352 On the Lower Limit of tlie Prismatic Spectrum, 



edges of the hole. In experiments such as the present, ^^ hero 

 intense spectra are required, it is often better to dispense with 

 a slit ; but it then becomes more important than ever to got 

 rid of the apparent motion of the sun. 



For fixed points of reference I followed Herschel in using 

 the nearly definite rays insulated by cobalt glass, of which the 

 principal are the extreme red, and a ray in the yellow nearly 

 halfway from D towards E *. After the spectrum had im- 

 pressed itself the glass was held to the eye, and the positions 

 of the centres of the red and yellow images of the sun 

 marked with a fine pricker or pencil. 



When the light is intense, the thermal spectrum begins to 

 show itself ons moked paper in ten or twenty seconds ; and the 

 experiment is one of great beauty, which I desire to recom- 

 mend to those who have access to a clear sun and are in pos- 

 session of the necessary (very simple) apparatus. The second 

 method, with the iron salts, has the advantage of giving per- 

 manent results, which can be examined at leisure. Both me- 

 thods agreed in giving thermal spectra ending on the less- 

 refrangible side in a more or less detached spot, whose posi- 

 tion, however, does not coincide with any of Herschel's. On 

 the contrary, its centre falls almost exactly in the gap between 

 Herschel's $ and 7. The gap by which the final spot in my 

 experiments is separated from the remainder of the spectrum 

 is almost exactly as far below the centre of the red cobalt 

 image of the sun as that is below the centre of the yellow 

 image, coinciding therefore with the centre of Herschel's spot 

 p. The most careful attention, in combination with frequent 

 repetition of the experiment, has failed to give me any indica- 

 tion of still less-refrangible rays. In order to vary the condi- 

 tions I have used a glass prism of 60°, a rock-salt prism and 

 lens (in conjunction with the silver-on-glass mirror), and, finally, 

 a large bisulphide-of-carbon prism of 45° (made for me by 

 Ladd) which gives a beam nearly 16 square inches in area, 

 but without success. 



At the same time I cannot pretend to feel entire confidence 

 in these results. Herschel was an experimenter of great skill 

 and experience, whose carefully recorded conclusions cannot 

 be put aside without explanation. It has occurred to me as 

 barely possible that a source of error lay in light reflected in- 

 ternally at the base of Herschel's prisms. It is known that 

 the regular spectrum of an isosceles prism v/ith a polished 

 base is accompanied by a white image, due to light three 



* A certain thidmess of tlie absorbing medium is necessary. The piece 

 I use is sufficiently wedge-sliaped to allow me to choose the most suitable 

 thiclvuess, which varies to some extent with the liglit. 



