Chemical Equilibrium of Solids. 



31 



P to Q. With the vapour-bath in the position bd, pressure 

 would thus cool the end A and heat the end B of the glass 

 tube A B ; but the cooling would exceed the heating. Hence 

 an error of the sign of the data, Table I. &c. Now let the 

 vapour-bath be shifted to the position ac and the thermal 

 state be steady. The B-half of the tube A B is now exposed 

 to cold, and compression can only have the effect of heating 

 these parts, seeing that the A-half is protected. Hence an 

 error of a sign opposite to the data of Table I. &c. Finally, 

 when the bath has the position c e, the error first considered 

 would be accentuated. 



I made experiments in accordance with this scheme at 310°, 

 but found that this severe test did not change the character 

 of the results of Table I. Data are superfluous. Again, de- 

 creasing the length of the tube A B had no appreciable effect. 

 Similar tests were made with reference to Table II. 



33. It is clear at the outset, since pressure increases both the 

 resistance of glass and the electromotive force NaHg/glass/Hg, 

 that changes of the latter will act in a way tending to wipe 

 out the change of resistance. Let e, I, R denote the electro- 

 motive force, current, and resistance of the uncompressed 

 circuit. Let I p be the current corresponding to pressure p, 

 and A'r and k e be respectively the pressure-coefficients of 

 resistance and electromotive force. Then 



hn={l{l+Kp)-l p )lp.I r (1) 



Table II. gives k e = 10/10 6 , say, whence the values of Table XL 

 result from Table I., series III. cf. § 11. In the final column 

 Z/rssSR/R. Sp, the former coefficient. 



Table XL — Allowance for Increments of Electromotive 

 Force. (9 = 215°. 



p> 



Wp.Jce. 



l(\+pJc e ). 



h- 



10 6 x£k. 



10 6 x£'r. 



523 



995 



1411 



5-2 

 9-9 



141 



21-18 

 21-28 

 2137 



19-47 

 18-11 

 17-27 



168 

 175 

 168 



157 

 164 



156 



512 

 1016 

 1403 



5-1 

 10-2 

 140 



21-18 

 21-28 

 21-37 



19-39 

 18-07 

 17-28 



180 

 172 

 169 



169 

 163 

 156 



Thus the data underlying §§ 22, 23, 24 are accentuated. 

 At 310° a permanent pressure-coefficient h e could not be de- 

 tected, and at 100° it is certainly very much in error (§ 34). 

 Since therefore k e is never fully vouched for, I have omitted 

 it in the computation of /c-r altogether. 



