1920] on The Blue Sky and the Optical Properties of Air 155 



practical experimenting. An enormous multiplication of the length 

 and tightness of the cotton-wool filter makes no difference at all ; a 

 filter of modest dimensions doing all there is to do. 



The dust particles which are originally present in the air, near the 

 ground or in a room, are large, being in some cases individually 

 visible to the naked eye ; thus they do not fulfil the condition for 

 scattering «a preponderance of blue light. The molecules of air are of 

 course amply small enough, and the band of light seen stretching 

 across the mouth of the dark cave is, to my eyes at least, of a full 

 blue colour. In exhibiting the effect to individual friends (and un- 

 fortunately it is not bright enough to be shown to an audience), I 

 have been surprised and somewhat disconcerted to find that they do 

 not all see it blue as I do, but some, for example, describe it as 

 lavender. This is undoubtedly due to a peculiarity of colour-vision 

 where faint lights are concerned. The ultimate test is the spectro- 

 scope. Photographs of the scattered light taken with this instrument 

 clearly show that the maximum of intensity is shifted towards the 

 blue, as compared with the original exciting light. 



Polarisation of Light Scattered by Pure Air. 



A very important point to examine in connection with the 

 scattered light is its state of polarisation. Visual examination with 

 a Xicol's prism soon showed that the polarisation was very nearly 

 complete. For closer examination I had recourse to photography. 

 It may perhaps be thought an easier and more effective plan to look 

 at a phenomenon than to photograph it, and no doubt it is so in 

 many cases ; not however where the light is very faint, but admits of 

 long exposure. It has long been recognised that photographs of the 

 nebulae will show much more than can be detected visually by the 

 keenest and most discriminating eye. In this work on the scattering 

 of light, I have found it positively less trouble to take a photograph 

 than to make a visual observation, even when the latter was feasible. 

 The time required to rest the eye in darkness and the effort of atten- 

 tion required in observing a faint effect cost the experimenter more 

 effort than the exposure and development of a plate. 



When the scattered beam in pure air is photographed, with a 

 double image prism of Iceland spar mounted over the photographic 

 lens, it is found that the polarisation is nearly complete, but not 

 absolutely so. However carefully the instrumental adjustments are 

 made, and however carefully the air is filtered, I have found that 

 there is a slight residual polarisation indicating vibrations parallel to 

 the direction of the original beam. The intensity of this residual 

 polarisation, in what may be called for convenience the wrong direc- 

 tion, is about 4 per cent, of the whole. Now, as the theory shows, 

 there are two causes to which failure of complete polarisation may 

 be attributed. One, which we may dismiss in this case, is that the 



