44 Messrs. Holman, Lawrence, and Barr on the 



denoted by %* e or 2je, is the algebraic sum of the thermal 

 E.M.F. proper of the junctions, of the Thomson E.M.F. in the 

 wires, and of any " stray " or local thermal E.M.F. in the 

 circuit. The last was found to be negligible throughout the 

 work. 



To observe the melting-point, the furnace containing the 

 metal is heated more or less rapidly until the melting-point is 

 approached. The blast-lamp is then adjusted to give a slowly 

 rising temperature. The thermal circuit, with the couple 

 previously fused into the metal, is connected to a suitable 

 section of a, b, c, d. The rheostats are continually adjusted 

 for zero deflexions of the galvanometer G, and the corre- 

 sponding readings of A are taken. These will show gradually 

 increasing values ; but the rise will presently be interrupted 

 by a series of constant readings, after which the readings will 

 again steadily increase. This period of constant, or nearly 

 constant, readings of A is that in which the latent heat of 

 fusion is being absorbed, and its duration is frequently several 

 minutes. The temperature at that time is, of course, that of the 

 melting-point. The reverse process, starting with the metal 

 in a molten state and cooling it gradually, shows a similar 

 period of solidification. 



]No difference was discovered between the ascending and 

 descending readings when a sufficient amount of the metal 

 and a slow rate of heating and cooling were employed. With 

 small amounts the steady reading was more or less masked by 

 phenomena which were clearly due to inequality in distri- 

 bution of temperature throughout the mass of mixed liquid 

 and solid metal. In the case of aluminium, however, some- 

 thing more than this irregularity was observed, as elsewhere 

 stated, but the time at command did not permit a study 

 beyond the point of satisfying ourselves that the point observed 

 was unquestionably the true melting-point. 



This work was done chiefly as the thesis work of Messrs. 

 Lawrence and Barr. The efficient assistance of Mr. C. L. 

 Norton contributed materially to its progress and success. 



The computation of temperatures ifrom the observed electro- 

 motive forces %e involves a knowledge of the function con- 

 necting the two, i. e. of the function 



S!«=/W, or t = F(2j«). 

 This problem has been elsewhere discussed by one of the 

 authors of this paper *. 



In that article two interpolation formulae were developed. 

 They were respectively of the following forms, applying to the 

 * Phil. Mag. xli. p. 465, 



