from their other experiments contains a short age of CaO compared with theSiOj 

 content. Moreover the vaporization was very weak. In their experiments the vapo- 

 rization of KjO became only detectable after addition of 50 "y GaO. We, on the 

 contrary, worked at much higher temperatures, so that, though using silicates 

 that contained little CaO, we found still vaporization of the alcaline. But this becomes 

 more difïicult according as more alcaline has been liberated from the glass ; this is 

 ot great importance for the discussion of the process mentioned sub D. 



D. Reaction betineen the tungsten and the glass. 



We have seen, that from the glass a metal is liberated soluble 

 in water and wliicli can easily be sublimated. This can only be 

 potassium or (and) sodium. When the air has been let in, hardly 

 anything- but K^O, Na^O and WO^ can be found on the wall by 

 means of the flame-spectra and microchemically. 



We have tried to separate the alcali-metal from the alcali-oxide 

 formed by the fractional vaporization of the glass by sublimating 

 in vacuo the metal by heating to a separate colder spot (a glass 

 appendix made for that purpose). This proved however to be 

 impossible. The alcali-metal is vaporized, but for the greater part 

 not condensed again. It is solved in the rest of the precipitate. 



Thus we were compelled to apply tlie following indirect method 

 of determination. 



into a bulb A a filament (fig. 3) had been sealed. Over the TF-wire a glass 

 capillary had been slipped. The lamp carried two side-tubes, one of which a 

 contained a small bulb g (with capillary terminals) free from gas and filled 

 with water, while the other one b conducted via a ?7-shaped tube h, immersed in 

 liquid air, and a stopcock d to the TöPLER-pump i, to which also a GAËDE-pump 

 could be connected for the evacuation. 



First the lamp A is burnt in vacuo, while Ji is immersed in liquid air, so that 

 A is free from any vapours. Then, d being shut, the lamp is loosened at the ground 

 joint e and the small bulb g in a is shaken to pieces. Now the developed hydro- 

 gen acts on the alkali-metal, by which action hydrogen is liberated. Afterwards 

 the apparatus is built up again, h immersed in liquid air, and by means of the 

 TöPLER-pump the developed hydrogen is pumped to the calibrated volume f. By 

 means of the micro-analysis-apparatus described in another paper ^) the volume is 

 measured and the purity of the obtained hydrogen is controlled. 



Experiment 1. Developed 412 mm^ gas. 



Analysis 98 o/g hydrogen. 



This quantity of hydrogen corresponds to about 1 mg tungsten. 



In fact a quantity of the order of 1 mg tungsten was found on the wall by 

 the colorimetric method. 



It is of course necessary that the water in the bulb a has been freed first in 

 a high vacuum from solved gases. This is done most easily by connecting 

 beforehand the small bulb with a high-vacuum-pump, freezing the boiled water 



>) L. Hamburger and VV. Koopman, Chem. Weekbl. 14, 742 (1917). 



