440 SCIENCE PROGRESS 



reversed ; that is to say, other things being equal, emission and 

 absorption lines are similarly and equally affected. The dis- 

 placement has also been found to be independent of the nature 

 of the gas surrounding the arc. This indicates that the pheno- 

 menon is primarily due to a change of density of the metallic 

 vapour rather than to a change of pressure. The two terms 

 have been used rather loosely and indiscriminately, but the 

 distinction is of some importance from the theoretical point of 

 view. All the facts point to the displacement as arising from 

 the closeness of the packing of the radiating molecules. It is 

 certain that with increase of pressure of the surrounding gas 

 the density of the metallic vapours in the arc increases, because 

 the electrodes are found to burn away faster, reversals become 

 more frequent, and there is an increase in the brilliancy of the 

 arc. Moreover, as Larmor has observed, " mechanical pressure 

 arises merely from the translatory motions of the molecules, 

 and these are so slow as hardly to count in connection with 

 radiation periods." 



All the experiments agree in proving that the displacement 

 is proportional to the increase in pressure, at least for pressures 

 up to a limit of ioo atmospheres ; and also that it increases with 

 the wave length for lines of the same series. There has been 

 some diversity of opinion as to the actual law of its dependence 

 upon wave length, and a knowledge of the law is of great import- 

 ance as a means of testing the theory. Humphreys's experiments 

 seemed to indicate a linear relation, and indeed Sanford, 1 in 

 discussing this subject, has used the fact that certain theories do 

 not give a linear relation as an argument against them. More 

 recent investigations have, however, negatived this result. 

 Duffield a found that a linear law would not hold in the case of 

 the spectra of iron, gold or silver, and that the displacement 

 varied with a higher power of the wave length than the first. 

 Rossi 3 showed that, in the case of vanadium, " the displacement 

 seems to be roughly proportional to the square or a higher 

 power of the wave length." The more complete investigation 

 of the iron spectrum by Gale and Adams 4 has shown that, in this 

 case, the displacement varies as the cube of the wave length. 

 By plotting their results upon a large scale, with wave lengths 

 as abscissae and displacements as ordinates, they found that the 



1 Astroph. Journ. 35, p. 3, 1912. 3 Astroph. Journ. 34, p. 21, 191 1. 



2 I^oc. tit. ante. K Loc. tit. ante. 



