silver to the litre of boiling wash-water would certainly prevent all action.

A few drops of hydrochloric acid added to the wash-water greatly diminishes its solvent action on flaky chloride of silver, but does not wholly prevent it, as is evident from the fact shown in the table below, that hydrochloric acid does not precipitate chloride of silver from its solution in water nearly as effectually as nitrate of silver; and, as is well known, hydrochloric acid, if in any considerable excess, exerts a strong solvent action on the precipitated chloride.

As shown by Stas, the precipitation of chloride of silver, from its solution in hot water by the reagents we have named, depends solely on the change which the reagents produce in the solvent. That the action is an example of simple solution is shown by the fact that a considerable portion of the chloride of silver dissolved in boiling water is deposited when the solvent cools. This phenomenon is a striking one, and can easily be observed by pouring into a glass crystallizing pan some of the clear solution obtained in the experiment described above. As the water cools it becomes cloudy, and deposits a granular powder, which adheres to the bottom of the glass. The grains are usually very small; but if the solution cools slowly the crystalline form can readily be distinguished under the high powers of a good microscope, and the little cubes present all the characteristics of the native crystals of chloride of silver. It is evident, therefore, that the granular condition of chloride of silver is a crystalline condition, and this experiment may elucidate the manner in which the native crystals are produced.

We have thus far only spoken of the solubility of chloride of silver in boiling water. As is evident from the crystallization just described, the solubility rapidly diminishes as the temperature falls; but even at the ordinary temperature the solubility is distinctly marked. Lukewarm water poured on and off freshly precipitated chloride of silver, becomes decidedly opalescent on the addition of nitrate of silver, and even if cold water is used the opalescence is perceptible.

In order to obtain an approximate measure of the effects we have described, Mr. Hyams precipitated about fifteen grammes of chloride of silver, and, after thoroughly washing it, boiled the precipitate with a large volume of water in a glass flask. At the end of an hour he decanted through a filter about one litre of the boiling water, and, having divided the filtrate into two portions, he added to one portion nitrate of silver, and to the other, hydrochloric acid. The precipitated chloride of silver was in each case collected and weighed. At the end

of two hours' boiling, two other portions were filtered off and treated in a similar way. These determinations were then repeated with a fresh quantity of chloride of silver, and afterwards taking a third quantity of chloride of silver, the boiling water was simply poured on twice in succession, and the similar portions thus obtained treated as before. The results in every case were nearly the same as shown in the following table. In this table

1 and 2 are results after one hour's boiling of 1st quantity.

9 and 10 after simply pouring on boiling water.

If we assume that the amount of chloride of silver precipitated by nitrate of silver under the conditions described above is a correct measure of the solubility of the chloride, it appears from the above determinations that about two milligrammes of chloride of silver are dissolved by each litre of boiling water, and further that only about one half of the amount thus dissolved is precipitated by hydrochloric acid.

In making chlorine determinations, it is a very common practice to wash with very hot water, in order to secure the prompt settling of the chloride of silver, or to wash away any occluded material, and it was the chief object of this investigation to determine the extent to which the solubility of the chloride in distilled water might affect the result. For this purpose we made two series of determinations

of the chlorine in chloride of antimony; in both cases precipitating with nitrate of silver the chlorine from a solution of the chloride of antimony in tartaric acid and water with the usual precautions. But, while in the first series the precipitated chloride of silver was washed with boiling hot distilled water to about the rooooo according to Bunsen's scheme; in the second series, although hot water was also used in washing, one decigramme of nitrate of silver per litre was added to each successive portion of the wash-water poured upon the precipitate, until the last two portions, which were poured on cold. By this simple device, the advantages of washing with hot water may be secured, while its solvent action is prevented. The results are given in the following table:

It is evident from these results that when great accuracy is required, the solubility of chloride of silver may become a very serious source of error in determinations of chlorine, and in our first paper on the atomic weight of antimony, this was the chief cause of the discrepancy between the analyses of chloride of antimony on the one hand, and the bromide, iodide, and sulphide of antimony-analyses of which closely agreed among themselves on the other hand. It was shown in the paper just referred to that, although the greatest care was taken in purifying the material, the chloride of antimony used actually left

behind on evaporation a sufficient amount of oxichloride to reduce the per cent of chlorine 0.116 (loc. cit. page 64). The mean results which we actually obtained from seventeen analyses of chloride of antimony was 46.620; and when to this we add 0.212 and 0.116, the sum is 46.948, which differs from 46.985 the theoretical value when Sb 120, and Cl=35.457 - by only 0.037, or if we take Cl35.5 by 0.072. In this estimate we leave out of the account the known solvent action on chloride of silver of the tartaric acid used to keep the antimony in solution. This must equally affect both of the series of determinations given above, and fully accounts for the small difference that remains to be explained. This whole discussion, however, only serves to confirm the conclusion previously expressed, that chloride of antimony is a most unsuitable material for the basis of an atomic weight determination; and, having explained the anomaly to which we gave prominence in a previous paper, we shall here take leave of the subject. We have also studied the solubility of bromide of silver in water, but this is so small that we found it difficult to measure the amount. In water at the ordinary temperature, or even in tepid water, bromide of silver is practically insoluble. In boiling water it is perceptibly soluble, but not more so than is chloride of silver in water at the ordinary temperature. Hence the determination of bromine does not require the same precautions, and is susceptible of greater accuracy than that of chlorine; and on this account, as well as for other reasons which we have previously discussed, the atomic weight of antimony can be deduced from the analyses of the bromide of antimony with as great accuracy as can at present be reached in such determinations. In the seven determinations of the per cent of bromine in bromide of antimony, whose results are given beyond (p. 18), the maximum difference from the mean value 66.6651 was only 0.0045, and Professor Mallet, in his analyses of bromide of aluminum, has obtained with the same method a similar degree of accuracy.*

In conclusion, we would again express our obligations to Mr. G. M. Hyams, who has assisted us in the work of this investigation.

IV. ADDITIONAL EXPERIMENTS ON THE ATOMIC WEIGHT OF

ANTIMONY.

In our previous paper on this subject, we gave our reasons for the opinion, since fully confirmed, that the bromide of antimony is the

* Philosophical Transactions, Part III. 1880, 1022.

most suitable compound of this element, as yet known, for determining its atomic weight; and the results of fifteen analyses of five different preparations of the bromide were published, which gave for the atomic weight in question the mean value 120.00 with an extreme variation between 119.4 and 120.4 for all the fifteen analyses, and between 119.6 and 120.3 for the six determinations in which we placed most confidence. The antimonious bromide used in these determinations was purified first by fractional distillation, and secondly by crystallization from a solution in sulphide of carbon. In the crystallized product thus obtained, the bromine was determined gravimetrically as bromide of silver in the usual way. Although it seemed at the time that the results were as accordant as the analytical process would yield under the unfavorable conditions, which the presence of a large amount of tartaric acid in the solution of the bromide of antimony necessarily involved; yet it was obvious that the agreement was far from that which was desirable in the determination of an atomic weight, and our chief confidence in the accuracy of the mean value — independently of its remarkable agreement with previous resultswas based on the fact that the known sources of error tended to balance each other. Hence our conclusions were stated with great caution, and the hope was expressed that, after a more thorough investigation of the subject, we might be able to return to the problem with such definite knowledge of the relations involved as will enable us to obtain at once more sharp and decisive results than are now possible."

In our previous paper, we described a simple apparatus which we devised for subliming iodide of antimony; and in a note to the paper we stated that we were applying the same process to the preparation of bromide of antimony, and that it promised excellent results. Our expectations in this respect have been fully realized, and the prod uct leaves nothing to be desired, either as regards the beauty or the constancy of the preparation. The fine acicular crystals are perfectly colorless, and have a most brilliant silky lustre. With ordinary precautions, they can be kept indefinitely without change, and it is easy therefore to determine the weight of the material analyzed to the tenth of a milligramme.

The material used in the following determinations was first prepared as described in our previous paper. It was then repeatedly distilled from a small glass retort rejecting at each distillation the first and the last portion. Lastly, it was twice sublimed in a slow current of absolutely dry carbonic dioxide. As it was only possible to sublime a