SEPTEMBER If, 1902] 
or solid air, if itdid occur, would necessarily be of a very tenuous 
description. In any case, the dense gases tend to accumulate in 
the lower strata, and the lighter ones to predominate at the higher 
altitudes, always assuming that a steady state of equilibrium has 
been reached. It must be observed, however, that a sample of 
air taken at an elevation of nine miles has shown no difference in 
composition from that at the ground, whereas, according to our 
hypothesis, the oxygen ought to have been diminished to 17 per 
cent., and the carbonic acid should also have become much less. 
This can only be explained by assuming thata large intermixture 
of different layers of the atmosphere is still taking place at this 
elevation. This is confirmed by a study of the motions of 
clouds about six miles high, which reveals an average velocity 
of the air currents of some seventy miles an hour ; such violent 
winds must be the means of causing the intermingling of 
different atmospheric strata. Some clouds, however, during 
hot and thundery weather, have been seen to reach an elevation 
of seventeen miles, so that we have direct proof that on occasion 
the lower layers of atmosphere are carried to a great elevation. 
The existence of an atmosphere at more than a hundred miles | 
above the surface of the earth is revealed to us by the appear- | 
ance of meteors and fireballs, and when we can take photo- | 
| krypton or xenon, should make their appearance in the spec- 
graphs of the spectrum of such apparitions we shall learn a 
great deal about the composition of the upper air. In the 
meantime Pickering’s solitary spectrum of a meteor reveals an 
atmosphere of hydrogen and helium, and so far this is corrobora- 
tive of the doctrine we have been discussing. It has long been 
recognised that the aurora is the result of electric discharges 
within the limits of the earth’s atmosphere, but it was difficult 
to understand why its spectrum should be so entirely different 
from anything which could be produced artificially by electric 
discharges through rarefied air at the surface of the earth. 
Writing in 1879, Rand Capron, after collecting all the recorded 
observations, was able to enumerate no more than nine auroral 
rays, of which but one could with any probability be identified 
with rays emitted by atmospheric air under an electric 
discharge. Vogel attributed this want of agreement between 
nature and experiment, in a vague way, to difference of tempera- 
ture and pressure ; and Zollner thought the auroral spectrum to 
be one of a different order, in the sense in which the line and 
band spectra of nitrogen are said to be of different orders. 
Such statements were merely confessions of ignorance. But 
since that time observations of the spectra of auroras have been 
greatly multiplied, chiefly through the Swedish and Danish 
Polar Expeditions, and the length of spectrum recorded on the 
ultra-violet side has been greatly extended by the use of photo- 
graphy, so that, in a recent discussion of the results, M. Henri 
Stassano is able to enumerate upwards of one hundred auroral 
rays, of which the wave-length is more or less approximately 
known, some of them far in the ultra-violet. Of this large 
number of rays he is able to identify, within the probable limits 
of errors of observation, about two-thirds as rays, which Prof. 
Liveing and myself have observed to be emitted by the most 
volatile gases of atmospheric air unliquefiable at the tem- 
perature of liquid hydrogen. Most of the remainder he 
ascribes to argon, and some he might, with more probability, 
have identified with krypton or xenon rays, if he had been 
aware of the publication of wave-lengths of the spectra of 
those gases, and the identification of one of the highest rays 
of krypton with that most characteristic of auroras. The rosy 
tint often seen in auroras, particularly in the streamers, 
appears to be due mainly to neon, of which the spectrum is 
remarkably rich in red and orange rays. One or two neon rays 
are amongst those most frequenuy observed, while the red ray of 
hydrogen and one red ray of krypton have been noticed only 
once. The predominance of neon is not surprising, seeing that 
from its relatively greater proportion in air and its low density 
it must tend to concentrate at higher elevations. So large a 
number cf probable identifications warrants the belief that we 
may yet be able to reproduce in our laboratories the auroral 
spectrum in its. entirety. It is true that we have still to account 
for the appearance of some, and the absence of other, rays of 
the newly discovered gases, which in the way in which we 
stimulate them appear to be equally brilliant, and for the 
absence, with one doubtful exception, of all the rays of nitrogen. 
If we cannot give the reason of this, it is because we do not 
know the mechanism of luminescence—nor even whether the 
particles which carry the electricity are themselves luminous, or 
whether they only produce stresses causing other particles which 
NO. 1715, VOL. 66] 
NATURE 
475 
encounter them to vibrate ; yet we are certain that an electric 
discharge in a highly rarefied mixture of gases lights one element 
and not another, in a way which, to our ignorance, seems 
capricious. The Swedish North Polar Expedition concluded 
from a great number of trigonometrical measurements that the 
average above the ground of the base of the aurora was fifty kilo- 
metres (thirty-four miles) at Cape Thorsden, Spitzbergen ; at this 
height the pressure of the nitrogen of the atmosphere would be 
only about one-tenth ofa millimetre, and Moissan and Deslandres 
have found that in atmospheric air at pressures less than one 
millimetre the rays of nitrogen and oxygen fade and are 
replaced by those of argon and by five new rays which Stdssano 
identifies with rays of the more volatile gases measured by us. 
Also Collie and Ramsay’s observations on the distance to which 
electrical discharges of equal potential traverse different gases 
explosively throw much light on the question ; for they find 
that, while for helium and neon this distance is from 250 to 300 
mm., for argon it is 454 mm., for hydrogen it is 39 mm., and 
for air and oxygen still less. This indicates that a good deal 
depends on the very constitution of the gases themselves, and 
certainly helps us to understand why neon and argon, which 
exist in the atmosphere in larger proportions than helium, 
trum of auroras almost to the exclusion of nitrogen and oxygen. 
How much depends, not only on the constitution and it may be 
temperature of the gases, but also on the character of the electric 
discharge, is evident from the difference between the spectra at 
the cathode and anode in different gases, notably in nitrogen and 
argon, and not less remarkably in the more volatile compounds 
of the atmosphere. Paulsen thinks the auroral spectrum wholly 
due to cathodic rays. Without stopping to discuss that question, 
it is certain that changes in the character of the electric dis- 
charge produce definite changes in the spectra excited by them. 
It has long been known that in many spectra the rays 
which are inconspicuous with an uncondensed electric dis- 
charge become very pronounced when a Leyden jar is in 
the circuit. This used to be ascribed to a higher tempera- 
ture in this condensed spark, though measurements of that 
temperature have not borne out the explanation. Schuster and 
Hemsalech have shown that these changes of spectra are in part 
due to the oscillatory character of the condenser discharge which 
may be enhanced by self-induction, and the corresponding 
change of spectrum thereby made more pronounced. Lightning 
we should expect to resemble condensed discharge much more 
than aurora, but this is not borne out by the spectrum. 
Pickering’s recent analysis of the spectrum of a’ flash obtained 
by photography shows, out of nineteen lines measured by him, 
only two which can be assigned with probability to nitrogen 
and oxygen, while three hydrogen rays most likely due to water 
are very conspicuous, and eleven may be reasonably ascribed to 
argon, krypton and xenon, one to more volatile gas of the neon 
class, and the brightest ray of all is but a very little less re- 
frangible than the characteristic auroral ray, and coincides with 
a strong ray of calcium, but also lies between, and close to, an 
argon and a neon ray, neither of them weak rays. There 
| may be some doubt about the identification of the spectral 
rays of auroras because of the wide limits of the probable 
errors in measuring wave-lengths so faint as most of them 
are, but there is no such doubt about the wave-lengths of the 
rays in solar protuberances measured by Deslandres and Hale. 
Stassano found that these rays, forty-four in number, lying be- 
tween the Fraunhofer line F and 3148 in the ultra-violet, agree 
very closely with rays which Prof. Liveing and myself measured 
in the spectra of the most volatile atmospheric gases. 
It will be remembered that one of the earliest suggestions 
as to the nature of solar prominences was that they were solar 
auroras. This supposition helped to explain the marvellous 
rapidity of their changes, and the apparent suspension of bril- 
| liant self-luminous clouds at enormous heights above the sun’s 
| surface. 
Now the identification of the rays of their spectra 
with those of the most volatile gases, which also furnish many 
of the auroral rays, certainly supports that suggestion, A 
stronger support, however, seems to be given to it by the 
results obtained at the total eclipse of May, 1901, by the 
American expedition to Sumatra. In the Astrophysical Journal 
for June last is a list of 339 lines in the spectrum of the corona 
photographed by Humphreys, during totality, with a very large 
concave grating. Of these no fewer than 209 do not differ from 
lines we have measured in the most volatile gases of the atmo- 
