certain High-temperature Boiling-points. 



153 



be entered here. Little can be learned from them, as might 

 have been expected. To get comparable values for the con- 

 stants of a complex equation, the data must either be very 

 fine or in great number. 



13. Bearing in mind therefore that it is my object to 

 detect possible relations, and that my high-temperature 

 boiling-points are necessarily somewhat crude, I will facilitate 

 computation by observing that in Dupre's data the prevailing- 

 value of A is between 15 and 20. Assuming A=20, the 

 following set of constants were obtained, in which, of course, 

 there is greater uniformity. Water is added from Dupre 

 and Bertrand. 



Table VII. — Constants for Boiling-point and Pressure. 



Substance. 



A. 



B. 



C. 



Water* ,.. 



19324 

 20 

 20 

 20 



2795 

 4379 



7467 

 8433 



3-868 

 4-217 

 3643 

 3603 



Cadmium 



Zinc 





* Bertrand (I.e. p. 93) inadvertently puts A =r 1744324, in which case, 

 however, pressures are measured in atmospheres instead of millimetres of 

 mercury, which is his usual standard. The same constant is repeated by J. J. 

 Thomson. 



14. From an inspection of Table VII., I was led to suspect 

 better agreement in making C constant throughout, and A 

 variable. This step is further suggested by assuming, con- 

 formably with the indications of Table VIL, that for any two 

 substances S and & the boiling-points 6 and 6', corresponding 

 to a given pressure j?, will follow the relation B/B' = #/#' = 1/m, 

 where n is constant for the given pair of substances. This is 

 virtually the principle of Groshans, and postulates a funda- 

 mental equation of the form (1), from which all others are 

 derived by substitution as follows : — 



logp = A-B/0-Clog<9, (1) 



= A-?iB/»0-C log n$+ C log w, 



= A'-B'/6"-Clog<9' (2) 



The constants of Table VIII. are obtained by supposing 

 C to be constant for all the substances. As to a choice of 



