570 PROCEEDINGS OP THE AMERICAN ACADEMY. 



made fast to the upper end of a b, and to the lower end of a fine glass 

 rod d x which carried a horizontal magnet m 2 . The rod d x was in turn 

 suspended by a short fiber to a steel pin e, which could be raised or 

 lowered in the bearing h. The whole was carried by a bent glass tube t, 

 firmly fastened to a solid brass foot F, resting on a plane _ground-glass 

 plate P, cemented to a brass platform mounted on three levelling screws 

 not shown. A bell-jar B, 25 cms. high and 11 cms. in diameter, covered 

 the balance. The flange of the bell-jar was ground to fit the plate P. 

 A ground-in hollow glass stopper fitted the neck of the bell-jar, which 

 could thus be put in connection with a system of glass tubes leading to a 

 Geissler mercury pump, a MacLeod pressure gauge, and a vertical glass 

 tube dipping into a mercury cup and serving as a rough manometer for 

 measuring the larger gas pressures employed during the observations. 

 The low pressures were measured on the MacLeod gauge in the usual 

 way. A semicircular magnet M, fitted to the vertical curvature of the 

 bell-jar, was used to direct the suspended magnet m 2 and thus to control 

 the zero position of the torsion balance. By turning M through 180°, 

 the opposite faces of the vanes G and D could be presented to the light. 



The Arrangement of Apparatus. 



A horizontal section of the apparatus through the axis of the light 

 beam is shown in Fig. 3. The white-hot end of the horizontal carbon 

 Si, of an A. T. Thompson 90° arc lamp, fed by alternating current, 

 served as a source. The arc played against the end of the horizontal 

 carbon from the vertical carbon which was screened from the lenses L x 

 and L 2 by an asbestos diaphragm d 2 . A lens, not shown, projected an 

 enlarged image of the arc and carbons on an adjacent wall, so that the 

 position of the carbons and the condition of the arc could be seen at all 

 times by both observers. 



The cone of rays passing through the small diaphragm d 2 fell upon the 

 glass condensing lenses L x , L 2 . At d z a diaphragm, 11.25 mm. in diam- 

 eter, was interposed, which permitted only the central portion of the 

 cone of rays to pass. Just beyond d 3 , the beam passed to a shutter at S 2 . 

 This shutter was worked by a magnetic escapement, operated by the 

 seconds contact of a standard clock. The observer at T x might choose 

 the second for opening or closing the shutter, but the shutter's motion 

 always took place at the time of the seconds contact in the clock. Any 

 exposure was thus of some whole number of seconds' duration. The 

 opening in the shutter was such as to let through, at the time of expo- 

 sure, all of the direct beam which passed through d 3 , but to shut out 



