1919] on Energy Distribution in Spectra 523 



the light in a curious manner, and apparently an irregular one not 

 following any simple law. The sensitivity of the eye is also different 

 for different colours. 



When the line is outside the visible spectrum, in the infra-red or 

 dark heat region, measurements of intensity can be made with some 

 accuracy by a thermopile or a bolometer. But they are needed 

 more urgently in the visible region at present, not only for the infor- 

 mation they will afford regarding the nature of the atom, but also 

 for application to other problems. The subject is very important, 

 for instance, in the interpretation of celestial spectra, and more par- 

 ticularly those spectra of great complexity and variability which are 

 associated with the birth of new stars, from which most of our 

 knowledge regarding such stars must be constructed. 



Previous knowledge of changes in spectral intensity under vary- 

 ing conditions was of necessity limited to the great changes. Those 

 changes which are of especial value in connections such as I have 

 mentioned are liable to be of a less conspicuous type, not readily 

 capable of detection by the ordinary photographic or the visual 

 method, and if detected, not capable of accurate measurement. 



In the visual region of the spectrum, observations with the bolo- 

 iiaeter are not satisfactory. The source of light must be very intense 

 in order to produce large deflections in the galvanometer, and only 

 the brightest lines could be dealt with even in this way. Only one 

 line in the spectrum can be experimented upon at a time, and the 

 source of light cannot be maintained constant over a protracted 

 period. The method is in fact quite unsuitable, and the spectro- 

 photometer has been tried instead, but no great accuracy is possible, 

 -and its use is confined to a very narrow region of wave-length. 

 Moreover, the variability of the source of light is again present. 



In adopting any photographic method for quantitative work, we 

 must remember that not only does the sensitivity of the plate vary 

 with the wave-length, but also that there is no very definite relation 

 between the density of a photographic image and either the intensity 

 of the light or the time of exposure. If we halve the former and 

 double the latter, we do not get the same density of the image, but 

 another which depends on the particular plate used. The grain of a 

 plate also scatters light, and the critical size of the image thus 

 depends on the exposure and the intensity of the light. We were 

 early compelled to conclude that accurate measurements of intensity 

 by a photographic method involve the necessity of an equal exposure 

 on the same plate for all the sources of light to be compared, and 

 the method to be described satisfies this necessity. 



The spectrograph for producing and photographing the lines of a 

 spectrum is set up in the usual way, which requires no description. 

 A wedge of neutral-tinted glass, cemented to another of clear glass 

 so as to form a plane-parallel plate, is mounted in front of the slit. 

 The image of the slit formed by light of any wave-length is thus 



