THERMO-ELECTRIC QUALITY UNDER PRESSURE. 309 



pressure between 0° and 100° of the commercial aluminum was 

 0.004307. According to this test it would seem, therefore, only sliglitly 

 less pure than the pure aluminum from the American Aluminum Co. ; 

 its coefficient as found in the resistance paper was 0.00434. The 

 relation between resistance and temperature of commercial alumi- 

 num was sensibly linear. 



At atmospheric pressure the thermo-electric behavior of pure 

 aluminum against lead is given by the formulas; 



E = (-0.416^ + 0.00008 r--0.00001 f) X 10-6 volts, 



P = (-0.416 + 0.00016^-0.00003^2) (^ _|_ 273) x lO'^ volts, 



a = (0.00016-0.00006 H + 273) X KT^ volts/°C. 



For commercial aluminum the corresponding formulas are: 



E = (-0.378^-0.00005/- + 0.0000094 Z'^) X lO'^ volts, 



p = (_o..378-0.0001/ + 0.0000282/2) {t + 273) X 10"^ volts, 



a = (-0.0001 + 0.0000564 /) (/ + 273) X 10"^ volts/°C. 



The numerical values of the differences between the constants for these 

 two different specimens are small, thus bearing out the observation 

 on temperature coefficient. 



Under pressure, the effects are more complicated, and for the most 

 part very small. It would not have been possible to carry through 

 these readings except for the perfect steadiness of the galvanometer, 

 allowing readings to be made to the limit of sensitiveness. This was 

 true both for the commercial and the purer samples. Wagner re- 

 marked also on the same behavior of his specimen. This is somewhat 

 surprising, because one would be inclined at first to explain the com- 

 plicated nature of the effects, which involves reversals of sign, by 

 assuming two modifications in varying proportions. But there was 

 no evidence whatever for any incompleteness of internal equilibrium; 

 if there is such an effect, the change of equilibrium must occur 

 immediately. 



Measurements on the purer sample were made first. This was 

 exposed twice to 9000 kg. at room temperature, then the apparatus 

 taken apart to remedy a leak, and pressure then applied again twice 

 to 12000 at 25° before beginning readings. Regular runs were made 

 at 25°, 50°, and 75°, but at 95° the wire broke with increasing pressure, 

 evidently because the transmitting medium was too stiff, being one 

 half kerosene and one half ether, instead of entirely ether as usual. 

 It w^as set up again with a new piece of wire, a contiguous piece from 

 the same length as the first piece, and this was seasoned by two appli- 



