MR. W. CROOKES ON REPULSION RESULTING PROM RADIATION. 
99 
attenuated gas in the radiometer, and on consideration it seemed likely that by a 
slight modification in the shape of the vanes and screens, much that was obscure 
and contradictory might be rendered clear and harmonious. 
Instead of cups I employed hemi-cylinders for the vanes, and the screens were 
square plates of mica held at right angles to the supporting arm, so that they should 
always be in a plane parallel to a tangent to the curvature of the bulb which they 
faced. Thus any reaction which might take place between the bulb and the surface 
of the screen could have no tendency to cause rotation in either direction. 
Eig. 12. 
Fig. 12 shows the radiometer in plan and elevation, a a are half cylinders of bright 
aluminium; b b are the mica screens on an adj ustible arm, so that by tapping they 
can be shifted into any relative position in respect to the hemi-cylinders, from touching 
the convex surfaces, as shown by the black lines, to touching the concave surfaces, 
as shown by the dotted lines. The radiometer was kept attached to the pump 
during the experiments, so that the degree of exhaustion could be varied, whilst 
the distance between the screens and vanes was at the same time altered. 
409. The vanes and mica screens were set as shown in fig. 13 a, A. the screens 
Fig. 13 a. 
touching the concave surfaces. A candle was placed 3 inches from the bulb, and 
the pump was set to work. At a pressure of 540 M, positive rotation commenced 
in the direction of the arrow. A tendency to positive rotation was observed at a 
lower exhaustion, but it is best to take the first point of permanent rotation. When 
the screen was moved away from the vane, rotation stopped. 
o 2 
