1067 



vapour bubbles are developed in it. When (lie voltage is kept con- 

 stant, tills ceases after some time. When the voltage is raised further, 

 the phenomenon of the boiling is repeated ; more of the gray glass 

 layer is vaporized. By this vaporized material the bulb wall becomes 

 hot, especially the parts opposite to the wire. Now this material is 

 seen to vanish gradually from the hottest spots and to be condensed 

 in the colder ones, where it is observed as a condensate oï metallic 

 drops. By local heating to some hundreds of degrees these may be 

 sublimated again. 



By later experiments we have found that the metal drops are able to attack 

 momentaneously the water in which the formed compound is solved. 



When the tension is raised still further the whole lamp becomes 

 so hot, that the above mentioned metal drops are all vaporized, 

 at least they vanish. On cooling down the tube they are however 

 not found back. They are solved colloidally in the precipitate on 

 the bulb; this lias obtained a light-brown coUuir. 



The condition.s under which we made these experiments were such that the 

 quantity of glass was so great compared with that of the tungsten that not all 

 the glass could be removed. This was however principally due to the following 

 cause. When the tempeiature of the wire is sufficiently liigh the glass on the 

 horizontal wire forms small spheres, which remain a long time at rest. When 

 however the temperature of the wire is raised still further, these molten spheres 

 become very mobile. They dance on the wire as if the spheroidal etïect played a 

 part and finally move towards the colder spots. These are found where the wire 

 is carried by the supports. Through these much heat is namely conducted away, 

 so that the temperature of the wire is lower at those spots. From this moment 

 hardly any glass is vaporized 



When now on one hand the precipitate on the wall is examined 

 and on the other hand the residue fin the neighbourhood of the 

 supports) on the wire, the different experiments give the results 

 collected in the following table. 



(See table following page). 



From this we may doubtlessly conclude that two processes have 

 taken place. 



a. Fractional distillation of the glass. 



b. Decomposition of the alkaline-oxide in the glass by the tungsten of 



the wire {W -\- '6 t/\,= WO, + 6 5)- 

 Na,0 Na 



C. Fractional distillation of the glass. 



Let us consider this process a more closely. Then it will no longer 

 astonish us. It confirms the view, that the glass is an undercooled 

 fluid, in which a great numbei' of substances (more or less volatile; 



