GRAVITATION AS AFFECTED BY TEMPERATURE. 
371 
for several days after the rough handling incidental to the adjustments. Since the fibre 
has a factor of safety of only 50 per cent, or so, great care is taken to avoid any jolts 
to the apparatus; the movements given to the beam are due not to jolts, but to the 
occasional contact of the torsion system with the vacuum tube. 
A drift of zero is found to occur, perhaps due to small elastic after-effect following 
the recent heating. I have found that in some unsymmetrical systems in vacuo , 
the zero never comes to rest, since the violent tremors set up by recurrent outside 
vibrations cause either constant strains in the fibre, or constant unbalanced pendular 
oscillations. However, under favourable conditions, readings can commence in 
three days. 
The next step is to ascertain the best angle at which the turn-table must be set for 
Fig. 11. This plan section shows the disposition of the large masses in the A and B jiositions. 
maximum deflection of the beam in the A and B positions (see fig. ll). For spherical 
masses the formula given by C. V. Boys* is very useful, viz. : — 
cos 2 0 + a +C cos 6 — 3 = 0, 
ac ■ 
where 0 is the angle for maximum attraction, 
c is distance between centre of beam and that of large mass ; 
a is half the length of the beam. 
Determine 6 in terms of a and c. But as it is most important both for sensitiveness 
and accuracy to have the exact positions of maximum attraction, the matter is settled 
by trial, and when the best positions on the dial of the swing table are found stops are 
placed so that in the coming experiment the same angle is used in every case. 
Before commencing actual readings, it must be decided whether or not the carbon 
tube is to be used cooled by liquid air. In the earlier experiments, liquid air, boiled 
under reduced pressure to give a temperature of —200° C., was used. Later this was 
given up, as the very exhausted carbon at ordinary temperature acted well enough. 
* ‘ Proc. Roy. Soc.,’ 46, 1889. 
