PRESSURE ON RESISTANCE OF METALS. 619 



Within the Hmits of error the deviation from linearity is symmetrical 

 and parabolic; the numerical values are given in Table XVII. 



The temperature coefficient found above between 0° and 100° 

 was 0.00297. Kaye and Laby give 0.0033, evidently taken from W. 

 von Bolton. On consulting the original paper/' however, it will be 

 found that the value is given with only one significant figure, 0.003. 

 The agreement of our results is, therefore, within the linn'ts of error. 

 The General Electric Co. was not able to give any information about 

 the purity apart from that afforded b}' the temperature coefficient. 



There seem to be no previous measurements of the pressure coeffi- 

 cient for comparison. 



When scaled to the same initial resistance, the resistance-pressure 

 curves are found to be steeper at the higher temperatures, but less 

 curved. 



Tungsten. This was obtained from the General Electric Co. 

 through the kindness of Dr. W. D. Coolidge. It is bare and 0.0004 

 inch diameter. It was seasoned for temperature at the same time as 

 Mo and Ta and for pressure by a single application of 12000 kg. at 0°. 

 There was no perceptible change of ^ero after the very first application 

 of pressure. It was wound loosely in two or three turns on a smooth 

 bone cylinder, and held loosely in place with silk thread. Its initial 

 resistance at 0° was 79.0 ohms. Connections were made with gold 

 solder to platinum by arcing in hydrogen. 



The smoothed results are collected in Table XVIII, and the experi- 

 mental values of mean coefficient and departure from linearity are 

 shown in Figure 20. The zero drift during a run was not large, being 

 respectively 0.25%, 0.0%, 0.56%, 0.08%^, and 0.054% of the total 

 effects at 0°, 25°, 50°, 75°, and 100°. The individual results were 

 irregular, however. At 75° and 100° in particular, there are large 

 irregularities in a direction opposite from that of hysteresis. The 

 maximum departure of any point from a smooth curve was 0.7% of 

 the total effect, and the numerical average 0.15%. The deviation 

 from linearity is slight; within the limits of error it is symmetrical 

 and parabolic. The characteristic numerical values are given in 

 Table XVIII. 



The average temperature coefficient found above is 0.00322. This 

 is low compared with the value 0.0050 of Somerville.^^ He found a 

 rather unusually large departure from linearity with temperature. 

 The source of Somerville's tungsten was the same as mine, the General 



17 W. von Bolton, ZS. Elektrochem. 11, 45 (1905). 



