666 PROCEEDINGS OF THE AMERICAN ACADEMY. 



X^ — KH^, or A— — , 



by p 



a- being the specific heat of electricity." 



Moreau, then, seems to admit that Voigt's F is a real thermo-electric 

 force, 



d®^dT a© 



dTdi/' 3/ 



directed along the F-axis at any point in the metal 31, though not 

 the force which he considers significant in the Nernst efi"ect. But it 

 seems to me that this supposed total electromotive force at any point 

 in the metal 31 is a fiction. The only electromotive forces which we 

 have reason to suppose existing in an unequally heated piece of metal 

 in open circuit, as a metal is when tested for the Nernst effect, are 

 those found in the Thomson effect. When we have two unequally 

 heated metals united in a thermo-electric circuit, the total electromo- 

 tive force at any point in either metal is likely to be something differ- 

 ent from that which the Thomson effect alone at that point would 

 account for, but we have no sufficient reason for supposing it to be 



actually the ^rr^ ^- mentioned above. If, for example, we are con- 



sidering iron and if, for simplicity, we assume that dT/di/ = 1, the 

 thermo-electric e. m. f. in the iron at temperature 7' is, according to 

 Voigt, represented by the length of the line TI, taken as a negative 

 quantity, in Figure 1 1 of our paper. Now it is true that, if we make 

 this assumption and then integrate around the whole circuit, which we 

 will suppose to be of copper and iron, we shall get the area CoCIIoCo 

 as the total e. m. f. of the circuit, which result will be correct. But we 

 should arrive at precisely the same correct integral result if we placed 

 the T„T line of Figure 11 indefinitely far to the left or to the right of 

 its present position, which would have the effect of increasing or de- 

 creasing indefinitely the value of the expression d®„JdT, the sup- 

 posed force at any individual point in either metal. I can see, then, 

 no objective reality in Voigt's supposed force Y. 



Moreau was, I think, the first to put Voigt's formula for the value 

 of the Nernst coefficient to the numerical test. He gives, ^3 the three 

 bottom lines being from his own observations, 



23 C. R., 130, 564 (1900).- 



