452 Professor 0. W. Richardson [May 7, 



of the latter hypothesis has, generally speaking, been greatly over- 

 estimated, it would take too long to discuss this question with the 

 completeness which it demands. I shall therefore content myself 

 with directing your attention to some experiments with tungsten 

 filaments which prove that only an insignificant fraction, if any, of 

 the emission from this substance can be attributed to chemical 

 action. 



Tungsten is peculiarly suited to these experiments on account of 

 its great refractoriness. It can be heated in a vacuum for con- 

 siderable periods at temperatures so high that all known impurities 

 are volatilized out of it. The preliminary treatment of the experi- 

 mental lamps furnishes some novel features which may prove of 

 interest. The ductile tungsten filaments are electrically welded to 

 the supporting leading wires in an atmosphere of hydrogen. After 

 mounting, the lamps are exhausted in a vacuum furnace (with an 

 external air pressure of about 1 cm.) at 550-600° C. for about 

 24 hours until the evolution of gas becomes very small. A Gaede 

 pump is used for the internal exhaust at first, and, later on, liquid 

 air and charcoal in addition. In the final stages the tungsten is 

 glowed at about 3000° absolute, and, for the best results, the anode 

 is heated by subjecting it to an intense electron bombardment from 

 the hot wire. The conditions as to freedom from gaseous contamina- 

 tion which have been attained in this way are far superior to those 

 which result from any other method of treatment. 



With lamps thus prepared I have carried out simultaneous 

 measurements of the rate of emission of electrons on the one hand, 

 and either of the variation of the pressure of the gas present or of 

 the rate of loss of matter by the filament on the other. Particular 

 experiments have led to the following numbers : — 



(1) For each molecule of gas given off the number of electrons 

 emitted by the filament may be as high as 260,000,000. 



(2) At each impact of a gas molecule with the filament 15,000 

 electrons would have to be emitted, and 



(3) Each atom of tungsten which disappears from the filament 

 would have to cause the emission of 984,000 electrons. 



The magnitude of these numbers entirely precludes the possibility 

 that chemical action plays any significant part in this emission. 

 Again, the mass of the electrons lost by a filament may exceed the 

 mass of tungsten lost in the same interval, proving that the emitted 

 electrons are not furnished at the expense of the tungsten. They 

 must, therefore, flow in from outside points of the circuit. Thus 

 these experiments furnish a direct proof that the electric current in 

 metals is carried by moving electrons. The mechanism of metallic 

 conduction becomes more mysterious every day, but this, at any rate, 

 is a fact which has to be reckoned with. 



Perhaps I can drive these matters home to you more effectually 

 by means of a simple experiment, which shows that these electron 



