136 ANNUAL REPORT SMITHSONIAN INSTITUTION, 1937 



will be the presence of a significant proportion of hydrogen. The 

 other way is to try various combinations of hydrogen and not-hydrogen 

 until we find the proportion which gives precise agreement of the cal- 

 culated and observed brightness. That is the method we generally 

 employ nowadays; the observed brightness of a star tells us what 

 proportion of its mass consists of hydrogen. 



Dr. Bengt Stromgren found in this way that the sun, Capella, and 

 other typical stars contain 33 percent of hydrogen. My own calcula- 

 tions agreed precisely. This agreement is rather specially interesting 

 because we adopted different composition for the rcmammg 67 per- 

 cent of the mass. Stromgren used a mixture of rather light elements, 

 familiarly known as "llussell's Mixture," believed to agree with the 

 composition of the outer layers determined with the spectroscope; I 

 used a mixture about three times heavier. Our precise agreement 

 confirms what I have already said — that it makes no difference what 

 land of stellar material you assume, so long as it is not-hydrogen. It is 

 still doubtful to what extent the proportion of hydrogen varies in 

 different stars; there is some evidence that it is greater in the most 

 massive stars, but the evidence is not very good. An important paper 

 ])resented by Prof. JI. N. Kussell to the Tercentenary Conference was 

 partly devoted to this question. 



I must say a word about the agreement of theory and observation. 

 Since we determine the proportion of hydrogen so as to make the 

 observed and calculated brightness agree, we obviously cannot claim 

 that the agreement is a confirmation of the theory. Nevertheless it 

 does furnish a fauly efficient check. Unless the theory were pretty 

 near the truth, we should find that for some of the stars which we 

 try it would be impossible to find any proportion of hydrogen that 

 would bring about agreement. It is satisfactory, therefore, that all 

 the stars give a reasonable proportion. If Strujugren had found, 

 instead of 33 percent, an answer which involved the square root of 

 — 1, as might easily have happened, we should have concluded that 

 there was something fishy about the theory. 



The recognition of white dwarf stars with density far transcending 

 that of any terrestrial matter is one of the more spectacular develop- 

 ments of the study of stellar constitution. A cubic inch of the mat- 

 ter of the companion of Sirius weighs about a ton, and some of the 

 more recently discovered white dwarfs appear to have higher den- 

 sities even than that. In order to explain a new point which has 

 arisen in connection with the theory of these stars, I must go back 

 to past history. In 1924 the mass-luminosity relation — that is, the 

 fonnula expressing the result of the calculation I have been describ- 

 ing — was worked out; and, on comparing with observation, it turned 

 out that it was obeyed not only by the diffuse giant stars for which 

 it was intended but also by the dwarf stars with densities greater 



