4 1 Boltzmann Equation
df (r, p, t) ∂f dr dp
= + · ∇r f + · ∇p f
dt ∂t dt dt
∂f p ∂f
= + · ∇r f − [∇r V (r)] · ∇p f = , (1.1)
∂t m ∂t coll
where V (r) is a single-particle potential. The left-hand side of (1.1) describes
the dynamics of a single particle. The influence of the other particles will give
rise to a further change of the distribution function ∂f /∂t|coll which describes
the effect of the collisions in the gas. We will not proceed historically, but
include directly the proper quantum statistics for quantum gases, so that
we are not limited to nondegenerate gases. This extension is necessary for
the application of the Boltzmann kinetics to electron gases in semiconductors
which are often degenerate, whether they are produced by doping, injection, or
optical excitation. Fermi’s golden rule gives us the transition probability per
unit time and thus the wanted change of f due to collisions. For an interacting
Fermi gas we calculate this change by considering approximately free-particle
collisions in which the particle is scattered from a momentum state p to a
momentum state p and simultaneously another particle is scattered from
state p1 to p1 , as well as the inverse process
∂f (p)
= − w(p, p1 ; p , p1 ) f (p)f (p1 )[1 − f (p )][1 − f (p1 )]
∂t coll
p ,p1 ,p1
− [1 − f (p)][1 − f (p1 )]f (p )f (p1 ) , (1.2)
where the intrinsic transition probalility per unit time is given by
w(p, p1 ; p , p1 ) = 2 |Wp,p1 ;p ,p1
1
− Wp,p1 ;p1 ,p |2
2π
× δp+p1 ,p +p1 δ(εp + εp1 − εp − εp1 ) . (1.3)
h̄
Here
Wp,p1 ;p ,p1 = pp1 |W |p p1 (1.4)
is the interaction matrix element and εp is the energy of the particle. The
second matrix element in (1.3) is the exchange term in which p1 and p are
interchanged. This form of the intrinsic transition probability is called the first
Born approximation. The population factors take care that the initial states in
the scattering event are populated and that the final states are empty in accor-
dance with the Pauli principle. The scattering p + p1 → p + p1 is a loss term
which reduces f (p), while the inverse process p + p1 → p + p1 increases the
distribution function. For shortness of notation, the parametric dependencies
on the spatial coordinate r and time t are not shown in the collision integral.
The form of the collision integral leads to five conservation laws for: (a) the