. 
S. P. Langley—Determination of Wave-lengths. 187 
uce a constant multiplier n, writing the equation of the inter- 
rh 
- 
rm ! 
I have drawn in this way (on a smaller scale than that of the 
normal or prismatic curves and following the smooth curve in 
the former as my original) four different schemes for the distri- 
bution of the energy. Curve A, fig. 5, represents the distribu- 
tion of solar energy (after absorption by our atmosphere) on the 
normal scale. Curve B, fig. 5, represents the same distribution 
on the scale of wave-frequency (general equation of interpolating 
polating curve y= —, so that the multiplying factor becomes 
n 
curve c= 7? proposed by Mr. Stoney.) Curve OC, fig. 5, repre- 
sents the distribution according-to a proposal (w=log A) of 
Lord Rayleigh. 
Curve D (y=C) is quite different from any of the preceding. 
It gives the distribution on a scale I have never seen proposed, 
but which I have found useful. In this, the bounding curve is 
a straight line parallel to the axis of X. This construction is not 
well suited to exhibit the cold bands, but if we consider only the 
general distribution of the energy we shall find that curve D is 
not merely suggestive as illustrating what has already been re- 
marked here as to the conventional character of the methods of 
showing this distribution, but that it has more practical uses, for 
in this last construction, it is easily seen that the sums of the 
energies between any two wave-lengths whatever, are directly 
proportional to the distance between their ordinates, measured 
on the axis of X. If then we desire (for instance) to know 
what relation the invisible bears to the visible heat, or to 
enquire about what point in the spectrum the energy is equall 
distributed, these and similar problems are solved throug 
curve D by simple inspection. : ; 
I have not been able yet to repeat the preceding determina- 
tions upon the lower part of the spectrum as often as I could 
wish. They are susceptible of improved accuracy by still 
longer experiment, but I think that within the limits of error — 
indicated ee may already be useful. I should add that 
