Lecture 8

WKB Approximation, Variational Methods and the

Harmonic Oscillator

Reading:
Notes and Brennan Chapter 2.5 & 2.6

Georgia Tech ECE 6451 - Dr. Alan Doolittle

 Wentzel-Kramers-Brillouin (WKB) Approximation
•The WKB approximation is a “semiclassical calculation” in quantum mechanics in
which the wave function is assumed an exponential function with amplitude and phase
that slowly varies compared to the de Broglie wavelength, λ, and is then semiclassically
expanded.
•While, Wentzel, Kramers, and Brillouin developed this approach in 1926, earlier in
1923, a mathematician, Harold Jeffreys, had already developed a more general method of
approximating linear, second-order differential equations (the Schrödinger equation is a
linear second order differential equation). Jeffreys is rarely given his proper credit.
•While technically this is an “Approximation method” not an “Exact solution” to the
Schrödinger equation and thus should be covered when we discuss Approximation
Methods in Chapter 4, it’s very close relationship to the simple Plane Wave solutions
warrants us discussing it now.
•The WKB method is most often applied to 1D problems but can be applied to 3D
spherically symmetric problems as well (see Bohm 1951 for example).
•The WKB approximation will be especially useful in deriving the Tunnel Current in a
tunnel diode (see Brennan section 11.6 for example). This topic will be covered in ECE
6453 in detail and only briefly introduced here.

Georgia Tech ECE 6451 - Dr. Alan Doolittle

 Wentzel-Kramers-Brillouin (WKB) Approximation
The WKB approximation states that since in a constant potential, the wave function
solutions of the Schrodinger Equation are of the form of simple plane waves,

2 m (E − U )
Ψ ( x ) = Ae ±ikx where k = 2π/λ = 2
=constant
h
if the potential , U→U(x), changes slowly with x, the solution of the Schrodinger
equation is of the form,
(*)
Ψ ( x ) = Ae iφ ( x )

Where φ(x)=xk(x). For the constant potential case, φ(x)=±kx so the phase changes
linearly with x. In a slowly varying potential, φ(x) should vary slowly from the linear
case, ±kx .
For the two cases, E>U and E<U, let k(x) be defined as (so we only have to solve the problem
once),
2 m (E − U ( x ) )
k (x ) = 2
for E > U(x)
h
2m(U ( x ) − E )
k ( x ) = −i 2
= - iκ (x) for E < U(x)
h
Georgia Tech ECE 6451 - Dr. Alan Doolittle
 Wentzel-Kramers-Brillouin (WKB) Approximation
Using the normalized version of (*) the Schrodinger Equation,
− h2 ∂2
Ψ ( x ) + U (x )Ψ ( x ) = EΨ ( x )
2m ∂x 2

Becomes,
2
∂ 2φ  ∂φ 
i 2 −   + (k ( x )) = 0
2

∂x  ∂x 
The WKB approximation assumes that the potentials are slowly varying. If this is the
case, k(x) is also slowly varying and so φ(x) slowly varying. Thus, the 0th order WKB
Approximation assumes, ∂ 2φ
≈0
∂x 2
and thus,
Note that the subscripts of
2
 ∂φ0  zero, 0, emphasize that this
 = (k ( x ))
2
 or solution comes from the 0th
 ∂x  order approximation.
φ0 ( x ) = ± ∫ k ( x )dx + C0 and thus,
Ψ (x ) = e i (± ∫ k ( x )dx +C0 )
Georgia Tech ECE 6451 - Dr. Alan Doolittle
 Wentzel-Kramers-Brillouin (WKB) Approximation
If a more accurate solution is required, we can rewrite our previous starting point as,
2 2
∂ 2φ  ∂φ   ∂φ  ∂ 2φ
i 2 −   + (k ( x )) = 0 ⇒   = (k ( x )) + i 2
2 2
or
∂x  ∂x   ∂x  ∂x
∂φ ∂ 2φ
= (k (x )) 2
+i 2 so,
∂x ∂x
∂ 2φ
φ ( x ) = ± ∫ (k ( x )) + i 2 dx + C1
2

∂x
So far, no approximation has been made (i.e. this is an exact solution).
The 1st order WKB approximation assumes that since,
2
 ∂φ0 
 = (k ( x )) from the 0th order solution,
2

 ∂x 
∂φ0
= ± k ( x ) so, Note that the subscripts of 1,
∂x
emphasize that this solution
∂ 2φ0 ∂
= ± k ( x ) thus, comes from the 1st order
∂x 2
∂x approximation.
∂
φ (x ) = ± ∫ (k ( x ))2 ± i k (x )dx + C1 and thus,
∂x
 
i  ± ∫ (k ( x ))2 ±i ∂ k ( x )dx +C1 
Ψ (x ) = e  ∂x 
Georgia Tech ECE 6451 - Dr. Alan Doolittle
 Wentzel-Kramers-Brillouin (WKB) Approximation
Some important points about the WKB Approximation:
1) We only need to know the shape of the potential to estimate the wave function since,
 
i  ± ∫ (k ( x ))2 ±i ∂ k ( x )dx +C1 
∂
U (x ) ⇒ k (x ) ⇒ φ (x ) = ± ∫ (k (x ))2 ± i k ( x )dx + C1 ⇒ Ψ (x ) = e  ∂x 
∂x
2) Since we are limited to slowly varying U(x) (compared to λ), then we can say,
∂
k ( x ) << (k ( x ))
2

∂x
and thus, the 1st order approximation is only slightly different from the 0th order approximation
3) The WKB approximation breaks down at regions where E~U (points when classical particles will
turn around and change directions – Classical Turning Points). In this case, the wavevector, k(x),
approaches zero but its derivative does not.
∂k ( x ) ∂  2 m (E − U ( x ) ) 
k ( x ) = 0 but = −i  ≠0

∂x ∂x  h2 
∂
k ( x ) << (k ( x )) does not hold true!
2
In this case, clearly
∂x
In these special cases, connection formulas must be used to tie together regions on either side of
the classical turning point. In all other regions WKB is valid (See Merzbacher 1970 for details).

Georgia Tech ECE 6451 - Dr. Alan Doolittle

 Wentzel-Kramers-Brillouin (WKB) Approximation
Example:
Consider the tunneling probability at a finite width E<U(x) U(x)
potential barrier.
Ψ * (L )Ψ (L )
T= * where Ψ (x ) = Ψ (0)e i (± ∫ k ( x )dx +C1 ) x=0 x=L x
Ψ (0)Ψ (0) Region I Region III

But for tunneling to occur, E<U so,

2m(U ( x ) − E )
k ( x ) = −i for E < U(x) = U
h2
 x 2 m (U ( x ) − E ) 
i  ± −i
∫ dx
 
Ψ (x ) = Ψ (0 )e
2
 0 h 

 x 2 m (U − E ) 
− dx 
 ∫ 
Ψ (x ) = Ψ (0 )e  0 h2 

  2 m (U − E )  
 − x
   
Ψ (x ) = Ψ (0 )e   h2  

Thus,
  2 m (U − E )     2 m (U − E )  
 − L  − L   2 m (U − E )  
       
Ψ (L )Ψ (L ) Ψ (0 )e Ψ (0 )e
*   h2     h2    −2 L 
  h 2  
T= = =e    

Ψ * (0)Ψ (0 )  
 −
 
2 m (U − E )  
0
 
  2 m (U − E )  
 −
 
0 
 
Ψ (0 )e   Ψ (0 )e
h2     h 2
 
Georgia Tech ECE 6451 - Dr. Alan Doolittle
 Variational Methods of Approximation
The concept behind the Variational method of approximating solutions to the Schrodinger
Equation is based on:
a) An educated guess as to the functional form of the wave function. Often this is based
on a similar problem that has an exact solution.
b) A “Variational parameter” that will be adjusted to obtain a minimum in the eigen
energy.
c) Recognition that all natural systems seek the lowest energy state.
Using all of the above, one minimizes the expected energy by iterative methods.

Given,
E = Ψ ( x, α ) H Ψ ( x, α )
Find the value of α for which E is minimum. Thus,
d d
E = Ψ ( x, α ) H Ψ ( x, α ) = 0 and
dα dα
d2 d2
E = Ψ ( x, α ) H Ψ ( x, α ) > 0
dα 2
dα 2

Georgia Tech ECE 6451 - Dr. Alan Doolittle

 Harmonic Oscillator
In many physical systems, kinetic energy is continuously traded off with potential energy.
Thus, as kinetic energy increases, potential energy is lost and vice versa in a cyclic
fashion. When the equation of motion follows,

d 2x
ma = m 2 = F = − kx
dt
1 2
V(x) = kx
a Harmonic Oscillator results. 2

The term -kx is called the restoring force.

The radian frequency of such an oscillation is,
k
ω= while the potential can be found by integrating the force, F = -kx
m
1
V(x) = kx 2
2
Thus, the Schrodinger equation for this potential is,
h2 d2 1 2
- 2
Ψ ( x ) + kx Ψ ( x ) = EΨ ( x )
2m dx 2
h2 d2 1
- 2
Ψ ( x ) + m ω 2 x 2 Ψ ( x ) = EΨ ( x )
2m dx 2
Georgia Tech ECE 6451 - Dr. Alan Doolittle
 Harmonic Oscillator Solution
The power series solution to this problem is derived in Brennan, section 2.6, p. 105-113 and
is omitted for the sake of length. Instead we will only discuss the operator based
solution. There is an infinite series of possible solutions described by:

 y2 
mω
Cn −  
πh mω
Ψn ( x ) =

 2 
n
hn ( y )e where C n ≡ and y ≡ x
 
2
n! h
2
Which has Eigen Energy solutions of,
 1
E n =  n + hϖ for n = 0, 1, 2, 3...
 2

The functions, hn(y) are Hermite polynomials defined by,

h0 ( y ) = 1
h1 ( y ) = 2 y
h2 ( y ) = 4 y 2 − 2

And the recursion relation,

d
hn +1 ( y ) = 2 yhn ( y ) − hn ( y ) for n ≥ 2
dy
Georgia Tech ECE 6451 - Dr. Alan Doolittle
 Harmonic Oscillator Operator Solution
Let us define two operators* as,
1  d 
a− =  y + 
2 dy 

which for reasons that will become obvious, we will call the annihilation or lowering
operator and,

1 d 
a+ =  y − 
2 dy 
which we will call the creation or raising operator.
Consider the product of these two operators:
Lets focus on this
1 d  d  expression for a
a − a + =  y +  y − 
2 dy  dy  moment.

− 1  2 d2
+ d d 
a a =  y − 2 − y + y 
2 dy dy dy 
− 1 2 1 d 2 1 d 1 d 
+
a a = y − − y + y
*Note: a+ and a- are 2 2 dy 2 2 dy  2 dy 
NOT Hermitian!
Georgia Tech ECE 6451 - Dr. Alan Doolittle
 Harmonic Oscillator Operator Solution
Cont’d:
− 1 2 1 d 2 1 d 1 d 
+
a a = y − 2
− y + y
2 2 dy 2 dy  2 dy 

Using the fact that

d
( ya ( y ) ) = a ( y ) dy + y d (a ( y ) ) = 1 + y d a ( y )
dy dy dy  dy 
− + 1 2 1 d 2 1 d 1 1 d 
a a = y − − y + + y
2 2 dy 2 2 dy  2 2 dy 
− + 1 2 1 d2 1
a a = y − +
2 2 dy 2 2
Let H' be defined as,
1 2 1 d2
H' = y − 2
such that,
2 2 dy
1
a − a + = H'+ and since it can be shown that the commutation,
2
a − a + − a + a − = 1 then,
1
a + a − = H'− (alternatively the same approach could be taken to directly evaluate a + a −
2
Georgia Tech ECE 6451 - Dr. Alan Doolittle
 Harmonic Oscillator Operator Solution
What we seek to do now is to eliminate the non-physical variable, y, and cast these results in
terms of our physical variables x, ω, m, and momentum p.
mϖ d h d
y= x and =
h dy mϖ dx
Thus,
1  d 
a− =  y + 
2 dy 
1  mϖ h d 
a− =  x + 
2  h mϖ dx 
h d
or using the fact that p x =
i dx
1  mϖ h ip x 
a− =  x + 
2 h m ϖ h 

1  mϖ 1 
a− =  x + i p 

2 h mhϖ 
x

Similarly,
1  mϖ 1 
a+ =  x − i p 

2 h mh ϖ x

Georgia Tech ECE 6451 - Dr. Alan Doolittle
 Harmonic Oscillator Operator Solution
Cont’d:
1  mϖ 1  1  mϖ 1 
a− =  x + i p  and, a+ =  x − i p 

2 h mh ϖ x 
2 h mh ϖ x
 
mϖ 2 1 mϖ 1 mϖ 1
a −a + = x + p x2 + i px x - i xp x
2h 2mhϖ h mhϖ h mhϖ
mϖ 2 1 i
a −a + = x + p x2 + (p x x - xp x )
2h 2mhϖ 2h
1  p x2 1  i
− +
a a =  + mϖ 2 x 2  + [p x , x ]
hϖ  2 m 2  2 h

[p x , x ] = h
i
1
U ( x ) = mϖ 2 x 2
2
 p x2 
H =  + U ( x ) 
 2m 

H 1
Thus, a −a + = +
hϖ 2

Georgia Tech ECE 6451 - Dr. Alan Doolittle

 Harmonic Oscillator Operator Solution
Cont’d:
H 1 1
Comparing, a −a + = + and H' = a − a + - one sees,
hϖ 2 2
H
H' = which we called the normalized Hamiltonian (i.e. unitless)
hϖ
Thus, solving for H and inserting in the Schrodinger equation,
 − + 1  − + 1
 a a − hϖ = H ⇒  a a − hϖΨ (x ) = EΨ (x )
 2  2
or
 − + 1 E
 a a −  Ψ (x ) = Ψ (x )
 2  h ϖ

But since the commutator, a − a + − a + a − = 1,

a −a + = a +a − + 1
 − + 1  + − 1
a a −  = a a + 
 2  2
We arrive at two new forms of the Schrodinger Equation :
 − + 1 E  1 E
 a a −  Ψ (x ) = Ψ (x ) and  a + a − +  Ψ (x ) = Ψ (x )
 2  h ϖ  2  h ϖ

Georgia Tech ECE 6451 - Dr. Alan Doolittle

 Using the new form of the Schrodinger Equation:
Harmonic Oscillator Operator Solution
Cont’d:
The new forms of the Schrodinger Equation :
 − + 1 E  1 E
 a a −  Ψ (x ) = Ψ (x ) and  a + a − +  Ψ (x ) = Ψ (x )
 2 hϖ  2 hϖ

Acting on the Schrodinger Equation with the creation operator,

 1 E
a +  a − a + −  Ψ (x ) = a + Ψ (x )
 2  h ϖ
 + − 1 + E +
 a a − a Ψ (x ) = a Ψ (x )
 2  h ϖ
Adding a + Ψ (x ) to both sides,
 + − 1 + (E + hϖ ) a + Ψ (x )
 a a + a Ψ (x ) =
 2 hϖ
Thus, if “Ψ“ is an eigenfunction of (a+a- + 1/2 ) with energy eigenvalue, E/ħω, then so is
“a+Ψ“ but with an energy eigenvalue of (E+ħω)/ħω. Thus, the raising/creation operator , a+,
raised the energy of the state by exactly one quantum, ħω.
Similarly, the lowering/annihilation operator, a-, lowers the energy of a state by one
quantum, ħω. (This is easily proved by the same procedures using a- instead of a+ - see
Brennan, page 117).

Georgia Tech ECE 6451 - Dr. Alan Doolittle

 Using the new form of the Schrodinger Equation:
Harmonic Oscillator Operator Solution
Cont’d:
Since the energy of the Harmonic Oscillator is never less than zero, some “ground state” Ψ0
must exist for which lowering the energy with the lowering operator, a-, must give zero
energy.
 − + 1 − (E 0 − hϖ ) a − Ψ (x )
 a a − a Ψ 0 (x ) =
hϖ
0
 2
If a − Ψ 0 (x ) ≠ Null State (0) then E 0 is not the lowest (ground state). Thus,
a − Ψ 0 (x ) = 0 and,
 + − 1 E
 a a +  Ψ 0 (x ) = 0 Ψ 0 (x )
 2 hϖ

 a {a Ψ 0 (x )}+ Ψ 0 (x ) = 0 Ψ 0 (x )
 + − 1  E
 2  hϖ
 + 1  E0
 a {0} + Ψ 0 (x ) = Ψ 0 (x )
 2  h ϖ
1 E
Ψ 0 (x ) = 0 Ψ 0 (x )
2 hϖ
hϖ
E0 =
2

Georgia Tech ECE 6451 - Dr. Alan Doolittle

 Using the new form of the Schrodinger Equation:
Harmonic Oscillator Operator Solution
Cont’d:
To arrive at all the other higher energy states, we simply apply the raising operator and
require normalization. Thus,

Ψ n (x ) = C n (a + ) Ψ 0 (x ) Where C n is the normalization coefficient (previously defined)

n

Similarly, the energy of the higher states is merely nhϖ larger than the ground state energy
1 
E n = nhϖ + E 0 =  + n  hϖ
2 

But what is Ψ0? a − Ψ0 ( x ) = 0

1  d 
a− =  y + 
2 dy 
1  d 
 y +  Ψ0 ( x ) = 0
2 dy 
dΨ0 ( x )
= − yΨ0 ( x )
dy
 y2 
- 

 2
Ψ0 ( x ) = Ae 

Georgia Tech ECE 6451 - Dr. Alan Doolittle

 Using the new form of the Schrodinger Equation:
Harmonic Oscillator Operator Solution
Cont’d:
Finding the successive excited states is merely just applying the raising operator. For
example the first excited state is:
Ψ1 ( x ) = a + Ψ0 ( x )
 y2 
- 

 2
Ψ1 ( x ) = a + A 0 e 

 y2 
- 
1  d   2

Ψ1 ( x ) =  y -  A 0 e 

2  dy 
 y2 
- 
2  2

Ψ1 ( x ) = A 0 ye 

2
 y2 
- 

 2
Ψ1 ( x ) = A1 ye 

Other examples left as homework problems.

Georgia Tech ECE 6451 - Dr. Alan Doolittle