III.

G Conservation Laws
Approach to equilibrium: We now address the third question posed in the introduction,
of how the gas reaches its nal equilibrium. Consider a situation in which the gas is
perturbed from the equilibrium form described by eq.(III.56), and follow its relaxation to
equilibrium. There is a hierarchy of mechanisms that operate at dierent time scales.
(i) The fastest processes are the two body collisions of particles in immediate vicinity.
Over a time scale of the order of c , f2 (q1 , q2 , t) relaxes to f1 (q1 , t)f1 (q2 , t) for separa
tions |q1 q2 | d. Similar relaxations occur for the higher order densities fs .
(ii) At the next stage, f1 relaxes to a local equilibrium from, as in eq.(III.53), over the
time scale of the mean free time . This is the intrinsic scale set by the collision term
on the right hand side of the Boltzmann equation. After this time interval, quantities
conserved in collisions achieve a state of local equilibrium. We can then dene at each
point a (time dependent) local density by integrating over all momenta as
n(q, t) =

d3 pf
1 (
p, q, t),

(III.69)

as well as a local expectation value for any operator O(

d3 pf
1 (
p, q, t)O(
p, q, t).

(III.70)

(iii) After the densities and expectation values have relaxed to their local equilibrium forms
in the intrinsic time scales c and , there is a subsequent slower relaxation to global
equilibrium over extrinsic time and length scales. This nal stage is governed by the
smaller streaming terms on the left hand side of the Boltzmann equation. It is most
conveniently expressed in terms of the time evolution of conserved quantities according
to hydrodynamic equations.
Conserved quantities are left unchanged by the two body collisions, i.e. satisfy
(p1 , q, t) + (p2 , q, t) = (p1 , q, t) + (p2 , q, t),

(III.71)

where (p1 , p2 ) and (p1 , p2 ) refer to the momenta before and after a collision, respectively.
For such quantities, we have
J (q, t) =


df1 
d p (
p, q, t)
(
p, q, t) = 0.
dt coll.
3

64

(III.72)

 Proof: Using the form of the collision integral, we have

Z
J = d3 p1 d3 p2 d2b |v1 v2 | [f1 (p1 )f1 (p2 ) f1 (p1 )f1 (p2 )] (p1 ).

(III.73)

(The implicit arguments (q, t) are left out for ease of notation.) We now perform the same
set of changes of variables that were used in the proof of the H-theorem. The rst step is
averaging after exchange of the dummy variables p1 and p2 , leading to
Z
1
J =
d3 p1 d3 p2 d2b |v1 v2 | [f1 (p1 )f1 (p2 ) f1 (p1 )f1 (p2 )] [(p1 ) + (p2 )] . (III.74)
2
Next, change variables from the originators (p1 , p2 , b ), to the products (p1 , p2 , b ) of the
collision. After relabeling the integration variables, the above equation is transformed to
Z
1
J =
d3 p1 d3 p2 d2b |v1 v2 | [f1 (p1 )f1 (p2 ) f1 (p1 )f1 (p2 )] [(p1 ) + (p2 )] .
2
(III.75)
Averaging the last two equations leads to
Z
1
d3 p1 d3 p2 d2b |v1 v2 | [f1 (p1 )f1 (p2 ) f1 (p1 )f1 (p2 )]
J =
4
[(p1 ) + (p2 ) (p1 ) (p2 )] ,

(III.76)

which is zero from eq.(III.71).

3
p, q, t) t +
+ F
p, q, t) = 0,
f1 (
J ( q, t) = d p (
m
p

(III.77)

where we have introduced the notations t /t, /q , and F = U/q . We

3
d p t +
+ F
+ F
(f1 ) f1 t +
= 0.
m
p
m
p

(III.78)

The third term is zero, as it is a complete derivative. Using the denition of expectation
values in eq.(III.70), the remaining terms can be rearranged into


E
 D p E
Dp

t (n ) + n
n t n
nF
= 0.
m
m
p
65

(III.79)

As discussed earlier, for elastic collisions, there are 5 conserved quantities: particle
number, the three components of momentum, and kinetic energy. Each leads to a corre
sponding hydrodynamic equation, as constructed below:
(a) Particle number: Setting = 1 in eq.(III.79) leads to
t n + (nu ) = 0,

(III.80)

where we have introduced the local velocity

p
m

(III.81)

This equation simply states that the time variation of the local particle density is due to
a particle current Jn = nu.
(b) Momentum: Any linear function of the momentum p is conserved in the collision, and
we shall explore the consequences of the conservation of
c

u.
m

(III.82)

Substituting c into eq.(III.79) leads to

F
= 0.
m

(III.83)

Taking advantage of c = 0, from eqs.(III.81) and (III.82), leads to

F
1

P ,
m
mn

(III.84)

where we have introduced the pressure tensor,

(III.85)

The left hand side of the equation is the acceleration of an element of the uid du/dt,
which should equal Fnet /m according to Newtons equation. The net force includes an
additional component due to the variations in the pressure tensor across the uid.
(c) Kinetic energy: We rst introduce an average local kinetic energy

mc2
2

p2
mu2
p u +
2m
2
66

(III.86)

and then examine the conservation law obtained by setting equal to mc2 /2 in eq.(III.79).
Since for space and time derivatives = mc c = mc u , we obtain
 

mc2
+ nmt u c + nm u (u + c )c nF c = 0.
t (n) + n (u + c )
2
(III.87)
Taking advantage of c = 0, the above equation is simplied to

 
mc2
t (n) + (nu ) + n c
+ P u = 0.
2

(III.88)

We next take out the dependence on n in the rst two terms of the above equation, nding
t n + nt + (nu ) + nu + h + P u = 0,

(III.89)

where we have also introduced the local heat ux

and the rate of strain tensor


h nm c c2 ,
2

u =

1
( u + u ) .
2

(III.90)

(III.91)

Eliminating the rst and third terms in eq.(III.89) with the aid of eq.(III.80) leads to
1
1
t + u = h P u .
n
n

(III.92)

Clearly to solve the hydrodynamic equations for n, u, and , we need expressions for P
and h, which are either given phenomenologically, or calculated from the density f1 , as in
the next sections.

67

MIT OpenCourseWare
http://ocw.mit.edu

8.333 Statistical Mechanics I: Statistical Mechanics of Particles

For information about citing these materials or our Terms of Use, visit: http://ocw.mit.edu/terms.