486 



SCIENCE. 



[N. S. VOL. VIII. No. 198. 



gauge adapted to measure small pressures 

 with very great precision, and a barometric 

 gauge for measuring larger pressures. 



In using this apparatus the gas to be 

 tested was introduced at atmospheric pres- 

 sure ; the ice tank was raised, and the fall- 

 ing temperature of the thermometer, which 

 could lose heat only by radiation, conduc- 

 tion and convection through the surround- 

 ing gas, was observed through a telescope. 

 The time required for the temperature to 

 fall through a given range, usually from 

 15° to 10° Centigrade, was carefully noted. 

 Then the ice tank was lowered, permitting 

 the thermometer to regain the temperature 

 of the laboratory ; some of the gas was 

 pumped out, and the cooling of the ther- 

 mometer again observed at this reduced 

 pressure. This process was repeated many 

 times, until the pressure was reduced to the 

 lowest point attainable. 



The results obtained with each gas were 

 plotted in a curve showing its heat con- 

 ductivity at all pressures from atmospheric 

 down ; the ordinatos representing the recip- 

 rocals of the time of cooling in seconds, 

 while the abscissas represented the pressure. 



The present chart shows curves repre- 

 senting the heat conductivity of several 

 gases, from fifty millionths of atmospheric 

 pressure downward. The data for all of 

 these, except the helium curve, are taken 

 from last year's paper ; but the scale is dif- 

 ferent. 



I am indebted to Professor Ramsay for 

 the helium used in obtaining the curve 

 here shown. 



The ordinates of each curve measured 

 from A B as a base line represent the total 

 rate of heat transmission by the ether and 

 the gas at the pressures indicated by the 

 abscissas, while ordinates measured from 

 the line C D represent the heat transmitted 

 by the gas alone. 



It will be observed that the curves of all 

 the gases named vanish together at the 



point of zero pressure D. Repeated ex- 

 periments have shown this condition to be 

 always strictly true within the narrow 

 limits of errors of observation, provided 

 that, before the introduction of the gas to 

 be tested, the whole apparatus has been 

 kept highly exhausted for several days ; or, 

 better still, has been heated several hours 

 by means of air and water baths, while 

 kept exhausted. Without one of these 

 precautions I was never able to get any 

 gas curve quite down to the point D, for 

 reasons which will appear. The point D 

 represents a period of three hundred sec- 

 onds required for the temperature of the 

 thermometer to fall from 15° to 10°, with 

 the pressure of the surrounding gas reduced 

 to one twenty-millionth or less. 



A very brief account of the circumstances 

 which led to the discovery of the new gas 

 may not be out of place. I had long been 

 engaged in high vacua experiments, and 

 had observed that glass apparatus, when 

 highly exhausted and heated, evolved gas 

 for an indefinite length of time, rapidly at 

 first, then slower, but never stopping until 

 the temperature was reduced. On cooling, 

 rapid reabsorption always took place, but 

 was never complete ; indicating that two or 

 more gases had been evolved by heating, 

 one of which was not absorbed by cooling. 

 In other words, the absorption was selec- 

 tive. The truth of this conclusion was 

 abundantly demonstrated subsequently. 

 However, the percentage of reabsorbed gas 

 was so large that I used a small quantity 

 of pulverized glass in several experiments 

 to absorb a part of the residual gas after 

 the highest attainable exhaustion was 

 reached. The pulverized glass was always 

 lead glass like that of the apparatus, and 

 was heated nearly red-hot for several hours 

 before and during the final exhaustion. 



During these experiments a curious phe- 

 nomenon was noted — the pulverized glass 

 lost its snowy whiteness. This I thought 



