572 



PROCEEDINGS OP THE AMERICAN ACADEMY. 



of the rotation axis, as well as the positions of the images of d z , when the 

 lens L 3 was against the stops. For the latter measurements, the vanes 

 could be moved out of the way by turning the suspension through 90° 

 by the control magnet M (Fig. 2). 



To make sure that the balance as used was entirely free from any 

 magnetic moment or disturbance, the small magnet m 2 was clamped in 

 one position to maintain a constant zero, and the period of the balance 

 was accurately measured with the axis of the large magnet M in the ver- 

 tical plane of the vanes and again when the axis was at right angles 

 to the plane of the vanes. Several series of this sort failed to show a 

 difference of 0.1 second in the period of the balance for the two positions 



of the magnet. 



Togalv. G % 



To galv. 



Figure 4. 



The bolometer at R (Fig. 3) was of sheet platinum 0.001 mm. thick, 

 rolled in silver. The strip was cut out in the form shown in Fig. 4, and 

 mounted on a thin sheet of slate S. Two windows had been cut in the 

 slate behind the strips at A BOD where the silver had been removed 

 leaving the thin platinum. The platinum surfaces were blackened by 

 Kurlbaum's process. The image from Z 5 (Fig. 3), fell at D. The sil- 

 ver ends between A and C were connected with E and F respectively. 

 On the heavy wire EF a sliding contact c served to balance the bridge, 

 all four arms of which are shown in the figure. 



Methods of Observation. 



The observations leading to the results given later were of three differ- 

 ent kinds : (1) The calibration of the torsion balance; (2) the measure- 



