Mr. Gr. J. Stoney on Crookes's Radiometer. 309 



tion. Let v be the average velocity of molecules of air at the 

 temperature t. Then this air presses against the walls of the 

 apartment with a force of p grammes per square centimetre, 

 because its molecules strike the walls with velocities of which 

 the average is v, and rebound from them with velocities of 

 which the average is also v. Let n be the number of mole- 

 cules which on the average strike a square centimetre of the 

 walls per second. Then, if we picture to ourselves a plane 

 drawn anywhere through this medium, n molecules per second 

 fly through every square centimetre of this plane from left to 

 right, and an equal number fly through it from right to left. 

 Having acquired these conceptions, let us next suppose a disk 

 like the disk of a radiometer, to be situated in the middle of 

 this apartment, the front of the disk being kept at temperature 

 t -h A£, and the back at temperature t. Let us further suppose 

 that there is no gravity, in which case there will be no con- 

 vection-currents. But a complete gradient would promptly 

 form in front of the disk ; in other words, the air in front of 

 the disk would be warmed to a considerable distance, and 

 somewhat dilated. Throughout this gradient the temperature 

 would vary from t+ At on the one side, to t on the other. 

 The density would also vary, and in such a way that the ten- 

 sion would be evexywhere the same as in the rest of the apart- 

 ment, viz. p grammes per square centimetre. In this case 

 the disk would be pressed equally in front and behind. Let 

 us next form the conception of a spherical surface, with its 

 centre at the middle of the disk, large enough to contain the 

 disk and extend beyond it, but not so large as to reach through 

 the gradient. The hemisphere in front of the disk would thus 

 contain air belonging to the gradient, and the hemisphere be- 

 hind the disk would contain air of the same kind as in the rest 

 of the apartment. Through each square centimetre of the 

 surface of the hemisphere behind the disk there will be n mole- 

 cules per second flying into the sphere, and n molecules per 

 second flying out ; and these molecules will be endowed with 

 velocities of which the average is v. But through any square 

 centimetre of the hemisphere in front of the disk there will be 

 a number of molecules flying in and flying out per second 

 which will be less than n, while their average velocity will be 

 greater than v. Let us next suppose that this spherical sur- 

 face, while it still allows the molecules to pass through it and 

 retain the direction of their flight, is given the property of 

 being able to cut down the velocities of molecules whose aver- 

 age is above v, so that the average velocity of the molecules 

 after passing through the surface becomes v, whatever it may 

 have been before. This new property will not make any dif- 



