FLAME SPECTRA AT HIGH TEMPERATURES. 
1049 
visible, but on one occasion the spectra of the sun and of the flame from the converter, 
during four periods in one “ blow,” were secured on one plate. With this plate it was 
easy to see that a large number of lines were coincident with lines in the solar 
spectrum. Upon enlarged prints, some of which were 10 X 12 or 12 X 15 inches, 
made with gelatine-silver bromide paper, it was easy to record the position of the 
lines and edges of bands with respect to the sodium line, as these were measured with 
a micrometer screw and microscope in the .manner already described (‘Phil. Trans.,’ 
1894). It became necessary, however, to obtain enlargements of greater dimensions, 
and, accordingly, the Autotype Company were requested to make such, measuring 
36 X 24 inches, in which the spectra were magnified ten diameters. These were 
easily examined and compared with Rowland’s first map of the solar spectrum and 
with Maclean’s photographed spark spectra of metals. These enlargements had 
another advantage than facility for recognising coincident lines, for, with a standard 
brass metre, the bands in the spectra were more easily measured than with the 
micrometer. Several interpolation curves were drawn by which linear measurements 
were reduced to oscillation-frequencies. These were necessary because the portion of 
the spectra less refrangible than H was differently focussed on some of the plates. 
The fiducial lines selected were 110 lines in the spectrum of iron, and in the solar 
spectrum, lying between D and P. 
It was found to be almost impossible to measure the same bands on different 
spectra and obtain measurements giving identical wavedengths. This will appear on 
referring to Plate 6, Crewe, and comparing the measurements of the first, second, and 
third spectra. Although there is some difficulty in obtaining measurements of bands 
so precise as we are accustomed to in line-spectra, this does not account for the 
discrepancy. It is, in point of fact, due to the bands altering in width, or, in some 
cases, becoming less distinct at the edges; the bands are also much obscured on some 
spectra by the continuous rays being strong. There is some difficulty also, it may be 
remarked, in measuring the broad lines visible in some of the banded metallic spectra 
(‘Phil. Trans.,’ Part I., 1894). 
Observations made at Crewe and at Doivlais in 1893. 
I am much indebted to Mr. F. W. Webb, of the Locomotive Department of the 
London and North-Western Railway, and to Mr. E. P. Martin, the Manager of the 
Dowlais Iron Works, for facilities afforded me in carrying out a series of observations 
at Crewe and at Dowlais, in January and in April, 1893. In order to photograph 
the Bessemer flame, the instrument was placed on a strong, low bench, sufficiently 
near to the mouth of the eonverter and in a position of safety as regards sparks 
projected from the vessel, when the blow at times became somewhat wild. While 
my assistant made exposure of the plates at different periods according to instruc¬ 
tions, the phenomena observed during the “blow” were noted by me. During the 
first seven minutes there is only a continuous spectrum. The sodium line then 
MDCCCXCIV.—A. 6 S 
