( 20 ) 



parallel light (which was necessary also now) have been described 

 already more than once. ^) 



The light from an arc-lamp or from the sun passed successively 

 through a aS^icol's prism, whose plane of vibration was inclined at 

 an angle of 45'" to the horizon, the magnetic field with its lines of 

 force normal to the beam, a second Nicol at right angles to the 

 first. Betw^een the Nicols the Babinet compensator was placed, the 

 edges of the two prisms being horizontal. An image of the com- 

 pensator was formed on the slit of the spectral apparatus; in the 

 middle of this image the central dark interference fringe, surrounded 

 by the coloured ones, w^as seen. In the spe(;triim a ])air of dark 

 interference fringes are observed and with the field off, only the line 

 absor})tion lines of the vapour are seen. Generally the reversed 

 sodiumline is observed already in the spectrum of the arc-light 

 itself and then the presence of sodium vapour between the poles 

 makes of course no difference at all. In order to obtain the degree 

 of sharpness of the interference fringes, necessary for tins part of the 

 investigation, we tried several compensators. Sufficient results were 

 obtained with a Babinkt compensator of which the prisms had angles 

 of about 50', obtained from I Ik» firm Stp:kü & Reutkk. 



The light passed the flame (a gas flame fed with o.xygen) ovei' 

 a length of nearly i^/^ cm. If the field had an intensity of about 

 23000 C. G. S. units, the quantitj' of sodium in the flame being very 

 small, the image observed was very similar to 

 that represented in Fig. 1. The latter is constructed 

 with the aid of photographic negatives and of eye 

 ol)ser Nations. The ^^■hole phenomenon is of course 

 \ery delicate as it only extends to the region of 

 the magnetically broadened D^ line; moreover it 

 depends very much on the (piantity of sodium 

 present. We did not yet succeed in getting negatives, 

 ^vllich showed the parts which are of very unequal 

 intensity all equally well. 

 Already some time ago Prof. Voigt was so kind 

 to inform one of us of the result, which according to his theory may 

 be anticipated in the case of a quadruplet. 



This conclusion is easily arri\ed at, if the calculation be simplified 

 by applying a certain approximation, the soundness of which cannot 

 be judged a priori, because constants appear whose numerical value 

 is not yet known. With this reservation the behaviour predicted 



Fig. 1. 



1) Zeevan I.e. and Arch. Néerl. (:>) 5. 237. J 900. 



