Quantum Mechanics

2019 fall lecture 6:

WKB Method and Semi-Classical Approaches

Daw-Wei Wang
Physics Department National Tsing Hua University, Taiwan
In this lecture, we will discuss the following three subjects:

I. Basic Concept of Semi-Classical Theory

II. The Wentzel-Kramers-Brillouin (WKB) Method

III. Local Density Approximation (Thomas-Fermi Approximation)

for an Interacting Condensate
I. Basic Concept of Semi-Classical Theory
1. In the early stage of the formation of Quantum Mechanics,
physicists try to understand some quantum phenomena in the
frame work of classical theory, for example, spectrum of H atom.
2. In order to explain the quantization of spectrum, Bohr assumes
that the quantization comes from the angular momentum, which
reflects something similar to “uncertainty principle”.
Of them implies the other. It means that
momentum and position can be
measured precisely at the same time,
but cannot be arbitrary value, due to
the standing wave picture.
3. Therefore, the spirit of semi-classical approach includes:
(1) to introduce the concept of de Broglie matter wave,
(2) to assume a position-dependent wavelength (or momentum),
(3) to make quantization from the boundary condition of wave

4. We could start from the energy conservation of a particle moving

in 1D system in the classical picture:

where the wavelength is

5. Let us consider an infinite square well, , for |x|>a.

Exact !
6. To understand how the wavefunction behaves in such semi-
classical approach, we start from a plane wave in uniform space:

7. As a result, when considering a smooth potential, V(x),

where in semi-classical approach,
we will have

which is so-called Sommerfeld’s Quantization Rule ( 0<γ <1

is a constant, depending on the boundary condition).

8. Although above result may give good approximation on the

energy quantization, but the boundary condition of the phase
and the amplitude of wavefunction are still unknown.
II. The Wentzel-Kramers-Brillouin (WKB) Method
1. In order to get a better description on the wavefunction in a
non-uniform potential, we have to apply such semi-classical
approximation to the original Schrodinger equation  WKB.
(These are two-way directions: Classical  Quantum)
2. The static 1D Schrodinger equation is :

Inspired by previous study, we could always assume that the

wavefunction can be also written as following form:

Where the “phase function”, , can be also complex. This

does not limit any possibility of the wavefunction.
3. Therefore, we can insert it into the differential equation
above, and get

4. Above “wave equation” shows a hierarchy of .

When , we get it back to the zeroth order semi-classical
solution:

5. If we expand in terms of (measurement of QM)

Such that we have

6. As a result, we can easily get (to the first order of ):

7. Therefore, after integration, we have

and hence the wavefunction becomes

The correction of amplitude comes

from the second order of .
8. The criteria for such approximation is to make sure the
coefficient of term in is much smaller than the constant
term, i.e

Why not ?
9. The position dependent wavelength comes from potential !

This means WKB is good in either one of the condition:

(1) Smooth potential ( small), or (2) High energy ( large)
10. Therefore, we understand that WKB is a good approximation
only when the potential gradient is so small (i.e. smooth) compared
to the local kinetic energy. In other words, every position looks
like a uniform plane. As a result, it is equivalent to “Local
Density Approximation (LDA)” in condensed matter theory,
where the interaction effects is also treated locally, depending on
local density (rather than density correlation with other places
where a full quantum mechanics should have). This is because
position (x) and momentum (p) could not be measured precisely in
QM, but is approximated allowed in semi-classical approximation.

11. So far we find that although the WKB wavefunction and phase
is derived in 1D case but the concept of Semi-Classical Approach
or Local Density Approximation still works in any dimension.
12.We now apply WKB to study the bound state problem by
considering the boundary condition of a soft boundary. In other
words, the potential can be linearized near the turning point with
a finite slope

Regime I Regime II Regime III

13. After linearizing the potential Schrodinger equation can be
transformed into Airy equation, which has a known solution:
14. The solution, Airy equation has the following properties:
15. Therefore, we can obtain the following asymptotic form of
the wavefunction inside the classical allowed regime (Regime II):

16. In Regime I and III (classical forbidden), the wavefunction is

17. As a result, to match the wavefunction in the middle upto a
sign (which means the symmetric or anti-symmetric of the
wavefunction with nodes or not), we must have

where , or we have the following quantization rule:

Sommerfeld’s
Quantization Rule
18. Note that we can rewrite the quantization rule by integrating the
close classical path about the turning points, and get

which is exactly the Sommerfeld’s Quantization Rule, and can be

extended to higher dimension space:

19. It is easy to see that this quantization of orbits can be easily

interpreted as the quantization of energy:

This explain how to connect the quantization

of energy to the classical motion.
20. Example I: Consider

Where Therefore the quantized

energy is
21. Example II: For a half-infinity potential well,
22. Example III: Bound state in Central Potential. Although WKB
applies to 1D problem only, but we could still apply it after
transforming the higher dimensional problem to 1D problem.
Consider a central force potential,

Using separation of variables, , we have

Define , the problem becomes the same form as 1D

Therefore, for a short-ranged attractive potential, for example,
ven der Waal potential or nuclear force, we expect the following
“effective 1D potential”

Repulsive
centrifugal
potential
23. WKB can be also applied in tunneling case. Consider a
simple model to describe the emission of alpha particle in the
following potential well with a barrier in the nucleus of a reactive
atom.
24. In general, consider the tunneling problem of a moving
particle from left to a potential barrier. WKB can be applied to
estimate the transmission rate easily

a b

a b
25. In summary, we can find the following features for WKB
methods:

Advantages:
(1) Simple calculation (wavefunction form is ready)
(2) Physics intuition (for local density approximation)
(3) Good result for excited states (in classical limit)
(4) Not bad for ground state and lowly excited states
(5) Can be used for both bound state and unbound states

Disadvantages:
(1) Useful only in one-dimensional system
(2) Cannot be applied for interacting systems
(3) Not easily extended to higher order
III. Local Density Approximation for an
Interacting Condensate
1. Recall that the ground state energy of interacting bosons can be
described by the following energy functional:

where is the many-body wavefunction within

meanfield approximation and can be normalized to particle
number

2. The dynamics of such many-body wavefunction (condensate)

is given by its energy variation with respect to wavefunction:
3. The static non-linear Schrodinger equation can be derived from
the fact that for a many-body system, the deviation of the total
energy with respect to the change of particle number is
defined to be “chemical potential”, i.e. we have

4. One could solve such static non-linear Schrodinger equation

(also called Gross-Pitaevskii Equation) for a given , and then
change its value to satisfy the normalization condition,
This is a grand-canonical ensemble.
Note that, if we want to apply variational method for the
many-body wavefunction, we have to minimize the total
energy, not the chemical potential.
See Lecture Notes in Variational Method.
5. Such Gross-Pitaevskii equation (GPE), do not have an analytic
solution in general. However, since the interaction term is usually
dominant the single particle (non-interacting part, including
kinetic and potential energy), it is reasonable to assume the
kinetic energy can be neglected for a repulsive interaction, so
that the wavefunction is vary flat and hence has less kinetic
energy:

Kinetic Part Potential Part Interaction Part

6. This is so-called Thomas-Fermi Approximation, which is

equivalent to semi-classical approximation by assuming
momentum can be defined (~zero) in each position. We could get
the following analytic solution:
7. We may again consider a parabolic potential trap in the system
of ultracold atoms, for example:

8. It is easy to calculate the “radius” in the three directions, and

we could obtain

9. For an isotropic potential, it is easy to calculate the total

number of particle for the determination of chemical potential
 Single particle
(Gaussian)

Trial wavefunction
(Gaussian)

Fully numerical
solution of GPE

Thomas Fermi
approximation