PRESSURE ON RESISTANCE OF METALS. 599 



The mean temperature coefficient of the sample above was 0.00434, 

 which is considerably higher than the best vahie of Jaeger and Diessel- 

 horst,^ 0.0039 (uncorrected). Their aluminum showed by analysis 

 0.5% Fe, and 0.4% Cu, and was therefore much less pure than the 

 sample above. The initial pressure coefficient at 0° is given as — O.OgS? 

 by Williams ^^ against — O.O54I6 from the deviation curves above. 

 Williams does not give the analysis or temperature coefficient of his 

 sample. 



The results are somewhat unusual. When the resistance-pressure 

 curves for different temperatures are scaled to the same initial resist- 

 ance, the curves for the higher temperatures are very slightly less 

 steep, and ver\' much less curved. 



Silver. This was obtained in the form of wire O.OG inch in diameter 

 from the United States Mint at Philadelphia. The original material 

 from which the wire was drawn was "proof" silver, in which no impur- 

 ity whatever could be detected by any chemical test. The drawing 

 was done with steel dies. After drawing, it was found that 0.03% of 

 impurit\' had been introduced, which was probably mostly steel from 

 the dies. I owe this examination of the effect of drawing to the interest 

 and kindness of the late Dr. D. K. Tuttle. It seems safe to assume 

 that the impurity so introduced was in the outer layers. I therefore 

 removed the outer layers by scraping with glass, reducing the diameter 

 from 0.06 to 0.04 inch. It is very probable that nearly if not all the 

 0.03% of impurity was removed in this way, although I made no 

 analysis of the wire after scraping. From the size 0.04 inch it was 

 further drawn to 0.003 inch through diamond dies and wound with 

 single silk insulation by the New England Electrical Works. It was 

 annealed after the final drawing. For the pressure experiment it 

 was wound on itself into a coreless toroid of 31.5 ohms resistance 

 at 0°. It was seasoned in the same way and at the same time as the 

 cadmium. Connections to the electrodes were made with silver solder. 



The smoothed results are shown in Table VIII and the experimental 

 points in Figure 10. The permanent change of zero after each run 

 averaged 0.1% of the total effect; . the maximum departure of any 

 point from a smooth curve, except for one point at 0° where the vis- 

 cosity was high, was 0.5%, and the average numerical departure was 

 0.1%. The departures from smoothness were almost entirely due to 

 the viscosity of the kerosene used as a transmitting medium. This 

 run was made before petroleum ether had been tried. The curves of 

 deviation from linearity are not symmetrical or parabolic although 

 nearly so. They also are shown in Figure 10. 



