308 BRIDGMAN. 



The pressure effect on thermal e.m.f. could be measured without 

 the sources of error which made the resistance measurements inaccu- 

 rate. It is very difficult to make good electrical connections to mag- 

 nesium, since it cannot be soldered, and any mechanical connection 

 encounters the resistance of the rapidly formed layer of oxide. Any 

 change of resistance introduces, however, no error in any potentio- 

 meter method of measuring e.m.f. like the present, so that it was 

 possible to make e.m.f. measurements over the entire range, as for any 

 other metal, whereas the resistance measurements could be made 

 only at 25°. More readings were made on magnesium than on other 

 metals because of irregularities in the earlier runs due to insufficient 

 stirring of the ice bath. The final runs were satisfactory. The maxi- 

 mum zero correction was 0.75% of the maximum effect and the 

 maximum divergence of any single observation from a smooth curve 

 was 3.1% of the maximum, but the average numerical departure was 

 only 0.22%. In passing from the curves at constant temperature to 

 those at constant pressure an adjustment of the 75° curve of 3% of 

 the maximum effect had to be made; elsewhere the readjustment was 

 very slight. 



The numerical values are shown in Tables XI and XII and Figures 

 12 and 13. The e.m.f. is negative and comparatively large, rising to 

 nearly 1.3 X lO'^ volts at 100° and 12000 kg. The curves of e.m.f. 

 at constant temperature against pressure are concave upwards, show- 

 ing an increasing proportional effect at higher pressures, contrary to 

 the normal behavior. At constant pressure, e.m.f. against tempera- 

 ture is linear. The Peltier heat is also negative, increasing numerically 

 with both temperature and pressure. The Thomson heat is unaffected 

 by pressure; this of course follows directly from the linearity of the 

 effect with temperature at constant pressure. 



There are no other measurements for comparison. 



Aluminum. The material was freshly extruded wire of 0.02 inch 

 diameter from the same rod as the wire on which the resistance meas- 

 urements were made. The original resistance wire was too small for 

 this purpose. The wire was annealed after extrusion by several 

 hours at 120° in an electric oven. The effects with aluminum were 

 so unusual that the measurements were repeated on a second speci- 

 men, this time of commercial wire. As purchased, this wire was 

 about 16 inch diameter; it was drawn down for this experiment through 

 steel dies to 0.0265 inches, and annealed by heating in boiling water 

 for about an hour. 



The average temperature coefficient of resistance at atmospheric 



