1889.1 On certain Ternary Alloys. 469 



5 10 15 . . 20 85 30 35 40 



J'f rr ru/ f.tfjr of y>// in HeuviC7~ Alloii. 



representing the first and the dotted line the second. Fig. 2 repre^ 

 sents the corresponding curves of the second kind, and fig. 3 those 

 of the third kind. Obviously, in each case there is little difference 

 between the continuous and dotted lines, so that it may be fairly 

 concluded that the effect of variation in temperature from 565 to 

 689 is negligible as compared with the experimental errors, more 

 especially those due to imperfect separation by gravitation of the two 

 alloys from one another. 



The curves representing the distribution (fig. 3) are remarkable. 

 As long as the tin percentage in the total mass is less than about six- 

 teen the lighter alloy contains more tin than the heavier one ; at 

 about this point (representing some 14 per cent, in the heavier and 

 18 per cent, in the lighter alloy) the difference becomes a maximum 

 after which the difference diminishes, until at about 28 per cent, the 

 same percentage of tin is contained in both alloys. After this the 

 heavier alloy contains more tin than the lighter, the difference con- 

 tinually increasing. 



Causes retarding Separation of Lighter and Heavier Alloy. 



Before proceeding further we thought it desirable to trace out the 

 causes rendering separation incomplete, even after some hours' 

 standing. At first we attributed it to the influence of traces of im- 

 purities in the metals used, but on repeating the observations in 

 pipe-bowls with the purest samples of each metal obtainable, we still 

 occasionally got irregular results, showing that this was not the sole 

 cause. Next we thought that the error might be due to the partial 

 formation of eutectic alloys during solidification, in such a way that 



