Fig. 1.

Deville of an application on a large scale of the laboratory method just described. He first put it up at the chemical works of M. du Sussex, at Javel, and later at the works of MM. Rousseau Bros., at Glacière. It has at present only an historic interest, as it was soon modified in its details so as to be almost entirely changed, but I give it here so as to show the different phases through which the industry has passed. The text is not given in full as Deville describes it, which would be unnecessary; but the condensed account gives a clear idea of the process. The full description may be found in Deville's book, or in the Ann. de Chem. et de Phys.' [3] xlvi. 445, where it first appeared.

The crude Al2C1o, placed in the cylinder A, is vap

orized by the fire and passes through the tube to the cylinder B containing 60 to 80 kilos of iron nails heated to a dull-red heat. The iron retains as relatively fixed ferrous chloride, the ferric chloride and hydrochloric acid which contaminate the A12C1, and likewise transforms any sulphur dichloride (SC12) in it into ferrous chloride and sulphide of iron. The vapors on passing out of B through the tube, which is kept at about 300°, deposit spangles of ferrous chloride, which is without sensible tension at that temperature. The vapors then enter D, a cast-iron cylinder in which are three cast-iron boats each containing 300 grms. of sodium. It is sufficient to heat this cylinder barely to a dull-red heat in its lower part, for the reaction once commenced disengages enough heat to complete itself, and it is often necessary to take away all the fire from it. There is at first produced in the first boat some aluminium and some sodium chloride, which latter combines with the excess of A12C16 to form the volatile chloride Al2C16.2NaCl. These vapors of double chloride condense on the second boat and are decomposed by the sodium to aluminium and sodium chloride. A similar reaction takes place in the third boat when all the sodium of the second has disappeared. When on raising the cover it is seen that the reactions are over, the boats are taken out, immediately replaced by others, and are allowed to cool covered by empty boats. In this first operation

the reaction is rarely complete, for the sodium is protected by the layer of NaCl formed at its expense. To make this disappear, the contents of the boats are put into cast iron pots or earthen crucibles, which are heated until the AlCl begins to volatilize. Then the pots or crucibles are cooled and there is taken from the upper part of their contents a layer of NaCl, almost pure, while underneath are found globules of aluminium, which are separated from the residue by washing with water. Unfortunately, the water in dissolving the Al2Cl of the flux exercises on the metal a very rapid destructive action, and only the globules larger than the head of a pin are saved from this washing. These are gathered together, dried, melted in an earthen crucible, and pressed together with a clay rod. The button is then cast in an ingot mould. It is important in this operation to employ only well purified sodium, and not to melt the aluminium if it still contains any sodium, for in this case the metal takes fire and burns as long as any of the alkaline metal remains in it. In such a case it is necessary to remelt in presence of a little Al2C16.2NaCl."

Deville says later, "Such was the detestable process by means of which we made the ingots of aluminium which were sent to the Exposition.'

Deville, after this, tried some experiments in which he used sodium vapor, and he thus reports his results in his book: "This process, which I have not perfected, is very easy to operate, and gave

me very pure metal at the first attempt. I operate as follows: I fill a mercury bettle with a mixture flask of chalk, carbon, and carbonate of soda, in the proportions best for generating sodium. An iron tube about ten centimetres long is screwed to the bottle, and the whole placed in a wind furnace, so that the bottle is heated to red-white and the fas tube is red to its end. The end of the tube is then introduced into a hole made in a large earthen crucible about one-fourth way from the bottom, so that the end of the tube just reaches the inside. surface of the crucible. The carbonic oxide (CO) disengaged burns in the bottom of the crucible, heating and drying it; afterwards the sodium flame appears, and then pieces of AlCl are thrown into the crucible from time to time. The salt volatilizes and decomposes before this sort of tuyere from which issues the reducing vapor. Al2C16 is added when the vapors coming from the crucible cease to be acid, and when the flame of sodium burning in the atmosphere of A12C16 loses its brightness. When the operation is finished, the crucible is broken and there is taken from the walls below the entrance of the tube a saline mass composed of NaCl, a considerable quantity of globules of aluminium, and some sodium carbonate, which latter is in larger quantity the slower the operation was performed. The globules are detached by plunging the saline mass into water, when it becomes necessary to notice the reaction of the water on litmus. If the

water becomes acid, it is renewed often; if alkaline, the mass impregnated with metal must be digested in nitric acid diluted with three or four volumes of water, and so the metal is left intact. The globules are reunited by melting with the precautions before given."

Deville modified these methods in various ways. AlCl is a deliquescent salt, difficult of preservation, and so was soon replaced by A12C16.2NaCl, which does not present these inconveniences. The double chloride, however, does draw some moisture and holds it energetically, from which it results that at a high temperature it will give rise to some alumina, which encloses the globules of metal with a thin coating and so hinders their easy reunion into a button. Deville remarked that the presence of fluorides facilitated the reunion of these globules, which he attributed to their dissolving the coat of Al2O3 on them. Since then, the employment of a fluoride as a flux is considered necessary to overcome the effect produced primarily by the Al2C1o.2NaCl holding moisture so energetically. The first fluoride employed by Deville was fluorspar, which was soon replaced by cryolite. This opens up the subject of the reduction of aluminium from cryolite, and since Percy and Rose both preceded Deville in using it, I will first give their investigations, following with those which Deville published in 1859.