296 Proceedings of Societies. 
same metal in air, and in a continuous current of pure hydrogen. Iron, 
for example, gave in hydrogen, a spectrum in which a bright orange and a 
strong green band were visible, besides a few faint lines in the blue part of 
the spectrum. Although the light produced by the action of the coil was 
allowed to fall for ten minutes upon a sensitive collodion surface, scarcely a 
trace of any action was procured; whilst, in five minutes, in the air, a power- 
ful impression of numerous bands was obtained. It was remarked by Mr 
Talbot, that in the spectra of coloured flames, the nature of the acid did not 
influence the position of the bright lines of the spectrum, which he found 
was dependent upon the metal employed, and this remark had been con- 
firmed by all subsequent observers. But the case was very different in the 
absorptive bands produced by the vapours of coloured bodies,—there the 
nature of both constituents of the compound was essentially connected 
with the production of absorptive bands. Chlorine, combined with hydro- 
gen, gave no bands by absorption in any moderate thickness. Chlorous 
acid and peroxide of chlorine both produced the same set of bands, while 
hypochlorous acid, although a strongly coloured vapour, and containing 
the same elements, oxygen and chlorine, produced no absorptive bands. 
Again, the brownish red vapour of perchloride of iron produced no ab- 
sortive bands; but when converted into vapour in a flame, this gave out 
bands independent of the form in which it occurred combined. These 
anomalies appeared to admit.of an easy explanation on the supposition 
that, in any case, the compound is decomposed in flame, either simply by 
the high temperature, just as water is, as shown by Grove, or, in all 
other cases of the production of bright lines by the introduction of a 
metallic salt into a flame of burning bodies (as shown by Deville). In 
the voltaic pile, the decomposition must of necessity take place by electric 
action. The compound gases, protoxide and binoxide of nitrogen, gave, 
when electrified, the same series of bright bands (as Pliicker had shown) 
which their constituents when combined furnish. Aqueous vapour always 
gives the bright lines due to hydrogen, and hydrochloric acid the mixed 
system of lines which could be produced by hydrogen and chlorine. The re- 
ducing influence of the hydrogen and other combustible constituents of the 
burning body would decompose the salt, liberating the metal, which would 
immediately become oxidized, or carried off in the ascending current. There 
was obviously a marked difference between the effect of intense ignition 
upon most of the metallic and the non-metallic bodies. The observations 
of Pliicker upon the spectra of iodine, bromine, and chlorine, show that 
they give, when ignited, a very different series of bands to those which 
they furnished by absorption, as Dr Gladstone had already pointed out ; 
but it was interesting to remark that, in the case of hydrogen, which, 
chemically, was so similar to metal, we have a comparatively simple 
spectrum, in which the three principal bright lines correspond to Fraun- 
hofer’s dark lines, C F and G. It was, however, to be specially noted, 
that the hydrogen occasioned no perceptible absorptive bands at ordinary 
temperatures in such thickness as we could command in our experiments ; 
and the vapour of boiling mercury was also destitute of any absorptive 
action, although when ignited by the electric spark it gave a characteristic 
and brilliant series of dark bands. The following experiment suggested 
itself as a direct test of Kirchoff’s theory. Two gas-burners, into which 
were introduced chloride of sodium on the wick of the spirit-lamp, were 
placed so as to illumine equally the opposite sides of a sheet of paper 
partially greased. The rays of the electric light screened from the pho- 
tometric surface, suitably protected, were made to traverse one of the 
flames. If the yellow rays of the light were absorbed by the sodium 
flame, the light emitted laterally by the flame should be sensibly in- 
creased. The experiment, however, failed to indicate any such increase 
