SOLAR ECLIPSE, MAY 0, ISSIi. 



105 



Fig. 2 



The essential feature of the spectroscope, however, and that whicli adapted it to its particular 

 pnrpose, was an arranoenient of two total reflecting prisms just in front of tlie collimator slit. The 

 oflice of these prisms was to form two virtual images, each being that of one-half the slit, separated 

 at their nearest ends by an interval equivalent to the diameter of the solar image in the equatorial 

 used. The construction will be rendered evident by the diagram (Fig 2), which represents a section 

 of the apparatus by a plane passing through the slit and the line of colliniation. The dimensions 

 were sochosen that if light from the right side of the solar image were transmitted, after two reflections 

 by the prism «' i', to the central point of the slit S, then light from the left edge of the solar image 

 would be transmitted to S by the prism a b. It is evident that 

 light from any other portion of the sun would not fall upon the lens 

 of tlie collimator, and consequently not appear in the speetroscoi>e, 

 but that light from regions to the right and left of the sun, up to a 

 distance determined by the length (in this case 1.5') of the slit or the 

 dimensions of the prism, would so appear; moreover, it is obvious 

 that the spectra formed of the two halves of the slit would be adja- 

 cent. In fact, these two spectra, one that of the region to the right 

 of the sun and the other that of the region to the left, were separated by a narrow black line 

 about as distinct as the G line of the spectrum.. Although all angles of reflection in the prisms 

 were greater than the critical angle, the sides a, b, a', and b' were silvered, to avoid trouble in 

 keeping them clean. 



It will be recognized that this spectroscope, which was specially devised for the work in view, 

 was most carefully considered in all points so as to be of the highest efficiency, i^or do I see now, 

 after Iiaving used it, how it could be bettered if the same large scale of 1 .6 inches available aperture 

 were not to be surpassed. 



OBSBEVATIONS. 



The observations made with all the instruments described, except the last, are given in the in- 

 dividual reports of the observers. Here I shall give my own only, though I shall have occasion to 

 discuss some of the others. 



A few minutes before the first contact the spectroscope slit was adjusted for parallelism with 

 the utmost care, since the essence of the method was the comparison of two spectra formed by 

 different halves of the slit. xVn ideally more perfect way of eliminating errors due to the slit would 

 be to rotate the spectroscope by ISOo on the axis of collimation of the equatorial at each observation, 

 but as I found no difficulty in so adjusting that it was quite imi>ossible to distinguish a difference 

 in the twohalves of the spectrum with any useful slit- width, I definitively abandoned the inconvenient 

 reversals. After this adjustment the slit was given the maximum width compatible with sharp 

 definition ; that is, the spectrum was made as bright as possible under the limitation that the line D 

 should be just recognizable as double. The group b was of course under these conditions, triple. 

 It was expected that it would be unnecessary to alter this adjustment during the eclipse, and it was 

 so found. 



A short time before second contact the spectroscope was rotated on the axis of its collimator 

 until the direction of its slit became sensibly normal to the edge of the solar image at the point 

 of second tangency. As I was engaged in this way I noticed on the white card-board which cov- 

 ered the slit-plate a sudden alteration in the aspect of the image by the growth of arcs of light 

 from the cusps of the sun, which quickly joined and formed a complete ring. The phenomenon 

 occurred, as I found from Mr. Fletcher's counting,Jl seconds before totalit\% It is only of scieu. 

 S. Mis. 1 10 14 



