Professional Documents
Culture Documents
David Foster
October 28, 2012
For a system governed by a linear differential equation, not only can f can be constructed as a sum
of point sources, but the resulting field φ is a sum of all the fields (all the G(r, r 0 )s) produced by those
individual point sources (at points r 0 ). Since f gives the weight of each point source, f is also the weight of
each resulting point-source-generated field. That is,
Z
φ(r) = f (r 0 )G(r, r 0 )dr 0 (3)
1
G to use is determined by boundary conditions. If the boundary is at infinity and φ → 0 on the boundary,
then the only sources in the universe are included in f . Then G is taken to be the point-source response
that decays at infinity, for example 1/|r − r 0 | for an electrostatic problem.
Again, in classical physics, a Green’s function is the field/potential response to a point source.
The (unperturbed) Green’s function for the time independent Schrödinger equation is defined as
Here we have used the notation G0 because the Green’s function is related to the unperturbed problem in
Eq. 5. In fact this equation means that G0 is an inverse of E − H (0) :
G−1
0 (r, E) = E − H0 (r) (8)
Equation 7 is an integral equation, giving the exact solution ψ in terms of itself in the integral on the
RHS. In the first Born approximation (I believe this is the B.A.), the perturbing potential V acts on the
unperturbed (incident) wavefunction, which we know. This amounts to replacing the ψ on the RHS of Eq. 7
1 If we had started with the Schrödinger equation having ψ instead of ∆ψ on the left hand side of Eq. [5], we would have the
2
(0) (0)
with ψk0 . In fact we can calculate the solution to Eq. 7 by interation. The first step is to put ψk0 on the
right hand side (the Born approximation) and then integrate to obtain ψ. This ψ is first order in V . We
then plug this ψ in to the RHS again and solve for a new ψ, which is second order, and so on.
We can also define a different, “full” Green’s function G, by
[E − H (0) (r) − V (r)]G(r, r 0 ) = δ(r − r 0 ) (9)
Here we are even further from the original concept of Green’s function, because the left hand operater acting
on the true solution ψ will give zero. So what is left that needs to be expanded in terms of the Green’s
function? I don’t know, but the advantage of this type of Green’s function is that, if we knew G, we could
obtain ψ easily through
Z
(0) (0)
ψ(r) = ψk0 (r) + dr 0 G(r, r 0 )V (r 0 )ψk0 (r 0 ) (10)
3
then
Z
ψ(r, t2 ) = dr 0 G(rt2 , r 0 t1 )ψ(r 0 , t1 ) (17)
for t1 < t2 . To verify, one can see that operating on each side of this equation with the parabolic differential
operator [i~∂t2 − H] yields the true equation 0 = 0. We can also go the other way: Since only one initial-type
boundary equation is necessary to uniquely specify a solution of the parabolic differential equation (14), then
if we have a G that satisfies both Eq. (14) and the physical requirement of Eq. (16), then it is a unique
solution. The G implicitly defined by Eq. (17) satisfies both (when one includes the physical requirment
that limt2 →t1+ ψ(r, t2 ) = ψ(r, t1 )). Thus Eq. (14) and Eq. (16) imply Eq. (17), which says that G is the
propagator of the wavefunction.
The Green’s function we have discused above is called the retarded Green’s function, because it propagates
forward in time and thus the result is delayed from the input. The time propagation operator for a time
independent Hamiltonian is
−i
U (t, t0 ) = exp[ (t − t0 )H] (18)
~
The retarded Green’s function is then
−i
GR (rt, r 0 t0 ) = θ(t − t0 )hr| exp[ (t − t0 )H]|r 0 i (19)
~
My extremely introductory introduction ends here, and you may continue it with the chapters of most
texts on Green’s function theory in quantum mechanics and many-body physics. Refs. [1, 2] are excellent
for condensed matter many-body theory.
References
[1] Henrik Bruus and Karsten Flensberg. Many-Body Quantum Theoy in Condensed Matter Physics. Oxford,
2004.
[2] Gabriele F. Giuliani and Giovanni Vignale. Quantum Theory of the Electron Liquid. Cambridge, 2005.