350 Dr. B. Hasselberg on the 



pump. But to this objection may be replied, first, that if 

 this were the case the carbon-monoxide spectrum must also 

 show itself, since, as is well known, this spectrum is nearly as 

 ubiquitous in highly exhausted spectral tubes as that of sodium 

 is in flames. In all the tubes prepared by me there was, how- 

 ever, no trace of it to be seen. Secondly, it is to be observed 

 that if our spectrum belonged to acetylene proceeding from 

 such impurity, we must expect to find the same more or less 

 completely in every tube prepared with the same pump. But 

 this is not the case ; for when the tube No. 3 was filled with 

 dry air and exhausted to about 1 millim., I observed a perfectly 

 pure and beautifully developed band-spectrum of nitrogen, 

 whilst the admission of moist air after removal of the phos- 

 phoric anhydride at once gave rise to the above spectrum. 



It is therefore impossible to attribute it to any carbonaceous 

 gas derived from the pump. 



As the result of all these experiments, I confidently express 

 the opinion that the spectrum described here certainly belongs 

 to hydrogen and not to acetylene. If we examine the series 

 III/ and IV. of Table A as well as Table B more closely, we 

 find that in both cases the number of observed (that is, of 

 visible) lines is very nearly the same ; in other words, that 

 increasing the layer of radiating gas about sixty times has 

 not caused any perceptible change in the spectrum. This 

 appears to be in direct opposition to the experiments described 

 above on page 338, since the longitudinal use of the tube 

 there described showed the spectrum perfectly distinctly, 

 whilst only traces were otherwise to be seen. The apparent 

 contradiction disappears, however, at once when we consider 

 that in the latter case the temperature was decidedly lower. 

 We easily see from Zollner's well-known formula 



E={l-(l-A^e, 



which gives the intensity of the light of wave-length X radi- 

 ated from an incandescent layer of gas of the thickness 8, that 

 this intensity will be less affected by an increase in the value 

 of £ the higher the temperature is. 

 It follows from the two equations, 



^ = -elog(l-A A )(l-A^ 

 5 -=6(l-A A y-^l + Slog(l-A A )}, 



dhdA 



A 



that t<t increases with A A so long as A A <1— e *, but de- 

 ad 



