THE RADIOMETER. 297 



very fine quartz fiber must be selected — the present one was quite in- 

 visible except by dififraction in sunlight. Here one must sacrifice per- 

 iod for sensitiveness. In the present work the maximum of the deflec- 

 tion was reached in 45 to 50 seconds. Since the radiation is weak, the 

 shifting of the zero was not very great while making observations. But 

 as soon as the making of observations in the region of 4.75 /x ceased, 

 the deflection moved off the scale, and had to be brought back by 

 turning the torsion-head. By using a double window of rock salt 

 (zv zv, fig. 133) some of the shifting of the zero due to changes in tem- 

 perature of the outer window was avoided. In addition to this the 

 pressure in the radiometer was kept at o.oi mm., since at the point of 

 maximum sensitiveness, 0.05 mm., heat conduction and convection cur- 

 rents were much greater. 



For the emission spectra of the metals, the radiometer vanes had an 

 area of about 2 by 15 mm. each, and the short period was due to the 

 heavy fiber suspension. The vacuum-tube radiation being very weak 

 required great sensitiveness, which was obtained in part by reducing 

 the size of the vanes (of mica) to i by 10 mm. and selecting a fine 

 fiber. The behavior of such a vane is entirely different from a heavier 

 one. At low pressures the vane before the window was suddenly re- 

 pelled from it, due apparently to the radiation through the window. 

 This repulsion occurred when both vanes were black and when the 

 unexposed one was not covered with lampblack. It was not due to 

 electrification, and throughout the vacuum-tube work it was necessary 

 to use a torsion-head (t, fig. 133) to keep the deflection on the scale. 



All greased joints were covered with beeswax and painted several 

 times with dilute shellac, which dries quickly and forms the most satis- 

 factory protection against leaking yet found. The rock-salt window 

 was secured in the same manner. The radiometer was packed in wool. 

 The window and spectrometer were inclosed in one continuous box, and 

 no radiation could enter except at the slit. As described elsewhere,^ the 

 apparatus stood in an inner room with an intervening door, which 

 could be closed at will. In the region of 4.75 /x, where the deflections 

 were large, the spectrum was explored in the daytime; but in the 

 visible spectrum, and just beyond it, the observations were usually made 

 at night, when the temperature could be more easily controlled. It was 

 then necessary to enter the room and stay a while until the heat from 

 the body established a new equilibrium. It was then possible, at times, 

 to make readings for a quarter of an hour without detecting a shift of 



^ Phys. Rev., 16, p. 35, 1903. 



