258 ANNUAL, REPORT SMITHSONIAN INSTITUTION, 1911. 



questions were discussed and plans for future work outlined. The 

 work of the union is carried on by several committees, winch report 

 at the triennial meetings. 



Probably the most important action taken was the adoption of a 

 new system of wave lengths of light. The system in use for the 

 last 20 years was introduced by Rowland, the values being obtained 

 from measurements of spectra made with concave gratings. This 

 system was far in advance of previous ones and was for a long time 

 considered practically perfect. More recent investigations have 

 shown, however, that not only were his absolute values in error, 

 every wave length being too great by about 1 part in 30,000, which 

 is not a matter of much moment, but that also — a much more serious 

 question — there were relative errors of the order of about 1 part in 

 100,000 among the different lines. These errors, due to unknown 

 defects in the gratings, were only discovered when new measure- 

 ments were made by a different method, that of interference. The 

 new primary standard was first determined by ]\Lichelson in 1892 by 

 actually counting the number of waves of the red line in the spec- 

 trum of cadmium in a known fractional part of the standard meter. 

 He found that there were 1,553,163.5 waves in a meter, equivalent 

 to a wave length 0.00064384722 mm., or, as it is usually written^ 

 6438.4722A. This value has more recently been confirmed by Fabry 

 and Perot and is accepted as the primary standard of the new sys- 

 tem. Secondary standards are composed of the wave lengths of 

 50 lines, in the arc spectrum of iron between ^4282 and A6495, which 

 have been independently measured by interference methods by three 

 observers — Fabry and Buisson at Marseille, Eversheim at Bonn, and 

 Pfund at Baltimore. The accordance of these measures is so good 

 that the range is generally less than one part in a million, and the 

 mean of the three is certainly correct, considerably within that 

 margin of error. From these secondary standards tertiary standards 

 are to be obtained by interpolation from grating spectra, and after 

 these tertiary standards have been obtained new measures of the 

 wave lengths of all lines in solar and terrestrial spectra will be 

 required. 



The importance of this work in solar and stellar investigations 

 can not be overestimated, as many important results depend on the 

 accuracy of wave-length values, and incorrect values may lead to 

 erroneous conclusions. Tins is an instance of what I previously 

 said of the necessary interrelation of astronomy and physics and 

 the impossibility of successfully attacking modern astronomical prob- 

 lems without the aid of the allied sciences. 



One of the important conclusions reached by the committee on 

 sun spots was the practically unchanging character of sun-spot 

 spectra. To this may be added the fact, conclusively proved by 



