results of this paper may be summarized briefly as follows.

1. The equilibria between two partially miscible liquids, and a consolute liquid follow the Mass Law.

2. There are four sets of equilibria corresponding to four different series of solutions.

3. If the two liquids are practically non-miscible, there are only two sets of equilibria.

4. The reacting weights of the liquids studied were not functions of the concentration, possibly with one exception.

5. There is a fundamental difference between the solute and the solvent.

6. The solubility curve of a substance in a varying mixture of two liquids at constant temperature has a break.

A GLANCE at published results shows that the atomic weight of
strontium has not been investigated for thirty-five years. The early
determinations, good enough for their time, show variations which
render them quite unsatisfactory to-day; and the case is parallel in
every respect to that of barium, which has formed the subject of two
recent papers."

*

The oldest experiments of any note upon the atomic weight of stron-
tium are those of Stromeyer,† who measured, in 1816, the gas evolved
from strontic carbonate upon its decomposition by an acid. The
result, which is only of interest historically, gives Sr 87.3, if a litre
of carbon dioxide weighs 1.977 grams under normal conditions.

=

At about the same time Rose ‡ found that 181.25 parts of argentic
chloride could be obtained from a hundred parts of strontic chloride,—
data which indicated Sr = 87.31. Twenty-seven years afterward, in

*These Proceedings, XXVIII. 1; XXIX. 55.

† Schweig. J., XIX. 228; Meyer u. K. Seubert's Atomgewichte, p. 123.
Poggendorff's Annalen, VIII. 189.

1843, Salvetat determined by loss of weight the carbon dioxide in strontic carbonate, and concluded that the metal must be 88.0, — a result which scarcely improved the situation.

Subsequently, in 1845, Pelouze † found the amount of silver necessary to precipitate a weighed amount of ignited strontic chloride; his results give the value Sr = 87.70. Thirteen years later Marignac ‡ repeated these experiments, determining also the amount of crystal water in crystallized strontic chloride, as well as the amount of strontic sulphate obtainable from the salt. Thus he found that 15.000 grams of crystallized strontic chloride yielded 8.9164 § grams of the anhydrous salt and 10.3282 grams of strontic sulphate; moreover, 15.000 grams of hydrated strontic chloride required 12.1515 grams of silver for precipitation. Another similar series of experiments upon the water of crystallization made its amount appear three milligrams more than before. These data give basis for a number of possible values for the atomic weight of strontium, ranging from 87.17 to 87.55, the individual figures being tabulated below.

In 1859 Dumas || published another determination of the ratio of strontic chloride to silver, the salt having been fused in a stream of hydrochloric acid. Altogether, 27.3435 grams of strontic chloride. required in his hands 37.252 grams of silver, the individual values for strontium varying from 87.3 to 87.8. Since this time the subject has remained untouched.

Below is tabulated a list of the various determinations, grouped according to the ratios determined.

§ Corrected by L. Meyer u. K. Seubert, Atomgewichte, pp. 78, 79.
Liebig's Annalen, CXIII. 34.

A critical review of the list reveals a great lack of trustworthiness in all the figures. The values deduced from the carbonate, and those involving water of crystallization, may all be thrown out at once; and the results yielded by the displacement of hydrochloric by sulphuric acid are but little better. The series upon which most chemists have relied the one based on the titration of the chloride by means of silver is hopelessly vitiated by the imperfect execution of the method of analysis. If any further proof of this uncertainty were needed, the following table, giving a comparison of the work of different experimenters upon other chlorides, would furnish it.

MOLECULAR WEIGHTS OF CHLORIDES BY THE METHOD OF GAY-LUSSAC.

*Much assistance in preparing this list has been obtained from the well known works of these authors. The figures have all been based upon the most recently accepted atomic weights.

†These Proceedings, XXIX. 80 et seq.

Thus, Pelouze, Marignac, and Dumas all obtained low results with the method of Gay-Lussac; in fact, the error sometimes exceeded the tenth of one per cent. The cause of this error, which appeared also in the work of these experimenters upon barium, has already been pointed out in another paper.

We are thus led to infer that the true molecular weight of strontic chloride must exceed the usually accepted value 158.4 by about one tenth of one per cent, and that the true atomic weight of strontium must be nearly 87.7. This inference is confirmed by the result of the investigation now to be described.

The balance and weights, and the methods of weighing and of tabulating results employed in the work recounted below have already been described in sufficient detail.† The balance seems to have increased slightly in sensitiveness during its four years' work, owing perhaps to the smoothing of microscopic roughnesses in the bearings. It is almost needless to say again that the weights were carefully standardized from time to time, and the small, surprisingly constant corrections were always applied. The correction to the vacuum standard was calculated by the usual formula:

0.001293

H 273°

(sp. gr. substance — 0.000156);
56) 760 273° +1°

= correction in grams for 1 gram of substance.‡

The values thus calculated for the appropriate substances at 20° and 760 mm. were as follows:

Correction to be applied to One Gram of Substance.

The general plan of the following work was similar to that adopted in the case of barium. For obvious reasons the bromide of strontium was chosen as the starting point; and the investigation began with a study of the properties of the salt, in order to determine its fitness for the purpose.

The atomic weight of silver is assumed to be 107.93, and that of bromine 79.955, unless a definite statement to the contrary is made.

*These Proceedings, XXIX. 80.

†These Proceedings, XXVI. 242; also XXVIII. 5.

H = atmospheric pressure; t° = atmospheric temperature at the time of weighing; 0.000156 standard weight of air displaced by 1 gram of brass.

=