210 



CARNEGIE INSTITUTION OF WASHINGTON. 



observations at the polar limb. Emphasis was placed upon the selec- 

 tion of the lines used, with a view to precision of measurement and 

 freedom from blends. The wave-lengths were measured at the sun's 

 center and limb and in the carbon arc in terms of identical iron 

 standards, and their solar wave-lengths were redetermined on the 

 Rowland system. The displacements, sun minus arc, were deter- 

 mined at the center by direct comparison and by three indirect 

 methods; the displacements, limb minus arc, were derived by two 

 methods, with results as shown by the table. 



The observations of greater weight, those upon low-level Unes of 

 intensity 00 to 1, indicate a slight upward movement of the vapor over 

 the center of the disk. A possible explanation from the relativity 

 point of view is that the displacement toward the red has been over- 

 balanced by a Doppler effect; but at the limb, where the Doppler 

 effect of such a motion would be zero, the solar and arc wave-lengths 

 are equal. 



If the means for all lines be considered, the displacement at the 

 center of the disk is zero and at the limb 0.0018 A toward the red. 

 Within the limits of error there is accordingly no evidence in these 

 observations of displacements in the direction of longer wave-lengths, 

 either at center or at limb, of the order of the 0.008 A required by 

 the equivalence principle of relativity as developed by Einstein. 



ERRORS IN THE MEASUREMENT OF CLOSELY GROUPED LINES. 



A further investigation of the errors incidental to the measurement 

 of pairs of closely adjacent hues is being carried on by Mr. St. John 

 with the aid of artificial complexes in which the actual separation and 

 relative intensities of the components are known. This promises to 

 yield results valuable for solar work, as the appearance of combinations 

 of solar lines can be simulated with a high degree of perfection. Pre- 

 liminary data show that for a doublet in which the maximum of one 

 component coincides with the first minimum of the other (the usually 

 accepted limit of spectrographic resolution), the measured separation 

 is 0.430 mm., while the actual separation is 0.380 mm. The combina- 

 tion reproduces approximately the appearance and scale of a solar pair 

 in a fifth-order spectrogram of the 30-foot spectrograph separated 

 0.1 A. The error represents an over-separation of 0.013 A, a confirma- 

 tion of the previous determination of the errors in the Rowland table 



