Chapter 3

Motion of a pointlike particle with

a single degree of freedom
3.1

Continuous observables

In this chapter, we study basic quantum physics of the simplest mechanical system: translational
motion of a pointlike particle with a single degree of freedom. In classical mechanics, this motion
is described by two canonical variables, position and momentum. Accordingly, in our quantum
treatment, we introduce two observable operators: position x
and momentum p.
In a Hilbert space of nite dimension, we expect (see Ex. 1.76) the set of eigenstates of any
physical observable to form an orthonormal basis. However, in the present case this rule can be
applied only partially. This is because, although the geometric space containing the particle is onedimensional, the associated Hilbert space is of innite dimension: for example, there are innitely
many position eigenstates |x, and all these eigenstates are incompatible1 . Furthermore, position
eigenstates form a continuum: for every real value of x there exists an associated eigenstate |x.
The same is true for the momentum observable.
The continuous nature of these observables implies that most mathematical rules (state and
operator decomposition, normalization, basis conversion, etc.) derived for nite-dimension Hilbert
spaces have to be modied: summation must be replaced by integration. This is our task for this
section. In order to reproduce these rules in the form that closely resembles those for the discrete
case, we need to dene a special normalization convention for continuous observable eigenstates.
Instead of normalizing these states to one, as we would do in the discrete case [cf. Eq. (1.8)], we
write:
x| x = (x x );

p| p = (p p ).

(3.1)
(3.2)

This appears strange at rst. According to Eq. (3.1), the inner product of the position eigenstate
|x with itself is x| x = (0), so this state has innite norm. The same applies to momentum
eigenstates. How is this consistent with the First postulate of quantum mechanics, which says that
all physical states must have norm 1? We answer this question by saying that continuous-observable
eigenstates are unphysical : it is impossible to set a particle at an absolutely precise location or make
it move at an absolutely precise velocity. Therefore, the First postulate does not apply to these
states; they are just a mathematical abstraction2 . This notwithstanding, all physically realistic
1 One may argue that one can also introduce continuous observables in nite-dimensional Hilbert spaces: for
example, the polarization angle of a photon can take on any value from 0 to 180 . However, this would not be an
observable in the quantum mechanical sense, because the relevant eigenstates are not orthogonal: a photon polarized
at 45 is a superposition of the 0 and 90 polarization states. On the other hand, a particle located at x = 4m is not
a superposition of the |x = 3m and |x = 5m position eigenstates. A particles position is a valid observable while a
photons polarization angle is not.
2 To treat this matter more rigorously, one introduces a special construction called rigged Hilbert space.

51

52

Quantum Mechanics I

states, which have some uncertainty both in the position and momentum, do have norm one in
accordance with the First postulate.
Any quantum states | can be decomposed into the basis associated with a continuous-variable
observable according to
+
| =
(x) |x dx.
(3.3)

This equation replaces Eq. (1.1) for the decomposition of a state into a discrete basis. The function
(x) is called the wavefunction of the state | in the x-basis (-representation) and is the continuousobservable analog of the column representation of a vector in a Hilbert space of a nite dimension.
Taking the adjoint of both sides of Eq. (3.3),

| =

(x) x| dx.

(3.4)

we also nd that the wavefunction of | is (x).

Exercise 3.1 Show that we can construct the following continuous analogues to important discretecase relations:
a) Instead of Eq. (1.11), we can write
(x) = x| ;

(3.5)

b) Instead of Eq. (1.31), we can write

|xx| dx = 1;

(3.6)

c) Instead of Eq. (1.9), we can write

1 | 2 =

1 (x)2 (x)dx;

(3.7)

Exercise 3.2 Show that, if the normalization rule were, as in the nite-dimensional case, x| x =
{
1 if x=x
0 if x=x , the above relations would be invalid.
Exercise 3.3

Show that, for physical states,

|(x)|2 dx = 1.

(3.8)

Exercise 3.4 Calculate the normalization factor A for the states with the following wavefunctions:
a) a top-hat function

{
tophat (x) =

0 if x < a or x > b;
A if a x b;

(3.9)

b) a Gaussian wavefunction
x2

Gauss (x) = Ae 2d2 .

Exercise 3.5

(3.10)

Find the wavefunction of the state of denite position |x0 in the position basis.

As in the discrete case, operators associated with continuous observables are given by

x
=

x |x x| dx.

(3.11)

3.1. CONTINUOUS OBSERVABLES

53

The operator functions are naturally dened as

f (x) |x x| dx.

f (
x) =

(3.12)

the two-dimensional function

For an arbitrary operator A,
A(x, x ) = x| A |x ,

(3.13)

is referred to as the operators matrix element.

As we see below and similarly to the discrete-variable case, the information about the matrix
element x| A |x for all x and x is sucient to reconstruct the operator. More generally, we can
perform operations with states and operators represented by one- and two-dimensional functions,
respectively, just as we operate with matrices in the discrete case, only replacing summation with
integration.
Exercise 3.6

Show that x
|x = x |x.

Exercise 3.7

Prove that:

a) any operator A can be written as

A =

A(x, x ) |xx | dx dx ;

(3.14)

b) if the operator is a function of x

,

| f (
x) | =

(x)f (x)(x)dx;

(3.15)

and any two states |, |

c) for any operator A,

| A | =

(x)A(x, x )(x )dx dx ;

(3.16)

d) the wavefunction of the state A | is

 
 
x A =

A(x, x )(x )dx ;

(3.17)

(x )A(x , x)dx ;

(3.18)

e) the wavefunction of the state | A is

 
 
 A x =

f) matrix elements of an operator A and its adjoint A are related as

(A )(x, x ) = A (x , x);

(3.19)

can be written in terms of their matrices as

g) the product of operators A and B


 
x AB  x =

A(x, x )B(x , x )dx .

(3.20)

54

Quantum Mechanics I

Figure 3.1: Approximate measurement of the position by projecting state | onto a state with a
top-hat wavefunction.
Let us now reformulate the Second postulate of quantum mechanics for measurements in continuousvariable bases. Suppose observable x
is measured in quantum state | with wavefunction x| =
(x). Acting similarly to the discrete case, we assign value prx = |(x)|2 as the probability to
observe a particular value of x. A subtlety arises, though, because if we associate this (or any other
nite) probability value to each specic x out of the continuum, the total probability to detect any
x will be innite. This is not surprising because, as we discussed earlier, eigenstates of the position operator are unphysical, and so is a projection measurement involving these states. It is even
theoretically not feasible to realize a perfectly precise measurement of the position observable.
As an alternative, let us model the measurement of x project state | onto the state with a
top-hat wavefunction (3.9) that takes a constant, non-zero value in the interval between x and
x + x, where x is arbitrarily small (Fig. 3.1). We nd in the rst order approximation
1
| tophat =
x

x+x

(x)dx

x(x)

(3.21)

and hence the probability to detect state |tophat in state | equals

2

prx x = || tophat | = |(x)|2 x.

(3.22)

Of course, not all apparata that measure the particles position involve projection onto the tophat function. However, the above result is universally applicable: the probability to detect a particle
in state | within a small interval [x, x + x] is given by Eq. (3.22). Accordingly, the quantity
prx = |(x)|2 has the meaning of the probability density associated with the state at position x.
If we apply Eq. (3.22) to nd the probability to detect any value of x between to +, we
nd unity in accordance with Eq. (3.8). This is not surprising: the particle can with certainty be
found somewhere, even though its precise position is not known.
Upon which state | will project after the measurement? The self-suggesting answer |x is, as
we discussed, unphysical. Yet it is useful as an approximation for many theoretical applications, as
long as we do not forget to take the normalization issue into account. The more physically realistic
answer will depend on the specics of the measurement apparatus; generally, one would obtain some
superposition or statistical mixture of multiple position eigenstates within a certain narrow interval.
For example, in the case studied above, the answer is |tophat .
Closely related to measurements is the notion of the expectation value. Denition 1.26 can
be reformulated for the continuous domain as follows: the expectation value associated with a
measurement of a random continuous variable x with probability density prx is given by

x =

xprx dx.

(3.23)

Under this denition, the expression (1.36) for the quantum expectation value remains the same. I
invite the reader to verify this independently in the following exercise.

3.2. DE BROGLIE WAVE

55

Exercise 3.8 Show that the expectation value of a continuous observable x

[dened as x =
+
x pr(x) dx] in a state | is given by x = | x
|.

Exercise 3.9 Show that the eigenstates of a Hermitian operator corresponding to dierent eigenvalues are orthogonal3 .

Table 3.1: Comparative summary of rules for operating with discrete-and continuous-variable bases
discrete basis {|vi }
continuous basis {|x}
orthonormality
vi | vj = ij
x| x = (x x )
+

decomposition
| = i |vi
| = (x) |x
i

of a state
second postulate
decomposition
of an operator

decomposition of 1
product of operators

i = vi |
pri = | vi | |2
(probability)
 
 
Aij = vi  A vj

A = Aij |vi vj |
i,j

1=
|vi vi |
i

(AB)ij = Aik Bkj

k

3.2

(x) = x|
prx = | x| |2
(probability
 
density)
 

A(x, x ) = x A x
+ +
A = A(x, x ) |xx | dxdx

= + |xx| dx
1

(AB)(x, x ) = A(x, x )B(x , x)dx

De Broglie wave

In the previous section, we studied the general mathematical machinery for handling the basis of
the Hilbert space formed by the eigenstates of a continuous observable. Now it is time to bring some
physics into the picture. We begin by postulating the relation between the position and momentum
eigenstates:
px
1
x| p =
ei ~ .
(3.24)
2~
Relation (3.24) states that the wavefunction of the state with denite momentum is an innite wave
of wavelength
dB = 2~/p,
(3.25)
which is known as the de Broglie wave. It manifests one of the central concepts of quantum mechanics
wave-particle duality, i.e. the property of all quantum matter to exhibit both particle and wave
features.
The de Broglie wave cannot be derived from the quantum mechanics postulates we studied so far.
Rather, it is a generalization of a multitude of experimental observations and theoretical insights.
The historical path towards the de Broglie wave can be briey outlined as follows.
In 1900, Max Planck explained the experimentally observed spectrum of blackbody radiation
by introducing the quantum of light, now known as the photon4 . He found that a good agreement
between theory and experiment can be obtained if one assumes that the energy of the photon is
proportional to the frequency of the light wave. The proportionality coecient, ~, became known
as the Planck constant.
Subsequently, in 1913, Niels Bohr has used the Planck constant in developing his model of the
atom, according to which, the electrons orbital is stable if its angular momentum equals an integer
number of ~. Based on this assumption one can calculate the spectrum of optical transitions of
the hydrogen atom, which exhibits very good agreement with the experiment. Similarly to Plancks
3 This appears to follow from Ex. 1.76. However, the statement of that Exercise was proven for a Hilbert space of
a nite dimension. In the present Chapter, the dimensions are innite.
4 The term photon has been introduced much later, in 1926, by the physical chemist Gilbert Lewis.

56

Quantum Mechanics I

theory of light, Bohrs model was empirical: it seemed to explain experimental results, but the
physics behind it remained a mystery.
The next major step towards the discovery of the de Broglie wave was made in 1923 by Arthur
Holly Compton, who provided theoretical explanation of diraction of X rays on free electrons. In
Comptons theory, photons were treated as ultrarelativistic particles, whose energy follows the famous Einsteins relation E = mc2 . Since we also have E = ~ due to Planck, we can calculate the
mass of the photon, m = ~/c2 , and its momentum5 , p = mc = ~/c. Expressing the electromagnetic wave frequency in terms of the wavelength, = 2c/, we nd p = 2~/ [cf. Eq. (3.25)]. The
collision between a photon and an electron can then be treated using the laws of momentum and
energy conservation, and an experimentally veriable dependence of the wavelength of the scattered
waves as a function of the scattering angle can be obtained. Again, this dependence showed excellent
agreement with the experiment.
Louis De Broglie in his 1924 PhD thesis hypothesized that Comptons theory did not have to
be limited to light particles. In fact, any particle that is moving with a certain momentum can be
associated with a wave with the wavelength given by Eq. (3.25). Under this assumption, one can
reinterpret Bohrs rule as a standing wave condition: the electrons model is stable if its circumference
ts an integer number n of de Brolie waves:
2r = ndB ,

(3.26)

where r is the radius of the orbit6 . Substituting expression (3.25) into Eq. (3.26), we nd pr = n~.
This is equivalent to Bohrs condition, because the electrons angular momentum is the product of
its momentum and the orbits radius.
The hypothesis of de Broglie was quickly conrmed by experiment. In 1927 at Bell Labs, Clinton
Davisson and Lester Germer observed diraction of a ux of electrons on the crystalline lattice
of nickel and measured the associated wavelength, which turned out to be in agreement with de
Broglies calculations.
We now proceed to derive some of the main consequences of the relationship between the position
and momentum eigenstates dened by the de Broglie wave.
Exercise 3.10

Estimate the de Broglie wavelength for

a) a car;
b) air molecules at the room temperature;
c) electrons in an electron microscope with a kinetic energy of 100 keV;
d) rubidium atoms in a Bose-Einstein condensate ate a temperature of 100 nanokelvin.
Exercise 3.11 Show that, according to the property (3.24), the position and momentum eigenstates can be expressed through each other as follows:
|p

|x

+
px
1

ei ~ |x dx;
2~
+
px
1

ei ~ |p dp.
2~

(3.27)
(3.28)

Exercise 3.12 Relying on the relation x| x = (x x ), show explicitly that Eq. (3.24) is
consistent with the orthonormality condition p| p = (p p ).
Hint: use Eq. (3.7).
5 The relation p = E/c between the momentum and energy of the photon is consistent with the expression for the
radiation pressure, which, according to Maxwells theory, equals the intensity of that radiation divided by the speed
of light. Radiation pressure has been experimentally observed by Peter Lebedev in 1900.
6 Smearing of the wave function was not known at that time.

3.3. POSITION AND MOMENTUM REPRESENTATIONS

57

As we discussed earlier, the convention to normalize continuous-variable eigenstates to a delta

function is mathematically convenient, but unphysical. For this reason, it is not correct to interpret
the square absolute value of the de Broglie wave, x| p = 1/2~, as a probability density in the
sense dened in Sec. 3.1. Actually, the deBroglie wave has innite extent in space, and thus the
probability to detect the particle within any nal interval is innitesimal.
The wavevector of the de Broglie wave is
k=

p
2
= .
dB
~

(3.29)

Sometimes it is convenient to handle momentum eigenstates |p in the physically equivalent form of

wavevector eigenstates |k = p/~, because then we need not worry about the Planck constant in the
exponent.
A minor subtlety associated with wavevector eigenstates is their normalization. The de Broglie
wavefunction for these states is given by
1
x| k = eikx ,
2

(3.30)

i.e. there is no factor of ~ in the denominator, in contrast to Eq. (3.24). Indeed, recalculating
Ex. 3.12 with Eq. (3.30), we obtain k| k = (k k ) as we would expect from eigenstates of a
continuous observable. Accordingly,

|k = ~ |p .
(3.31)
We see that two physically equivalent states, |p and |k, have a dierent norm. This is another
consequence of the unphysical character of normalization for continuous observable eigenstates.

3.3

Position and momentum representations

This section is dedicated to converting the representations of various states and operators between
the position and momentum bases. As in the discrete case, the primary tool for such conversion is
the method of inserting the identity operator, i.e. utilizing the fact that the operator
+
+
=
1
|xx| dx =
|pp| dp
(3.32)

can be inserted into any inner product expression.

Exercise 3.13 Find the explicit formulae for converting between the position (x) and momentum

(p)
representations of a given quantum state |.
Answer:
+
1
i px

~ dp;
(p)e
(3.33a)
(x) =
2~
+
px
1

(x)ei ~ dx
(p)
=
(3.33b)
2~
For reference, we also write these formulae in the wavevector representation:
+
1
ikx

(x) =
(k)e
dk;
2
1

(k)
=
2

(x)eikx dx.

(3.34)

(3.35)

As we see, conversion between the position and wavevector representations and back is simply the
direct and inverse Fourier transformation, respectively.

58

Quantum Mechanics I

In this course, we will always use a tilde [e.g. (p)

or (k)]
to denote wavefunctions in the
momentum or wavevector representations. Comparing Eqs. (3.33b) and (3.35) we nd the relation
between these two representations:

(k)

(p)
=
(3.36)
~
for p = k~.
Exercise 3.14 Consider a function V (
x) of the position operator. Write the matrix element of
this operator in the momentum basis.
Answer:
+

i
1

V (p , p ) = p | V (
x) |p =
e ~ x(p p ) V (x)dx.
(3.37)
2~
x) =
Exercise 3.15 Find the momentum representation of the position operator function V (
V0 (x).
Answer: Constant V0 /(2~).
Exercise 3.16

For a normalized Gaussian wavefunction

(x) =

1
(d2 )1/4

ei

p0 x
~

(xa)2
2d2

(3.38)

a) calculate the corresponding wavefunction in the momentum basis.

(Hint: use the standard rules for the Fourier transformation);

Verify normalization.

b) calculate the expectation values of the position and momentum.

Answer:

(k)
=

d2

)1/4

ei(kk0 )a e(kk0 )

2 2

d /2

(3.39)

where k0 = p0 /~. x = a, p = p0 .
Exercise 3.17 Show that the matrix element x| p |x of the momentum in the position representation is given by
d
d
x| p |x = i~ (x x ) = i~ (x x )
(3.40)
dx
dx
Exercise 3.18

Show that, for an arbitrary state |,

x| p | = i~

Exercise 3.19

d
d
x| = i~ (x).
dx
dx

(3.41)

Show that x| p2 | = ~2 d2 (x)/dx2 .

Equation (3.41) is an important relation demonstrating the action of the momentum operator
upon a quantum state in the position basis. If state | has wavefunction (x) in the position basis,
state p | has wavefunction i~d(x)/dx.
Exercise 3.20 Obtain the analogues of the two expressions above for the position operator in the
momentum representation.
a) Show that the matrix element is
p| x
|p = i~

d
(p p )
dp

(3.42)

d
(p).
dp

(3.43)

b) Show that, for an arbitrary state |,

p| x
| = i~

3.3. POSITION AND MOMENTUM REPRESENTATIONS

59

An important operator in one-dimensional quantum mechanics is the position displacement operator, which translates the wavefunction in the position basis by a certain distance. As we see below,
0 /~
this operator is given by eipx
, where x0 is the displacement distance (Fig. 3.2). The momentum displacement operator is dened similarly and given by eip0 x/~ (x), p0 being the displacement
momentum.

Figure 3.2: Eect of the position displacement operator on a wavefunction.

Exercise 3.21
a)

Show that
0 /~
eipx
|x = |x + x0 ;

(3.44)

0 /~
b) if the wavefunction of state | is (x), then the wavefunction of state eipx
| is (xx0 );

c) application of the position displacement operator to a state adds x0 to the mean position value,
but does not change the mean momentum value;
d) application of the position displacement operator to a state does not change the position and
momentum uncertainties.
Exercise 3.22 Show that the momentum displacement operator eip0 x~ has similar properties with
respect to the momentum as does the position displacement operator with respect to the position.
Exercise 3.23

Do the position and momentum displacement operators commute?

Exercise 3.24 Show that for a real wave function (x), pr(p) = pr(p). Show that the expectation
value of the momentum observable is zero.
Now that we have practiced switching between the position and momentum bases, we are ready
to introduce the uncertainty relation between these observables. As we know from Sec. 1.14, the
uncertainty relation of any two observables is determined by their commutator.
Exercise 3.25

Show that, for any state |,

a)
d
(x);
dx

(3.45)

d
(x) i~(x);
dx

(3.46)

x| x
p| = i~x
b)
x| px
| = i~x
c)

[
x, p] = i~.

(3.47)

Exercise 3.26 Show that the Heisenberg uncertainty principle for the position and momentum
operators has the form
2 2 ~2
.
(3.48)
x
p
4

60

Quantum Mechanics I

Exercise 3.27

For a Gaussian wavefunction,

Gauss (x) =

x2

e 2d2 ,
1/4 d

calculate x2 , p2 and verify the uncertainty principle.

Hint:
+

2
x2 ex dx = /2

(3.49)

(3.50)

Answer: x2 = d2 /2, p2 = ~2 /2d2 .

We have thus obtained the uncertainty principle in its original form, as dened by Werner
Heisenberg in 1927: simultaneous, precise measurement of a particles position and momentum
is not possible. From the last example above we see that the position-momentum uncertainty
can be interpreted as a consequence of one of the properties of the Fourier transformation: if the
wavefunction in the position basis becomes narrower, its Fourier transform, i.e. the wavefunction
in the momentum basis, becomes wider. A reduction in the position uncertainty of a state entails
an increase in the momentum uncertainty, and vice versa.
Now let us reproduce in its original form another research masterpiece, the 1935 EinsteinPodolsky-Rosen paradox.
Exercise 3.28 Suppose each of the two observers, Alice and Bob, holds a one-dimensional pointlike particle. The two particles are prepared in entangled state |AB whose wavefunction is
(xA , xB ) = (xA xB )

(3.51)

a) Express the state of the two particles in the momentum representation. Neglect normalization.
b) Suppose Alice performs a measurement of her particles position and obtains some result x0 .
Onto which state will Bobs particle project?
c) Suppose Alice instead performs a measurement of her particles momentum and obtains some
result p0 . Onto which state will Bobs particle project?
A , pB ) = (pA + pB ); b)|x0 ; c)|p0 .
Answer: a)(p
Here is how Einstein, Podolsky and Rosen describe this paradoxical situation:
...either one or the other, but not both simultaneously, of the quantities P and Q can be predicted,
they are not simultaneously real. This makes the reality of P and Q depend upon the progress of the
measurement carried out on the rst system, which does not disturb the second system in any way.
No reasonable denition of reality could be expected to permit that.7
In other words, by choosing to measure in the position or momentum basis, Alice can create
one of two mutually incompatible physical realities associated with Bobs particle: either a state
with a certain position and uncertain momentum or the other way around. This appears unphysical
because the physical reality at Bobs location is changed through a remote action by Alice without
any interaction.

3.4

The free space potential

For the remainder of this chapter, we will solve the Schrodinger equation for the motion of a pointlike
particle in a eld of some conservative force. In classical physics, this motion is determined by the
Hamiltonian, i.e. the sum of the kinetic and potential energies expressed as a function of the
particles position and momentum. The quantum expression for the Hamiltonian is identical to
classical except that the canonical observables are written as operators:
p2
= V (
.
H
x) +
2m
7 Einstein,

(3.52)

Podolsky and Rosen use symbols Q and P for the position and momentum observables, respectively.

3.4. THE FREE SPACE POTENTIAL

61

Here m ia the particles mass, p2 /2m is the kinetic energy operator and V (
x) is the potential energy
operator, which is the function of the position operator. Our goal is to solve the Schrodinger equation
associated with this Hamiltonian.
Because both canonical observables can be expressed in the position basis (see Ex. 3.19), it is
also convenient to write the Schr
odinger equation in this basis that is, take the inner product of
x| with both sides of Eq. (1.76).
Exercise 3.29

Show that in the x-basis, the Schrodinger equation (1.76) takes the form
[
]
d(x, t)
i
~ 2 d2
= V (x)
(x, t).
dt
~
2m dx2

(3.53)

We have thus rewritten the Schr

odinger equation, describing the quantum evolution of an abstract
object the quantum state as something much more concrete: a partial dierential equation.
Let us now solve this equation for the simplest case, V (x) 0 (the free space evolution). Under this
condition, any eigenstate |p of the momentum operator with eigenvalue p is also an eigenstate of
the Hamiltonian (3.52) with the eigenvalue E = p2 /2m.
Exercise 3.30 Show that the wavefunction describing the evolution of state |p under Hamiltonian
(3.52) with V (x) 0 is given by
p
p2
1
p (x, t) =
ei ~ xi 2m~ t .
2~

(3.54)

Find the phase velocity of this wave.

According to the above, the time-dependent behavior of the wavefunction of the momentum
eigenstate is similar to that of a traveling wave of wavevector k = p/m and angular frequency
= p2 /2m~ = ~k 2 /2m.

(3.55)

The evolution of this wave consists of translation with the phase velocity vph = dB /T = /k =
p/2m, where T = 2/ is the period associated with the wave motion.
Note that the phase velocity is dierent from the value p/m expected classically. The explanation
is that in the momentum eigenstate, the position is completely uncertain and the probability of
nding the particle anywhere in space is the same, prx |p (x, t)|2 = 1/2~. This probability does
not change with time. Accordingly, the phase velocity of the de Broglie wave does not have any
direct correspondence to the motion of matter8 .
In order to understand how the Schr
odinger evolution translates into motion, we have to study
a state whose wavefunction is to some extent localized in space (we use the term wavepacket for
such wavefunctions). For example, let us investigate the Gaussian wavepacket with a nonzero mean
momentum.
Exercise 3.31
with a = 0.

Consider a wavefunction that at the moment t = 0 has a Gaussian form (3.38)

0) in the wavevector representation. Find its evoa) Find the corresponding wavefunction (k,

lution (k, t) under the free space Hamiltonian.

8 In fact, the phase velocity of the de Broglie wave is a matter of convention rather than physics. If we assume
a constant potential V (x) = V0 and calculate the Hamiltonian eigenstate with energy E + V0 , its time-dependent
wavefunction will be of the form
E+V0
p
1
ei ~ xi ~ t .
p (x, t) =
2~
The spatial behavior of this wavefunction is the same as that in Eq. (3.54), because the energy is related to the
momentum according to E = p2 /2m. But the time evolution will be faster because the frequency of the wave now
equals (E +V0 )~ rather than E/~. Accordingly, the wavefunction, which describes the physical setting that is identical
to that of a free particle of energy E, will have a dierent phase velocity.

62

Quantum Mechanics I
b) Find the wavefunction (x, t) in the position basis.
Hint: express k = k0 + k. Perform the inverse Fourier transformation with respect to k and
then use Eqs. (C.24) and (C.25) for the Fourier transformation of a shifted function.
c) Find the mean value of the position x and its variance x2 as functions of time. Compare
the velocity of the wavepacket with that expected classically and the phase velocity of the de
Broglie wave.
(
)
2
~ 2 t2
Answer: x = (p0 /m)t, x2 = d2 1 + m
2 d4

We see that the evolution of the Gaussian wavepacket consists of three features: (1) multiplication
by an overall phase factor (which has no physical consequences), (2) translation with the eective
velocity vgr = p0 /m and (3) spreading in space. As we see below, the spreading is a secondary
eect that can often be neglected. Aside from that spreading, the shape of the Gaussian wavepacket
remains the same; it travels as a single unit, reproducing the classical motion of a pointlike particle.
Exercise 3.32
a) Show that if p0 greatly exceeds the momentum uncertainty of the initial
wavepacket, the distance traveled by the center of the wavepacket during time t is much
greater than the length over which it spreads.
b) Estimate the time required by a wavepacket associated with a single electron with the position
uncertainty on the scale of 1
A, to spread over a length of 1 mm.
It is instructive to draw an analogy with wave optics. In its many aspects, the evolution of the
wavepacket resembles that of a short laser pulse. Such a pulse can be decomposed, by means of
the Fourier transform, into a set of plane waves with dierent frequencies. If the pulse propagates
in vacuum, such that all its plane wave components travel with the same phase velocity (c), the
group velocity of the wavepacket is also c and the wavepacket does not spread. But if the pulse
enters a refractive medium in which the phase velocity depends on the wavelength (i.e. dispersion
is present), then the wavepacket motion will be determined by its group velocity,
vgr = d/dk,

(3.56)

and the wavepacket may experience spreading. This is exactly what we observe with our matter
wavepacket.
Exercise 3.33 Apply expression (3.56) to Eq. (3.55) to show that the group velocity of a matter
wavepacket with a mean momentum p0 and momentum uncertainty p p0 equals vgr p0 /m.
Classical features associated with the motion of wavepackets can be revealed in an even more
general form by way of the following calculations.
Exercise 3.34 Show that for any arbitrary state | that evolves in accordance with the Schrodinger

equation and any arbitrary operator A,

d A
i
=
[H, A] ,
(3.57)
dt
~

where averaging is meant in the quantum-mechanical sense, i.e. A = | A |.
Exercise 3.35

For the Schr

odinger evolution of a state of a pointlike particle, show that

a)
x
=
p/m.
b)
d
p
=
dt

dV (x)
(x) (x)
dx

(3.58)


 dV 


dx = 
 ,
 dx 

where V (x) is the potential as a function of the position.

(3.59)

3.5. MULTIDIMENSIONAL MOTION AND PROBABILITY FLUX

63

Relation (3.59) is known as the Ehrenfest theorem. In order to understand its meaning, let us
remember that, in classical mechanics, we associate a potential energy with a conservative force eld
by means of the following expression (here specialized to one-dimensional motion):
x2
U (x1 ) U (x2 ) =

F (x)dx.

(3.60)

x1

Using the Newton-Leibniz axiom, we can rewrite it as

F (x) =

dV (x)
.
dx

(3.61)

But the right-hand side of the above equation is identical to that entering Eq. (3.59). In other words,
Eq. (3.59) tells us that the time derivative of the particles mean momentum equals the mean force
acting on that particle. But this is simply the second Newtons law!

3.5

Multidimensional motion and probability flux

Exercise 3.36

Recalling that pr(x) = (x) (x), derive the following equation:

dpr(x)
dj
= ,
dt
dx

where
j = i

~
2m

d
d

dx
dx

(3.62)
)
.

(3.63)

Definition 3.1 The quantity j in the above equation is the probability current density indicating
the ow of the probability density per unit time. Eq. (3.62) is called the continuity equation.
j governs the
Note 3.1 The continuity equation, whose multidimensional form is dpr(x)
=
dt
ow of any conserved quantity and is present in many elds of physics, for example, electro- and
hydrodynamics.
Exercise 3.37 If the wavefunction can be made real through multiplying by an overall phase
factor, the probability ow vanishes.
Exercise 3.38 Show that the probability density current for the de Broglie wave is proportional
to its momentum.
Exercise 3.39 Verify the continuity equation explicitly for a spreading Gaussian wavepacket at
rest (Ex. 3.31).
Exercise 3.40 Consider a wavefunction which is a superposition of two wavepackets, one dened
by Eq. (3.38) and one being its mirror image around the point x = 0. Assume p0 = 0 and a d,
i.e. the two wavepackets have almost no spatial overlap.
a) What is the correct normalization factor?
b) Find the corresponding probability density pr(p) in the momentum space.
c) Using the result of Exercise 3.31, nd the probability density in the position space after a very
long time t. Discuss the analogy with two-slit diraction in optics.
Exercise 3.41

Derive, for a three-dimensional geometric space,

a) the Schr
odinger equation in the x-basis;
b) the quantum-mechanical continuity equation.

64

Quantum Mechanics I

3.6

Time-independent Schr
odinger equation

| = E |. This
Frequently, our goal is to nd the energy eigenstates, i.e. such states | that H
equation is referred to as the time-independent Schr
odinger equation. In the x-basis, it takes the
form (cf. Eq. (3.29)):
[
]
~2 d2
V (x)
(x) = E(x)
(3.64)
2m dx2
This is a second-order ordinary dierential equation, which can be readily solved, particularly if
the potential is a nice function of the position. For example, let us nd the solution for a constant
potential.
Exercise 3.42 Find the general solutions to Eq. (3.64) for V (x) = V0 . Consider the cases (a)
E > V0 (b) E < V0 .

Answer: (a) Aeikx + Beikx ; (b) Aekx + Bekx where k = 2m|E V0 |/~ and A, B are random
coecients.
We see that the solutions are fundamentally dierent for the energies above and below the
potential level. In the former case, we obtain an oscillatory solution akin to the de Broglie wave. In
the latter case, the solutions are exponentially growing or falling as a function of the position. If the
potential is constant for all positions, such a solution will blow up at x a behavior which
implies innite probabilities and thus cannot occur in a physical state (or even in an approximation
thereof). Hence there are no Hamiltonian eigenstates with eigenvalues below the potential energy
level. However, situations where the energy is lower than the potential for a part of the position
axis are possible, as is the case, for example, with quantum tunneling.
Before we proceed to nding the energy eigenwavefunctions for specic potentials, let us derive
a few general properties of these wavefunctions that will assist us in this task.
Exercise 3.43 Show that, if (x) is a solution of the time-independent Schrodinger equation, then
both (x) and d(x)/dx must be continuous if both the potential V (x) and the wavefunction are
nite for all x.
This result will turn out extremely useful for many problems in which the potential is given by
a piecewise function, i.e. a set of dierent elementary functions each dened on its own interval of
positions. It is relatively easy to nd the solution for each interval, but then these solutions must
be stitched together to form a physically meaningful wavefunction. Exercise 3.43 provides us with
the guideline for this stithcing.
Exercise 3.44
E. Show that:

Consider set SE consisting of all Hamiltonian eigenstates with energy eigenvalue

a) There exists a spanning set of SE which consists only of states with real wavefunctions so
any element of SE can be written as a linear combination of energy eigenstates with the same
energy eigenvalue and real wavefunctions;
b) if V (x) is an even function of the position, ehere exists a spanning set of SE consisting only of
states with wavefunctions that are either even or odd.
Exercise 3.44 simplies our search for solutions of the time-independent Schrodinger equation.
In particular, it is sucient to only look for wavefunctions that are real. Then we are guaranteed
not to miss any solution that cannot be written as a linear combination of those solutions that
we have found.
For example, the de Broglie wave
px
1
p (x) =
ei ~ ,
2~

(3.65)

3.7. BOUND STATES

65

associated with momentum eigenstate |p, is a solution of the time-independent Schrodinger equation
with energy eigenvalue E = p2 /2m. The same is true for the wavefunction
px
1
p (x) =
ei ~ ,
2~

(3.66)

which is the de Broglie wave for momentum eigenstate |p. But then the real wavefunctions
p,+ (x) =

1
p (x) + p (x)
px
=
cos
2
~
2~

(3.67)

p, (x) =

p (x) p (x)
1
px
=
sin
2i
~
2~

(3.68)

and

also represent energy eigenstates with the same eigenvalue. The de Broglie wavefunctions (3.65) and
(3.66) and hence can any other wavefunction corresponding to the same energy can be written
as linear combinations of these real wavefunctions.

3.7

Bound states

Bound states are characterized by a wavefunction that tends to zero at |x| , so the particle
exhibits some degree of localization. This property is typical for energy eigenstates in well-shaped
potentials, i.e. such elds in which the particle is attracted towards a certain location or a set of
locations. Physical examples include a pea inside a glass, a ball on a spring (harmonic oscillator)
or an electron within an atom. For this type of potentials, we usually take advantage of Ex. 3.44(a)
and look for solutions of the time-independent Schrodinger equation in the real domain.
Exercise 3.45 Consider a potential V (x) that approaches a particular value V0 at |x| .
Show that an energy eigenstate is bound if and only if its energy does not exceed V0 .
As we shall see in the examples below, the boundary conditions imposed on the behavior of the
wavefunction at |x| can be satised only for specic, discrete values of the energy. In other
words, a well-shaped potential features a discrete or quantized spectrum of energy eigenvalues.
V0

V0

?a / 2

0
a/2

Figure 3.3: Potential for Ex. 3.69

Exercise 3.46
tential:

Find the energy eigenvalues and eigenwavefunctions for the nite square well po{
V0 for |x| > a/2
V (x) =
(3.69)
0 for |x| a/2

a) Write the general solution for each region where the potential is constant. Eliminate unphysical
terms that grow at innity.
b) Apply the statement of Ex. 3.43 to stitch these results together. Derive the transcendental
equation dening the energy eigenvalues.

66

Quantum Mechanics I
Hint: use the result of Ex. 3.44(b).
Answer: Even wavefunctions:
tan =
odd wavefunctions:

(0 /)2 1;

(3.70a)

cot = (0 /)2 1,

where
=

(3.70b)

2mEa/2~; 0 = 2mV0 a/2~.

(3.71)

c) Solve the equation for the energy eigenvalues graphically.

a)

V0 = 2.25(2h 2 / ma 2 )

-a/2

V0 = 12.25(2h 2 / ma 2 )

a/2

-a/2

V0 =

a/2

-a/2

a/2

b)

10
20
30
40
V0 , in units of 2h 2 / ma 2

50

Figure 3.4: Solution for Ex. 3.69. a) Wavefunctions for the lowest energy eigenstates with dierent
well depths. The well on the left supports only one bound state; the well in the middle, three bound
states, the well on the right supports innitely many. The zero of each wavefunction corresponds to
the associated energy eigenvalue. b) Three lowest energy solutions of the transcendental equations
(3.70) as a function of the well depth. At least one bound state exists for all V0 ; the existence of
further bound states is conditional on V0 exceeding certain threshold values.
It is instructive to plot the wavefunctions for a few lowest energy eigenstates of potential (3.69).
As we can see in , the wavefunction extends outside the box, so there is a nite probability to nd
the particle in the region where the potential is higher than the particles energy. This is, of course,
an expressly nonclassical phenomenon. The deeper the well, the smaller the part of the wavefunction
outside the well, and hence the probability to nd the particle in that region. In the limit V0 ,
this probability tends to zero. In this case, the problem permits an analytic solution, as we see in
the following Exercise.
As expected from Ex. 3.42, we see exponential decay outside the well and oscillatory behavior
inside. For each subsequent energy eigenstate, the number of times the wavefunction crosses the x
axis increments by one. The increasing number of crossings is associated with a faster oscillation,
a higher wavevector, and hence a higher energy value. Accordingly, for each number of crossings
there exists a certain minimum potential, below which the bound state no longer exist [Fig. 3.4(b)].
A well of a nite depth can thus support only a nite number of bound states. However, no matter
how shallow the well, it does support at least one bound state [Fig. 3.4(b)].

3.7. BOUND STATES

67

Exercise 3.47 Find the energy eigenvalues and bound stationary states for Ex. 3.46 in the case
V0 (the potential referred to as the potential box ).
Answer: A discrete energy spectrum with
En =

n (x)

n (x)

~ 2 2 n2
;
2ma2

(
)

a2 sin nx
, even n
a

, a/2 x a/2;
(
)

a2 cos nx
,
odd
n.
a

(3.72)

(3.73)

0, |x| > a/2.

These wavefunctions are displayed in Fig. 3.4(a), right panel.

The above wavefunctions exhibit a number of interesting features:
(x) = 0 outside the box;
d(x)/dx exhibits discontinuities at x = a/2;
(x) is continuous for all position values.
In order to better understand them, let us rewrite the time-independent Schrodinger equation (3.64)
as follows:
~2 d2

(x) = [E V (x)](x).
(3.74)
2m dx2
Outside the box, V (x) = . Therefore, if (x) is nonzero, the right-hand side of the above equation
becomes innite, and so does d2 (x)/dx2 over a continuous interval of positions. This is not possible
for any regular function.
The innite value of the potential outside the box also implies that the conditions of Ex. 3.43 are
not fullled, so neither the wavefunction nor its derivative have to be continuous at x = a/2. In the
present case, we observe that the wavefunction is continuous but the derivative is not. Intuitively
this can be understood as follows. A discontinuity in d(x)/dx implies that the right-hand side of
Eq. (3.74) is singular, i.e. it takes on innite values which is exactly what we have in our problem.
But if the wavefunction itself is not continuous, the potential must exhibit singularity of the second
order i.e. be a derivative of a function that is already singular. Such potentials are extremely
exotic and do not occur within the scope of basic quantum mechanics. Therefore, in solving the
time-independent Schr
odinger equation in the position basis, it is generally safe to assume that the
wavefunction behaves continuously.
Exercise 3.48 For the eigenstates of the previous problem, nd the uncertainties of the position
and momentum and verify the uncertainty principle.
{

for |x|<a/2
Exercise 3.49 Consider the state (x) = Ax
(A = 2 3/a3/2 being the norm) in the
0 for |x|a/2
potential of the previous problem. Find the energy spectrum of this state, i.e. the probabilities
pr(En ) to measure each energy eigenvalue. Show that these probabilities sum up to 1. (Hint:

1/n2 = 2 /6.)
Exercise 3.50

For the nite well potential (3.69):

a) Find analytically the approximate corrections to the rst two energy levels of an innitely deep
potential well (Ex. 3.47) when it is replaced by a nite well with a V0 E1 , where E1 is given
by Eq. (3.72).
b) Find numerically all energy eigenvalues for k0 a = 10. How many energy levels are there
altogether? Is your result for n = 1, 2 consistent with the result of part (a)?

68

Quantum Mechanics I
c) What is the minimum depth that is required in order for the well to contain a given number
N of bound eigenstates?

Exercise 3.51 Find the transcendental equation for the energy eigenvalues associated with the
bound stationary states of the potential

+ for x 0;
0 for 0 < x a;
V (x) =

V0 for x > a.
Compare your result with that of Ex. 3.69.
?

V0

0
a

Figure 3.5: Potential for Ex. 3.51

Exercise 3.52 Solve the time-independent Schrodinger equation for the nite well potential (3.69)
anlytically in the limit of an innitely deep and narrow potential well: a 0, V0 = W0 /a, with
W0 = const. How many bound states can this well contain?
Exercise 3.53
W0 (x)

Find the energy eigenvalues and the bound stationary states of the potential V (x) =

a) (a) using the position basis; (b) using the momentum basis.
Verify the consistency of your solutions with one another and with the result of Ex. nitewellEx2
Hint:
+
+
1
1
dx
=
;
dx = /2
(3.75)
2
1
+
x
(1
+
x2 )2

Answer: Single eigenstate with E = W02 m/2~2 . Wavefunctions:

(x) =

{
k0

ek0 x at x > 0
;
ek0 x at x 0
3/2

where k0 =

2mE/~ = W0 m/~2 .

2k0

(k)
=
,
2(k02 + k 2 )

Exercise 3.54 A particle is in the bound state of the potential V (x) = W0 (x). The potential
suddenly changes to V (x) = 2W0 (x). Find the probability that the particle will remain in a
bound state.
Exercise 3.55

Investigate the bound states of the potential V (x) = W0 (x a) W0 (x + a).

a) Find the transcendental equation for the energy eigenvalues (consider both the even and odd
case).
b) Show that in the limit a this equation becomes identical to that for a single well.

3.8. UNBOUND STATES

69

c) Find the rst order correction to the solution of the transcendental equation for the case of a
being very large (k0 a 1), but not innite.
Exercise 3.56 Under the conditions of the previous problem (distant wells), suppose the particle
at the moment t = 0 is localized at the rst well (i.e. its wavefunction is that of Ex. 3.53 centered
at x = a). What is the probability of nding it in the second well as a function of time?
Exercise 3.57 (Kronig-Penney solid state model). Find allowed ranges (zones) of energy eigenvalues
for a particle in a periodic potential eld
a)

b)

~2 k0
V (x) =
(x na);
m
n=

(3.76)

~2 k0
(x na).
V (x) =
m
n=

(3.77)

Consider both limiting cases k0 a 1 and k0 a 1.

Exercise 3.58 Show that bound energy eigenstates states of a pointlike particle with a single degree
of freedom cannot be degenerate.

3.8

Unbound states

Unbound wavefunctions take nite, nonzero values at innity. In contrast to bound states, stationary
unbound states possess the following properties.
Their energy exceeds the value of the potential at x or x + or both.
They are not normalizable.
Their associated energy eigenvalues form a continuous spectrum.
They can be degenerate: more than one linearly independent wavefunctions may correspond
to a specic energy eigenvalue.
The most straightforward example of the unbound state is the momentum eigenstate |p in free
space. The associated energy eigenvalue, E = p2 /2m, exceeds
the potential V (x) = 0 at innity.

Since there exists a momentum eigenstate |p, where p = 2mE, for every energy value, these values
form a continuum. For every energy value, states |p are degenerate.
Let us now study the unbound energy eigenstates that correspond to more complicated potential
energy functions.
Exercise 3.59

(the single-step potential). Find the eigenstates of the Hamiltonian

{
0 for x 0
V (x) =
V0 for x > 0

(3.78)

(Fig. 3.6) corresponding to a given energy E > V0 . Show that, for every value of E, there exist
two linearly independent solutions. Choose the solutions that correspond to the following physical
situations: (1) a wave approaching the step only from the left and (2) a wave approaching only from
the right.
Answer (in the notation of Fig. 3.6). First solution (D = 0):
B=A

2k0
k0 k1
; C=A
,
k0 + k1
k0 + k1

(3.79a)

70

Quantum Mechanics I

Figure 3.6: Notation for the de Broglie component waves in a single-step potential (Ex. 3.59 and
3.62)
and (2) an initial de Broglie wave approaching from the right (A = 0)
B=D

2k1
k1 k0
; C=D
.
k0 + k1
k0 + k1

(3.79b)

Now let us try and relate our calculation to the intuitive physical picture of a material particle
encountering a potential step. Making this connection may be dicult because the energy eigenstate
is time independent, whereas the collision of a particle with a step, as we intuitively imagine it, is
an intrinsically time-dependent phenomenon.
Some intuitive connection can be drawn in we recall that even stationary states experience
quantum evolution, which consists of the time-dependent phase factor eiEt/~ . Accordingly, the
de Broglie waves associated with amplitudes A and C (we will call them the A- and C-waves,
respectively) move to the right while the B-wave is moving to the left.
This situation resembles a continuous laser beam which propagates from air into glass, experiencing partial reection in accordance with the Fresnel equations. Similarly to the situation with
the quantum particle, the reection is not an instantaneous event but a stationary process which
comprises motion of the electromagnetic waves in space and time. In fact, if we compare the Fresnel
equations for the eld amplitudes with Eqs. (3.79), and take into account that the optical wavevector is proportional to the index of refraction, we will nd these two sets of equations completely
identical!
An interesting feature of the result (3.79) is that the amplitude C of the transmitted de Broglie
wave is higher than the amplitude A of the incident one. The particle appears more likely to be
found behind the step than within the same interval in front of it. Doesnt that contradict the law
of conservation of matter?
We can gain some insight for answering this question by looking at optical waves again. According
to Fresnel equations, the amplitude of the light wave inside glass is also higher than that in the air.
However, this does not conict with the law of energy conservation because the wave inside the glass
travels at a lower speed. Accordingly, the ux of energy (Poynting vector) carried by the transmitted
wave is lower than that of the incident wave.
A similar argument can be drawn in the case of a quantum particle. It is not the probability
density associated with the wavefunction that determines the conservation of matter but rather the
probability density current studied in Sec. 3.5. As we learned in Ex. 3.38, this current is proportional
not only to the squared absolute value of the de Broglie wave amplitude, but also to its momentum.
If we take this into account, we will observe that the conservation of matter is perfectly sustained.
Exercise 3.60 Find the probability density currents for each wave in Eqs. (3.79a) and (3.79b).
Find the reection and transmission coecients for these currents and show that their sum is one.
What is the behavior of these coecients for E V0 and E ?
Exercise 3.61

Solve Ex. 3.59 for energies below V0 . Verify that the reection coecient is one.

3.8. UNBOUND STATES

71

An even better overview of the quantum reection of a particle on a potential step can be obtained
if we combine multiple de Broglie waves into a Gaussian wavepacket and then study its evolution in
a way that is similar to Ex. 3.31, but in the potential given by Eq. (3.78).
Exercise 3.62 Find the evolution of a Gaussian packet initially described by Eq. (3.38) with a
positive p0 and negative a in the single-step potential eld (Fig. 3.6). Assume that
|a| d so the wavepacket is at rst entirely to the left of the step;
d2 ~t/m so spreading of the wavepacket (Ex. 3.31) can be neglected;
the initial average energy of the particle E = p20 /2m is greater than V0 ;
the rms momentum
uncertainty of the wavepacket ~/2d is small compared to the average
momentum ~k1 = 2m(E0 V0 ) behind the step.
As we see, upon encountering the step the initial wavepacket splits. A part of the wavepacket
continues to propagate past the step with a lower group velocity, and another part reects o the
step and begins to propagate in the backward direction.
As the nal comment about the potential step problem let us note that the fact that the particle
has a probability to bounce o a potential step that is lower than the particles energy or even
negative (as in the case described by Eq. (3.79b)) is expressly quantum. Any classical particle
will simply y above the potential step, reducing or increasing the speed but never reversing the
direction of its motion.
Even more nonclassical is the eect of quantum tunneling, which we study next.

Figure 3.7: Tunnelling through a barrier (Ex. 3.63)

Exercise 3.63

Consider the potential in Fig. 3.7, i.e.

{
0 for x 0 or x > L
V (x) =
V0 for 0 < x L

(3.80)

a) What is the degeneracy of energy levels?

b) Find the solution of the time-independent Schrodinger equation corresponding to a de Broglie
wave entering from the left and 0 < E < V0 .
c) Find the transmission and reection coecients for the probability current. Is their sum equal
to one?
d) Investigate the boundary cases E 0 and E V0 .
Answer.

 2
E 
4k02 k12
T =   = 2 2
;
2
A
4k0 k1 + (k1 + k02 )2 sinh(k1 L)
 2
B 
(k 2 + k 2 )2 sinh(k1 L)
.
R =   = 2 2 1 20 2 2
A
4k0 k1 + (k1 + k0 ) sinh(k1 L)

(3.81a)
(3.81b)

72

Quantum Mechanics I

We observe that a particle encountering a nite potential barrier which is higher than the particles kinetic energy has a nite probability of tunnelling through this barrier. This phenomenon
has, of course, no analogy in classical physics.
Exercise 3.64 Repeat Ex. 3.63 for E > V0 . Is the sum of the reection and transmission coecients equal to one in this case?
Exercise 3.65 Calculate the reection and transmission for scattering on a delta-potential V (x) =
W0 (x). Compare your results with those obtained from Eqs. (3.81) for an innitely thin and high
rectangular potential barrier (L 0, V0 = W0 /L).
Exercise 3.66 Perform numerical investigations of the propagation of a Gaussian wavepacket in a
potential shown in Fig. 3.7. Make simplifying assumptions as necessary. What time does it take the
wavepacket to penetrate the barrier?
Solving the previous exercise, we nd that the wavepacket spends no time inside the barrier.
The tunnelling occurs instantaneously: the transmitted wavepacket emerges behind the barrier
simultaneously with the initial wavepacket being absorbed. This can be traced back to the fact that
the C- and D-waves have a constant phase, so the complex argument Arg(x) of the wavefunction
at points x = 0 and x = L is the same. Schrodinger evolution of a de Broglie wave is equivalent to a
phase shift, and the lack thereof translates into a zero delay between the incident and transmitted
waves, resulting in an innite group velocity.
In Chapter 2 we already encountered a quantum phenomenon that appeared to enable fasterthan-light communication, but, after careful analysis, found this to be only an illusion. The same
can be said about the present situation, although the reason is dierent. Let us ask ourselves: at
what moment does an observer behind the barrier learn that the particle is entering the barrier? Is
it when a half of the wavepacket has emerged from the barrier, three-fourths or nine-tenths?
The correct answer is, much earlier than that. From complex analysis, we know that the Gaussian
function is analytic: any fragment of this function allows one to reconstruct its behavior in the entire
complex plane. Therefore, any observer anywhere in space is aware of the presence of the particle
with a Gaussian wavefunction, and can predict its evolution, from the initial moment of our analysis.
With this in mind, it makes no sense to talk about instant communication.
What if we instead tried a dierent wavefunction, for example, of the top-hat shape (3.9), which
takes on nonzero values only within a nite spatial region? The problem with such wavefunctions
is that, in contrast to the Gaussian case, their decompositions into the momentum basis are not
narrow: for example, the Fourier transform of the top-hat function is the sinc function. For such
functions, we cannot apply the approximations used to calculate the evolution of Gaussian functions
(see Ex. 3.62) for example, because they have signicant components corresponding to energies
above the barrier. This tremendously complicates the calculations; however, if the analysis is perform
thoroughly, it will exclude any possibility of superluminal propagation.

3.9
3.9.1

Harmonic oscillator
Annihilation and creation operators

The harmonic oscillator is a physical system of primary importance. Its applications extend far
beyond the motion of simple mechanical systems. For example, the quantum description of light,
not limited by the single-photon subspace studied in Chapters 1 and 2, is identical to that of the
harmonic oscillator. Other applications include solid state physics, molecular spectroscopy and even
atomic physics.
Figure 3.8(a) displays a classical harmonic oscillator a ball on a spring. The oscillation
frequency is related to the spring constant k in accordance with k = m 2 . Because the potential
energy of the spring is U = kx2 /2, its full Hamiltonian is given by
H=

m 2 x2
p2
+
2m
2

(3.82)

3.9. HARMONIC OSCILLATOR

73

a)

b)
k

pmax = mw xmax

m
xmax

Figure 3.8: A classical harmonic oscillator. a) The physical model; b) motion in the phase space.
The trajectory is elliptical with the half-axes pmax = mxmax .
The harmonic oscillator is a typical potential well. Therefore its energy eigenstates are bound and
nondegenerate (see Ex. 3.58). It is possible to nd the wavefunctions of these states by solving the
time-independent Schr
odinger equation (3.64) in the position basis. However, because the harmonic
oscillator is such an important physical system, we choose to develop its quantum theory in a more
general fashion. We begin by rescaling the position and momentum observables so they are more
convenient to operate with.
Exercise 3.67 Rescale the position and momentum observables, i.e. dene new observables (X =
Ax, P = Bp) with proportionality constants A and B such that (a) in the new variables (X, P ) the
phase space trajectory is circular [Pmax = Xmax , see Fig. 3.8(b)] and (b) for corresponding quantum
P ] = i.
operators, [X,
Answer:

m
p
=x
X

(3.83)
; P =
~
m~
Exercise 3.68

and P are dimensionless.

Show that the rescaled observables X

Exercise 3.69 If a certain quantum state has wavefunctions (x) = x| and (p)
= p| , what

are the corresponding wavefunctions (X) = X| and (P ) = P | in the rescaled variables?

and P - bases.
Write the relations analogous to (3.33) for converting wavefunctions between XExercise 3.70

Show that
X| P | = i

Exercise 3.71
Answer:

d
(X);
dX

| = i
P | X

d
(P ).
dP

(3.84)

and P
Write the Hamiltonian in terms of X
)
(
= 1 ~ X
2 + P 2 .
H
2

(3.85)

We now proceed to dening and studying the properties of the two operators which, as we shall
see in the next subscection, enact transitions between adjacent energy eigenstates.
Definition 3.2 The annihilation operator is dened as follows:
)
1 (
a
= X
+ iP ;
2
The operator a
is called the creation operator.
Exercise 3.72

Show that:

(3.86)

74

Quantum Mechanics I
a) the creation operator is

)
1 (
a
= X
iP ;
2

(3.87)

b) the creation and annihilation operators are not Hermitian;

c) their commutator is
[
a, a
] = 1;

(3.88)

d) position and momentum can be expressed as

)
(
= 1 a
X
+a
;
2

)
1 (
P = a
a
;
i 2

(3.89)

e) the hamiltonian can be written as

(
)
1

H = ~ a
a
+
;
2

(3.90)

f) the commutators
[
a, a
a
] = a
;

3.9.2

[
a , a
a
] =
a .

(3.91)

Fock states

Our next goal is to nd eigenvalues and eigenstates of the Hamiltonian. Because of Eq. (3.90), they
are also eigenstates of operator n
=a
a
.
Exercise 3.73

Suppose some state |n is an eigenstate of the operator a

a
with eigenvalue n:
a
a
|n = n |n

(3.92)

Show that
a) the state a
|n is also an eigenstate of a
a
with eigenvalue n 1;
b) the state a
|n is also an eigenstate of a
a
with eigenvalue n + 1.
Hint: Use Eq. (3.91).
As we know, energy spectra of bound states are nondegenerate, i.e. for each value of n there
exists no more than a single energy eigenstate |n. Therefore Ex. 3.73 shows that the states a
|n
and a
|n are proportional to states |n 1 and |n + 1, respectively. The proportionality coecient
can be found from the requirement that the energy eigenstates must be normalized.
Exercise 3.74

Show that

a)
a
|n =
b)
a
|n =
Hint: use n| a
a
|n = n.

n |n 1 ;

(3.93a)

n + 1 |n + 1 ;

(3.93b)

3.9. HARMONIC OSCILLATOR

75

We found that, if state |n with energy ~(n + 1/2) exists as a physical state (i.e. is a normalized
element of the Hilbert space), so does the state |n 1 with energy ~(n 1/2). Similarly, states
|n 2, |n 3, and so on must also exist. Continuing this chain for suciently many steps, we will
end up with energy eigenstates with negative energy values.
How can we resolve this contradiction? The only way is to assume that n must be nonnegative
integer so the chain is broken at n = 0, in which case
a
|0 = |zero .

(3.94)

Then (provided that the state |n = 0 exists), energy eigenstates form an innite set with equidistant
eigenvalues ~(n + 1/2).
Definition 3.3 Energy eigenstates of a harmonic oscillator are called Fock or number states. The
state |0 is called the vacuum state.
Exercise 3.75
Answer:

Express |n through |0.

( )n
a

|0
|n =
n!

(3.95)

Exercise 3.76 Calculate the wavefunction of the vacuum state in the position and momentum
representations.
Hint: use Eqs. (3.84), (3.86) and (3.94).
Answer:
2
2
1
1
0 (X) = 1/4 eX /2 ; 0 (P ) = 1/4 eP /2 .
(3.96)

Note that the above wavefunctions are unique up to an arbitrary overall phase factor. For the
vacuum state, by convention, we choose this factor so as to obtain a real and positive denite
wavefunction in the position basis. It then automatically follows that the wavefunction in the
momentum basis is also real and positive. Furthermore, as we shall see below, this convention
ensures that the wavefunctions of all other Fock states are also real.
By having explicitly found the wavefunction of the vacuum states, we have proven its existence,
and thus, automatically, the existence of all other Fock states because these states are obtained
from the vacuum state by applying the creation operator.
Now the physical meaning of the eigenvalue n of operator n
becomes also clear: this is the
number of excitation quanta in the harmonic oscillator. The number of such quanta is always
integer and, because energy levels are equidistant, all of them have the same energy, ~. In this
way, the excitation quanta in a harmonic oscillator resemble particles and, in many cases, they do
behave like particles. Among the examples are photons in an optical pulse phonons in a vibrational
mode of a solid state.
Exercise 3.77
|2.

a) By applying Eq. (3.95), calculate the wavefunctions of Fock states |1 and

b) Show that the wavefunction of an arbitrary Fock state |n is given by

n (X) =

Hn (X) X 2 /2

e
,
2n n!

1/4

(3.97)

where Hn (X) are the Hermite polynomials,

Hn (X) = (1)n ex

/2

dn x2 /2
e
.
dX n

(3.98)

As we have proven in Ex. 1.76, eigenstates of a Hermitian operator form a basis of the relevant
Hilbert space. Since the proof was made for Hilbert spaces of nite dimensions, one should generally
be careful applying that result to innite-dimensional spaces, such as those studied in this chapter.
For example, all the eigenstates of the innite potential box (Ex. 3.47) have wavefunctions that

76

Quantum Mechanics I

n=3
n=2
n=1
n=0
0

Figure 3.9: Wavefunctions of the rst three energy levels of a harmonic oscillator.
vanish outside the box, so they do not span the space of all possible wavefunctions. On the other
hand, these wavefunctions do span the space of all quantum states that are localized within the
box, i.e. are physically allowed under this potential. Even more complicated is the situation with a
nite well (Ex. 3.69), whose energy eigenstates combine a discrete (for E < V0 ) and continuous (for
E V0 ) spectra.
The situation with the Harmonic oscillator is convenient in this respect because, on the one
hand, there are no quantum states that are not physically allowed and, on the other hand, the entire
energy spectrum is discrete. As a result, the Fock states do form an orthonormal basis in the Hilbert
space. This statement also follows from mathematical properties of the Hermite functions (3.97),
which are known to form a basis in the Hilbert space of all normalizable functions dened on the
real axis.
It is instructive to compare the wavefunctions of the Fock states with those of energy eigenstates
of the nite potential well (shown in Fig. 3.4). In both cases, the wavefunctions exhibit oscillatory
behavior inside the well and exponentially fall o outside. The dierence is that the energy levels
are equidistant for the harmonic oscillator, but not equidistant for the rectangular well. Also, each
eigenwavefunction of the well is dened in a piecewise fashion [see Eqs. (8.19) and (8.28)] while for
the harmonic oscillator potential it is a single elementary function.
Exercise 3.78 For an arbitrary |n, calculate X, X 2 , P , P 2 and verify the uncertainty
principle.
Hint: Rather than integrating the wavefunctions, it is more convenient to employ Eqs. (3.89) and
(3.93).
Note 3.2 The vacuum state is the only minimum-uncertainty Fock state.
Exercise 3.79 Find the evolution of the state |0 + |1; calculate the time dependence of X,
P and plot the trajectory in the phase space.
Answer (for real , ):

X = 2(cos t); P = 2(sin t).

This last result is remarkable. The Fock states themselves are stationary and thus exhibit no
time variation of the mean position and momentum. But in their linear combinations, the mean
position and momentum oscillate similarly to those of a classical harmonic oscillator. Of course,
this behavior is still very dissimilar to classical: the mean values of the position and momentum are
microscopic, and their scale is compatible to that of their own uncertainties. In the next subsection
we will study those states whose behavior under Schrodinger evolution resembles classical much
more signicantly.

3.9.3

Coherent states

The coherent state is the closest quantum approximation of the classical picture of the harmonic
oscillator motion. As we shall see, in this state the mean position and momentum vary as functions of

3.9. HARMONIC OSCILLATOR

77

time in the same way as do those of a classical ball on a spring. The amplitude of this oscillation can
be arbitrarily high, but the position and momentum uncertainties remain as low as in the vacuum
state. Because of their classical-like behavior, coherent states frequently occur in nature, not only
in mechanical oscillators, but also in other incarnations of the harmonic oscillator. For example,
the quantum state of light in a laser pulse is also a coherent state.
Definition 3.4 The coherent state | is an eigenstate of the annihilation operator with eigenvalue
:
a
| = | .
(3.99)
The quantity || is called the amplitude, Arg the phase of the coherent state.
Exercise 3.80 For a coherent state |, show that its wavefunctions in the position and momentum
bases are given by
(XX )2
P X
1
2
(X) = 1/4 ei 2 eiP X e
;
(3.100)

(P ) =
where
X =

1
1/4

ei

P X
2

2Re ;

eiX P e

P =

(P P )2
2

2Im .

(3.101)

(3.102)

The overall phase factors eiP X /2 are included into Eqs. (3.100) and Eq. (3.101) as a matter
of convention in order to make these two equations (which are obtained from one another by means
of a direct or inverse Fourier transform) look similar. We will see shortly that this convention is in
fact pretty useful.
By comparing Eq. (3.100) withEq. (3.38) we observe that the coherent state wavefunction is a
Gaussian wavepacket centered at 2Re. Aside from a linearly varying phase factor and the shift
of the center, the shape of the wavefunction is identical to that of the vacuum state. This is not
surprising given that, according to the above denition, the vacuum state is a coherent state with
zero amplitude. In fact, one can show the following:
Exercise 3.81 Show that the coherent state can be obtained from the vacuum state by applying
displacement operators
|

= eiP X /2 eiP X eiX P |0

= e
Exercise 3.82
| are

iP X /2 iX P iP X

(3.103)

|0 .

Show that the expectation values of the position and momentum in coherent state
X =

2Re ;

P =

2Im .

(3.104)

Show that the uncertainties of these observables equal

1
X 2 = P 2 = .
2

(3.105)

One can picture the coherent state in the

phase space as a circle positioned at points ( 2 Re , 2 Im )
(Fig. 3.10). The radius of the circle, 1/ 2, symbolically represents the standard deviation of the
position and momentum, which are independent of the coherent amplitude9 . In all coherent states,
similarly to the vacuum state, the position-momentum uncertainty product takes the minimum
possible value.
Now let us nd the Schr
odinger evolution of the coherent state in time. To this end, we decompose
the coherent state into energy eigenstates and apply the evolution operator to each of these states.
9 In fact, this circle has more than just a symbolic value. The behavior of uncertainties in the phase space is
described by the so-called Wigner function, which is the analog of the classical phase-space probability density.

78

Quantum Mechanics I
P

2 Ima

Arg a
2 Rea

Figure 3.10: The phase space picture and evolution of the position and momentum observables in
the coherent state.
Exercise 3.83 Show that the coherent state can be written as
| = e||

/2
a a

|0 .

(3.106)

Show that Eq. (3.106) can be simplied as follows:

| = e||

/2
a

|0 .

(3.107)

Hint: use Eq. (3.103) and the Baker-Hausdor-Campbell formula (1.70).

Exercise 3.84 Find the decomposition of the coherent state into the number basis (make sure
your answer is properly normalized).
Answer:
n
2
|n .
| = e|| /2
(3.108)
n!
n
Now the role of the convention for the overall phase in Eqs. (3.100) and (3.100) becomes clear.
With this convention, the decomposition (3.108) of the coherent state in to the Fock basis takes an
extremely simple form.
If one performs an energy measurement on a coherent state, one will nd, in accordance with
Eq. (3.108), that that the probability to project onto a particular Fock state is given by:
prn = |n| | = e||
2

||2n
.
n!

(3.109)

This probability distribution is very well known in mathematical statistics. It is the so-called Poisson
distribution, which expresses the probability of a number of events to occur within a given period
of time if these events occur with a known average rate and independently of each other. Let us
distract ourselves from pure quantum physics for a few minutes and study the properties of this
distribution.
Exercise 3.85 Find n, n2 for a coherent state.
Answer: n = n2 = ||2 .
For example, in a certain city, 25 babies are born per day on average. The exact number of
babies that are born on each day varies: sometimes there are exactly 25, sometimes (for example)

3.9. HARMONIC OSCILLATOR

79

22, and sometimes 32. The probability that a specic number n of babies are born on some specic
2
day is then given
by Eq. (3.109)
with n = || = 25. The mean square uncertainty in the number
of babies equals n2 = n = 5, i.e. on a typical day it is much more likely to see 20 or 30
babies rather than 10 or 40 (Fig. 3.11).

prn

0.15

0.10

0.05

10

20

30

40 n

Figure 3.11: the Poisson distribution with n = 4 (empty circles) and n = 25 (lled circles).

Although the absolute uncertainty of n increases with n, the relative uncertainty n2 / n

decreases. In our example above, the relative uncertainty is 5/25 = 20%. But in a smaller town,
where n = 4, the relative uncertainty is as high as 2/4 = 50%.
Returning to the quantum oscillator, the higher the amplitude of the coherent state, the lower
is the relative uncertainty in the number of excitation quanta within that state. For example, if we
measure the energies of laser pulses (which are in the coherent state), the observed value will vary
from pulse to pulse the phenomenon known as the shot noise. For a higher intensity laser beam,
with a higher average pulse energy, the relative magnitude of the shot noise will decrease. The more
macroscopic the coherent state becomes, the less signicant are its quantum features.
Exercise 3.86 A laser of wavelength = 800 nm emits pulses with a repetition rate f = 100
MHz. The average power of the laser is P = 1 Watt. Calculate
a) the mean energy of each laser pulse;
b) the mean number of photons in each pulse;
c) the relative uncertainty in the energy of the pulses due to the shot noise.
How will the answer in part (c) change if the average laser power is P = 1 W?
Exercise 3.87

Show that | = e||

/2| |2 /2+

In regards to the last exercise, one may ask how come the coherent states associated with different s are not orthogonal. Didnt we prove in Ex. 1.76 that eigenstates of an operator form an
orthonormal basis? The answer is that the statement of Ex. 1.76 only applies to Hermitian operator
and the annihilation operator is not Hermitian. Coherent states do form a spanning set, but they
are not orthogonal.
Exercise 3.88 Do eigenstates of the creation operator exist and if so, what is their decomposition
into the number basis?
Exercise 3.89
given by

Show that the action of the evolution operator exp(iHt/~)

upon the state
is


exp(iHt/~)
| = eit/2 eit .

(3.110)

80

Quantum Mechanics I

This result is remarkable. Neglecting the overall quantum phase factor eit/2 , a coherent
state evolves into another coherent state with the same amplitude, but dierent phase. Note the
distinction between the quantum phase factor outside the ket in Eq. (3.110) (which does not reect
in any measurement results) and the coherent phase factor eit inside the ket, which indicates a
dierent eigenvalue of the annihilation operator and hence a physically dierent quantum state.
The role of the coherent phase is evident from Fig. 3.10. The Hamiltonian evolution of the
coherent state entails linear growth of the phase and hence the circular motion of the expectation
values in the phase space. Assuming real, we nd:

X =
2Re(eit ) = 2 cos t;

P =
2Im(eit ) = 2 sin t.
(3.111)
Aside from a microscopic uncertainty, the behavior of the position and momentum observables in
the coherent state is identical to classical.
Exercise 3.90

Are coherent states stationary?