﻿Fixed-Arm Spectroscopes. 345 



40 millim. long. Probably this degree of accuracy could be 

 secured if necessary by the use of a minimum deviation attach- 

 ment of pure linkwork, like that shown in fig. 4 (PL IX.), 

 which is far superior in every respect to the ordinary sliding 

 link-form usually placed on a spectroscope*. It may also be 

 secured by the system first described, because in it any 

 required degree of accuracy may be reached by an increase 

 in size of the drums a, b. 



It was at first thought that a return to this system would 

 be necessary ; but while considering how this could best be 

 done, the idea of another simple system, which did away with 

 all mechanical arrangements, presented itself. It is derived 

 from the first modification of the earlier system (fig. 1, 

 PL VIII.) very easily by suppressing entirely the movable 

 spectroscope-arm and securing the reflecting mirror rigidly 

 to the prism. It will evidently have the same angular motion 

 as before, and will therefore reflect a ray which passes through 

 the prism at minimum deviation in the same direction for all 

 positions of the latter, or vice versa. With this arrangement 

 it is evident that there will be no error due to mechanical 

 inaccuracies of the moving parts because there is no longer 

 any relative motion between them, both prism and mirror 

 being fixed relatively to each other. But it would appear at 

 first sight that for large angles of deviation the light would fall 

 upon a very different part of the reflecting mirror to that which 

 it would for small angles, and that a large lateral displacement 

 of the reflected ray might be expected in consequence. The 

 lateral shifting is no great detriment if a collimator is used, 

 and the beam which passes through the prism is a strictly 

 parallel one, for then its effect is simply to necessitate the 

 loss of some light, or else the use of larger lenses to provide 

 for this shifting. But where only one objective is used, as in 

 the simpler forms of instruments, or for some special reason 

 in the better onesf (see fig. 7, PL IX.), and the beam 

 through the prism is in consequence conical, the lateral dis- 

 placement gives rise, as already explained, to an angular 

 error depending on the radius of curvature of the wave-front 

 of the incident beam. This error, or, more properly, cor- 

 rection, was especially objectionable in the particular work 

 then in hand, in which it was very desirable to maintain an 

 exact relationship between the angular rending of the spectro- 

 meter-circle (in which prism and mirror were mounted), and 

 the wave-length of the radiation falling on what corresponded 



* Kempe, ' Lecture on Linkages/ p. 40. 



t In this case it was desirable in order that there might be as little 

 loss by reflexion and absorption as possible. 



