276 BRIDGMAN. 



used before in the resistance measurements partly mixed with kero- 

 sene. This was always now poured pure into the lower part of the 

 apparatus without admixture of kerosene. The upper of the three 

 cylinders was initially filled with a mixture of equal parts kerosene 

 and ether. This liquid remains so thin under even the highest pres- 

 sures that somewhat more trouble was found from minute leaks than 

 ever with undiluted kerosene. The bearing surfaces of the washers 

 must be in very good condition, otherwise leak takes place along 

 scratches on the steel too minute to be sufficiently closed by the 

 plastic flow of the steel at high pressures. However, all these minute 

 leaks closed up automatically under sufficiently long duration of 

 pressure, and never could introduce any error in the readings, which 

 were made only after leak had disappeared. Leak merel}' made the 

 manipulation somewhat more inconvenient. The use of petroleum 

 ether made it impossible, because of its low boiling point, to make 

 measurements at atmospheric pressure at the higher temperatures. 

 Above 25°, therefore, the initial readings were made at pressures of 

 300 kg., and extrapolation made to atmospheric pressure. 



The effect of viscous drag was further increased by the necessity 

 for insulating the wire AB from the steel tube. The wire was in most 

 cases bare, and insulation was provided by slipping over the wire 

 short lengths (| inch long) of thin glass tube. It was necessary to 

 choose these of the best diameter. If they >vere too large, the viscous 

 resistance between the glass and the wall of the steel tube was so great 

 that all the pieces moved together bodily, and might cut off the wire 

 at the upper end ; whereas if the glass tube was too small the viscous 

 resistance between the glass and the wire was so great that the stress 

 on the wire was no longer truly hydrostatic. The best size for the 

 glass was found after several trials. That the effect of viscous drag 

 was finally successfully eliminated is shown by the fact that even for 

 metals as soft as lead and thallium the permanent change of zero 

 after a run to 12000 kg. was less than 1/10% of the total effect. 



The inner wire was attached to the lower insulating plug with soft 

 solder, except in a few cases such as magnesium or aluminum, when 

 connection was made by wrapping the joint with small copper wire 

 and smothering over the outside with soft solder. Slight variations 

 of resistance at the contacts evidently were of no importance, and it 

 therefore was not necessary to take the precautions used in the re- 

 sistance measurements. At the upper end, a flexible connection to 

 allow for inequalities of compression and relative motion between 

 the cylinder and the core of the insulating plug was made b}- soft 



