1451 
wanting in the accuracy of his calculation. Among other things he 
assigns equal velocity to all the molecules. I will, therefore, derive 
by a somewhat different course an expression for the mean deviation, 
which only differs from the above mentioned one in numerical 
coefficient, but in which Maxwell’s distribution of velocity is taken 
into account. 
Let us suppose that a spherical “Brownian particle’ with mass 
M = executes elastic collisions with a number of spherical molecules 
with mass 7, the MAXWeLLIAN distribution of velocity prevailing. 
We will call the particle shortly J/. 
If we represent by PA and PB (fig. 1) the velocities of the 
particle and the colliding molecule before the collision, and bring 
a plane W through Q, the end point of the velocity of the 
centre of gravity, parallel to the plane of collision, we shall find 
the velocities PA’ and PB’ after the collision by reflecting the 
velocities relative to the centre of gravity with respect to W and 
by adding them to the velocity of the centre of gravity. When W 
changes its position, A’ moves on a sphere round Q with radius 
QA. Maxwerr *) showed that all the points of this sphere have equal 
probability for A’, so that the mean value of QA’ for all the 
collisions of Min every direction is 0. The velocity of the centre 
of gravity, however, has on an average a projection in the direction 
of PA. From this Jeans *) derives that a molecule on collision with 
other similar ones on an average retains a part @s of its original 
velocity s. For collisions between particles of different mass @ 
assumes a value which I have calculated, and which has already 
been communicated by Prof. KUENEN *). 
1) J. GC. MAXwELt. Phil. Mag. 4 19. p. 19 (1860). 
2) J. H. Jeans. Phil. Mag. (6) 8 p. 700. (1904). 
8) J. P. Kuenen. These Proc. XVII, p. 1068, 
