132 BARUS— ELLIPTIC INTERFERENCE [April 21, 



screened off, whereas the whole grating g (about one half inch 

 square) is without a screen, all the fringes from the maximum size 

 to complete evanescence beyond the range of visibility are pro- 

 ducible. Naturally if the edge of g' is quite dark everything 

 vanishes. 



It follows therefore that pairs of corresponding rays are always 

 in question. These corresponding rays are at a definite ND, apart 

 where D is the grating space and A^ the number of lines per cm. of 

 the grating in question. This distance ND is greater as the fringes 

 are smaller and may be of the order of a cm. when the fringes pass 

 beyond the range of visibility. Again ND is equal to the width of 

 the crack when the largest fringes vanish. Finally when ND is 

 zero, as in the original unbroken grating, the size of the fringes is 

 infinite. 



It has been stated that the use of the slit or a laterally limited 

 objective is advantageous because all the lines are much sharper. 

 Inert or harmful illumination is cut off. If the slit is over the 

 objective of the telescope only a small part of the field of view shows 

 the lines; if placed over the objective of the collimator, the fringes 

 are of extreme clearness throughout the spectrum. It may be ulti- 

 mately of advantage to use the edge near the crack g' only, together 

 with the whole of g. For if a small strip of g' at the crack 5" is 

 used with the whole of g, the smaller fringes are weakened or wiped 

 out. Thus the inner edge of the nearer grating with successive 

 parts of the further grating is chiefly effective in the production of 

 these interferences. 



To bring the two edges quite together was not possible in my 

 work, as they were rough and the apparatus improvised. 



7. Data. — Some measurements were attempted, with the view 

 only of checking the equations presently to be deduced. The adjust- 

 ment on an ordinary spectrometer is not firm enough and the fringes 

 being very fine (a few minutes of arc) are difficult to follow unless 

 quite stationary. 



Table I., however, gives both the values of de/dn, displacement 

 per fringe, for different angles of incidence i and of diffraction 6, 

 and dO/dii, the angular deviation per fringe at the D line. In meas- 



