410 Dr. C. Chree. Collimator Magnets and the 



We have P = when X'/X = J2/3 = 0*8165, approx., 

 Q = „ X'/X = 0*467, approx. 

 We can thus make either P or Q vanish, but not both simultaneously. 

 When P = 0, 



Q = -9X4/2. 



§ 46. Both collimator and mirror magnets are comparatively " short," 

 i.e., their lengths are only about ten times their diameters. In such 

 magnets we should expect, according to Coulomb, 



2X = 21/3, 2 A.' = 2l'/3, 



where I and I' are the lengths of the magnets. The formulae by Green, 

 Jamin, and others are more complicated. 



Perhaps all we are justified in saying, a priori, is that X and X' cannot 

 exceed 1/2 and l'/2, and that the collimator and mirror magnets are 

 sufficiently similar in pattern to make it probable that the assumption 



X'/X = I'll 



is not far wrong. 



To throw some light on the question, I had measurements made of 

 the magnets of the Kew unifilar and of a unifilar of class E. The Kew 

 unifilar belongs to group B, and its P is negative but exceptionally 

 small. The instruments of group E are, as we have already seen, 

 exceptionally uniform in character, and invariably have their P large 

 and positive. 



The results were as follows, lengths being in centimetres, 



Unifilar. 1. V . I'/l. 



Kew, group B 9*35 7*60 0*81 



„ E 9-17 6-35 0-69 



Supposing X'/X = I'jl, we should have in these two cases 



P/\ 2 . Q./\ 4 . 



0-81 0-00 -4-4 



0-69 +0-57 -2-9 



These figures, taken in conjunction with our previous data, are on 

 the whole distinctly favourable to the hypothesis 



2Xjl = 2X'/r = n (a constant) (18) 



Accepting this hypothesis provisonally, I had the curiosity to see what 

 value we should obtain for n by ascribing to P the mean value found 

 for unifilars of group E, taking for I and I' the values quoted above for 

 a unifilar of that group. 



