202 MESSRS. C. T. HEYCOCK AND F. H NEVILLE 



Owing to the rapid oxidation of the aluminium, even when a current of coal gas was 

 led into the crucible, the synthetical method of arriving at the composition of the 

 alloys from the weights of metal added was impracticable. It was therefore neces- 

 sary, between every two readings of the freezing point, to extract a sample of the 

 alloy for analysis. This was done by sucking out a portion of the thoroughly molten 

 and stirred metal in a pipette of Jena combustion glass. If rapidly performed, this 

 process is practicable, even at the melting point of gold. The samples were thus 

 obtained in the form of rods, or occasionally as tubes, about 4 millims. in diameter. 

 After breaking off the glass these rods could easily be cut up for analysis and for 

 microscopical study. The order of operation was as follows : 



(1.) A dose of aluminium was added, and the alloy well stirred. 

 (2.) The freezing point was determined roughly. 

 (3.) The metal was re-melted, stirred, and a sample extracted. 

 (4.) The alloy was again re-melted, if necessary, and the freezing point deter- 

 mined accurately. 



The cycle was then repeated. 



Except when we approached the composition AuAl 2 , each sample extracted was a 

 rod of very uniform composition, so that analyses of the upper and lower portions gave 

 identical results. But at the very high temperatures of extraction needed for alloys 

 near Au A1 2> the operation had to be conducted very rapidly on account of the softening 

 of the Jena glass, and it was not always possible to obtain solid rods. In such cases 

 the upper and lower portions of the same extract sometimes differed in their com- 

 position to the extent of almost 1 per cent. This and other causes, which will be 

 referred to later, made the determination of the curve in the neighbourhood of 

 AuAlo a difficult matter. 



The analyses were conducted as follows : A weighed quantity of the alloy, from 

 2 to 4 grammes, was digested with aqua regia in a covered porcelain dish, until all 

 soluble matter had dissolved. It was then evaporated with excess of hydrochloric 

 acid, diluted and filtered. The residue on the filter paper seldom weighed more than 

 a few milligrams, except when portions of glass adhered to the alloy. This residue 

 was ignited and weighed ; it consisted, partly at all events, of graphite. The weight 

 of. the residue was subtracted from the weight of the alloy taken for analysis, and 

 the difference regarded as pure alloy. The filtrate containing the chlorides of gold 

 and aluminium was considerably diluted, and the gold precipitated by sulphurous acid 

 gas. The gold was filtered off, ignited and weighed, and the percentage of gold in 

 the pure alloy thence calculated. In the majority of cases we obtained the percentage 

 of aluminium by difference, assuming that the pure alloy contained only gold and 

 aluminium. This procedure was justified for all parts of the curve except those near 

 AuAl 2 , by the test analyses given in square brackets in the tables, in which a complete 

 determination, both of the gold and of the aluminium was made. In these test 



