Stoney — Analysis of the Spectrum of Sodium. 217 



direction of revolution in the elliptic partial. If, on tlie other 

 hand, it is found that the more refrangible constituent of the double 

 line is the stronger, this will go far to prove that the direction of 

 the apsidal shift is independent of the sign of n. It is therefore 

 desirable to ascertain by observation which constituent is the 

 brighter. 



9. Finally, our investigation makes it probable that there is 

 some connexion between Series D and Series S in the spectrum of 

 sodium, and that the same relation prevails between the two series 

 of triple lines in the spectra of Zn, Cd, and Hg. This is suggested 

 by the circumstance that the line which we now know to be the 

 first term of Series S stands in a position in all these spectra 

 which appears to be related in the same way to the positions in 

 them of the lines of Series D. But what the connexion is we do 

 not yet know. 



Description of the Diagrams. 



A sketch on a very small scale of the primary curves of the 

 three sodium series is given in PI. xvii., fig. 4, and a sketch of 

 their derived curves is given in fig. 5. In both these figures, as in 

 the diagrams of the corresponding hydrogen curves in PL xvi., the 

 horizontal lines represent the oscillation-frequencies of the successive 

 lines of each series, when measured from the vertical line to the 

 curve belonging to that series. A small circle is placed round 

 those dots that correspond to lines that have not yet been observed, 

 and the cross on the upper line between - 500 and - 600 indicates 

 the distance to which Professor Hartley has succeeded in photo- 

 graphing in the ultra-violet. 



To judge what the approximation is that has been attained, 

 imagine each of the diagrams enlarged, until the vertical line 

 becomes ten metres long. Each of the diagrams would then occupy 

 the side of a large house. Even on this immense scale the greatest 

 deviation of the observed ends of the horizontal lines from the 

 curve, would be less than six millimetres in the case of Series S, 

 would be under three millimetres in Series P, and would be only a 

 fraction of a millimetre in Series D. Although these deviations 

 are very small, modern spectroscopic work is carried out with such 

 accuracy that they may not be attributed to errors of observation, 

 and, accordingly, we are justified in drawing the first and second 

 of the inferences on p. 215. 



