1881.] Investigations on the Spectrum of Magnesium. 191 



2850, is always seen, very frequently reversed. Of the quadruple 

 group in the spark to which Cornu has assigned the wave-lengths 

 2801-3, 2797-1, 2794-5, and 2789'9, the first and third are strongly 

 developed in the arc, the other two not at all. Next follows a set of 

 five nearly equidistant lines, well-defined and strong, but much less 

 strong than the two previously mentioned, wave-lengths about 2782" 2, 

 2780-7, 2779-5, 2778-2, 2776*9. The middle line is a little stronger 

 than the others. The same lines come out in the spark. 



Beyond these follow a series of pairs and triplets ; probably they 

 are triplets in every case, but the third, most refrangible, line of the 

 triplets is the weakest, and has not in every case been noticed as yet. 

 These succeed one another at decreasing intervals with diminishing 

 strength, and are alternately sharp and diffuse, the diffuse triplets 

 being the strongest. The positions are shown in fig. 2. The series 

 resembles in general character the sodium and the potassium series 

 described by us in a former communication, and we cannot resist the 

 inference that they must be harmonically related, though they do not 

 follow a simple harmonic law. The most refrangible line in the 

 figure at wave-length 2605 represents a faint diffuse band which is 

 not resolvable into lines ; it belongs, no doubt, to the diffuse members 

 of the series, and, to complete the series, there should be another 

 sharp group between it and the line at wave-length 2630. This 

 belonging to the weaker members of the series is too weak to be seen. 



The approximate wave-lengths found by us for these lines are as 

 follows:— 2767-5, 2764'5, 2736, 2732-5, 2731, 2698, 2695, 2693-5, 

 2672-5, 2670, 2668'5, 2649, 2646, 2633, 2630, 2605. 



It is worthy of remark that the line at wave-length 5710, described 

 by us in a previous communication (" Proc. Boy. Soc," vol. 30, 

 p. 98), is very nearly the octave of the strong line at 2850. Moreover 

 the measures we have taken of the wave-length of this last line, with 

 a Rutherford grating of 17,296 lines to the inch, indicate a wave- 

 length 2852 nearly, which is still closer to the half of 5710. 



In Cornu's map of the solar spectrum a line is ascribed to mag- 

 nesium with the wave-length 3278. Although a line at this place 

 appears in many of our photographs of the arc, we have not been able 

 to identify it as a line due to magnesium. It does not show any 

 increased strength when magnesium is introduced into the arc. When 

 metallic magnesium is dropped into a crucible of magnesia or lime 

 through which the arc is passing, the electric current seems some- 

 times to be conducted chiefly or entirely by the vaporised metal, so 

 that the lines of other metals almost or wholly disappear ; but the 

 line at wave-length 3278 does not in such cases appear, though the 

 other magnesium lines are very strongly developed. The line at 

 wave-length 2850 is often, under such circumstances, enormously 

 expanded and reversed, those at wave-lengths 2801, 2794, and the 



