1115 
ral the curve in the first graphical representation will deviate 
from a straight line. If we now connect two points from this 
graphical representation, the slope of this straight line will indi- 
cate the value of energy which belongs to a temperature which 
lies between those of the two connected points. It will then be 
clear that if we wish to determine the energy value in a similar 
way, the found value will differ the less from that which corre- 
sponds to the two observation temperatures as Xe, is smaller. Hence 
the energy value will also be found with the greater relative accu- 
racy as the energy value itself is greater, ie. the energy found 
graphically will then proportionally differ only little from the energy 
values at the observation temperatures. If we now fill in the graphi- 
cally found value in equation 12 and if we apply equation 12 to 
the two observation temperatures, a too great value for the energy 
difference will have been chosen for the one temperature, a toc 
small value for the other. For a temperature between the two 
temperatures of observation the energy value is then chosen exactly 
right; hence the correct entropy value has therefore been yielded 
by equation 12 for this temperature. Therefore when equation 12 
is used the found values of entropy will deviate somewhat from the 
real ones at the two temperatures of observation. 
If we denote the two temperatures of observation by 7 and 7, 
and the temperature for which the graphically found value of the 
energy holds, by 7, and if we imagine the value of energy found 
at 7, and the corresponding entropy substituted in equation 10, we 
may question what deviation equation 10 gives us for the values 
of K,. at the temperatures 7’, and 7’. The error made in the enerev 
c p 1 2 8) 
T. 
3 
is a=nk |, 
when we apply equation 10 as 7,, amounts to — dT, that in 
C 
T 
Ts 
dnH 
—dT. If we now consider 
the entropy-term amounts to sl 
1 
that the energy and the entropy occur with opposite sign in the 
second member of equation 10 and that 
dn re dEnH 
ANT 
we see that these two errors cancel each other for the greater part 
in the second member of equation 10, and that therefore in spite 
of these approximations a pretty accurate value of A, ean be found. 
This fact explains why notwithstanding an appreciable value of the 
dT 
