ATOMIC HYDROGEN 107 



a 2000-fold increase in the rate of evaporation, and such an increase could 

 certainly not be compensated for by the effect of the gas in retarding the 

 evaporation. 



With filaments of much larger diameter, however, the effect of the 

 gas in decreasing the efficiency was not nearly so marked. We therefore 

 constructed lamps having filaments of large diameter in the form of a 

 single loop and filled these lamps with nitrogen at atmospheric pressure. 

 We ran these lamps with a filament temperature so high that, in spite of 

 the gas, the efficiency corresponded to about 0.8 watt per candle instead 

 of the usual i watt per candle at which we tested our vacuum lamps. We 

 were disappointed to find that these lamps blackened much more rapidly 

 than vacuum lamps of similar efficiency so that the total useful life of the 

 lamp was short. 



This result, which is what most lamp engineers would have expected, 

 seemed to indicate that the rise in temperature necessary to offset the 

 heat losses by the gas increased the evaporation by more than the amount 

 of the reduction in evaporation due to the gas. If I had not previously 

 become so familiar with the behavior of various gases, this discouraging 

 result might easily have stopped further experimenting in this direction. 

 However, I noticed that the bulb had blackened during the short life of the 

 lamp, whereas from my knowledge of the interaction of tungsten and 

 nitrogen I had expected a deposit of a clear brown color. I felt that the 

 black deposit, therefore, could mean only one thing, water vapor, not- 

 withstanding the fact that to avoid this water vapor we had taken pre- 

 cautions which were greater, I believe, than had ever been used before for 

 the preparation of moisture- free gases and glass surfaces. We were thus 

 led to take still greater precautions and use still larger bulbs, so that the 

 glass surfaces could not become overheated by the convection currents in 

 the gas that rose from the filament. We were then soon able to make lamps 

 having a life of over 1000 hours with an efficiency about 30 to 40 per cent 

 better than could have been obtained with filaments in vacuum. 



As I look back upon these experiments I think that we were very 

 fortunate at that time in not having had at our disposal a supply of argon 

 gas. From theoretical reasons I had concluded that argon should be better 

 than nitrogen, and if I had had argon I should probably have tried it first. 

 If these lamps had blackened because of traces of water vapor, I would 

 naturally have attributed this to the increase in evaporation caused by the 

 high temperature, and would have had no reason for suspecting that water 

 vapor was the cause of the trouble, for, of course, in argon a brown deposit 

 would not be expected in any case. 



The lamps that we were able to make in this way, with an improved 

 efficiency, were limited to those which took a current of 5 amperes or more, 



