Experimental Researches on Gravitation. 499 



a way practically perfect. Even after twenty-four hours 

 the inner pressure does not rise above 7/10 mm. of mercury,, 

 which represents the vapour tension of the mastics used 

 to seal the balance. While the tests are going on, the 

 mercury's rotatory pump must be kept going to reduce 

 the pressure to less than 1/10 mm. of mercury. Under such 

 conditions the perturbations of temperature that result from 

 the mercury surrounding the V and V capsules are com- 

 pletely avoided. 



The observations are made with a beam of light reflected 

 by S on a scale at the distance of 12 m. ; it is possible to de- 

 tect 1/10 mm. on this scale. The sensibility of the balance 

 can be brought in this way to a deviation of 170 mm. for 

 the beam of light per mg. Hence it becomes possible to 

 estimate about 1/1700 mg. in a direct reading, and reach a 

 superior preciseness with many observations. 



However, a doubt arises, is not such a preciseness illusive, 

 and will not the slightest causes of perturbation hide com- 

 pletely a deviation of a few millimetres ? 1 have nevertheless 

 been able to remove all the more considerable causes of error. 

 The most important of these consisted of the external 

 mechanical shakes produced by the life of the town. I 

 avoided them either by making my observations in the night 

 or by availing myself of the days of the general strike, in 

 this respect very useful. 



Observation on the variation of weight. — The effect of the 

 presence of mercury around the ball m was ascertained, in 

 the following way. In the recipient U determinations of 



tffr. 4. 



220 



the balanceposition of rest were made rapidly and alternately 

 with and without mercury. Fig. 4 shows the diagram of one 

 of this series of observations, carried out on July 20-21, 1919. 



