( 348 ) 



saturated sodium vapour is about — -— of that of tlie atmospheric air 

 at O"* and 76 cms. 



Observations. 



If we now regulate the intensity of the current in the furnace in 

 such a manner that the thermo-couple indicates a steady temperature 

 (in many of our experiments 390"), then within the tube the density 

 of the vapour is not everywhere the same, to be sure, for the 

 temperature falls from the middle towards the ends, but since the 

 surfaces of equal temperature are practically perpendicular to the 

 beam of light, all rays pass nearly rectilinearly through the vapour. 

 Accordingly the spectrum is only little changed ; the two Z)-lines 

 have become somewhat stronger, which we shall, for the present, 

 ascribe to absorption by the sodium vapour in the tube. 



We now blow a feeble current of aii' through the tube A which 

 thus is slightly cooled, so that sodium condenses on it, the vapour 

 density in its neighbourhood diminishing. We soon see the sodium 

 lines broaden considerably^ This cannot be the consequence of in- 

 creased absorption, since the average va]:»our density has decreased 

 a little. The reason is that rays of light with very great refractive 

 indices are now bent towards cf (fig. 3), rays with very small indices 

 towards q; hence in the image of the slit P which is formed on >S, 

 rays belonging to regions on both sides of tlie Z>-lines no longer 

 occur, while yet this image remains perfectly shai'p since the course 

 of all other rays of the spectrum has not been perceptibly altered. 

 If now at the same time the tube B is heated by a current of e.g. 

 20 Ampères, by which the density gradient^ in the space between 

 the tubes is increased, the breadth of the lines becomes distinctly 

 greater still. The heat geuerated in the tube by the current is about 

 1 caloiy per second ; it is, howevei', for the greater pari conducted 

 away to the cooled ends of the tube, so that tlie i-isc of leiiiperalure 

 can only be small. 



By switching a current key and a cock, A and 7> can be made 

 to suddenly exchange parts, so that A is heated, B cooled. The dark 

 bands then shrink, pass into sharp Z)-lines and then expand again, 

 until, after a few minutes, they have recovered their original breadth. 



P'ine and sharp, however, the lines in the transition stage are 

 only if the temperature of the furnace is very constant. If it rises 

 or sinks the minimal breadth appears to be not so small. In this 



