this region in a direction, making an angle <p with the normal to 
the photosphere, will leave the cell as a diverging beam. The 
spreading is wider in proportion as the refraction constant R, n for 
the kind of light considered is greater. (If R m is negative, the rays 
will first converge, but afterwards diverge just as well). 
With small values of y> and moderate refraction, all the incident 
rays will leave the cell in such directions, that they continue to move 
away from the photosphere. But with larger values of y as well 
as of R m there is increasing chance, that some of the dispersed rays 
are directed back to the photosphere and thus prevented from leaving 
the sun. 
The total quantity of light entering our cell is found by integra¬ 
ting round the normal, and with <p between the limits 0 and 
(It is true that into our cell some light will enter, which comes from 
higher levels, where it has been refracted downward by other cells, 
but that contribution is small in comparison with the supply from 
neath). All this light would be emitted by the celestial body if 
t ere were no incurvation of rays j but now a part returns to the 
p otosphere. That part is greater in proportion as R m has greater 
vpositive or negative) values. 
The cells are not spherical indeed, but it is clear that if other 
s apes are considered, our result must on the average continue to 
o . nd a cell enclosing a maximum of density, makes the rays 
ist converge, then diverge if R m is positive, and makes them always 
diverge if R m i s negative. 
So all the inequalities of the density co-operate to prevent a 
part of the light of the photosphere from passing through the 
In the solar spectrum the waves , which belong to the nearest vid- 
m y °f ihe ^sorption lines , appear with less intensity than the other 
waves, not because they are absorbed in the solar'atmosphere, hut 
oecause they are partly directed back to the interior. The Fraunhofer 
tmes are absorption limes, enveloped in dispersion bands. 
Asymmetry of the Fraunhofer lines. 
Prom onr modified conception of the solar spectrum — which 
ni? supplem ents Kikchhoff’s interpretation — it follows that all 
') The same refraction effect must also cause the brightness of the disk to 
