THERMO-ELECTRIC QUALITY UNDER PRESSURE. 379 



metal. This effect was often looked for in the early days of the sub- 

 ject, and the concensus of opinion was that the effect does not exist; 

 whenever an effect like it ^\^ls obtained, it was always explainable by 

 imperfect homogeneity in the metal. At the same time there seems 

 to be no reason why the effect might not exist, and it occurred to me 

 that with present apparatus we are in a position to push the limits 

 within which the effect must lie much further than before. The 

 circuit on which I experimented consisted of liquid mercury, and was 

 therefore completely homogeneous. The mercury was contained 

 in a quartz capillary; at the center of the circuit the capillary was 

 drawn down with thin walls to perhaps 0.5 mm. diameter and 0.25 

 mm. thickness of wall. Over the neck in the capillary was slipped 

 a piece of tightly fitting mica. A simple arrangement allowed a jet 

 of water to be directed against the quartz on one side of the mica 

 and on the other side a small gas flame. In the mercury underneath 

 the mica there was, therefore, an intense temperature gradient. The 

 two ends of the quartz capillary were led to an ice bath, and from this 

 connection made through copper leads to a galvanometer. There 

 was therefore no chance for thermo-electric action unless there were 

 one due to the temperature gradient. The galvanometer was a very 

 sensitive Thomson galvanometer constructed by Coblentz; I owe the 

 opportunity to use it to the kindness of Dr. I. Gardner. With this 

 I could establish that there was no e.m.f. in the mercury of as much 

 as 3 X 10"^^ volts, up to temperature gradients steep enough to vapo- 

 rize the mercury at one end of the constriction in the capillary. Care 

 had to be taken in establishing this result; for example, if the part 

 of the circuit containing the gradient is not horizontal, an effect will 

 be found due to the effect of pressure differences on thermo-electric 

 quality. 



The effect, if it exists, is therefore exceedingly minute. The 

 equations above show that if the effect exists consequences with 

 regard to other effects are involved. We have the equation 



Hence if (ta for example, depends on the temperature gradient, then 

 Pab niust also, or the Peltier heat would depend on the total quantity 

 of heat flowing into the junction. Such an effect has not been ob- 

 served, and therefore inferentially the Thomson heat is independent 

 of temperature gradient. However, the observations on Peltier heat 

 have presumably been no more accurate than those on the freedom of 

 total e.m.f. from effects of gradient. 



