20 Mr. C. H. Lees on 



o 



the temperature,* and that Angstrom's solution (2) can only 

 be taken as a rough approximation to the actual variation 

 of temperature throughout the bar. 



To get rid of this difficulty of the variation of h with the 

 temperature, and at the same time to do away with the 

 necessity for a knowledge of the values of c and p for the 

 crystal, I began experiments with the bars and crystal 

 packed in a cylinder of saw-dust, and used only the "steady 

 state." The bars were of brass, 2 cms. diam. and 26 cms. 

 long. The ends of each bar were amalgamated, and along 

 the curved surface four thermo-j unctions of iron and brass 

 wire were soldered, two being as near the ends as possible. 

 The two bars were held in position within a vertical paper 

 cylinder 8 cm. diam. by means of six set-screws, which 

 enabled the two amalgamated surfaces, between which the 

 crystal disc was to come, to be set parallel. After putting in 

 the crystal, the space between the bars and paper cylinder 

 was packed with saw-dust, the screws withdrawn, and con- 

 tacts made at top and bottom with a water and a steam 

 can respectively. The ends of the wires from the thermo- 

 junctions came outside the paper cylinder and could be put 

 in succession in circuit with a galvanometer. Observations 

 of deflections of the galvanometer and a previous determina- 

 tion of the constants of the junctions, give the temperatures 

 at eight points of the bars, and from these the values oidvjdx 

 below and above the crystal respectively can be calculated. 

 These, combined with a knowledge of the conductivity k of 

 the brass bar, give the flow of heat into and out of the 

 crystal disc respectively through its plane surfaces. From 

 the eight observations of temperature, the temperatures of 



* It is evident, moreover, that the solution (2) can only apply to a bar of 

 infinite length heated and cooled at one end. For a finite bar, and especially 

 for the case of a bad conductor interposed between two lengths of bar, 

 it is necessary to add to the expression (2) a corresponding expression in which 

 the sign of x is altered. This makes the calculation of k and h much more 

 complicated. 



