A.— MATHEMATICAL AND PHYSICAL SCIENCES. 41 



takes the excitation potential as less than ITS volts. This condition 

 excludes very many possibilities. Indeed, if we are to be bound by the 

 selection rules, it excludes all the possibilities. So, with the example of 

 the nebula? before him, McLennan waives these rules, and assigns the 

 green line to a transition from one or other of the low-lying metastable 

 states which the theory indicates. 



The lowest state of all should be a triplet, and owing to the absence 

 of companions to the green line this may very probably be excluded. 



If so, only one alternative remains, and the successful determination 

 of the Zeeman effect carried out in McLennan's laboratory is in harmony 

 with the choice so arrived at. An independent investigation by L. H. 

 Sommer, published immediately afterwards, covered exactly the same 

 ground, and led him to the same choice. This is satisfactory so far, but 

 the position will be much strengthened and consolidated when we have 

 an independent determination of the levels in question, giving the means 

 of calculating a theoretical wave-length for comparison with that observed. 

 To do this will require a fuller survey of the Schumann region of the arc 

 spectrum than has yet been made. For the aurora line we have the experi- 

 mental production from oxygen but not the numerical spectroscopic 

 relation. For the nebular lines our position is exactly the reverse. 



The origin of the green auroral line has thus been definitely cleared up, 

 at all events in so far that it is attributable to the arc spectrum of oxygen. 

 There are, however, other features of the auroral spectrum which are still 

 obscure. I will limit myself to discussion of one of them — -the red line 

 of the aurora. Red auroras are comparatively rare, and when they do 

 occur the distribution of colour presents very curious features. In some 

 cases the ends of the streamers are tipped with red, while the greater part 

 of the length is green. The only reddish aurora which I have been privi- 

 leged to observe at my home in the south of England (May 14, 1921) was 

 of a different character, the colour ranging rapidly through various shades 

 of purple. The light was distributed in irregular patches high up near the 

 zenith, though predominantly in the north. At the same time its position 

 was highly unstable, and the general impression produced was reminiscent 

 of high potential discharges in highly exhausted vacuum tubes. Vegard 

 has described cases where the whole sky suddenly turned crimson. He 

 has obtained good small-scale spectrograms of the red line, which give the 

 position as X6322, which, however, is subject to a probable error of at 

 least ±1A°. A determination by V. M. Slipher of the Lowell Observatory 

 gave X6320. 



So far as can be judged from the evidence available, no pair of the low- 

 lying levels of the oxygen arc scheme which McLennan has discussed in 

 connection with the aurora are suitably placed to yield this red line by 

 combination. We naturally turn to the consideration of nitrogen spectra, 

 whicji, as is well known, are represented in the blue and violet regions of 

 the auroral spectrum. 



I described in 1922 a spectrum in which one of the first positive bands 

 of nitrogen X6323 was very much intensified relative to the neighbouring 

 red bands, which ordinarily are of comparable brightness. This spectrum 

 was produced by adding a large excess of helium to the nitrogen afterglow, 

 and the source had a visual red colour dominated by this band. In 



