April 15, 1921] 



SCIENCE 



369 



data^- ^' ' yielded more consistent results and 

 results in agreement witli theory. Using the 

 lines near the head of the CN" 3883 band, the 

 shift of solar wave-lengths, compared to ter- 

 restrial, should be 0.0082 A to the red, equiv- 

 alent to the Doppler effect of a descending 

 current on the sun of 0.634 km./sec. 



It appears to the author that a spectral line 

 must rigidly fulfill three conditions, in order 

 to be suitable for use in this work. (1) It 

 must show no pressure shift, pole-effect, or 

 other variation of frequency with physical 

 condition of the source (excluding gravita- 

 tional effects), (2) it must be a single, sharp, 

 symmetrical line, (3) it must, in the solar 

 spectrum, be quite free from other " foreign " 

 lines. 



Band lines are used because they seem to 

 fulfill condition (1). In the early work 

 proper attention was not paid to condition 

 (3). Grebe and Bachem,"* by obtaining micro- 

 photometric curves, have attempted to rigidly 

 satisfy condition (3), and in doing so have 

 had to discard all but eleven of the 36 lines 

 formerly measured by them. But no investi- 

 gators have made an attempt to rigorously 

 satisfy condition (2). Now the author, in 

 arriving at a new formula for band series,^ 

 obtained very fine spectrograms of the 3883 

 band, and made an extended investigation of 

 its structure, supplementing the work of 

 Uhler and Patterson." There are a number 

 of different series in this band (twenty in 

 all, commonly classified as ten series of 

 doublets), but without exception the indi- 

 vidual members of the various twenty series 

 are sharp, symmetrical lines. This is a note- 

 worthy characteristic of most band series, 

 differentiating them from line series, the 

 members of which are all complex, according 

 to the Bohr-Sommerfeld theory. Moreover 

 these latter are quite susceptible to changing 

 electrical conditions. 



But the ten doublet series of the 3883 band 



5 Zeit. f. Phys., 1, 51, 1920. 



6 Zeit. f. Phys., 2, 415, 1920. 

 ^ Phys. Zeit., 21, 662, 1920. 



s Astro. Jour., 46, 85, 1917; Phys. Bev., 11, 136, 

 1918; 13, 360, 1919. 



9 Astro. Jour., 42, 434, 1915. 



have different spacing and so are continually 

 crossing, resulting very frequently in an ap- 

 parent broad, unsymmetrical line, even with 

 the best resolving power at our disposal. But 

 this complex is really only the superposition, 

 or partial superposition of two or more 

 sharp, symmetrical lines. It is self-evident 

 that the apparent center of gravity of such a 

 complex depends on the length of exposure, 

 etc., while the position of the " peak " of a 

 micro-photometric curve depends on the rela- 

 tive intensity and position of the component 

 members of the complex. It is known that 

 the relative intensity of certain series in the 

 3883 band changes with physical conditions, 

 and there is evidence that their relative 

 intensity in the sun is different from that in 

 the ordinary carbon are. Hence any apparent 

 " line " which is really a complex is entirely 

 unsuitable for the detection and measurement 

 of so small a shift as that predicted by Ein- 

 stein. This is especially true as the solar 

 lines are in absorption, while the arc lines 

 are in emission. 



The author, in his analysis, has identified 

 many series lines, not previously identified, 

 and by obtaining accurate formulas for the 

 stronger series, has been able to compute 

 " theoretical " positions for all lines of these 

 series, including those entering into com- 

 plexes. In all cases tested, the actual appear- 

 ance of the complex was in agreement with 

 the theoretical structure thus built up. Also, 

 many complexes have been recognized which 

 may not previously have been suspected as 

 such, and the presence of several extraneous 

 lines in the normal arc spectrum (carbon 

 lines, but not band lines) has been detected. 

 Thus material is at hand for a rigid investi- 

 gation of the eleven lines finally used by 

 Grebe and Bachem. The details of this work 

 will be published elsewhere. Only the results 

 are given here. 



Of the eleven lines only two (A 3873.504 

 and A 3858.822, on the Eowland system) fully 

 satisfy condition (2). Even this is not 

 strictly correct, for the two lines are unre- 

 solved doublets, the 31st and 49th member, 

 respectively, of the A^ series. But the two 



