THERMIONIC ELECTRON EMISSION 435 



elementary curves, each one of which Is shifted along the V axis by the 

 amount of the potential drop along the filament. Such a sum curve 

 consists of a straight line having the correct slope at sufficiently great 

 retarding potentials ; the sharpness of the break point in the curve is, 

 however, destroyed and the slope of the curve for small retarding 

 potentials is decreased, thus simulating the ideal curve for a higher 

 temperature. The best way to obviate this difficulty is to work with 

 equipotential cathodes which are heated indirectly. This makes the 

 construction of the tube more difficult and has been used only by 

 Demski.22 Most of the work has been done on filaments which were 

 heated intermittently by means of a mechanical or electrical com- 

 mutator. 



In this way Germer,-" Demski and others have shown that the dis- 

 tribution of thermionically emitted electrons is Maxwellian and cor- 

 responds to a temperature which is equal to the temperature of the 

 cathode to within less than 5 per cent. Germer worked with tungsten 

 for a series of temperatures between 1440 and 2475° K. Demski 

 worked with tungsten and with oxide-coated filaments. He used a 

 mechanical and an electrical commutator and also worked with equi- 

 potential cathodes. Nottingham ^^ and others have reported that for 

 thoriated tungsten and oxide-coated filaments the temperature com- 

 puted from the shape of the log i versus V curve for small retarding 

 potentials was about 1.5 times the temperature of the cathode. Not- 

 tingham explains this as due to a sharp peak in the potential distance 

 curve through which a part of the wave electrons can penetrate. In 

 my opinion it is much more likely that these observations are due to 

 non-uniformities in the work function of the cathode and the anode. 



If the work function of the cathode is non-uniform, the observed 

 curve should result from the summing up of the currents for a series of 

 curves somewhat similar to curves 1, 4 and 5 in Fig. 6. The sum 

 curve will have the correct slope at sufficiently great retarding poten- 

 tials ; but at low values of Vr the slope should be too small correspond- 

 ing to too high a temperature. The break point will be less sharp. 



If the work function of the anode is non-uniform, the elements of 

 the sum curve will consist of a series of ideal curves shifted parallel to 

 the V axis. The sum curve will again yield correct temperatures at 

 large values of Vr but too high temperatures at small values of Vr. 

 That the work function of cathodes is usually non-uniform will be 

 shown in the next section. It is to be expected that the anode work 

 function will also be non-uniform since the anode is more difficult to 

 heat treat than the cathode. However, when one takes into account 

 the effect of these non-uniformities, it is seen that the experiments 



