200 ANNUAL REPOKT SMITHSONIAN INSTITUTION, 19 31 



by year. The great outstanding puzzles of the nebular and auroral 

 spectra yielded a few years ago, and the solar corona alone remains 

 uninterpreted. 



The proof that iron (for example) is present in the sun has been 

 for two generations a standard laboratory experiment — which ought, 

 by the way, to be exhibited to the general public in our great science 

 museums whenever the sun shines. The analysis of the sun would 

 therefore appear to be a very simple matter, but, as is usual in all 

 precise work, many complications arise. 



With such strong lines as those of iron the simpler " method of 

 coincidences " is conclusive, and many other elements — hydrogen, 

 calcium, magnesium, sodium, aluminum — can be found at once in 

 the same manner. Evidently, too, in a general way, an abundant 

 element will give strong lines and a rare one weak lines, so that our 

 analysis can be roughly quantitative as well as qualitative. 



But if we wish to make it complete, we must record and measure 

 as accurately as possible all the lines of the solar spectrum and also 

 the emission lines of all the elements. 



The solar part of this great task was first adequately attempted 

 by Rowland. Developing the difficult technique of constructing dif- 

 fraction gratings to a point which has hardly since been equalled and 

 certainly not surpassed, he obtained photographs of the spectrum 

 which showed the faintest lines with a distinctness which is still the 

 envy if not the despair of other workers. His great " Table of solar 

 spectrum wave lengths " covers the whole range from the ultraviolet 

 to the red — as far as the best plates then in existence were sensitive — 

 and includes more than 20,000 lines. Within this region it is sub- 

 stantially complete. The recent revision of Rowland's table, under- 

 taken at Mount Wilson, has added only seven faint lines, not counting 

 those which appear only in sun spots, and so fell outside his purview. 

 The principal reason for the revision (apart from the practical 

 one that the earlier work was out of print) was that Rowland's scale 

 of wave lengths, though a great improvement on anything which pre- 

 ceded it, requires correction on the average by 1 part in 28,000, and 

 also by smaller fluctuati'ng amounts not exceeding one two hundred 

 thousandth of the whole, but too large to be ignored in precise work. 

 The corrected wave lengths are good to one or two parts in a million- 

 Measures of comparable precision have been made in the laboratory 

 for many, though by no means all, emission spectra. It is a danger- 

 ous thing, however, merely to compare a set of laboratory wave 

 lengths, however precise, with the solar tables and conclude the 

 presence of an element because a number of its lines appear. Allow- 

 ance must be made, of course, for the small shift to the red in the sun, 

 which, on the average (though not in all details) agrees with Ein- 



