EFFECT OF PRESSURE ON CONDUCTIVITY OF METALS. 93 



of error in this method, and it will be seen later that the results are 

 consistent with this explanation. In some of the later measurements 

 the attempt was made to cut down the motion possible in the thermo- 

 couple by threading into the hole along with the couple a wire 0.002 

 inch diameter. In some cases this seemed to produce good effects, 

 and in others apparently not. There is danger that too close a fit 

 may introduce tensions into the thermo-couple, and so change its 

 constants. It did not seem feasible to try for a much closei* fit be- 

 tween the couple and the hole because of the difficulty of getting a 

 perfectly smooth coating of enamel on the wire of the couple. 



The specimens of the two soft metals lead and tin were cast, and the 

 holes for the thermo-couple were cast with the specimens, using a 

 0.007 inch wire for the core, held in a proper position in the mold. 

 For all the other metals the holes were drilled. A small jig was made 

 and the holes were drilled in a jeweller's lathe. In spite of the use of 

 the jig it was not possible to make all the specimens exactly alike, and 

 the distance apart of the holes was independently measured in every 

 specimen with a microscope. The extreme variation in their distance 

 apart was from 0.190 to 0.210 cm. I found that a very convenient 

 drill may be made of a piece of tungsten wire, pointed so as to form the 

 conventional flat drill. The advantage of this over steel is that it is so 

 tough that it is almost impossible to twist off, which may very easily 

 happen to a steel drill in a clinging metal like gold, for example. 

 While not as hard as steel, tungsten is hard enough to drill all the 

 metals used here, which included iron, nickel, and platinum, all in the 

 annealed condition. 



Because the hole containing the couple is comparatively short, 

 there is greater danger of error from heat leakage along the thermo- 

 couple wire than in the first method. To eliminate this source of 

 error, the wire of the couple was bent sharply at the entrance of the 

 hole, at either end, and was wrapped once, or in some cases twice, 

 about the specimen, and held in close contact with the external sur- 

 face by a wrapping of fine silk thread. This ensures that the wire for 

 some distance from each junction shall have the same temperature as 

 the specimen at the junction, and so should eliminate error from leak. 

 As an additional control with regard to heat leakage along the wire, 

 the experiment was tried with two specimens of first using a couple 

 constructed as above (constantan between two lengths of copper) and 

 then a couple constructed of constantan between two lengths of 

 "therlo." "Therlo" is an alloy made by the Driver Harris Co. with 

 very nearly the same e.m.f. against constantan as copper, but with 



