648 PROCEEDINGS OF THE AMERIC.VN ACADEMY. 



to an actual displacement of the electric current in our plate, the flow, 

 under the action of the magnet, being stronger along one edge than 

 along the other edge of the plate 1 This seems very unlikely, especially 

 in view of the fact that, in Plate 2 at least, the resistance appears to 

 be slightly, very slightly, decreased by magnetization. Indeed, the 

 whole amount of heat generated per minute per unit length of the 

 plate by our current Cp was about 0.6 calorie, and the change in this 

 amount due to the magnetization was probably not so much as 0.00002 

 calorie. 



There is at least one other point of view which is worth trying. 

 The transverse potential-gradient set up in the Hall effect is, appar- 

 ently, like the Thomson potential -gradient, static, maintained without 

 flow of electricity do\vn the gradient, though there may be, if the cir- 

 cuit is properly closed, flow of electricity up this gradient. In the 

 Thomson effect this static condition is attended by, and is apparently 

 due to, a temperature-gradient of the opposite slope. Is it perhaps 

 true that wherever such a static potential-gradient exists in a metal 

 there is an accompanying temperature-gradient, and are the two 

 gradients in iron always of opposite slope 1 The Ettingshausen tem- 

 perature-slope is opposite in our iron to the potential slope of the Hall 

 effect, and to this extent encourages our question ; but when we come 

 to the ratio of the two slopes and compare this with the ratio of the 

 two slopes in the Thomson effect, we find a great difference. In the 

 latter case the ratio is our s. The exhibit follows : 



(V.) 



We have at least the satisfaction of finding that the ratio of the two 

 ratios, although each changes a good deal, nearly 50 per cent, remains 

 pretty nearly constant, ^^ especially at the lower temperatures, where 

 the values of « T^ are probably more accurate than for the higher ones. 

 Indeed, the whole change from 20° to 100° is less than 6 per cent. 

 But the transverse temperature-gradient is only one two-hundredth 

 part of what it should be to make the transverse ratio equal to 5. 



It would be difficult to go farther in this line of inquiry without 



^^ It should be remembered, however, tliat the temperature-coefficient of 

 eTh cannot be regarded as accurately determined. 



