ON THE SPECTRUM OF HYDROGEN. 
385 
of a rapid disappearance of the gas. In the case of mercury an explanation based on 
some action of the walls of the capillary is even less satisfactory, as it would he necessary 
to assume that the glass walls were an almost inexhaustible source of mercury which 
was rapidly absorbed by the electrodes ; for in this case also there is no evidence of 
any approach to a steady state. 
It has not been possible to make any quantitative comparison of the rates at which 
the Hydrogen lines spread into, and the mercury lines out of, the capillary ; but the 
speed of the Hydrogen lines relative to those of mercury suggests forcibly that the gases 
actually move in opposite directions, and that the Hydrogen and mercury do not appear 
at the centre and the ends of the capillary respectively immediately after the condenser 
is cut out, for the -simple reason that they are not there. In the case of lines due to 
sulphur, &c., there is no doubt that the glass capillary of the vacuum tube may be a 
source of these impurities, and that then appearance immediately after the condenser 
is cut out may be due in part to their being liberated from the walls of the capillary 
by the powerful condensed discharge.* The evidence for a separation of the gases is 
therefore not clear in such cases. The Angstrom bands behaved like the IT^-drogen 
lines, but in this case the spectrum is due to a compound which is certainly broken up 
by the condensed discharge, and which would therefore require some time to reform 
or accumulate in the capillary when the condenser is cut out. 
With this exception all lines due to the heavier elements appear in the centre of the 
capillary when the condenser is cut out. If a partial separation of the gases takes place 
it is clear that, whatever the mechanism by which this occurs may be, the degree of 
separation is not proportional to the total energy flowing through the tube in a given 
time but must increase rapidly with the current density ; for the total energy of the 
condensed discharge was no greater than that of the uncondensed discharge, and was 
in fact somewhat smaller. It is, however, to be expected that some separation should 
be effected by the uncondensed discharge, and in addition to the effect recorded above 
in the case of the mercury line, we have often noticed that when heavy, uncondensed 
discharges are passed through Hydrogen tubes containing a little water vapour, the 
series lines of oxygen appear exclusively in the central portions of the capillary.! 
These observations seem to provide an explanation of the greater part of the pheno¬ 
mena described by Wood ( loc. cit.), who found that in long Hydrogen tubes which were 
not absolutely free from impurities, the Balmer series appeared strongly in the central 
portions of the capillary whilst the secondary spectrum was more strongly developed at 
* It lias been noticed that when very powerful condensed discharges are employed, “ arc ” lines of the 
constituents of the glass walls of the tube appear with great brilliance for a short space of time after the 
condenser is cut out, and experiments which are now being made seem to show that this may be developed 
into a convenient method of producing the spectra of many substances. 
f The phenomena are evidently of an entirely different character to those recorded by Sir J. J. Thomson 
(‘ Roy. Soc. Proc.,’ 58, p. 247, 1895). In the latter experiments the discharge was, in the main, unidirec¬ 
tional, and differences were observed in the spectra at the two poles. In our experiments the tubes were 
excited by high potential alternating discharges, and the spectra at the two electrodes were identical, 
VOL. CCXXII.—A. 3 H 
