Relation to Pressure and Temperature. 487 



the said error of fi is never larger than 2 per eent. By com- 

 paring table 3 with tables 4 to 18, ($), it will be Been that the 



results at high temperatures are much more favorable. Sim- 

 ilar inferences were obtained with water, which may therefore 

 omitted here. 



13. TeoiKermals and isendrqpics. — To obtain the isothermals 

 of the compressed liquids rigorously, it is necessary to wait a 

 long time after each compression. By so doing the tube is 

 seriously weakened, and at high temperatures, in virtue of the 

 diminished viscosity (cf. § 8), it is probably even distorted. 

 It appears, however, that for liquids like the present, the 

 distinction between isothermal and adiabatic compression is 

 beyond the scope or sensitiveness of the present method of 

 volume measurement. Hence, I made my observations by 

 varying pressure from 20 atm. to the maximum, and then from 

 the maximum back to 20 atm., with only such allowance of 

 time as was necessary to take the observations at the successive 

 stages of pressure. Near the critical temperature this mode 

 of procedure is no longer warranted. 



Throughout the work the chief reliance was placed on the 

 constancy of the fiducial zero. The volume decrements cor- 

 responding to a given pressure must be identical, no matter 

 whether they are observed in the pressure increasing or pres- 

 sure decreasing march of the measurements. 



14. Thermal exjxinsiun. — Measurements of heat expansion 

 are necessary for the coordination of the results. Without a 

 thermometer tube, such measurements are not very accurate ; 

 and the inaccuracy is much increased by an unavoidable break- 

 ing off of the ends of the mercury thread, whenever the sub- 

 stance operated on solidifies. In passing from one constant 

 temperature to another, an exchange of boiling tubes is nec- 



iry. It is difficult to keep the substance fused during 

 the whole time of readjustment. On solidification the charge 

 contracts as much as 10 per cent. The mercury thread is 

 therefore forced into the vacuities in the axis of the solid 

 thread. After fusing again it will often be found that the 

 mercury thread is broken into distinct pieces, sometimes in 

 great number, alternating w^ith small threads of substance. 

 All of these must be measured, an operation which is not only 

 excessively tedious, but makes the expansion measurement, in- 

 accurate. For this reason I added no correction for the ex- 

 pansion of glass. The errors thus introduced may amount to 

 2 or 3 per cent in unfavorable cases. See tables 5 to 19. At 

 high temperature such corrections would be arbitrary for the 

 reasons given in §11. Here however, it is probably quite 

 negligible, because of the large volume expansions. 



