REVERSED AND NON-REVERSED SPECTRA. 47 



to have coalesced in a way which defies description. In their place appeared 

 a wide strip of equidistant parallel crescents, as shown in figure 28. The 

 Fraunhofer lines had long vanished and the appearance of the spectrum was 

 whitish and intense. The fringes in question are thus in a measure achro- 

 matic. The strips appear quite regular through the breadth of the spectrum 

 and its width may be one-third of the length of the spectrum. The fringes 

 move with the normal displacement of M (interferometry) and the range is 

 large (0.5 cm. without adjustment), provided M does not require readjust- 

 ment by rotation. Simultaneously the strip is displaced longitudinally in 

 the spectrum in the usual way. 



On closing the slit the ellipses break up into sharp strands again without 

 offering a systematic clue as to the manner in which this is done. The strands 

 usually trend more or less vertically with two sharp, strong groups, flanked 

 by one or more weak groups on each side. 



On removing the condenser these crescents became more slender but much 

 sharper, so that in spite of the diminished light they could be well seen. They 

 were then found to be like the approximately confocal ellipses of displacement 

 interferometry, though not subject to the same laws. They embraced over 

 one-third of the visible overlapping (green -yellow through red) spectra, ter- 

 minating in very fine hair-lines on one side but coarse lines on the other. 

 On opening the slit to about o.i mm. the evolution was curious. With a 

 very fine slit a relatively narrow strip of strong slanting lines was seen in 

 the yellow. As the slit widened these developed curvature, adding the more 

 slender complements of the ellipses on the red side, until this part of the spec- 

 trum was filled with confocal half-ellipses having a transverse major axis. 

 The range of displacement of M is practically indefinite, depending simply 

 on the degree to which the spectra overlap; 3 to 4 cm. were tried. A hori- 

 zontally wide mirror at M is needed ; for the ellipses travel through the spec- 

 trum and the pencil along the mirror, from end to end. Both sides of the 

 ellipses may be traversed by rotating the plate compensator, which succes- 

 sively accentuates (in a transverse strip) a definite part of their contours. 

 In this way the thick apices or either of the hair-like lateral ends may be 

 clearly brought out. Thus the two lines a, in figure 28, limit the strong part 

 of the ellipses. When a moves to right or left, the hair-lines appear more 

 and more strongly until they terminate, showing that the inclusive strip is 

 also limited. 



To further study this result, the grating* G' was successively rotated in 

 small amounts on a normal axis with adjustment at M. It was thus 

 possible to find both the ends of the ellipses, as well as the central parts. 

 As a result the form figure 29 was definitely brought out. The con- 

 focal ellipses are extremely eccentric, with very turgid apices, so that the 

 central part, if in the spectrum, consists of transverse straight lines. Motion 

 of M shifts the fringes to and from the center where they originate or evanesce. 



* When nearly centered, rotation of M about a horizontal axis is also sufficient to com- 

 plete the centering of the ellipses. 



