Quantum Information (WiSe 22/23) Last updated: 2022-10-26

Chapter 1.1: Basic Concepts in Quantum Information

Single Systems
Instructor: Dr. Christian Schilling Scriber: Ignacio Cirac, Sirui Lu, and Rahul Trivedi

1 Introduction
In this chapter, we will begin to develop the basic tools to describe a quantum system. We will
keep our treatment abstract, and make no reference to a specific physical system. We will also
restrict ourselves to finite-dimensional systems — a number of experimentally relevant physical
systems can be approximated well by finite-dimensional systems, and making this approximation
makes the mathematics surrounding their analysis a lot simpler. Before delving into a description
of a finite-dimensional quantum system, we briefly define classical systems with finite configuration
spaces.

Classical configurations A classical system with a configuration space of size d < ∞ is one which
can only be in d distinct configurations.

Quantum theory In quantum theory, to describe a system with d configurations, the first step
is to associate it with a finite-dimensional Hilbert space H = Cd by identifying each configuration
with the vector of an ONB of the Hilbert space.

Construction of Hilbert space Given a system with d configurations labelled as {1, 2 . . . d}, a
(quantum) Hilbert space H ∼ = Cd associated with the system identifies the configuration i with a
vector |ii ∈ H such that {|1i , |2i . . . |di} is an orthonormal basis for H.
Some examples of such systems are provided below:
1. A optical pulse as described by its polarization is a system with two configurations cor-
responding to two orthogonal polarizations in the plane perpendicular to the direction of
propagation.
2. An electron spin is a system with two configurations, spin up or spin down.
3. A finite-level atom is an example of a system with a finite configuration space, with its
different energy levels being the different possible configurations.

Equipped with this Hilbert space, we will now present the postulates of quantum mechanics,
which will deal with how to: (i) specify possible states of a quantum system, (ii) describe and
compute the results of measurements on the system, and (iii) describe the dynamics of the system.

2 Quantum States
2.1 Pure states
Postulate I A pure state of a quantum system is associated to a vector |ψi ∈ H, a Hilbert space,
with k|ψik = 1. Additionally, two vectors related by a global phase describe the same quantum
state, i.e. |ψi and eiϕ |ψi describe the same quantum state1 . Thus, we will often ignore the global
phases of states.

1A mathematical consequence and detail is that in quantum mechanics we are in fact working with projective
Hilbert space.

1
Chapter 1.1 Quantum Information (WiSe 22/23)

Superposition An important and physically relevant consequence of the quantum mechanical

description of a system’s state is that of superposition, which can be illustrated with pure states
alone. Recall that a classical system with a finite configuration space with d configurations can only
be in one of the d configurations. Consider two such configurations, 1 and 2, then as we described
above, these configurations correspond to two states |1i and |2i in the Hilbert space corresponding
to the quantum system. Since any vector |ψi ∈ H describes a physically valid (pure) state of
the system, we can construct states of the form |ψi = c1 |1i + c2 |2i ∈ H for c1 , c2 ∈ C and
|c1 |2 + |c2 |2 = 1 — these states are called ‘superpositions’ of the states |1i and |2i and have no
classical counterparts. We will see the consequences of this quantum mechanical feature several
times throughout this course.

Single qubit states The qubit is the “quantum” version of a classical bit that can be described
by two configurations 0 and 1. The Hilbert space of the qubit H = C2 , and the two classical
configurations are associated with |0i , |1i ∈ H. This is one of the most important quantum
systems that we will consider and routinely use throughout this course. A general pure state of
the qubit can be described by the vector

|ψi = c0 |0i + c1 |1i , (2.1)

where c0 , c1 ∈ C with |c0 |2 +|c1 |2 = 1 (normalization). However, this representation is often tedious
since c0 and c1 are constrained by the normalization condition, and there is also a redundancy of
the global phase in describing a pure state. A convenient representation that gets around these
issues and also provides a geometrically visualizable representation of the qubit is the Bloch sphere
representation.

Bloch sphere representation A pure state of a qubit can be uniquely identified with a vector
|ψ(θ, ϕ)i with θ ∈ [0, π] and ϕ ∈ [0, 2π) and
   
θ θ
|ψ(θ, ϕ)i = cos |0i + e sin
iϕ
|1i . (2.2)
2 2

In order to show that, let us consider a qubit pure state described by a vector |ψi = c0 |0i + c1 |1i
where c0 , c1 ∈ C and |c0 |2 +|c1 |2 = 1. Note that since multiplication of this vector by a global phase
yields the same pure state, we can assume c0 ∈ [0, ∞). Furthermore, since |c0 |2 = 1 − |c1 |2 ≤ 1, it
follows that c0 ∈ [0, 1] — consequently, there is a unique angle θ ∈ [0, π] such that c0 = cos(θ/2).
This choice also implies that |c1 |2 = 1−|c0 |2 = sin2 (θ/2), and therefore for some unique ϕ ∈ [0, 2π),
c1 = sin(θ/2)eiϕ . 

Some remarks about the Bloch sphere:

1. The Bloch state |ψ(θ, ϕ)i can be visualized to lie on the surface of a 3D sphere of radius 1
(called the Bloch sphere), with (θ, ϕ) being the angles of the point in the spherical coordinates.
Conversely, due to the uniqueness of the Bloch representation, every point on the surface of
the Bloch sphere corresponds to a unique pure state.
2. Some important pure states that we will encounter in practice:
√ √
• |+i ≡ |0ix = (|0i + |1i)/ 2, |−i = |1ix = (|0i − |1i)/ 2. These correspond to the
points of intersection between the x axis and Bloch sphere.
√ √
• |0iy = (|0i + i |1i)/ 2, |1iy = (|0i − i |1i)/ 2. These correspond to the points of
intersection between the y axis and the Bloch sphere.

Vector representation We can represent the states as column vectors with d complex components.
For instance, for a qubit
cos(θ/2)
   
c0
|ψi → → . (2.3)
c1 sin(θ/2)eiϕ

2
Chapter 1.1 Quantum Information (WiSe 22/23)

2.2 Mixed states

Motivation In some settings we can only associate a vector |ψi i ∈ Cd with some classical prob-
ability pi . Imagine we put in a box 1 qubit in state |0i and another one in state |1i, and then
we take one out of the box. What is its state? With probability 1/2 it will be |0i and with the
same probability, |1i. Such states are called mixed states, and are mathematically described by an
operator X
ρ= pi |ψi i hψi | . (2.4)
i

Postulate I’ To any state of a system we associate a density operator, ρ ∈ L(H), fulfilling

1. It is Hermitian (ρ = ρ† ).

2. It is positive semi-definite (ρ ≥ 0).

3. It has trace 1 (tr(ρ) = 1).
This postulate is more general than the previous one.

Remarks
1. A pure state corresponding |ψi ∈ Cd is described by a density matrix ρ = |ψi hψ| — thus, a
pure state has a density matrix representation which is a rank 1 projector.

2. Each operator ρ satisfying the conditions in Postulate I’ is a different state, i.e. this represen-
tation has no redundancy. Thus, if ρ1 6= ρ2 , then the states they describe are also physically
different and can be distinguished by measurements (we will see how in the section on mea-
surements).

3. Consider a mixed state ρ, then since ρ ≥ 0, ρ = ρ† and tr(ρ) = 1, it can be diagonalized and
expressed as
Xd
ρ= λi |ψi i hψi | , (2.5)
i=1

where λi ∈ [0, 1), |ψi i ∈ Cd are normalized eigenvalues and eigenstates of ρ with i λi =
P
1. This representation can be given a physical interpretation — it is to be understood
as an ensemble of pure states, as described by the vectors |ψ1 i , |ψ2 i . . . , with probabilities
λ1 , λ2 . . . . More concretely, this is the state that we would obtain as a result of an experiment,
which with probability λi prepares the state |ψi i.

4. Alternatively, consider an experiment that prepares a pure state |φi i with probability pi for
i ∈ {1, 2 . . . N }. The resulting state of the quantum state can be written as
N
X
ρ= pi |φi i hφi | . (2.6)
i=1

It can be verified that ρ is a valid density matrix by verifying the conditions of Postulate I’.
Furthermore, we point out that the states |φi i need not be orthogonal to each other for ρ
to be a valid density matrix (which was the case for the representation of ρ as an ensemble
in the previous point with its eigenstates). This implies that there can be many different
representations of a mixed state as an ensemble of pure states.

Convexity of mixed states The set of density matrices is convex i.e. if ρ1 , ρ2 ∈ D(H) are mixed
states and p ∈ [0, 1], then (1 − p)ρ1 + pρ2 is also a valid mixed state.

3
Chapter 1.1 Quantum Information (WiSe 22/23)

Remarks

1. This result can easily be extended to convex combination of morePthan 2 density PMmatrices
M
i.e. ∀ density matrices ρ1 , ρ2 . . . ρM , and p1 , p2 . . . pM ∈ (0, 1) with i=1 pi = 1, i=1 pi ρi is
also a valid density matrix.
2. The convexity of the set of density matrices can be given a physical interpretation — suppose
we consider an experiment that prepares the state ρ1 with probability p1 , ρ2 with probability
p2 and so on, then the state that is prepared by the experiment is a convex combination of
the states ρ1 , ρ2 . . . , which is again a valid quantum state.

Mixedness Recall that a state ρ is mixed (i.e. not pure) if it has a rank > 1 — the rank of ρ
can be considered as an indicator of how mixed the state is. Equivalently, if it has more than one
non-zero eigenvalues. However, an issue with this measure is that there could be a state which is
very close to pure but still have a rank > 1 (for instance, the state 0.999 |0i h0| + 0.001 |1i h1| has
a rank 2), which makes the rank not a good measure of the purity of a typical state. Below, we
define two measures of how pure or mixed the state is that are used frequently in applications.
Pd
• Purity: The purity of a state ρ is given by tr(ρ2 ) = i=1 λ2i where λ1 , λ2 . . . λd are eigenvalues
of ρ.
• Von Neumann entropy: The von Neumann entropy of a state ρ ∈ D(H) is given by S(ρ) =
Pd
− tr(ρ log2 (ρ)) = − i=1 λi log2 (λi ) where λ1 , λ2 . . . λd are the eigenvalues of ρ and 0 log2 0 =
0.
Both trace purity and von Neumann entropy measure how pure or mixed the state is. In particular,

1. For all mixed states ρ, 0 ≤ S(ρ) ≤ log2 (d) and 1/d ≤ tr(ρ2 ) ≤ 1. We leave a proof of these
bounds as an exercise for the reader.
2. The lower bound for the von Neumann entropy and the upper bound on trace purity is
achieved if the state ρ is pure i.e. S(ρ) = 0 and tr(ρ2 ) = 1. The upper bound of von
Neumann entropy and the lower bound on trace purity is achieved if the state ρ = 1/d (this
is also called the maximally mixed state). Thus a larger von Neumann entropy and a lower
trace purity indicate a more mixed state.

2.3 Pauli operators

An ONB of operators One can define

P0 = |0ih0|, σ+ = |1ih0|,
(2.7)
P1 = |1ih1|, σ− = |0ih1|.

The first two are projectors (i.e., Pi2 = Pi ), whereas the second two are called excitation (σ+ )
and deexcitation (σ− ) operators, respectively. The identity can be written as 1 = P0 + P1 , and
therefore one can express any operator acting on the two–level system as a linear combination of
these operators, in particular the density operator , e.g.
X
ρ= ρi,j |iihj| (2.8)
i,j

Additionally, these operators are orthonormal (with respect to the Hilbert-Schmidt inner product),
C
and thus they form an ONB in L(C2 ), the space of linear operators acting on 2 . One can write
them in the matrix form as in (2.3)

1 0 0 0
   
P0 = , σ+ = , (2.9)
0 0 1 0
0 0 0 1
   
P1 = , σ− = , (2.10)
0 1 0 0

and therefore any operator can also be expressed as a 2 × 2 matrix.

4
Chapter 1.1 Quantum Information (WiSe 22/23)

ẑ = |0i
<latexit sha1_base64="ORFMS3kH8UwPV1xcD+x+jag44+o=">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</latexit>

| i

✓
ŷ
'
x̂

ẑ = |1i

Figure 1: Bloch sphere representation of single qubit state.

Pauli operators Sometimes it is convenient to express operators in terms of a basis that includes
the Pauli matrices σx , σy , σz , which are complex 2 × 2 matrices given by

1 0 0 1 0 −i 1 0
       
σ0 = 1 = , σx = , σy = and σz = . (2.11)
0 1 1 0 i 0 0 −1

Some useful properties of Pauli matrices:

1. They are all Hermitian (σα = σα† for all α ∈ {0, x, y, z}), satisfy σα2 = 1 ∀ α ∈ {0, x, y, z}.
2. Anticommutation relations between the Pauli matrices are given by {σα , σβ } = 2δα,β 1 for
α, β ∈ {x, y, z}.

3. Commutation relations between the Pauli matrices are given by [σα , σβ ] = 2i γ εα,β,γ σγ for
P
α, β ∈ {x, y, z}.
4. Orthogonality relations between the Pauli matrices holds with respect to the trace inner
product i.e. tr(σα σβ ) = 2δα,β for all α, β ∈ {0, x, y, z}. The Pauli matrices are thus linearly
√
independent and form a basis for 2 × 2 complex Hermitian matrices. In fact, the set {σα / 2}
forms an ONB.

Pauli operators in arbitrary directions Given a vector in R3 , ~n = (nx , ny , nz ), we define σ~n =

~n · ~σ = nx σx + ny σy + nz σz — this operator is often called the Pauli vector along ~n. It can be
noted that the eigenvalues of σ~n are simply given by ±|~n|. If |~n| = 1 we say that σ~n is a Pauli
vector in the direction defined by ~n.

Bloch representation of qubit mixed states Let ρ be a density operator of a qubit; then there
is a unique ~n ∈ R3 such that |~n| ≤ 1 and ρ = (1 + ~n · ~σ )/2. This can be shown as follows. We
easily note that since ρ = ρ† , it is Hermitian and thus expressible as a linear combination of Pauli
matrices and thus using the orthogonality relations between the Pauli matrices
X 1 1 + ~n · ~σ
ρ= tr(σα ρ)σα = , (2.12)
2 2
α∈{0,x,y,z}

where we have used that tr(1ρ) = 1 and identified nα = tr(σα ρ) for α ∈ {x, y, z}. This immediately
also yields that ~n = (nx , ny , nz ) for a given density operator ρ is unique. Furthermore, the
eigenvalues of ρ are (1 ± |~n|)/2, thus if ρ  0, then |~n| ≤ 1.
Every single-qubit state can be associated with a 3D vector of length less than or equal to 1, and
can thus be associated with a point on or inside the Bloch sphere. If it is on the Bloch sphere,
then it is a pure state. If it is inside the Bloch sphere, then it is a mixed state.

5
Chapter 1.1 Quantum Information (WiSe 22/23)

3 Observables and measurements

3.1 Observables
Postulate II An observable is associated with a Hermitian operator O = O† ∈ L(H) acting on
the system Hilbert space H ∼
= Cd .

Qubit operator For a qubit, any of these operators O can be written in terms of the operators
(2.7), since
O = 1O1 = (P0 + P1 )O(P0 + P1 )
(3.1)
= O00 |0ih0| + O11 |1ih1| + O01 |0ih1| + O10 |1ih0|,
where Oij = hi|O|ji = hj|O|ii. In a matrix form we can write
 
O00 O01
O= . (3.2)
O10 O11
Alternatively, we can also write O in any other ONB, such as the Pauli basis,
X
O= cα σ α , (3.3)
α=0,x,y,z

where cα = 21 tr(Oσα ) ∈ R.
3.2 Measurements
Given an observable O = O† , we can always write the spectral decomposition as
X
O= oi Pi . (3.4)
i

Here oi are real numbers, whereas Pi are orthogonal projector operators onto the (possibly degen-
erate) eigenspace corresponding to the eigenvalue oi . They fulfill
X
Pi Pi0 = δi,i0 Pi , Pi = 1. (3.5)
i

With all this, we can introduce the postulate of quantum mechanics regarding the measurement
of such an observable.

Postulate III Given a state, ρ, when we measure an observable O, we obtain as an outcome

one of its eigenvalues, oi . The probability pi of obtaining this outcome and the state ρi after the
measurement are
pi = tr(Pi ρ),
1 (3.6)
ρi = Pi ρPi .
pi
Note that one can destroy the particle in some measurements, so it does not make sense to talk
about the state after the measurement. The measurements that do not destroy the particle are
typically called ”filtering measurements”.

Special cases After given the general description, let us elaborate on how to apply the measure-
ment postulate to some specific cases.
• Measuring pure states: Measurement of O on a pure state ρ = |ψihψ| leads to outcome oi
with probability pi = hψ|Pi |ψi the state becomes Pi |ψi /kPi |ψi k.
• Non-degenerate measurement: When the eigenvalue oi is non-degenerate, Pi = |φi ihφi |, where
O|φi i = oi |φi i. If the state of the system is described by ρ, in a filtering measurement of
O one obtains the result oi with probability pi = hφi |ρ|φi i, and the state of the system is
projected (collapses) onto |φi i. If ρ is a pure state |ψi hψ|, the formula for the outcome
probability further simplifies to pi = |hφi |ψi|2 .
Example 3.1. Consider a simple pure state |ψi = |+i and O = σz . The probabilities of
obtaining +1 and -1 are p+ = 1/2 and p− = 1/2. The state after the measurement is |0i and
|1i, respectively.

6
Chapter 1.1 Quantum Information (WiSe 22/23)

Remark The measurement postulate of quantum mechanics tells us that it is possible that re-
peating exactly the same experiment give different measurement outcome. This is different in the
classical setup, where if one repeats an experiment in exactly the same conditions, the result is
the same. Furthermore, the state of our system is changed by the measurement, something which
is not necessarily so in the classical setup. This change of the state after the measurement is
sometimes referred to as the collapse of the wave function.

Expectation values If one prepares a large number of systems in the same state described by ρ,
and measures the observable O in all these systems, we will obtain a series of measurement results
o(1) , · · · , o(N ) and their average approaches the expectation (averaged) value hOi = tr(Oρ).
1  (1)  X
o + o(2) + . . . + o(N ) −→ pi oi = tr(ρO). (3.7)
N N →∞
i

√
Example 3.2. Consider ρ = 41 |0i h0| + 3
|1i h1| and the observable O = (σx + σz )/ 2. The
√ 4
expectation value tr(Oρ) = − 2/4.

Measuring a single system Note that the measurement postulate implies a very fundamental
difference between classical and quantum information. Whereas it is possible to determine whether
the state of a classical bit is 0 or 1, it is impossible to do so for a qubit. More precisely, it is
impossible to determine an unknown state of a qubit since any measurement perturbs the state
of the system (in general). For instance, suppose we measure an observable to determine whether
the qubit, prepared in the state |ψi = c0 |0i + c1 |1i, is in the state |0i or |1i . If the outcome is
0 (1), the states of the qubit after the measurement is |0i (|1i), respectively. Thus, if we obtain
the outcome 0, we only know that the probability for the system to be in the state |0i did not
vanish prior to the measurement, i.e., |c0 |2 6= 0. Since the state of the system is changed after
the measurement, we cannot learn any other information. It also implies that it is not so easy to
copy quantum information. In contrast to classical information, we cannot simply “look” at the
state and reproduce it. In fact, we will see later on that it is impossible to copy unknown quantum
information (no–cloning Theorem). Note that, however, if one would have infinitely many qubits
prepared in the same state, then one could indeed determine the state, i.e., one could determine
c0 and c1 . This process is called state tomography.

Quantum state tomography If one is able to prepare a system in a given state repeatedly, then
one can completely determine its state, that is, ρ. Quantum state tomography is the procedure
that determines a state ρ through a sequence of measurements in different basis. As an example,
for a qubit one can measure hσx i, hσy i, and hσz i, and with these results calculate λ1,2,3 in (2.12).
Quantum state tomography can also P be applied to higher dimensional systems, say qudits.
Consider a qudit density matrix ρ = ρn,m |ni hm|. The matrix elements ρn,m can be retrieved
by calculating the expectation value of On,m = |ni hm|. However, On,m = |ni hm| is not Hermitian
and thus is not an observable we can measure. A solution is to separate the Hermitian and anti-
Hermitian part of On,m . A complete set of Hermitian operator is given by,

On,m + Om,n |ni hm| + |mi hn|

H
On,m = = ,
2 2 (3.8)
On,m − Om,n |ni hm| − |mi hn|
A
On,m = = .
2i 2i
for n < m, and On,n . With the measurement in this basis, we can fully determine the state ρ if
you can repeat your measurements. In practice, one doesn’t have infinite measurements, so the
density operator can only be determined with finite accuracy. Note that in order to completely
determine a d × d density matrix, we need to measure d2 − 1 expectation values, which is equal to
the number of free parameters in ρ. The number P of operators we introduced above is d2 , but we
do not need to determine Od,d since we have 1 = n On,n .

Definition of density operators Now, equipped with the measurement postulate, we can ar-
gue why we use density operators instead of pure states whenever we have them with different

7
Chapter 1.1 Quantum Information (WiSe 22/23)

probabilities. Indeed, if we have certain states with some probabilities, we could simply compute
expectation values as X
hOi = pi hψi | O |ψi i , (3.9)
i

i.e., it is given by the probability that we have the state ψi times the expectation value of O in
that state. However, we can write hOi = tr(ρO), where ρ is given in (2.4). In fact, if we compute
once ρ in this way, we can then easily compute expectation values, so that we do not need to carry
out the sum in (3.9) every time. Furthermore, if we perform measures and have to keep track
of all probabilities, then it will become more and more difficult. In summary, the use of density
operators is extremely useful in quantum information. Additionally, when we study entanglement,
we will see that it is necessary in that case.

Indistinguishablity Consider two density matrices ρ1 = i pi |ψi i hψi | and ρ2 = i qi |φi i hφi |.
P P
The measurement postulate tells us what happens with the measurement only depends on the
density matrix. Therefore if ρ1 = ρ2 , then no measurement can distinguish between these two
ensembles. This means that even if two states are prepared in a very different way, if the resulting
density operator is the same, we cannot distinguish them at all.

Uncertainly If the state to be measured is an eigenstate of the measurement of O, the measure-

ment will always give a definite result with probability 1. If not, the measurements will have some
uncertainty, quantified by
2
(∆O)2 = hψ| (O − hOi)2 |ψi = hO2 i − hOi . (3.10)

Measurements in a basis This probability and the state after the measurement are completely
determined by the operators Pi and the density operator of the system. Since in Quantum Infor-
mation we are typically not interested in the eigenvalues of the observables we measure, we could
describe all possible measurements in terms of these operators [that is, we do not need to consider
observables but just projection operators fulfillingP(3.5)]. Consequently, in order to specify a mea-
surement we just have to specify {Pi } such that Pi = 1, and Pi Pj = δij Pi ≥ 0. When there is
no degeneracy in a measurement, {Pi } can be expressed through an orthonormal basis {|ψi ihψi |}.
We say that in this kind of measurement we are measuring in the basis {|ψi i}di=1 .
Example 3.3. Measurement in the Z basis: {|0i , |1i}. Measurement in the X basis: {|0ix , |1iz }.

Generalized measurements In fact, there is a more general scenario representing a measurement.

Such a measurement is characterized by a set of operators {Ai }Ni=1 which satisfies Ai ≥ 0 and
P †
i Ai Ai = 1. The probability of obtaining certain outcome i and the state of the system after
this outcome are given by  
pi = tr Ai ρA†i ,
1 (3.11)
ρi = Ai ρA†i .
pi

POVM If we are only intersted in the probabilities, we just need to specify Ei = A†i Ai , so
that pi = tr(Ei ρ). This concept is closely related to a mathematical concept, Positive Operator-
Valued Measures (POVM). Colloquially, one calls a POVM measurement. The probabilities of
each outcome is characterized by a set of operators {Eα }, fulfilling
X
Eα ≥ 0, Eα = 1, (3.12)
α

such that the probability of obtaining α for a state ρ is pα = tr(Eα ρ).

♣♣♣

Acknowledgement.— We thank Bennet Windt, Yilun Yang, and Adrian O. Paulus for providing
their handwritten notes. In particular, we are grateful to Bennet Windt for providing us the
sketches of Figure 1 to 3. We thank Franz Silva-Tarouca, Adrian O. Paulus, and Jan Geiger for

8
Chapter 1.1 Quantum Information (WiSe 22/23)

their proofreading.

These lecture notes were kindly provided by Ignacio Cirac, Sirui Lu and Rahul Trivedi.