54 Lord Rayleigh's Investigations in Optics. 



effect was sometimes so persistent that I began to think I had 

 evidence of a tendency to separate into distinct layers. When 

 the edges of the prisms (and slit) were horizontal, the tube 

 being also horizontal as before, the loss of definition after 

 shaking still occurred, but could be remedied in great mea- 

 sure by a change of focus. Further experience, however, led 

 me to attribute these effects to temperature-differences, caused 

 perhaps by the ceiling of the laboratory being warmer than 

 the floor. At any rate, they were greatly mitigated by wrap- 

 ping round the tube strips of copper and numerous folds of 

 cloth ; and they could be produced with considerable persist- 

 ency by touching the top of the naked tube for a few seconds 

 with the hands. A difference of even T J^ of a degree Cent, 

 between the upper and lower halves of the prisms might be 

 expected to make itself apparent when the edges are vertical. 

 The advantages of this construction are the elimination of re- 

 flection and the almost absolute immunity from defects due 

 to errors in the separating surfaces ; but they are rather dearly 

 purchased. As might be expected, the best results as to defi- 

 nition are obtained when the tube is vertical ; but such an ar- 

 rangement is inconvenient, as it involves the aid of reflectors. 

 Spectroscopes on this plan may perhaps be useful for special 

 purposes ; but the want of ready adaptability to different parts 

 of the spectrum is a serious objection. 



The general result of this discussion would seem to be in 

 favour of a spectroscope with simple glass prisms of such 

 angle that the reflected light is wholly polarized, the number 

 of prisms being increased up to the point at which mechanical 

 difficulties begin to interfere. With the aid of reflection, at 

 least six prisms may be used twice over. When it becomes 

 necessary to increase the size of the prisms, considerable eco- 

 nomy may be effected by the introduction of half-prisms at the 

 ends of the train, as already explained. 



When the surfaces are not quite flat, or when the focus of 

 the collimator is imperfectly adjusted, it becomes important to 

 secure a rather exact perpendicularity between the edges of 

 the prisms and a plane passing through their middle points ; 

 otherwise the linear image of a point may not be parallel to 

 the length of the slit. Even when the plane of symmetry is 

 obtained, there may remain considerable errors, due to curva- 

 ture of the image, dependent upon the quantity denoted by 7 

 in § 6. Much information as to the cause of bad definition 

 may often be obtained by replacing the slit by a simple hole 

 and examining the character of the resultant image. In many 

 instruments a great improvement is effected by narrowing 

 the vertical aperture whenever the light will bear diminu- 



