905 
1 
would be perfectly accurate. If rs > 7 x 2(1 + y) > s, and not before 
1 
Psi 7 p72 + y)<s. 
one Pi 
The rule at which we arrive when we put y= ace viz. 
by ft 
bin 3 
is satisfied for substances for which 4 is invariable. Then of course 
b, N + jl = _ 
——_=1 and f= Ber = 1. For all other substances 7 > 4, and 
lim 
——— > 1; the first member of the equation, viz —— is then, of 
3 lim 
course, also always greater than 1. Later on we shall set ourselves 
the task to inquire into the theoretical reason for this relation. But 
for the present we shall accept it as perfectly accurate, and see to 
f ee i 
what conclusions it leads. If we write =r s’ for f—1, we get: 
by q oes \2 
= SS Sasi (a 
Dim 64 8 
2 3 ki by 
The value of s can, therefore, not be smaller than De For ae in 
Nim 
Beso: for sf == 18 = a ie = 3,77; a. value which moreover already 
s° 64 - by 
follows from the equaliy ——- = — put above. For — =83, to 
“ f—1 21 bin 
3 be t 8 fe » 
which f==10 would belong, s would be = —V3, or s= 4,62. 
But so high a value of f or s has only seldom been found. If 
before in the absence of a leading idea, I assumed a still greater 
nd 
ratio for —~, this was a mistake. 
lim 
From: 
follows 
b, fe 9 /sr Olim . 
i F(a) tee 
biim v l im 8 5 
Of course we find back the rule from which we have started 
but with a determined value for the factor ¢. As I showed before 
