THERMO-ELECTRIC QUALITY UNDER PRESSURE. 275 



This was the same apparatus as that used in much previous work,^ 

 and has been described so thoroughly that no further description is 

 needed here. The pressure was measured throughout with the same 

 coil of manganin wire as that used for the measurements of pressure 

 coefficient of resistance. It was calibrated from time to time. The 

 pressure part of all this work went more smoothly than ever before; 

 the complete set of measurements on 24 different specimens was made 

 without a single rupture, except one pinching off of the stem of an 

 insulating plug, and with only a few leaks due to defective washers. 



Various details of manipulation had to be carefully observed in 

 order to get good results. It is well known that thermal e.m.f. is 

 exceedingly sensitive to slight mechanical changes. For instance, 

 the wire of an ordinary constantan thermo-couple cannot be bent 

 without changing its constants. I was prepared, therefore, to expect 

 very large irregularities at these high pressures, and was most pleas- 

 antly disappointed to find that, with some precautions, reliable 

 measurements could be obtained. It is above all else necessary to 

 avoid all permanent distortion. The viscosity of most liquids be- 

 comes so great under high pressure that permanent distortion may 

 easily result. The connecting tube is about 18 inches long, and the 

 hole ^ inch diameter. When pressure is increased or decreased the 

 entire liquid in the tube moves bodily one way or the other, because 

 of change of volume in the liquid in the lower cylinder, and if the liquid 

 is sufficiently viscous will carry the wire with it, and in some cases 

 may easily break it. The magnitude of this A-iscous drag depends 

 on the diameter of the wire; the smaller the diameter the larger 

 the ratio of surface to cross section, and the larger the effect. It 

 was desirable for this reason to use in many cases a wire larger than 

 that on which resistance measurements were made, and I therefore 

 prepared a number of new wires from the same samples as the old 

 wires. Viscous drag may result either in a permanent distortion of 

 the wire, showing itself as a permanent change of the zero reading, 

 or it may produce an effect wuthin the elastic limit which disappears 

 with release of pressure. Such an elastic effect was shown by several 

 of the smaller wires; it may be eliminated by reversing the direction 

 of the change of pressure. The viscous effect can be reduced in most 

 all cases to negligible proportions by using a very thin liquid to trans- 

 mit pressure, changing pressure slowly, and by running pressure back 

 and forth several times over a small range before making a reading. 

 The liquid found suitable was petroleum ether, which had also been 



3 P. W. Bridgman, Proc. Amer. Acad. 49, 627-643, 1914. 



