
624 Chapter 20. Phenomenological Theory
=
N
e
-
l0J
"
\l) +
J2
f dt'liy-^^'-^^^-e-^'-^"'] (20.55)
Take U(t') to have time dependence exp(—iuit'); to must have a small positive imaginary
part
77
for the time integral converge, or else take üj
t
i to have an imaginary part —y
t
i.
f\l\<\^\l'\
V I^IO
e lUJ
' \ -iuj.t Excluding / from the sum takes
= » t/J' /w ; 7 (
e
■ care of the normalization. (2U.56)
L
,Y/ Δ(w,-o;//+a;)J
If, on the other hand, the time dependent potential were to have the form
U*
exp[ico*t],
then one would have instead
<
«-^ (l'\Û*\l)e
iu}
*' 1
i'('>Hi'>
+
gi^-!^.)R
w
<
2
°-
57
>
Imaginary
Frequencies.
For purely formal reasons, so that the integrals in (20.56)
converge, it is necessary to take the frequencies
to
or
UJ/I
to have imaginary parts;
u)
needs a small positive imaginary part
r],
while
U[i
needs an imaginary part
—7//.
The constant
r\
does not have much physical interest. It corresponds to turning on
the interaction potential U very slowly, a long time in the past, so that the interact-
ing system can adiabatically adjust itself to the new potential and reach a steady
state.
The constants 7// are more important. By making them nonzero, one is able
to describe transitions into metastable states. The initial state of a system is usu-
ally very stable; the system has been there for a long time, and it would remain if
not perturbed by U. However, the intermediate states indexed by /' are usually not
true eigenstates. In a formally exact description of a quantum system, the eigen-
values and frequencies should all be real. The common situation is, however, that
the quantum states /' used in describing excited states of a system are not exact,
both because one has only an approximate solution of the Hamiltonian and be-
cause the Hamiltonian actually has many more degrees of freedom than one is able
to describe. Not being exact eigenstates, these states decay, and the decay can be
modeled by giving them complex energies. All Bloch states
ip
n
-^
should really be
described in this way; they decay because electron-electron interactions have not
been treated exactly, because the real wave functions involve many electrons not
one,
because the crystal in which they live has impurities, or because of interac-
tions with phonons. Localized states, described for example by Wannier functions,
should be taken to decay for the same sorts of
reasons.
The important point is that
many different types of ignorance and error can be accommodated with a few con-
stants 7/', and only in this way does detailed comparison with experiment become
possible. The initial states |/) usually do not decay. They are, after all, states the
system chose to go to itself in the absence of outside interference.
Interaction with Electromagnetic Field. The next step is to use Eq. (20.56) to
describe a situation in which a collection of electrons interacts with an electromag-
netic field. As in Section 16.3 the sensible way for an electrical field to coexist with
periodic boundary conditions is to introduce it through a vector potential. Special-
izing to the case described by Eq. (20.24) where the wavelength of light is much
greater than unit cell dimensions or electron mean free paths, one can treat the light