94 DISPLACEMENT INTERFEROMETRY BY 



ments were therefore made by simulating the glass cylinder GG' by a thin 

 cylindrical glass shell, closed below and above and containing a solution of 

 mercury potassic iodide, with an index at pleasure between 1.5 and 1.7. It 

 was not difficult to meet the conditions of figure 88 so far as mere refraction is 

 concerned, and certain incidental results obtained in this w r ork are given 

 elsewhere (Chapter X, 79). 



The active slit in this experiment is the image within the cylinder, G, of the 

 slit of the collimator and the former is sufficiently fine to show Fraunhofer 

 lines even when the latter is a millimeter or more broad, so that there is no 

 deficiency of light. 



But in relation to the detection of the interferences, the two reversed spectra, 

 strongly divergent in their homogeneous rays, introduced certain grave diffi- 

 culties. For it will appear that the spectrum issuing at d', figure 88, passes 

 over the distance b farther than the spectrum issuing at d, before they reach 

 the telescope together. The result is that the two spectra lie in different focal 

 planes, unless the telescope T is very remote. In other words, there must 

 be parallax between the apices of the spectrum wedges. This makes the ad- 

 justment very difficult and I failed after long searching and with many devices 

 to obtain any results, though I see no reason why the fringes should not occur. 

 Identical spectra issue at d and d'. 



To obviate this annoyance the symmetrical adjustment (Michelson and 

 Morley) with an additional mirror at d, figure 88, corresponding symmetric- 

 ally to N, and a symmetrically placed cylinder G, suggests itself. In such a 

 case the spectra lie in the same focal plane, and since they have undergone 

 two and three reflections, respectively, before reaching T, the interferences 

 of non-reversed spectra are obtained without much difficulty. In my experi- 

 ments, owing to the irregularity of the glass cylinder used, the fringes were 

 also irregular, but otherwise clear and strong, as a wide slit is admissible. To 

 adjust for coincidence it suffices to rotate the half-silver N on a vertical and 

 a horizontal axis. The fringes are modified as to size, etc., by rotating N on a 

 vertical axis and displacing it at the same time. They lie rather sharply in a 

 definite focal plane. 



The same difficulty is attached to the designs in figures 86 and 87, which 

 also give reversed spectra and a path excess at T, of b for one of them, after 

 issuing from the cylinder. 



The case of two internal reflections is complicated by the occurrence of 

 multiple images from N, figure 89, even when one side is half -silvered. This 

 is particularly the case when the cylinder G contains water, as in my first 

 experiments; for the glancing angle at b is then but 25. There is an advant- 

 age, however, inasmuch as N may be placed at a correspondingly large distance 

 from G. In spite of the duplicated images, the fringes were found more easily 

 and were less irregular than anticipated. They are liable to be reproduced 

 usually in a different size and orientation in each of the images. They will 

 be found in the colored edge and even in the white glare (caustic) which 

 emanates from the cylindrical surfaces. They could be made quite large, clear, 



