Sie sind auf Seite 1von 7

Quantum Mechanics: bits and pieces 3

Notes by Sergei Winitzki


DRAFT March 5, 2007
Contents
1 Quantization of a harmonic oscillator 1
1.1 Classical description . . . . . . . . . . . . . . 1
1.2 Quantization in Heisenberg picture . . . . . . 1
1.3 The measurement postulate . . . . . . . . . . 2
1.4 Quantum evolution . . . . . . . . . . . . . . 2
1.4.1 Evolution operator . . . . . . . . . . . 2
1.4.2 Schrdinger picture . . . . . . . . . . 3
1.4.3 Operators of the type exp(
x
) . . . . 3
1.5 Energy spectrum . . . . . . . . . . . . . . . . 4
1.5.1 Creation and annihilation operators . 4
1.5.2 Energy eigenstates . . . . . . . . . . . 4
1.5.3 Selecting the Hilbert space . . . . . . . 4
1.5.4 Coordinate representation . . . . . . . 4
1.5.5 *Nonexistence of eigenvectors of a

. 5
2 *Coherent states 5
2.1 Basic properties . . . . . . . . . . . . . . . . . 5
2.1.1 Occupation number representation . . 5
2.1.2 Approximate orthogonality . . . . . . 6
2.1.3 Coordinate representation . . . . . . . 6
2.1.4 Decomposition of unity . . . . . . . . 6
2.2 Holomorphic representation . . . . . . . . . . 6
2.2.1 Representation of states . . . . . . . . 6
2.2.2 Representation of operators . . . . . . 6
2.3 Evolution of coherent states . . . . . . . . . . 7
2.3.1 Quasi-classical behavior . . . . . . . . 7
2.3.2 Excitations of coherent states . . . . . 7
1 Quantization of a harmonic oscilla-
tor
Quantization is a procedure whereby a physicist creates a
quantum theory of some physical system out of the clas-
sical theory of that system. In this section we shall briey
outline the procedure of quantization in general and con-
sider a simple harmonic oscillator as a specic example.
1.1 Classical description
The classical theory of a harmonic oscillator with mass m
and frequency is based on the Lagrangian
L(q, q) =
1
2
m q
2

1
2
m
2
q
2
, (1)
which yields the equation of motion for the trajectory q(t),
q =
2
q. (2)
This equation predicts the position and the velocity of the
oscillator at any time t if initial conditions q(t
0
), q(t
0
) are
known.
In the Hamiltonian formalism, one computes the canon-
ical momentum p = m q and the Hamiltonian of the har-
monic oscillator,
H (q, p) =
p
2
2m
+
1
2
m
2
q
2
. (3)
This yields the equations of motion
dp
dt
=
H
q
= m
2
q, (4)
dq
dt
=
H
p
=
p
m
. (5)
These equations are of course equivalent to Eq. (2).
1.2 Quantization in Heisenberg picture
The quantum-mechanical description of the harmonic os-
cillator in the Heisenberg picture is based on the Hamilton
equations (4)-(5). Instead of the classical position q(t) and
momentum p(t), one introduces the position operator q(t)
and the momentum operator p(t). It is postulated that the
operators p, q satisfy the following commutation relation at
equal times,
[ q(t), p(t)] = i

1. (6)
These operators evolve according to the Heisenberg equa-
tions
d p(t)
dt
=
1
i
_
p,

H
_
= m
2
q, (7)
d q(t)
dt
=
1
i
_
q,

H
_
=
1
m
p, (8)
where

H = H ( q, p) =
1
2m
p
2
+
1
2
m
2
q
2
(9)
is the Hamiltonian operator (i.e. the energy operator).
The operators p, q,

H act in a Hilbert space Hof quantum
states of the oscillator. A quantum state is thus a (time-
independent) vector [) H. The precise choice of the
space H will be made below after we analyze the Hamil-
tonian

H.
The construction described here, namely the replace-
ment of the classical variables p, q by operators p, q in some
Hilbert space H, is called canonical quantization of a clas-
sical Hamiltonian system. In principle, this procedure can
be applied to any theory for which a Hamiltonian descrip-
tion exists. The choice of a Hilbert space is an important
step of the quantization procedure and must be made from
physical considerations, i.e., to make the theory conform to
the experimental data.
A standard example of operators p, q and a Hilbert space
is given by the operators q = x, p = i/x, acting on
functions (x), where x is a real variable representing the
coordinate. This representation of quantum operators in
the space of functions (x) is called the coordinate repre-
sentation. The functions (x) representing quantum states
are called wave functions. However, the coordinate rep-
resentation is only one of innitely many possible repre-
sentations. It is not convenient to think of quantum states
always as wave functions. Calculations are often simpler
with more abstract states [) and operators p, q.
1
1.3 The measurement postulate
This subsection applies to general quantum systems and is
not specic to the harmonic oscillator.
In quantum mechanics, all physical observables (such as
the momentum p or the coordinate q) are represented by
operators ( p, q, etc.). The interpretation of a quantum state
[) is provided by the measurement postulate: If the sys-
tem is in the state [), the probability for measuring a value
a of an observable quantity represented by an operator

A is
Prob
_

A a; [)
_
=
[a[)[
2
a[a) [)
, (10)
where [a) is an eigenvector of the operator

Awith the eigen-
value a. (The symbolic notation

A a means a mea-
surement of the quantity represented by the operator

A
yielded the value a. For now we assume that [a) is a non-
degenerate eigenvector of

A with eigenvalue a.) If a mea-
surement yields a value a
0
, the state of the system after the
measurement is the eigenvector [a
0
). In quantum mechan-
ics, eigenvectors are frequently also called eigenstates.
Remarks:
1. The quantity

Amight be the position q, the momentum
p, the energy

H, or some other operator-valued func-
tion of p and q; such operators

A are called dynamical
observables.
2. All measurements of a quantity A yield only eigenval-
ues of the corresponding operator

A. Normally, dy-
namical observables are represented by Hermitian op-
erators having only real eigenvalues.
3. The number dened by Eq. (10) is always between 0
and 1 (this follows from the Cauchy-Schwartz inequal-
ity).
4. Usually the state vector [) and the eigenvectors [a)
are normalized so that a[a) = [) = 1. In that case,
Eq. (10) is simplied to just [a[)[
2
. But in principle
the formalism of quantum theory does not require that
all vectors [) be normalized to [) = 1.
5. If the operator

A has a continuous spectrum, then usu-
ally the eigenvectors [a) are generalized vectors and
a[a) is undened. In that case Eq. (10) cannot be ap-
plied directly; indeed, the probability to measure one
particular value a in a continuous spectrum is zero.
A realistic device can measure a continuous quantity
(such as the position q) only to a certain nite accu-
racy. Therefore, a physically motivated question is
to nd the probability density for measuring

A within
an interval [a, a + a]; we may denote this event by

A [a, a + a]. If the generalized eigenvectors [a) are


normalized to the delta function,
a
1
[a
2
) = (a
1
a
2
) , (11)
the probability density is
lim
a0
Prob
_

A [a, a + a] ; [)
_
a
=
[a[)[
2
[)
. (12)
6. The state [) is always normalizable, [) < . A
physical system cannot be prepared in a generalized
eigenstate such as [q) or [p).
Some direct consequences of the measurement postulate:
1. If the systemis in an eigenstate [a), then the probability
for measuring

A a is equal to 1 and the probabilities
for all other values is equal to zero.
2. If the state of the system is not an eigenstate of

A, then
the theory does not predict the measured value of A
with certainty but only gives the probability for various
outcomes. This is in a stark contrast with classical the-
ories which always predict the results of all measure-
ments with certainty.
3. Experimentally, one should test the theory by prepar-
ing many identical systems in identical states [) (this
is called preparing an ensemble) and by measuring
the same quantity A for all those systems. This allows
one to determine the probability distribution of mea-
sured values a and to compare that distribution with
the prediction of quantum theory. The theory uniquely
predicts the probability distribution for measured values
(even though there is no denite prediction for the re-
sult of an individual measurement).
4. Since the operator

A might be a function of time such
as q (t), the probabilities for measurement outcomes
may depend on time. However, the state [) is con-
sidered to be time-independent. (This is true in the
Heisenberg picture, whereas the Schrdinger picture,
explained below in Sec. 1.4.2, involves time-dependent
states.)
5. If the eigenstates of

A form an orthonormal basis, the
expectation value of A in a normalized state [) is
found as
A)
|

a
a Prob
_

A a; [)
_
=

a
[a) a a[)
=

a
[

A[a) a[) = [

A[) . (13)
Note that for a Hermitian operator

A, the expectation
value is always a real number.
1.4 Quantum evolution
This subsection applies to general quantum systems and is
not specic to the harmonic oscillator.
1.4.1 Evolution operator
The evolution of the operators p and q is described by the
Heisenberg equations. It is easy to prove that if A(p, q) is
any function, then the time dependence of the operator

A
A( p, q) is described by the equation
d

A
dt
=
1
i
_

A,

H
_
. (14)
(To prove this, we may rst consider

A = p
m
q
n
for integer
m, n and then represent

A as a power series in p and q.)
Equation (14) suggests that all dynamical observables

A(t) evolve in a similar way. This similarity is made for-


mally clear by introducing the evolution operator

U(t, t
0
)
such that for any dynamical observable

A we have

A(t) =

U

(t, t
0
)

A(t
0
)

U (t, t
0
) . (15)
Each Hamiltonian

H determines the corresponding evolu-
tion operator

U. The evolution operator

U can be found
from the equation
i

t

U (t, t
0
) =

H

U (t, t
0
) ,

U (t
0
, t
0
) =

1 (16)
2
(a derivation of this equation is given below). It follows
from this equation that

U is unitary if

H is Hermitian.
When the Hamiltonian

H is time-independent, the solu-
tion of Eq. (16) can be formally written as

U (t, t
0
) = exp
_
1
i
(t t
0
)

H
_
. (17)
We call this equation formal because the implied innite se-
ries is not guaranteed to converge. This form of the evo-
lution operator also does not hold for systems with time-
dependent Hamiltonians, e.g.

H = H ( p, q, t) =
p
2
2m
+
1
2
W (t) q
2
F (t) q,
where W(t) and F(t) are some functions of time.
Derivation of Eq. (16). We need to show that if

U(t) is
such that Eqs. (14)-(15) hold for an arbitrary operator

A,
then

U(t) satises Eq. (16).
Substituting

A(t
0
) =

1, we nd

A(t) =

1 and therefore


U =

1, i.e. the operator

U must be unitary. It follows that
d
dt
_


U
_
= 0 =
_
d
dt

_

U +

U

d
dt

U. (18)
Now we can compute the time derivative of Eq. (15),
d
dt

A(t) =
_
d
dt

_

A(t
0
)

U +

U


A(t
0
)
d
dt

U
=
_
d
dt

_

U

A(t) +

A(t)

U

d
dt

U
=
_

A(t) ,

U

d
dt

U
_
. (19)
A substitution into Eq. (14) yields
_

A,

H i

d
dt

U
_
= 0 (20)
for an arbitrary operator

A. An operator that commutes
with all other operators must be proportional to

1. There-
fore

H i

d
dt

U = (t)

1, (21)
where (t) is an arbitrary function of time. This equa-
tion is equivalent to Eq. (16) except for the presence of the
extra term (t)

1. However, we can redene the operator

U(t)

U(t)e
i(t)
, where (t) is such that

(t) = (t); this


redenition has no effect on Eq. (15). After the redenition,
we nd that the new evolution operator satises Eq. (16).
1.4.2 Schrdinger picture
As an alternative to the Heisenberg picture, the
Schrdinger picture uses time-dependent states [ (t))
and time-independent operators q, p.
The Schrdinger picture is related to the Heisenberg pic-
ture by applying the evolution operator

U (t, t
0
) to the
Heisenberg states. Suppose that in the Heisenberg picture
the system was in a state [), then we dene
[ (t))

U (t, t
0
) [) , (22)
where t
0
is some initial time. The dynamic observables

A are now considered time-independent,



A

A(t
0
), and
equal to the Heisenberg operators at time t
0
. With these
denitions, it is easy to see that the matrix elements of the
operator

A coincide in both pictures:
[

A(t) [) = (t)[

A[ (t)) . (23)
Thus we reformulated the quantum-mechanical de-
scription in terms of time-dependent states and time-
independent operators.
It follows fromEqs. (16) and (22) that the time-dependent
state [ (t)) satises the Schrdinger equation
i

t
[ (t)) =

H[ (t)) . (24)
In the coordinate representation, [ (t)) is represented by a
function (x, t) and q and p by the operators x and i
x
.
Then the Schrdinger equation for the Hamiltonian (9) is
i
(x, t)
t
=

2
2m

2
(x, t)
x
2
+
1
2
m
2
x
2
(x, t) . (25)
To perform calculations in the Heisenberg picture, one
has to determine the evolution operator

U(t, t
0
), while in
the Schrdinger picture one needs to solve the Schrdinger
equation and obtain the wave function (x, t). Thus the
Schrdinger picture is easier to use for ordinary quantum-
mechanical problems such as the quantization of a har-
monic oscillator.
1.4.3 Operators of the type exp(
x
)
A cautionary note: operators such as exp (
x
), if under-
stood literally as
exp (
x
) f (x) =

n=0

n
n!
d
n
f (x)
dx
n
, (26)
are dened only on entire analytic functions f (x). Entire an-
alytic functions f (x) coincide with the sum of their Taylor
series at all values of x, and therefore
exp (
x
) f (x) =

n=0

n
n!
d
n
f (x)
dx
n
= f (x +) (27)
for all (complex) . Non-analytic functions, e.g. f (x) = [x[,
do not everywhere coincide with the sum of their Taylor
series; for instance, the operator exp (
x
) applied to the
function [x[ does not yield [x +[.
However, the shift operator f (x) f (x +) is dened
for all functions f (x). It is often easier to perform calcu-
lations with the operator exp (
x
) and not with the shift
operator. Therefore one usually performs the calculations
with operators of the form exp (
x
) even though one re-
ally means to use the shift operator. The intermediate steps
in such calculations are valid only when the operators are
applied to analytic functions; however, the nal results are
usually valid for all functions.
Similar considerations apply to the evolution operator

U (t t
0
) written in the exponential form (17). In the coor-
dinate representation, the Hamiltonian operator

H usually
contains derivatives and thus the operator exp(i

Ht/)
can be applied only to analytic functions (x). If the ini-
tial state [) is represented by a non-analytic function (x),
the operator exp(i

Ht/) cannot be applied. However, the
actual evolution operator

U (t t
0
) is of course dened for
all functions (x), not only for analytic functions. For in-
stance, the action of the evolution operator on an arbitrary
wavefunction can be expressed as an integral,

U (t t
0
) (x) =
_
dy K (x, y) (y) dy, (28)
3
where the function K (x, y) is called the propagator.
1
The
integral representation (28) is valid for all integrable func-
tions (x); the problem is with the exponential in the for-
mula (17). This exponential denes an operator that can be
applied only to analytic (x).
1.5 Energy spectrum
Up to this point, our considerations applied to all systems
and not only to the harmonic oscillator. Now we shall ap-
ply the formalism to the Hamiltonian (3).
1.5.1 Creation and annihilation operators
For a quantized oscillator with frequency , we dene the
auxiliary operator a,
a =
_
m
2
q +
i

2m
p, a

=
_
m
2
q
i

2m
p. (29)
As shown, the operator a is not Hermitian and its conjugate
is a

,= a. These operators are called the creation ( a

) and
the annihilation ( a) operators. One can easily verify that
[ a, a

] = 1. (30)
The Hamiltonian (9) is expressed through the new oper-
ators as

H =
_
a

a +
1
2
_
=
_
a a

1
2
_
=
_

N +
1
2
_
.
(31)
The operator

N a

a is called the occupation number op-


erator.
1.5.2 Energy eigenstates
The spectrum of the operator

H can be derived from the
assumption that there exists at least one eigenstate of

H.
The derivation is standard; it is based on the easily veried
property,

H[) = E [)
_

H a [) = (E ) a [) ,

H a

[) = (E + ) a

[) .
(32)
It follows from Eq. (31) that all eigenvalues of

H are
bounded from below by
1
2
. The lowest-energy state is
called the ground state. If we denote by [0) the assumed
ground state such that

H[0) = E
0
[0), it follows from
Eq. (30) that a [0) would be an eigenstate of lower energy
E
0
< E
0
unless
a [0) = 0. (33)
Therefore

H[0) =
1
2
[0) E
0
=
1
2
. (34)
It also follows that ( a

)
n
[0) are eigenstates of

H with eigen-
values E
0
+n and that there are no other possible eigen-
values. (In Sec. 1.5.5 it is shown that the operator a

cannot
have any eigenvectors, so in particular ( a

)
n
[0) , = 0 for all
n.) Therefore the spectrum of

H consists of the eigenvalues
E
n

_
1
2
+n
_
, n = 0, 1, 2, ... (35)
1
An explicit expression for the function K(x, y) can be derived in cer-
tain simple cases, in particular for the harmonic oscillator. At this moment
we only need to know that this function exists.
We denote by [n) the n-th eigenstate of

H,
[n)
1

n!
( a

)
n
[0) . (36)
One can verify that the factor

n! in Eq. (36) is necessary
for the correct normalization m[n) =
mn
.
1.5.3 Selecting the Hilbert space
To specify the Hilbert space H of oscillator states uniquely,
it is usually assumed that the occupation number eigenvec-
tors [n) are nondegenerate and form a complete basis of
the space H. This is equivalent to assuming that:
1. The ground state [0) exists and is unique. It follows
that all eigenstates [n) of

H are nondegenerate.
2. There are no other states [) H orthogonal to all [n).
The rst assumption eliminates alternative vacuum states
[0

), [0

) etc. Without the second assumption, there may


exist some states [) that are orthogonal to all [n) and are
not eigenstates of a or

H.
It is important to realize that these assumptions govern
the choice of the Hilbert space of quantum states and do not
follow from the algebraic properties of the operators a, a

.
Ultimately, these are physical assumptions, i.e. hypothe-
ses justied by experimental evidence. The quantum the-
ory of the harmonic oscillator built upon these assump-
tions is the simplest one possible and agrees with exper-
iments. The correct Hilbert space H is the space of se-
quences (z
0
, z
1
, z
2
, ...) such that

n=0
[z
n
[
2
< . A se-
quence (z
0
, z
1
, z
2
, ...) H represents a normalizable state

n=0
z
n
[n).
1.5.4 Coordinate representation
The coordinate representation involves the Hilbert space of
square-integrable functions (x) and is the simplest way
to represent the operators p, q. Once we assume that all the
quantum states of the oscillator belong to this space, then
it follows automatically that the ground state [0) is unique
and that the basis [n) is complete. The crucial assump-
tion is that the physics of the one-dimensional quantum
oscillator is adequately described by the coordinate rep-
resentation with wave functions (x) depending on one
real-valued coordinate x (rather than, say, wave functions
(x, y, z) depending on more variables).
Let us compute the ground state [0) in the coordinate rep-
resentation. By denition, a [0) = 0. The annihilation oper-
ator is
a
_
m
2
q +
i

2m
p =
_
m
2
x +
_

2m

x
,
and the state [0) is represented by the function (x) such
that a(x) = 0, i.e.
_
_
m
2
x +
_

2m

x
_
(x) = 0. (37)
This differential equation is easily solved,
(x) = C exp
_

mx
2
2
_
, (38)
where C is a constant of integration. The value of C is de-
termined from normalization of [0),
0[0) =
_
+

[(x)[
2
dx = 1, (39)
4
which yields
[C[ =
_
m

_
1/4
.
This determines C up to a complex phase; this ambiguity
is usually irrelevant in calculations. Therefore, the wave
function of the ground state is
(x) =
_
m

_
1/4
exp
_

mx
2
2
_
. (40)
1.5.5 *Nonexistence of eigenvectors of a

We now prove that the creation operator a

has no eigen-
vectors in the space H dened above. The operator a

can-
not have eigenvectors [) with zero eigenvalues because
otherwise we would have a contradiction,
0 = [ a a

[) = [) +[ a

a [) > 0. (41)
Suppose that [) is an eigenvector with a nonzero eigen-
value, i.e. a

[) = [) with ,= 0, then
0[ a

[) = 0 = 0[) , (42)
thus 0[) = 0. Since for n 1,
n[ a

[) = n 1[

n[) , (43)
it follows by induction that n[) = 0 for all n 0. Since
the set [n) is a complete basis in H, there are no nonzero
vectors orthogonal to every [n), so we must have [) = 0.
The absence of normalized eigenvectors of a

can also
be derived without assuming the completeness of the basis
[n). In other words, the properties of the operators a and
a

are such that a

cannot have eigenvectors in any Hilbert


space. A proof follows.
Suppose to the contrary that there exists a normalized
eigenvector [) such that a

[) = [) and [) = 1.
Then we can derive an inequality for [[,
[ a a

[) = [[
2
= [ a

a [) + 1 1. (44)
Similarly, by considering a
2
( a

)
2
we nd
[[
4
= [ a
2
( a

)
2
[) =
= [ ( a

)
2
a
2
[) + 4 [ a

a [) + 2 2. (45)
More generally, one can show by induction that
[[
2n
= [ a
n
( a

)
n
[) n!, (46)
for n = 1, 2, ... However, for any there exists a large
enough n such that the inequality (46) is violated. Therefore
the operator a

cannot have any normalized eigenvectors.


2 *Coherent states
A coherent state of a harmonic oscillator is an eigenstate
of the annihilation operator a. Thus the denition of the
coherent state [z) is
a [z) = z [z) , (47)
where z is a complex number. Sometimes the coherent
states are normalized to z[z) = 1; however, the states [z)
are not an orthogonal basis and it is not necessary to nor-
malize them. Moreover, in many calculations it is more con-
venient to choose a different normalization; below we shall
use z[z) = exp([z[
2
).
In this section we denote coherent states by [z), [z
1
), etc.,
and arbitrary states by [), [
1
), etc., while the occupation
number eigenstates are [0), [1), ..., [n), ... It should be clear
from the context which state is being referred to. (Note
that the coherent state [z = 0) is the same as the occupation
number eigenstate [0).)
2.1 Basic properties
2.1.1 Occupation number representation
A coherent state [z) can be dened by the explicit formula
[z) = exp(z a

) [0) , (48)
where [0) is the vacuum state of the oscillator, a [0) = 0. A
simple calculation shows that the state (48) satises Eq. (47).
The formula (48) can be derived from Eq. (47) by the fol-
lowing algebraic method (which is simpler than trying to
use the coordinate representation). The state [z) is a super-
position of the occupation number eigenstates [n),
[z) = z
0
[0) +z
1
[1) +..., (49)
where the coefcients z
n
are so far unknown. Any such
superposition can be obtained by acting with appropriate
powers of the creation operator a

on [0), so we can write


[z) =
_
z
0
+z
1
a

2!z
2
( a

)
2
+...
_
[0) f( a

) [0) , (50)
where we introduced an unknown function f (s),
f (s) z
0
+z
1
s +

2!z
2
s
2
+..., (51)
instead of the power series in brackets in Eq. (50). We shall
now determine the function f (s). Due to Eq. (30), the op-
erators a and a

are a canonically conjugate pair and so


[ a, f( a

)] = f

( a

), (52)
where we can treat a

as a normal variable (the argument


of the function f). It follows from this property and from
Eq. (47) that
z [z) = a [z) = af( a

) [0) = f

( a

) [0) , (53)
therefore
df
ds
= zf (s) f (s) = C (z) e
zs
, (54)
where C (z) is an arbitrary normalization constant. The
choice C (z) 1 yields Eq. (48). [This choice is made only
for simplicity and does not necessarily yield a state [z) nor-
malized to 1; see below for normalization factor.]
It follows from Eq. (48) that the coherent state [z) is ex-
pressed through the occupation number eigenstates [n) as
[z) =

n=0
z
n

n!
[n) . (55)
The denition (48) entails the following (easily veried)
normalization of coherent states,
z
1
[z
2
) = exp (z

1
z
2
) . (56)
Therefore the vector
[z
norm
) e
|z|
2
/2
[z) = exp
_
z

a +z a

_
[0) (57)
is normalized, z
norm
[z
norm
) = 1. (Note that the operator
z

a + z a

is anti-Hermitian, thus the exponential of this


operator is unitary.) Below it will be convenient the de-
nition (48) of the coherent state, rather than its normalized
version [z
norm
).
5
2.1.2 Approximate orthogonality
Equation (56) shows that coherent states are not orthogonal
to each other, namely z

[z) ,= 0 for any z and z

. However,
the states with very different values of the parameter z are
approximately orthogonal. The angle (z, z

) between the
directions of [z) and [z

) can be formally dened by


cos (z, z

) =
[z[z

)[
_
z[z)
_
z

[z

)
. (58)
A simple calculation yields
cos (z, z

) = exp
_

1
2
[z[
2
+ Re (z

)
1
2
[z

[
2
_
= exp
_

1
2
[z z

[
2
_
. (59)
Therefore cos is very close to 0 when [z z

[ 1. Since
cos 0, it follows that

2
, that is, the vectors [z) and
[z

) are close to being orthogonal.


2.1.3 Coordinate representation
In the coordinate representation, states [) are described by
functions (x) = x[), where [x) is the generalized eigen-
basis of the position operator x.
A coherent state [z) dened by Eq. (48) is represented by
the function

z
(x) =
_
m

_
1/4
exp
_

z
2
2

mx
2
2
+zx
_
2m

_
.
(60)
This formula can be derived by solving the equation a [z) =
z [z) in the x representation, and then xing the normaliza-
tion. It follows that the wavefunction (60) is a Gaussian
wave packet.
2.1.4 Decomposition of unity
Despite the non-orthogonality of coherent states, they sat-
isfy the property
1

__
d (Rez) d (Imz) e
zz

[z) z[ =

1, (61)
where the integration is performed over all complex z. This
identity is analogous to the decomposition of unity through
an orthogonal basis (despite the fact that the states [z) are
not orthogonal!).
To derive the identity (61), it sufces to prove that for any
occupation number eigenstates [m) and [n) we have

mn
= m[n) =
1

__
d (Rez) d (Imz) e
zz

m[z) z[n) .
(62)
This can be veried by a direct integration using Eq. (48).
2.2 Holomorphic representation
Coherent states provide a representation of quantum states
[) of a harmonic oscillator by analytic functions of an aux-
iliary complex variable z. This representation is useful in
certain applications (for instance, in quantum optics).
2.2.1 Representation of states
In the holomorphic representation, an arbitrary state vec-
tor [) is represented by the function
(z) z[) , (63)
where z is now considered to be an auxiliary complex vari-
able. With the denition (55), we nd
(z) =

n=0
(z

)
n

n!
n[) , (64)
where [n) are the (normalized) occupation number eigen-
states. Since [n[)[ 1 for all n, the series in Eq. (64) abso-
lutely converges for all complex z

and therefore the func-


tion (z) is always an entire analytic function of z

, i.e. an
analytic function without poles or singularities in the en-
tire complex plane. In mathematics such functions are also
called holomorphic, hence the name holomorphic repre-
sentation.
Examples of holomorphic representations:
1. A coherent state [z
0
) is represented by the function
z[z
0
) = e
z

z
0
. (65)
2. An occupation number eigenstate [n) is represented by
the function
z[n) =
1

n!
(z

)
n
. (66)
To restore the state vector [) from its representative (z),
we use the decomposition (61) and nd
[) =
1

__
d (Rez) d (Imz) e
zz

[z) z[) . (67)


In particular, setting [) = [z

) we get
[z

) =
1

__
d (Rez) d (Imz) e
zz

[z) z[z

) , (68)
which demonstrates an explicit linear dependence between
coherent states [z). This is consistent with the fact that the
set of all coherent states is an overcomplete basis: any state
[) can be expressed as a linear combination of [z), but
there are too many vectors [z). So a decomposition of
a state [) in the coherent state basis is not unique, i.e. in
the expression
[) =
__
d (Rez) d (Imz) A(z) [z) (69)
there are many ways to choose the function A(z). However,
the holomorphic representation is unique, i.e. each state [)
is described by a unique holomorphic function (z) de-
ned through Eq. (63).
2.2.2 Representation of operators
Operators

A are represented by their matrix elements
z
1
[

A[z
2
) A(z
1
, z
2
). It follows from Eq. (55) that
A(z
1
, z
2
) =

m,n=0
n[

A[m)
(z

1
)
n
z
m
2

m!n!
, (70)
therefore A(z
1
, z
2
) is an entire analytic function of z

1
and
(separately) of z
2
.
For example, the identity operator

1 is represented by the
function

m,n=0
n[

1 [m)
(z

1
)
n
z
m
2

m!n!
= e
z

1
z
2
= z
1
[z
2
) . (71)
The annihilation operator is represented by
z
1
[ a [z
2
) = z
2
z
1
[z
2
) = z
2
e
z

1
z
2
. (72)
6
The occupation number operator is represented by
z
1
[

N [z
2
) = z
1
[ a

a [z
2
) = z

1
z
2
e
z

1
z
2
. (73)
More generally, any operator

A can be expressed through
the creation and annihilation operators as

A =

m,n=0
A
mn
( a

)
m
a
n
, (74)
where A
mn
are some coefcients and all operators a are to
the right of all a

. (This is called normal ordering.) Any


combination of a and a

can be normal-ordered by using


the commutation relations. Then one can show that the
holomorphic representation of the operator

A is
z
1
[

A[z
2
) = z
1
[z
2
)

m,n=0
A
mn
(z

1
)
m
z
n
2
. (75)
Now we can compute the result of an operator

A acting
on a state [) in the holomorphic representation. An oper-
ator

A represented by a function A(z
1
, z
2
) acts upon a state
[) represented by a function (z) and yields a different
state represented by the following function of z
1
,
z
1
[

A[) =
1

__
d (Rez) d (Imz) e
zz

z
1
[

A[z) z[)
=
1

__
d (Rez) d (Imz) e
zz

A(z
1
, z) (z) .
(76)
2.3 Evolution of coherent states
In the Schrdinger picture, states evolve with time. A sim-
ple calculation yields the following result: If the harmonic
oscillator is in a coherent state [z) at an initial time t
0
, the
state [(t)) at a later time t is (up to a phase) also a co-
herent state with the parameter z exp (i(t t
0
)). More
precisely,
[(t)) =

U (t t
0
) [z) = e

i
2
(tt
0
)

ze
i(tt
0
)
_
. (77)
To prove this, we shall not try to compute the wave func-
tion in the position representation but instead use a trick.
We shall nd an operator-valued function f( a

, t) such that
[(t)) =

U (t t
0
) [z) = f( a

, t) [0) . (78)
Since initially the state is [z) and we have the relation (48),
the function f must satisfy the condition
f
_
a

, t
0
_
= exp
_
za

_
. (79)
The Schrdinger equation for the state [(t)) becomes
i

t
f( a

, t) [0) =

Hf( a

, t) [0) = ( a

a +
1
2
)f( a

, t) [0) ,
(80)
which yields a differential equation for f,
i
f
t
=

2
f +a

f
a

, (81)
where we now consider a

to be an ordinary variable. The


general solution of this partial differential equation is
f
_
a

, t
_
= e

i
2
t
g
_
e
it
a

_
, (82)
where g (q) is an arbitrary function. This function is xed
by the initial condition (79) and we obtain the desired re-
sult,
f
_
a

, t
_
= e

i
2
t
exp
_
ze
i(tt
0
)
a

_
. (83)
2.3.1 Quasi-classical behavior
Coherent states exhibit the smallest possible uncertainty in
the position and momentum (the Heisenberg inequality be-
comes an equality). However, coherent states are not the
only states with this property. More general states with the
minimum uncertainty are squeezed states that also nd
their use in quantum optics and in quantum eld theory.
(We shall not prove these statements here; the required cal-
culations are straightforward and can be found in many
textbooks.)
In addition to having the minimum uncertainty, coher-
ent states correspond to the motion of the oscillator with
expectation values q(t)) and p(t)) that obey the classical
equations of motion. For instance, we can easily compute
z[ q [z) =
_

2m
z[ ( a + a

) [z) =
_

2m
(z +z

) , (84)
therefore z(t)[ q [z(t)) = q(t
0
) cos (t t
0
). This is the fa-
miliar classical trajectory of an oscillator.
In a coherent state, the expectation values of the coordi-
nate and momentum can be macroscopically large. Thus
a coherent state is the quantum-mechanical representation
of an almost classical motion of the harmonic oscilla-
tor, where the classical trajectory is dened as precisely as
quantum uctuations allow.
2.3.2 Excitations of coherent states
Even though the oscillator in a coherent state behaves sim-
ilarly to a classical system, quantum uctuations are still
present. One can dene excited coherent states,
[z; n) =
1

n!
( a

)
n
[z) , (85)
which describe an oscillator executing a quasi-classical mo-
tion and additionally having certain quantum excitations.
Note that the operators exp(z a

) commute for all z, and


thus any coherent state [z) (including the vacuum [0)) can
be expressed through any other coherent state [z

) by
[z) = exp
_
(z z

) a

_
[z

) . (86)
Therefore the operator exp(z a

) is invertible and the set of


states [z; n), where n = 0, 1, ... and z is xed, is a complete
basis of the Hilbert space.
7

Das könnte Ihnen auch gefallen