RECENT PROGRESS IN ASTRONOMICAL RESEARCH. 163 



which he made for the purpose, Rowhmd photographed the spectrum 

 of the sun and that of numerous metals as jiroduced in the electric 

 arc, giving the wave lengths generally to thousandths of an Angstrom 

 unit or to seven places of figures. The i^ublication of these results, 

 including thousands of spectrum lines hitherto unrecorded, together 

 with very numerous statements of the coincidences of solar and 

 metallic lines, was thought at the time to be the last word on the 

 subject of standard wave lengths which would be needed for many 

 years. Rowland, however, did not determine the absolute scale of 

 his system of wave lengths, but based it on the measures of the wave 

 length of the D lines made by Bell and others, and himself observing 

 by the method of coincidences, as above explained, stood sponsor only 

 for the relative accuracy of the measures throughout the spectrum. 



In 1893 Michelson and Benoit, by the aid of the Michelson inter- 

 ferometer, which, though based finally on the same princii3les as the 

 grating, is yet wholly difi'erent in its make-up and use, determined the 

 absolute wave length of the red ray of cadmium in terms of the 

 standard meter. These results proved that Bell's value of the wave 

 length of D, adopted as the basis of Rowland's system, was too large 

 by about tw^o-tenths of an Angstrom unit. A little later it was shown 

 by JcAvell that there is a lack of exact coincidence between solar and 

 metallic spectrum lines, and about the same time it Avas shown by 

 Humphreys and Mohler that pressure played a part in determining 

 the apparent place of lines. About 1900 Kayser discovered that 

 Rowland's tables were not exactly consistent in themselves, for it 

 made a difference, whether the observer measured to one or another 

 of Rowland's lines, what wave length he would assign to some line 

 intermediate between them. 



About this time Michelson proved mathematically that certain 

 errors in the ruling of gratings are possible, which destroy the exact 

 validity of the method of coincidences, so that a line of given wave 

 length in the first order spectrum does not necessarily absolutely 

 coincide with one of half that wave length in the second-order 

 spectrum. Kayser immediately tested this conclusion practically 

 with two large Rowland gratings in his possession, and found dif- 

 ferences of as much as 0.03 Angstrom units traceable to this cause. 



Thus in the course of fifteen years after the publication of 

 Rowland's spectrum maps it was found that the absolute scale of wave 

 lengths he had adopted was in error by 0.2 Angstrom unit and, what 

 is far more serious, there exists a source of error unsuspected till 

 recently, which has introduced inconsistencies of several hundredths of 

 an Angstrom unit in the relative accuracy of Rowland's tables. 



These conclusions are independently verified by comparison of 

 Rowland's numbers with the apparently highly accurate determina- 



