Efect of Pressure upon the Spectrum of an Element. 121 



struction of the molecule apparently ; and this quantity is 

 measured to a certain degree by the coefficient of expansion 

 of the element in the solid form. Therefore it might be 

 expected that the measured shift would vary in the same 

 direction as the coefficient of expansion of the solid. This is 

 actually the case, with no exception. Again, since on any 

 theory of emission of waves the wave-length varies directly 

 as the linear dimension of the portion of matter producing 

 the waves, it would be expected that the measured shift 

 would vary directly as the coefficient of linear expansion, 

 which is found to be the case. If a is this coefficient, it may 

 be stated as an experimental law that for different elements 

 (3 = ea, where c is a constant which in some way connects 

 the ordinary changes in sizs of a solid due to temperature 

 changes with the hypothetical changes in the size of the mo- 

 lecules of the vapour due to pressure changes. 



It is not difficult to see that one would expect, as conse- 

 quences of the above ideas, that the shift would be proportional 

 to the total increase of pressure, regardless of its mode of 

 production ; and that it should also vary directly as the 

 wave-length in any one series or group of lines, for in such 

 a case the longer waves indicate greater linear dimensions of 

 the vibrating segments, if the term may be used. These are 

 observed phenomena, as is stated above. 



The fact that the shift characteristic of a principal series is 

 less than that of the first subordinate, and this in turn less 

 than that of the second subordinate, would be expected, in 

 accordance with these ideas, if the molecules producing the 

 principal series were of a simpler structure than those pro- 

 ducing the first subsidiary ; and if the molecules producing 

 the second subsidiary were the most complex of all. For, 

 since the shift of a series depends upon the looseness of the 

 molecular structure, it would be expected that, if these mole- 

 cules split up in any way, the fragments would be more stable 

 and firm than the original molecules, and therefore the shift 

 of the original molecules would be greater than that of the 

 fragments. It is difficult, however, to see any reason why 

 the shift of the different series should vary according to any 

 simple law, or why the shift of the same series of different 

 elements should be in accordance with any formula so simple 

 as that of the cube root of the atomic weight. If it could 

 be assumed that the shift was proportional to the linear 

 dimensions of the segment producing the waves, most inter- 

 esting deductions might be drawn ; but there seems to be no 

 justification for the assumption. 



