Production of Vacua, and their Spectroscopic Examination. 233 



point, or 60 absolute, gave the value of 26 mm., and a ratio of the 

 tensions of nitrogen to oxygen of 6 to 1, whereas from the curves the 

 value ought to be 6 -7 to 1. Olszewski gives the tension of nitrogen at 

 - 214 as 60 mm., and as at this temperature the oxygen tension is 

 3'8 mm., the ratio of the saturated pressures of the two gases at the 

 melting point of nitrogen would be as 16 to 1, which is far too high. 

 Probably the oxygen value will be nearest the truth, seeing it has the 

 lowest melting point. The tension is about a ten millionth of an 

 atmosphere. In the case of nitrogen, the maximum theoretical pressure 

 would be one five-hundred-thousandth of an atmosphere. It is 

 safe to infer that the vacuum left after liquefying the air out of a 

 vessel by means of liquid hydrogen cannot exceed the millionth part 

 of the atmospheric pressure, exclusive of the pressure resulting from 

 any incondensable material other than nitrogen and oxygen. This is 

 just about the pressure of the vapour of mercury at the ordinary tem- 

 perature in the Torricellian vacuum, so that as good an exhaustion 

 ought to result as can be got by boiling out a space with mercury. 

 There is another way in which the question may be put. Assuming 

 the molecular latent heats are approximately proportioned to the abso- 

 lute boiling points, then we can, from a comparison with the oxygen 

 value, deduce that of hydrogen, and thereby get the constants in a 

 two-term formula for the vapour pressures. For pressures below an 

 atmosphere, the following approximate formulae were deduced : 



009. ( 

 Oxygen ...... logp = 7'2058 - mm ....... (4). 



. 

 Hydrogen ... log^ = 7-2428 - i^jL' mm ......... (5). 



From these expressions it follows that at its boiling point, or 35 

 absolute, hydrogen has 7/852000 times the pressure of oxygen, or the 

 latter pressure is about the eight millionth of an atmosphere. A 

 similar formula, calculated from the critical and boiling point data, 

 gives substantially the same order of quantities. Formula (4) for 

 oxygen tensions must be fairly accurate, seeing it gives a theoretical 

 latent heat of about 56 units per gram of liquid evaporating at the 

 boiling point, whereas direct determinations result in 55 units. To test 

 this inference, the following plan of experimenting was adopted : 

 Ordinary shaped vacuum tubes, like A, B, used for the spectroscopic 

 examination of gases, with and without electrodes (figs. 1 and 2), 

 having a capacity ranging from 15 to 25 c.c., had pieces of quill tubing 

 about a foot long sealed on. The tubes were contracted at D to about 

 1 mm., so that they could be sealed off with rapidity. The end C 

 sometimes terminated in a small bulb (fig. 3), in order to give increased 

 cooling surface, and, when necessary, to allow many times the volume 



