232 PROFESSORS G. D. LIVEING AND .T. DEWAR 



2748-58 . .at 4th order. 



2748-27 . . 



2748-20 . . 5th 

 2748-16 , 6th 



Mean . 2748'30 



Six pairs taken with a ROWLAND grating and the same goniometer gave us 



2748*43 . . at 3rd order. 



2748-41 . . ,-, 4th 



2748-50 . . 5th 



2748'35 . . 6th 



2748-41 . . 7th 

 2748-29 , 8th 



Mean . 2748'40 



BELL'S* value of the wave-length of the same line is 2748'45. The photographs 

 were all taken with a collimating eyepiece. Considering that the cadmium lines, 

 though strong, are very diffuse, so that there is room for considerable error in esti- 

 mating the centre of the photographic image, and that an error of 0'013 mm. in 

 measuring the distance between the two images of the line on the photograph would 

 make an error of 1" in the deviation, or very nearly "03 in the value of the wave- 

 length, the values above are fairly concordant. However, besides the difficulty of 

 exact measurement arising from the ill-defined character of the cadmium lines, there 

 is a source of error arising from the gratings themselves, and affecting both of them. 

 It is that the two images of any given line, on the two sides of the normal to the 

 grating, are not in focus at the same distance from the object-glass of the telescope. 

 Hence, if the photographic plate is adjusted to the focus of a particular line when the 

 grating is turned in one direction, it will not receive a sharp image of the same line 

 when the grating is turned in the other direction. 



With a collimating eyepiece it is comparatively easy to place the grating so that 

 the normal to its plane may coincide with the axis of the collimator. Hence, fair 

 measures of the wave-length of any line may be obtained from the deviation for that 

 line, as measured on one side only of the normal. But, since every change of focus 

 makes it necessary to take a fresh reading of the position of the grating when it is 

 normal to the axis of the collimator, it is better, as a rule, because errors of reading 

 thereby become of less consequence, to measure the angle between the two images, 

 one on one side and the other on the other side of the normal. If this is done, the 

 difficulty arising from difference of focus on the two sides must be faced, as any 

 alterations of adjustment between the times of taking the pair of photographs would 

 be wholly inadmissible. 



* ' Amer. Journ. Sci.,' vol. 31, 1886, p. 429. 



