An Investigation of the Spectra of Bessemer Flames. 95 



the instruments and observers. A flame was visible at the mouth of 

 the converter as soon as the cloud of dust had cleared away ; this had 

 a yellow or yellowish-red colour. The flame grew rapidly in length 

 and remained clear as in the " acid " process, until it dropped and the 

 second stage began. In this stage the flame was very short, and a 

 large quantity of fume was expelled from the vessel ; the flame grew 

 longer and the quantity of fume increased as the "blow " proceeded. 



Twenty-six plates of spectra were photographed ; some of these 

 were very sharp and gave a complete record of substances present 

 in the flame at intervals of one minute throughout the blow. Care- 

 ful measurements of the best spectra have been made, and the wave- 

 lengths of the lines and bands recorded. The others, not measured in 

 detail, have been compared with these, but no lines or bands occur in 

 them which do not also occur in the best plates. A plate of spectra 

 was usually taken by giving the same time of exposure to each 

 spectrum of the series until the flame dropped ; two further exposures 

 were then made on the flame of the over-blow. The spectra increase 

 in intensity as the blow proceeds in the first stage, and this can only 

 result from a corresponding increase in the temperature of the bath 

 of metal and of the flame. 



Much detail was lost in many of the spectra, by the interference of 

 the light reflected from a large quantity of white dust and smoke in 

 the air in the neighbourhood of the converters. The converter nearest 

 the observers was the only one of the four from which delicate detail 

 was obtainable, and this was only secured by working in the evening 

 when the sun was very low, or after it had set. 



Considerable difficulty was experienced in the identification of some 

 of the lines and bands. This was due partly to the comparatively 

 small dispersion in the less refrangible portion of the green and red rays, 

 by which lines and the sharp edges of bands were almost indistinguish- 

 able on the strong continuous spectrum. In other cases, lines were 

 present which had not been observed in flame spectra before, some due 

 to uncommon elements, and others were relatively much stronger 

 than a study of the oxyhydrogen flame and other spectra of the same 

 metals led us to expect. 



Conclusions. 



(1.) Line spectra are not observed in the open-hearth furnace. This 

 is attributed mainly to the fact that the atmosphere of the furnace 

 is oxidising, and under these conditions, as Gouy has shown,* only 

 sodium gives a spectrum approaching in intensity that which it gives 

 in a reducing flame. The D lines were observed by eye observation, 

 but did not appear on the photographs. 



* ' Phil. Mag.,' vol. 2, 1877, p. 156. 



