1898.] on Experiments on Certain Elements in relation to Heat. 741 



ences of melting point, which in these cases can have no effect on the 

 result. 80 large a difference must be due to peculiarities inherent in 

 the atoms themselves ; and differences of atomic heat are to a certain 

 extent comparable with the differences observed in other physical 

 properties, which, like specific volume, specific refraction, &c., are 

 approximately additive. 



If we try to think what is going on in the interior of a mass of 

 solid when it is heated, the work done is expended not only in setting 

 the atoms into that kind of vibration which corresponds to rise of 

 temperature, that is, it makes them hotter, but partly in separating 

 the molecules or physical units from one another (= expansion) and 

 partly in doing internal work of some kind, the nature of which is not 

 known. A difference between metals and non-metals has been brought 

 out by the researches of Heycock and Neville, who find that metals 

 dissolved in metals are generally monatomic ; whereas it is generally 

 admitted that iodine, sulphur and phosphorus in solution are poly- 

 atomic. It is moreover remarkable that, although in respect to specific 

 heat each element in a solid seems to be independent of the other 

 elements with which it is associated, when the elementary substances 

 are vaporised some rise in separate atoms like mercury, some in 

 groups of atoms like iodine, sulj)hur, arsenic and phosphorus, and as 

 the temperature is raised these groups are simplified with very vary- 

 ing degrees of readiness. 



The two metals, cobalt and nickel, with which I began my inquiry, 

 have very nearly the same atomic weight, the value, 58 • 24 for nickel 

 and that for cobalt 58 • 49, being calculated by F. W. Clarke from the 

 results of a great many analyses by many different chemists. They 

 are so close together that for a long time they were regarded as 

 identical, and Mendeleef does not hesitate even to invert the order by 

 making Co = 58*5 and Ni = 59. These metals, nevertheless, differ 

 from each other in several very important chemical characters. Nickel, 

 for example, forms the well known and highly remarkable compound 

 with carbonic oxide discovered by Dr. Mond. Cobalt, on the other 

 hand, produces many ammino-compounds to which there is nothing 

 corresponding among the compounds of nickel. 



Having put aside the common excuses for the observed diverg- 

 ences from the constant of Dulong and Petit, we are compelled to 

 look round for some other hypothesis to explain them. 



The constitution of carbon compounds is now accounted for by a 

 hypothesis concerning the configuration of the carbon atom introduced 

 by Van't Hoff and Le Bel twenty-five years ago, and which is now 

 accepted by the whole chemical world. It seems not unreasonable to 

 apply a similar idea to the explanation of those cases of isomerism 

 which have been observed in certain compounds of the metals, notably 

 chromium, cobalt and platinum. This has already been done by Pro- 

 fessor Werner, of Ziirich. If the constitution of comi^ounds can be 

 safely explained by such hypothesis, this implies the assumption of 

 peculiarities in the configuration of the individual constituent metals 



Vol. XV. (No. 92.) 3 



