Chapter 9: The WKB Approximation
Back to time independent problems. WKB stands for Wentzel, Kramers, Brillouin.
WKB is a technique to obtain approximate solutions to time
independent problems, mainly in 1D or where only “r” matters in 3D.
Main intuitive idea: suppose you have a potential V(x) totally constant,
no imperfections. Then, the solution if is:
Of course, here A is constant, k is constant, λ is constant.
However, a perfect flat potential is unlikely. Suppose V(x) is
“nearly” flat but changes very slowly with x, i.e. over distances
much larger than λ. Then, the solution cannot be too different:
A, k,… will now be smooth slowly varying functions of x.
1
9.1: The “Classical” Region
Let us first consider the case E>V(x), i.e.
the classical region. First, we will not
make any approximation and find exact
equations for amplitude and phase. Then,
we will make the WKB approximation.
Exactly, this can be written where
Propose , which is generic for any wave function.
Here both A(x) and φ (x) are real functions, dependent on x.
2
Real part: Imaginary part:
Cannot be solved unless we
assume A’’~0, i.e. amplitude
varies slowly with x. Exact:
3
Again, the two
exact eqs. are:
If A’’~0, then:
We started with then we arrive to:
This is the WKB approximation to the wave function.
Note φ(x) is an indefinite integral i.e. x dependent.
We will need boundary conditions.
4
Example 9.1: Potential well with two vertical walls.
E2
Assume E > V(x) for all values of x
(this may or may not be right, we have
E1 to be careful).
We found before:
In general, we have to make a linear combination:
where
5
Boundary conditions:
(1) at
This means C2 =0
(2) at
This means
6
means
where E is the unknown for each “n”.
The integral can be done analytically and an
equation for E will be found, or we can find E
numerically.
7
If V(x)=0 inside the well, then of course:
which is the
exact result.
8
9.2: “Tunneling”
Now consider regions that
are NOT classical i.e. E<V(x).
Indefinite integral,
i.e. x dependent
We can repeat all the
same and we find:
Note: no “i” in phase
and |..| in p(x)
Typical result
9
The result of previous page is a general result:
2
2
2
Example,
Problem 8.3: E
2a
This is the
where
exact result exact
from Ch. 2:
10
General WKB approx. for tunneling
through barrier of width a:
We pretend we do not know
the exact result and try to
use the WKB approximation:
Width of barrier is 2a here:
Then, WKB prediction
for tunneling is:
exact
exact
Dominant exponential is
exact properly reproduced by WKB 11
Famous example: Gamow’s
theory of alpha decay (1928)
size r1 ~ 10-15 m
12
Turns out, the integral can be done exactly, and moreover it
can be simplified considerably if
Full integral After some approximations
where
1 fm = 10-15 m is As “m” we use the mass of an alpha
the size of a particle ~ 4 proton masses
typical nucleus
Z is the positive charge of the nucleus
13
If alpha particles have an average velocity “v”
inside the well, then to travel from r=0 to r=r1 it
takes t=r1/v, i.e. hits the walls with a period 2r1/v.
At each collision the probability of remaining
trapped is e+2γ (or prob. of escape is e-2γ )
The lifetime then is τ = (2r1/v) e+2γ.
The energy of the alpha Then, ln(τ)= ln(2r1/v) +2γ with
particle is not arbitrary
but resembles that of a γ~
square well.
Z=92
~109 years
Experiments confirm that lifetime ~1000 years
depends ~ linearly on 1/sqrt{Eα} on ~1 year Z=88
a range of lifetimes from 109 years
~1 minute
to tiny fractions of seconds! Z=84
(Geiger-Nuttall law) ~ tiny fractions
of a second
14
Lecture ended here. The next three pages
are only for completeness. It is not
material that you need to know for Test 3
(final exam).
15
9.3: The connection region
In many examples we use the WKB approximation in cases
V(x) has vertical walls.
But in most real situations, this is not the case, such as in
alpha decay. We may try the “usual” procedure:
Naively we may simply be tempted to try to match
coefficients at the boundary.
16
Expected result
However, at exactly the “x” where we switch from
classical to non-classical then p(x) = V(x)-E is zero.
Then, WKB wave functions explode. Not realistic!
In practice a “patching procedure” is
followed, where a “third region” is
introduced where the potential is
linearized
17
A problem with a linear potential is exactly solvable and
leads to the Airy functions, complicated functions usually
given in an integral form. They are oscillatory on one side
and exponential on the other.
If we had a sharp wall on one
side (not the actual problem at
hand) the shape of the Airy
functions is as shown (leading
to bound states):
The WKB patching procedure would be too complicated to
describe in detail, just be aware of its existence.
18
