416 Prof. Wood on tlie Magneto- Optics of Sodium Vapour 



quantity (X 2 — \ 2 n ) f which changes sign on crossing the 

 absorption-hand) is squared. An attempt was first made to 

 verity the simplified formula by writing X, ra = 5893, *, <?., 

 assuming the D lines to be a single absorption-band. The 

 value of the constant could then be determined from a single 

 observation, and the rotations for different values of X cal- 

 culated. The calculated values agreed very closely with the 

 observed on one side of the double absorption-band, but the 

 agreement on the other side was very poor. This was at once 

 seen to be due to the fact that the rotatory power of the two 

 lines was very different, and the following formula was con- 

 sequently adopted : — 



._ a\ 2 _J\P__ 



The constants a and b were calculated from two observations 

 of X, one greater than D 1? the other less than D 2 . In almost 



every case the ratio 7- was found to lie between the limits 1'8 



and 2-03. Writing Di = 5896 and D 2 = 5890, the values of 

 the constants a and b for vapours of different densities are 

 recorded in the following table. The densities are defined by 

 the magnitude of the rotation for a wave-length midway 

 between D 1 and D 2 . 



Density .... 



Centre 8= 50° 



„ „ 90° 



„ ., 180° 



„ „ 360° 



„ „ 450° 



„ „ 540° 



„ ,, 720° 



„ „ 960° 



a. 



6. 



a/b. 



1123 



587 



1-9 



2217 



1350 



1-65 



4309 



1860 



2-3 



5111 



2571 



1-98 



9279 



4545 



2-03 



10326 



4607 



2-24 



12750 



7100 



1-8 



23873 



10333 



26 



Centre 

 Rotation Cale, 



51° 



98° 

 161° 

 L ; 05° 

 369° 

 392° 

 538° 

 893° 



The "centre*"' as defined above does not in fact fall at 

 X = 5893, but at 5893*5, i.e., when the bright bands unite, 

 the point of unison lies nearer D x than D 2 . This is owing to 

 the fact that the expansion of the bands from D 3 is more rapid 

 than from Di. 



It was found that if the constants a and b were calculated 

 from different pairs of readings, there was considerable 

 difference between the values, a circumstance which indicates 

 either that the formula is at fault, or that the observations 

 were not sufficiently accurate. We are dealing with very 

 steep curves, as will be seen by reference to fig. 3, and 



