103 
92, 93] Atomic Weight of Stellar Matter 
generation unaltered, we should, as we have seen, halve the coefficient of 
opacity of the star. This would necessitate a change in the star’s build : in 
actual fact its radius would increase fourfold while its surface temperature 
would be halved. We could follow the progress of the demon’s work by 
watching the changes in either the radius or the surface-temperature of the 
star. Hence from the observed surface-temperature or the calculated radius 
of any star the mass and luminosity of which are known, it must be possible 
to estimate the atomic weight of the atoms of which the star is composed. 
Formula (93T) provides the means. 
Table XIII gives the values of N-/A calculated from this formula. The 
third column contains the bolometric luminosity in terms of the sun’s bolo- 
metric luminosity as unity, and the fourth contains the value of G measured 
in ergs per gramme per second. 
Table XIII. Atomic Weights of Stellar Matter. 
(Calculated on the supposition that the gas-laws are obeyed.) 
Star 
Mass 
(0 = 1) 
Bolom. 
Lum. (Q =1) 
G 
K 
T 
M c 
N 2 
A 
V Puppis 
19-2 
11,000 
1,100 
0-67 
160x10° 
190 
u Herculis 
7'6 
1,250 
300 
1-20 
100x10° 
210 
Sirius A 
2-45 
38 
29 
3-53 
80 x 10« 
350 
Sun 
100 
1-00 
1-90 
12-2 
55 x 10° 
440 
60 Kruger A 
0-25 
001 
0 06 
150 
42 x 10° 
100 
„ B 
0-20 
0 003 
0-021 
230 
65 x 10° 
130 
Betelgeux 
[40] 
6,000 
T3001 
0-42 
10 x 10° 
320 
Capella A 
4-18 
105 
48 
2-08 
19x10° 
250 
a Cent. B 
0-97 
0-46 
0-90 
12-9 
45 x 10° 
750 
The values of N 2 /A in the last column cannot lay claim to any high 
degree of accuracy. There are two principal sources of error, namely the value 
assumed for G and the value assumed for A. c in equation (93'1). 
In the case of stars of very high or very low temperature, the value of 
G cannot be determined at all exactly. It is determined from the bolometric 
absolute magnitude of the star, and even if the visual absolute magnitude is 
known, the bolometric correction is generally very uncertain. An error of only 
half a magnitude in the bolometric correction would throw the value of N' 2 /A 
into error by a factor of T58. 
This source of error almost vitiates the entries for V Puppis, u Herculis, 
60 Kruger and a Orionis, but hardly affects those for such stars as the Sun 
and Capella, for which the bolometric correction is small, and the absolute 
emission of radiation is well known. But the entries for these stars are affected 
by errors in the value of \ c . For stars of the mass of our sun, equation (911) 
shews that X c varies inversely as /a -4 so that \ C (X C -1- 1) varies approximately as 
/a s . As the value of T c also varies approximately as g (equation (90 2)) it