696 Professor Dewar [June 2, 



Kelvin and Prof. Tail's thermo-electric diagram. If the lines of 

 copper and platinum were prolonged in the direction of negative 

 temperature, they would intersect at — 95° (J. Similarly, the copper 

 and palladium lines would cut one another at — 170° C. Now, if 

 this diagram were correct, the E.M.F. of the thermo-electric junc- 

 tions of these two pairs of metals should reverse at these points. 

 A Cu — Pt junction connected to a reflecting galvanometer was then 

 placed in oxygen vapour and cooled down. At —100° C. the spot 

 of light stopped and reversed. A Cu — Pd junction was afterwards 

 placed in a tube containing liquid oxygen, and a similar reversal 

 took place at about —170° C. 



Liquid oxygen is a non-conductor of electricity : a spark, taken 

 from an induction coil, one millimetre long in the liquid requires a 

 potential equal to a striking distance in air of 25 millimetres. It 

 gave a flash now and then, when a bubble of the oxygen vapour in 

 the boiling liquid came between the terminals. Thus liquid oxygen 

 is a high insulator. When the spark is taken from a Wimshurst 

 machine the oxygen appears to allow the passage of a discharge to 

 take place with much greater ease. The spectrum of the spark taken 

 in the liquid is a continuous one, showing all the absorption bands. 



As to its absorption spectrum, the lines A and B of the solar 

 spectrum are due to oxygen, and they came out strongly when the 

 liquid was interposed in the path of the rays from the electric lamp. 

 Both the liquid and the highly compressed gas show a series of five 

 absorption bands, situated respectively in the orange, yellow, green 

 and blue of the spectrum. 



Experiments prove that gaseous and liquid oxygen have sub- 

 stantially the same absorption spectra. This is a very noteworthy 

 conclusion considering that no compound of oxygen, so far as is 

 known, gives the absorptions of oxygen. The persistency of the absorp- 

 tion through the stages of gaseous condensation towards complete 

 liquidity implies a persistency of molecular constitution which we 

 should hardly have expected. The absorptions of the class to which 

 A and B belong must be those most easily assumed by the diatomic 

 molecules (O2) of ordinary oxygen; whereas the difi'use bands above 

 referred to, seeing they have intensities proportional to the square of 

 the density of the gas, must depend on a change produced by com- 

 pression. This may be brought about in two ways, either by the 

 formation of more complex molecules, or by the constraint to which 

 the molecules are subjected during their encounters with one 

 another. 



When the evaporation of liquid oxygen is accelerated by tlie 

 action of a high expansion pump and an open test-tube is inserted 

 into it, the tube begins to fill up with liquid atmospheric air, pro- 

 <luced at the ordinary barometric pressure. 



Dr. Janssen had recently been making prolonged and careful ex- 

 periments on Mont Blanc, and he found that these oxygen lines disap- 

 peared more and more from the solar spectrum as he reached higher 



