Resonance Spectra of Iodine Vapour. 479 



work. The light from the arc was first rendered parallel by 

 means or! a large condenser, then passed through a Nicol 

 prism of about 13 cm. aperture, and finally foeussed at the 

 centre of the bulb by a second condenser. A large Savart 

 plate was mounted to one side o£ the bulb, and the fringes 

 viewed through a second large nicol. It was first necessary 

 to get rid of all reflexions of polarized light from the walls 

 of the bulb and room. This was accomplished by means of 

 screens suitably placed, and a background of black vel\ et. 

 To make sure that the fringes resulted wholly from the 

 fluorescent iodine vapour, the wall of the bu!b was touched 

 with a bit of cotton soaked with liquid air. This immediately 

 condensed the vapour, destroyed the fluorescence and all 

 trace of the fringes, the background appearing perfectly 

 black. It is of course necessary to get the bulb in such a 

 position that none oE the internal reflexions comes from the 

 region where the fringes are observed. 



On PI. V. fig. 1 we have a photograph of the fluorescence 

 of the vapour excited by the mercury arc without the nicols 

 or Savart plate, and (fig. 2) a photograph of the Savart fringes 

 crossing the cone of fluorescent light. It was thought that 

 possibly the percentage of polarization could be increased by 

 reducing the density of the iodine vapour. The oblique 

 plates by which the polarization was compensated were set 

 at such an angle as to just cause the disappearance of the 

 fringes, under which condition any increment in the per- 

 centage of polarization would cause their reappearance. 



The outside of the bulb was now touched with cotton 

 soaked with liquid air, which caused a rapid condensation 

 of the iodine vapour. The cone of fluorescent light faded 

 gradually awa}~, but the Savart fringes did not reappear, 

 though the results obtained with sodium vapour made me 

 think that at the lowest vapour densities consistent with 

 visible fluorescence, increased polarization might be ex- 

 pected. Warming the bulb to increase the vapour density 

 above the value which it has at room temperature did not 

 appear to decrease the amount of polarization, though the 

 intensity of the fluorescence was considerably diminished. 



The investigation of the polarization of the resonance lines 

 was next investigated. This I regard as one of the most 

 important points in the whole investigation, for if the bright 

 lines excited by polarized monochromatic light are themselves 

 polarized, it indicates that the directed motion passed over 

 by the light-waves to the resonant electron is passed on as 

 directed motion to the other electrons, or whatever the centres 

 of emission of the bright lines of the resonance spectrum 



