1672 
the proportion of argon, but the proportion of oxygen obtained 
agrees well with Baty’s analyses. 
The results obtained with liquid air (from five series of obser- 
vations) are contained in the following table: 
t') J) T 
I 79,52 0,083960 21,028 
ll 79,55 0,03957 21,030 
106 79,60  0,03994 __ 21,028 
Iv 79,60  0,03980 21,026 
V 79,59 0,03989 _ 21,028 
Mean 79,57 + 0,03976 21,028 
To the temperature of 79°.57 corresponds a density of 0.909 and 
a composition: 35°/, O,, 63°/, N,, 2°/, Ar. 
For the calculation of the coefficient ZL’ by means of equation 
(28’) the time of swing 7, of the system is needed when not 
subjected to any friction at all; this time of swing is not identical 
with that given in the last line but one of the table in IV $ 5, 
because in that measurement the sphere was at normal temperature, 
whereas here 7’, refers to the system with the sphere at the tem- 
perature of liquid air and must therefore be smaller on account of 
the smaller moment of inertia of the sphere*). It would have been 
possible by a reduction of the moment of inertia to 80° K. (which 
will in any case be needed presently) to calculate from the former 
value the value of 7’, corresponding to the actual experiment, but, 
on account of the changes referred to in the time of swing from 
one day to the next, this result would still have been comparatively 
uncertain, especially with a view to the smallness of 7-—7,; it 
thus seemed preferable to determine 7—-7’, directly. We have tried 
to do this by tirst making a determination of 7’ with the sphere 
submerged in the liquid air and then drawing out the liquid, until 
the sphere was just above the liquid surface, and in that condition 
1) The temperatures as given are nol the observed temperatures, but the boiling 
ELT 
points reduced to normal pressure (‘skin in 0°,012 per ofem. meron ; the 
ap 
correction applied in this manner was inside the limits of the possible errors of 
observation in the measurement of the temperatures. 
2) From a cause which is not fully explained yet the values of 5 in the different 
series of observations mutually differ by more than was to be expected from the 
degree of accuracy of each measurement. 
5) It is true that the times of oscillation as given in the table are not the same 
as they would be in the absence of friction, but to all probability the differences 
are inside the limits of accuracy of the observations, 
