corresponding to certain Points in the Solar Spectrum. 489 
numbers to any determinate law. The only relation hitherto 
recognized is one subsisting in the second set, in which it is 
found that, if the reciprocals of the numbets be compared, the 
square of that of F is equal to half the sum of the squares of 
those of B and H, or F;=1(B?+H}), thé corresponding num- 
bers being 
Bey. wiry. 44 == 6107100 
1(B24H2) . =3107068 
Difference . 0000032 
In the first set this relation does not subsist with nearly the 
. same degree of accuracy. 
There is, however, a relation of more importance which may 
be established by a very slight alteration upon the values of B 
and D. Although the observed values of these two agree in 
both series, neither of them can be regarded as correct to within 
a ten thousandth part of its magnitude. Now by altering each 
to a smaller amount than this proportion, we may establish the 
remarkable relation B°=D®. Thus the observed value of B° 
(taking seven effective numbers) is 1594052 
And that of D6is 1593958 
Difference only . 0000099 ; 
Dividing this difference, we make each 1594025, which gives 
B=2540844 (log 4049780), differing from the observed value 
by only ‘0000156, and D=2175112 (log 3374816), differing 
from the observed value by only ‘0000112. As these alterations 
are so trifling, and this relation of B°=D® is very convenient, 
there need be no hesitation in adopting it, and regarding the 
above as the settled values of B and D. 
There is a similar close approximation by which the value of 
the wave-length of i may be deduced from those of B and D. 
It is BPD=E". Thus— 
The observed value of E in the first series is . 1943000 
The corrected log B 4049780 x 7 =2°8348460 
Add corrected log D . . . . 138374816 
Divide by ll. . . . 4s 81788276 
Gives forlogE. . . . . . 0:2883934= 1942645 
Difference only . . . . ‘0000855 
This difference being much within the limits of probable errors 
of observation, the above may be regarded as a convenient rela- 
tion by which to connect the value of the wave-length of E with 
those of B and D, and the true logarithm of EK may accordingly 
be assumed as 0:2883934. The nearer approach to this value 
exhibited by the first observed series may be viewed as one ad- 
2G2 
