You are on page 1of 12

The operator method for solving

the harmonic oscillator


in quantum mechanics
Anders Blom
Division of Solid State Theory
Department of Physics, Lund University
Solvegatan 14 A, S223 62 Lund, Sweden
Anders.Blom@teorfys.lu.se
March 19, 2003
1 Introduction
This paper is a comprehensive review of the operator method formalism for the
linear, one-dimensional harmonic oscillator in quantum mechanics. Any basic level
textbook on quantum mechanics contains similar discussions, and we do not claim
to bring forth any new material here; it is merely a compilation, with comments,
of the material found in books such as J. J. Sakuari, Modern Quantum Mechanics
(Addison-Wesley, New York, 1994).
The harmonic oscillator is a problem of fundamental importance in quantum
mechanics
1
. Apart from being one of the few exactly solvable quantum mechanical
problems, its physical relevance reaches far beyond the most obvious interpretation
of the oscillator as an analogue of the classical spring force problem. Any potential,
of arbitrarily complicated form, which possesses a minimum or equilibrium can to
lowest non-trivial order be treated as a harmonic oscillator. Higher order terms in
the expansion of the physical potential can then be added as perturbations.
1
Of course the harmonic oscillator is a very fundamental problem also in classical mechanics;
see e.g. the extensive discussions of the classical oscillator in H. Goldstein, Classical Mechanics,
2nd Ed. (Addison-Wesley, New York, 1980).
1
To solve the Schrodinger wave equation with the harmonic oscillator potential,
and thereby obtain the wave function and the eigenenergies, is by no means ex-
tremely complicated. It does however involve some dierentiating and integrating,
and one needs to master the properties of Hermite polynomials, in order to make
order of things. In contrast, the operator method oers an almost eortless road to
obtaining the eigenstates and their properties, and is also rather straightforward to
apply to perturbation theory. In general, the operator formalism oers many advan-
tages over the wave picture, by focusing on the physical properties of the quantum
mechanical states, instead of any particular representation of them, as the wave
function is
2
.
2 Transforming the Hamiltonian
The potential energy of the harmonic oscillator is often written as
1
2
kx
2
, where k is
the force constant. However, since there are many occasions where there is no real
force acting, but yet the relevant potential is that of an oscillator, we shall prefer
to express things in the mass m of the particle and the oscillator frequency . Thus
the Hamiltonian we set out to diagonalize is
H =
p
2
2m
+
1
2
m
2
x
2
, (2.1)
where p and x are the momentum and position operators, respectively.
The usefulness of the operator method lies in the denition of two new, dimen-
sionless operators,
a =
_
m
2
x + i
_
1
2m
p, (2.2)
a

=
_
m
2
x i
_
1
2m
p. (2.3)
Note that these two operators are non-Hermitian, since obviously (a

= a. By
using the canonical commutation relation [x, p] = i, it can immediately be shown
from the denitions above that
_
a, a

= 1 (2.4)
2
See the rst chapters of Sakuarais book for a most enlightening discussion, in which the wave
function is not given any particular status, as is the case in Schrodingers wave mechanics, but
instead is shown to be just the representation of the state in the basis of the position eigenstates.
2
which, as will become apparent during the following discussion, is the property
which denes almost all the important physical properties of the harmonic oscillator.
It is generally the case that the commutation relationships between the operators
determine the physics or perhaps one should say that it is the opposite which is
true. Nevertheless, once the commutators are known, the other properties of the
eigenstates and eigenvalues usually follow, almost automatically.
The question one may ask at this point is, how do we know that a (and a

),
dened above, will help us solve the problem? Of course, as we will see, this is
in fact so, but how does one arrive at these denitions? Without claiming to be
rigorous or complete, let us consider how one can get at least a reasonable idea for
a new operator that may be helpful.
The rst observation to make is that the Hamiltonian (2.1) contains a sum of
squares, and our general experience tells us that if we can complete the square, we
may be in better shape to solve the problem. Let us therefore dene a new operator
b =
x
x + i
p
p,
since b
2
then will contain both x
2
and p
2
, as desired. Complex numbers arise more
naturally in quantum mechanics than in classical mechanics, and we know that the
two wave functions and

must be treated as separate quantities, for instance


when we deal with probability currents. Thus the Hermitian conjugate of this op-
erator,
b

=
x
x i
p
p,
is yet another, independent operator. We have now actually made a canonical
transformation from p and x to b and b

without losing anything, meaning that


the volume of the phase space remains intact.
If we the relationships,
x = (b + b

)/2
x
, (2.5)
p = (b b

)/2i
p
. (2.6)
and insert this into the Hamiltonian (2.1), we get after some re-arrangements
H =
1
8
_
bb

+ b

b
_
_
1
m
2
p
+
m
2

2
x
_

1
8
_
b
2
+ b

2
_
_
1
m
2
p

m
2

2
x
_
.
Thus, if we take

x
= m
p
(2.7)
the Hamiltonian simplies to
H =
1
4m
2
p
_
bb

+ b

b
_
. (2.8)
3
Now we evaluate the commutator
_
b, b

= i
x

p
[x, p] + i
x

p
[p, x] = 2m
2
p
.
To keep things as simple as possible (and to make b and b

dimensionless) we let

p
=
_
1
2m
, (2.9)
which gives [b, b

] = 1. At this point it is clear that the operators b and b

are
identical to a and a

above, and in result we have also shown that the Hamiltonian


can be written
H =
_
a

a +
1
2
_
, (2.10)
where we used the commutator (2.4) and Eq. (2.9) to come from Eq. (2.8) to the
end result.
3 Determining the eigenenergies;
the number operator
Apart from additive and multiplicative constants, the Hamiltonian (2.10) depends
only on the operator a

a, which turns out to be convenient to bestow with a name


and a symbol. We therefore dene the number operator
N a

a. (3.1)
The reason for this name will soon be obvious. The number operator N commutes
with the Hamiltonian H, and they therefore have a complete, common set of eigen-
states. Thus the eigenvalues of N, which we denote n, are good quantum numbers
for the harmonic oscillator, and we will use them to label the eigenstates n, which
fulll
N n = n n , (3.2)
H n = E
n
n , (3.3)
where E
n
are the energy eigenvalues. The basis spanned by the energy eigenstates
n is denoted the energy representation.
We can now immediately write down an expression for E
n
; since H =
_
N +
1
2
_
we must have
E
n
= (n + 1/2). (3.4)
4
Note, that so far we have not in any way determined the eigenvalues n of the number
operators; this is however our next immediate task.
As we already have said, all the physics lies in the commutation relations, and
we therefore consider
[N, a] = [a

, a]a = a. (3.5)
We here used the rule [AB, C] = A[B, C] + [A, C]B and the trivial fact that a
commutes with itself.
Let us act with the commutator [N, a] on an eigenstate n of the number oper-
ator, using Eq. (3.5) on the one hand, and the basic denition of the commutator
on the other:
[N, a] n =
_
a n
Na n aN n = (N n)a n
Thus we see that
Na n = (n 1)a n ,
which means that a n is an eigenstate of N with eigenvalue n 1. The operator a
is therefore known as the lowering or destruction (annihilation) operator.
In an equivalent way one shows that [N, a

] = a

and that the application of a

to a state n yields a state with eigenvalue n + 1; thus a

is the raising or creation


operator. Similar stepping or ladder operators appear in the quantum mechanical
treatment of angular momentum, where they are equally useful for obtaining the
eigenstates and eigenvalues as in our present discussion.
Still we have no idea what the eigenvalues n are, but we do know that applying
the operator a to an eigenstate n of N steps down the eigenvalue by an amount
1. If we then apply a again, we get an eigenstate with the eigenvalue n 2, and
so on. One may ask, if we can continue to apply a forever, producing states with
eigenvalues which eventually will become very large negative numbers? The answer
is no, and the reason is that the norm of the eigenstates must be positive.
Consider the state corresponding to the eigenvalue n1. Its norm can be written
as an inner product of the vector n 1 in ket-space and its dual correspondent in
bra-space, the vector n 1 . The norm of all eigenstates of N must be positive, as
for all quantum mechanical eigenstates. Thus
0 n 1 n 1 =
_
n 1
_

_
n 1
_
=
_
n a

_
a n
_
=

n a

a n
_
= n N n = nn n ,
(3.6)
which shows that n 0 (as long as the state n has a positive norm, which we
require).
5
Hence there must exist an eigenstate with an eigenvalue n
min
such that when we
apply a we do no longer step down the eigenvalue, but instead we hit the vacuum :
a n
min
= . (3.7)
We can determine this lowest possible eigenvalue by acting with N on the state
n
min
,
N n
min
= a

a n
min
= 0.
Thus n
min
= 0, which means that the eigenvalues n of the number operator are
non-negative integers, since all other eigenstates can be reached from the ground
state 0 by successive application of a

:
n (a

)
n
0 .
Assuming that the ground state is normalized, we can also determine the constant
of proportionality, by using Eq. (3.6) which states that n 1 n 1 = nn n.
From this, it immediately follows that
n =
1

n!
(a

)
n
0 . (3.8)
Note that there is no upper bound on n; any non-negative integer is an eigenvalue
of N.
We have already related the energy eigenvalues to those of the number operator
in Eq. (3.4), so once the values of n are known, we also obtain
E
n
= (n + 1/2), n = 0, 1, 2, . . . , (3.9)
and thus the ground state energy is E
0
= /2. The fact that this is non-zero, as
opposed to the classical case, is sometimes referred to as quantum mechanical zero-
point vibration. It is a direct consequence of Heisenbergs uncertainty principle; we
cannot simultaneously specify the position (exactly at the equilibrium position) and
momentum (exactly zero) of the particle with innite precision, as is required if the
energy is exactly zero.
4 Various properties of the eigenstates;
matrix elements and real-space wavefunctions
Not only the energy eigenvalues can be obtained with the operator formalism. Any
problem involving the harmonic oscillator can be considered using this approach, and
for such purposes it is often useful to know the matrix elements of various operators
6
with the energy eigenstates n. The matrix elements of the two operators H and
N are trivial, since they are diagonal in the energy representation:
n H n

= E
n

nn
, (4.1)
n N n

= n
nn
. (4.2)
Eigenstates belonging to dierent eigenvalues are naturally orthogonal,
n n

=
nn
. (4.3)
From this follows readily that, since n 1 a n, we must have
n a n


n,n

1
.
The proportionality constant can be found from considering
a n = a
1

n!
(a

)
n
0 =
1

n!
aa

(a

)
n1
0 =
1

n!
(1 + a

a)(a

)
n1
0
=
1

n
(1 + N)
_
1
_
(n 1)!
(a

)
n1
0
_
where we used Eq. (3.8), the commutator (2.4) and identied the number operator
N = a

a. Now, from Eq. (3.8), the quantity in square brackets is nothing but the
state n 1, and thus
a n =

n n 1
which gives the matrix element
n a n

n
n,n

1
. (4.4)
In a similar way, using
a

n =
1

n!
a

(a

)
n
0 =

n + 1
(n + 1)!
(a

)
n+1
0 =

n + 1 n + 1
we obtain

n a

_
=

n + 1
n,n

+1
. (4.5)
One can also, using Eqs. (2.5) and (2.6), easily write down the matrix element
for the position and momentum operators. The details are left as an exercise, but
the resulting matrix representation of x becomes
_

2m
_
_
_
_
_
_
_
0

1 0 0 0

1 0

2 0 0
0

2 0

3 0
0 0

3 0

4
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
_
_
_
_
_
_
_
. (4.6)
7
Of course neither a, a

, x nor p are diagonal in the energy representation, since they


do not commute with the Hamiltonian H (or the number operator N).
As a further exercise it is instructive to consider the operators x x x
and p p p, where denotes expectation values. Taking the expectation
values of these operators with the energy eigenstates, one nds that

(x)
2
_
(p)
2
_
= (n + 1/2)
2

2
(4.7)
which shows that all states fulll the uncertainty relation x p /2 (here
A
_
A
2
) and the ground state satises the equal sign (minimum uncertainty
product). The latter is to be expected, since in the x-representation the ground
state wave function is a Gaussian.
The wave functions

n
(x) x n (4.8)
of the energy eigenstates in the x-representation can be derived from Eq. (3.7). If
we substitute in the denition of a, Eq. (2.2), the position and momentum operators
x and i/x in the x-representation, we obtain
a
x
=
1

2m
_
mx +

x
_
, (4.9)
where the index x on the operator indicates the x-representation. Things become a
bit clearer if we dene a dimensionless quantity

_
m

x, (4.10)
whence
a
x
=
1

2
_
+

_
. (4.11)
Multiplying Eq. (3.7) from the left with x we now get x a 0 = 0, or in the
x-representation
1

2
_
+

0
() = 0. (4.12)
This dierential equation is simple to solve, and we obtain the normalized wave
function

0
() =
_
m

_
1/4
e

2
/2
. (4.13)
Thus the ground state is indeed a Gaussian, as we already commented on above.
We can now obtain the wave functions
n
of all other eigenstates n by apply-
ing the operator (a

)
n
to the ground state. Keeping in mind that the Hermitian
conjugate of d/dx is d/dx, we have
a

x
=
1

2
_

_
, (4.14)
8
and thus the normalized eigenfunctions are

n
() =
_
x
(a

)
n

n!
0
_
=
(m/)
1/4

2
n
n!
_

_
n
e

2
/2
. (4.15)
This can be written as

n
() =
(m/)
1/4

2
n
n!
H
n
()e

2
/2
, (4.16)
where H
n
(x) is the n-th Hermite polynomial.
5 Bosons and fermions
There are many aspects of the harmonic oscillator which are relevant and interesting.
One may wish to consider the time evolution of the eigenstates or mixtures of them,
and applications in perturbation theory are particularly useful for the treatment of
anharmonic eects, for instance. We will however leave it to the interested reader
to further pursue these paths, and nish o this presentation by instead consid-
ering the connection between statistics and the creation and annihilation operator
commutation relations.
An important application of the harmonic oscillator potential is in the treatment
of vibrational modes in crystals. In reality, the vibration of the atoms may be
a very complicated superposition of many frequencies. For most purposes it is
however sucient to consider the separate modes, each with a xed frequency which
is related by a dispersion relation to the wave vector describing the propagation of
the vibration in the crystal.
For each mode, if m is the mass of the vibrating atom (consider a mono-atomic
crystal for simplicity), and the mode frequency, the system is described, to lowest
order, by the Hamiltonian (2.1), which we now know how to solve. We thus obtain
a series of eigenenergies E
n
(), separated from each other by a constant interval
(see Eq. (3.4)), the lowest one being E
0
() = /2. In order to separate the
dierent modes, the eigenstates, energies and operator must also be labeled by the
frequency, as n

and a

etc.
By studying the corresponding real-space wave functions derived in the previous
section, one nds that the higher the eigenvalue n is, the larger is the oscillation
amplitude, measured by (x)
2
. Since the oscillation frequency however remains
unchanged, the particle velocity must become larger, and thus both the kinetic
and potential energy of the oscillator increase as we go to higher eigenstates. The
quantum mechanical harmonic oscillator in fact fullls the same virial theorem as
the classical oscillator does,
_
p
2
2m
_
=
_
m
2
x
2
2
_
=
H
2
. (5.1)
9
The n-th eigenstate of each mode has an energy
E
n
() = E
0
() + n. (5.2)
Using this relation, we now introduce the concept of phonons, by stating that the
excitation of the oscillator from state n

to n + 1

corresponds to the creation of


a phonon with frequency (or energy ). This gives a clear interpretation of the
operator a

(a

), as the creation (annihilation) operator of a phonon of frequency


. Note that it is crucial to separate the eigenstates of the harmonic oscillator from
the phonons. An eigenstate n

of the oscillator contains a number of phonons,


but these phonons are all in the same state .
Phonons are not particles, but quasi-particles, meaning that they corresponds to
elementary excitation of the system. By adding energy, e.g. in the form of increased
temperature, to the crystal, we may introduce new modes into the vibration (i.e.
create phonons of other frequencies), and we may also excite each mode into a higher
eigenstate, i.e. create more phonons of the same frequency. In the eigenstate n

the system contains n phonons with energy . There is no upper boundary for
the quantum number n, and so there is no limit how many such equivalent phonons
we may create. It is therefore clear that phonons are bosons, since we can place an
innite number of them in the same state.
Let us now turn to a completely dierent problem, and imagine that we wish to
place a number of electrons in an innite square quantum well
3
. It will be assumed
that the electrons do not interact with each other. We know, since the electrons
are fermions, that we can only place two of them (one of each spin projection) in
each well eigenstate n. We shall however for simplicity ignore spin, and thus each
eigenstate can hold only a single electron.
We now dene the operator c

n
(c
n
) to be a creation (annihilation) operator,
which places (removes) an electron in (from) the well eigenstate n. Note that we
are not relating the operators c
n
and c

n
to physical observables such as x and p,
like we did for the harmonic oscillator. In fact, there is really no need to do that in
the oscillator case either, except when we wish to switch to the x-representation, as
long as we dene the Hamiltonian in terms of the number operator, as in Eq. (2.10).
What should the properties of our new creation and annihilation operators be?
First of all, as in the case of bosons, we will need a number operator F
n
which
counts how many fermions are in the well eigenstate n. Since we are dealing
3
We could, in principle, instead consider a parabolic potential; a physical example of this situa-
tion is when a magnetic eld is applied to a crystal. The corresponding eigenstates are known
as Landau levels. It may however be somewhat confusing to talk about electrons in a harmonic
oscillator, after we just nished talking about phonons. It is nevertheless very useful for the
ambitious reader to repeat the following discussion with Landau levels instead of quantum well
levels, to clarify the dierence between the operators relating to the conning potential and its
eigenstates on the one hand, and the lling of the levels with fermions (or bosons) on the other.
10
with fermions, we already know that there can be only two eigenvalues of this
(unspecied) operator, namely 0 and 1. Keep in mind that the eigenstates f
n
of
F
n
(i.e. 0 and 1) are not to be confused with the eigenstates of the quantum well,
in the same way the harmonic oscillator eigenstates are n

and not , which are


the phonon eigenstates. In each harmonic oscillator state n

we have n phonons
of frequency ; in each eigenstate f
n
of F
n
we have f electrons in the n-th energy
level of the quantum well.
Note also that we can choose to study the occupation of each individual quan-
tum well level, in the same way as we initially considered only a single mode of
phonons, which a xed frequency. From now on we therefore omit the label n on
the eigenstates f and the operators.
We cannot have a negative number of particles in an eigenstate, or more than
one fermion in the same state, and hence
c 0 = , c

1 = . (5.3)
Here denotes the vacuum. Furthermore, the denitions of c and c

as creation and
annihilation operators immediately imply
c 1 = 0 , c

0 = 1 . (5.4)
One may wonder if c 1 is normalized (if 1 is). We can consider using the same
matrix elements as for bosons, in which case indeed 0 c 1 = 1 (see Eq. (4.4)). On
the other hand,
0 0 =
_
1 c
_

_
c

1
_
=

1 c

c 1
_
,
and since we will soon dene c

c to be the number operator F, giving

1 c

c 1
_
=
1 1 1, we may simply take (5.4) as denitions which imply a certain normalization
condition that will be reected in the eigenvalues of the operator c

c.
Taking the above expressions together, we nd that
c
2
= 0 (5.5)
in the Hilbert space spanned by the two eigenstates of F. So far we have not specied
the number operator, but in analogy with the bosonic case, it seems reasonable to
dene
F c

c, (5.6)
which clearly is Hermitian. We must now check that we indeed obtain F f = f f,
f = 0, 1, but this follows immediately from the normalization implied in Eqs. (5.4):
F 1 = c

c 1 = c

0 = 1 ,
F 0 = c

c 0 = c

= 0 0 .
11
(In the last line we used the fact that the number 0 acing on any ket produces the
vacuum.)
We will nally derive the commutation relation between c and c

. Since cc

1 =
0 and c

c 0 = 0 we have
0 = c 1 = cc

0 = (cc

+ c

c) 0 ,
1 = c

0 = c

c 1 = (c

c + cc

) 1 ,
which proves that in the Hilbert space spanned by the two eigenstates of F, we have
c

c + cc

= 1. (5.7)
Dening the anti-commutator [A, B]
+
AB + BA (the anti-commutator is some-
times also denoted {A, B}), we can write
1 = [b, b

=
_
_
_
bb

+ b

b, (fermions),
bb

b, (bosons).
(5.8)
The use of creation and annihilation operators is an extremely powerful tool in
quantum eld theory, and is based exactly on the simple relationships (5.8). Instead
of imposing the very dierent properties of BoseEinstein or FermiDirac statistics
on all aspects of the calculations, one can use a completely unied approach, based
only on these fundamental commutation relations, which merely introduce a few
sign changes here and there. Of course, the physics exhibited by particles obeying
either statistics is still radically dierent, as it should be.
Note that it is not customary to derive the commutation relations as we have
done here; instead Eq. (5.8) is postulated (along with Eq. (5.5) for bosons). From
these postulates one can then show that the eigenvalues of c

c can only be 0 and


1 for fermions (and any non-negative integer for bosons), and furthermore derive
Eqs. (5.3) and (5.4) (for fermions).
12

You might also like