682 



hut protected against contact with dust, before the observations took 



place. In botli ca«es the mercury had been distilled in vacnuni. 



With mercury that had been purified by being- shaken with 



potassium hydroxide and nitric acid and had then been dried, we 



found in the same way some days after the formation of the meicury 



surface : 



/==78°14' , // = 36°r2', 



which is in satisfactory agreement witli llie second determination. 

 The sign of (he difference between tiie vabies of / and // in the two 

 first determinations is the same as is to be derived from the tlieoretical 

 research by Drude. A surface layer greatly diminishes the angle of 

 l)rincipal incidence and enlarges the principal azimuth but little^). 

 The following observations show the influence of the layer of air 

 very clearly. The pure mercury is in a well cleaned receptacle of 

 Lkybold with plane parallel side walls, as is very often used for light- 

 filters. The observations have been made at an angle of incidence 

 of 78°, with the analyzer at an azimuth of 45°'). 



Adjustment of the 

 Compensator Polarizer 



56.46 ') 43°3:r 



Two days later, during which time the mercury was protected 



from dust, 



56.71 43°47' 



After the mercury in the receptacle had been shaken and a new 

 surface had been formed : 



56.44 43°33' 



After one day 56.64 43° 44' 



After shaking 56.42 43°34' 



Ii» a following set of observations, in which a pure surface was 

 obtained by means of the method of overflowing of Röntgen *), which 

 has also been used by Rayi.eigh, tliQ observations yielded : 



Adjustment of the 



Compensator Polarizer 



56.43 43°44' 



After two days 56.74 43°41' 



1) P. Drude, Wied. Ann., 36, 532, 865, 1889; 43, 112(5, 1891. 



') At this azimuth tlie error in the determination of the phase-ditrereuce intro- 

 duced by the metallic reflexion and of the azimuth of the restored plane polarisation, 

 whicli henceforth will be referred to as restored azimuth, is a minimum. Gf. R. 

 SissiNGH, Thesis for the doctorate, p. 57 ; Arch. Néerl., 20, p. 188. 



^) These and all the further values are every time the mean of four readings. 



*) RöNTGEN. Wied. Ann., 46. 152. 1892; Rayleiüh. Phil. Mag.. (5), 30, 398, 1890. 



