MAGNETIC EFFECTS ON NOISE IN A GERMANIUM FILAMENT 657 



is the ratio 



-. _ ^gl {JDrli+a) + {Jl)Tl{-a) 

 ly H — ~" 



{Jiy„=o i+a) + {Jl)rUi-a)' 



(11) 



ASf/y=0 



Therefore we also need the lifetimes at zero H (that is, zero $). But at 

 $ = 0, the r's are indeterminate, and we therefore have to take limits. 

 Expansion in powers of ^ shows that 



It T„(+a) = ^ 



It Th{-0) = — — 



(-^) 



1 + 



Y- + -V 



l + 2>- 



Thus we finally get 



Nj, = 



^9l 





{Ji} 



V^ L 



Oil + 0:26* 



{jVi 



Oil 



{^-^) 



+ 1 



q;i6 



+ «2 



(12) 



(Jl) 



F(-^)^f('n^) 



There remains one small difficulty in the way of comparing experiment 

 with theory: We do not know (Jl), (Jl)- As suggested by H. C. Mont- 

 gomery, we are able to overcome this difficulty as follows : We first draw 

 a number of curves of 



rU+a) 



th 



(-a) 



versus <i>, 



rli+a) rK-a) 



for various sets of parameters (^i , ^2.) (See figures 3, 4). Then we 

 contrive to match a linear superposition 



Ci 



Tlr(+a) . . rli-a) 



rl{+a) 



+ C2 



To 



i-a) 



where Ci + C2 = 1 to the experimental curve. This will be possible only 

 for one particular set of values (^1 , ^2). From Ci and C2 = 1 — Ci and 



