17 



see that the relation (18) is a close approximation to (17). 



Finally let us investigate what approximation is made in computing 



T^/FxL = t — _ v"* » the quantity of greatest interest. 

 Writing for the moment log t.. = - U, and log t 8 = V, we have 



*1 e- U - 1 



^r=^ e v_ e -u ;tuW7- 



But by (6) and (10) we see that U+V involves terms that are a second order 

 approximation to U° + V° = log t 2 ° - log t-° (by 19) and (20)) while the remaining 

 terms are not only small but also act to cancel each other, by (21). Hence we 

 conclude that a close approximation to T.. is obtained by using equations (6a), (8a), 

 (10a) and (12a) instead of the set of equations (6), (8), (10) and (12). 



