Rotation of Sodium Vapour at the D Lines. 147 



720° &c. As the bands widen and recede from the positions of 

 the D lines, the point midway between the D's becomes bright 

 and dark in succession, as I showed in the earlier paper. 

 Rotations as high as 1350° at this point have been observed, 

 the rotation increasing by 180° as the centre passes from 

 dark through bright to dark again. 



Even with the electric oven considerable difficulty was 

 found in keeping the density of the sodium vapour constant 

 during the time necessary to secure a complete set of readings. 

 It w T as finally found, however, that by putting a rheostat in 

 circuit with the magnet, the slight changes in the density of 

 the vapour could be completely compensated by altering the 

 magnetic field. In this way, the centre of the system could 

 be kept absolutely black for half-an-hour. This being the 

 case, it became at once possible to obtain satisfactory photo- 

 graphs of the phenomenon, from which the wave-length 

 determinations could be made at leisure. A small camera 

 was attached to the eyepiece tube, the rotation spectrum 

 being watched in the first order spectrum during the 

 exposure. With the centre of the system dark, the slightest 

 change in the density of the vapour became at once manifest 

 by the appearance of a faint line exactly in the centre of the 

 central dark band. This could be at once abolished by 

 making a slight change in the intensity of the magnetic 

 field. This is of course only true when the change consists 

 of a slight cooling of the tube, which causes the bright bands 

 bordering the central dark band to retreat towards the right 

 and left towards the D lines. If a slight rise in temperature 

 occurs, the bright bands push in towards the centre, and the 

 change cannot be so quickly noted as in the case before 

 mentioned, when a faint line appears on the black back- 

 ground. On this account, the tube was always heated a little 

 above the point at which it was desired to work, the current 

 being then reduced a fraction of an ampere. It is of course 

 not as easy to keep things absolutely constant when the 

 centre of the system is bright, for in this case a slight change 

 is not instantly recognized. 



An exposure of ten minutes was found sufficient, and a 

 number of very satisfactory records were obtained in this way. 

 Enlargements of two of them are reproduced (PI. VIIL), 

 the rotations at the centre being 360° and 720° respectively. 



Inasmuch as sunlight was employed, the Fraunhofer lines 

 in the vicinity of the D lines appear in the broad flares of 

 light, for which the rotation is 90°, and these lines served as 

 standards in determining the w T ave-lengths by measuring the 

 plates on the dividing-engine. 



L2 



