98 H. A. Rowland—Concave Gratings for Optical Purposes. 
spectrum. Again* it is found impossible to obtain interference 
‘between two rays whose paths differ by much more than 
50,000 wave-lengths: 
All the methods of determining the limits seem to point to 
about the 150,000th of the wave-length as the smallest distance 
at which the two lines can be separated in the solar spectrum 
by even a spectroscope of infinite power. As we can now 
nearly approach this limit I am strongly of the opinion that we 
have nearly reached the limit of resolving power, and that we 
can never hope to see very many more lines in the spectrum 
than can be seen at, present, either by means of prisms or 
gratings. 
It is not to be supposed, however, that the average wave- 
length of the line is not more definite than this, for we can 
easily point the cross hairs to the center of the line to perhaps 
1 in 1,000,000 of the wave-length. The most exact method of 
detecting the coincidences of a line of a metal with one in the 
solar spectrum would thus be to take micrometric measure- 
ments first on one and then on the other; but I suppose it 
would take several readings to make the determination to 1 in 
1,000,000. 
Since writing the above I have greatly improved my appa- 
ratus and can now photograph 150 lines between the Hl and K 
lines, including many whose wave-length does not differ more 
than 1 in about 80,000. I have also photographed the 1474 
and b, and 6,, widely double, and also E just perceptibly 
double. With the eye much more can be seen, but I must say 
that I have not yet seen many signs of reaching a limit. The 
lines yet appear as fine and sharp as with a lower power. If 
my grating is assumed to be perfect, in the third spectrum | 
should be able to divide lines whose wave-lengths differed, in 
about 150,000, though not to photograph them. 
The E line has components, about 45},5th of the wave-length 
apart. I believe I can resolve lines much closer than this, say 
1 in 100,000 at least. Hence the idea of a limit has not yet 
been proved. 
However as some of the lines of the spectrum are much wider 
than others we should not expect any definite limit, but a grad- 
ual falling off as we increase our power. At first, in the short 
wave-lengths at least, the number of lines is nearly proportional 
to the resolving power, but this law should fail as we ap- 
proached the limit. 
* This method of determining the limit has been suggested to me by Prof. Cc. 
S. Hastings, of this University 
- 
