286 Mr. J. N. Lockyer on Spectrum- Analysis. [May 8, 



in such a manner that they could at will be surrounded with any gas or 

 vapour. Hydrogen was used in some of these experiments ; it was puri- 

 fied in the usual manner by drying and freeing from traces of sulphu- 

 retted hydrogen ; it was then passed over clean cut pieces of sodium, and 

 admitted to the poles. An induction-spark from 5 one-pint Grove cells 

 was used, the circuit being without the Leyden jar. 



The lead compounds behaved (in air) as follows : — 



The fluoride gave the eleven longest lines of the metal ; but four were 

 very faint. 



The chloride gave nine lines ; one of these was very short. 

 The bromide gave six lines ; but one was a mere dot on the pole. 

 The iodide gave four lines distinctly and two as dots, one of which was 

 scarcely visible. 



It is pointed out that the decrease in length and number of lines 

 follows the increase in the atomic weight of the non-metallic element, 

 the lines dying out in the order of their leugth. 



Barium was next experimented on, the same series of salts being used. 

 A marked departure from the results obtained in the case of the lead 

 compounds was observed, especially in the case of the fluoride, its 

 spectrum being much the simplest ; in fact it consisted of only 4 lines. 

 Strontium behaved like barium ; and so did magnesium fluoride. This 

 anomalous behaviour was found to be most probably due to the ex- 

 ceedingly refractory nature of these fluorides, all of them being quite in- 

 fusible., and non-volatile in any spark that was used. 



Sodic fluoride, sodic chloride, sodic bromide, and sodic iodide exhi- 

 bited a behaviour exactly the reverse of that of lead ; i. e. the iodide 

 showed most of the metallic spectrum. 



The difference between flame-spectra and those produced by a weak 

 electric discharge are then discussed. Beads of the chlorides &c. were 

 heated in a Bunsen gas-flame ; Bal 2 gave a " structure " spectrum (since 

 proved to be due to the oxide) and the line at wave-length 5534*5, 

 by very far the longest metallic line of barium. The bromide be- 

 haved like the iodide ; and so did the chloride, except that its spectrum 

 was more brilliant. Baric fluoride gave scarcely a trace of a spectrum, 

 the oxide structure being scarcely visible, and 5534*5 very faint indeed. 

 The strontium salts follow those of barium — 4607*5, the longest strontium 

 line, appearing in conjunction with an oxide spectrum. The strontic 

 fluoride, however, refused to give any spectrum whatever. These 

 results are compared with those obtained with the weak spark, and it 

 is shown that the difference is one of degree : e. g. baric bromide gives 

 25 lines in the spark ; these are the longest lines. In the flame it gives 

 but one line ; but this is the longest of all the barium lines, and indeed 

 very far exceeds all the others in length. When the flame-spectra are 

 compared with those produced by the low-tension spark, the spectra of 

 the metals in the combination are in the former case invariably more 

 simple than in the latter, so that only the very longest line or lines are left. 



