of Metallic Spectra in the Ultra- Violet. 345 



as fiducial marks, by means of which wave-lengths of spectral 

 lines on the plate can be calculated. In the case of diffraction- 

 spectra obtained by deflecting a bundle of parallel rays at the 

 angle of incidence, *, with a deviation of order n, An is con- 

 nected with the wave-length X, and with a certain constant, a, 

 of the grating by the formula 



_ . An (. An\ 



2a sin -^- cos I i 5- \=n\. 



It is evident that at least two errors can arise in the use of 

 this formula ; one from defective graduation of the circle of 

 the spectrometer, another from the process of referring from 

 the photographs of the slit on the plate to the photographs of 

 the metallic lines. 



We select the work of Hartley and Adeney* as perhaps the 

 best type of this method of using a camera with a spectrometer. 

 Their work is characterized by great care and thoroughness, 

 and no one could probably attain better results by the use of 

 a flat grating, with its concomitants of collimator, photo- 

 graphing-lens, and camera. These observers state that they 

 were not troubled by the underlying spectrum of a higher 

 order than that which they photographed, for it was not 

 brought to a focus with the latter. In the new method we 

 propose to illustrate, all the spectra are in focus together, and 

 this fact, instead of being an obstacle, can be turned to great 

 advantage. In the absolute measurements of the wave-length 

 of light, the spectrometer method with eye-observation and 

 with a micrometer is unquestionably more accurate than any 

 photographic method. We have in this determination to deal 

 with comparatively large quantities, and with well-defined 

 directions, which can be made to coincide with optical axes 

 of the instrument ; this is not the case, however, with the 

 majority of the spectral lines on a photographic plate placed 

 in the camera, which replaces the observing-telescope of the 

 spectrometer. The photograph contains possible errors, and 

 any shifting or movement of the spectrometer-circle to deter- 

 mine intervals on the photographic plate is apt to introduce 

 other errors. 



The ideal arrangement would seem, therefore, to be a pho- 

 tographic apparatus which should remain in focus for all the 

 spectra of the different orders, in which distances between 

 successive lines on the photographs of the spectra should be 

 closely proportional to wave-lengths ; so that, the constant 

 being known for a certain position of the sensitive plate, the 



* Philosophical Transactions, clxxv. (1884) pp. 63-137. 



