THERMO-ELECTRIC QUALITY UNDER PRESSURE. 277 



soldering the wire to a helix of copper wire, which was in turn soldered 

 to the connecting tube. Connection between the tube and the cyl- 

 inder was amply provided by the mechanical contact through the 

 threads, which supported the entire thrust of the internal ])ressure. 

 The sohlered connection at the top of the wire was not made until 

 after the bottom connection had been made and screwed into position, 

 so that all initial strains of tension or torsion were avoided. The 

 outer wire was protected over its entire length by a loosely fitting 

 rubber tube, which slipped over the outer end of the insulating plug 

 and prevented contact with the water of the bath. The connections 

 at E and D to the copper leads to the electrical part of the apparatus 

 were made with soft solder. The wire FE and part of the copper lead 

 in the upper bath were also protected with rubber tubing. The two 

 connections E and D were placed close together and tied to the out- 

 side of the upper cylinder. They were protected from direct contact 

 with the water of the bath with a covering of felt 1 inch thick. The 

 temperature of these junctions was therefore controlled by the large 

 mass of the cylinder and remained constant in spite of small fluctua- 

 tions in the bath liquid. The copper leads from E and D were taken 

 from the same spool and were so homogeneous that no undesirable 

 e.m.f. was ever introduced by this part of the circuit. The connec- 

 tion at F was a mechanical connection. A small hole was bored from 

 the outer surface of the cylinder to within | inch of the inner surface. 

 Against the flat bottom of this hole a steel tube which contained the 

 wire was forced by a nut. This wire was lead axially through the 

 tube, and insulated from it, except at the very end, where it was 

 soldered in place with enough soft solder to make a little mound at 

 the bottom of the tube. When forced into place by the screw, this 

 mound was flattened out, making at the same time good electrical 

 contact and preventing contact between the water of the bath and the 

 wire. This wire was brought as close as possible to the inside of the 

 cylinder in this way in order to reduce to a minimum any disturbance 

 arising from slow fluctuations in the temperature of the l)ath. Any 

 irregularity introduced by flow of heat into the bath along the pressure 

 tubing, either above or below, may be shown by a simple computation 

 to be absolutely negligible; any such efl^ect is certainly less than 10~^ 

 degrees. The upper bath was the variable bath, and was controlled 

 by a thermostat within 2 or 3 thousandths of a degree. In ^•irtue of 

 the precautions just described all appreciable errors due to tempera- 

 ture control were eliminated. In the lower bath, slow variations in 

 temperature would be of greater effect, because of the long stem of the 



