106 PHENOMENA, ATOMS, AND MOLECULES 



loss by convection, I was able to prove that the abnormal rate of heat loss 

 previously observed with tungsten filaments at high temperatures in 

 hydrogen was due to actual dissociation ; in fact I was thus able to calculate 

 the heat of dissociation and the degree of dissociation at different tem- 

 peratures. 



However, to make sure of these conclusions I wished to make measure- 

 ments of the heat losses in gases which could not possibly dissociate, and 

 therefore undertook experiments with heated tungsten wires in mercury 

 vapor at atmospheric pressure. A little later I experimented with nitrogen 

 to see if this gas dissociated at high temperatures, but found that it did 

 not do so. In both of these gases the filaments could be maintained at tem- 

 peratures close to the melting point for a far longer time than if heated in 

 vacuum at the same temperature. Thus the rate of evaporation was greatly 

 decreased by the gas, many of the evaporating tungsten atoms being 

 brought back to the filament after striking the gas molecules. 



By this time I was familiar with all the harmful effects which gas can 

 produce in contact with filaments and knew under what conditions these 

 bad effects could be avoided. In particular, I realized the importance of 

 avoiding even almost infinitesimal traces of water vapor. Thus, when I 

 found a marked effect of mercury vapor and nitrogen in reducing the rate 

 of evaporation, it occurred to me that it might be possible to operate a 

 tungsten filament in gas at atmospheric pressure and obtain a long useful 

 life. Of course, it would be necessary to raise the temperature far above 

 that at which the filament could be operated in vacuum in order to com- 

 pensate for the serious loss in efficiency due to convection by the improved 

 efficiency resulting from the rise in filament temperature. Whether or not 

 the increased rate of evaporation, due to this increase in temperature, would 

 more than ofifset the decrease in the rate due to the gas was a matter that 

 could only be tested by experiment. 



In connection with my studies of the heat losses from filaments of 

 various diameters at incandescent temperatures, I had found that the heat 

 loss increased only very slowly with the diameter, so that the loss 

 per unit area from a small filament was enormously greater than from 

 a large filament. Calculations showed that it was hopeless to get practical 

 lamps with filaments in nitrogen, if these filaments were of very small 

 diameter. For example, a filament i mil in diameter, which corresponds to 

 an ordinary 25-watt lamp, if run in nitrogen at atmospheric pressure would 

 consume 4.8 watts per candle at a temperature of 2400° K., which would 

 give I watt per candle with a filament in vacuum. This great loss in 

 efficiency is due to the cooling effect of gas. To bring back the efficiency 

 of the gas-filled lamp to that of the vacuum lamp, it would be necessary 

 to raise the temperature from 2400° to 3000° K., which would have caused 



