direction is generated. When this stream is superposed on the first, 
n n 
the numbers of molecules become a, — — (a, + a) anda, — * (a, + a,) 
n n 
and the coefficient of diffusion D 
1 
D= = (n,u,l, + n,u,l,). 
According. to this formula. D would vary strongly with the com- 
position of the mixture, when m, and m, differ much. In order to 
show this we put successively n, =O and n, =O and find for the 
limiting values of D: 
Dee yo 
A FalS ee m,+m, 
ENE 
La = 
; a 
3 no m,+m, 
D(n, =”) == 
4 
Using the relation u,’m, = um, = xh’ where / is the constant 
Jt 
in Maxwett’s law of distribution, we can also write 
Ne as 8 = 
320no? V ah m,m,+m, 
i ane es Babe 
Zana? V ah m,m,-+m, 
The two values of D are to each other as m,:m, e. g. for car- 
bon dioxide and hydrogen as 2: 44. 
The experimental evidence ') is in favour of a coefficient which 
varies with n, and n,, but only to a very small extent, so that a 
variation as given by Mrysr’s formula is out of the question. 
The coefficient of diffusion according to STEFAN *) is: 
D= 3 1 eae 
16no? V xh m,m, 
therefore independent of the composition of the mixture, which agrees 
approximately with experiment. The same expression follows from 
Maxwett’s second theory when applied to elastic molecules; this 
was proved by LANGEVIN ®). The only simplifying supposition which 
1) Compare A. Lonius, Ann. d. Ph. (4) 29 p. 664, 1909 
2) J. Sreran, Wien. Sitz.ber. 65 p. 323. 1872 | 
3) P. Langevin, Ann. chim. phys. (8) 5 p. 245. 1905. Maxwett himself had used 
the same method (Nature 8. p. 293. 1873): his result given without proof differs 
by the factor */; from that of Langevin, 
15% 
