SAUNDERS. — ARC SPECTRA OF THE ALKALI MKTALS. 453 



Wave-number diQerences. 

 Calculated from maia Hue. Calculated from satellite. 



531.9 553.7 



542.3 553.9 



547.2 554.7 



]t will be seen at once that the differences calculated from the satellites 

 agree beautifully with those of tiie second suljordiuate series. Ap- 

 parently, then, the wave-number difference for the first subordinate 

 series, as usually calculated, should vary slightly, increasing with de- 

 creasing wave4ength. A glance at the table of dififerences for rubidium 

 shows the same effect in that spectrum also, and there, too, if we use the 

 satellite at 7759.5 (the only one yet observed) with the line 7619.2, we 

 obtain a difference of 237.8 instead of 234.7, thus bringing this value 

 into agreement with those for the second subordinate series. This 

 principle may be used to calculate the positions of the other satellites in 

 the rubidium spectrum. They should be at 9213.6 (using Lehmann's 

 line 8766.10 for the calculation), 6299.6, and 5725.1, thus being within 

 very short distances of their parent lines. As the latter are broadened 

 toward the greater wave-lengths they cover up the satellites, so that these 

 have not yet been observed. The satellites in the spectra of lithium, 

 sodium, and potassium are also, probably, too close to their parent lines to 

 be distinguishable. 



The writer wishes to express his indebtedness to the committee in 

 charge of the Rumford Fund for a grant covering the expenses of this 

 investigation. 



Syracuse Univkrsity, 

 June, 1904. 



