5 Solving Schrodinger’s Equation in One Dimension

1. Write down Schrodinger’s Eqn

2. ???
3. PROFIT

Except for the original Star WarsTM Trilogy, the middle portion of all Trilogies is always boring9 . And
so, in our lectures on Quantum Mechanics, we arrive at the sagging middle.
In this section, we solve Schrodinger’s Equation for a wide variety of potentials U (x) in one dimension.
So unfortunately we will spend some time mangling with partial differential equations, which may or may
not be your favorite cup of caffeinated beverage. Having said that, solving this equation will illustrate
some of the features of Quantum Mechanics of which we have been making assertions about so far, and
some which you may have heard about.

• Section 5.1 : Quantization of Energy States

• Section 5.2 : Scattering. Transmissivity and Reflectivity

• Section 5.3 : Tunneling

• Section 5.4 : The Gaussian Wave Packet and Minimum Uncertainty

• Section 5.5 : Parity Operator

• Section 5.6 : Bound and Unbounded States

Note : in this section we will almost exclusively be working in the Stationary States basis, i.e. χE of
the Hamiltonian, so we will drop the subscript E from χ and ψ. When there is an ambiguity, we will
restore it. Also, sometimes we refer to χ(x) as the “wavefunction”, although technically we are really
solving for the eigenfunctions of the Hamiltonian.

5.1 Quantization of Energy Eigenstates : The Infinite Potential Well

Consider the infinite potential well (Fig 10)


 0 , 0<x<a
U (x) =


∞ , otherwise

Using Stationary States Eq. (164) in one dimension

� �
−iEt
ψ(x, t) = χ(x) exp , (172)
�
we obtain the time-independent Schrodinger’s Equation
�2 d 2 χ
−+ U (x)χ = Eχ. (173)
2m dx2
We now look at the behavior of χ(x) inside and outside the well.

• Outside Well:
U (x) = ∞ ⇒ χ(x) = 0 (174)
otherwise E = ∞ from (173). Thus, as in classical physics, there is zero probability of finding the
particle outside the well.
9 The prequels of Star Wars is the exception: they are all terrible.

35
 U

n=1
n=2
n=3

0 a x
Copyright © 2008 University of Cambridge. Not to be quoted or reproduced without permission.

Figure
Figure16:
10: Lowest energy
A few lowest wavefunctions
energy eigenfunctionsofofthe
the infinite squarewell.
infinite square well.
U

• Inside Well:
�2 d 2 χ
U (x) = 0 ⇒ − = Eχ (175)
2m dx2
It is a bother to carry around all the constants in Eq. (175), so we define
�
2mE
k= > 0, (176)
0 �2a x

with a note to emphasise the fact that k depends on the eigenvalue E, and we get
Figure 17: Wavefunction at large n.
d2 χ
= −k 2 χ. (177)
dx2
• The corresponding energy levels are,
Eq. (177) has the general oscillatory solution
�2 k 2 �2 π 2 n2
Eχ(x)
= E=n A sin(kx)
= =
+ B cos(kx) 2 (178)
2m 2ma
withn two
for = 1,arbitrary
2, . . . complex constants A and B which depends on the boundary conditions (since
Eq. (177) is second order in x).
– Like
To find A andBohr
B we atom,
match energy
solutionslevels
at the are quantized
boundaries x = 0 , x = a by imposing continuity:

– Lowest energy level or groundstate,

χ(0) = χ(x) = 0 (179)

so χ(0) = 0 ⇒ B = 0 and χ(a) = 0 ⇒ A sin(ka) E= 0 with �π

ka = nπ,
2 2
1 = 2
> which
0 implies a set of solutions
 � nπx � 2ma

 An sin a , 0<x<a
• The resulting wavefunctions
χn (x) =are illustrated in Figure 16


0 , otherwise
– Wave functions alternate between even and odd under reflection,
Since the eigenfunctions χn are discrete, it turns out that they are normalizable. Applying the normalizing
condition � +∞ � ax → �
a�
2 nπx
− x |A |2 a
2
|χn | dx = |An | 2
sin 2 dx = n = 1, (180)
−∞ 0 a 2
so we get– Wavefunction χn (x) has n + 1 zeros or nodes where ρn (x) = |χn |2 vanishes.
2
|An | = , ∀ n. (181)
a
– n → ∞ limit. Correspondence principle: probability density approaches a con-
Note that while An can be a complex number it will not matter since it is a irrelevant phase in this case.
stant ≡
Hence the energy classical result
eigenfunctions of the(see Figure
infinite square17).
well is a discrete series of sin functions labeled by
n (Fig. 10). An eigenfunction labeled by n has n − 1 zero crossings, called nodes.
Some properties:
26
36
 • Quantization of Energy Levels: Since n is a discrete spectrum, this means that a quantum
particle inside a infinite square well can only take specific, quantized, values of E, unlike in Classical
Mechanics. The energy levels are found by using Eq. (176)

�2 k 2 �2 π 2 n2
E = En = = , for n = 1, 2, . . . (182)
2m 2ma2
The difference between two energy levels Em and En is

�2 π 2
∆E = Em − En = (m2 − n2 ). (183)
2ma2

• Ground State has non-zero Energy: Also, unlike Classical mechanics, the lowest energy state
is non-zero
�2 π 2
E1 = > 0. (184)
2ma2
The lowest energy state of any system is called its Ground State or Vacuum State.

• Odd/Even Solutions and Parity Operator If we draw a vertical line through x = a/2 and
imagine it to be a mirror, the eigenfunctions alternate signs under reflection on this mirror in the
following way
Even : χn (x) = χn (a/2 − x) , n = 1, 3, 5, . . . (185)

while
Odd : χn (x) = −χn (a/2 − x) , n = 2, 4, 6, . . . (186)

i.e. the eigenfunctions χn naturally fall into Even and Odd sets. We hasten to apologize for the
unfortunate fact that even n are odd eigenfunctions and vice versa.
We can be precise about what we mean by reflection by defining the Parity Operator P̂ whose
action is to change the sign of the argument of a state around an axis. In the above example, the
axis is x = a/2, so the P̂
P̂ ψ(x) = ψ(a/x − x). (187)

It is clear that the eigenfunctions χn are also (simultaneous) eigenfunctions of P̂

P̂ χn = (−1)n+1 χn , (188)

with eigenvalues +1 for even eigenfunctions and −1 for odd eigenfunctions.

We will discuss the Parity Operator in section 5.5, but can you see why the eigenfunctions are also
Parity eigenfunctions?

5.2 Scattering : Transmissivity and Reflectivity

Consider the Step Potential (Fig 11)


 0 , x<0 Region I
U (x) =


U0 , x>0 Region II

Consider the behavior of a particle sent from the left of the plot with energy E. Classically, we know
the answer simply: if E > U0 then the particle goes over the barrier and if E < U0 the particle is reflected
back. What happens quantum mechanically? To find out, we have to solve the Schrodinger’s Equation.

37
 U0

Region I Region II

x
x�0

Figure 11: The step potential.

Again, using Stationary states, the time-independent Schrodinger’s Equation is

�2 d 2 χ
− + U (x)χ = Eχ. (189)
2m dx2

Like the previous problem, we consider the solutions independently in Regions I and II. There are
two cases, when E > U0 and when E < U0 .

• Case 1: E > U0 : Region I : U = 0, Eq. (173) becomes

−�2 d2 χ
= Eχ, (190)
2m dx2
�
and defining as usual k = 2mE/�2 > 0, we get

d2 χ
= −k 2 χ (191)
dx2
which we can write the solution Eq. (177) as before when we consider the infinite potential well.
However, let’s write it in a more physically motivated way as follows

−ikx
eikx + Ae
χ(x) = ���� � �� �, (192)
incoming reflected

where A quantifies the amplitude of a reflected wave10 . In other words, we set up the incoming right
moving wave exp(ikx) (secretly we have imposed boundary conditions), and we want to calculate
how much of the wave is reflected back as left moving exp(−ikx). They are called right/left moving
as they are eigenfunctions of the momentum operator p̂ with ±�k eigenvalues respectively.
You can think of this as a probability current heading in the +x direction
� �
−i� dχ dχ†
jI = χ† − χ
2m dx dx
�k
= ( 1 − |A|2 ) (193)
m ���� ����
jinc jref

where jinc is the current carried by the right moving wave and jref the left moving wave. If |A| = 1
then the total current is zero, i.e everything is reflected back.
Region II : U = U0 , Eq. (173) becomes

−�2 d2 χ
= (E − U0 )χ, (194)
2m dx2
10 Some books call this the reflectivity, and gave it a name R.

38
since E > U0 , let’s define �
2m(E − U0 )
q= > 0, (195)
�2
so we get
d2 χ
= −q 2 χ, (196)
dx2
which is also oscillatory. Since there is no left-moving wave (again secretly imposing boundary
conditions), the general solution is
iqx
χ(x) = Be� �� � , (197)
transmitted

where B quantifies the amplitude of a transmitted wave. The current can be calculated as usual,
and it is
�q
jII = jtrans |B|2 . (198)
m
We want to now solve A, B as functions of k, q by matching solutions at x = a. Since U (x) is
discontinuous but finite, this can get a bit tricky. We will use the following result:
Continuity of χ at Discontinuous potential: Suppose U (x) is discontinuous but finite at x = a,
then χ(a) and dχ/dx|x=a are continuous, but d2 χ/dx2 |x=a is discontinuous.
Proof : From the Schrodinger’s Equation, since U (a) is discontinuous, d2 χ/dx2 |x=a is also discon-
tinous. Now integrate the time-independent Schrodinger’s Equation over the interval [a − �, a + �],
� a+� � a+�
−�2 d2 χ
dx = dx (E − U (x))
a−� 2m dx2 a−�
� � �
dχ �� dχ �� 2m a+�
− = − dx (E − U (x)). (199)
dx �a+� dx �a−� �2 a−�

Taking the limit of � −→ 0, the RHS of Eq. (199) vanishes, so this implies
� � � �
dχ �� dχ ��
lim = , Continuity of first derivative (200)
�→0 dx � dx �a−�
a+�

which also implies that χ(a) is continuous. �.

Using our result above, we can then continuity conditions at the boundary x = 0 to get

χ(x = 0) ⇒ 1 + A = B, (201)

and
� �
lim ik(e−ik� − Aeik� ) = iqBeik� ⇒ ik(1 − A) = iqB. (202)
�→0

Using the two equations Eq. (201) and Eq. (202), we can solve for R and T
k−q 2k
A= , B= . (203)
k+q k+q
Now if E � U0 , q → k and hence A → 0 and B → 1, i.e. if the incoming wave is very energetic,
everything is transmitted and nothing is reflected as we expect classically.
Although A and B are real quantities here, this is not always the case. And don’t fall into the temp-
tation of comparing the absolute amplitudes of A and B as the wavefunctions are not normalizable!
The right way to think about this is to compare probability currents,
� �2
Reflectivity , R = jjinc
ref k−q
= k+k (204)
� �2
Transmissivity , T = jjtrans inc
4kq
= k+k (205)

39
 So as E � U0 , then q → k, and R = 0 and T = 1 as expected. Note that even when E > U0 , there
is a non-zero chance of particles being reflected, unlike the classical case.
We can check for the conservation of probability, using equation Eq. (157), which in one dimension
is
∂ρ ∂j
+ = 0. (206)
∂t ∂x
For Stationary States, ρ is independent of time, so then this becomes
∂j
= 0 ⇒ jI = jII , (207)
∂x
and using the results we have
�k �q 2
(1 − |R|2 ) = |T | . (208)
m m
• Case 2: E < U0 : The results in Region I is as before, but for Region II, we define
�
2m(U0 − E)
κ= > 0, (209)
�2
so Schrodinger’s Equation becomes
d2 χ
= κ2 χ. (210)
dx2
This equation has the solution
χ(x) = Ceκx + De−κx . (211)
The growing mode C is non-normalizable, so we set C = 0 hence the final solution is

χ(x) = De−κx (212)

i.e. the wavefunction decays as it penetrates the barrier. Note that we can simply use our previous
solution, and substitute q → iκ, to find the coefficients
k − iκ 2k
A= , D= . (213)
k + iκ k + iκ
The current in Region II vanishes
� �
−i� dχ dχ†
jII = jtrans = χ† − χ = 0, (214)
2m dx dx
meaning that no particle is transmitted. What about the reflectivity? Since |A|2 = 1,

jref = jinc , (215)

the reflectivity is unity.

*Now you may feel a bit uncomfortable – the wavefunction inside the barrier, even though expo-
nentially small, does not vanish. Does this mean that, via Born’s Rule, we should have a small but
non-vanishing probability of finding the particle? How do reconcile the fact that the probability
current is zero yet there is non-zero wavefunction inside the barrier? To resolve this paradox will
take us too far afield, but we can do a “word calculus” version of it.
If we follow Born’s Rule logic, then there is a finite probability of finding a particle with negative
kinetic energy – corresponding to the fact that the “momentum” of the particle is imaginary.
But nobody has seen a negative kinetic energy particle before, so this must not be the answer11 .
Physically, to observe such a particle, we need to shine a light on it and then collect the scattered
light to deduce the location of the particle. To fix the exact location of the particle, we need
the wavelength λ of the light to be much smaller than the characteristic penetration depth of the
wavefunction, but this means collision of the light particle with the particle will give it sufficient
energy, 2π�/λ � (U − E) to kick it out of the barrier!*
11 Physicists give them a name, ghosts, because nobody has seen them before.

40
 U0

Region I Region II Region III

x
x�0 x�a

Figure 12: The barrier potential.

5.3 The Barrier Potential : Tunneling

In the previous step potential, we see that even if E < U0 , the wavefunction penetrates into the barrier,
and decay exponentially as long as the barrier remains in place. Now, what happens, if after some distance
a, the potential drops again to zero Fig. 12 ? The wavefunction decays exponentially until it hits x = a,
and then suddenly it is no longer suppressed by the potential and is free to propogate. Physically, this
means that there is a no-zero probability of finding a particle on the right side of the barrier – we say
that the particle has tunneled through the barrier, and this phenomenon is knonw as Tunneling. We
will consider the case when E < U0 . As before, we define the variables
� �
2mE 2m(U0 − E)
k= > 0 , κ = > 0, (216)
�2 �2
and as should be familiar to you, we write down the following ansatz


 exp(ikx) + A exp(−ikx) x<0 Region I
χ(x) = B exp(−κx) + C exp(κx) 0 < x < a Region II


D exp(ikx) x>a Region III
At boundary x = 0, continuity conditions imply

1 + A = B + C , ik(1 − A) = κ(−B + C), (217)

whilat at x = a, we get

B exp(−κa) + C exp(κa) = D exp(ika) , κ(−B exp(−κ)a + C exp(κ)) = ikD exp(ika). (218)

We then do a bunch of tedious algebra to solve for A, B, C and D. Since we are interested in the
transmitted current, we look for D which has the following horrible form
2kκe−2ika
D= . (219)
i(k 2 − κ2 ) sinh(2κa) + 2kκ cosh(2κa)
The incident and transmitted flux are then
�k �k
jinc = , jIII = jtrans = |D|2 (220)
m m
hence the transmissivity is
jtrans
T = = |D|2
jinc
4k 2 κ2
= > 0, (221)
(k 2 + κ2 )2 sinh2 (κa) + 4k 2 κ2

41
i.e. the transmissivity is positive and non-zero. There is a chance that you will find a particle of
momentum p = �k on the right side of the barrier x > a.

5.4 The Gaussian Wavepacket

In the previous problems, you may have feel a bit uncomfortable that we are using momentum eigenstates

up = Aeikx (222)

as “particle states”, even though we have argued forcefully in the section 3 that such states are not
normalizable. Your discomfort is well-founded – indeed in realistic situations, say when we want to do
an experiment by sending a particle into a barrier and check whether it tunnels or not, we set up the
particle whose position we “roughly” know, say at x0 . A good model of such a set-up is to specify the
probability density of the particle to be a Gaussian at some fixed time t0 = 0, i.e.
� �
−(x − x0 )2
ρ(x, t0 ) ∝ exp , (223)
2σ̃ 2
where σ̃ 2 is the dispersion12 of the particle, i.e. it measures how “spreaded out” the particle is. Those
who have studied statistics might recall that the above probability density means that the particle can
be found within x0 ± σ̃ is 66%. Of course, the Gaussian is square integrable, so this wavefunction is
normalizable. The question we want to ask now is: if we set this system up and let it evolves freely, i.e.
U (x) = 0, what happens to it?
As before, we want to work in the Stationary State basis. But since we are interested in time evolution,
we will keep the time dependence. From Eq. (164), and using E = p2 /2m = �2 k 2 /2m for a free particles
and χ(x) = exp(ikx) this becomes
� �
−i�k 2 t
ψk (x, t) = exp(ikx) exp ,. (224)
2m
Since the spectrum for free particles is continuous, we can construct any arbitrary real space wavefunction
by an integral Eq. (169),
� ∞ � �
−i�k 2 t
ψ(x, t) = dk C(k) exp(ikx) exp (225)
−∞ 2m
where C(k) is some smooth function. What C(k) should we choose such that we obtain the probability
density Eq. (223) which possess some average momentum p0 = �k0 ? Now we cheat a little, and assert
that this corresponds to the choice
� σ �
C(k) = exp − (k − k0 )2 , (226)
2
which (not surprisigly) is also a Gaussian in k-space with a dispersion of 2σ −1/2 . Now we want to evalute
the integral Eq. (225). Collecting the terms proportional to k 2 and k in the exponential,
� ∞ � �
σ i�t
ψ(x, t) = dk exp[− + k 2 + (σk0 + ix) k −σk02 /2]. (227)
−∞ 2 2m � �� � � �� �
� �� � β δ
α/2

Copmleting the square we get

� �� �2 �
∞
1 β β2
ψ(x, t) = dk exp − α k − + +δ (228)
−∞ 2 α 2α
� 2 �� ∞ � � �2 �
β 1 β
= exp +δ dk exp − α k − . (229)
2α −∞ 2 α
12 In statistics, this is known as the variance but we will follow standard jargon.

42
 2� Σ

k0

Figure 13: The Gaussian Wavepacket in k-space with mean momentum �p� = �k0 .

�
The integral is a usual Gaussian Integral13 , and gives 2π/σ so we finally get the wavefunction
� � 2 �
2π β
ψ(x, t) = exp +δ . (230)
σ 2α
√
This wavefunction is not normalized, so let’s normalized is by setting ψ̃ = ψ/ N , where
� ∞
N = dx |ψ|2
−∞
� ∞ � � �2 �
σ k0 �t
= dx exp − 2 x −
−∞ |α| m
�
π
= 2π (231)
σ

where again we have used the Gaussian integral.

The properly normalized probability density function is then
� � �
2 1 σ −σ(x − x̄(t))2
|ψ̃(x, t)| = exp (232)
π σ 2 + x̄2 /k02 σ 2 + x̄2 /k02

where the mean position is moving with a constant velocity

�k0 p0
x̄(t) = t = t. (233)
m m
This is consistent with our classical intuition that the particle move with velocity ẋ = p0 /m. At t = 0,
we recover the promised Gaussian Wavepacket in position space we wrote down earlier Eq. (223).
However, more interestingly, the effective dispersion
� � �
1 x̄2
σ̃(t) = σ+ > σ̃(0) (234)
2 σk02

is also increasing! In other words, not only is the Gaussian wave is moving, it is also spreading, becoming
more and more delocalized. Since the dispersion measures the uncertainty in the location of the particle,
the particle’s position is becoming more and more uncertain. In fact, it can be shown that the Gaussian
Wavepacket is the state of minimum uncertainty.
�∞ 2 �
13 Gaussian Integral is eax = π/a.
−∞

43
 v�p0 �m

x�0� x�t�

Figure 14: The probability density function ρ(x, t) in x-space, with its mean x̄(t) moving at velocity
p0 /m. The dispersion of the particle increases with time, and hence the particle’s position become less
certain.

Minimum Uncertainty Wavepacket: From our discussion of the momentum eigenfunctions in

section 3 and continuous spectrum, we showed that the probability density of momentum is Eq. (122),
i.e. � �
−σ(p − p0 )2
ρ(p, t) = |C(k)|2 ∝ exp (235)
�2
�
which is a Gaussian with dispersion σp = �2 /2σ. Then the product of the two dispersions yield
� �
�2 � 2 t2 �2
(∆x)2 (∆p)2 = σ̃ 2 σp2 = 1+ 2 2 ≥ (236)
4 m σ 4

which is to say that at t = 0, our knowledge of the wavepacket in both the position and momentum is at
the minimum but is non-zero. You might have seen this relation before, and it is called the Heisenberg
Uncertainty Relation. Since we have secretly used Schrodinger’s Equation when we wrote down the
Stationary States as basis and used Born’s Rule to interprete the results, it is a consequence of quantum
mechanical nature of the particle. We will discuss this in much greater detail in section 7.

5.5 Parity Operator

When we consider the infinite potential in section 5.1, we showed that its energy eigenfunctions naturally
subdivided into odd and even states. We also introduced the Parity Operator whose action is to change
the sign of the argument of a state around an axis. In 1 dimension, this is simple (taking the axis x = 0
for simplicity)
P̂ ψ(x) = ψ(−x). (237)

In higher dimensions, we have to specify which of the spatial arguments we want to flip as they are
different operations. For example in 2 dimensions

2 dimensions : P̂x f (x, y) = f (−x, y) , P̂y f (x, y) = f (x, −y) (238)

are two different operators: P̂x flips around x = 0 while P̂y flips around y = 0.
In 3 dimensions, and an operator which flip all arguments would have the form

P̂ f (x) = f (−x), (239)

44
 Figure 15: Parity Operators:
Alice discovering that the Won-
derland is not what it cut out
to be.

so all the points go through the origin to its diagrammatic opposite position – such an operation is
sometimes called an inversion.
Some properties of the Parity Operator:

• P̂ is Hermitian. (You will be asked to show this in the Example Sheet.)

• P̂ 2 = 1. Proof is trivial : P̂ P̂ f (x) = P̂ f (−x) = f (x) �.

• Eigenfunctions and Eigenvalues of P̂ : Let λ be an eigenvalue of P̂ and φ(x) is its eigenfunction so

P̂ φ(x) = λφ(x) = φ(−x) (240)

and now, applying P̂ on both sides from the left, we get

P̂ 2 φ(x) = λ2 φ(x) = φ(x) (241)

and hence φ± (x) eigenfunctions of P̂ with eigenvalues λ = ±1. We call φ+ parity even and φ−
parity odd solutions.

• Simultaneous eigenfunctions of Ĥ and P̂ : Other than the infinite potential well, what other
potentials U (x) also admit parity eigenfunctions? Consider the time-independent Schrodinger’s
Equation with eigenfunction χ(x)

�2 ∂ 2
Ĥχ(x) = − χ(x) + U (x)χ(x), . (242)
2m ∂x2
If we flip the coordinate x → −x in Eq. (242), ∂ 2 /∂x2 remains invariant, and so if the potential is
invariant under the same reflection U (x) = U (−x), then it’s clear that χ(−x) is also an eigenfunction
of Ĥ and χ(−x) = ±χ(x). We say that U (x) is symmetric under reflection x → −x, and that
χ(x) = ±χ(−x) is symmetric/antisymmetric under x → −x.
In other words, if U (x) obeys the same symmetry14 as the Parity Operator, then eigenfunctions of
the associated Hamiltonian Ĥ are also eigenfunctions of P̂ . Of course, there exist a large degeneracy
in the eigenfunctions of P̂ .
14 In slicker language we will introduce in section 7, we say that P̂ and Ĥ commute.

45
 U0

x = −a x=a x

Figure
Figure 16: The finite potential well. 20: The finite
The Hamiltonian of thissquare well.
potential possess both bound and unbounded
Copyright © 2008 University of Cambridge. Not to be quoted or reproduced without permission.

states.
The finite potential well
5.6 Bound and Unbound States : The Finite Potential Well
Potential,
Consider the finite potential well (Fig 16)
Region I : 
U (x) = 0 −a<x<a

 0 , −a < x < a
Region II U : (x) =  = U0 otherwise (31)

U0 , otherwise
as shown in Figure 20.
Bound and Unbounded States: It is clear from our discussion on the step potential in section
5.2, if the energy of the states E ≤ U0 , then the wavefunction have exponentially decaying solutions in
Stationary
Regions states
I and III. obey,words, they don’t propagate outside Region II and hence are “trapped” inside
In other
the well. We call such states bound�states.
2 2
d χ In general, bound states (like the infinite potential well
−
case) are discrete. If, on the other hand, E> + U (x)χ = Eχ (32)
2 U0 , even in Region I and III the wavefunction propagates
2m dx
and hence is unbounded.
consider eventheparity
Consider boundedboundstates
case E ≤ U0 . We define the constants as usual
� �
2mE 2m(U0 − E)
k= χ(−x)
> 0 , κ == χ(x) 2 > 0. (243)
�2 �
Region
obeying 0≤II :EThe
≤ Schrodinger’s Equation
U0 Define real becomes
constants
� d2 χ �
2mE 2
= −k 2 χ 2m(U0 − E) (244)
k= ≥ 0 dx κ= ≥0 (33)
�2 �2
which has solutions
• Region I The SchrödingerχIIequation becomes,
(x) = A cos(kx) + B sin(kx). (245)

d2 χso there must

This potential is clearly symmetric under x → −x,
2
exist even and odd solutions. It is clear
that A cos(kx) is the even solution and B sin(kx) is =
the −k
odd χ
solution.
dx2
Consider the even solution, so setting B = 0 we have
The general solution takes the form,
χII (x) = A cos(kx). (246)
χ(x) = A cos(kx) + B sin(kx)
Regions I and III : The solution are
✘
even parity condition, F e✘
Region I : χI (x) = Ceκx + ✘ −κx
(247)
−a �
where we drop the F term as it is not normalizable, i.e. dx = −∞ and similarly
χ(−x) = χ(x) ⇒ B = 0 −∞ exp(−κx)
⇒ χ(x) = A cos(kx)
Region III : χIII (x) = De−κx + ✘ ✘.
Ge✘
κx
(248)
31
46
 y
Π�2 Π 3Π�2 2Π 5Π�2

Figure 17: Bounded even solutions of the finite potential well. The thick line is tan y while the dotted
�
lines are (λ − y 2 )/y with increasing λ to the right. The crossings between lines indicates a possible
solution. One can see that as we increase the width a2 of the potential, or the depth U0 , there are more
bounded solutions.

Since χ(x) is even, we know that χ(x) = +χ(−x), so C = D. Imposing continuity at x = a, we get

A cos(ka) = D exp(−κa) , − kA sin(ka) = −κD exp(−κa). (249)

Combining all these equations, we find the transcendental equation

k tan(ka) = κ (250)

which cannot be solved in closed form, but we can plot out the solution to see its features. Define
2mU0 a2
λ= > 0 , y = ka, (251)
�2
then Eq. (250) becomes �
λ − y2
tan y = (252)
y
which we plot in Fig. 17. From this figure, we can see

• As the width a and depth U0 increases, the number of solutions grow.

• Bounded states have discrete energy spectrum as advertised

�2 yn2
En = (253)
2ma2
where yn is a solution to the Eq. (252) with n labeling each crossing, and from fig 17 we see that
� �
1
(n − 1)π < yn < n − π. (254)
2

• Again there exist a non-zero probability density in the classical forbidden Regions I and III.

You will be asked to look for odd solutions in an Example Sheet.

47