Review of Quantum Mechanics

Jhih-Sheng Wu

2024

Contents

1 Wavefunction 3

2 Dirac Notations 4

3 Postulates of Quantum Mechanics 6

4 Quantum Dynamics: Schrödinger, Interaction, Heisenberg Pictures 11

4.1 Schrödinger Picture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

4.2 Heisenberg Picture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

4.3 Interaction Picture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

5 Harmonic Oscillators 14

5.1 Number States in the Position Bases . . . . . . . . . . . . . . . . . . . . . . . . . 16

5.2 Dynamics of a Harmonic Oscillator . . . . . . . . . . . . . . . . . . . . . . . . . 17

6 Coherence and Decoherence at a Glance 18

6.1 Two Waves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18

6.2 Many Waves and Ensemble . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19

1
7 Density Operator Formulation 19

7.1 Density Operators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20

7.2 Dynamics of Density Operators . . . . . . . . . . . . . . . . . . . . . . . . . . . 26

2
Quantum mechanics is a modern mathematical theory used to describe the quantum phenom-
ena. Although many scientists think it is not the ultimate theory, it is the most accurate theory
today that describes the experiments. Quantum mechanics is formulated under the postulates,
which are made by the scientists to explain the experiments. In classical mechanics, a physical
system consists of physical quantities which have definite values. For examples, the position x
and the momentum p of a particle at any given time t are assumed to be some numbers. On
the contrary, a physical system in quantum mechanics is described by a state |ψi. The notation
|ψi is called a ket. In a closed system, the state |ψi contains all the information of the systems.
The exotic part of quantum mechanics is that even |ψi is complete, the outcome of observed
quantities are still probabilistic.

1 Wavefunction

Let’s use the wavefunction to elaborate the nature of probability. The wavefunction of a particle
is obtained by writing |ψi in the x basis |xi,
ψ(x) ≡ hx|ψi. (1)
For a given wavefunction ψ(x), the probability to find the particle to be at x is |ψ(x)|2 dx. Since
the total probability is one, the normalization of a state requires that
Z
|ψ(x)|2 dx = 1. (2)

The average position hxi (expectation value) of the particle is

Z
hxi = x|ψ(x)|2 dx. (3)

With the definitions, the well known Schrödinger’s equation reads

∂ ~2 ∂2
i~ ψ=− ψ + V (x)ψ. (4)
∂t 2m ∂x2
The representation of a state |ψi is not unique. For example, we can use the momentum basis
|pi to write |ψi,
φ(p) ≡ hp|ψi, (5)
and
Z
|φ(p)|2 dp = 1. (6)

Eq. (4), the Schrödinger’s equation, is only one example of the Hamiltonian. In quantum me-
chanics (or classical mechanics), a Hamiltonian is roughly speaking the total energy of a system.
Thus, the Schrödinger’s equation only describes the kinetic energy and the potential energy of
an electron. If we want to describe the other energies, we need to use the other Hamiltonians.
This will be elaborated in Postulate 2.

3
 Example 1: Plane Wave
Let |pi be the eigenstate of the momentum operator p̂ such that

p̂|pi = p|pi.

The state |pi has a well-defined momentum p. A plane wave is indeed the projection of
|pi onto the x basis.
px
hx|pi = ei ~ .
If instead, we project the state |pi onto the p basis,

hp0 |pi = δ(p − p0 ).

That is, a plane wave in the p space is a delta function. This relation is analogous to the
Fourier transform of a single frequency signal eiωt is a delta function.

2 Dirac Notations

In quantum mechanics, the Bra-Ket notations are convenient tools. Any states are written as
kets |ψ1 i, |ψ2 i, |ψ3 i,... You can think a ket as a column vector. However, the representation of
a column vector depends on the bases. For example, in the position basis, a ket can be defined
as:
 
 ψ(x1 ) 
 ψ(x2 ) 
|ψi =  ..  . (7)
 
 . 
 
ψ(xN )

whereas in the momentum basis

 
 φ(p1 ) 
 φ(p2 ) 
 
|ψi =  .. 
 (8)
 . 
 
φ(pN )

The role of a bra is similar to row vectors in linear algebra. A bra is defined as the hermitian
conjugate of a ket. For examples, in the position basis, a bra can be defined as:
 
hψ| = ψ ∗ (x1 ) ψ ∗ (x2 ) . . . ψ ∗ (xN ) . (9)

whereas in the momentum basis

 
hψ| = φ∗ (p1 ) φ∗ (p2 ) . . . φ∗ (pN ) . (10)

4
The inner product of two states |ψi and |φi is
hψ|φi, (11)
which is a complex number. The inner product hψi |ψi i is the probability to find the particle in
the ith state. The outer product of two states |ψi and |φi is
|φihψ| (12)
which is a matrix.

Note 1: Continuous and discrete variables

A state |ψi can be a finite or an infinite column vector. This will depend on which basis
we express the state in. For example, the state of an electron ψ in the x basis is an
infinite vector, and in the spin basis, it is a 2-component vector. Here, x is the continuous
variable, and spin is the discrete variable. In the cases of light, the photon number is a
discrete variable, and the amplitude of the electric field is a continuous variable. More
specifically, a state |ψi is fully categorized by all its commutable observables. For example,
an electron is fully described by x and spin. Its Hamiltonian is block-diagonalized, where
each block has a spin value.

Exercise 1: Calculation of bras and kets

Let
 
 1 
|ai = 2i  (13)
 
3
 
 
 i 
|bi = 0 (14)
 
2
 

1. What are ha| and hb|?

2. Calculate ha|ai, hb|bi, ha|bi and hb|ai?

3. Calculate |aihb| and |biha|. Are they complex conjugate of each other?

If |ψi is to describe a single particle, the normalization of a state requires the inner product

hψ|ψi = 1 (15)
or in a specific basis
X
|ψi |2 = 1, (16)
i

5
and for a continuous variable like x,
Z
dx|ψ(x)|2 = 1. (17)

In the position basis, the position is a operator x̂ (a matrix).

 
x1 0 0 0 
 0 x2 0 0 
 
x̂ =  ..
 .. . . ..  (18)
 .
 . . . 
0 0 · · · xN


As it should be, the operator x̂ is a diagonal matrix in the position basis. In the Dirac’s notation,
the expectation value of x is

hxi = hψ|x̂|ψi (19)

X
= xi |ψi |2 (20)
i
Z
= x|ψ(x)|2 dx. (21)

3 Postulates of Quantum Mechanics

Postulate 1: State Vector

A physical system is completely described by a complex state vector |ψi in the Hilbert
space. The Hilbert space is a vector space constructed by all the state vectors |ψi whose
inner products are finite.

The state vector |ψi contains all the information. The state vector can be written as a sum of
other (basis) vectors.

X
|ψi = αi |ψi i (22)
i

The probability to find the system in the ith state is |αi |2 . The simplest example is the qubit,

|ψi = α|0i + β|1i. (23)

6
 Figure 1: Bloch sphere.

Without losing the generality, the qubit can be written as (|α|2 + |β|2 = 1)
θ θ iφr
 
iφg
|ψi = e cos |0i + sin e |1i , (24)
2 2
where φg is the global phase, and φr is the relative phase between the |0i and |1i states. Without
comparing with another qubit, the phase ψg does not have much meaning. The degrees of
freedoms of a qubit are given by θ and φr , which correspond to a surface of a sphere. The
space of a qubit is called the Bloch sphere.

Postulate 2: Temporal Evolution

The evolution of a closed quantum state is described by the unitary transformation.

The state |ψ(t 0 )i is related to the state |ψ(t)i by

|ψ(t 0 )i = Û (t, t 0 )|ψ(t)i, (25)

where U (t, t 0 ) is a unitary operator (a matrix), U † U = 1. 1

The postulate comes from the conservation of total probabilities,

hψ(t 0 )|ψ(t 0 )i = hψ(t)|ψ(t)i = 1 (26)

The unitary operator can be written as

1 For the sake of simplicity, I won’t use a hat for an operator all the time unless there will be confusion.

7
 H 0
U (t, t 0 ) = e−i ~ (t −t) , (27)

where H has to be hermitian, H = H† , to make U (t 0 , t) unitary.

Exercise 2: Exponential Function of Matrices

Show that the operator defined by Eq. (27) is unitary. Use the following facts,

• The matrix exponential of a matrix M is defined eM = 1 + M + M2! + ...

2

• eA eB = eA+B if [A, B] ≡ AB − BA = 0. This can be proved by using the above

definition. This equation is a special case of the Baker-Campbell-Hausdorff formula,
which reads

eX eY = eZ
1 1 1
Z = X + Y + [X, Y ] + [X, [X, Y ]] − [Y , [X, Y ]] + · · · ,
2 12 12

• H = H†

The evolution of a state is then given by

H 0
|ψ(t 0 )i = e−i ~ (t −t) |ψ(t)i, (28)
or more generally, if H(t) is time-dependent,
R t0
t H(τ)dτ
0
|ψ(t )i = e −i ~ |ψ(t)i. (29)
Equation (29) is simply a consequence of the unitary evolution. By differentiating Eq. (29), one
can obtain the general form of the Schrödinger’s equation,
∂
i~ |ψi = H|ψi. (30)
∂t
The operator H is called the “Hamiltonian” of the system. The Hamiltonian, coming from the
classical mechanics, typically is the total energy of the system. For example, for a particle, the
p2
Hamiltonian H = 2m +V (x). The case of a particle is only one of the examples. If the systems are
discrete and finite (energy levels), the Hamiltonian is a finite-dimensional matrix. For example,
classically, the energy of a magnetic moment µ in a magnetic filed B is E = −µ·B. The magnetic
moment µ is related to the angular momentum L by µ = γL, where γ is the gyromagnetic ratio.
Quantumly the Hamiltonian is H = −µ · B, and the angular momentum L becomes an operator.
In quantum mechanics, the angular momentum operators do not commute, and they satisfy the
relation,
[Li , Lj ] = iijk ~Lk . (31)

8
In the case of an electron, the angular momentum operator is σ = (σx , σy , σz ), where σi are the
Pauli matrices (2 by 2 matrices). Hence, the dimension of the Hamiltonian is two.

Let’s consider a system with N levels of the energies E1 , E2 , ..., EN . The energy eigenstates,
|Ei i, satisfy
H|Ei i = Ei |Ei i. (32)
The Hamiltonian in the energy bases |Ei i is diagonal
 
E1 0 0 0 
 0 E2 0 0
 
(33)

H =  .. .. . . ..  .
 .
 . . .


0 0 · · · EN


The solution of the time-dependent Schrödinger’s equation (Eq. (30)) is

X Ei
|ψ(t)i = αi e−i ~ t |Ei i, (34)
i

where αi are the coefficients of the initial state in terms of |Ei i.

Postulate 3: Measurement
Quantum measurement (collapse). A measurement makes a system |ψi collapse randomly
into some state |ψi i. The possible outcome states |ψi i depend on the measurements.
For example, if we measure the position of a particle, the outcome states are |xi with
−∞ < x < ∞. A measurement is described by a set of operators {Mm }, where m denotes
all the possible outcome states. After a measurement, the state becomes

Mm |ψi
q (35)
†
hψ|Mm Mm |ψi

with the probability

†
p(m) = hψ|Mm Mm |ψi. (36)

The completeness theorem requires that

X
†
Mm Mm = 1. (37)
m

For example, the measurement operators on a qubit are

M0 = |0ih0| (38)
M1 = |1ih1| (39)

9
 Exercise 3: Qubit Measurement
√
3
The initial qubit state is 12 |0i + 2 |1i.

1. What are the two possible states after a measurement of {M0 , M1 }?

2. What are the probabilities to be the above two states, respectively?

Heisenberg uncertainty principle

Postulate 4: Hermiticity
Any physical observables are Hermitian operators. For example, in the position basis, the
position and momentum operators are

x̂ = x (40)
∂
p̂ = −i~ (41)
∂x

Let  be the physical observable operator. The expectation value of  of a state |ψi is

hAi = hψ|Â|ψi. (42)

The state |pi is the eigenvector of the momentum operator p̂,

p̂|pi = p|pi, (43)

and for the position operator x̂,

x̂|xi = x|xi. (44)

Note that the eigenvectors of a Hermitian operator form a complete set of bases of the
space.

The eigenvectors |Ai i of A forms a complete set of bases of the state space. The eigenstates are
orthogonal and normal,

hAj |Ai i = δij . (45)

That is, any state |ψi can be written as

X
|ψi = αi |Ai i. (46)
i

10
The completeness implies that the identity 1 is,
X
1= |Ai ihAi | (47)
i

The standard deviation of A is σ (A),

p
σ (A) ≡ hA2 i − hAi2 . (48)

Two operators A and B are compatible if their commutator [A, B] ≡ AB − BA = 0. Otherwise,

they are incompatible. If

[A, B] = c (49)

and c is a number, the general uncertainty principle reads

|hψ|[A, B]|ψi| |c|
σ (A)σ (B) ≥ = . (50)
2 2

Exercise 4: Uncertainty Principle

Prove the Heisenberg uncertainty principle, Eq. (50). Hint: use the Cauchy-Schwarz
inequality.

hψ|ψihφ|φi ≥ |hψ|φi|2 , (51)

where |ψi and |φi are two states.

The most classical example of the uncertainty principle is about x and p,

[x, p] = i~. (52)

The uncertainty principle reads

~
σ (x)σ (p) ≥ . (53)
2

4 Quantum Dynamics: Schrödinger, Interaction, Heisenberg

Pictures

The state |ψi of a physical system contains all the informations, but is not the direct observable.
We do not directly see or measure the state |ψi, but rather we measure the physical quantities

11
such position, momentum, and so on. These physical quantities are the expectations of the
observable operators (such as x̂ and p̂). For example, when we measure the position of a state,
the averaged outcome is
x(t) = hψ(t)|x̂|ψ(t)i.
This outcome is called the observable.

More generally, let’s consider a physical observable operator Â, and its expectation is

A(t) = hψ(t)|Â|ψ(t)i (54)

The number A(t) is what we measured in the experiment. To obtain the evolution of A(t)
belongs to the subject of quantum dynamics. In quantum dynamics, there are three main
pictures to solve the problems.

4.1 Schrödinger Picture

Consider that the observable operator A is static and the states |ψ(t)i is evolving.
iHt
|ψ(t)i ≡ |ψ(t)iS = e− ~ |ψ(0)i (55)

The expectation value in this picture is

hA(t)i = hψ(t)|A|ψ(t)i (56)

4.2 Heisenberg Picture

Consider that the observable operator A(t) is dynamic and the states |ψ(t)i is static.
iHt iHt
Ah ≡ A(t) = e ~ Ae− ~ , (57)

and the expectation value is

hA(t)i = hψ(0)|Ah |ψ(0)i. (58)

The evolution of Ah follows the Heisenberg’s equation,

∂Ah
i~ = [Ah , H]. (59)
∂t

12
 Exercise 5: Proof of the Heisenberg’s equation
iHt
Let U (t) = e− ~ so that Ah = U † AU . Differentiating Ah with respect to t gives

∂Ah ∂U † ∂U
= AU + U † A (60)
∂t ∂t ∂t
First, show that the derivative of U (t) is

∂
i~ U (t) = HU (t). (61)
∂t
Use the two above equations to prove the Heisenberg’s equation.

Example 2: Harmonic Oscillator in Heisenberg Picture

The Hamiltonian of a harmonic oscillator is
p2 mω2 x2
H= + .
2m 2
In this case, the position operator in the Heisenberg Picture is
iHt iHt
x̂h = e ~ x̂e− ~ .

At t = 0, x̂h (0) = x̂(0). The operator x̂h (t) at a later time, can be obtained by

∂x̂h
i~ = [x̂h , H].
∂t
You can show that
∂x̂h
= −mω2 x̂h .
∂t

4.3 Interaction Picture

When the Hamiltonian includes two terms: one is the original Hamiltonian H0 and the inter-
action with the external system V (t), it is convenient to use the interaction picture, where both
the states and the operator are evolving. The total Hamiltonian is H = H0 + V (t). The state
|ψiI is
H0 t
|ψiI = ei ~ |ψ(t)iS , (62)

13
and the operator AI is
H0 t H0 t
AI = ei ~ Ae−i ~ , (63)

The Schrödinger equation becomes

∂
i~ |ψi = VI (t)|ψiI , (64)
∂t I
H0 t H0 t
VI (t) ≡ ei ~ V (t)e−i ~ . (65)
VI
Note that the solution to Eq. (64) is not |ψ(t)iI = e−i ~ t |ψ(0)iI because the VI (t) is time-
dependent. The solution to to Eq. (64) is

|ψ(t)iI = UI (t, t0 )|ψ(t0 )iI (66)

i t
Z
UI (t, t0 ) = 1 − VI (t 0 )UI (t 0 , t0 )dt 0 (67)
~ t0

The Heisenberg’s equation becomes

∂AI
i~ = [AI , H0 ]. (68)
∂t

5 Harmonic Oscillators

The Hamiltonian of a simple harmonic oscillator is

p2 mω2 x2
H= + , (69)
2m 2
√
where ω = k/m and k is the spring constant. We define the creation operator a† and the
annihilation operator a,
r
mω ip
 
a= x+ , (70)
2~ mω
r
mω ip
 
†
a = x− . (71)
2~ mω

Exercise 6: Commutation Relation

Show that

[a, a† ] = 1. (72)

14
 Use the relation [x, p] = i~.

The Hamiltonian is rewritten as

1
 
H = ~ω a a + †
(73)
2
1
= ~ω(N + ) (74)
2
where N = a† a is the number operator. The eigenvector of N is |ni

N |ni = n|ni, (75)

where n is the eigenvalue. The number states are orthonormal

hm|ni = δmn . (76)

Important indentities are

[N , a] = −a, (77)
[N , a† ] = a† , (78)

As a result, we have
 
N a† |ni = a† N + a† |ni = (n + 1)a† |ni, (79)
N a|ni = (aN − a) |ni = (n − 1)a|ni, (80)

These eqautions imply that

a|ni = c− |n − 1i, (81)

a† |ni = c+ |n + 1i, (82)

The constants c− and c+ can be fixed by noting that

hn|a∗ a|ni = n = |c− |2 , (83)

hn|aa∗ |ni = n + 1 = |c+ |2 . (84)
√ √
Taking c− and c+ to be positive by convention, c− = n and c+ = n + 1. We have the important
relations which explain the names, creation and annihilation,

√
a|ni = n|n − 1i, (85)
√
a† |ni = n + 1|n + 1i. (86)

15
 Note 2: Representation in the number basis
The number n is the number of the energy quanta. The smallest number of n is n = 0.
The physical meaning of |ni is a state containing n energy quanta. Thus, |ni is called the
number state. The energy of a harmonic oscillator is
1
 
En = n + ~ω (87)
2

The 12 ~ω is interpreted as the vacuum energy since it exists even when n = 0. Applying
√
a creation operator on the |ni, the state |ni becomes n + 1|n + 1i, that is, the a† will
create one single quantum to the original state. Similarly, the a will annihilate one energy
quantum from the system. We can also prove that

(a† )n
|ni = √ |0i. (88)
n!
The position operator x and momentum operator p can be expressed as
r
~
x= (a + a† ) (89)
2mω
r
mω~
p=i (−a + a† ) (90)
2

5.1 Number States in the Position Bases

As the familiar wave function ψ(x), we can express the |ni in the x bases. The wavefunctions
are ψn (x) ≡ hx|ni. Let’s solve the ground states first ψ0 (x). We start with

a|0i = 0 (91)
⇒hx|a|0i = 0 (92)
r
mω ip
 
⇒ x x+ 0 =0 (93)
2~ mω
!
~ ∂
⇒ x+ ψ (x) = 0 (94)
mω ∂x 0
 2
1 − 12 xx
⇒ψ0 (x) = 1/4 √ e 0 , (95)
π x0
q
where x0 = ~
mω

16
 Figure 2: Wavefunction ψn (x).

Exercise 7: Uncertainty of the ground state

Show that for the ground state ψ0 (x), the uncertainty relation has a equal sign, that is,
the state has the minimum uncertainty,

~
σ (x)σ (p) = . (96)
2

Using Eqs. (88) and (95), we obtain the expression for φn (x),
  !n  2
1 2 ∂ − 21 xx
(97)
 
ψn (x) =  √  x − x0
 e 0 .
π1/4 2n n!x0n+1/2 ∂x

5.2 Dynamics of a Harmonic Oscillator

The Heisenberg’s Equations of a and a† (t) are

da†
i~ = [a† , H] = −~ωa† , (98)
dt
da
i~ = [a, H] = ~ωa, (99)
dt
(100)
whose solutions are
a(t) = a(0)e−iωt , (101)
iωt
† †
a (t) = a (0)e . (102)

17
In terms of x and p, Eqs. (101) and (102) read
!
ip(t) ip(0) −iωt
a(t) = x(t) + = x(0) + e , (103)
mω mω
!
ip(t) ip(0) iωt
†
a (t) = x(t) − = x(0) − e . (104)
mω mω

Solving the equations for x(t) and p(t), we have

p(0)
x(t) = x(0) cos ωt + sin ωt, (105)
mω
p(t) = −mωx(0) sin ωt + p(0) cos ωt. (106)

Note 3: Heisenberg picture of x and p of a harmonic

oscillator
Equations (105) and (106) are exactly the same as the equations of motion derived from the
classical mechanics. In contrast, x(0) and p(0) are operators. If we take the number state
|ni, the expectation value hn|x(t)|ni vanishes. We will not observe an expectation value
hx(t)i obeying the classical motion. It turns out that the state mostly close to a classical
state is the coherent state |λi, which is the eigenvector of the annihilation operator a,

a|λi = λ|λi. (107)

We will talk more about the coherent states later.

6 Coherence and Decoherence at a Glance

Coherence refers to many meanings in different circumstances. We consider its usages in the
context of physics. Roughly speaking, coherence means that two (or more than two) states
(waves, particles) have a well-defined correlation as time t or positions x change.

6.1 Two Waves

For instance, there are two waves with the complex amplitudes ψ1 (x, t) and ψ2 (x, t), respec-
tively. In terms of the absolute amplitudes A and phases φ, they can be written as ψ1 (x, t) =
A1 (x, t) exp(iφ1 (x, t)) and ψ2 (x, t) = A2 (x, t) exp(iφ2 (x, t)). Coherence means that the differ-
ence between their phases φ1 − φ2 is constant as t or x changes. In the two-slit experiments,
coherent light source is required to produce interference patterns. In this definition, two plane
waves exp(ik1 x − ω1 t + φ1 ) and exp(ik2 x − ω2 t + φ2 ) are coherent if k1 = k2 and ω1 = ω2 . The

18
reason for the same frequencies or the same wavenumber is as following. When we measure the
interference, frequently we collect the data for a long time over many periods . The interference
signal is the time-average of the product ψ1∗ (t)ψ2 (t). |ψ1 (t)|2 and |ψ2 (t)|2 are the background
intensity. The time-averaged interference is
RT h i
∗
2Re ψ1 (t)ψ2 (t)
Iinterference = lim 0 . (108)
T →∞ T
If the two waves have different frequencies, the time-average vanishes.

Another question is that are any two waves fully coherent if they have the same frequencies.
The answer is not necessary. Why? It is because the phases φ1 and φ2 can fluctuate. The
coherence implies that δ = φ1 − φ2 is a constant as time t and position x changes. In practical
situations, as the waves propagate, the environment provides noises to the phases. As a result,
the time-average becomes smaller. This process is called “decoherence” Typically, a system
gradually loses its coherence as t increases or traveled length x increases.

6.2 Many Waves and Ensemble

A more realistic system consists of many waves (states, particles)2 ,

ψ(x, t) = ψ1 (x, t) + ψ2 (x, t) + ψ3 (x, t) + ... (109)

The interference involves all the cross-product terms ψ1∗ (t)ψ2 (t), ψ2∗ (t)ψ3 (t), ψ3∗ (t)ψ4 (t), and so
on. To deal with a system containing a large number of waves (particles), it is more convenient
to use a statistical tool than listing all the states. The idea is to use probabilities to describe
distributions of states. Such a tool in quantum mechanics is called the density operator or
matrix. Basically, a density matrix contains the information of the probability in each state.

7 Density Operator Formulation

Let’s introduce the concepts of an ensemble.

Note 4: Ensemble
An ensemble is a statistical tool to describe a system of many particles. An ensemble
consists of a large number (ideally, infinite) of virtual copies of a particle. Each copy
represents a possible state that a particle can be in. A specific ensemble is specified

2 We use the terms “particle”, “waves”, and “states” interchangeably.

19
 by assigning the probability in each state. For example, a photon state |photoni is
decomposed as

|photoni = α|Li + β|Ri, (110)

where |Li (|Ri) is the left(right)-polarized state. In an ensemble, there are many photons.
Let pL and pR be the probabilities of the left-polarized state and the right-polarized state,
respectively, where

pL + pR = 1. (111)

The probabilities pL and pR specify the ensemble where there are infinite particles, and
each has the probability pL (pR ) in the state |Li (|Ri).
However, we can not use the following expression to describe an ensemble,
hhhh (
((((
|ensemblei = p |Li + p |Ri, (112)
hh( h ( (
( ( h L hhh
h R
(((( hh
since this expression is used for one single state. The mathematical tool to describe an
ensemble is the density matrix ρ̂,

ρ̂ = pL |LihL| + pR |RihR|. (113)

7.1 Density Operators

In a general case, the density matrix ρ̂ can be defined as

X
ρ̂ = pi |ψi ihψi |, (114)
i
X
pi = 1, (115)
i
0 ≤ pi ≤ 1, (116)
where |ψi i are the basis states, and pi is the probability to find a particle in the ith state.
Most times, |ψi i are chosen to be orthonormal vectors. In the definition by Eq. (114), the density
matrix is a diagonal matrix. In general, a density matrix can have nonzero off-diagonal elements
in another basis. Consider a new set of orthonormal basis vectors |ai i, obtained by the unitary
transformation
X
|ai i = Uij |ψj i. (117)
j

where U † U = 1. The matrix element Uij can be obtained explicitly by multiplying hψj 0 | on the
both sides of Eq. (117),
Uij = hψj |ai i. (118)

20
The inverse transforms are
X
|ψi i = Uij† |aj i, (119)
j
X
hψi | = Uj 0 i haj 0 |. (120)
j0

In the new basis,

X
ρ̂ = pi |ψi ihψi | (121)
i
  
X X  X 
= pi  Uij† |aj i  Uj 0 i haj 0 | (122)
  
  0 
i j j
 
X X 
= 
 pi Uij† Uj 0 i  |aj ihaj 0 | (123)
j,j 0 i
X
≡ ρjj 0 |aj ihaj 0 |, (124)
j,j 0

where the element ρjj 0 is given by

X
ρjj 0 = Uj 0 i pi Uij† (125)
i
 T
= UP U† , (126)

where P is a diagonal matrix whose diagonal elements are pi . In the new bases |aj i, the off-
diagonal element ρjj 0 can be nonzero. Indeed, the off-diagonal element ρjj 0 is related to the
correlation between the two states |aj i and |aj 0 i.

Example 3: Two-Level System

Let’s work out an example of a two-dimensional density matrix. Consider an ensemble of
the density matrix

ρ = 0|LihL| + 1|RihR|, (127)

or in the matrix form

!
0 0
ρ= . (128)
0 1

21
Keep in mind that the matrix in Eq. (128) is written in the basis vector defined by
!
1
≡ |Li, (129)
0
!
0
≡ |Ri. (130)
1
Now we consider the new bases |Xi and |Y i (linear polarized states)
1
|Li = √ (|Xi − i|Y i) , (131)
2
1
|Ri = √ (|Xi + i|Y i) . (132)
2
The unitary transformation is
! !
|Li † |Xi
=U , (133)
|Ri |Y i
where
1 −i 

 √ √
U † =  12 i
2 , (134)
√ √ 

 12 2
√1 

 √ 
U =  i2 −i
2  . (135)
√ √ 
2 2

In the new bases, using Eq. (126) the density matrix is

−i T
 1
√1 
 ! 1 
 √ 0 0  √2 √
ρ̂ =  i2 −i
2 2 (136)
0 1  √1 i

√ √  √ 
 
2 2 2 2
1 −i
!
= 2
i
2
1 . (137)
2 2

Note that the matrix in Eq. (137) is written in the new basis (see Eq. (124)). For this matrix,
the column vectors (1, 0)T and (0, 1)T represent the new basis vectors |a1 i and |a2 i, which
in this case are
!
1
≡ |Xi, (138)
0
!
0
≡ |Y i. (139)
1

The off-diagonal element ρY X = √i reflects the fact that for a right-circular-polarized

2
state |Ri, the phase difference between |Xi and |Y i is π/2 (the phase factor exp iπ
2 = i).

22
 Exercise 8: Density Matrix
Consider an ensemble of the density matrix
1 3
ρ = |LihL| + |RihR|. (140)
4 4
Calculate the density matrix in the bases |Xi and |Y i.

If we measure an observable A of the ensemble, the expectation value is called “ensemble

average”,
X
hAi = pi hψi |A|ψi i (141)
i
X
= pi hψi |ψj ihψj |A|ψi i (142)
i,j
X X
= hψj | A |ψi ipi hψi |ψj i (143)
j i
| {z }
This term is Aρ
X
= hψj |Aρ|ψj i (144)
j
= Tr (Aρ) . (145)

Although we derive the ensemble average Eq. (145) in the |ψi i bases, the trace of a matrix is
independent of the bases. Thus, Eq. (145) is valid in any basis. This basis-free property is the
advantage of using a trace. One direct application is when A = 1,
X
Tr (ρ) = pi = 1, (146)
i

which tells the trace of a density matrix is the total probability.

Exercise 9: Properties of Density Matrix

Some important properties of density matrix are listed below:

(a) ρ = ρ†

(b) Tr (ρ) = 1
 
(c) 0 < Tr ρ2 ≤ 1

Prove that the above properties are true in any set of bases.

23
Note 5: Pure and Mixed Ensemble
We start with the bases |ψi i, where ρ is diagonal. A pure ensemble is specified by pi = 1
of some state |ψi i and all other pj = 0 for j , i. The equivalent condition of a pure
ensemble is
 
Tr ρ2 = 1, (147)

which applies to a density matrix in any basis. The condition of a mixed ensemble is
 
Tr ρ2 < 1. (148)

One particle state is always a pure ensemble. One common mistake is to be confused by
the superposition of one particle and the mixed ensemble.3 Consider a one-particle state
(qubit) composed of the superposition of |0i and |1i.

|ψi = α|0i + β|1i. (149)

One might think that this state has a density matrix ρ = |α|2 |0ih0| + |β|2 |1ih1|. But, this
is wrong! The correct density matrix is

ρ = |ψihψ| (150)
= (α|0i + β|1i) (α ∗ h0| + β ∗ h1|) (151)
= |α|2 |0ih0| + |β|2 |1ih1| + αβ ∗ |0ih1| + α ∗ β|1ih0| (152)
!
|α|2 αβ ∗
= ∗ , (153)
α β |β|2

where
!
1
≡ |0i, (154)
0
!
0
≡ |1i. (155)
1

It is possible to find the bases where ρ is diagonal, since ρ is a hermitian matrix. The
off-diagonal elements in Eq. (153) describe the correlations between the states |0i and |1i.
An example of a mixed ensemble of qubits is

ρ = |α|2 |0ih0| + |β|2 |1ih1| (156)

!
|α|2 0
= , (157)
0 |β|2

where both |α|2 and |β|2 are nonzero. In this mixed ensemble, the off-diagonal elements
are zero. This means that there is no correlation between the states |0i and |1i.

24
If the number of bases is N , the most random mixed ensemble is
N
1X
ρMR = |ψi ihψi | (158)
N
i=1
1
= 1N ×N , (159)
N
where 1N ×N is the N -by-N identity. The off-diagonal elements of the ensemble ρMR are
always zero, i.e., there is not any correlation between the basis states. It can be shown
that an entangled system must be in a mixed ensemble. The extent of entanglement of
an ensemble is somewhat related to von Neumann entropy.

Exercise 10: Pure Ensemble

Which density matrices are pure ensembles?

(a)
!
0.5 0
ρ=
0 0.5

(b)

cos2 θ
!
cos θ sin θ
ρ=
cos θ sin θ sin2 θ

(c)
!
cos2 θ cos θ sin θeiφ
ρ=
cos θ sin θe −iφ sin2 θ

(d)
1
!
cos2 θ 2 cos θ sin θ
ρ= 1 2
2 cos θ sin θ sin θ

3 In many places, mixed states are called instead of mixed ensemble, although the latter is properer.

25
7.2 Dynamics of Density Operators

First, the density operator is not an observable, so we can not use the Heisenberg’s picture to
obtain its dynamics. Let’s begin with a density matrix in the diagonal form,
X
ρ(t) = pi |ψi (t)ihψi (t)|, (160)
i

where the dynamics of the states can be obtianed with Schrödinger Picture

∂
i~ |ψ (t)i = H|ψi (t)i, (161)
∂t i
∂
−i~ hψi (t)| = hψi (t)|H. (162)
∂t
Using Eqs. (160), (161) and (162), we obtain

∂ρ(t) i
= [ρ(t), H] . (163)
∂t ~
This equation is known as the von Neumann equation
 or quantum Liouville equation. Equa-
tion (163) describes a closed system where Tr ρ is a constant in time. This means that the
2

coherence of the system is not changed. How could a system have dissipation and decoherence?
When a system is open to the environment, the interaction between the system and the environ-
ment leads to dissipation and decoherence. The idea is to write H = Hsys + Henv and to derive
a equation only about the reduced density matrix

∂ρ(t)sys ih i
= ρ(t)sys , Hsys + environment terms, (164)
∂t ~
where the reduced density matrix is obtained by the partial trace

ρ(t)sys = Trenv (ρ(t)). (165)

There is not a unique answer how to write the environment terms since that depends on what
kind of environment it is and the interaction. The discussions of the environment terms belong
to the subject “Open Quantum Systems”, which is beyond the scope of the note. We will adopt
the phenomenological methods later.

Exercise 11: Quantum Liouville Equation

Derive the von Neumann equation, Eq. (163). The first step is to differentiate Eq. (160).

26
Example 4: Dynamics of a Two Level System
Let the unperturbed Hamiltonian be
!
Ec 0
H= , (166)
0 Ev
and write the density matrix in this basis
!
ρcc ρcv
ρ= . (167)
ρvc ρvv
Using the von Neumann equation, Eq. (163), we can obtain four first-order differential
equations. Two of them are redundant because ρcc + ρvv = 1 and ρcv = ρvc
∗ . We need only

two equations
∂
ρ = 0, (168)
∂t cc
∂ i
ρcv = ρcv (Ev − Ec ) , (169)
∂t ~
with the solutions
ρcc (t) = ρcc (0), (170)
ρvv (t) = ρvv (0), (171)
ρcv (t) = ρcv (0)e−iωcv t , (172)
(173)

with ωcv = Ec −E
~ . The populations ρcc and ρvv are unchanged in an unperturbed system.
v

The off-diagonal element ρcv has a constant amplitude and a linearly-growing phase in
time. This means that the coherence of the system is unchanged. In a realistic situation,
the system will be dephased. A phenomenological way to add the dephasing is to add
−γρcv in Eq. (169),
∂ i
ρcv = ρcv (Ev − Ec ) − γρcv , (174)
∂t ~
with the solution
ρcv (t) = ρcv (0)e−iωcv t−γt , (175)
(176)
and γ is called the dephasing rate. The time T2 = γ1 is the characteristic time of dephas-
ing. If the system is open, the probability of the top level ρcc can decay. This can be
described as
∂
ρ = −γr ρcc , (177)
∂t cc
where γr is the relaxation rate. The time T1 = 1
γr is the characteristic time of relaxation.

27
 Exercise 12: Quantum Liouville Equation
Derive Eqs. (168) and (169).

References
[1] R. Shankar, Principles of Quantum Mechanics, 1994

[2] J. J. Sakurai, Modern Quantum Mechanics, 1994 and 2010

[3] Michael A. Nielsen and Isaac L. Chuang, Quantum Computation and Quantum Information,
2000

28