FLAME SPECTRA AT HIGH TEMPERATURES. 
1087 
spectrum be accounted for in all cases if they are not due to carbon or to carbonic 
oxide but to manganese ? On an examination of the curves drawn from the analyses 
of Sneltts it will be observed that all the manganese is burnt out of the molten metal 
within the first fev/ minutes and converted into slag. It is obvious that the 
manganese bands cannot in these circumstances proceed from the bath of metal up 
to the end of the “ fining stage ” if the quantity required to give a marked indication 
of them is comparatively large. The sole source of the manganese bands must be a 
quantity of metal reduced from the slag, this reduction cannot take place at once, but 
is the result of a continously increasing temperature and the chemical action of 
reducing materials such as ferrous carbide, carbonic oxide, and possibly metallic iron. 
In order to ascertain whether slag could be reduced and give rise to the charac¬ 
teristic spectrum of the Bessemer flame, a piece of slag from the works at Crewe was 
heated by the oxyhydrogen blow-pipe both in the inner and outer flames, that is to say, 
where in the latter case reduction could take place by dissociation by heat alone, but 
in the former it might be aided by chemical action of the excess of hydrogen. The 
photographs obtained were strikingly like those from the Bessemer flame at Crewe. 
Comparing it with ferro-manganese, we have band for band belonging to manganese, 
and line for line in the iron spectrum, exactly reproduced. There can be no doubt 
whatever that both iron and manganese are freely volatilized from the slag. Con¬ 
sidering the small proportion of manganese in haematite pig, and the fact shown 
in Diagram I, that the manganese in such metal is all converted into slag during the 
first five or six minutes, it is evident that the continued brilliancy of the manganese 
spectrum during the “ boil” must be entirely due to its vaporization from the slag. 
In order to connect the disappearance of this spectrum with the chemical change 
involved in the decarburization of the iron, we must consider the falling off in intensity 
of the line spectra during the close of the “ fining stage,” before the final drop of the 
flame, and it will be seen that this can clearly be due to no other cause than a fall of 
temperature, consequent upon a reduction in the quantity of carbon burnt. The 
thinness, transj^arency, and want of brilliancy in the flame at this period is due to 
the comparatively small quantity of carbonic oxide in the issuing gas; the final drop 
being caused by an escape of oxygen into the vapours and gases within the converter, 
which is signalized by a cloud of fume. The removal of the carbon from the metal 
causes the disappearance of the manganese bands. 
Let us now consider the case of the Dowlais “ blow.” 
The spectrum of the flame in this case resembles strongly the spiegel-eisen spectrum, 
and those of metallic manganese and ferro-manganese. There can be no doubt that 
manganese is vaporized in the bath of metal, and hence the large number of bands, 
their distinctness, and great brilliancy. The diagram of Kupelwieser’s analysis will 
make this plain, for here we have an excess of manganese in the iron, which, though 
oxidized during the “ slag-forming period,” continues to be vaporized during the 
