406 H. Draper—Dvffraction Spectrum Photography. 
then found to have cast its shadow on H, of the 8d order, and 
the point at 5, of the 2d order had impressed itself near O of 
the 3d order. 
By a simple calculation it was thus rendered evident that a 
given ray in the compound line H, was of the wave-length 
3930-71 ten millionths of a millimeter, and that another line 
near O had the wave-length 3444°6. By looking at the photo- 
graph, the reader will see that 3930 falls upon a fine division 
in H,, which is beautifully shown in both the spectrum with 
the scale and the enlarged proof below. Of course, the ray @ 
of the 3d order, the wave-length of which is known, had im- 
pressed itself photographically on the collodion. 
aving thus ascertained the wave-lengths of three fixed 
points in the photograph, the next step was to apply a scale 
reading to a single ten-millionth of a millimeter, and, if possi- 
ble, fractions thereof. After many abortive attempts to use that 
part of Angstrém’s map which lies between G and H, and to 
attach thereto an additional length of scale sufficient to extend 
to the end of the ultra-violet region, I was compelled to resort 
to a linear dividing engine, and rule a scale which was about 
twice the length of the photographic reduction shown 11 the 
accompanying plate. Of course, this necessitated drawing m 
by hand the same systems of lines and lettering as are shown 
on Angstrém’s chart, and this I did as carefully and faithfully 
as I could. 
_ It only remained to reduce this divided scale to the proper 
size to fit the spectrum photograph; after many trials 1t was 
d 
it was found that the third fixed point was not attained, a 
that there was an error of about two divisions. But if a 
ncor- 
or if 
this small error should be partly attributed to D,, and partly 
, my scale would be correct. Future measures of me 
wave-lengths of these rays, and of 6,, can alone settle this deli 
cate point, for the determinations of Mascart and Angstrom th 
Thalen differ nearly to the extent mentioned above. © ‘ 
same remark is true of Angstrém compared with Ditscheiners 
while the difference between Angstrém and van der vie 
is more than three times the amount necessary to remove sh 
discrepancy. In any case the photograph is correct, as 1t 18 
