( -^ItJ ) 



positive. (I found for instance a^ positive with tin-amalgams). It 

 therefore seems, that when one of the components (water for instance) is 

 an associating substance, or when tlie other component is electroJyticaUy 

 dissociated, we must certainly not follow D. Berthelot in writing- 

 a^a^ = a%2, independent of the tact, that in such cases neither 

 «1 and «2, nor b^ and b.^ are constants. 



The formule (8) in my previous communication now becomes: 



1 a,.v' 



T—T. 



or with — = «, " =: a 



9o 9o 



1 + 



1 "%(l-.^■) 



9o 



lUV 



^■=^«ï3^k^ («'-) 



Fi-oni obser^'ations, where the values of x are less Ihan 0,1, the 

 value of was found to be exactly 0,396 ^). If we now further 

 acce])t for the values of the coefficients a and r : 



a = 0,0453 ; /• = — 0,74. 



Avhich are calculated from other observations Avith higher values of./', we 

 obtain {T, = 273,15 + 231,63 = 504,8) the survey on the next page. 

 We notice, that in this table the agreement is an excellent one ; the 

 average deviation is about 0,9'', whilst in the case of the empirical 

 formula with ^^v^ and y.i-' (see previous communication), if the last 

 value is not counted, it amounted to 0.85''. van der Waals's expression 

 for the correctionterm, therefore, represents at least equally well the 

 course of the mcltingpoint-line over the portion, observed from 212° 

 to 65''. But what is still more important, is the fact, that whilst my 

 foi'mer empirical formula does not very accurately' represent the two 

 last observations (the difference in the last even amounted to 10°), 

 VAN DER Waals's expression not only satisfactorily represents these, 

 but also the four obserA'ations at still lower temperatures (compare 

 p. 22 of VAN Het eren 's Dissertation). In this observations the values 

 of .V and t were determined by analysis of the liquid phaFe, which 

 is in equilibrium with the solid phase at a given temperature. 



200 

 1) In the previous communication 0.400 was accepted, but — = 0,396 is some- 



505 



what more accurate. 



