( 546 ) 



of dots, while for the continuous rotation of the spectrometer an 

 intermittent one has been substituted. In this way for any recorded' 

 radiation-intensity the corresponding position of the prism can be found, 

 not by measuring abscissae, but by counting dots. Since moreover nc$ 

 only the deflections of the galvanometer but each time also the zero- 

 positions are recorded, it is possible to determine on the spectograms 

 i lie radiation-intensities also when during the observations the surround- 

 ing temperature, and consequently the zero-position, was variable. 

 The principal advantages of this method of observation over the 

 usual one are : 



1. the absolute reliability of the observations, 



2. the very short time required for a set of observations, 



3. the accuracy with which interpolation is possible when the 

 zero-position shifts, 



4. the non-existence of disturbances, caused by the proximity of 

 the observer. 



5. the complete comparability of the different observations, 



6. the possibility of estimating the probable error from the shape 

 of the zero-line. 



The short time in which a set of observations is made, is of 

 importance when e.g. heat-sources arc investigated which, like the 

 arc. show slow changes in radiation-intensity. A spectrum, ranging 

 from 0,7 to 6 fi was recorded with 200 displacements of the spectro- 

 meter in two hours. 



In the spectrograms a spectral line is represented by 5 to 6 dots. 

 With one displacement of the spectrometer namely the line is 

 shifted over a distance amounting to 2 /s °f *' ie breadth of the image 

 of the slit, or of the equal breadth of the thermopile. Hence the 

 same kind of radiation will strike the thermopile during five successive 

 displacements. From the mutual position of the dots, the place where 

 the radiation-intensity has its maximum may be accurately determined. 

 In order to derive from this the place occupied by the line in the 

 spectrum, it is sufficient to know one fixed point in the spectrum. 

 This fixed point was as a rule taken from a comparison spectrum, 

 for which the carbonic acid emission of a Bunsen flame was chosen, 

 the maximum of which, according to very accurate measurements of 

 Paschbn, lies at 4.403ft. Part of the flame spectrum was for this 

 purpose recorded simultaneously with the spectrum to be studied. 



A simple calculation then gives the refractive index for the 

 unknown ray. In order to derive from this the wave-length of the 

 line, a dispersion formula must be used. I became aware that the 



