10 ox THE Ari'LHATION OF INTERFERENCE METHODS 



ratio of the wave-length to the width of the rectangular aperture, the above 

 formula becomes 



From this tabic it appears that if the visibility is to be estimated by observa- 

 tions with a telescope of 12 """■ aperture (or with a cii'cular apeiiure about one 

 fourth greater), an error in the adjustment of the sm-faces of a second of are 

 would produce a diminution of four or five per cent, in the visibility. Ac- 

 cordingly, if the ways on which the miiTor-cariiage moves are not tnie to this 

 degree, it is necessary to make the adjustment for every obsei-vation. 



This can be done with very great acciu-acy by moving the beam of light 

 from side to side and adjusting the niiiTor until there is no perceptible alter- 

 ation in the size of the i-ings. Since the admissible error in adjustment is 

 inversely proportional to the a])erture, the observations may be facihtated 

 by making this as small as possible if there he light to spare. This is all the 

 more necessaiy for the same reasons, if the surfaces be not true. However, 

 the error due to this source may be easily con-ected (since aU the observations 

 are affected alike) by niultij)l>ing by a constant factor. 



In ordei- that the visibihty curve may extend as far as possible, it is 

 neces.sar>' that the vapor should be very rare. Accordingly, in all but a few 

 cases to be mentioned later, the su])stance to l)e investigated was inclosed 

 in a vactnnn-tiib(> which was heated to drive off any moisture or occluded 

 gases. 



