602 



BRIDGMAN. 



run was 0.1 G% of the total effect. Except for several points at 0°, 

 where the viscosity is high, the maximum departure of any point from 

 a smooth curve was 0.3% of the total effect, and the average numeri- 

 cal departure was 0.08%. The curves of deviation from linearity 

 are distinctly not symmetrical, and show an unusually large progres- 

 sive change in the location of the maximum with increasing tempera- 

 ture. The deviation curves are shoAvn in Figure 11 also. 



20'' 40= 60' 80 

 Temperature 



100° 

 - Gold 



2 4 6 8 



Pressure, Kg. /Cm.' X 10 



10 



12 



Figure 11. Gold, results for the measured resistance. The deviations 

 from linearity are given as fractions of the resistance at kg. and 0°C. The 

 pressure coefficient is the average coefficient between and 12000 kg. 



The average temperature coefficient between 0° and 100° listed 

 above, 0.003968, is for the coil which had been subjected to pressure. 

 Another coil of the same material, not subjected to pressure, gave 

 0.004009. The curves of resistance of each of these specimens against 

 temperature were equally smooth, the departure from smoothness at 

 any point not being more 1/7000 of the total change between 0° and 

 100°. It would seem, therefore, that 0.004009 should be taken as the 

 best value for pure gold under normal conditions. This figure is to be 

 compared with 0.00368 given b}' Jaeger and Diesselhorst ^ for "pure" 

 gold. But the temperature coefficient of gold is unusually sensitive 

 to impurity; Jaeger and Diesselhorst found that gold with .0.1% Fe, 

 and 0.1% Cu and a trace of Ag had a coefficient of only 0.00203. 

 Undoubtedly the lower value of Jaeger and Diesselhorst for "pure" 

 gold is to be explained by slight impurities in their sample. 



The initial pressure coefficient at 0° found graphically from the 

 deviation curves above is —0.03117, which is to be compared with 

 — O.O527 of Beckman.^ He had trouble with change of zero in his 

 measurements, and states that the error may be over 3%. The mean 



