338 Profs. Liveing and Dewar, On the use of [April 3U, 



direction is as easily attained by means of a lens in front of the 

 slit fitted into a tube which is a prolongation of the telescope 

 tube. When the source of light is so placed that the image 

 produced by the lens is just above the centre of the slit the best 

 results are obtained. 



Prisms may be used as well as gratings with the instrument 

 if the train terminate with a half prism with its face silvered, 

 and we have tried such a plan with success. The dispersion is 

 of course double what it would be with the same prisms used in 

 the ordinary way, as the light passes twice through them. The 

 light is however enfeebled though the dispersion is increased, 

 and the gain in the use of prisms appears to be less than it is 

 in the use of gratings where the only extra cause of loss of light 

 is the reflecting prism which makes but a very trifling difference. 



The advantages which we think we attain by this method are: 



1. Great dispersion without loss of definition. 



The ordinary formula connecting the wave-length, grating 

 space, and inclinations of the collimator and telescope to the normal 

 is a (sin 6 + sin <f>) = nX, where a is the space between successive 

 lines of the grating, 0, cj) the inclinations of the axes of the colli- 

 mator and the telescope respectively to the normal, n the order, 

 and \ the wave-length. 



In ordinary instruments the difference between 6 and <f> 

 must be something considerable, and for the spectra of the 

 higher orders either the incident or the diffracted ray is very 

 oblique and the brightness as well as the definition suffers. For 

 any particular position of the grating with reference to the 

 collimator the angular separation of two rays in the field of 

 view will be measured by d<f>/d\, which is proportional to n sec <£, 

 and the maximum value of n is obtained hy making both 6 and <£ 

 as large as possible. It Avill therefore happen frequently, when 

 both 6 and </> can be made as near 90° as we please, that we are 

 able to observe a particular ray in a spectrum of a higher order 

 than when there must be a considerable difference between them. 

 The definition is also as good as it can be made for- any given 

 grating because the eye-piece ensures the parallelism of the 

 incident rays. As a test we may mention that with Rowlands 

 grating having 14,438 lines to the inch the less refrangible of the 

 E lines is very distinctly divided in the oth order, though we 

 were not able to resolve this line with a Rutherfurd Prating 

 having 17,296 lines to the inch used in the old way, notwith- 

 standing that this latter grating is in some respects the better of 

 the two. The E group in the 5th order bears the magnification 

 of a Ramsden eye-piece, with lenses of one inch focal length, very 

 well. This is the highest power we have used ; and it will be seen 



