1887.] on Light as an Analytic Agent. 87 



much expanded as to be lost in the continuous spectrum ; for it 

 showed in a photograph afterwards taken. Other lines were, how- 

 ever, seen — the sodium yellow lines, the calcium indigo line, a group 

 of other blue lines, and a group of green lines, amongst which 

 one line was conspicuous, and this line, by comparison with the 

 solar spectrum, was identified with E. We had not expected to 

 see any lines of iron, as iron and its compounds give no lines in the 

 flame of a Bunsen burner, and we supposed that it would only be 

 volatilised at a much higher temperature. But the appearance of E 

 suggested that other of the green and blue lines might be due to 

 iron ; so we proceeded to compare the positions of these lines with 

 those of the electric spark between iron electrodes. For this purpose 

 one of the spark-lines was first brought carefully on to the pointer, 

 or cross wires, in the eye-piece of the observing telescope, and then, 

 the passage of the spark being stopped, the flash of the exploding 

 gases was observed. It was not difficult to see whether any line was 

 on the pointer ; and the observation could be repeated as many times 

 as was desired without any shift of the apparatus. Nine of the most 

 conspicuous green and yellowish-green lines in the flash were thus 

 identified with lines of iron. For the blue and violet we adopted the 

 photographic method as much less trying to the eyesight. Eight to 

 twelve flashes were taken in succession without any shift of the 

 apparatus, so as to accumulate their effects on the photographic plate. 

 Eight flashes were found enough in 'general to produce a good im- 

 pression, and more than twelve could not well be taken without 

 turning out the water which accumulated in the tube, and cleaning 

 the glass which closed its end. After the flashes had been taken, 

 but without shifting the photographic plate, the slit of the spectro- 

 scope was partly covered, and the electric spark between iron points 

 passed in front of the slit. We had thus on the plate the photograph 

 of the flash as well as of the sj)ark. Fourteen more lines in the in- 

 digo and violet were thus identified with iron lines ; and on extending 

 the photographs into the ultra-violet, and substituting quartz lenses 

 and prisms for the glass ones hitherto used, a much larger number of 

 lines were identified. There could be no doubt, then, that we had 

 iron vapour in the flash. We supposed that it must be derived from 

 dust of oxide shaken by the explosion off the sides of the tube, and 

 we had the tube bored out clean and bright like a gun-barrel. This 

 made no diminution in the brightness or number of the lines ; and we 

 came to the conclusion that the explosion detached particles of iron 

 from the tube, and converted them into vapour. This was confirmed 

 by finding that, however carefully the tube had been cleaned, the 

 glass ends always became clouded with a rusty deposit after ten or 

 twelve flashes. Altogether 68 lines of iron have been identified in 

 the flash, of which about 40 lie in the ultra-violet between H and 0. 

 Only one iron line above O has been definitely identified, and that in 

 only a few photographs. It is T. 



As iron gave so many lines in the flash it was reasonable to sup- 



