THE AID OF THE ACHROMATIC FRINGES. 



107 



a pressure gradient, would fall under the same objections. I can only conclude 

 vaguely, therefore, that in some way the local vortices evoked by therma- 

 distribution resolve themselves into a persistent ordered rotation of the cylin- 

 der of liquid around its vertical axis, with the regressive motion specified 

 confined to one or two relatively thin layers. In other words, the conditions 

 of hydrostatic equilibrium imply an inclined surface of the liquid, with its 

 maximum head in the region of the illuminated part. But such a structure, 

 with its forces oblique to the surface, is gravitationally unstable. It is difficult 

 to see, however, why the flow which must result should be an orderly rotation 

 of nearly the whole cylinder of liquid. 



80. Separated Jamin plates. In the course of my work the desirability of 

 an interferometer like figure 104, in which the mirrors M, M', N, N', are par- 

 allel (all but M' being half-silvered) and an auxiliary mirror, m, m' , has often 

 presented itself. In such a case the observer at the telescope at T can control 

 the slit, being within reach of the collimator at L, and at the same time con- 

 trol the micrometer- screw at n, normal to A r . If this screw is insufficient, 



or 



71" 



104 



' ; v//-ip/ 

 <Jf*J 



tf 



c/t 





/. 



'/ ^. 



N. 



105 



^ 

 c> j 



106 



since the auxiliary mirror is made of two parts, m and m', one of these may be 

 on a normal micrometer- screw n' '. There is no difficulty in obtaining fringes 

 when m, m', is removed and observation made at T'; or at least the difficulties 

 may be overcome if M and N f are parts of the same sheet of plate glass, as in 

 figure 105, the parallelogram of rays having been adjusted for a suitable angle 

 (cf. Carnegie Inst. Wash. Pub. No. 249, Part II, 60). In a test made at 

 random, this gave the spectrum fringes and the achromatics superbly. To 

 equalize the glass-paths in the presence of m, m', in order that the achromatics 

 may appear in the minimum number, the half-silvered plate may be set 

 (all at 45) as indicated in figure 104, where each component ray traverses 

 the plates of equal thickness four times. In the absence of m, m', the silver 

 faces of M', figure 104, should be reversed, so that each ray penetrates the 

 glass twice, as shown in figure 105. If this is not done, special glass compen- 

 sators will have to be used, which usually are an annoyance. 



The case of figure 104, however, is apt to resist the endeavors to find either 

 the spectrum fringes or the achromatics, even if N' is made continuous as in 

 figure 1 06. In each case, on reversing M' and observing at T r , the fringes 

 will be found at once. The reason for this is not far to seek. True, there are 

 supernumerary rays, quite vivid and therefore giving a strong spectra. If the 

 position for fringes were found, they would not, however, escape detection 

 for this reason, and I have often seen good fringes under worse conditions. 



