FLAME SPECTRA AT H[CH TEMPERATURES. 
167 
On the other hand, the evidence that it is due to the metal is of tlie following 
character:— 
(1.) Tt may be produced from the metal in a reducing' flame, and it disappears 
when an excess of oxygen is present. (2.) Although it may be produced by heating 
manganic oxide containing 66 per cent, of manganese, the spectrum is weak. 
(3.) A stronger spectrum is obtained by heating spiegel-eisen containing 18 to 20 
per cent, of manganese, and by heating ferro-manganese, containing 80 })er cent, 
of manganese, than that which it is possible to obtain by heating, to the same 
temperature and during the same period, manganic oxide containing 66 per cent, of 
manganese. Silico-spiegel containing 10 per cent, of silicon and 18 to 20 per cent, 
of manganese did not yield the manganese bands so strongly as the spiegel-eisen 
containing the same proportion of metal, probably because the manganese is converted 
into silicate. Even Turton’s tool steel yields a fairly strong indication of the man¬ 
ganese bands. 
If we examine the spectrum of air of the first order as obtained by spai-ks 
uncondensed, it appears to consist of bands only, but a more minute examination of 
spectra taken with an instrument giving considerable dispersion and excellent 
definition has shown that the bands are composed of three over-lapping series of 
lines. Such a chaiacter is usual with degraded band spectra of elements. If the 
pressure be reduced from the normal of 760 millims. to something like 5 millims. or 
less, then the bands disappear, and the strongest edge of each band remains as a line 
to represent the spectrum of the element at diminished pressure. Now, this change 
is one which is observed in the case of those metals which give band spectra, but, if 
they give bands and lines together, then the lines remain after the bands have 
vanished. This is to be observed in the spectra of silver, lead, bismuth, and 
tellurium. 
The most interesting case, however, is that of silver, for the spectrum is composed 
of a number of regularly disposed and closely placed lines. 
The bands are degraded towards the rays of lesser relrangibility, that is to say, in 
this dh'ection the lines are of diminishing intensity, and they arc of increasing width 
apart. When the quantity of silver diminishes, and consequently the vapour exerts 
less pressure, being mixed with the vapour of other metals, the bands become narrower 
until at last nothing but lines remain, and these are the strongest lines belonging to 
the strongest bands. They correspond to those on the spark spectrum with wave¬ 
lengths 3382'3 and 3280’1. 
Thus we see how the line spectra are related to band spectra, and that there is 
really no essential difference between the constitution of the matter which enters 
into the vapours of metals and metalloids; there is, in fact, something in their 
constitution common to both, which is appai’ently dependent on their vapour 
pressures and probalfly due to the action of the molecules upon one another when 
