PHYSICS. 403 



This theoretical hmit can be reached, however, only on the condi- 

 tion of an extraordinary degree of accuracy in the spacing of the lines. 

 Several methods for securing this degree of accuracy have been 

 attempted, but none has proved as effective as the screw. This must 

 be of uniform pitch throughout and the periodic errors must be extremely 

 small. For a short screw, for example, one sufficient for a grating 2 

 inches in length, the problem is not very difficult, but as the length of 

 the screw increases the difficulty increases in much more rapid propor- 

 tion. It was solved by Rowland in something over two years. 



Since this time many problems have arisen which demand a higher 

 resolving power than even these gratings could furnish. Among these 

 is the resolution of doubles and groups of lines whose complexity was 

 unsuspected until revealed by the interferometer and amply verified 

 by subsequent observations by the echelon and other methods. 



Others that may be mentioned in this connection are the study of 

 the distribution of intensities within the spectral "fines" ; their broaden- 

 ing and displacement with temperature and pressure; the effect of 

 magnetic and electric fields, and the measurement of motions in the 

 line of sight, as revealed by corresponding displacement of the spectral 

 lines in consequence of the Doppler effect. All of these have been 

 attacked with considerable success by observations with the echelon, 

 the interferometer, and the plane-parallel plate. These methods have 

 a very high resolving power, but labor under the serious disadvantage 

 that adjacent succeeding spectra overlap, making it difficult to inter- 

 pret the results with certainty. 



Some twelve years ago the construction of a rufing engine was under- 

 taken with the hope of ruling gratings of 14 inches — for which a screw of 

 something over 20 inches is necessary. This screw was cut in a specially 

 corrected lathe, so that the original errors were not very large, and 

 these were reduced by long attrition with very fine material until it 

 was judged that the residual errors were sufficiently small to be auto- 

 matically corrected during the process of ruling. The principal claim 

 to novelty of treatment of the problem fies in the apphcation of inter- 

 ference methods to the measurement and correction of these residual 

 errors. For this purpose one of the interferometer min-ors is fixed to 

 the grating carriage, while a standard, consisting of two mirrors at a 

 fixed distance apart, is attached to an auxifiary carriage. When the 

 adjustment is correct for the front surface of the standard, interference 

 fringes appear. The grating carriage is now moved through the 

 length of the standard (0.1 mm. if the periodic error is to be investi- 

 gated; 10 or more milfimeters if the error of run is to be determined) 

 when the interference fringes appear on the rear surface. This opera- 

 tion is repeated, the difference from exact coincidence of the central 

 (achromatic) fringe with a fiducial mark being measured at each step 

 in tenths of a fringe (twentieths of a light-wave). As a whole fringe 



