1.

1 Dirac Notation and rules of Quantum Mechanics 1

1.1 Dirac Notation and rules of Quantum Mechanics

1.1.1 States and operators
A quantum state is represented by the ket |ψi. The Hermitian conjugate is the bra hψ|. The
inner product is
hφ|ψi = c (a number). (1.1)
If c = hφ|ψi then the complex conjugate is c∗ = hφ|ψi∗ = hψ|φi. Kets and bras exist in
a Hilbert space which is a generalization of the three dimensional linear vector space of
Euclidean geometry to a complex valued space with possibly infinitely many dimensions.
The inner product is linear

hφ| (a1 |ψ1i + a2 |ψ2i) = a1hφ|ψ1 i + a2 hφ|ψ2i. (1.2)

Operators are denoted by a hat Â.

 (c1|ψ1i + c2 |ψ2i) = c1 Â|ψ1i + c2 Â|ψ2i.

The matrix element of an operator is

hφ|Â|ψi = hφ| (Â|ψi) = (hφ|Â) |ψi = c (a number). (1.3)

The expectation value of an operator for a system in state |ψi is

hÂi = hai = hψ|Â|ψi. (1.4)

The complex conjugate of the matrix element is

hφ|Â|ψi∗ = hψ|†|φi = c∗ (1.5)

where † is the Hermitian conjugate of Â. When  is represented by a matrix the Hermitian
conjugate is found by transposing the matrix and then taking the complex conjugate of
each matrix element. The operation of taking the Hermitian conjugate of a combination of
numbers, states, and operators involves changing c → c∗ , |ψi → hψ|, hψ| → |ψi,  → †
and reversing the order of all elements. For example
 †
c1 †hφ|B̂|ψihξ| = c∗1|ξihψ|B̂ † |φiÂ.

1.1.2 Observables
Observables are represented by Hermitian operators which satisfy † = Â. The expectation
value of a Hermitian operator is real:

a∗ = hÂi∗ = hψ|Â|ψi∗ = hψ|†|ψi = hψ|Â|ψi = a. (1.6)

Denote the eigenstates of a Hermitian operator by |ni. The eigenvalues are real since

hm|Â|mi = hm|am |mi = am hm|mi = am

February 4, 2010 M. Saffman

2 1 Quantum Mechanics

and
a∗m = hm|Â|mi∗ = hm|†|mi = hm|Â|mi = am .
States corresponding to different eigenvalues are orthogonal. We assume the states are
normalized so that hm|ni = δmn . To prove orthogonality we calculate

hm|Â|ni = hm|an |ni = an hm|ni

and
   †
hm|Â|ni = hm| |ni = †|mi |ni = (am |mi)† |ni = a∗m hm|ni = am hm|ni.

Thus
(am − an )hm|ni = 0
so hm|ni = 0 if m 6= n.

The eigenstates |ni of a Hermitian operator form a complete set. Therefore for an arbitrary
ket |ψi
X∞
|ψi = cn |ni (1.7)
n=0

where
X
∞ X
∞ X
∞
hn|ψi = hn| cj |ji = cj hn|ji = cj δnj = cn .
j=0 j=0 j=0

Thus !
X
∞ X
∞ X
∞ X
∞
|ψi = cn |ni = hn|ψi|ni = |nihn|ψi = |nihn| |ψi.
n=0 n=0 n=0 n=0

Since this is true for arbitrary |ψi we can write the identity operator as
X
∞
Iˆ = |nihn|. (1.8)
n=0

A component of |ψi can be found by operating with the projection operator P̂n = |nihn|.
We have
X
∞ X
∞ X
∞
P̂n |ψi = |nihn| cj |ji = |ni cj hn|ji = |ni cj δnj = cn |ni. (1.9)
j=0 j=0 j=0

The projection operator is idempotent:

(P̂n )2 = P̂n P̂n = (|nihn|) (|nihn|) = |ni (hn|ni) hn| = |nihn| = P̂n . (1.10)

The inner product of two P

states can be expressed
P in terms of the coefficients of their decom-
position. We write |ψi = n cn |ni, |φi = n bn |ni. Then
X X XX X
hφ|ψi = b∗m hm| cn |ni = b∗m cn δmn = b∗n cn . (1.11)
m n m n n
1.1 Dirac Notation and rules of Quantum Mechanics 3

The spectral representation of an operator is found from

! !
X X XX
 = IˆÂIˆ = |mihm|  |nihn| = |mihm|an |nihn|
m n m n
XX X
= an |miδmn hn| = an |nihn|.
m n n

Thus X
 = an |nihn|. (1.12)
n

The representation (1.12) is diagonal since we have expressed  in a basis of the eigen-
vectors of Â. If we choose some other set of basis vectors {|mi} (not the eigenvectors of Â)
then the representation will not be diagonal. Thus
X
 = an |nihn|
n
! !
X X X
= an |mihm| |nihn| |m0ihm0 |
n m m0
X
= an |micmn cnm0 hm0 |
n,m,m0
X
= umm0 |mihm0|.
m,m0

P
where cmn = hm|ni and umm0 = n an cmn cnm0 .

1.1.3 Degeneracy
Each eigenvalue aα may be associated with a subspace of dimension nα > 1. The nα degener-
ate eigenvectors can be orthonormalized to span the subspace. In this case we label the eigen-
vectors with an additional parameter r as |α, ri where r = 1, 2...nα and hβ, s|α, ri = δαβ δrs.
The eigenvalue relation is then

Â|α, ri = aα|α, ri for r = 1, ...nα. (1.13)

The identity operator can be written as

X
∞ X
nα
Iˆ = |α, rihα, r|. (1.14)
α=0 r=1

The spectral decomposition of the operator is

X
∞ X
nα
 = aα|α, rihα, r|. (1.15)
α=0 r=1

February 4, 2010 M. Saffman

4 1 Quantum Mechanics

1.1.4 Continuous Basis

A Dirac ket |ψi should be thought of as an abstract symbol for a quantum state. It is not
tied to any particular representation. By taking inner products we can find a representation
in terms of a discrete set of basis states as in Eq. (1.7). This can also be generalized to a
continuous basis |ξi using the orthogonality condition
hξ 0 |ξi = δ(ξ − ξ 0 )
and the representation of the unit operator
Z
Iˆ = dξ |ξihξ|.

An arbitrary state |ψi can be expanded as

Z  Z Z
ˆ
|ψi = I|ψi = dξ |ξihξ| |ψi = dξ hξ||ψi|ξi = dξ ψ(ξ)|ξi

where we have introduced the wavefunction ψ(ξ) = hξ|ψi. The wavefunction ψ(ξ) gives the
amplitude of the decomposition of the state |ψi into the basis ket |ξi.
Using a continuous basis we can calculate the matrix elements of operators as follows.
ˆ ∂ ) that is some function of ξˆ and ∂ . For arbitrary states
Consider a general operator Â(ξ, ∂ ξ̂ ∂ ξ̂
|ψi, |φi we have
Z
hφ|Â|ψi = ˆ ∂ )|ξ 0 ihξ 0 |ψi.
dξ 00 dξ 0 hφ|ξ 00 ihξ 00 |Â(ξ, (1.16)
∂ ξˆ
Now the matrix element in the middle of the last expression is
ˆ ∂ )|ξ 0 i = hξ 00 |A(ξ 0 , ∂ )|ξ 0 i ≡ δ(ξ 00 − ξ 0 )A(ξ 0 , ∂ ).
hξ 00 |Â(ξ,
∂ ξˆ ∂ξ 0 ∂ξ 0
Thus (1.16) becomes
Z
∂
hφ|Â|ψi = dξ 00 dξ 0 hφ|ξ 00 iδ(ξ 00 − ξ 0 )A(ξ 0 , )hξ 0 |ψi
∂ξ 0
Z
∂
= dξ 0 hφ|ξ 0 iA(ξ 0 ,
0
)hξ 0 |ψi
∂ξ
Z
∂
= dξ φ∗ (ξ)A(ξ, )ψ(ξ). (1.17)
∂ξ
Equation (1.17) gives the general formula for evaluating a matrix element in terms of an
expansion in a continuous basis.

1.1.5 Representation of Derivatives

Given a ket |ψi we can define another ket |dψ/dξi whose representation is the derivative of
the original one. This new ket is the result of transforming the original one with an operator
and we write the transforming operator as ddξˆ so
d dψ
|ψi = | i.
dξˆ dξˆ
1.1 Dirac Notation and rules of Quantum Mechanics 5

The matrix element of the differential operator is

Z Z Z
d d 0 0 d 0 dψ(ξ 0 )
hφ| |ψi = dξ hφ| |ξ ihξ |ψi = dξ hφ| |ξ iψ(ξ ) = dξ 0 φ∗(ξ 0 )
0 0 0
. (1.18)
dξˆ dξˆ dξˆ dξ 0

Assuming the wavefunctions vanish at infinity an integration by parts gives

Z 0 Z ∗ 0
0 ∗ 0 dψ(ξ ) 0 dφ (ξ )
dξ φ (ξ ) = − dξ ψ(ξ 0 ). (1.19)
dξ 0 dξ 0

Comparing (1.18) and (1.19) we get

d 0 dφ∗ (ξ 0 ) dφ 0
hφ| |ξ i = − = −h |ξ i
dξˆ dξ 0 dξˆ
so
d dφ
hφ| = −h | (1.20)
dξˆ dξˆ
 †  †
ˆ
and d/dξ = −d/dξ. Thus id/dξ = id/dξˆ and e.g. the momentum operator −i~d/dx̂
ˆ ˆ
is a Hermitian operator.
If you find the steps from (1.18) - (1.20) unnecessarily complicated an alternative is to
simply note that
 
d d
hφ| |ψi = hφ| |ψi
dξˆ dξˆ
Z
dψ
= dξ φ∗
dξ
Z
dφ∗
= − dξ ψ
dξ
dφ
= −h |ψi
dξ
 
d
= hφ| |ψi.
dξˆ

Thus
d dφ
hφ| = −h |.
dξˆ dξˆ

1.1.6 Position and momentum representations

We can use the above results for a continuous basis to find the relation between the position
and momentum wavefunctions ψ(x) and φ(p). Consider position and momentum eigenkets:
x̂|xi = x|xi and p̂|pi = p|pi. The matrix element of p̂ between different basis kets is

hx|p̂|pi = hx|p|pi = phx|pi.

February 4, 2010 M. Saffman

6 1 Quantum Mechanics

Using the position representation of the momentum operator we can also write this as
 
∂ ∂
hx|p̂|pi = hx| −i~ |pi = −i~ hx|pi. (1.21)
∂ x̂ ∂x

This follows also from the general expression for matrix elements Eq. (1.17) 1 .
Comparing we see that
∂ ip
hx|pi = hx|pi
∂x ~
which has solution
hx|pi = Neıpx/~
with N a normalization constant.
We then observe that
Z  Z
ψ(x) = hx|ψi = hx| dp |pihp| |ψi = N dp eıpx/~ ψ(p)

and Z  Z
∗
ψ(p) = hp|ψi = hp| dx |xihx| |ψi = N dp e−ıpx/~ ψ(x).
R R
Requiring that dx |ψ(x)|2 = dp |ψ(p)|2 gives |N |2 = 1/h. We choose N to be real and
obtain
Z
1
ψ(x) = √ dp eıpx/~ ψ(p)
2π~ Z
1
ψ(p) = √ dx e−ıpx/~ ψ(x),
2π~
which demonstrates that ψ(x) and ψ(p) are related by a Fourier transform. For clarity we
will sometimes write φ(p) instead of ψ(p).

1.1.7 Commuting observables

If a state |ψi is an eigenstate of two observables  and B̂ with eigenvalues α and β then

ÂB̂|ψi = αβ|ψi = βα|ψi = B̂ Â|ψi. (1.22)

A necessary and sufficient condition for  and B̂ to have a common complete set of eigen-
states is that  and B̂ commute:

[Â, B̂] = ÂB̂ − B̂ Â = 0. (1.23)

The uncertainty (variance) of an operator is

h(∆Â)2i = hÂ2 i − hÂi2 . (1.24)

R R
1
To see this use |xi = dξ |ξihξ|xi = dξ δ(ξ − x)|ξi, so (1.17) gives hx|p̂|pi = −i~hx| ∂∂x̂ |pi =
R ∂ ∂
−i~ dξ δ(ξ − x) ∂ξ hξ|pi = −i~ ∂x hx|pi.
1.1 Dirac Notation and rules of Quantum Mechanics 7

The generalized uncertainty relation for two noncommuting observables is

1
(∆A)(∆B) ≥ |h[Â, B̂]i| (1.25)
2
q q
where ∆A = h(∆Â) i, ∆B = h(∆B̂)2 i.
2

There are many ways to prove this, see for example the proof in Zettili p. 95 which
invokes the Schwarz inequality. Here is an alternative approach. Define Â0 = Â − hÂi so
hÂ0 i = 0,
h(∆Â0)2 i = hÂ02i = (∆A)2
and similarly for B̂ 0 = B̂ − hB̂i. We note that [Â0, B̂ 0] = [Â, B̂].
Consider an arbitrary ket |ψi and the quantity (Â0 + iλB̂ 0)|ψi with λ real. The square
of the norm is

||(Â0 + iλB̂ 0)|ψi||2 = hψ|(Â0 − iλB̂ 0 )(Â0 + iλB̂ 0)|ψi

= hψ|Â02|ψi + λ2 hψ|B̂ 02|ψi + iλhψ|[Â0, B̂ 0]|ψi
= λ2 (∆B)2 + iλhψ|[Â, B̂]|ψi + (∆A)2
≥ 0.

The last term is the commutator of two Hermitian operators which is anti-Hermitian (it
has imaginary eigenvalues), so i times the commutator is a Hermitian operator. The above
expression is therefore a quadratic function of λ with real coefficients, and is therefore strictly
real. It is non-negative for all values of λ, and therefore must not have two different real
roots, since if that were the case the expression would have to change sign and be somewhere
negative. For this to be so the discriminant must be negative or zero which corresponds to
1
(∆A)(∆B) ≥ |hψ|[Â, B̂]|ψi|.
2
Since this is true for any |ψi we can write
1
(∆A)(∆B) ≥ |h[Â, B̂]i|
2
as desired.
As an example of application of the commutation relation (1.25) consider x̂ and p̂. Their
commutator can be evaluated in the position representation as

hψ|[x̂, p̂]|ψi = hψ|(x̂p̂ − p̂x̂)|ψi

Z  
∗ ∂ ∂
= −i~ dx ψ x − x ψ
∂x ∂x
Z  
∗ ∂ψ ∂ψ
= −i~ dx ψ x −ψ−x
∂x ∂x
Z
= i~ dx ψ ∗ψ
= hψ|i~|ψi.

February 4, 2010 M. Saffman

8 1 Quantum Mechanics

Since |ψi is an arbitrary ket we have [x̂, p̂] = i~ and using (1.25)
~
(∆x)(∆p) ≥
2
which is the usual position-momentum form of the uncertainty relation.

1.1.8 Measurements
P
An arbitrary state is |ψi = ∞ n=0 cn |ni. The result of a measurement of the observable A is
X X XX X
hÂi = hψ|Â|ψi = c∗m hm|Â cn |ni = c∗m cn an hm|ni = |cn |2 an . (1.26)
m n m n n

The probability of the measurement result being the eigenvalue an is

P (an ) = |cn |2 = |hn|ψi|2 = ||P̂n |ψi||2 = hψ|P̂n |ψi. (1.27)

After  has been measured and given the result an the new state of the system is

P̂n |ψi cn
|ψ 0i = = |ni. (1.28)
||P̂n |ψi|| |cn |

A subsequent measurement of hÂi will return an with unit probability.

1.1.9 Evolution in time

Time evolution is governed by the Hamiltonian Ĥ and
d
i~ |ψ(t)i = Ĥ|ψ(t)i. (1.29)
dt
~ 2
For a particle of mass m in the coordinate representation Ĥ = − 2m ∇2 + V (r) and we get
the Schrödinger equation

∂ψ(r, t) ~2 2
i~ =− ∇ ψ(r, t) + V (r)ψ(r, t). (1.30)
∂t 2m
The formal solution for time evolution is

|ψ(t)i = Û (t, t0)|ψ(t0)i

where the time evolution operator is

 Z t 
i 0 0
Û (t, t0) = exp − dt Ĥ(t ) (1.31)
~ t0 +
h i
and [....]+ indicates time ordering of operator products: Â(t1)...Â(tn ) = Â(t1 )...Â(tn )
+
when t1 ≥ t2 ≥ ... ≥ tn .
1.1 Dirac Notation and rules of Quantum Mechanics 9

When Ĥ is not explicitly time dependent this simplifies to

 
i
Û (t, t0) = exp − Ĥ(t − t0 ) . (1.32)
~

The time evolution operator is unitary which means Û † = Û −1 so

Û † Û = Û Û † = Î. (1.33)

The time evolution of an initial state |ψ(t0)i can be determined explicitly by expanding in
energy eigenfunctions Ĥ |ni = En |ni. Assume the initial state
X X
|ψ(t0)i = hn|ψ(t0)i|ni = cn0 |ni
n n

and the time dependent state

X X
|ψ(t)i = hn|ψ(t)i|ni = cn (t)|ni.
n n

Plugging in to the Schrödinger equation gives

X dcn X X
i~ |ni = Ĥcn |ni = En cn |ni.
n
dt n n

Projecting out the mth component by operating on both sides with hm| gives

dcm
i~ = Em cm
dt
which is solved by
cm (t) = cm0 e−ıEm (t−t0 )/~ .
Thus X X
|ψ(t)i = cn (t)|ni = cn0 e−ıEn(t−t0 )/~ |ni.
n n

1.1.10 Schrödinger and Heisenberg pictures

There are several different ways of working with time evolution.
In the Schrödinger picture the state |ψ(t)iS is a function of time while all observables
ÂS are constant. The state vector evolves according to the Schrödinger equation i~ d|ψi
dt
S
=
ĤS |ψiS and the expectation value of an operator at time t is

a(t) = hÂS i
= S hψ(t)|ÂS |ψ(t)iS .

Here the subscripts S refer to the Schrödinger picture.

February 4, 2010 M. Saffman

10 1 Quantum Mechanics

We can cast this in a different form as follows. The expectation value of an operator at
time t is

a(t) = hÂS i
= S hψ(t)|ÂS |ψ(t)iS
= hψ(0)|Û † (t)ÂS Û (t)|ψ(0)i.

We can define a time dependent operator by ÂH = Û † (t)ÂS Û (t), and time independent
states by |ψiH = |ψ(0)i such that

a(t) = H hψ|ÂH |ψiH .

This is the Heisenberg picture in which the kets are stationary but the operators evolve in
time. The equation of motion for the operator is found from

dÂH (t) d  †  ∂ Â
H
= Û (t)ÂS Û (t) +
dt dt ∂t
dÛ † (t) dÛ (t) ∂ ÂH
= ÂS Û (t) + Û † (t)ÂS +
dt dt ∂t
   
i † † −i ∂ ÂH
= Û Ĥ ÂS Û (t) + Û (t)ÂS Ĥ Û (t) +
~ ~ ∂t
i ∂ ÂH
= [Ĥ, ÂH ] +
~ ∂t

where the last lines follows from [Û , Ĥ] = 0, and for completeness we have included the
possibility of an explicit time dependence (not due to Û (t)) in ÂH . Thus

dÂH i ∂ ÂH
= − [ÂH , Ĥ] + , (1.34)
dt ~ ∂t
which is referred to as the Heisenberg equation (although it was first written down by Dirac).
This equation is usually more difficult to solve than the Schrödinger equation. However, it
provides a clear picture of the correspondence between quantum and classical mechanics.

1.1.11 Quantum - Classical correspondence and Ehrenfest’s theo-

rem
In classical mechanics we can describe a dynamical system by coordinates qj and momenta
pj . The Hamiltonian of the system is a function H(qj , pj ) of the coordinates and momenta
which satisfy Hamilton’s equations
∂qj ∂H ∂pj ∂H
= , =− .
∂t ∂pj ∂t ∂qj

For a particle in a potential V (r), H = p2 /2m + V (r) so Hamilton’s equations give dr/dt =
p/m and dp/dt = −∇V (r).
1.1 Dirac Notation and rules of Quantum Mechanics 11

For any two dynamical quantities A, B we define the Poisson bracket {A, B} by
X  ∂A ∂B ∂A ∂B

{A, B} = − .
j
∂qj ∂pj ∂pj ∂qj

Clearly {A, B} = −{B, A}.

The time derivative of the quantity A is
dA X  ∂A ∂qj ∂A ∂pj

∂A
= + +
dt j
∂qj ∂t ∂pj ∂t ∂t

X ∂A ∂H 
∂A ∂H ∂A
= − +
j
∂qj ∂pj ∂pj ∂qj ∂t
∂A
= {A, H} + . (1.35)
∂t
The similarity between the quantum equation (1.34) and the classical equation (1.35) is
remarkable. We see that the quantum equation of motion for the operator  can be found
by writing down the classical equation for the dynamical variable A in terms of Poisson
brackets and making the substitution
i
{A, H} → − [ÂH , Ĥ]. (1.36)
~
Alternatively, we may think of quantum mechanics as the more fundamental theorem
which classical mechanics is an approximation to. In the limit of ~ → 0 the transformation
i
− [ÂH , Ĥ] → {A, H} (1.37)
~
reveals the limiting classical equation of motion. Note that while (1.36) should always lead
to a valid quantum equation, the substitution (1.37) may be meaningless since there are
quantum problems for which no classical analog exists. A prime example is the spin of an
electron.
Finally, we note that the time evolution of the expectation value of a quantum variable has
a close analogy with the time evolution of the corresponding classical quantity. This analogy
is already apparent in equations (1.34,1.35), which when written in terms of expectation
values is referred to as Ehrenfest’s theorem.
To see this note that for any differentiable operator function F̂ = F (q̂j , p̂j ) we have the
following commutators

∂ F̂ ∂ F̂
[q̂j , F̂ ] = i~ , [p̂j , F̂ ] = −i~ .
∂ p̂j ∂ q̂j

From (1.34)with F̂ = Ĥ we obtain

* + * +
d ∂ Ĥ d ∂ Ĥ
hq̂j i = , hp̂j i = − .
dt ∂ p̂j dt ∂ q̂j

February 4, 2010 M. Saffman

12 1 Quantum Mechanics

p̂2
Thus for a particle in potential V (r̂) with Ĥ = 2m
+ V (r̂) we get

d hp̂i d
hr̂i = , hp̂i = −h∇V (r̂)i. (1.38)
dt m dt
The first of Eqs. (1.38) is identical to the corresponding classical equation obtained by
putting hr̂i → r, hp̂i → p. However, the second equation differs from the corresponding
classical equation since, in general,

h∇V (r̂)i 6= ∇V (r)|r=hr̂i .

The difference betweens these two expressions is negligible provided the potential is close to
constant over the extent of the deBroglie wave packet of the particle. This condition must
be satisfied for the evolution of the expectation values of the quantum variables to agree
with the results of a classical description.

1.2 Principles of quantum mechanics

We can summarize the formalism given above in a small set of principles.

1. A physical system is associated with a Hilbert space EH containing ket vectors. At time
t the physical state is completely described by a ket |ψ(t)i residing in EH .

2. Any physical quantity A is associated with a Hermitian operator  that acts on

kets in EH . The result of a measurement of A is always one of the eigenvalues an of Â.
The probability of measuring an is P (an ) = ||P̂n |ψi||2 where P̂n = |nihn| is the projec-
tor onto the ket |ni. After the measurement the system will be in the new state |ψ 0i = ||P̂P̂n |ψi
|ψi||
.
n

3. Time evolution is governed by the Hamiltonian according to i~ d|ψ(t)i

dt
= Ĥ |ψ(t)i.

How to translate these simple rules into explicit procedures is often far from obvious, and
learning how to do so constitutes a course in Quantum Mechanics.