236 PROCEEDINGS OF THE AMERICAN ACADEMY. 



aggregate and not a single molecule. It is also possible that hy- 

 drogen atoms are associated with the sodium, for the work on the 

 dispersion indicated that there was present always a small trace of 

 some gas other than sodium, which no amount of pumping would 

 remove; that is, it appeared to be tangled up in the sodium vapor, 

 condensing with it in the cooler parts of the tube. All of this is, 

 however, irrelevant, for we are for the present merely engaged with 

 the study of a certain remarkable vibratory mechanism, and for the 

 present need not concern ourselves whether it is a molecule, a cluster 

 of molecules, or a compound molecule. 



We will begin by a description of the various spectra which we shall 

 study and compare in the present paper. The spectrum region with 

 which we are concerned lies between wave-lengths 4600 and 5700, i. e. 

 the region of the green-blue channelled absorption spectrum. 



The Absorption Spectrum. 



Photographed with the twelve-foot concave grating, the absorption 

 spectrum is found to consist of a multitude of fine lines, to the num- 

 ber of about 1500 in the region specified. Its appearance has been 

 found to be profoundly affected by the presence of hydrogen or any 

 inert gas. It is shown on Plate 1, Figure 1, g, photographed in the 

 second order with a twelve-foot concave grating.' In hydrogen at 

 atmospheric pressure its appearance is shown by spectra /and h, the 

 fluted appearance being very marked. In a high vacuum its appearance 

 is shown by spectrum b, the flutings having entirely disappeared. 

 The chief change appears to lie in the increased absorbing power of 

 certain lines or groups of lines. Spectrum d is taken under nearly the 

 same conditions, the sodium vapor being less dense, however. A care- 

 ful study of the effect of the inert gas has not yet been made, and it is 

 mentioned here only on account of its relation with the subject of the 

 paper. Comparatively few of the absorption lines have any relation 

 with the fluorescent and magnetic rotation spectra, the ones concerned, 

 however, being those which are affected by the presence of the inert gas. 



Moreover, both the fluorescence and magnetic rotation disappear, 

 i. e. cannot be excited, when the sodium vapor is formed in an atmos- 

 phere of hydrogen or other inert gas. Of this matter more will be said 

 later on. 



The Magnetic Rotation Spectrum. 



It was found last year that a number of vapors, showing fine and 

 sharp absorption lines, when placed in a powerful magnetic field, rotate 



