QM-III Term Paper

The Quantum Double Well

Potential and its Applications

Author:
Supervisor:
Vivek Pareek
Dr. Siddharth Lal
10MS46

December 7, 2013
 Abstract

Double well potential problem for bound state is a very interesting problem
in Quantum Mechanics. This problem is a model problem and is used to
describe phenomenon like ammonia flipping, macroscopic tunneling, macro-
scopic coherence and Bloch waves. Here in our term paper we have tried
to solve this problem using WKB Approximation and Instanton Method.
Using the results, we have then tried to put light on the above mentioned
phenomenons.
Contents

1 Intoduction 2

2 Perturbation Approach 3

3 WKB Approximation 5
3.1 General Formalism . . . . . . . . . . . . . . . . . . . . . . . . 5
3.2 Connection Formulas . . . . . . . . . . . . . . . . . . . . . . . 7
3.3 Solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
3.3.1 Eigenfunction . . . . . . . . . . . . . . . . . . . . . . . 8
3.3.2 Eigenvalues (Energy levels) . . . . . . . . . . . . . . . 11

4 Istanton Method 15
4.1 The Path Integral Formulation . . . . . . . . . . . . . . . . . 15
4.2 Imaginary Time Formalism . . . . . . . . . . . . . . . . . . . . 17
4.3 Solving Double well . . . . . . . . . . . . . . . . . . . . . . . . 18

5 Phenomenons 22
5.1 Ammonia Inversion . . . . . . . . . . . . . . . . . . . . . . . . 22
5.2 Macroscopic Quantum Coherence . . . . . . . . . . . . . . . . 22
5.3 More Applications . . . . . . . . . . . . . . . . . . . . . . . . 24

1
Chapter 1

Intoduction

We can write a symmetric double well potential as

V (x) = λ2 (x2 − a2 )2 (1.1)
where λ is some real constant.
V(x)

x
-a +a

Figure 1.1: Double-well potential

If we consider the classical picture, a particle with energy less than the
central potential barrier in one of the well can never reach to the other side.
But quantum mechanics allows it to tunnel through the barrier and appear
in the other well.
We will be dealing with a particle with E < V (0) in a double well potential
and would try to find out its energy eigen states in this report .
One can solve this problem using the time-independent Shrodinger differ-
ential equation. But this potential has quartic terms in position and which
makes it extremely difficult to solve analytically exactly. Thus, we have to
use approximate methods to solve this. The first method that comes to the
mind is perturbation. Let’s try to solve using that and see why it fails here.

2
Chapter 2

Perturbation Approach

As discussed above our first line of thought was to apply perturbation to

the problem and see if we can get any sensible result out of it. In perturba-
tion technique we calculated the form of the Hamiltonian and the correction
terms near one of the minima. The double well potential is given by

V (x) = λ2 (x2 − a2 )2 (2.1)

where λ is a real constant and is considered to be small . We can see that the
potential has a reflection symmetry about the y axis and has two equivalent
classical ground states at x=±a(see Fig 2.1). We expand the potential about

V(x)

x
-a +a

Figure 2.1: In the double well potential we have reflection symmetry under
x → −x

one of the minima x = a and introduce a new variable y = x−a. Substituting

3
we get,

V (y) = λ2 (y + 2a)2 y 2
= 4λ2 a2 y 2 + 4λ2 ay 3 + λ2 y 4
1
= mω 2 y 2 + 2λωy 3 + λ2 y 4 (2.2)
2
1
where we have substituted mω 2 = 4λ2 a2 to eliminate a. We observe here
2
that the potential has the form of a harmonic oscillator with higher order
terms contributing to the perturbation about x = a.
Similar calculations can be made about x = −a and we get the same
form of the above potential. This suggests that we have a doubly degenerate
system (to all orders in energy) but this violates a well established theorem
in QM saying, there is no degeneracy for one dimensional systems. Hence
our initial assumption of two distinct harmonic oscillator has failed. This
is because of the fact that we did not consider the tunneling effect of the
particle while using perturbation.
But our problem does not end here. With our earlier assumptions we can
say that the expectation value of x for particle starting out in the right well
will be +a and that of the left well will be -a. This is because of the fact
that the two potential well would have distinct ground state wave function.
But we know that for symmetric potential we get an even symmetric ground
state wave function. This means that the expectation value of x should be
zero.
We can now see that the perturbation technique gives us incorrect results
for both the wave function as well as the energy spectrum. Hence we have to
search for a technique that takes care of the problems faced in this formalism.

Note: Degenerate Perturbation technique can not be used for the above
case. Degenerate Perturbation is used for degeneracy in a single system but
here we have degeneracy for two different systems, namely, one expanded
around x = a and other expanded around x = -a. Most importantly we will
not be able to capture the tunneling effect through any order perturbative
expansion.

4
Chapter 3

WKB Approximation

3.1 General Formalism

Our aim is to solve the following equation,
d2 ψ 2m
+ 2 (E − V (x))ψ = 0 (3.1)
dx2 ~

or
d2 ψ p2 (x)
+ 2 ψ=0 (3.2)
dx2 ~

where p(x) = (2m[E − V (x)])1/2 . We know for a constant V (x) we have a

solution of the form
ψ = exp(ip(x)/~) (3.3)
For our case we will be considering a potential which is very slowly varying
with x. So we consider the following ansatz to our problem of slowly varying
potential,
ψ = exp(iφ(x)/~) (3.4)
where φ(x) can be complex. Putting the solution into the equation, we get,
 ′ 2
φ iφ′′ p2 (x)
− + + 2 =0 (3.5)
~ ~ ~
We now expand φ in a power series in ~. The logic behind the expansion of
ψ is as follows, if ~ → 0, the wavelength λ = 2π~/p tends to zero. This is
to ensure that any potential can be considered slowly varying in this limit
which was our initial assumption.Looking at the expansion,
φ = φ0 + ~φ1 + ~2 φ2 + ... (3.6)

5
where φ1 is the first term, φ2 is the second term in the power series expansion
and so on. The WKB approximation consists of keeping just the first two
terms in the expansion. Putting the truncated expansion into Eq 3.5 and
grouping terms with the same ~ dependence, we get

−(φ′0 )2 + p2 (x) iφ′′0 − 2φ′1 φ′0

+ + O(~0 ) = 0 (3.7)
~2 ~
In the first approximation we consider the ~−2 term, which gives
Z x
φ0 (x) = ± p(x′ )dx′ (3.8)

Now we consider the ~−1 term. For the ~−1 to vanish, we need

iφ′′0 = 2φ′1 φ′0

φ′′0
⇒ ′
= −2iφ′1
φ0
⇒ ln φ′0 = −2iφ1 + c
⇒ φ1 = +i ln(φ′0 )1/2 + c/2i = i ln p1/2 + c1

where c and c1 are constants.Hence to this order in ~ we get

  Z x 
1/2 i ′ ′
ψ = exp(iφ(x)/~) = A exp(ln[p(x)] ) exp ± p(x )dx
~
  Z x 
A i ′ ′
= exp ± p(x )dx (3.9)
[p(x)]1/2 ~

Finally we can see the wave function of the following form

  Z x 
A i ′ ′
ψ(x) = exp ± p(x )dx (3.10)
[p(x)]1/2 ~

where A is the normalization constant. We can see that the probability

density has 1/p(x) dependence which is classically intuitive. A particle with
a greater momentum at a point is less likely to be found at that point.
Another point to be noted is that the condition for the validity of the WKB
approximation is
φ′′ φ′
| 0 | << | 0 |2 (3.11)
~ ~
Physical interpretation: The change of momentum per unit wavelength
should be much less than the momentum it self at that point.

6
3.2 Connection Formulas
All is well and good in WKB approximation unless we hit classical turning
point i.e, the point where E = V (x) , we see that ψ blows up. We need to
fix this to be able to use Eq as the general solution of Eq. We now need to
’patch’ up the wave function in the vicinity of the turning points. This is
done by approximating the potential by a straight line,
V (x) = E + V0′ (x) (3.12)
This is basically keeping first two terms of the Taylor series expansion of the
potential about the turning point. Putting it into the Eq we get,
d2 ψp
= zψp (3.13)
dz 2
where z ≡ αx and  1/3
2m ′
α≡ V (0)
~2

This is Airy’s Equation and its solutions are Airy’s functions. Now the task
that remains is, to match it to the WKB solution in the overlap regions on
either sides of the turning point. The only trouble is to find these overlap
regions where ψp is a good approximation(close to turning points) and yet
WKB solution holds good(yet far enough from turning points).

The mechanism to carry out the patch up will not be discussed here but
the technique will be used extensively to find the wave function for the double
well potential. Still we would like to mention the form of wave function for
right-hand and left hand turning points,
Right-hand turning point:
 Z x0 
2D 1 ′ ′ π
ψ(x) = p sin p(x )dx + ; f or x < x0
p(x) ~ x 4
1 x
 
D
Z
′ ′
= p exp − |p(x )|dx ; f or x > x0 (3.14)
p(x) ~ x0
Left-hand turning point:

D′ 1 x0
 Z 
′ ′
ψ(x) = p exp − |p(x )|dx ; f or x < x0
p(x) ~ x
 Z x
2D ′

1 ′ ′ π
= p sin p(x )dx + ; f or x > x0 (3.15)
p(x) ~ x0 4

7
3.3 Solution
3.3.1 Eigenfunction
Now we are in a position to tackle our problem of a bound state in double
well potential. Consider the x > 0 portion of the potential. We can see that
we have two classical turning points x1 and x2 where we need to ’patch’ the
wave function.
Mean while let us consider the general WKB solution for x > 0,

1 x
 
D
Z
′ ′
ψW KB (x) = p exp − |p(x )|dx ; x > x2
p(x) ~ x2
 Z x2
i x2
   
1 i
Z
′ ′ ′ ′
= p B exp p(x )dx + C exp − p(x )dx ; x1 < x < x2
p(x) ~ x ~ x
 Z x1
1 x1
   
1 1
Z
′ ′ ′ ′
= p F exp |p(x )|dx + G exp − |p(x )|dx ; 0 < x < x1
p(x) ~ x ~ x

From connection formula we know for right side turning point i.e, x2 for this
case, for x1 < x < x2 we have,
 Z x2 
2D 1 ′ ′ π
ψW KB = p sin p(x )dx + (3.16)
p(x) ~ x 4

Now to effect the join at x1 we re-write Eq 3.6 as,

 Z x2
1 x

2D 1 π
Z
′ ′ ′ ′
ψW KB = p sin p(x )dx − p(x )dx +
p(x) ~ x1 ~ x1 4
 Z x1 
2D 1 π
= −p sin p(x′ )dx′ − θ − (3.17)
p(x) ~ x 4

1 R x2
where p(x′ )dx′ . Now we shift the origin to x1 , and we get,
~ x1
 Z 0
1 0
   
1 1
Z
′ ′ ′ ′
ψW KB (x) = p F exp |p(x )|dx + G exp − |p(x )|dx ; x<0
p(x) ~ x ~ x
 Z x1 
2D 1 ′ ′ π
= −p sin p(x )dx − θ − ; x>0 (3.18)
p(x) ~ x 4

From connection formula we get,

ψp (x) = aAi(−αx) + bBi(−αx) (3.19)

8
 r
3 2mV ′ (0) √
with α = 2
and a,b are constants and p(x) = ~α3/2 x. Now we
~
will match the wave functions for overlap regions in x < 0 and x > 0.
R0 2
Overlap Region 1 (x < 0): x |p(x′ )|dx′ = ~(−αx)3/2
3
    
1 2 3/2 −2 3/2
ψW KB ≈ √ F exp (−αx) + G exp (−αx)
~α3/4 (−x)1/4 3 3
   
a −2 3/2 b 2 3/2
ψp ≈ √ exp (−αx) +√ exp (−αx)
2 π(−αx)1/4 3 π(−αx)1/4 3
(3.20)

Now equating the coefficients of ψW KB and ψp we get,

r
π
a = 2G (3.21)
~α
r
π
b=F (3.22)
~α

Rx 2
Overlap Region 2 (x > 0): 0 p(x′ )dx′ = ~(αx)3/2
3
 
2D 2 3/2 π
ψW KB ≈ √ sin (αx) − θ −
~(α3 x)1/4 3 4
   
a 2 3/2 π b 2 3/2 π
ψp ≈ √ sin (αx) + +√ cos (αx) +
π(αx)1/4 3 4 π(αx)1/4 3 4
(3.23)

Now we can write,

 
2D 2 3/2 π
ψW KB ≈√ sin (αx) − θ −
~(α3 x)1/4 3 4
2 3/2 2 3/2
 
D i (αx) −i (αx)
ψW KB = √ e 3 e−iπ/4 e−iθ − e 3 eiπ/4 eiθ 
3
i ~(α x) 1/4

(3.24)

9
Similarly we can write,
   
a 2 3/2 π b 2 3/2 π
ψp ≈ √ sin (αx) + +√ cos (αx) +
π(αx)1/4 3 4 π(αx)1/4 3 4
2 2
 
a i (αx)3/2 −i (αx)3/2
ψp = √ 1/4
e 3 e−iπ/4
−e 3 eiπ/4 
2i π(αx)
2 3/2 2 3/2
 
b i (αx) −i (αx)
+ √ 1/4
e 3 e −iπ/4
+ e 3 eiπ/4 
π(αx)
(3.25)

Now equating Eq 3.24 and Eq 3.25 we get,

r
π −iπ/4 −iθ
−2D e e = (a + ib)eiπ/4
α~
r
π −iθ
⇒ (a + ib) = 2D ie (3.26)
α~
and,
r
π iπ/4 iθ
2D e e = (−a + ib)e−iπ/4
α~
r
π iθ
⇒ (a − ib) = −2D ie (3.27)
α~
Now adding and subtracting Eq 3.26 and Eq 3.27,Finally we get,
r
π
a = 2D sin θ (3.28)
α~
r
π
b = 2D cos θ (3.29)
α~
Now combining all the results from Eq 3.21, Eq 3.22, Eq 3.28 and Eq 3.29
we get

G = D sin θ (3.30)
F = 2D cos θ (3.31)

So finally after such a tedious calculation we stand here with the solution for

10
the double well potential problem as
1 x
 
D
Z
′ ′
ψ(x) = p exp − |p(x )|dx ; f or x > x2
p(x) ~ x2
 Z x2 
2D 1 ′ ′ π
= p sin p(x )dx + ; f or x1 < x < x2
p(x) ~ x 4
 Z x
1 x
   
D 1
Z
′ ′ ′ ′
= p 2 cos θ exp |p(x )|dx + sin θ exp − |p(x )|dx ;
p(x) ~ x2 ~ x2
f or x < x1
where x1
1
Z
θ≡ p(x)dx
~ x2

3.3.2 Eigenvalues (Energy levels)

Since V(x) is symmetric the eigen-functions should be symmetric. Hence,
wave-functions would be either even or odd. Putting the condition ψ(0) = 0
for odd, we get
Rx
Rx

2 cos θ exp 1/~ 0 1 |p(x′ )|dx′ + sin θ exp −1/~ 0 1 |p(x′ )|dx′ = 0
=⇒ sin θe−φ/2 = −2 cos θeφ/2 =⇒ tan θ = −2eφ (3.32)
and ψ ′ (0) = 0 for even, we get
−D dp(x)
|0 [2 cos θeφ/2 + sin θe−φ/2 ] +
2|p(x)|3/2 dx
  Z x1 
D ′ ′
p 2 cos θ exp 1/~ |p(x )|dx (−|p(0)|/~) +
p(x) 0
 Z x1  
′ ′
sin θ exp −1/~ |p(x )|dx (|p(0)|/~) = 0
0
=⇒ sin θe −φ/2
= 2 cos θeφ/2 =⇒ tan θ = 2eφ (3.33)
where we have defined
x1
1
Z
φ≡ |p(x′ )|dx′ (3.34)
~ −x1
and used
p
dp d 2m(V (x) − E) √ 1 dV
= = 2m √ (3.35)
dx dx 2 V − E dx
dV dp
∵ |0 = 0 ∴ |0 = 0 (3.36)
dx dx
11
Assuming central barrier is quite high which makes φ large and therefore
exp φ huge. Thus,
1
θ ≡ (n + )π + ǫ, where |ǫ| ≪ 1 (3.37)
2

 
1
∴ tan θ = tan (n + )π + ǫ
2
 
1
sin (n + )π + ǫ
2
=  
1
cos (n + )π + ǫ
2
n
(−1) cos ǫ
=
(−1)n+1 sin ǫ
− cos ǫ
=
sin ǫ
1
≃ − (3.38)
ǫ
As a result,
1
− = ±2eφ (3.39)
ǫ
1 1
=⇒ θ = (n + )π ∓ e−φ (3.40)
2 2
To get the eigen-values, let’s assume our double well potential as

mω 2 (x + a)2
V (x) = ,x < 0 (3.41)
2
mω 2 (x − a)2
,x > 0 (3.42)
2

12
 1 1
We have for E = mω 2 (x22 − a)2 = mω 2 (x21 − a)2
2 2
s
1 x2
 
1
Z
θ = 2m E − mω 2 (x − a)2 dx
~ x1 2
s
1 z2
 
1
Z
= 2m E − mω 2 z 2 dz, z ≡ x − a
~ −z2 2
s
2 z2
 
1
Z
= 2m E − mω 2 z 2 dz
~ 0 2
Z z2 q
2mω
= z22 − z 2 dz
~ 0
mω
q
= [z z22 − z 2 + z22 sin−1 (z/z2 )]|z02
~
mω 2 −1
= [z sin (1)]
~ 2
mπω 2E πE
= 2
= (3.43)
2~ mω ~ω
Thus, we get
πE 1 1
= (n + )π + e−φ (3.44)
~ω 2 2
And finally we get the energy levels as follows,
 
± ∼ 1 ~ω −φ
f or φ ≫ 1, En = n + ∓ e (3.45)
2 2π
 
1
Now consider the case(φ → ∞) ⇒ En± ∼ = n+ which means we get back
2
two detached SHO with degenerate energy states. This is what we got while
working with perturbation method. One important point to note is that, the
even states ( ψn+ ) have slightly lesser energy than the odd (ψn− ) ones.[2]
We now consider the case of a particle being dropped in the double well.
The energy of the particle is E<V(0). This means particle has not enough
energy to climb over the hill classically.
Suppose the particle starts out in a state of the form
1
Ψ(x, 0) = √ (ψn+ (x) + ψn− (x)) (3.46)
2
If the wave-function of odd state is positive in the right well, then the particle
starts out in the right well. To find the state at a later time t, we just need

13
to attach their respective energies with the individual states:

iEn+ t iEn− t
    
1 + −
Ψ(x, t) = √ ψn (x) exp − + ψn (x) exp −
2 ~ ~
   
1
exp − n + iωt 
iωte−φ iωte−φ
   
2 + −
= √ ψn exp + ψn exp −
2 2π 2π

where equation 3.45 has been used. Since, at the start all other energy modes
are ‘dead’, throughout its evolution, the particle would be linear combination
of just these two energy modes. As a result, the particle keeps oscillating
back and forth between the wells. For instance, at half the time-period τ /2,
it is concentrated in the left well as state has become
 
1
exp − n + iωt
2
Ψ(x, τ /2) = √ (ψn+ − ψn− ) (3.47)
2
This becomes clearer with probability density –

iωte−φ iωte−φ
     
2 1 + 2 − 2 + −
|Ψ(x, t)| = (ψn ) + (ψn ) + ψn ψn exp + exp −
2 π π
−φ
 
1 iωte
= (ψn+ )2 + (ψn− )2 + 2ψn+ ψn− cos (3.48)
2 π

2π 2 φ
Thus, the time-period of oscillation is τ = e .
ω
Higher the barrier(represented by φ), longer it takes for the particle to
oscillate back and forth.
Having calculated the wave function and energy eigenvalues for the double
well potential we now look for a more elegant way to reach at the same result.
We will be discussing Instanton method next which provides us exactly what
we need.

14
Chapter 4

Istanton Method

4.1 The Path Integral Formulation

Before we get into looking at another method to solve the double well po-
tential problem, we try to understand Feynman’s path integral formalism as
briefly as possible. Suppose a particle starts from (x′ ,t′ ) in the space-time
and makes a journey to its final position(x,t) via some path. Then Feynman’s
recipe to find the propagator U(x,t;x′ ,t′ ) summarised in three simple steps is
as follows

• All paths in the x-t plane connecting (x,t) and (x′ ,t′ ) are drawn.

• For each path x(t), the action S[x(t)] is computed.

• U(x, t; x′ , t′ )=A all paths exp iS[x(t)]/~

P

where A is an overall normalization constant.

Figure 4.1: Paths connecting the initial position-time with final position

15
 Now we consider the propagator in some details. We can write,

U(x, t; x′ , t′ ) = hx|U(t − t′ )|x′ i

X
= A exp iS[x(t)]/~
all paths
= A[exp(iS1 /~) + exp(iS2 /~) + exp(iS3 /~) + ...] (4.1)

where S1 , S2 , S3 are actions corresponding to different paths.

Transition probability from one point to another would be given by prop-
agator’s amplitude square. This would give rise to all sorts of cross terms
(between the actions of different paths) that involve nearby as well as far
away terms.
Now in the classical limit i.e, ~ → 0, we can see that the exponential term
of the propagator becomes highly oscillatory. When we take a sum over all
the possible paths, only the terms near the stationary path would survive.
The stationary path is the classical action S which remains ’stationary’ even
under such highly oscillatory conditions. Having said that we now see only
the paths near the stationary path survive so we don’t have to worry about all
the possible cross terms between all the possible paths. To find U apart from
the above justification we have also used The superposition principle and the
composition property of translation operator from quantum mechanics.

It is important to note that classical path is important not only because

it contributes the highest but also because the paths close to it contibute
coherently.
Thus believing the above arguments to be a good approximation for semi-
classical limit, we may ignore all but the classical path and its neighbors in
the calculation of U(t). Thus,

U(x, t; x′ , t′ ) ≃ A′ eiSCl /~ (4.2)

where A′ “measures” the number of paths in the range where they add con-
structively.
Now the trouble is that for our problem of double well potential we don’t
have a classical path corresponding to the tunneling effect. The solution to
this problem is provided by the imaginary time formalism to be dealt in the
next section.

16
4.2 Imaginary Time Formalism
In order to see tunneling in the path integral
 formalism
 we transform t →
iHt
t′ = −iτ in the usual propagator U(t)=exp − , to get the imaginary
~
time propagator  
1
U(τ ) = exp − Hτ (4.3)
~
In other words, if Shrodinger equation had been
d
−~ |Ψ(t)i = H|Ψ(t)i (4.4)
dτ
the propagator would have been given by Eq 4.3. The propagator is no more
unitary . Thus, we have

lim hx|U(τ )|x′ i = lim hx|nihn|x′ ie−En τ

τ →∞ τ →∞
≃ ψ0 (x)ψ0∗ (x)e−E0 τ (4.5)

Now, we can find the ground state of the system by evolving the propagator
for sufficiently long time. But the importance of the transformation is the
following,
Let’s see how the action S gets changed.
"   #
2
1 dx
Z Z
iS = i dtL = dt − V (x) (4.6)
2 dt
"  2 #
1 dx
Z
= dτ − − V (x) (4.7)
2 dτ
"   #
2
1 dx
Z
= − dτ + V (x) (4.8)
2 dτ
Z
= − LE (x, ẋ)dτ (4.9)

where "   #
2
1 dx
LE = + V (x) (4.10)
2 dτ
Thus, now S would be minimized by path given by

d2 x
= +V ′ (x) (4.11)
dτ 2
17
This means we have to just invert the potential and solve the problem. But
we must be careful while choosing the paths. We need to consider only those
paths that has finite action.

−→

Credit:R Shankar

We can see that now for the inverted potential we can have classically allowed
paths unlike the case earlier for the real potential where there was no classical
path corresponding to the tunneling.

4.3 Instanton Solution

We will now use the imaginary time technique to solve the problem of double
well. The particle can be found on either of the maximas.We can see that
the ground state is degenerate. Moreover in the semiclassical regime we
can approximate the distribution of position to be gaussian. Since, we are
interested only in ground and first excited state, we can truncate the Hilbert
space to be left with only these degenerate states. In the basis of | ± ai,
Hamiltonian(H) matrix can be written as
 
E0 0
H=
0 E0

But we know that h−a|H|ai 6= 0 because of tunneling.So we need to reset the

reference energy so that diagonal elements E0 = 0. Regarding off-diagonal
elements, h−a|H|ai = ha|H| − ai = E as the probability for the particle to
go from left-well to the right -well is same as the other way round due to
symmetry and therefore,
 
0 E
H= (4.12)
E 0

18
 Now we need to find its eigen-values and eigen-states. We can see that H
matrix is simply a constant E multiplied by Pauli sigma x(σx ) matrix whose
eigen-values are ±E and eigenstates are
1
|Si = √ (|ai + |-ai) (4.13)
2
1
|Ai = √ (|ai − |-ai) (4.14)
2
We need to find E using the propagator. For that, let’s expand it

ha|U(τ )| − ai = ha| exp(−Hτ )| − ai

1
≃ 0 − τ ha|H| − ai + Oτ 2
~
we just consider the linear term in τ to find E. Now from previous discussion
we know,
 
1
ha| exp(−Hτ )| − ai = exp − SCl (4.15)
~
To find the action, we need to find the classical path followed by the particle
in inverted potential. We can see that energy is constant of the motion and
can be used to determine the system’s dynamics. Thus,

d2 2
ẍ = λ2 (x − a2 )2 (4.16)
dx2

 2
m dx
Ee = − V (x) = 0 (4.17)
2 dτ
For ground state, we take E=0. This means it would take infinite time to
reach from one well to the other. We need not integrate the differential
equation to get action as shown below
Z τ0 /2
SCl = lim (T + V )dτ
τ →∞ −τ0 /2
Z τ0 /2
= 2T dτ
−τ /2
Z a 0p
= 2mV (x)dx (4.18)
−a

19
[5] Eq 4.18 can be obtained from Eq 4.17. Once we have action S, propagator
is

1 ap
 Z 
ha|U(τ )| − ai ≃ exp − 2mV (x)dx (4.19)
~ −a

This is in agreement with WKB. But there is no linear term in τ . That

was the term which would have given us the off-diagonal element. So what’s
wrong here?
Ah!We were forgetting A′ in U(t) ≃ A′ eiSCl /~ . We have to take into
account other paths as well that that have the same action and that contibute.
Now lets solve Eq 4.16, we get
"r #
2
x(τ ) = a tanh λτ (4.20)
m

We find that particle spends most of its time at ±a, and jumps ’instantly’

Figure 4.2: Solution

x ^

< >
τ

-a
>

from -a to +a in a short time centered around τ = 0. This is the reason,

solution is called instanton. There are many terms in the final solution but
they have the same action. They only differ in tunneling instants τ0 . We
sum over all these paths, i.e, integrate over the instant of tunneling τ0 . Since
−τ /2 < τ0 < τ /2, thus

ha|U(τ )| − ai
1 ap
 Z 
≃ τ × exp − 2mV (x)dx
~ −a

20
 Comparing the above with
1
ha|U(τ )| − ai ≃ 0 − τ ha|H| − ai + Oτ 2 (4.21)
~
we get  a 
1
Z p
E = − exp − 2mV (x)dx (4.22)
~ −a

Thus, we conclude

1 ap
 
1
Z
|Si = √ (|ai + |-ai) E = − exp − 2mV (x)dx (4.23)
2 ~ −a
1 ap
 
1
Z
|Ai = √ (|ai − |-ai) E = exp − 2mV (x)dx
2 ~ −a

The symmetric solution has lower energy than anti-symmetric. This is be-
cause more curvature in the wave function or more change of sign implies
higher energy.
Moreover, we have totally ignored the pre-factors in our calculation of
the propagator. Including that, we would have improved the precision of our
statements but would have missed the physical aspect of the solution.

21
Chapter 5

Phenomenons

5.1 Ammonia Inversion

One of the applications of the double well potential involves the working
of Ammonia MASER. The mechanism of oscillating wave function as dis-
cussed in section is a very good explanation for the inversion of ammonia
molecule.Hydrogens in ammonia form a plane and at a given time the ni-
trogen can be above the plane or below it. The time period of oscillation
between these two states can be derived from the formalism discussed earlier
in section . The oscillation is dependent on the energy gap between the two
states corresponding to the position of nitrogen above the plane or below the
plane of hydrogen. The working of MASER is similar to that of the LASER
with the only difference being the frequency or energy of EM wave required
to stimulate emission in both the cases. Discussion about the working of
MASER is beyond the scope of this report but we would like to point out
that some important feature of double well potential can be used to study
various interesting phenomenon like stimulated emission.

5.2 Macroscopic Quantum Coherence

So far in our discussion we have assumed that the system under consideration
is detached from the environment, which is of course far from reality. So ac-
tually what happens when we have interaction with the environment? As we
know that the very process of observation forces my system to collapse into
one of the eigenstates, similar arguments can be extended for interactions as
observation is also a form of interaction. Thus interaction forces the system to
choose one of the states which leads to the breakdown of quantum coherence.

22
 Credit: nature.com

With the help of the figure above we have tried to show the fate of a
dissipative system. Instead of going from figure a to d, we would urge you to
do the reverse.In fig d we have a double well potential which has no interac-
tion with the environment. The wave function keeps on oscillating from one
well to another as discussed earlier and still retains its wave front.The sys-
tem is said to be exhibiting MQC(Macroscopic Quantum Coherence). Now
as we start to introduce interaction with the environment we see that the
wave front looses its form but still there is some tunneling. Now the system
is exhibiting MQT(Macroscopic Quantum Tunneling) which is shown in fig
(c). Finally we can see in fig (b) that increasing the dissipative forces to very
high values can cause the system to stick to only one potential.

Credit:nature.com

The oscillation for the MQC case is illustrated in the figure above. We can

23
see that the wave function retains its wavefront even after tunneling through
the potential barrier.

5.3 More Applications

Double well is a widely used model in various fields. Some of the application
are

• Arrhenius Law: It says that chemical reaction constant k is propor-

tional to e−Eact /kB T , where Eact is the activation energy. Modeling a
chemical reaction as a process in which reactant goes from less stable
position to more stable in asymmetrical double well potential, we can
“derive” this result using diffusion equation. [1]

• Quantum information processing: It has been proposed that the system

of particles in a double well potential can be used as quantum logic gates
for ultracold atoms confined in optical lattices which are fundamental
building blocks. In these proposals, a regular array of cold atoms in an
optical lattice serves as the quantum register. [?]

• Band formation in crystalline solids due to uplifting of degeneracy.

24
Conclusion

In this report we have tried to solve the double well potential problem using
WKB approximation as well as Instanton method. It was quite evident that
the former required mathematical calculations to a great extent where as
the later deals with the problem in an elegant way. But caution must be
exercised in applying the imaginary time formalism as in general we may
not be able to get away with intuitively straight forward solution. We then
discussed few phenomenons with the help of the formalism developed earlier.
During the discussion about MQC we understood how interaction with the
environment can determine the fate of the particle. And finally we concluded
with a small discussion about further application of double well potential in
various other areas.

25
Acknowledgement

We would like to thank our instructor Dr.Siddharth Lal for organizing such
a wonderful term paper session where we were able to learn about different
areas in Quantum Mechanics.
I would also like to thank my partner, Mohit Pandey, for our discussion
sessions and for helping me to relate the mathematical and the physical
aspects of the problem
I would also like to thank all my friends who have helped me in completing
my term paper successfully.

26
Bibliography

[1] W Bialek. Biophysics: Searching for principles. 2011.

[2] David Jeffery Griffiths. Introduction To Quantum Mechanics, 2/E. Pearson Education India, 2005.

[3] Riccardo Rattazzi. The path integral approach to quantum mechanics lecture notes for quantum
mechanics iv. 2009.

[4] J J Sakurai. Modern Quantum Mechanics. Pearson, revised edition, 1994.

[5] R Shankar. Principles of quantum mechanics. Springer, second edition, 1994.

[6] A.I Vainshtein V.I Zakharov V.A Novikov, M.A Shifman. Chapter iii, the abc of instantons.

27