Atomic Laws of Thermochemistry . 17 



shown there that if n is the valency of the metal R in a haloid 

 compound RS», then w{Wl)* is the sum of a constant for R 

 and n times a constant for S. By determining* mean values 

 for these parts and then recalculating values of M.H, it would 

 be possible to smooth out some of the irregularities in the 

 values of M.H given in the table, but for the present I prefer 

 to leave the values just as derived from the experimental data. 

 Where a melting-point has been wanting in the last table, I 

 have divided the value of M 2 Z derived from the boiling-point 

 by 1*2 and enclosed it in brackets as a value derived from the 

 melting-point, the value of Mlp/J or L derived from this is 

 also enclosed in brackets. 



To determine how the values of the latent heat of vaporiza- 

 tion per gramme-molecule due to molecular force stand in 

 relation to experimental values of the total heat of vaporization, 

 we have the following four values of molecular latent heats 

 at boiling-point taken from Berthelot's Mecanique Chimique, 

 namely, 7-2 for CC1 4 , 6*3 for SiCl 4 , 7-6 for SnCl 4 , and 8'4 for 

 AsCl 3 ; these should be diminished by about one eleventh of 

 their value to remove the part due to external work, and then 

 increased by the latent heat of fusion to give the molecular 

 latent heat of vaporization of the solid without performance 

 of external work. Let us suppose these two corrections to 

 neutralize one another as latent heats of fusion are known to 

 be small, then we have the comparison : — 



- 0C1 4 . SiCl 4 . SnCl 4 . AsCl 3 . 

 Experimental molecular latent heat ... 7*2 6*3 7'6 8*4 



Calculated „ „ L. ... 8-4 77 9*1 8-1 



This shows a fair general agreement between the two sets of 

 values of L : in the first three cases the theoretical value is 

 about 1*2 times the experimental ; while in the case of the 

 one metal Hg, for which we could make the comparison, the 

 experimental was found to be 1*6 times the theoretical. 

 Accordingly it seems safest to multiply all our theoretical latent 

 heats for the metals by 1*6, and divide the theoretical latent 

 heats for compounds by 1*2, in order to be as little at variance 

 as possible with the few direct measurements that are available. 

 The next table contains the values of the latent heats per 

 gramme-equivalent thus derived from those in Tables I. and 

 III, by the factors 1*6 and 1/1*2 : — 



Phil. Mag. S. 5. Vol. 40. No. 242. July 1895. C 



