dx
(x)(x) = 
Dirac notation
For a complete basis set,
i
, we can dene the expansion
 =
i
i
where j  =
i
j i
..
ij
=
j
.
For example, in the real space basis,  =
_
dx (x)x.
Then, since xx
= (x x
),
x
 =
_
dx (x) x
x
. .
(xx
)
= (x
)
In Dirac formulation, real space representation recovered from inner
product, (x) = x; equivalently (p) = p.
Operators
An operator
A maps one state vector, , into another, , i.e.
A = .
If
A = a with a real, then  is said to be an eigenstate (or
eigenfunction) of
A with eigenvalue a.
e.g. plane wave state
p
(x) = x
p
= Ae
ipx/
is an eigenstate of
the momentum operator, p = i
x
, with eigenvalue p.
For every observable A, there is an operator
A which acts upon the
wavefunction so that, if a system is in a state described by , the
expectation value of A is
A = 
A =
_
dx
(x)
A(x)
Operators
Every operator corresponding to observable is linear and Hermitian,
i.e. for any two wavefunctions  and , linearity implies
A( +) =
A +
A
For any linear operator
A, the Hermitian conjugate (a.k.a. the
adjoint) is dened by relation

A =
_
dx
A) =
_
dx (

Hermiticity implies that
A
=
A, e.g. p = i
x
.
Operators
From the denition,
 = 
= 
A,
i.e. (
 =
A, (
=
A
2
From 
B =
B =
,
it follows that (
B)
=
B
.
Operators are associative,i.e. (
B)
C =
A(
B
C),
but not (in general) commutative,
B =
A(
B) = (
B) =
B
A .
Operators
A physical variable must have real expectation values (and
eigenvalues) physical operators are Hermitian (selfadjoint):

H
=
__
(x)
H(x)dx
_
=
_
(x)(
H(x))
dx =
H
i.e.
H = 
H =
, and
H
=
H.
Eigenfunctions of Hermitian operators
Hi = E
i
i form complete
orthonormal basis, i.e. i j =
ij
For complete set of states i , can expand a state function  as
 =
i
i i 
In coordinate representation,
(x) = x =
i
xi i  =
i
i 
i
(x),
i
(x) = xi
Resolution of identity
 =
i
i i 
If we sum over complete set of states, obtain the (useful) resolution
of identity,
i
i i  = I
i
x
i i x = x
x
i.e. in coordinate basis,
i
(x)
i
(x
) = (x x
).
As in 3d vector space, expansion  =
i
b
i
i and  =
i
c
i
i
allows scalar product to be taken by multiplying components,
 =
i
b
i
c
i
.
Example: resolution of identity
Basis states can be formed from any complete set of orthogonal
states including position or momentum,
_
dxxx =
_
dppp = I.
From these denitions, can recover Fourier representation,
(x) x =
_
dp xp
. .
e
ipx/
/
2
p =
1
2
_
dp e
ipx/
(p)
where xp denotes plane wave state p expressed in the real space
basis.
Timeevolution operator
Formally, we can evolve a wavefunction forward in time by applying
timeevolution operator.
For timeindependent Hamiltonian, (t) =
U(t)(0), where
timeevolution operator (a.k.a. the propagator):
U(t) = e
i
Ht/
follows from timedependent Schrodinger equation,
H = i
t
.
By inserting the resolution of identity, I =
i
i i , where i are
eigenstates of
H with eigenvalue E
i
,
(t) = e
i
Ht/
i
i i (0) =
i
i i (0)e
iE
i
t/
Timeevolution operator
U = e
i
Ht/
Timeevolution operator is an example of a Unitary operator:
Unitary operators involve transformations of state vectors which
preserve their scalar products, i.e.
 =
U
U = 
U
!
= 
i.e.
U
U = I
Uncertainty principle for noncommuting operators
For noncommuting Hermitian operators, we can establish a bound
on the uncertainty in the expectation values of
A and
B:
Given a state , the mean square uncertainty dened as
(A)
2
= (
A)
2
= 
U
2
(B)
2
= (
B)
2
= 
V
2
where
U =
A
A,
A 
A, etc.
Consider then the expansion of the norm 
U + i
V
2
,

U
2
+
2

V
2
+ i
U
V i
V
U 0
i.e. (A)
2
+
2
(B)
2
+ i [
U,
V] 0
Since
A and
U,
V] = [
A,
B].
Uncertainty principle for noncommuting operators
(A)
2
+
2
(B)
2
+ i [
A,
B] 0
Minimizing with respect to ,
2(B)
2
+ i [
A,
B] = 0, i =
1
2
[
A,
B]
(B)
2
and substituting back into the inequality,
(A)
2
(B)
2
1
4
[
A,
B]
2
i.e., for noncommuting operators,
(A)(B)
i
2
[
A,
B]
Uncertainty principle for noncommuting operators
(A)(B)
i
2
[
A,
B]
For the conjugate operators of momentum and position (i.e.
[ p, x] = i , recover Heisenbergs uncertainty principle,
(p)(x)
i
2
[ p, x] =
2
Similarly, if we use the conjugate coordinates of time and energy,
[
E, t] = i ,
(t)(E)
i
2
[t,
E] =
2
Timeevolution of expectation values
For a general (potentially timedependent) operator
A,
t

A = (
t
)
A +
t
A +
A(
t
)
Using i
t
 =
H, i (
t
) = 
H, and Hermiticity,
t

A =
1
i
H
A +
t
A +
1
A(i
H)
=
i
_

A 
H
_
. .
[
H,
A]
+
t
A
For timeindependent operators,
A, obtain Ehrenfest Theorem,
t

A =
i
[
H,
A] .
Ehrenfest theorem: example
t

A =
i
[
H,
A] .
For the Schrodinger operator,
H =
p
2
2m
+ V(x),
t
x =
i
H, x] =
i
[
p
2
2m
, x] =
p
m
Similarly,
t
p =
i
H, i
x
] = (
x
H) =
x
V
i.e. Expectation values follow Hamiltons classical equations of
motion.
Symmetry in quantum mechanics
Symmetry considerations are very important in both low and high
energy quantum theory:
1
Structure of eigenstates and spectrum reect symmetry of the
underlying Hamiltonian.
2
Transition probabilities between states depend upon
transformation properties of perturbation = selection
rules.
Symmetries can be classied as discrete and continuous,
e.g. mirror symmetry is discrete, while rotation is continuous.
Symmetry in quantum mechanics
Formally, symmetry operations can be represented by a group of
(typically) unitary transformations (or operators),
U such that
O
U
O
U
Such unitary transformations are said to be symmetries of a
general operator
O if
O
U =
O
i.e., since
U
=
U
1
(unitary), [
O,
U] = 0.
If
O
H, such unitary transformations are said to be symmetries of
the quantum system.
Continuous symmetries: Examples
Operators p and r are generators of spacetime transformations:
For a constant vector a, the unitary operator
U(a) = exp
_
a p
_
eects spatial translations,
U
(a)f (r)
U(a) = f (r + a).
Proof: Using the BakerHausdor identity (exercise),
e
Be
A
=
B + [
A,
B] +
1
2!
[
A, [
A,
B]] +
with e
= e
a
and
B f (r), it follows that
(a)f (r)
U(a) = f (r) + a
i
1
(
i
1
f (r)) +
1
2!
a
i
1
a
i
2
(
i
1
i
2
f (r)) +
= f (r + a) by Taylor expansion
Continuous symmetries: Examples
Operators p and r are generators of spacetime transformations:
For a constant vector a, the unitary operator
U(a) = exp
_
a p
_
eects spatial translations,
U
(a)f (r)
U(a) = f (r + a).
Therefore, a quantum system has spatial translation symmetry i
U(a)
H =
H
U(a), i.e. p
H =
H p
i.e. (sensibly)
H =
H( p) must be independent of position.
Similarly (with
L = r p the angular momemtum operator),
_
_
_
U(b) = exp[
i
b r]
U() = exp[
i
e
n
L]
U(t) = exp[
i
Ht]
eects
_
_
_
momentum translations
spatial rotations
time translations
Discrete symmetries: Examples
The parity operator,
P, involves a sign reversal of all coordinates,
P(r) = (r)
discreteness follows from identity
P
2
= 1.
Eigenvalues of parity operation (if such exist) are 1.
If Hamiltonian is invariant under parity, [
P,
H] = 0, parity is said to
be conserved.
Timereversal is another discrete symmetry, but its representation
in quantum mechanics is subtle and beyond the scope of course.
Consequences of symmetries: multiplets
Consider a transformation
U which is a symmetry of an operator
observable
A, i.e. [
U,
A] = 0.
If
A has eigenvector a, it follows that
Ua will be an eigenvector
with the same eigenvalue, i.e.
AUa =
U
Aa = aUa
This means that either:
1
a is an eigenvector of both
A and
U (e.g. p is eigenvector
of
H =
p
2
2m
and
U = e
i a p/
), or
2
eigenvalue a is degenerate: linear space spanned by vectors
U
n
a (n integer) are eigenvectors with same eigenvalue.
e.g. next lecture, we will address central potential where
H is
invariant under rotations,
U = e
i e
n
L/
states of angular
momentum, , have 2 + 1fold degeneracy generated by
L
.
Heisenberg representation
Schrodinger representation: timedependence of quantum system
carried by wavefunction while operators remain constant.
However, sometimes useful to transfer timedependence to
operators: For observable
B, timedependence of expectation value,
(t)
B(t) = e
i
Ht/
(0)
Be
i
Ht/
(0)
= (0)e
i
Ht/
Be
i
Ht/
(0)
Heisenberg representation: if we dene
B(t) = e
i
Ht/
Be
i
Ht/
,
timedependence transferred from wavefunction and
t
B(t) =
i
e
i
Ht/
[
H,
B]e
i
Ht/
=
i
H,
B(t)]
cf. Ehrenfests theorem
Quantum harmonic oscillator
The harmonic oscillator holds priviledged position in quantum
mechanics and quantum eld theory.
H =
p
2
2m
+
1
2
m
2
x
2
It also provides a useful platform to illustrate some of the
operatorbased formalism developed above.
To obtain eigenstates of
H, we could seek solutions of linear second
order dierential equation,
_
2
2m
2
x
+
1
2
m
2
x
2
_
= E
However, complexity of eigenstates (Hermite polynomials) obscure
useful features of system we therefore develop an alternative
operatorbased approach.
Quantum harmonic oscillator
H =
p
2
2m
+
1
2
m
2
x
2
Form of Hamiltonian suggests that it can be recast as the square
of an operator: Dening the operators (no hats!)
a =
_
m
2
_
x + i
p
m
_
, a
=
_
m
2
_
x i
p
m
_
we have a
a =
m
2
x
2
+
p
2
2m
i
2
[ p, x]
. .
i
=
1
2
Together with aa
=
H
+
1
2
, we nd that operators full the
commutation relations
[a, a
] aa
a = 1
Setting n = a
a,
H = ( n + 1/2)
Since operator n = a
H = (a
a + 1/2)
Ground state 0 identied by nding state for which
a0 =
_
m
2
_
x + i
p
m
_
0 = 0
In coordinate basis,
xa0 = 0 =
_
dx
xax
0 =
_
x +
m
x
_
0
(x)
i.e. ground state has energy E
0
= /2 and
0
(x) = x0 =
_
m
_
1/4
e
mx
2
/2
N.B. typo in handout!
Quantum harmonic oscillator
H = (a
a + 1/2)
Excited states found by acting upon this state with a
.
Proof: using [a, a
] aa
a = 1, if nn = nn,
n(a
n) = a
aa
..
a
a + 1
n = (a
a
..
n
+a
)n = (n + 1)a
n
equivalently, [ n, a
] = na
n = a
.
Therefore, if n is eigenstate of n with eigenvalue n, then a
n is
eigenstate with eigenvalue n + 1.
Eigenstates form a tower; 0, 1 = C
1
a
0, 2 = C
2
(a
)
2
0, ...,
with normalization C
n
.
Quantum harmonic oscillator
H = (a
a + 1/2)
Normalization: If nn = 1, naa
n+1
a
n!
(a
)
n
0, nn
=
nn
are eigenstates of
H with eigenvalue E
n
= (n + 1/2) and
a
n =
n + 1n + 1, an =
nn 1
a and a
] = 1.
Later in the course, we will nd that these commutation relations
are the hallmark of bosonic quantum particles and this
representation, known as second quantization underpins the
quantum eld theory of relativistic particles (such as the photon).
Quantum harmonic oscillator: dynamical echo
How does a general wavepacket (0) evolve under the action of
the quantum timeevolution operator,
U(t) = e
i
Ht/
?
For a general initial state, (t) =
U(t)(0). Inserting the
resolution of identity on the complete set of eigenstates,
(t) = e
i
Ht/
n
nn(0) =
i
nn(0)e
iE
n
t/
e
i (n+1/2)t
For the harmonic oscillator, E
n
= (n + 1/2).
Therefore, at times t =
2
m, m integer, (t) = e
i t/2
(0)
leading to the coherent reconstruction (echo) of the wavepacket.
At times t =
H,
B].
Therefore, making use of the identity, [
H, a] = a (exercise),
t
a = i a, i.e. a(t) = e
i t
a(0)
Combined with conjugate relation a
(t) = e
i t
a
+ a), p = i
_
m
2
(a a
)
p(t) = p(0) cos(t) m x(0) sin(t)
x(t) = x(0) cos(t) +
p(0)
m
sin(t)
i.e. operators obey equations of motion of the classical harmonic
oscillator.
But how do we use these equations...?
Quantum harmonic oscillator: timedependence
p(t) = p(0) cos(t) m x(0) sin(t)
x(t) = x(0) cos(t) +
p(0)
m
sin(t)
Consider dynamics of a (real) wavepacket dened by (x) at t = 0.
Suppose we know expectation values, p
2
0
=  p
2
, x
2
0
= x
2
,
and we want to determine (t) p
2
(t).
In Heisenberg representation, (t) p
2
(t) =  p
2
(t) and
p
2
(t) = p
2
(0) cos
2
(t) + (mx(0))
2
sin
2
(t)
m(x(0) p(0) + p(0)x(0))
Since (x(0) p(0) + p(0)x(0)) = 0 for (x) real, we have
 p
2
(t) = p
2
0
cos
2
(t) + (mx
0
)
2
sin
2
(t)
and similarly  x
2
(t) = x
2
0
cos
2
(t) +
p
2
0
(m)
2
sin
2
(t)
Coherent states
The ladder operators can be used to construct a wavepacket which
most closely resembles a classical particle the coherent or
Glauber states.
Such states have numerous applications in quantum eld theory and
quantum optics.
The coherent state is dened as the eigenstate of the annihilation
operator,
a = 
Since a is not Hermitian, can take complex eigenvalues.
The eigenstates are constructed from the harmonic oscillator ground
state the by action of the unitary operator,
 =
U()0,
U() = e
a
a
Coherent states
 =
U()0,
U() = e
a
a
The proof follows from the identity (problem set I),
a
U() =
U()(a +)
i.e.
U is a translation operator,
U
()a
U() = a +.
By making use of the BakerCampbellHausdor identity
e
X
e
Y
= e
X+
Y+
1
2
[
X,
Y]
valid if [
X,
Y] is a cnumber, we can show (problem set)
U() = e
a
a
= e

2
/2
e
a
a
i.e., since e
a
0 = 0,
 = e

2
/2
e
a
0
Coherent states
a = ,  = e

2
/2
e
a
0
Expanding the exponential, and noting that n =
1
n!
(a
)
n
0, 
can be represented in number basis,
 =
n=0
(a
)
n
n!
0 =
n
e

2
/2
n
n!
n
i.e. Probability of observing n excitations is
P
n
= n
2
= e

2 
2n
n!
a Poisson distribution with average occupation, a
a = 
2
.
Coherent states
a = ,  = e

2
/2
e
a
0
Furthermore, one may show that the coherent state has minimum
uncertainty x p =
2
.
In the real space representation (problem set I),
(x) = x = N exp
_
(x x
0
)
2
4(x)
2
i
p
0
x
_
where (x)
2
=
2m
and
x
0
=
_
2m
(
+) = Acos
p
0
= i
_
m
2
(
) = mAsin
where A =
_
2
m
and = e
i
.
Coherent States: dynamics
a = ,  =
n
e

2
/2
n
n!
n
Using the timeevolution of the stationary states,
n(t) = e
iE
n
t/
n(0), E
n
= (n + 1/2)
it follows that
(t) = e
i t/2
n
e

2
/2
n
n!
e
int
n = e
i t/2
e
i t