1880.] On Heat Conduction in Highly Rarefied Air. 243 



I have embodied these results in the preceding diagram. The ordi- 

 nates represent the number of seconds occupied during the rise of 

 each 5°, starting from 25° ; the abscissse represent the pressure. The 

 lower portion gives the total variation in time between pressures of 

 760 millims. and 1 millim. The upper and larger portion of the dia- 

 gram gives the absciss® in millionths of an atmosphere. At the right 

 ,side of the diagram, in the space ABC, I have drawn a series of 

 horizontal lines increasing in length from 0'25 millim. at 400 M., to 

 100 millims. at 1 M. These show the actual lengths of the mean free 

 path of the molecules of air at the degrees of exhaustion to which they 

 are opposite.* The parallelism between the curves formed by joining 

 the ends of these horizontal lines and the curves representing the rate 

 of cooling is sufficiently close to justify the inference that they are 

 associated phenomena. 



There are two ways in which heat can get from the glass globe to 

 the thermometer— (1) By radiation across the intervening space ; 

 (2) by communicating an increase of motion to the molecules of the 

 gas, which carry it to the thermometer. It is quite conceivable that 

 a considerable part, especially in the case of heat of low refrangibility, 

 maybe transferred by " carriage," as I will call it to distinguish it from 

 .convection which is different, and yet that we should not perceive 

 much diminution of transference, and consequently much diminution 

 of rate of rise with increased exhaustion, so long as we work with 

 ordinary exhaustions up to 1 millim. or so. For if, on the one hand, 

 there are fewer molecules impinging on the warm body (which is 

 adverse to the carriage of heat), yet on the other the mean length of 

 path between collisions is increased, so that the augmented motion is 

 -carried further. The number of steps by which the temperature 

 passes from the warmer to the cooler body is diminished, and accord- 

 ingly the value of each step is increased. Hence the increase in the 

 difference of velocity before and after impact may make up for the 

 .diminution in the number of molecules impinging. It is therefore con- 

 ceivable that it may not be till sach high exhaustions are reached that 

 the mean length of path between collisions becomes comparable with 

 the diameter of the case, that further exhaustion produces a notable fall 

 in the rate at which heat is conveyed from the case to the thermometer. 



The above experiments show there is a notable fall, a reduction of 

 pressure from 5 M. to 2 M., producing twice as much fall in the rate 

 ,as is obtained by the whole exhaustion from 760 millims. to 1 millim. 

 We may legitimately infer that each additional diminution of a 

 millionth would produce a still greater retardation of cooling, so that 

 in such vacua as exist in planetary space the loss of heat — which in 

 that case would only take place by radiation — would be exceedingly slow. 



* In the published diagram the lengths have been reduced by the engraver in the 

 proportion of 8 to 3. 



