on Spectrum Analysis. 105 



the air in this tube was probably a little contaminated with 

 coal-gas, introduced when the apparatus was last fused together. 

 After the discharge had gone on for some minutes, the pressure 

 was measured and found to be 0*0005 3. Hence we see that, 

 in spite of the persistent heating, gas still adhered to the wall 

 of the tube and was disengaged by the electricity. 



How obstinately air adheres to a glass surface is particularly 

 shown by the tube put up in the Physical Laboratory of the 

 University. This tube had been exhausted from time to time 

 for four months, each time to a few millionths of a milli- 

 metre ; yet, after some time, there was always a considerable 

 pressure. Thus, on the 8th of January the pressure was 

 0-0013 millim., on the 14th 00011 millim., on the 21st 

 0-0005 millim., and on the 22nd of February 0*0028 millim.* 

 This tube was provided with electrodes of a special kind. 

 Before the tube was closed, thin aluminium-foil was introduced 

 into the tube, corresponding to the external tinfoil coatings, 

 from which narrow strips projected along the axis of the tube 

 towards the middle, and served as electrodes. I was led to 

 this arrangement, not being able to melt the electrodes into 

 the tube ; the action may be considered as nearly the same as 

 that of the ordinary electrodes melted into the glass. This 

 tube, in fact, ceased to be luminous at a pressure of about 

 0*004 millim. and a spark-length of 5 millim. When the 

 pressure was still smaller (a few millionths of a millimetre), I 

 could not perceive any luminosity, even with the greatest 

 possible spark-length (15 millim.) ; but the luminosity com- 

 menced suddenly upon the bursting of one of the capillary 

 canals mentioned above. 



In the hydrogen-tube I obtained a very pure hydrogen 

 spectrum ; although the gas had been prepared in the usual 

 way from zinc and sulphuric acid, both free from arsenic, 

 the gas had a strong odour of hydrocarbons. The luminosity 

 began at a pressure of 30 millim. (with gas containing air at 

 43 millim.). At a pressure of 0*35 millim. the spectrum 

 showed the lines C ( = 656), F ( = 486), and 434 (united with 

 the mercury-line 436 to a band?), as well as numerous faint 

 lines (especially in the red and orange) of the second hydrogen 

 spectrum investigated by Hasselberg. The tube was luminous 

 at the highest vacuum attainable, at least at the positive 

 coating ; the light failed only in individual discharges. 

 Crookes's rays were seen even at a pressure of 0*008 millim. 



* I have not observed the phenomenon mentioned by Bessel-Hagen 

 (Wied. Ann. xii. p. 440, 1881), that the air liberated from the walls of 

 the tube by electric discharges condenses again, so that after some time 

 the pressure decreases again. 



