Polarization of the Light of the Sky. 503 



We see now that these maps disregard entirely the existence 

 of a negative region. For instance, the first line, corresponding 

 to polarization ll = l 0, 6, goes right round each point, through 

 both positive and negative regions. Hence each of the neutral 

 points has positive polarization both above and below it, which 

 is contrary to all the observations. 



Again, suppose a point taken anywhere in the positive region, 

 and another in the negative. It is clear that, on any line joining 

 these, the polarization must pass through a zero value on 

 changing sign. Hence there must be a neutral line or region 

 separating the positive and negative regions. The extreme diffi- 

 culty of the observations may be the reason why this has not 

 been established by observation. But it is clear that it must 

 be so ; and traces of the fact are found here and there in the ob- 

 servations. I will mention some cases. 



Edinburgh Trans, xxiii. p. 214, June 10 : — " At 3 h 47 m , when 

 the sky was everywhere pure and free of clouds, the neutral line 

 of the polarimeter was curved above the land horizon. At 6 h 32 m 

 the curvature was greater above the sea horizon, the deviation 

 increasing towards the horizon." 



1842, Feb. 21, at 12 h 39 m : — "Neutral line convex towards 

 the sun in western horizon." 



In the figure (p. 213) of the bands in the polarimeter, the 

 appearance indicated is that of a horizontal neutral line. 



Eubenson (p. 144) swept the horizon immediately after sun- 

 set ; at a very small distance from the sun he recognized a well- 

 defined neutral point. No azimuth is given ; and this is the 

 only observation he gives of the phenomenon. 



Probably the difficulty of the observations may be the reason 

 why no observer has searched methodically for neutral lines, but 

 always for points ; these may be merely the intersection of the 

 neutral line with the circle in which the instrument sweeps. 

 Ruben son was frequently unable to find any of the neutral points, 

 and has no detailed observations of them. We also find that 

 Babinet regarded the observation both of his own point and of 

 Brewster's as of extreme difficulty. 



Before proceeding to suggest a more consistent interpretation 

 of the data, I will advert shortly to the principal points estab- 

 lished in TyndalFs experiments, and endeavour to gain some 

 general principles for our guidance. 



Stage I. Vapour of minimum density. Blue colour. Pola- 

 rization positive. Light emitted normally to the illuminating 

 beam perfectly polarized. 



Stage II. Increased density of vapour, and probably increased 

 size of particles. Blue passes over into white. 



(a) Light emitted normally is partially polarized positively 



