( 293 ) 



1900 we showed that in the different glowing conditions the colours 

 must succeed each other in this order. If for a given high tempera- 

 ture we accept the colour to be white, we find that with decrease 

 of temperature the colours in the triangular diagram of colours make 

 a curve which from white first goes directly to yellow of X 587 but 

 which, as the colour becomes deeper, bends towards the red and 

 corresponds to light of greater wavelength. With increase of tem- 

 perature, on the contrary, the line of colour runs from white to 

 the opposite side, to the blue of X 466. 



Because the colours which are produced by white light after having 

 been subjected to different degrees of atmospheric absorption, also 

 follow about this same line, we may expect that the colours of 

 the self-luminous celestial bodies will in general he on this line 

 or near it; they are determined on this line by one coordinate, one 

 number. This renders it comprehensible why on the one hand the 

 designation by means of letters and words, or the measurement with 

 Zöllner's colorimeter, which produces quite different colours, has 

 given so few satisfactory results, and on the other hand why the scale 

 of Schmidt, who designates the colours by one series of figures, 

 where is white, 4 yellow, and 10 red has proved to be the best to 

 work with. After this method has been drawn up the best and most 

 complete list of stellar colours, published in 1900 by H. Osthoff at 

 Cologne, in the A. N. Bd. 153 (Nr. 3657—58). This list in which 

 the colours of all stars to the S''^ magnitude are given, down to a 

 tenth class of colour, and which was the fruit of systematic estimates 

 during 14 years, enables us to accurately determine the relation 

 between spectrum and colour. 



In a former paper ^) we remarked that we did not know where 

 in the continuous series of spectra of the Oriontype and the first 

 type we have to look for the highest temperature or at any rate the 

 greatest luminosity. We may assume that it will be there where the 

 colour is whitest; the spectral-photometric measurements, to which 

 we have alluded in that paper, are still wanting, but for this purpose 

 we can also avail ourselves advantageously of estimates of colour ; 

 this has been the reason for the investigation of which the results 

 follow here. 



In this case where we required a specification of the spectra, as 

 detailed as possible, to serve as an argument for the colour, we 

 have naturally used again Maury's classes. In order, however, to 

 determine a mean colour for each class we must correct the colours 



1) The luminosity of stars of different types of spectrum. Proceedings of June 

 30 1906 p. 134. 



