1086 
air by hydrogen (this especially) or the reverse, caused a great 
increase in the friction’), which, as we have said, gradually become 
less again’). 
Under these unfavourable circumstances, in order to be able to 
arrive at provisional results with the apparatus as it was constructed, 
we were obliged to change our method of working to some extent 
and to demand a much smaller degree of accuracy from the results. 
The determination of the times of oscillation by registration, in 
particular (see Comm. N°. 1495, IV, § 4), was a complication dis- 
proportional to the accuracy, and could be quite adequately replaced 
by a purely chronometric determination, by means of a stop-watch, 
which showed '/,® of a second*). Further, the sensibility of the wire 
to changes of condition made it necessary that the apparatus should 
remain unchanged during a whole series of experiments, that is, 
that the cylinders should not be exchanged; as this did away with 
the use of the cylinders altogether (see Comm. N°. 1495, IV, § 5) 
they could just as well be left out*). We, therefore, continued the 
work with a constant oscillating system; in consequence of which 
the friction of the gas upon that part of the apparatus not 
immersed in liquid had to be eliminated in a different way. 
We did this in the following manner: besides the experiments in 
which the sphere oscillated in the liquid, we also made experiments 
with the sphere just above the liquid oscillating in vapour at alow 
temperature; from the knowledge of the density and viscosity of 
this vapour, by means of the formulae (24’) and (28) in Comm. 
N°. 1485, the retarding couple could be determined which the sphere 
experienced by the friction in the vapour; this couple could be 
subtracted from the total moment in the experiment in the vapour; 
the difference. we considered might be taken as giving the 
couple which the oscillating system experiences by friction in the 
experiments in the liquid. 
3. After several unsatisfactory attempts, we succeeded in 
carrying out in one day (July 12 1916) a series of reliable, and 
h Presumably a consequence of occlusion of gases by the metal wire, We have 
not used quartz wires yet, which probably would not possess this unpleasant peculiarity. 
2) To avoid further trouble from these changes we left the apparatus always 
filled with hydrogen, 
3) By determining the duration of ten oscillations T could still he determined 
to within about ‘O1 sec. 
4) The moment of inertia of the oscillating system was thus, at ordinary tem- 
perature, K=372:5 + 27-8 = 400°3. Hereby the period of oscillation becomes 
smaller than before (17-22 sec.) it is true, but that was not a decisive objection. 
