REVERSED AND NON-REVERSED SPECTRA. 19 



These results are correct to 0.5 per cent and are as close as the estimation of 

 p, c, &, and fractions of a fringe will warrant. If results of precision were 

 aimed at, a long tube should of course be used. What was particularly 

 marked in these experiments was the motion of fringes in the passage from 

 any approximately adiabatic to isothermal conditions and on approaching a 

 plenum of air. 



Since the refraction depends on density, there should not (apparently) be 

 any motion at all; but the thin tube is always more nearly isothermal than 

 the much larger barrel of the air-pump. As a consequence there is residual 

 expansion from the former to the latter. 



10. Continuation. Babinet compensator. The behavior of an old Babinet 

 compensator, placed nearly normal to one of the beams, figure 3, was peculiar, 

 though the fringes were clear and easily controlled. The dimensions of the 

 right-handed quartz wedge were roughly calipered and found to be: length, 

 4.2 cm.; thickness at ends, 1.017 and 0.934 cm. Thus there is a grade of 

 0.083/4.2 =0.0193, or something over i of arc. A vertical displacement of 

 2.5 cm. of this wedge was available behind the stationary counteracting 

 left-handed wedge. 



The fringes were not uniform and they required an inclination < 



to the vertical of the rulings of the grating G' . The fringes were 

 evidently curved lines, intersected by the vertical strip within 

 which they are visible. Consequently they appeared as in figure 

 8, with linear elements in the middle, shortening into dots at either 

 end of the strip. On motion of the compensator wedge they 

 moved toward or from the center of symmetry, as is also indicated g 

 in the figure. Tiled forms were frequent. The most interesting 

 feature, however, was their alternate appearance and evanescence 

 in cycles. While the wedge was moved over 2.5 cm. of its length, 

 7 of these cycles appeared and vanished, each consisting of about y 

 36 to 40 fringes. The disappearance was not always quite com- 

 plete, but the fringes could not be restored by any adjustment for 

 coincidence of spectra. 



An attempt was made to find the angle of the quartz wedge by the first 

 method. Data, 0.0023, 0.0024, 0.0024 cm., were found for the displacement 

 of the micrometer per fringe. Hence, apart from dispersion, 



- 

 a= - = 0.022 radian 



2X0.5442X0.0024 



which, as in case of the glass plate, is again slightly above the calipered value. 



In another somewhat thinner Babinet compensator the constants were: 

 length, 3.35 cm.; thickness, small end, 0.494 cm., large end, 0.496 cm. The 

 prism angle is 01 = 0.062/3.35=0.0185 radian, also about i. 



In this case there was no periodic phenomenon, but in its place the degree 

 of longitudinal coincidence of the axes of the two spectra continually changed. 



