three globules of aluminium, tolerably large considering the size of the experiment, were obtained, along with a large number of very small ones. The larger ones were melted together under KCl. Some experiments made in iron crucibles were not attended with the same success as those of Rose, no globules of any considerable size remained in the melted fluorides; the metal seemed to alloy on the sides of the crucible, which acquired a color like zinc. It is possible that this difference may have arisen from using a higher temperature than Rose, as we made these experiments in a furnace, not over the blowpipe. Porcelain and clay crucibles were also tried, but laid aside after a few experiments, owing to the action of the fluorides upon them, which in most cases was sufficient to perforate them completely."

The above papers, Rose's and Dick's, contain all the published researches with cryolite until Deville's attention was turned towards it. He then took up the subject with his accustomed thoroughness. The following pages are taken from his 'De l'Aluminium,' the subject not being given in its entirety, but only the most important points. He published the first account of these researches in 'Ann. de Chem. et de Phys.' [3], xlvi. 451:

"I have repeated and confirmed all the experiments of Dr. Percy and H. Rose, using the specimens of cryolite which I obtained from London through the kindness of MM. Rose and Hofmann. I have,

furthermore, reduced cryolite mixed with NaCl by the battery, and I believe that this will be an excellent method of covering with aluminium all the other metals, copper in particular. Anyhow, its fusibility is considerably increased by mixing it with Al2C16.2KCl. Cryolite is a double fluoride of aluminium and sodium, containing 13 per cent. Al, 32.5 per cent. Na, and 54.5 per cent. F. Its formula is Al2F6.6NaF. I have verified these facts myself by many analyses."

Deville then gives a description of methods of making cryolite artificially, which is unnecessary to repeat here, for natural cryolite is so cheap that these methods are of no practical importance. He continues:

"In reducing the cryolite I placed the finely-pulverized mixture of cryolite and NaCl in alternate layers with sodium in a porcelain crucible. The uppermost layer is of pure cryolite, covered with NaCl. The mixture is heated just to complete fusion, and, after stirring with a pipe-stem, is let cool. On breaking the crucible, the aluminium is often found united in large globules easy to separate from the mass. The metal always contains silicon, which increases the depth of its natural blue tint and hinders the whitening of the metal by nitric acid, because of the insolubility of the silicon in that acid. M. Rose's metal is very ferruginous. I have verified all M. Rose's observations, and I agree with him concerning the return of metal, which I

have always found very small. There are always produced in these operations brilliant flames, which are observed in the scoria floating on the aluminium, and which are due to gas burning and exhaling a very marked odor of phosphorus. In fact, P20 exists in cryolite, as one may find by treating a solution of the mineral in sulphuric acid with molybdate of ammonia, according to H. Rose's reaction.

"The facility with which aluminium unites in fluorides is due without doubt to the property which these possess of dissolving the alumina on the surface of the globules at the moment of their formation, and which the sodium is unable to reduce. I had experienced great difficulty by obtaining small quantities of metal poorly united, when I reduced the Al2C16.2NaCl by sodium; M. Rammelsberg, who often made the same attempts, tells me he has had a like experience. But, I am assured by a scrupulous analysis that the quantity of metal reduced by the sodium is exactly that which theory indicates, although after many operations there is found only a gray powder, resolving itself under the microscope into a multitude of small globules. The fact is simply that Al2C16.2NaCl is a very poor flux for aluminium. MM. Morin, Debray, and myself have undertaken to correct this bad effect by the introduction of a solvent for the Al2O3 into the saline slag which accompanies the aluminium at the moment of its formation. At first, we found

it an improvement to condense the vapors of Al2C1o, previously purified by iron, directly in NaCl, placed for this purpose in a crucible and kept at a red heat. We produced in this way, from highly colored material, a double chloride very white and free from moisture, and furnishing on reduction a metal of fine appearance. We then introduced fluorspar (CaF2) into the composition of the mixture to be reduced, and we obtained good results with the following proportions:

The double chloride ought to be melted and heated almost to low red heat at the moment it is employed, the NaCl calcined and at a red heat or melted, and the CaF2 pulverized and well dried. The double chloride, NaCl and CaF2 are mixed and alternated in layers in the crucible with sodium. The top layer is of the mixture, and the cover is NaCl. Heat gently, at first, until the reaction ends, and then to a heat about sufficient to melt silver. The crucible, or at least that part of it which contains the mixture, ought to be of a uniform red tint, and the material perfectly liquid. It is stirred a long time and cast on a well-dried, chalked plate. There flows out first a very limpid liquid, colorless and very fluid, then a gray material, a little more pasty, which contains aluminium in little grains,

and is set aside, and finally a button with small, metallic masses which of themselves ought to weigh 20 grms. if the operation has succeeded well. On pulverizing and sieving the gray slag, 5 or 6 grms. of small globules are obtained, which may be pressed together by an earthen rod in an ordinary crucible heated to redness. The globules are thus reunited, and when a sufficient quantity is collected the metal is cast into ingots. In a well-conducted operation, 75 grms. of Na ought to give a button of 20 grms. and 5 grms. in grains, making a return of one Al from three of Na. Theory indicates one to two and a half, or 30 grms. of Al from 75 of Na. But all the efforts which have been made to recover from the insoluble slag the 4 or 5 grms. of metal not united but easily visible with a glass, have been so far unsuccessful. There is, without doubt, a knack, a particular manipulation on which depends the success of an operation which would render the theoretical amount of metal, but we lack it yet. These operations take place, in general, with more facility on a large scale, so that we may consider fluorspar as being suitable for serving in the manufacture of aluminium in crucibles. We employed very pure fluorspar, and our metal was quite exempt from silicon. It is true that we took a precaution which is necessary to adopt in operations of this kind; we plastered our crucibles inside with a layer of aluminous paste, the composition of which has been given in Ann. de Chem. et de Phys.,' xlvi.

6