320 Prof. Thomson, 



and not the density of the whole collection of particles. Thus 

 we might have a great density of luminous particles and yet a 

 small density of the particles giving out some particular line, in 

 this case we should not get a widening of the line : thus for 

 example if we have a small quantity of one gas A mixed with 

 a large quantity of another B, we might have a large current 

 density through the mixture and yet a small density of the 

 particles giving out the A lines — in such a case the lines of A 

 would not widen : we see in this way the explanation of the 

 well-known fact that the lines of an impurity do not widen. 



Different lines in the same spectrum widen to very different 

 amounts, some become quite broad whilst others hardly widen at 

 all : the difference being far greater than could be explained by 

 the factor 1/p in the expression for Sp. May not this difference 

 in the behaviour of the lines be due to different lines originating 

 in different systems ? We are accustomed to regard all the lines 

 in, say, the iron spectrum as corresponding to possible periods of 

 vibration in the normal atom of iron ; it seems to me more 

 probable that many of these lines originate from systems which 

 are not the normal atom, but systems — corpusculates we might 

 call them — formed by the combination of the atom with some of 

 the crowd of corpuscles by which it is surrounded in the electric 

 discharge. When a discharge passes the atoms are surrounded by 

 a crowd of corpuscles and it seems probable that compounds, not 

 permanent indeed, but lasting sufficiently long to give out their 

 characteristic vibrations might be formed between the atom and 

 the corpuscles. These compounds need not be all of the same 

 kind, nor need all kinds be equally numerous, the lines given out 

 by systems which were numerous would widen, while those given 

 out by the scarce systems would not. 



The absorption spectrum of the cold vapour would not show 

 the lines due to these temporary compounds of atoms and cor- 

 puscles, but these lines might be reversed by vapours through 

 which an electric discharge is passing or which are at such high 

 temperatures that they are good conductors of electricity ; in such 

 cases corpuscles are present and could form compounds with the 

 atoms. 



The lines most likely to be reversed would be those corre- 

 sponding to the most prevalent compound, as in the reversing 

 layer the systems able to absorb these lines would be more 

 numerous than the systems able to absorb any other lines ; and 

 since in the luminous gas the density of these systems would also 

 be greater than that of the others, these lines would be the most 

 easily widened. Thus we should expect lines easily widened to be 

 easily reversed. 



So far we have only considered the effect on the periods of the 



