the Nature of Spectra. 87 



double line of sodium appeared dark, but not thicker than in 

 a short tube under the same conditions, and much thinner 

 than when the density of the sodium was only a little increased. 

 What was here proved for sodium vapour may in an analo- 

 gous manner hold good for the other gases, such as nitro- 

 gen &c. At all events the above experiment shows that changes 

 of thickness must not be supposed to run parallel with changes 

 of density. In order that this should be the case, the curve 

 which represents the dependence of the coefficient of absorp- 

 tion would have to be perfectly identical with that which ren- 

 ders the disturbance (conditioned by the greater number of 

 the collisions in consequence of the greater density) of the 

 vibrations of the aether envelopes as a function of the number 

 of the collisions (which is extremely improbable), while the 

 widening of the lines would be fixed by the former curve at 

 increasing thickness, by the latter at increasing density of the 

 absorbing layer. 



Further experiments, by Lockyer, Schuster, and others, 

 teach besides that sodium, as well as the vapours of other 

 metals, possess different absorption-spectra at different tempe- 

 ratures, yet without the limits within which the thickness of 

 the absorbing stratum varies being very wide. From this 

 may also be explained the recent experiments of Liveing and 

 Dewar, who vaporized sodium in a vertical tube heated below 

 to a white heat, the bottom of which served at the same time 

 as the source of light ; they then introduced from above to 

 various depths a tube containing hydrogen, and observed the 

 changes of the spectrum. Thereby, on the one hand, the 

 thickness of the layer through which the rays passed was 

 varied, but, on the other, when the tube was sunk deeper the 

 most extremely absorbent layers had certainly a much higher 

 temperature than when it was lowered to a less depth*. 



Spectra of Solids and Liquids. — If we condense gases into 

 the solid or liquid state, the vibrations in the individual mole- 



* It might at first appear as if the action of the molecules in a body 

 upon a luminous ray which passes through it (as exemplified in disper- 

 sion, absorption, the rotation of the plane of polarization) were not pro- 

 portional to the number of the acting molecules in the unit of volume (as, 

 we learn from experiment, it is), but to the number in the unit of length, 

 and therefore to the third root of the former. But if we employ a method 

 analogous to that which serves for ascertaining the number of the collisions 

 of a molecule in a space filled with other molecules of gas, we see that 

 the vibrations of the luniiniferous aether are affected by all the molecules 

 which are present in a cylindrical space whose radius is equal to the 

 radius of action of a molecule upon an aether particle. But the number 

 of molecules in such a space increases proportionally to the total number 

 of the molecules. 



