Review of Literature and Research Proposal

Semiclassical Approximations in Quantum Mechanics

Report

Course code PAS 9106

Submitted to the Central University of Himachal Pradesh

for the Partial Fulfilment of the Degree of

Master of Science (Physics)

In the School of Physical and Material Sciences
In the Department of Physics and Astronomical Science

Antariksha Basu
(REG. NO. CUHP23PGPAS07)

Supervisor, Dr. Ayan Chatterjee

December, 2024
CENTRAL UNIVERSITY OF HIMACHAL PRADESH
 Abstract

In the discourse of solving Schrodinger equation, the WKB method holds a special
place. This is because of its efficiency to handle the most complicated of potentials with
relative ease compared to all other approximation schemes. In fact, recent developments
have indicated that the WKB approximation may even surpass the accuracy of results
obtained through perturbation techniques and high computational cost. Furthermore, the
WKB method also lead to some improved insights into values beyond the saddle point. In
this review, we describe the standard WKB method, and show how it may be improved.
Out of several such improvement techniques, we review the uniform WKB approximation,
and provide several examples.

5
1 Introduction

The postulates of quantum mechanics provided us with one of the most important equations in
all of physics, the Schrodinger equation. It emerged from the observations of the black body
radiation, the double slit experiment etc. The equation exactly predicted the spectral lines of
hydrogen atom, it laid the first theory and predicted the unfamiliar phenomenon of quantum
tunneling through a potential barrier. The limit of this grand equation was pretty simple, it
was a second order linear differential equation which had a room to express different potentials
of nature, if only nature was so simple. The physical world expresses itself in quite absurd
mathematical forms, hence complicated potential functions. It becomes intellectually clogged
to exactly solve the Schrodinger equation for those potential and therefore approximate
methods were looked upon. We came up with several beautiful approximations like the time
independent perturbation theory, the variational principle and so on. Then, the ideas of
Wentzel, Kramers, Brilloin and Jeffery, produced the well known JWKB approximations. The
method was constrained by validity conditions yet produced as promising result as the higher
orders of perturbation theory. At first we require to be familiar with the formal WKB
approach and its interpretations. We would look into the connection formulas and the validity
of the WKB solution. We shall also see how the WKB solutions forms through a perturbation
series in powers of ℏ, and as ℏ approaches zero, we obtain our classical solutions similar to the
Newton’s laws. Thus justifying its title of being a ”Semi-classical approximation”.

1.1 The Schrodinger equation

The Schrödinger equation is not the only way to study quantum mechanical systems and make
predictions. Other formulations of quantum mechanics include matrix mechanics, introduced
by Werner Heisenberg, and the path integral formulation, developed chiefly by Richard
Feynman. When these approaches are compared, the use of the Schrödinger equation is
sometimes called ”wave mechanics”.
The time-independent Schrödinger equation in one dimension is:
ℏ2 d2 ψ(x)
− + V (x)ψ(x) = Eψ(x) (1)
2m dx2
where: where ψ(x) is the wave function as a function of position, m is the mass of the particle,
V (x) is the potential energy as a function of position x, E is the energy eigenvalue of the
system and ℏ is the reduced Planck constant.

6
The time-independent Schrödinger equation for a particle moving in a region of constant
potential V (x) = V is given by:

d2 ψ(x) 2m
2
= 2 (E − V )ψ(x) (2)
dx ℏ

Let the wave number k be defined as:

r
2m
k≡ (E − V ) (3)
ℏ2

Thus, the solution to the differential equation becomes:

ψ(x) = Aeikx + Be−ikx (4)

where A and B are constants that can be determined by boundary conditions. The wave
number k is related to the particle’s momentum by:

p = ℏk

This represents the case where the energy of the particle is greater than the potential energy
E > V , and the solution is a combination of right- and left-moving plane waves.

1.2 The formal WKB method

The WKB method is a well known technique for solving the time independent Schrödinger
equation, preferably one dimension. It is quite useful in calculating the bound state energies
and the tunneling rates through potential barriers. Apart from that, the development of the
mathematical structure also generalises to solving most second order linear differential
equation as an approximation.
We proceed by modelling a solution of the schrodinger equation in terms of plane waves and
scaled by a functional amplitude
1
ψ(x) ∼ e ℏ S(x) (5)

7
we want to exploit the fact that at the classical level, ℏ is very close to being infinitesimal,

S(x) = S0 (x) + ℏS1 (x) + ℏ2 S2 (x) + ... (6)

Investigating the first derivative of Eq. (5)

1 ′ 1
ψ ′ (x) ∼ S (x)e ℏ S(x) (7)
ℏ
1 1
∼ [ S0′ (x) + S1′ (x) + ℏS2′ (x) + ...]e ℏ S(x) (8)
ℏ

and,

1 ′ 1 1
ψ ′′ (x) ∼ [ 2
S0 (x) + (S0′′ (x)) + 2S0′ (x)S1′ (x) + ...)]e ℏ S(x) (9)
ℏ ℏ
Now, we plug back Eq. (9) into Eq. (1) to obtain a solution that used ℏ as the perturbation
parameter,

ℏ2 1 ′ 1 1 1
− [ 2 S0 (x) + ((S0′′ (x)) + 2S0′ (x)S1′ (x)) + ...)]e ℏ S(x) = [E − V (x)]e ℏ S(x) (10)
2m ℏ ℏ
We have now expressed our formal solution as a perturbation series of ℏ, and would approach
further by comparing the powers of ℏ on both side to determine the S functions in the
perturbation series. This is one of the reasons to consider the method as a ’semi-classical’ one
since we have exploited the fact that the magnitude of ℏ to be relatively infinitesimal
compared to the energy and potential terms on the right hand side. Now, we define
Q(x) ≡ 2m(V (x) − E) and compare the terms,

O(1) : S02 (x) = Q(x), (11)

O(ℏ) : S0′′ (x) + 2S0′ (x)S1′ (x) = 0, (12)

O(ℏ2 ) : S1′2 + S1′′ (x) + 2S0′ (x)S2′ (x) = 0 · · · (13)

Now, determining the structure of the functions using Eq. (11) and Eq. (12), and neglecting
the higher order terms of ℏ,
Z x
S0 (x) = ± Q1/2 (t)dt, (14)
a
1 Q′ (x)
S1 (x) = − (15)
4 Q(x
(16)

Hence we obtain the complete solution by plugging in Eq. (14), and Eq. (15) into Eq. (5)

8
 Figure 1: This is a one turning point refering Eq. (17).

1 1 x 1/2
R
Q (t)dt − ℏ1 ax Q1/2 (t)dt
R
[C 1 e ℏ a + C 2 e ] (17)
Q1/4
We can observe that the WKB solution blows of for the turning points ie. when Q(x) = 0, and
thus we realise that the solution is very inaccurate around the turning point based on how
someone define being around the point. Let’s not worry about the validity of the approximation
and discuss the transition of the solution through either side of the turning point.

1.3 Connection Formulas of one turning point

A smooth craft discussion is present in Griffiths[1]. Nevertheless we would discuss some key
point from the literature.
Let’s consider a Q(x) for Q(x) > 0 for x < 0 and Q(x) > 0 for x > 0, and
1 i 0
R
|Q1/2 (t)|dt − ℏi x0 |Q1/2 (t)|dt
R
ψ(x) = [C 1 e ℏ x + C 2 e x<0 (18)
Q1/4
1 − ℏ1 0x |Q1/2 (t)|dt
R
[C 3 e x>0 (19)
Q1/4
We Taylor expand Q(x) to its first order to get,

9
 ℏ2 d2 ψ(x)
− + (E + V ′ (0)x)ψ(x) = Eψ(x) (20)
2m dx2
or
d2 ψ(x)
= αψ(x) (21)
dx2
where
2m ′ 1
α=[
2
V (0)] 3 (22)
ℏ
Eq. (21) is the well known Airy equation and,

ψ(x) = aAi(αx) + bBi(αx) (23)

The Ai(αx) and Bi(αx) are the Airy function that we will discuss in Chapter 2.

The Airy equations have the form,

1 ∞
Z  3 
s
Ai(z) = cos + sz ds (24)
π 0 3
1 ∞ −s3 /3+sz
Z   3 
s
Bi(z) = e + sin + sz ds (25)
π 0 3
And their asymptotic forms,
For x >> 0
1 2
−3x2
3
Ai(x) ∼ 1 1 e (26)
2π 2 x 4
1 2 3
Bi(x) ∼ 1 1 e3x2 (27)
π 2 x4
(28)

For x << 0
1 2 3 π
Ai(x) ∼ 1 sin [ (−x) 2 + ]
1 (29)
π (−x)2 43 4
1 2 3 π
Bi(x) ∼ 1 1 cos [ (−x)
2 + ] (30)
π 2 (−x) 4 3 4

(31)

Since, we have already seen that the WKB solutions are not valid around the turning points,
we use the asymptotic limits of the airy function considering it to be far away from the turning
point but sufficiently near for the Taylor approximation to be valid. Hence we have the
approximate ψ(x),
1 0 1
Z
2D π
ψ(x) = 1/4
sin [ Q 2 (t)dt + ] x<0 (32)
Q ℏ x 4
D [− 1 0 |Q 2 (t)|dt]
R 1
e ℏ x x>0 (33)
Q1/4

10
We have our first result connecting the either side of the turning point to the wave function
with only one parameter D, which can be determined using the boundary conditions.

11
2 Airy Functions

In this chapter, we would look at a qualitative discussion on solving the Airy equation and
with sufficiently rigorous mathematical details and go through its asymptomatic expansion
with the help of complex analysis. We would later have a glance at Laplace method of
stationary phase which would help us to approximate some special integrals.

2.1 The Airy equation

We begin by defining the Airy equation over complex variable z,

d2 ψ(z)
− zψ(z) = 0 (34)
dz 2
Using the Laplace transform for solving the differential equation, we address
Z
ψ(z) = dtetz Φ(t) (35)
Γ

Therefore, Z
′′
ψ (z) = t2 etz Φ(t)dt (36)
Γ

Plugging the definitions back into Eq. (34), we have

Z
(t2 − z)etz Φ(t)dt = 0 (37)
Γ

The term zψ(z) becomes Z

tz
[e Φ(z)]Γ − etz Φ′ (t)dt (38)
Γ

Thus, Z
tz
[e Φ(z)]Γ + (t2 Φ(t) − Φ′ (t))etz dt = 0 (39)
Γ

We choose the contour Γ for which

[etz Φ(z)]Γ = 0 =⇒ t2 Φ(t) − Φ′ (t) = 0 (40)

Hence, Z
t3
ψ(z) = C dtetz+ 3 (41)
Γ

We can clearly see that the integrand of Eq. (41)has no singularity and finite t, also an open Γ
with finite end point violate Eq. (40) ie. our imposed boundary condition. So, for the
integrand to vanish at Γ end points we let,
t3 |t|3 e3iα
t = |t|eiα ; e− 3 = e− 3 (42)

12
 Figure 2: The allowed open contours

we must have
Re[ei3α ] > 0 =⇒ cos3α > 0 (43)

Hence,
π π [ π 5π [ 7π 3π
α ϵ [− , ] [ , ] [ , ]
6 6 2 6 6 2
Now, the endpoints of our contour must lie in these angular regions as shown in Fig.( 2 )so to
satisfy the boundary conditions. Now, consider the close contour

Γ = C1 + C2 − C3 (44)

then, I Z Z Z
= + − =0 (45)
Γ C1 C2 C3

Hence, we have two independent solutions for the Airy equation Eq. (34). Now to compute the
integral form of the equation, we use the contracted contour Fig.( 3)

13
Figure 3: A contracted contour for C1

14
 Figure 4: Contracted contour for C3 and C2
Z
t3
I(z, C1 ) = dtetz− 3 (46)
C1

Let t=iτ , then Z ∞

τ3
I(z) = i dτ ei(τ z− 3 ) (47)
−∞

Hence,
∞
τ3
Z  
1
I(z, C1 ) = Ai(z) ∼
= cos + τz dτ (48)
π 0 3
Similarly, we use Fig. (4) to compute the other independent solution
Z ∞  3 
1 τ
I(z, C2 − C3 ) = Bi(z) ∼ −τ 3 /3+τ z
= e + sin + τ z dτ (49)
π 0 3

2.2 Laplace Method of stationary phase

We are now attempting to find approximate value of a particular for of integral which is
dependent on a parameter and we intend to find the value of the integral as the parameter
become infinitely large(see [2] and [3]),
Z b
I(s) = g(x)esf (x) dx; s→∞ (50)
a

15
We assume that f (x) has an extrema in the domain [a, b]. Therefore we can Taylor expand
f (x) around that point say x0 ,
1 ′′
f (x) = f (x0 ) + f ′ (x0 ) (x − x0 ) + f (x0 ) (x − x0 )2 + O (x − x0 )3 .


(51)
2!
1
f (x) ∼
= f (x0 ) + f ′′ (x0 )(x − x0 )2 (52)
2!
Suppose f (x0 ) is a maxima. Similarly, for g(x), we can approximate it to g(x0 ), because it is
being scaled by an exponential function. Then,
Z b
|f ′′ (x0 )|(x−x0 )2
∼
I(s) = g(x0 )esf (x0 )
e−s 2! dx; s→∞ (53)
a

Considering [a,b] for the whole plane,[-∞,+∞]

Z ∞ |f ′′ (x0 )|(x−x0 )2
I(s) ∼
= g(x0 )esf (x0 ) e−s 2! dx; s→∞ (54)
−∞

As we use the well known Gaussian integral,

Z ∞
2 π 1
e−αx dx = ( ) 2 (55)
−∞ α
We make the correspondence between Eq. (54) and Eq. (55) to get
2π 1
I(s) ∼
= g(x0 )esf (x0 ) ( )2 ; s→∞ (56)
s(|f ′′ (x 0 )|)

This is our required approximation, However, it might be so that our integral be define over
the complex plane and if f (z) is analytic, The Cauchy-Riemann condition implies a saddle
point along our contour instead of stationery points. So, we consider a special integral where
we have a phase exponential,
Z ∞
I(M ) = g(M, z)eiM f (z) dz (57)
−∞

In this case, we can use another well known integral for a Gaussian with phase term,
Z ∞
2 π 1 απ
eiαx dx = ( ) 2 ei |α| 4 (58)
−∞ |α|
Again me make the correspondence of Eq. (54), Eq. (58) with Eq. (57),
2π 1 π
I(M ) ∼
= g(M, x0 )eiM f (z0 ) ( ′′
) 2 ei 4 ; M →∞ (59)
M (|f (x0 )|)

2.3 Asymptotic forms of the Airy functions

Over the past subsections, we saw that our ideal contours can be deformed into simpler
contours along the axes lines and our analysis of the stationery phase approximation were for

16
such well behaved contours. Hence, we use the previous results to compute the asymptotic
forms of our Airy functions.
Lets consider the solution-like function as of Eq. (47)
Z ∞
1
ψ(x) = eiQ(t,x) dx (60)
2π −∞
where
t3
Q(t, x) = tx + (61)
3
Say Q(t,x) has a stationery point at t=t0 , then
dQ(t, x)
|t=t0 = 0, (62)
dt
1
t0 = ±(−x) 2 (63)

We expand Eq. (60), Z ∞

1 1 ′′ (t ,x)(x−x )2 )
ψ(x) = ei(Q(t0 ,x)+ 2! Q 0 0
dt (64)
2π −∞

Considering the case when x → −∞,

1
Q′′ (t0 , x) = ±2|x| 2 (65)

So, we accordingly have a maxima and minima based on which stationary point we choose, and
Eq. (53) takes care of it based on the sign although we derived the result for a maxima and
clearly the stationary point is not a point of inflection. We compare Eq. (59) to Eq. (64) plug
the value of Eq. (65), and finally we express the Ai(x) in terms of cosine term to obtain its
Asymptotic form,
1 2 3 π
Ai(x) ∼ 1 sin [ (−x) 2 + ]
1 (66)
π (−x)
2 34 4

(67)

Similarly, we compute the negative asymptotic form of Bi(x) from Eq. (49),
1 2 3 π
Bi(x) ∼ 1 cos [ (−x) 2 + ]
1 (68)
π (−x)
2 34 4

(69)

The same treatment is valid for the case where x →∞

1
t0 = ±i|x| 2 (70)

and
1
Q′′ (t0 , x) = ±2i|x| 2 (71)

17
To obtain,

1 2 3
Ai(x) ∼ 1 1 e− 3 x 2 (72)
2π x 2 4

1 2
x2
3
Bi(x) ∼ 1 1 e
3 (73)
π 2 x4
(74)

18
3 The uniform WKB for one turning point

In this chapter we will lay the groundwork for a method of solving the one dimensional
Schrodinger equation based on the work of Langer[4][5]. We would only focus on Airy equation
as of now and show that, for a Q(x) with one turning point can be approximated to Airy
functions. Let’s address a problem Fig. (5) that requires solution through the following
method(symbols to be discussed later)

Γ2 (x) = x2 − β 2 (75)

3.1 The modified Airy functions

We have the Schrodinger equation and the Airy equation respectively as per Langer’s[4][5]
requirement, we would go hand in hand with standard MAF formalism[6].

d2 ψ(x)
2
+ Γ2 (x)ψ(x) = 0 (76)
dx
d2 Ai(aζ)
− aζAi(aζ) = 0 (77)
d(aζ)2
Writing these you equations together immediately would draw a parallelism, a simple
introspection makes it realisable that we are trying to map the Schrodinger like equation to
that of airy function. A few glances at this suggests that if Γ2 (x) were to be expanded in its
1st order Taylor series, that would make up the Airy equation, which means although we are
interested in the asymptomatic forms of the solution, yet our first order Taylor expansion
suggests that x is very small. Such a contradictory idea can only mean the coefficients of the
higher order terms has to be abnormally small. This clearly suggests us that the potential
function must vary extremely slowly, so that all the higher derivation of our potential term
must be very small in magnitude. This simple observation sets our validity for the solution we
are gonna look upon. Consider ψ(x) in the form of Airy functions,

ψ(x) = F (x)Ai[aζ(x)] (78)

G(x)Bi[aζ(x)] (79)

where a = ±1. The reason for these possible choices of a is based on the nature around the
turning points. Its an important consideration, since this exact condition decides the behaviour

19
of our solution asymptotically. Now

dAi(x) ′
ψ ′ (x) = F ′ (x)Ai[aζ(x)] + aF (x) ζ (x) (80)
d(aζ(x))

dAi[xζ(x]) ′
ψ ′′ (x) = F ′′ (x)Ai[aζ(x)] + 2aF ′ (x) ζ (x)
d(aζ(x))
d2 Ai[aζ(x)] ′2 dAi[aζ(x)] ′′
+ a2 F (x) 2
ζ (x) + aF (x) ζ (x) (81)
d(aζ(x)) d(aζ(x))

Now, plugging Eq. (78) and Eq. (81) into Eq. (77), we have

dAi[xζ(x]) ′
F ′′ (x) = Ai[aζ(x)] + 2aF ′ (x) ζ (x)
d(aζ(x))
d2 Ai[aζ(x)] ′2 dAi[aζ(x)] ′′
+ a2 F (x) 2
ζ (x) + aF (x) ζ (x)
d(aζ(x)) d(aζ(x))
+ Γ2 (x)F (x)Ai[aζ(x)] = 0 (82)

For the third and fifth term in Eq. (82) , we can define a relation between ζ(x) and Γ(x) by,

Γ2 (x) + a3 ζ ′2 (x)ζ(x) = 0 (83)

which gives us, Z x

3 3 1 2
ζ(x) = [ (−a)− 2 (Γ(t)) 2 dt] 3 (84)
2 x0

Again, in Eq. (82) we can compute the rest of the terms neglecting F ′′ (x) terms which is our
approximation stating the effective amplitude of our solution changes very slowly.This is a
direct comparison of your previous statement, if the potential happens to change as an
infinitesimal ideal, we expect our amplitude not to vary at all. In Eq. (84), the x0 is chosen to
be the turning point which can be absorbed in the constant coefficients of the Airy functions.
Therefore,
dAi[xζ(x]) ′ dAi[aζ(x)] ′′
2aF ′ (x) ζ (x) + aF (x) ζ (x) = 0 (85)
d(aζ(x)) d(aζ(x))
giving us,
C
F (x) = (86)
ζ ′2
Hence we have our solution for Eq. (77),

C D
ψ(x) = Ai[aζ(x)] +
1 1 Bi[aζ(x)] (87)
(ζ ′ ) 2 (ζ ′ ) 2
Z x
3 − 32 1 2
ζ(x) = [ (−a) (Γ(t)) 2 dt] 3
2 x0

20
 Figure 5: The inverted harmonic oscillator potential

3.2 Approximating free state solutions for the inverted harmonic

oscillator like function

Now getting back to the problem Eq. (75), a problem that I worked upon

Γ2 (x) = x2 − β 2

Here, we are going to focus on the scattering solution for x approaching ∞, and also the right
choice would be a = −1. Then we construct the corresponding Eq. (84)
1
3 x(x2 − β 2 ) 2 1 1 β 2 lnβ 2
ζ(x) = [ ( − β 2 ln(((x2 − β 2 ) 2 + x) 2 ) + )] 3 (88)
2 2 2
As x approaches ∞,

3 x2 β 2 ln2x 2
ζ(x) = [ ( − )] 3 (89)
2 2 2
2 3 x 1 2 1
ζ ′ (x) = [ ( + )] 3 ζ − 2 (90)
3 2 2 4
Clearly ζ(x) increases as x increases.
Now, we plug in Eq. (89) and Eq. (90) into Eq. (87), with Eq. (66) and Eq. (68) to obtain the
form of asymptomatic expansion as per our required constraint x >> 0. Thus, we have our

21
Figure 6: Scattering state solutions for inverted harmonic oscillator For β=1, C1 =1.9 and C2 =3.4

uniform WKB solution

x 1 1
ψ(x) = [ + ]− 2 (C1 Sin(terms) + C2 Cos(terms)) (91)
2 4

As we expect a decline in amplitude of the sin and cosine terms, as x approaches ∞, yet this
not a good approximation asymptomatically, But there’s a keen observation to make .Unlike,
the Formal WKB method, this approximation doesn’t blow up at the turning point due to the
factor 1/4 as shown in Fig. (6), it means the uniform WKB accounts the connection on the
either side of the turning point and can be easily connected by the boundary conditions, if
given. From this example it is obvious that the uniform WKB provides a decent insightful
solution for a given problem.

22
4 Research proposal

In this particular literature, we have discussed briefly all the required toolkit to approximately
solve the Schrodinger equation. I now aim to solve several standard WKB problems and eigen
value problems like the quartic oscillator[7] to exercise and verify these methods with utmost
certainty. Then, I shall aim to apply the gained insights and the mathematical tools to
interpret spectral problems related to blackhole tunneling[8], since blackholes or general
relativity as such has always been a menace to quantum mechanics and unike SR, their
reconcilation is oblivious. So, we procced by using the results of GR and QM simultaneously
and gain as much insights as possible.

References

1. Griffiths, D. Introduction to Quantum Mechanics.

https://books.google.co.in/books?id=0h-nDAAAQBAJ (2017).

2. Ramkarthik, M. S. & Pereira, E. L. Airy Function Demystified - I. Resonance 26, 629–647

(May 2021).

3. Ramkarthik, M. S. & Pereira, E. L. Airy Function Demystified - II. Resonance 26, 757–789
(June 2021).

4. Berry, M. V. & Mount, K. E. Semiclassical approximations in wave mechanics. Rep. Prog.

Phys. 35, 315–397 (1972).

5. LANGER, I. E. On the Connection Formulas and the Solutions of the Wave Equation.
Physical Review 51 (April 1937).

6. A K Ghatak, R. L. G. & Goyal, I. C. Modified Airy FUnction and WB Solutions to the

Wave Equation. NIST Monograph 176, 180–184 (1991).

7. Delabaere, E. & Pham, F. Unfolding the Quartic Oscillator. Annals of Physics 261,
180–218. issn: 0003-4916.
https://www.sciencedirect.com/science/article/pii/S0003491697957377 (1997).

8. Hatsuda, Y. & Kimura, M. Spectral Problems for Quasinormal Modes of Black Holes.
Universe 7, 476. issn: 2218-1997. http://dx.doi.org/10.3390/universe7120476 (Dec.
2021).

23