MAPPING THE LEAST REFRANGIBLE END OF THE SOLAR SPECTRUM. <565 
due regard to my own experiments, and own to meeting serious difficulties. It can 
hardly be possible to doubt the accuracy of the measures made by Herschel, but it 
must be remembered that the length of spectrum with which he worked was very 
small, 4 inches being the extreme limit of the thermograph, whilst not 2 inches was 
the length of the visible spectrum. It must also be recollected that he fixed the 
position of the various parts of his thermograph by a reference to the absorption 
spectrum of a solar image through cobalt glass, using the centre of the yellow solar 
image so obtained as a starting point for his measurements. It is to this that I wish 
to draw special attention as being the cause of the probable difficulty in recognising 
the breaks in the continuity of the heating effect of solar radiation when examined by 
the aid of a prism. 
We may assume that the upper limit of visibility of the spectrum lies somewhere 
near H, and if, as is shown in the figure (Plate 32, fig. 2), we deduct the semi¬ 
diameter of the sun from the place shown as the limit of visibility, we have a very 
probable position for the H line. Now in the plate accompanying his paper (Philo¬ 
sophical Transactions, 1840) Herschel shows a photographic spectrum taken on what 
he calls bromuretted paper. Having experimented on similar paper, I found that the 
lowest part of the spectrum reached by his method of working was near X (Plate 30, 
fig. 2). 
Again, a measurement of the position of the lithium and sodium lines in regard to 
the position of the red images of the sun seen through cobalt glass (Plate 32, fig. 4) 
places the centre of it near the line B. 
Taking H, B, and X on Herschel’s scale of lengths as fixed by this discussion, and 
setting up as ordinates —, we find that the lines joining them lie nearly in a straight 
line, but that the fiducial line through the centre of the yellow solar image (Plate 32, 
fig. 4)* lies nearer to D than it should do. This point should be nearly half way 
between D and E, but slightly near D. The difference between the position given it 
by this graphic way, and that it should occupy according to Herschel’s drawing, is yy 
of an inch—a length which is certainly very small. If the bands and lines shown in 
the photograph are inserted by means of their ordinates, it will be seen that the 
breaks of continuity in the thermograph agree fairly well with the absorption as 
shown in the photograph on the same plate. I have shown in chain dotted lines the 
approximate positions that H, E, D, B, A, X, r, <£, and \p would occupy when the position 
of the spectrum is fixed by reference to images of the sun seen through the cobalt glass 
as given in the diagram. It will be noticed that there is still an agreement in the main 
between the thermograph and the photograph, but (first) that the upper limit of visi¬ 
bility is a long way in the ultra-violet, and (second) that § and e of the thermograph 
are beyond the theoretical limit of the spectrum. This last can scarcely be the proper 
* Eigs. 3, 4, and 5 are taken from Sir J. Herschel’s plate in the Phil. Trans., 1840. 
