1883.] a cullimatiiig eye-piece in spectroscopy. 341 



easily recognised line in the spark spectrum of iron. The D lines 

 are convenient from the facility with which they may be observed, 

 and though they are diffuse lines in the sodium flame they can be 

 read accurately by the reversed line when the dispersion is high. 

 The temperature correction is an important one whenever the 

 observations are carried on in a place of variable temperature, a 

 change of 4° in the temperature of the Rowland's grating we have 

 used making a difference of more than half a minute in the devia- 

 tion of the b group in the 3rd order. Mendenhall has determined 

 the coefficient of expansion of one of his gratings by observations 

 of the deviation at different temperatures, and his result agrees 

 very nearly with the coefficient of expansion of speculum metal 

 observed by Smeaton; and it is sufficient, with our present in- 

 strumental means, to correct the grating space for temperature 

 by either coefficient. 



It had occurred to us that instead of measuring directly the 

 angle through which the grating: is turned we might measure the 

 deviation of the ray reflected from its surface and so double the 

 angle to be measured and diminish errors of measurement. This 

 method, however, does not answer so well as the other when the 

 measures are taken on one side only, and the reasons are obvious. 

 If the two telescopes are not very accurately adjusted to the centre 

 of rotation sensible errors will affect the measured angles ; and 

 the zero reading cannot be obtained in the same way as the other 

 readings but has to be found by placing the telescope opposite the 

 collimator and taking the reading in that position. The method 

 answers well if the readings are taken on both sides. The ad- 

 justment of the axis of the telescope to the centre is a trouble- 

 some business, and is liable to small but sensible errors in change 

 of focus for the observation of different rays. We make the 

 adjustment by first replacing the object-glass by a lens of short 

 focal length so as to make the telescope into a microscope. A 

 small plumb-line consisting of a single thread of unspun silk is 

 then adjusted on the top of the instrument until it does not alter 

 its place in the field of view when the turn-table or the telescope 

 is rotated. The cross wires of telescope and collimator are then 

 brought on to it. After this has been done the readings on the 

 two sides differ so much that the double readings cannot be dis- 

 pensed with. By taking the mean of the readings on the two 

 sides very accurate results may be obtained, as both instrumental 

 errors and errors of reading are reduced as much as possible. 

 Thus three readings of the 1) lines on each side,, with a Rowland's 

 grating on which the distance between the lines had been gauged 

 by readings of the solar lines E, gave the wave-lengths 5895*06 

 and 588905, numbers which differ from Angstrom's by only -07. 



In order to photograph lines for the purpose of wave-length 



