of the Interrupted Spectra of Gases. 5 1 



riodic time, but that this period is exposed to perturbation when 

 two molecules pass sufficiently close to one another, and that 

 during the intervals between two perihelion passages it settles 

 down towards its mean value (Phil. Mag. vol. xxxvi. (1868) 

 p. 135). The perturbations, as a rule, seem to take place in 

 both directions, some of them increasing and others of them di- 

 minishing the periodic time ; for in most gases the lines widen 

 out in both directions on raising the temperature — that is, on 

 rendering the molecular collisions more violent. But it some- 

 times happens that the disturbed motion differs in character 

 from the normal motion so sensibly that it gives rise to a some- 

 what different pattern of spectrum. 



19. For example, this occurs in the case of the principal lines, 

 the D lines, of the sodium-spectrum. By introducing sodium- 

 carbonate into a suitable part of the oxyhydrogen flame, the bright 

 lines can be made to widen out to any desired amount ; while 

 the original positions of the lines are at the same time presented 

 to the eye by the dark absorption-lines, or reversed spectrum as 

 it is called, caused by the surrounding mantle of cooler sodium- 

 vapour. In this way the positions of the lines when narrow and 

 when wide can be directly compared, and will be found to differ. 

 In neither of the lines does the dark line lie in the middle of the 

 bright band : in Di (the more refrangible one) it inclines towards 

 the red end of the spectrum, and inD n (the less refrangible one) 

 it inclines towards the blue. Hence the middles of the broad 

 bands are further asunder than the narrow lines. 



20. A similar appearance is met with in the case of chloro- 

 chromic anhydride. The eye can easily detect that the lines are 

 not everywhere equally spaced, though the deviation of any one 

 line from its calculated position is so slight that in the measures 

 we have taken the amount cannot be separated from errors of 

 observation. The middles of the lines appear to be displaced by 

 small amounts, some to the left, others to the right. And since 

 those parent motions in the molecules of the vapour whose pe- 

 riodic times are somewhat longer than the normal amount must 

 have a predominance in the formation of the lines that deviate 

 towards the blue end of the spectrum, and vice versa, we can by 

 this property distinguish between the lines which belong to mo- 

 tions in the gas that are faster than, equal to, or slower than, 

 those of normal period ; and by treating each of the series of 

 lines so obtained as in section III. of this paper, information as 

 to the differences in the nature of the motion in these three 

 cases may perhaps be attainable. 



21. It is evident that we only need to obtain more accurate 

 measures to be able, by applying the method described above in 

 § 9 j to determine with precision by how much the periodic motion 



E2 



