On the Chemistry of the Hottest Stars. 159 



generally observed dropping- out of lines in the spark depends upon 

 the diminished quantity of incandescent vapour as compared with 

 that in the arc. 



With the brilliant spark obtained under the new conditions there 

 is little difference betweeen arc and spark with regard to the num- 

 ber of lines. Hence it may be concluded that the small number of 

 lines previously recorded in the spark spectrum was not an effect of 

 increased temperature, but one due to the small quantity of vapour 

 produced by the use of a small coil. 



The next point was to inquire if the photographic region of the 

 spectrum reveals lines which are the equivalents of those at 4924*1 

 and 5018*6 (Rowland's scale), which I had previously shown to be 

 enhanced in passing from the arc to the spark spectrum.* 



Seven additional lines were detected in the photographs at the 

 following wave-lengths on Rowland's scale. 



42333 

 4508-5 

 4515-4 

 4520-4 

 4522*7 

 4549-6 

 4584-0. 



These have been confirmed by a reference to the map of the 

 spark spectrum of iron published by Dr. McClean,f and attention 

 was drawn to those at 4584-0 and 4233*3 in my paper on the arc 

 spectrum of electrolytic iron.;£ 



All these appear as short lines in the arc spectrum, so that the 

 view that the short lines which appear in the arc spectrum can be 

 divided into two categories, one including the lines which are 

 brightened in the spark, and the other the lines which are not so 

 enhanced, is confirmed. 



Having thus established that there are differences between the arc 

 and spark spectrum, and that these differences are not due to the 

 different quantities of vapour in the two cases, it must be concluded 

 that a difference of temperature is the main cause of change. 



Including the flame spectrum then, four distinct temperature 

 stages are indicated by the varying spectrum of iron : — ■ 



(1) The flame spectrum, consisting of a few lines only, including 

 the well-known triplets and many strong lines in the ultra-violet. 



(2) The arc spectrum consisting, according to Rowland, of 2000 

 lines or more. 



* ' Koy. Soc. Proc.,' vol. 32, p. 204. 



f ' Monthly Notices K.A.S.,' vol. 52. 



X ' Phil. Trans.,' vol. 185, A, pp. 995, 996; 



