GRAVITATION AS AFFECTED BY TEMPERATURE. 
359 
case when m, m have the form of chains. But a low vacuum of I mm., or greater, 
acts well, especially with the work performed between 9 to 5 at night. 
It is important to attain the best conditions in the action of all these gas factors 
(which appear to be mutually incompatible). 
The accompanying graph shows gas pressure as abscissae on a logarithmic scale, 
while the ordinates indicate trouble due to any factor at any pressure. 
The proportions are conjectural, depending on the conditions of any particular 
experiment. Convection is roughly proportional to pressure down to a vanishing 
quantity below 1 mm. Damping, unlike convection, is nearly independent of pressure 
to about 100//. It is represented below the axis of abscissae as being a blessing 
rather than an evil in these experiments. I have worked in these experiments chiefly 
at about 14 mm. and about 20m/z (he., 20 x 10 _6 mm.). It will be seen that these two 
are favourable regions. Poynting and Barlow (P. and B.) used a very high vacuum 
in their research on recoil from light. Braun (B.) in the determination of the 
Newtonian constant had pressure of 4 mm., while Nichols and Hall (N. and H.) 
worked at 16 mm. in their research on radiation pressure. 
As distinct from the above we have Poynting and Phillips (P. and P.) working in 
their weight experiments at 16// which is unfavourable, as being in the thick of the 
radiometer region, and although convection which they specially feared would at that 
pressure be negligible deocclusion for the temperature 100° C. might be considerable. 
3. Radiation Pressure .—This is yet another factor which must be borne in mind. 
The practice is to maintain the lamp for illuminating the telescope scale steadily 
alight before and throughout the experiment and to keep it practically constant in 
position. No other lamp comes within view of the interior of the apparatus. 
Moreover, for the elimination of this and of the gas actions, the walls of the vacuum 
are maintained uniform in temperature. Taking these precautions it is expected 
that the radiation pressure on the mirror of the suspended system will be constant 
throughout and will thus introduce no error. 
Besides the many precautions mentioned, there are others incidental to the use 
rather than the design of the apparatus and will therefore appear in the detailed 
account following. 
V. Early Experiments. 
The Cavendish experiment has been developed to high excellence, but there is no 
previous investigation on the temperature effect to indicate the best way to avoid the 
many troubles sure to beset the investigation. The early methods, briefly mentioned 
below, failed for reasons given in each case, but they provided useful experience :— 
(l) The first apparatus used was made of brass of special purity. The large 
masses were rods of lead coated with gold and the small ones were of purest gold 
wire. These were all hung together in the brass vacuum vessel. This form was 
3 C 2 
