Você está na página 1de 4

Coherent states of the harmonic oscillator

It turns out to be possible to find normalizable eigenstates of the lowering operator


a, essentially because the spectrum of the number operator a a is bounded from below. (That is, there exists a state | 0 i that is annihilated by a, i.e., a state such that
a | 0 i = 0.) These state are called coherent states,10 and they have an enormous
number of very interesting and important properties. These states and their various generalisations play a pivotal role in all of quantum optics, among other areas.11
1. Let |i denote an eigenstate of a with eigenvalue , i.e., a |i = |i. The
possible values of will be determined shortly.
P
(a) Expand |i in the Fock basis, i.e., let |i = n=0 cn | n i. Substitute this in
the eigenvalue equation for |i, and determine the coefficients cn recursively,
by equating the coefficients of each individual Fock state on either side of
the equation. Impose the normalization condition h |i = 1, to arrive at
the following expression for |i (after choosing the overall phase factor to be
unity):

n
2 X
1
| n i.
|i = e 2 ||
n!
n=0
Note that no condition whatsoever has been placed upon the eigenvalue . Since
we are able to find a normalizable eigenstate |i for any arbitrary complex value of
the eigenvalue , we must conclude that the eigenvalue spectrum of the lowering
operator a is, in fact, the whole of the finite part of the complex plane. Recall
that a is non-self-adjoint. Its eigenvalues can therefore be complex numbers, in
general. We now find that its eigenvalue spectrum is actually doubly continuous
(both 1 Re and 2 Im can take on values in (, ).)
(b) Show that a similar argument does not work for the raising operator a : that
is, there can be no normalizable eigenstate of a , in stark contrast to the case
of a. Note the reason why: there is no state | n i such that a | n i = 0, i.e.,
there is no upper bound to the eigenvalue spectrum of a a.
You can verify the foregoing statements in an alternative (but equivalent) way, by
working in the position representation. Let (x) h x |i be the position-basis
wave function corresponding to the CS |i. The eigenvalue
equation then gives

h x | a |i = h x |i = (x). Put a = (x + ip)/ 2, and remember that p is


represented by id/dx in the position basis. This leads to a first-order differential
equation for (x),

d
+ (x 2 ) = 0.
dx
The solution is a shifted Gaussian in x, apart from a phase factor. It is clearly normalizable. In contrast, if (x) is the wave function corresponding to an eigenvalue
of the operator a , the differential equation satisfied by it is

d
(x 2 ) = 0.
dx
The solution is obviously not normalizable in (, ).
10 I

will use the abbreviation CS for coherent state.


particular, the state of the radiation field in an ideal single mode laser is a CS. We will not
be concerned here with this aspect of coherent states.
11 In

(c) Show that the normalized position-space wave function corresponding to the
CS |i is given by

2 1 )2

(x) = 1/4 e 2 (x

2 x)

ei2 (1

(d) Hence show that the momentum-space wave function corresponding to the CS
|i is given by

2 2 )2 /2

e (p) = 1/4 e(p

ei1 (

2 p2 )

The displacement operator: Using the fact that (a )n | 0 i =


be re-expressed as
2

1
|i = e 2 || e a | 0 i.

n! | n i, |i can

Now, since a | 0 i = 0, we have an | 0 i = 0 for all positive integer values of n. Hence

e a | 0 i = | 0 i. Therefore |i may be further re-written as

|i = e 2 || e a e

| 0 i.

Using the BCH formula, we find


1

e 2 || e a e

= e a

a def.

= D().

The operator D() is called the displacement operator, for a reason which will
become clear shortly.12 In terms of this operator, the CS |i is simply
|i = D() | 0 i.

But the adjoint of D is D = e( a a) . Since the operator a a commutes


with itself, it follows at once that D D = D D = I, i.e., that D() is a unitary
operator. The CS |i is therefore nothing but a unitarily transformed vacuum
state.13
This helps us understand why the position-space wave function (x) and the
momentum-space wave function e (p) of a CS are also Gaussians, apart from a
phase factor in each case. The significance of the parameters 1 and 2 also becomes
clear: the peak of the Gaussian
is displaced from 0 in the vacuum state to

2 1 in position space, and to 2 2 in momentum space.


2. Coherent states are minimum uncertainty states: Since |i is a unitarily
transformed vacuum state, you might expect certain properties of | 0 i to be carried
over to all coherent states.
(a) In the CS |i, compute the expectation values of
x=

a + a
a a
, p = , x2 and p2 .
2
i 2

12 We should write D(, ) because the operator D is parametrized by as well as (or the
two independent variables 1 and 2 ). But it is customary to write just D(), in a slight abuse
of notation.
13 The vacuum state itself is also a coherent state, of course, corresponding to = 0.


(b) Hence show that (in the units we have chosen) x = p = 1/ 2 in this state,
independent of . Therefore the uncertainty product is (x)(p) = 12 in any
CS, i.e., all of them are minimum uncertainty states.
3. Non-orthogonality of CS: The states {|i, C} are normalized to unity,
but they are not mutually orthogonal.
(a) Show that, if and are any two distinct complex numbers, then
1

h| i = e 2 (||

+||2 )+

(b) Hence show that


2

|h| i|2 = e|| .


The overlap |h| i|2 therefore decreases quite rapidly as the distance ||
between the two states increases, but it is not identically equal to zero.
Over-completeness of CS: It can be shown that
Z
d2 |ih| = I,
R
where I is the unit operator and d2 stands for an integral over the whole of the
complex plane, i. e., a double integral over all values of 1 and 2 in the ranges
1 (, ) and 2 (, ). The fact that the right-hand side has a factor
(> 1) implies that the set of CS is an over-complete set of states.
It is of interest to ask whether there exist subsets of the set of CS that are
complete sets. The determination of such sets is a nontrivial task. It is known, for
instance, that the set of states for which lies on the unit circle in the complex
plane form a complete set; so does the set of states for which = m + ni, where m
and n are integers; and so on.
4. The mean and variance of a a in a CS: The expectation value of the
number operator a a in the Fock state | n i is of course h n | a a | n i = n, and
hence h 0 | a a | 0 i = 0. In the case of the CS |i, we have a |i = |i and hence
h| a = h|. It follows immediately that the expectation value of the number
operator in the CS |i is
h| a a |i = ||2 .
Show that the variance of the number operator a a in the CS |i is also equal to
||2 . Hence the standard deviation of this quantity, i.e., the uncertainty in a a, is
equal to || in the state |i.
It turns out that all the higher cumulants of the number operator are also equal to
||2 in the CS |i. In the context of radiation, this is a consequence of the fact that
the photon number in ideal, single-mode laser light has a Poisson distribution.
5. Use Hadamards lemma to show that
D() a D () = a

and D() a D () = a .

Again, these relations tell us why D() is termed the displacement operator. Under
a unitary transformation by D(),
D()

D()

the state | 0 i |i, while the operator a D() a D ().


3

Hence the relation a | 0 i = 0 is transformed to


D() a | 0 i = 0, or D() a D ()D() | 0 i = (a ) |i = 0.
Viewed thus, the defining eigenvalue equation for a CS is nothing but a unitarily
transformed or displaced version of a | 0 i = 0.
D() as an element of a Lie group: We have mentioned already that the operators a, a and I are the generators of a Lie algebra, the Heisenberg algebra.
Exponentiating a linear combination of these generators will then yield the general
element of the corresponding Lie group. It is in this sense that D() is an element
of the Heisenberg group.14 The question of interest, then, is the group multiplication law.
6. Let and be any two complex numbers. Show that the displacement operator
satisfies the following group composition rule:
1

D() D() = D( + ) e 2 (

= D( + ) ei Im ( ) .

Generalizations: Various generalizations of coherent states are possible. This


topic has been studied in some detail, especially in the context of quantum optics. We have the generalized coherent state | n, i obtained by applying the
displacement operator D() to the Fock state | n i (rather than the vacuum state
| 0 i) and normalizing the result; the photon-added coherent state | , n i obtained by operating on |i with a n and normalizing the result; and so on. Other
transformations of the vacuum state are also of relevance. For instance,

14 It involves two real parameters and , in accord with the two nontrivial generators a and
1
2
a of the Lie algebra.