(817 ) 
contraction of the oxygen on solidification it is seldom that one does 
not see some empty space in the ellipsoid; the operation must then 
be repeated several times, since the oxygen that is still liquid at this 
temperature has a pretty great viscosity and flows with difficulty 
from the tube; we shall return to this point later. When the ellipsoid 
is completely filled with solid oxygen the adjusting tube may be 
lowered right down. A mark is made beforehand, so that the ellipsoid 
may be accurately adjusted to the centre of the gap when the 
silvered tube is again in its place. 
Errors, corrections, auxiliary measurements. 
1. Couples due to inhomogeneity. As will presently appear, we made 
measurements not only in liquid hydrogen (solid oxygen), but also 
keeping everything else the same, at two temperatures in a bath of 
liquid oxygen (i.e with the same ellipsoid of liquid oxygen). Since 
the susceptibility of the liquid oxygen was known, we had therefore 
two measurements of the couples due to inhomogeneity as a function 
of the field; they are given in Table VII. As a result of the some¬ 
what smaller dimensions of the ellipsoid, these corrections are com¬ 
paratively much less important. 
2. Purity of the oxygen. The oxygen was freed from nitrogen by 
vaporizing a large quantity of impure liquid oxygen under reduced 
pressure. 
3. Density of the solid oxygen. We have already mentioned the 
difficulty of completely filling the ellipsoid with solid oxygen. On 
account of the opaqueness of the oxygen that has already solidified 
one cannot with certainty assert that this condition has been fulfilled *). 
Since the specific susceptibility is determined from a known 
volume this error would have immediate effect upon the result. 
We tried to eliminate this error as well as possible by deter¬ 
mining the density with the same ellipsoid by filling it with solid 
oxygen under the same circumstances as those obtaining in the 
experiments and then measuring the quantity of gas formed from 
it on vaporization. We may assume that the small cavities that may 
form are pretty much the same in the various cases. Indeed, from 
two similar measurements the density measured in this way was 
found to vary by only about l 9 / 0 . By taking as the mean density 
that determined by these experiments, the eventual presence of 
cavities is allowed for. In this way we obtained 
p = 1.41. 
The absolute values of the couples due to inhomogeneity of the 
When there is an empty space af a few mm 3 , however, it can be seen quite 
well. 
