668 PROCEEDINGS OF THE AMERICAN ACADEMY. 



steel and the wrong sign in the other cases. I have not found any 

 perfectly explicit statement by Voigt himself of the relation between 

 the Nernst coefficient and the Hall coefficient, but I reason as follows: 

 If we consider iron and take a case in which dT/dy is positive, we 

 have, as we saw a little distance back, F as a negative quantity ; that 

 is, according to Voigt, there is a thermo-electric e. m. f. acting in the 

 direction of decreasing y. But with our iron in open circuit, as it 

 would be for the Nernst-efFect observations, we have no flow of elec- 

 tricity along the iron, the natural thermo-electric e. m. f. being balanced, 

 or, as Voigt says, compensated, by a static charge having a negative 

 potential-gradient in the positive direction of y. If, now, we think of 

 the magnetic field as rotating the equipotential lines of this potential- 

 gradient in the direction of the Amperian current, as the equipotential 

 lines of the primary electric current are rotated in the Hall effect in 

 iron, we get what in our paper has been called, and what Moreau 

 would call, a negative Nernst effect. 



According to this argument, which deals with a puzzling question 

 and may be incorrect in its conclusion, all the signs in the third column 

 of Moreau's table, as given above, should be changed, and in every case 

 for which the specific heat of electricity within a metal and the thermo- 

 electric height of the metal with respect to lead have the same sign 

 the Moreau formula and the Voigt formula will predict opposite signs 

 for the Nernst effect. 



The Moreau formula seems to me profoundly suggestive, though 



not strictly correct. But the fact that the Voigt formula, which I 



believe to be fundamentally wrong in one of its factors, gives results 



numerically so like those of Moreau, and so like the values given by 



direct observation, is enough to warn us to be cautious in examining 



every theory and testing every formula relating to the matters here 



dealt with. 



E. H. H. 



