_-,.< MKxxUs. A M-IUVIT-i; AND G. HKMSALKClI UN 



or rather vapour densities. Dr. FEDDERSEN was led by his researches to conclude 

 that the spark through air volatilised the metal, which afterwards, according to him, 

 took no further part in the discharge. If that is the case the process of molecular 

 diffusion should, at equal temperatures, be inversely proportional to atomic weight. 

 There is no doubt that the first luminosity of the discharge is entirely due to the 

 spark breaking through the air. The air lines are so little widened in our photo- 

 graphs that it is only close to the pole that any metal vapour can be present while 

 the air is luminous. During the interval between the initial current and the first 

 return, the metal vapour will diffuse chiefly, if not entirely, by reason of the 

 molecular velocities. The initial discharge starts a sound-wave which must leave for 

 a short time the air between the electrodes in a state of rarefaction, and it is perhaps 

 right to consider that the mass of metallic vapour suddenly formed is driven by its 

 own pressure into the partial vacuum formed by the heated air. It would seem more 

 correct, therefore, to compare the process with that of a gas under pressure flowing 

 into a vacuum than to that of pure thermal diffusion. There is not much difference 

 between these views, and we may take it that in our experiments we have approxi- 

 mately measured the velocity of sound in the metallic vapour. This gives a relation 

 between its temperature and density. Neglecting a possible difference in the ratio 



of specific heats, the relation 



V = 80 



should hold approximately, T being the absolute temperature, V the velocity of 

 sound, and p the vapour density referred to hydrogen. Thus, for cadmium, the 

 average molecular velocity found was 560, and substituting p = 56, we obtain 

 T = 2700, which seems a possible biit rather low value. We must conclude that the 

 molecule of cadmium in the spark cannot have a mass which is much smaller than 

 that determined directly near the boiling point of the metal. 



If we compare different metals with each other we are struck with the almost 

 identical numbers obtained for zinc and cadmium. It is possible, of course, that zinc 

 vapour may be diatomic, but it seems more probable that as the spectrum of zinc and 

 cadmium show homologous lines, the molecular constitution of the two vapours is the 

 same. Aluminium, with a small atomic weight, has a high velocity, and so has 

 magnesium, but the ratio of the velocity of aluminium to cadmium is roughly as 3:1, 

 while the ratio of the square roots of the atomic weights is only as 2 : 1. 



The uncertainty of our numbers is so great that we only wish at present to draw 

 the general conclusion that the two metals having the lowest atomic weights, which 

 have been examined by us, are also those showing the highest velocities. For the 

 same reason we forbear discussing the question as to the different behaviour of 

 different lines, especially with regard to other peculiarities possibly existing in the 

 behaviour of lines which show abnormal velocities. The data are at present too 

 scanty and uncertain to allow us to attach any value to the coincidence of such 

 peculiarities. 



