Quantum States and Ensembles2

Lecture Notes for Ph219/CS219:
Quantum Information
Chapter 2
John Preskill
California Institute of Technology
Updated July 2015

Contents
2 Foundations I: States and Ensembles 3

2.1 Axioms of quantum mechanics 3
2.2 The Qubit 7
2.2.1 Spin- 12 8
2.2.2 Photon polarizations 14
2.3 The density operator 16
2.3.1 The bipartite quantum system 16
2.3.2 Bloch sphere 21
2.4 Schmidt decomposition 23
2.4.1 Entanglement 25
2.5 Ambiguity of the ensemble interpretation 26
2.5.1 Convexity 26
2.5.2 Ensemble preparation 28
2.5.3 Faster than light? 30
2.5.4 Quantum erasure 31
2.5.5 The HJW theorem 33
2.6 How far apart are two quantum states? 36
2.6.1 Fidelity and Uhlmann’s theorem 36
2.6.2 Relations among distance measures 38
2.7 Summary 41
2.8 Exercises 43
2
2
Foundations I: States and Ensembles
2.1 Axioms of quantum mechanics

In this chapter and the next we develop the theory of open quantum
systems. We say a system is open if it is imperfectly isolated, and therefore
exchanges energy and information with its unobserved environment. The
motivation for studying open systems is that all realistic systems are open.
Physicists and engineers may try hard to isolate quantum systems, but
they never completely succeed.
Though our main interest is in open systems we will begin by recalling
the theory of closed quantum systems, which are perfectly isolated. To
understand the behavior of an open system S, we will regard S combined
with its environment E as a closed system (the whole “universe”), then
ask how S behaves when we are able to observe S but not E.
Quantum theory is a mathematical model of the physical world. For
the case of closed systems we can characterize the model by stating five
axioms; these specify how to represent states, observables, measurements,
and dynamics, and also how to combine two systems to obtain a composite
system.
Axiom 1. States. A state is a complete description of a physical

system. In quantum mechanics, a state is a ray in a Hilbert space.
What is a Hilbert space?
a) It is a vector space over the complex numbers C. Vectors will be

denoted |ψi (Dirac’s ket notation).
b) It has an inner product hψ|ϕi that maps an ordered pair of vectors to
C, and that has the properties:
i) Positivity: hψ|ψi > 0 for |ψi =
6 0.
3
4 2 Foundations I: States and Ensembles
ii) Linearity: hϕ|(a|ψ1 i + b|ψ2 i) = ahϕ|ψ1 i + bhϕ|ψ2 i.

iii) Skew symmetry: hϕ|ψi = hψ|ϕi∗ .
(The ∗ denotes complex conjugation.)
c) It is complete in the norm ||ψ|| = hψ|ψi1/2 .
(Completeness is an important proviso in infinite-dimensional function

spaces, since it ensures the convergence of certain eigenfunction expan-
sions. But mostly we will be content to work with finite-dimensional
inner-product spaces.)
What is a ray? It is an equivalence class of vectors that differ by
multiplication by a nonzero complex scalar. For any nonzero ray, we can
by convention choose a representative of the class, denoted |ψi, that has
unit norm:
hψ|ψi = 1. (2.1)
Thus states correspond to normalized vectors, and the overall phase of
the vector has no physical significance: |ψi and eiα |ψi describe the same
state, where |eiα | = 1.
Since every ray corresponds to a possible state, given two states |ϕi, |ψi,
another state can be constructed as the linear superposition of the two,
a|ϕi + b|ψi. The relative phase in this superposition is physically sig-
nificant; we identify a|ϕi + b|ϕi with eiα (a|ϕi + b|ψi) but not with
a|ϕi + eiα b|ψi.
We use the notation hψ| (Dirac’s bra notation) for a linear function (a
dual vector) that takes vectors to complex numbers, defined by |ϕi →
hψ|ϕi.
Axiom 2. Observables. An observable is a property of a physical

system that in principle can be measured. In quantum mechanics,
an observable is a self-adjoint operator.
An operator is a linear map taking vectors to vectors,
A : |ψi 7→ A|ψi, A (a|ψi + b|ϕi) = aA|ψi + bA|ϕi. (2.2)
(We will often denote operators by boldface letters.) The adjoint A† of

the operator A is defined by
hϕ|Aψi = hA† ϕ|ψi, (2.3)
for all vectors |ϕi, |ψi (where here A|ψi has been denoted as |Aψi). A
is self-adjoint if A = A† , or in other words, if hϕ|A|ψi = hψ|A|ϕi∗ for all
vectors |ϕi and |ψi. If A and B are self adjoint, then so is A+B (because
(A + B)† = A† + B † ), but (AB)† = B † A† , so that AB is self adjoint
2.1 Axioms of quantum mechanics 5
only if A and B commute. Note that AB + BA and i(AB − BA) are

always self-adjoint if A and B are.
A self-adjoint operator in a Hilbert space H has a spectral representa-
tion – its eigenstates form a complete orthonormal basis in H. We can
express a self-adjoint operator A as
X
A= an E n . (2.4)
n
Here each an is an eigenvalue of A, and E n is the corresponding orthog-

onal projection onto the space of eigenvectors with eigenvalue an . The
E n ’s satisfy
E n E m = δn,m E n .
E †n = E n . (2.5)
The orthogonal projector onto the one-dimensional space spanned by the

vector |ψi may be expressed as |ψihψ|, where hψ| is the bra that annihi-
lates vectors orthogonal to |ψi. Therefore, an alternative notation for the
spectral representation of A is
X
A= |nian hn|, (2.6)
n
where {|ni} is the orthonormal basis of eigenstates of A, with A|ni =

an |ni.
(For unbounded operators in an infinite-dimensional space, the defini-
tion of self-adjoint and the statement of the spectral theorem are more
subtle, but this need not concern us.)
Axiom 3. Measurement. A measurement is a process in which in-

formation about the state of a physical system is acquired by an
observer. In quantum mechanics, the measurement of an observ-
able A prepares an eigenstate of A, and the observer learns the
value of the corresponding eigenvalue. If the quantum state just
prior to the measurement is |ψi, then the outcome an is obtained
with a priori probability
Prob(an ) = kE n |ψik2 = hψ|E n |ψi ; (2.7)
if the outcome an is attained, then the (normalized) quantum state

just after the measurement is
E n |ψi
. (2.8)
kE n |ψik
If the measurement is immediately repeated, then according to this rule

the same outcome is obtained again, with probability one. If many iden-
tically prepared systems are measured, each described by the state |ψi,
then the expectation value of the outcomes is
X X
hai ≡ an Prob(an ) = an hψ|E n |ψi = hψ|A|ψi. (2.9)
n n
Axiom 4. Dynamics. Dynamics describes how a state evolves over

time. In quantum mechanics, the time evolution of a closed system
is described by a unitary operator.
In the Schrödinger picture of dynamics, if the initial state at time t is

|ψ(ti, then the final state |ψ(t0 )i at time t0 can be expressed as
|ψ(t0 )i = U (t0 , t)|ψ(t)i|ψ(t)i, (2.10)
where U (t0 , t) is the unitary time evolution operator. Infinitesimal time

evolution is governed by the Schrödinger equation
d
|ψ(t)i = −iH(t)|ψ(t)i, (2.11)
dt
where H(t) is a self-adjoint operator, called the Hamiltonian of the sys-
tem. (The Hamiltonian has the dimensions of energy; we have chosen
units in which Planck’s constant ~ = h/2π = 1, so that energy has the
dimensions of inverse time.) To first order in the infinitesimal quantity
dt, the Schrödinger equation can be expressed as
|ψ(t + dt)i = (I − iH(t)dt)|ψ(t)i. (2.12)
Thus the operator U (t + dt, t) ≡ I − iH(t)dt is unitary; because H is

self-adjoint it satisfies U † U = 1 to linear order in dt. Since a product of
unitary operators is unitary, time evolution governed by the Schrödinger
equation over a finite interval is also unitary. In the case where H is time
0
independent we may write U (t0 , t) = e−i(t −t)H .
Our final axiom relates the description of a composite quantum system
AB to the description of its component parts A and B.
Axiom 5. Composite Systems. If the Hilbert space of system A

is HA and the Hilbert space of system B is HB , then the Hilbert
space of the composite systems AB is the tensor product HA ⊗ HB .
If system A is prepared in the state |ψiA and system B is prepared
in the state |ϕiB , then the composite system’s state is the product
|ψiA ⊗ |ϕiB .
2.2 The Qubit 7
What is a tensor product of Hilbert spaces? If {|iiA } denotes an or-

thonormal basis for HA and {|µiB } a basis for HB , then the states
|i, µiAB ≡ |iiA ⊗ |µiB are a basis for HA ⊗ HB , where the inner product
on HA ⊗ HB is defined by
AB hi, µ|j, νiAB = δij δµν . (2.13)

The tensor product operator M A ⊗ N B is the operator that applies M A
to system A and N B to system B. Its action on the orthonormal basis
|i, µiAB is
X
M A ⊗ N B |i, µiAB = M A |iiA ⊗ N B |µiB = |j, νiAB (MA )ji (NB )νµ .
j,ν
(2.14)
An operator that acts trivially on system B can be denoted M A ⊗ I B ,
where I B is the identity on HB , and an operator that acts trivially on
system A can be denoted I A ⊗ N B .
These five axioms provide a complete mathematical formulation of
quantum mechanics. We immediately notice some curious features. One
oddity is that the Schrödinger equation is linear, while we are accustomed
to nonlinear dynamical equations in classical physics. This property seems
to beg for an explanation. But far more curious is a mysterious dualism;
there are two quite distinct ways for a quantum state to change. On the
one hand there is unitary evolution, which is deterministic. If we specify
the initial state |ψ(0)i, the theory predicts the state |ψ(t)i at a later time.
But on the other hand there is measurement, which is probabilistic.
The theory does not make definite predictions about the measurement
outcomes; it only assigns probabilities to the various alternatives. This is
troubling, because it is unclear why the measurement process should be
governed by different physical laws than other processes.
The fundamental distinction between evolution and measurement, and
in particular the intrinsic randomness of the measurement process, is
sometimes called the measurement problem of quantum theory. Seeking a
more pleasing axiomatic formulation of quantum theory is a worthy task
which may eventually succeed. But these five axioms correctly account
for all that we currently know about quantum physics, and provide the
foundation for all that follows in this book.
2.2 The Qubit

The indivisible unit of classical information is the bit, which takes one
of the two possible values {0, 1}. The corresponding unit of quantum
information is called the “quantum bit” or qubit. It describes a state in
the simplest possible quantum system.
The smallest nontrivial Hilbert space is two-dimensional. We may de-

note an orthonormal basis for a two-dimensional vector space as {|0i, |1i}.
Then the most general normalized state can be expressed as
a|0i + b|1i, (2.15)
where a, b are complex numbers that satisfy |a|2 +|b|2 = 1, and the overall
phase is physically irrelevant. A qubit is a quantum system described by
a two-dimensional Hilbert space, whose state can take any value of the
form eq.(2.15).
We can perform a measurement that projects the qubit onto the basis
{|0i, |1i}. Then we will obtain the outcome |0i with probability |a|2 , and
the outcome |1i with probability |b|2 . Furthermore, except in the cases
a = 0 and b = 0, the measurement irrevocably disturbs the state. If the
value of the qubit is initially unknown, then there is no way to determine
a and b with that single measurement, or any other conceivable measure-
ment. However, after the measurement, the qubit has been prepared in a
known state – either |0i or |1i – that differs (in general) from its previous
state.
In this respect, a qubit differs from a classical bit; we can measure a
classical bit without disturbing it, and we can decipher all of the infor-
mation that it encodes. But suppose we have a classical bit that really
does have a definite value (either 0 or 1), but where that value is initially
unknown to us. Based on the information available to us we can only say
that there is a probability p0 that the bit has the value 0, and a probability
p1 that the bit has the value 1, where p0 + p1 = 1. When we measure
the bit, we acquire additional information; afterwards we know the value
with 100% confidence.
An important question is: what is the essential difference between a
qubit and a probabilistic classical bit? In fact they are not the same, for
several reasons that we will explore. To summarize the difference in brief:
there is only one way to look at a bit, but there is more than one way to
look at a qubit.
2.2.1 Spin- 12
First of all, the coefficients a and b in eq.(2.15) encode more than just the
probabilities of the outcomes of a measurement in the {|0i, |1i} basis. In
particular, the relative phase of a and b also has physical significance.
The properties of a qubit are easier to grasp if we appeal to a geomet-
rical interpretation of its state. For a physicist, it is natural to interpret
eq.(2.15) as the spin state of an object with spin- 12 (like an electron).
Then |0i and |1i are the spin up (| ↑i) and spin down (| ↓i) states along
2.2 The Qubit 9
a particular axis such as the z-axis. The two real numbers characterizing
the qubit (the complex numbers a and b, modulo the normalization and
overall phase) describe the orientation of the spin in three-dimensional
space (the polar angle θ and the azimuthal angle ϕ).
We will not go deeply here into the theory of symmetry in quantum
mechanics, but we will briefly recall some elements of the theory that
will prove useful to us. A symmetry is a transformation that acts on
a state of a system, yet leaves all observable properties of the system
unchanged. In quantum mechanics, observations are measurements of
self-adjoint operators. If A is measured in the state |ψi, then the outcome
|ai (an eigenvector of A) occurs with probability |ha|ψi|2 . A symmetry
should leave these probabilities unchanged, when we “rotate” both the
system and the apparatus.
A symmetry, then, is a mapping of vectors in Hilbert space
|ψi 7→ |ψ 0 i, (2.16)
that preserves the absolute values of inner products
|hϕ|ψi| = |hϕ0 |ψ 0 i|, (2.17)
for all |ϕi and |ψi. According to a famous theorem due to Wigner, a
mapping with this property can always be chosen (by adopting suitable
phase conventions) to be either unitary or antiunitary. The antiunitary
alternative, while important for discrete symmetries, can be excluded for
continuous symmetries. Then the symmetry acts as
|ψi 7→ |ψ 0 i = U |ψi, (2.18)
where U is unitary (and in particular, linear).
Symmetries form a group: a symmetry transformation can be inverted,
and the product of two symmetries is a symmetry. For each symmetry op-
eration R acting on our physical system, there is a corresponding unitary
transformation U (R). Multiplication of these unitary operators must re-
spect the group multiplication law of the symmetries – applying R1 ◦ R2
should be equivalent to first applying R2 and subsequently R1 . Thus we
demand
U (R1 )U (R2 ) = Phase(R1 , R2 ) · U (R1 ◦ R2 ) (2.19)
A phase depending on R1 and R2 is permitted in eq.(2.19) because quan-
tum states are rays; we need only demand that U (R1 ◦ R2 ) act the same
way as U (R1 )U (R2 ) on rays, not on vectors. We say that U (R) provides
a unitary representation, up to a phase, of the symmetry group.
So far, our concept of symmetry has no connection with dynamics.
Usually, we demand of a symmetry that it respect the dynamical evolu-
tion of the system. This means that it should not matter whether we
first transform the system and then evolve it, or first evolve it and then
transform it. In other words, the diagram
dynamics
Initial - Final
transformation transformation
? ?
dynamics
New Initial - New Final
is commutative, and therefore the time evolution operator eitH commutes

with the symmetry transformation U (R) :
U (R)e−itH = e−itH U (R) ; (2.20)
expanding to linear order in t we obtain
U (R)H = HU (R). (2.21)
For a continuous symmetry, we can choose R infinitesimally close to the

identity, R = I + T , and then U is close to I:
U = I − iεQ + O(ε2 ), (2.22)
where Q is an operator determined by T . From the unitarity of U (to

order ε) it follows that Q is an observable, Q = Q† . Expanding eq.(2.21)
to linear order in ε we find
[Q, H] = 0 ; (2.23)
the observable Q commutes with the Hamiltonian.

Eq.(2.23) is a conservation law. It says, for example, that if we pre-
pare an eigenstate of Q, then time evolution governed by the Schrödinger
equation will preserve the eigenstate. Thus we see that symmetries imply
conservation laws. Conversely, given a conserved quantity Q satisfying
eq.(2.23) we can construct the corresponding symmetry transformations.
Finite transformations can be built as a product of many infinitesimal
ones:
θ N θ
R = (1 + T ) ⇒ U (R) = (I + i Q)N → eiθQ , (2.24)
N N
taking the limit N → ∞. Once we have decided how infinitesimal symme-
try transformations are represented by unitary operators, then it is also
2.2 The Qubit 11
determined how finite transformations are represented, for these can be

built as a product of infinitesimal transformations. We say that Q is the
generator of the symmetry.
Let us briefly recall how this general theory applies to spatial ro-
tations and angular momentum. An infinitesimal rotation by dθ (in
the counterclockwise sense) about the axis specified by the unit vector
n̂ = (n1 , n2 , n3 ) can be expressed as
~
R(n̂, dθ) = I − idθn̂ · J, (2.25)
where (J1 , J2 , J3 ) are the components of the angular momentum. A finite
rotation is expressed as
~
R(n̂, θ) = exp(−iθn̂ · J). (2.26)
Rotations about distinct axes don’t commute. From elementary proper-
ties of rotations, we find the commutation relations
[Jk , J` ] = iεk`m Jm , (2.27)
where εk`m is the totally antisymmetric tensor with ε123 = 1, and repeated
indices are summed. To implement rotations on a quantum system, we
find self-adjoint operators J 1 , J 2 , J 3 in Hilbert space that satisfy these
relations.
The “defining” representation of the rotation group is three dimen-
sional, but the simplest nontrivial irreducible representation is two di-
mensional, given by
1
J k = σk , (2.28)
2
where

0 1 0 −i 1 0
σ1 = , σ2 = , σ3 = , (2.29)
1 0 i 0 0 −1
are the Pauli matrices. This is the unique two-dimensional irreducible
representation, up to a unitary change of basis. Since the eigenvalues of
J k are ± 21 , we call this the spin- 21 representation. (By identifying J as
the angular-momentum, we have implicitly chosen units with ~ = 1.)
The Pauli matrices also have the properties of being mutually anticom-
muting and squaring to the identity,
σ k σ ` + σ ` σ k = 2δk` I ; (2.30)
σ )2 = nk n` σ k σ ` = nk nk I = I (where repeated indices
therefore (n̂ · ~
are summed). By expanding the exponential series, we see that finite
rotations are represented as
θ θ
U (n̂, θ) = e−i 2 n̂·~
σ = I cos
θ
− in̂ · ~
σ sin . (2.31)
2 2
The most general 2×2 unitary matrix with determinant 1 can be expressed
in this form. Thus, we are entitled to think of a qubit as a spin- 21 object,
and an arbitrary unitary transformation acting on the qubit’s state (aside
from a possible physically irrelevant rotation of the overall phase) is a
rotation of the spin.
A peculiar property of the representation U (n̂, θ) is that it is double-
valued. In particular a rotation by 2π about any axis is represented non-
trivially:
U (n̂, θ = 2π) = −I. (2.32)
Our representation of the rotation group is really a representation “up to
a sign”
U (R1 )U (R2 ) = ±U (R1 ◦ R2 ). (2.33)
But as already noted, this is acceptable, because the group multiplication
is respected on rays, though not on vectors. These double-valued repre-
sentations of the rotation group are called spinor representations. (The
existence of spinors follows from a topological property of the group —
that it is not simply connected.)
While it is true that a rotation by 2π has no detectable effect on a
spin- 12 object, it would be wrong to conclude that the spinor property
has no observable consequences. Suppose I have a machine that acts on
a pair of spins. If the first spin is up, it does nothing, but if the first spin
is down, it rotates the second spin by 2π. Now let the machine act when
the first spin is in a superposition of up and down. Then
1 1
√ (| ↑i1 + | ↓i1 ) | ↑i2 7→ √ (| ↑i1 − | ↓i1 ) | ↑i2 . (2.34)
2 2
While there is no detectable effect on the second spin, the state of the
first has flipped to an orthogonal state, which is very much observable.
In a rotated frame of reference, a rotation R(n̂, θ) becomes a rotation
through the same angle but about a rotated axis. It follows that the three
components of angular momentum transform under rotations as a vector:
U (R)J k U (R)† = Rk` J ` . (2.35)
Thus, if a state |mi is an eigenstate of J 3
J 3 |mi = m|mi, (2.36)
then U (R)|mi is an eigenstate of RJ 3 with the same eigenvalue:
RJ 3 (U (R)|mi) = U (R)J 3 U(R)† U (R)|mi

= U (R)J 3 |mi = m (U (R)|mi) . (2.37)
2.2 The Qubit 13
Therefore, we can construct eigenstates of angular momentum along the

axis n̂ = (sin θ cos ϕ, sin θ sin ϕ, cos θ) by applying a counterclockwise rota-
tion through θ, about the axis n̂0 = (− sin ϕ, cos ϕ, 0), to a J 3 eigenstate.
For our spin- 12 representation, this rotation is
θ 0 −e−iϕ

θ 0
exp −i n̂ · ~ σ = exp
2 2 eiϕ 0
θ
−e−iϕ sin 2θ

cos 2
= , (2.38)
eiϕ sin 2θ cos 2θ
and applying it to 10 , the J 3 eigenstate with eigenvalue 1, we obtain

−iϕ/2
cos 2θ

e
|ψ(θ, ϕ)i = , (2.39)
eiϕ/2 sin 2θ
(up to an overall phase). We can check directly that this is an eigenstate
of
e−iϕ sin θ

cos θ
n̂ · ~
σ = iϕ , (2.40)
e sin θ − cos θ
with eigenvalue one. We now see that eq.(2.15) with a = e−iϕ/2 cos 2θ ,
b = eiϕ/2 sin 2θ , can be interpreted as a spin pointing in the (θ, ϕ) direction.
We noted that we cannot determine a and b with a single measurement.
Furthermore, even with many identical copies of the state, we cannot
completely determine the state by measuring each copy only along the
z-axis. This would enable us to estimate |a| and |b|, but we would learn
nothing about the relative phase of a and b. Equivalently, we would find
the component of the spin along the z-axis
θ θ
hψ(θ, ϕ)|σ3 |ψ(θ, ϕ)i = cos2 − sin2 = cos θ, (2.41)
2 2
but we would not learn about the component in the x-y plane. The prob-
lem of determining |ψi by measuring the spin is equivalent to determining
the unit vector n̂ by measuring its components along various axes. Alto-
gether, measurements along three different axes are required. E.g., from
hσ3 i and hσ1 i we can determine n3 and n1 , but the sign of n2 remains
undetermined. Measuring hσ2 i would remove this remaining ambiguity.
If we are permitted to rotate the spin, then only measurements along
the z-axis will suffice. That is, measuring a spin along the n̂ axis is
equivalent to first applying a rotation that rotates the n̂ axis to the axis
ẑ, and then measuring along ẑ.
In the special case θ = π2 and ϕ = 0 (the x̂-axis) our spin state is
1
| ↑x i = √ (| ↑z i + | ↓z i) (2.42)
2
(“spin-up along the x-axis”). The orthogonal state (“spin down along the
x-axis”) is
1
| ↓x i = √ (| ↑z i − | ↓z i) . (2.43)
2
For either of these states, if we measure the spin along the z-axis, we will
obtain | ↑z i with probability 12 and | ↓z i with probability 21 .
Now consider the combination
1
√ (| ↑x i + | ↓x i) . (2.44)
2
This state has the property that, if we measure the spin along the x-axis,
we obtain | ↑x i or | ↓x i, each with probability 21 . Now we may ask, what
if we measure the state in eq.(2.44) along the z-axis?
If these were probabilistic classical bits, the answer would be obvious.
The state in eq.(2.44) is in one of two states, and for each of the two,
the probability is 12 for pointing up or down along the z-axis. So of
course we should find up with probability 21 when we measure the state
√1 (| ↑x i + | ↓x i) along the z-axis.
2
But not so for qubits! By adding eq.(2.42) and eq.(2.43), we see that
the state in eq.(2.44) is really | ↑z i in disguise. When we measure along
the z-axis, we always find | ↑z i, never | ↓z i.
We see that for qubits, as opposed to probabilistic classical bits, proba-
bilities can add in unexpected ways. This is, in its simplest guise, the phe-
nomenon called “quantum interference,” an important feature of quantum
information.
To summarize the geometrical interpretation of a qubit: we may think
of a qubit as a spin- 21 object, and its quantum state is characterized
by a unit vector n̂ in three dimensions, the spin’s direction. A unitary
transformation rotates the spin, and a measurement of an observable has
two possible outcomes: the spin is either up or down along a specified
axis.
It should be emphasized that, while this formal equivalence with a spin-
1
2 object applies to any two-level quantum system, not every two-level
system transforms as a spinor under spatial rotations!
2.2.2 Photon polarizations

Another important two-state system is provided by a photon, which can
have two independent polarizations. These photon polarization states also
transform under rotations, but photons differ from our spin- 21 objects in
two important ways: (1) Photons are massless. (2) Photons have spin-1
(they are not spinors).
2.2 The Qubit 15
We will not present here a detailed discussion of the unitary represen-

tations of the Poincare group. Suffice it to say that the spin of a particle
classifies how it transforms under the little group, the subgroup of the
Lorentz group that preserves the particle’s momentum. For a massive
particle, we may always boost to the particle’s rest frame, and then the
little group is the rotation group.
For massless particles, there is no rest frame. The finite-dimensional
unitary representations of the little group turn out to be representations
of the rotation group in two dimensions, the rotations about the axis de-
termined by the momentum. For a photon, this corresponds to a familiar
property of classical light — the waves are polarized transverse to the
direction of propagation.
Under a rotation about the axis of propagation, the two linear polariza-
tion states (|xi and |yi for horizontal and vertical polarization) transform
as
|xi → cos θ|xi + sin θ|yi

|yi → − sin θ|xi + cos θ|yi. (2.45)
This two-dimensional representation is actually reducible. The matrix

cos θ sin θ
(2.46)
− sin θ cos θ
has the eigenstates

1 1 1 i
|Ri = √ |Li = √ , (2.47)
2 i 2 1
with eigenvalues eiθ and e−iθ , the states of right and left circular polar-
ization. That is, these are the eigenstates of the rotation generator

0 −i
J= = σ2, (2.48)
i 0
with eigenvalues ±1. Because the eigenvalues are ±1 (not ± 21 ) we say

that the photon has spin-1.
In this context, the quantum interference phenomenon can be described
as follows. The polarization states
1
|+i = √ (|xi + |yi) ,
2
1
|−i = √ (−|xi + |yi) , (2.49)
2
are mutually orthogonal and can be obtained by rotating the states |xi
and |yi by 45◦ . Suppose that we have a polarization analyzer that allows
only one of two orthogonal linear photon polarizations to pass through,
absorbing the other. Then an x or y polarized photon has probability 21
of getting through a 45◦ rotated polarizer, and a 45◦ polarized photon
has probability 12 of getting through an x or y analyzer. But an x photon
never passes through a y analyzer.
Suppose that a photon beam is directed at an x analyzer, with a y
analyzer placed further downstream. Then about half of the photons will
pass through the first analyzer, but every one of these will be stopped
by the second analyzer. But now suppose that we place a 45◦ -rotated
analyzer between the x and y analyzers. Then about half of the photons
pass through each analyzer, and about one in eight will manage to pass all
three without being absorbed. Because of this interference effect, there
is no consistent interpretation in which each photon carries one classical
bit of polarization information. Qubits are different than probabilistic
classical bits.
A device can be constructed that rotates the linear polarization of a
photon, and so applies the transformation Eq. (2.45) to our qubit; it
functions by “turning on” a Hamiltonian for which the circular polar-
ization states |Li and |Ri are nondegenerate energy eigenstates. This
is not the most general possible unitary transformation. But if we also
have a device that alters the relative phase of the two orthogonal linear
polarization states
|xi → e−iϕ/2 |xi,
|yi → eiϕ/2 |yi (2.50)
(by turning on a Hamiltonian whose nondegenerate energy eigenstates are
the linear polarization states), then the two devices can be employed to-
gether to apply an arbitrary 2 × 2 unitary transformation (of determinant
1) to the photon polarization state.
2.3 The density operator

2.3.1 The bipartite quantum system
Having understood everything about a single qubit, we are ready to ad-
dress systems with two qubits. Stepping up from one qubit to two is a
bigger leap than you might expect. Much that is weird and wonderful
about quantum mechanics can be appreciated by considering the proper-
ties of the quantum states of two qubits.
The axioms of §2.1 provide a perfectly acceptable general formulation
of the quantum theory. Yet under many circumstances, we find that the
axioms appear to be violated. The trouble is that our axioms are intended
to characterize the quantum behavior of a closed system that does not
interact with its surroundings. In practice, closed quantum systems do
not exist; the observations we make are always limited to a small part of
a much larger quantum system.
When we study open systems, that is, when we limit our attention to
just part of a larger system, then (contrary to the axioms):
1. States are not rays.

2. Measurements are not orthogonal projections.
3. Evolution is not unitary.
To arrive at the laws obeyed by open quantum systems, we must recall

our fifth axiom, which relates the description of a composite quantum
system to the description of its component parts. As a first step toward
understanding the quantum description of an open system, consider a
two-qubit world in which we observe only one of the qubits. Qubit A is
here in the room with us, and we are free to observe or manipulate it any
way we please. But qubit B is locked in a vault where we can’t get access
to it. The full system AB obeys the axioms of §2.1. But we would like
to find a compact way to characterize the observations that can be made
on qubit A alone.
We’ll use {|0iA , |1iA } and {|0iB , |1iB } to denote orthonormal bases for
qubits A and B respectively. Consider a quantum state of the two-qubit
world of the form
|ψiAB = a|0iA ⊗ |0iB + b|1iA ⊗ |1iB . (2.51)
In this state, qubits A and B are correlated. Suppose we measure qubit

A by projecting onto the {|0iA , |1iA } basis. Then with probability |a|2
we obtain the result |0iA , and the measurement prepares the state
|0iA ⊗ |0iB ; (2.52)
with probability |b|2 , we obtain the result |1iA and prepare the state
|1iA ⊗ |1iB . (2.53)
In either case, a definite state of qubit B is picked out by the measure-

ment. If we subsequently measure qubit B, then we are guaranteed (with
probability one) to find |0iB if we had found |0iA , and we are guaran-
teed to find |1iB if we had found |1iA . In this sense, the outcomes of the
{|0iA , |1iA } and {|0iB , |1iB } measurements are perfectly correlated in the
state |ψiAB .
But now we would like to consider more general observables acting on

qubit A, and we would like to characterize the measurement outcomes for
A alone (irrespective of the outcomes of any measurements of the inac-
cessible qubit B). An observable acting on qubit A only can be expressed
as
M A ⊗ IB, (2.54)
where M A is a self-adjoint operator acting on A, and I B is the identity
operator acting on B. The expectation value of the observable in the
state |ψi is:
hM A i = hψ|M A ⊗ I B |ψi
= (a∗ h00| + b∗ h11|) (M A ⊗ I B ) (a|00i + b|11i)
= |a|2 h0|M A |0i + |b|2 h1|MA |1i (2.55)
(where we have used the orthogonality of |0iB and |1iB ). This expression
can be rewritten in the form
hM A i = tr (M A ρA ) , ρA = |a|2 |0ih0| + |b|2 |1ih1| (2.56)
and tr(·) denotes the trace. The operator ρA is called the density operator
(or density matrix) for qubit A. It is self-adjoint, positive (its eigenvalues
are nonnegative) and it has unit trace (because |ψi is a normalized state.)
Because hM A i has the form eq.(2.56) for any observable M A acting
on qubit A, it is consistent to interpret ρA as representing an ensemble of
possible quantum states, each occurring with a specified probability. That
is, we would obtain precisely the same result for hM A i if we stipulated
that qubit A is in one of two quantum states. With probability p0 = |a|2 it
is in the quantum state |0i, and with probability p1 = |b|2 it is in the state
|1i. If we are interested in the result of any possible measurement, we can
consider M A to be the projection E A (a) onto the relevant eigenspace of
a particular observable. Then
Prob(a) = p0 h0|E A (a)|0i + p1 h1|E A (a)|1i, (2.57)
which is the probability of outcome a summed over the ensemble, and
weighted by the probability of each state in the ensemble.
We have emphasized previously that there is an essential difference be-
tween a coherent superposition of the states |0i and |1i, and a probabilistic
ensemble, in which |0i and |1i can each occur with specified probabilities.
For example, for a spin- 12 object we have seen that if we measure σ 1 in
the state √12 (| ↑z i + | ↓z i), we will obtain the result | ↑x i with probability
one. But the ensemble in which | ↑z i and | ↓z i each occur with probability
1
2 is represented by the density operator
1 1
ρ= (| ↑z ih↑z | + | ↓z ih↓z |) = I, (2.58)
2 2
and the projection onto | ↑x i then has the expectation value

1
tr (| ↑x ih↑x |ρ) = h↑x |ρ| ↑x i = . (2.59)
2
Similarly, if we measure the spin along any axis labeled by polar angles θ
and ϕ, the probability of obtaining the result “spin up” is
h|ψ(θ, ϕ)ihψ(θ, ϕ)|i = tr (|ψ(θ, ϕ)ihψ(θ, ϕ)|ρ)

1 1
= hψ(θ, ϕ)| I|ψ(θ, ϕ)i = . (2.60)
2 2
Therefore, if in the two-qubit world an equally weighted coherent su-
perposition of |00i and |11i is prepared, the state of qubit A behaves
incoherently – along any axis it is an equiprobable mixture of spin up and
spin down.
This discussion of the correlated two-qubit state |ψiAB is easily gener-
alized to an arbitrary state of any bipartite quantum system (a system di-
vided into two parts). The Hilbert space of a bipartite system is HA ⊗HB
where HA,B are the Hilbert spaces of the two parts. This means that if
{|iiA } is an orthonormal basis for HA and {|µiB } is an orthonormal basis
for HB , then {|iiA ⊗ |µiB } is an orthonormal basis for HA ⊗ HB . Thus
an arbitrary pure state of HA ⊗ HB can be expanded as
X
|ψiAB = aiµ |iiA ⊗ |µiB , (2.61)
i,µ
where i,µ |aiµ |2 = 1. The expectation value of an observable M A ⊗ I B

P
that acts only on subsystem A is
hM A i = AB hψ|M A ⊗ I B |ψiAB
X X
= a∗jν (A hj| ⊗ B hν|) (M A ⊗ I B ) aiµ (|iiA ⊗ |µiB )
j,ν i,µ
X
= a∗jµ aiµ hj|M A |ii = tr (M A ρA ) , (2.62)
i,j,µ
where
X
ρA = trB (|ψihψ|) ≡ aiµ a∗jµ |iihj| (2.63)
i,j,µ
is the density operator of subsystem A.

We may say that the density operator ρA for subsystem A is obtained
by performing a partial trace over subsystem B of the density operator
(in this case a pure state) for the combined system AB. We may regard a
dual vector (or bra) B hµ| as a linear map that takes vectors in HA ⊗ HB
to vectors of HA , defined through its action on a basis:
B hµ|iνiAB = δµν |iiA ; (2.64)
similarly, the ket |µiB defines a map from the HA ⊗ HB dual basis to the
HA dual basis, via
AB hiν|µiB = δµν A hi|. (2.65)
The partial trace operation is a linear map that takes an operator M AB
on HA ⊗ HB to an operator on HA defined as
X
trB M AB = B hµ|M AB |µiB . (2.66)
µ
We see that the density operator acting on A is the partial trace
ρA = trB (|ψihψ|) . (2.67)
From the definition eq.(2.63), we can immediately infer that ρA has the
following properties:
1. ρA is self-adjoint: ρA = ρ†A .
2. ρA is positive: For any |ϕi, hϕ|ρA |ϕi = µ | i aiµ hϕ|ii|2 ≥ 0.
P P
2
P
3. tr(ρA ) = 1: We have tr(ρA ) = i,µ |aiµ | = 1, since |ψiAB is
normalized.
It follows that ρA can be diagonalized in an orthonormal basis, that the
eigenvalues are all real and nonnegative, and that the eigenvalues sum to
one.
If we are looking at a subsystem of a larger quantum system, then, even
if the state of the larger system is a ray, the state of the subsystem need
not be; in general, the state is represented by a density operator. In the
case where the state of the subsystem is a ray, and we say that the state is
pure. Otherwise the state is mixed. If the state is a pure state |ψiA , then
the density matrix ρA = |ψihψ| is the projection onto the one-dimensional
space spanned by |ψiA . Hence a pure density matrix has the property
ρ2 = ρ. A general density matrix, expressed in the basis {|ai} in which
it is diagonal, has the form
X
ρA = pa |aiha|, (2.68)
a
P
where 0 < pa ≤ 1 and a pa = 1. If the state is not pure,P there
Pare two
or more terms in this sum, and ρ2 6= ρ; in fact, tr ρ2 = p2a < pa = 1.
We say that ρ is an incoherent mixture of the states {|ai}; “incoherent”

means that the relative phases of the |ai’s are experimentally inaccessible.
Since the expectation value of any observable M acting on the subsys-
tem can be expressed as
X
hM i = tr M ρ = pa ha|M |ai, (2.69)
a
we see as before that we may interpret ρ as describing an ensemble of pure

quantum states, in which the state |ai occurs with probability pa . We
have, therefore, come a long part of the way to understanding how prob-
abilities arise in quantum mechanics when a quantum system A interacts
with another system B. A and B become entangled, that is, correlated.
The entanglement destroys the coherence of a superposition of states of
A, so that some of the phases in the superposition become inaccessible if
we look at A alone. We may describe this situation by saying that the
state of system A collapses — it is in one of a set of alternative states,
each of which can be assigned a probability.
2.3.2 Bloch sphere

Let’s return to the case in which system A is a single qubit, and consider
the form of the general density matrix. The most general self-adjoint
2 × 2 matrix has four real parameters, and can be expanded in the basis
{I, σ 1 , σ 2 , σ 3 }. Since each σ i is traceless, the coefficient of I in the
expansion of a density matrix ρ must be 12 (so that tr(ρ) = 1), and ρ
may be expressed as
1
ρ(P~ ) = I + P~ · ~
σ
2
1
≡ (I + P1 σ 1 + P2 σ 2 + P3 σ 3 )
2
1 1 + P3 P1 − iP2
= , (2.70)
2 P1 + iP2 1 − P3

where P1 , P2 , P3 are real numbers. We can compute detρ = 41 1 − P~ 2 .
Therefore, a necessary condition for ρ to have nonnegative eigenvalues is
detρ ≥ 0 or P~ 2 ≤ 1. This condition is also sufficient; since tr ρ = 1,
it is not possible for ρ to have two negative eigenvalues. Thus, there is
a 1 − 1 correspondence between the possible density matrices of a single
qubit and the points on the unit 3-ball 0 ≤ |P~ | ≤ 1. This ball is usually
called the Bloch sphere (although
it is really a ball, not a sphere).
~
The boundary |P | = 1 of the ball (which really is a sphere) contains
the density matrices with vanishing determinant. Since tr ρ = 1, these
density matrices must have the eigenvalues 0 and 1 — they are one-
dimensional projectors, and hence pure states. We have already seen
that any pure state of a single qubit is of the form |ψ(θ, ϕ)i and can be
envisioned as a spin pointing in the (θ, ϕ) direction. Indeed using the
property
σ )2 = I,
(n̂ · ~ (2.71)
where n̂ is a unit vector, we can easily verify that the pure-state density
matrix
1
ρ(n̂) = (I + n̂ · ~σ) (2.72)
2
satisfies the property
(n̂ · ~
σ ) ρ(n̂) = ρ(n̂) (n̂ · ~
σ ) = ρ(n̂), (2.73)
and, therefore is the projector
ρ(n̂) = |ψ(n̂)ihψ(n̂)| ; (2.74)
that is, n̂ is the direction along which the spin is pointing up. Alterna-
tively, from the expression
−iϕ/2
e cos (θ/2)
|ψ(θ, φ)i = , (2.75)
eiϕ/2 sin (θ/2)
we may compute directly that
ρ(θ, φ) = |ψ(θ, φ)ihψ(θ, φ)|
cos2 (θ/2) cos (θ/2) sin (θ/2)e−iϕ

=
cos (θ/2) sin (θ/2)eiϕ sin2 (θ/2)
sin θe−iϕ

1 1 cos θ 1
= I+ iϕ = (I + n̂ · ~
σ ) (2.76)
2 2 sin θe − cos θ 2
where n̂ = (sin θ cos ϕ, sin θ sin ϕ, cos θ). One nice property of the Bloch
parametrization of the pure states is that while |ψ(θ, ϕ)i has an arbitrary
overall phase that has no physical significance, there is no phase ambiguity
in the density matrix ρ(θ, ϕ) = |ψ(θ, ϕ)ihψ(θ, ϕ)|; all the parameters in ρ
have a physical meaning.
From the property
1
tr σ i σ j = δij (2.77)
2
we see that
hn̂ · ~
σ iP~ = tr n̂ · ~σ ρ(P~ ) = n̂ · P~ . (2.78)
We say that the vector P~ in Eq. (2.70) parametrizes the polarization of

the spin. If there are many identically prepared systems at our disposal,
we can determine P~ (and hence the complete density matrix ρ(P~ )) by
measuring hn̂ · ~
σ i along each of three linearly independent axes.
2.4 Schmidt decomposition

A bipartite pure state can be expressed in a standard form (the Schmidt
decomposition) that is often very useful.
To arrive at this form, note that an arbitrary vector in HA ⊗ HB can
be expanded as
X X
|ψiAB = ψiµ |iiA ⊗ |µiB ≡ |iiA ⊗ |ĩiB . (2.79)
i,µ i
Here {|iiA } and {|µiB } are orthonormal basis for HA and HB respectively,
but to obtain the second equality in eq.(2.79) we have defined
X
|ĩiB ≡ ψiµ |µiB . (2.80)
µ
Note that the |ĩiB ’s need not be mutually orthogonal or normalized.

Now let’s suppose that the {|iiA } basis is chosen to be the basis in
which ρA is diagonal,
X
ρA = pi |iihi|. (2.81)
i
We can also compute ρA by performing a partial trace,
ρA = trB (|ψihψ|)
X X
= trB ( |iihj| ⊗ |ĩihj̃|) = hj̃|ĩi (|iihj|) . (2.82)
i,j i,j
We obtained the last equality in eq.(2.82) by noting that

X
trB |ĩih j̃| = hk|ĩihj̃|ki
k
X
= hj̃|kihk|ĩi = hj̃|ĩi, (2.83)
k
where {|ki} is a complete orthonormal basis for HB . By comparing

eq.(2.81) and eq. (2.82), we see that
B hj̃|ĩiB = pi δij . (2.84)
Hence, it turns out that the {|ĩiB } are orthogonal after all. We obtain
orthonormal vectors by rescaling,
−1/2
|i0 iB = pi |ĩiB (2.85)
(we may assume pi 6= 0, because we will need eq.(2.85) only for i appearing
in the sum eq.(2.81)), and therefore obtain the expansion
X√
|ψiAB = pi |iiA ⊗ |i0 iB , (2.86)
i
in terms of a particular orthonormal basis of HA and HB .

Eq.(2.86) is the Schmidt decomposition of the bipartite pure state
|ψiAB . Any bipartite pure state can be expressed in this form, but the
bases used depend on the pure state that is being expanded. In general,
we can’t simultaneously expand both |ψiAB and |ϕiAB ∈ HA ⊗ HB in the
form eq.(2.86) using the same orthonormal bases for HA and HB .
It is instructive to compare the Schmidt decomposition of the bipartite
pure state |ψiAB with its expansion in a generic orthonormal basis
X
|ψiAB = ψaµ |aiA ⊗ |µiB . (2.87)
a,µ
The orthonormal bases {|aiA } and {|µiB } are related to the Schmidt
bases {|iiA } and {|i0 iB } by unitary transformations U A and U B , hence
X X
|iiA = |aiA (UA )ai , |i0 iB = |µiB (UB )µi0 . (2.88)
a µ
By equating the expressions for |ψiAB in eq.(2.86) and eq.(2.87), we find

X √
(UA )ai pi UBT iµ .

ψaµ = (2.89)
i
We see that by applying unitary transformations on the left and right,

any matrix ψ can be transformed to a matrix which is diagonal and non-
negative. (The “diagonal” matrix will be rectangular rather than square
if the Hilbert spaces HA and HB have different dimensions.) Eq.(2.89) is
√
said to be the singular value decomposition of ψ, and the weights { pi }
in the Schmidt decomposition are ψ’s singular values.
Using eq.(2.86), we can also evaluate the partial trace over HA to obtain
X
ρB = trA (|ψihψ|) = pi |i0 ihi0 |. (2.90)
i
We see that ρA and ρB have the same nonzero eigenvalues. If HA and

HB do not have the same dimension, the number of zero eigenvalues of
ρA and ρB will differ.
If ρA (and hence ρB ) have no degenerate eigenvalues other than zero,
then the Schmidt decomposition of |ψiAB is essentially uniquely deter-
mined by ρA and ρB . We can diagonalize ρA and ρB to find the |iiA ’s
and |i0 iB ’s, and then we pair up the eigenstates of ρA and ρB with the
same eigenvalue to obtain eq.(2.86). We have chosen the phases of our
basis states so that no phases appear in the coefficients in the sum; the
only remaining freedom is to redefine |iiA and |i0 iB by multiplying by
opposite phases (which leaves the expression eq.(2.86) unchanged).
But if ρA has degenerate nonzero eigenvalues, then we need more in-
formation than that provided by ρA and ρB to determine the Schmidt
decomposition; we need to know which |i0 iB gets paired with each |iiA .
For example, if both HA and HB are d-dimensional and Uij is any d × d
unitary matrix, then
d
1 X
|ψiAB = √ |iiA Uij ⊗ |j 0 iB , (2.91)
d i,j=1
will yield ρA = ρB = d1 I when we take partial traces. Furthermore, we

are free to apply simultaneous unitary transformations in HA and HB ;
writing
X X
|iiA = |aiA Uai , |i0 iB = |b0 iB Ubi∗ , (2.92)
a b
we have
1 X 1 X
|ψiAB = √ |iiA ⊗ |i0 iB = √ |aiA Uai ⊗ |b0 iB Uib†
d i d i,a,b
1 X
=√ |aiA ⊗ |a0 iB . (2.93)
d a
This simultaneous rotation preserves the state |ψiAB , illustrating that

there is an ambiguity in the basis used when we express |ψiAB in the
Schmidt form.
2.4.1 Entanglement
With any bipartite pure state |ψiAB we may associate a positive integer,
the Schmidt number, which is the number of nonzero eigenvalues in ρA
(or ρB ) and hence the number of terms in the Schmidt decomposition
of |ψiAB . In terms of this quantity, we can define what it means for a
bipartite pure state to be entangled: |ψiAB is entangled (or nonseparable)
if its Schmidt number is greater than one; otherwise, it is separable (or
unentangled). Thus, a separable bipartite pure state is a direct product
of pure states in HA and HB ,
|ψiAB = |ϕiA ⊗ |χiB ; (2.94)
then the reduced density matrices ρA = |ϕihϕ| and ρB = |χihχ| are pure.
Any state that cannot be expressed as such a direct product is entangled;
then ρA and ρB are mixed states.
When |ψiAB is entangled we say that A and B have quantum corre-
lations. It is not strictly correct to say that subsystems A and B are
uncorrelated if |ψiAB is separable; after all, the two spins in the separable
state
| ↑iA | ↑iB , (2.95)
are surely correlated – they are both pointing in the same direction. But
the correlations between A and B in an entangled state have a different
character than those in a separable state. One crucial difference is that
entanglement cannot be created locally. The only way to entangle A and
B is for the two subsystems to directly interact with one another.
We can prepare the state eq.(2.95) without allowing spins A and B to
ever come into contact with one another. We need only send a (classical!)
message to two preparers (Alice and Bob) telling both of them to prepare
a spin pointing along the z-axis. But the only way to turn the state
eq.(2.95) into an entangled state like
1
√ (| ↑iA | ↑iB + | ↓iA | ↓iB ) , (2.96)
2
is to apply a collective unitary transformation to the state. Local unitary
transformations of the form UA ⊗ UB , and local measurements performed
by Alice or Bob, cannot increase the Schmidt number of the two-qubit
state, no matter how much Alice and Bob discuss what they do. To
entangle two qubits, we must bring them together and allow them to
interact.
As we will discuss in Chapter 4, it is also possible to make the distinction
between entangled and separable bipartite mixed states. We will also
discuss various ways in which local operations can modify the form of
entanglement, and some ways that entanglement can be put to use.
2.5 Ambiguity of the ensemble interpretation

2.5.1 Convexity
Recall that an operator ρ acting on a Hilbert space H may be interpreted
as a density operator if it has the three properties:
(1) ρ is self-adjoint.
(2) ρ is nonnegative.
(3) tr(ρ) = 1.
It follows immediately that, given two density matrices ρ1 , and ρ2 , we can

always construct another density matrix as a convex linear combination
of the two:
ρ(λ) = λρ1 + (1 − λ)ρ2 (2.97)
is a density matrix for any real λ satisfying 0 ≤ λ ≤ 1. We easily see that
ρ(λ) satisfies (1) and (3) if ρ1 and ρ2 do. To check (2), we evaluate
hψ|ρ(λ)|ψi = λhψ|ρ1 |ψi + (1 − λ)hψ|ρ2 |ψi ≥ 0; (2.98)
hρ(λ)i is guaranteed to be nonnegative because hρ1 i and hρ2 i are. We
have, therefore, shown that in a Hilbert space H of dimension d, the
density operators are a convex subset of the real vector space of d × d
hermitian operators. (A subset of a vector space is said to be convex if
the set contains the straight line segment connecting any two points in
the set.)
Most density operators can be expressed as a sum of other density
operators in many different ways. But the pure states are special in this
regard – it is not possible to express a pure state as a convex sum of two
other states. Consider a pure state ρ = |ψihψ|, and let |ψ⊥ i denote a
vector orthogonal to |ψi, hψ⊥ |ψi = 0. Suppose that ρ can be expanded
as in eq.(2.97); then
hψ⊥ |ρ|ψ⊥ i = 0 = λhψ⊥ |ρ1 |ψ⊥ i
+ (1 − λ)hψ⊥ |ρ2 |ψ⊥ i. (2.99)
Since the right hand side is a sum of two nonnegative terms, and the
sum vanishes, both terms must vanish. If λ is not 0 or 1, we conclude
that ρ1 and ρ2 are orthogonal to |ψ⊥ i. But since |ψ⊥ i can be any vector
orthogonal to |ψi, we see that ρ1 = ρ2 = ρ.
The vectors in a convex set that cannot be expressed as a linear com-
bination of other vectors in the set are called the extremal points of the
set. We have just shown that the pure states are extremal points of the
set of density matrices. Furthermore, only the P pure states are extremal,
because any mixed state can be written ρ = i pi |iihi| in the basis in
which it is diagonal, and so is a convex sum of pure states.
We have already encountered this structure in our discussion of the
special case of the Bloch sphere. We saw that the density operators are a
(unit) ball in the three-dimensional set of 2 × 2 hermitian matrices with
unit trace. The ball is convex, and its extremal points are the points on
the boundary. Similarly, the d × d density operators are a convex subset
of the (d2 − 1)-dimensional set of d × d hermitian matrices with unit trace,
and the extremal points of the set are the pure states.
However, the 2 × 2 case is atypical in one respect: for d > 2, the points
on the boundary of the set of density matrices are not necessarily pure
states. The boundary of the set consists of all density matrices with
at least one vanishing eigenvalue (since there are nearby matrices with
negative eigenvalues). Such a density matrix need not be pure, for d > 2,
since the number of nonvanishing eigenvalues can exceed one.
2.5.2 Ensemble preparation

The convexity of the set of density matrices has a simple and enlightening
physical interpretation. Suppose that a preparer agrees to prepare one of
two possible states; with probability λ, the state ρ1 is prepared, and with
probability 1 − λ, the state ρ2 is prepared. (A random number generator
might be employed to guide this choice.) To evaluate the expectation
value of any observable M , we average over both the choices of preparation
and the outcome of the quantum measurement:
hM i = λhM i1 + (1 − λ)hM i2
= λtr(M ρ1 ) + (1 − λ)tr(M ρ2 )
= tr (M ρ(λ)) . (2.100)
All expectation values are thus indistinguishable from what we would

obtain if the state ρ(λ) had been prepared instead. Thus, we have an
operational procedure, given methods for preparing the states ρ1 and ρ2 ,
for preparing any convex combination.
Indeed, for any mixed state ρ, there are an infinite variety of ways to
express ρ as a convex combination of other states, and hence an infinite
variety of procedures we could employ to prepare ρ, all of which have
exactly the same consequences for any conceivable observation of the sys-
tem. But a pure state is different; it can be prepared in only one way.
(This is what is “pure” about a pure state.) Every pure state is an eigen-
state of some observable, e.g., for the state ρ = |ψihψ|, measurement of
the projection E = |ψihψ| is guaranteed to have the outcome 1. (For
example, recall that every pure state of a single qubit is “spin-up” along
some axis.) Since ρ is the only state for which the outcome of measuring
E is 1 with 100% probability, there is no way to reproduce this observ-
able property by choosing one of several possible preparations. Thus, the
preparation of a pure state is unambiguous (we can infer a unique prepa-
ration if we have many copies of the state to experiment with), but the
preparation of a mixed state is always ambiguous.
How ambiguous is it? Since any ρ can be expressed as a sum of pure
states, let’s confine our attention to the question: in how many ways
can a density operator be expressed as a convex sum of pure states?
Mathematically, this is the question: in how many ways can ρ be written
as a sum of extremal states?
As a first example, consider the “maximally mixed” state of a single

qubit:
1
ρ = I. (2.101)
2
This can indeed be prepared as an ensemble of pure states in an infinite
variety of ways. For example,
1 1
ρ = | ↑z ih↑z | + | ↓z ih↓z |, (2.102)
2 2
so we obtain ρ if we prepare either | ↑z i or | ↓z i, each occurring with
probability 21 . But we also have
1 1
ρ = | ↑x ih↑x | + | ↓x ih↓x |, (2.103)
2 2
so we obtain ρ if we prepare either | ↑x i or | ↓x i, each occurring with
probability 21 . Now the preparation procedures are undeniably different.
Yet there is no possible way to tell the difference by making observations
of the spin.
More generally, the point at the center of the Bloch ball is the sum of
any two antipodal points on the sphere – preparing either | ↑n̂ i or | ↓n̂ i,
each occurring with probability 12 , will generate ρ = 12 I.
Only in the case where ρ has two (or more) degenerate eigenvalues
will there be distinct ways of generating ρ from an ensemble of mutually
orthogonal pure states, but there is no good reason to confine our attention
to ensembles of mutually orthogonal pure states. We may consider a point
in the interior of the Bloch ball
1
ρ(P~ ) = (I + P~ · ~σ ), (2.104)
2
with 0 < |P~ | < 1, and it too can be expressed as
ρ(P~ ) = λρ(n̂1 ) + (1 − λ)ρ(n̂2 ), (2.105)
if P~ = λn̂1 + (1 − λ)n̂2 (or in other words, if P~ lies somewhere on the line

segment connecting the points n̂1 and n̂2 on the sphere). Evidently, for
any P~ , there is a an expression for ρ(P~ ) as a convex combination of pure
states associated with any chord of the Bloch sphere that passes through
the point P~ ; all such chords comprise a two-parameter family.
This highly ambiguous nature of the preparation of a mixed quantum
state is one of the characteristic features of quantum information that
contrasts sharply with classical probability distributions. Consider, for
example, the case of a probability distribution for a single classical bit.
The two extremal distributions are those in which either 0 or 1 occurs
with 100% probability. Any probability distribution for the bit is a convex
sum of these two extremal points. Similarly, if there are d possible states,
there are d extremal distributions, and any probability distribution has
a unique decomposition into extremal ones (the convex set of probability
distributions is a simplex, the convex hull of its d vertices). If 0 occurs
with 21% probability, 1 with 33% probability, and 2 with 46% probability,
there is a unique “preparation procedure” that yields this probability
distribution.
2.5.3 Faster than light?

Let’s now return to our earlier viewpoint — that a mixed state of system
A arises because A is entangled with system B — to further consider the
implications of the ambiguous preparation of mixed states. If qubit A has
density matrix
1 1
ρA = | ↑z ih↑z | + | ↓z ih↓z |, (2.106)
2 2
this density matrix could arise from an entangled bipartite pure state
|ψiAB with the Schmidt decomposition
1
|ψiAB = √ (| ↑z iA | ↑z iB + | ↓z iA | ↓z iB ) . (2.107)
2
Therefore, the ensemble interpretation of ρA in which either | ↑z iA or
| ↓z iA is prepared (each with probability p = 21 ) can be realized by
performing a measurement of qubit B. We measure qubit B in the
{| ↑z iB , | ↓z iB } basis; if the result | ↑z iB is obtained, we have prepared
| ↑z iA , and if the result | ↓z iB is obtained, we have prepared | ↓z iA .
But as we have already noted, in this case, because ρA has degenerate
eigenvalues, the Schmidt basis is not unique. We can apply simultaneous
unitary transformations to qubits A and B (actually, if we apply U to A
we must apply U ∗ to B as in eq.(2.92)) without modifying the bipartite
pure state |ψiAB . Therefore, for any unit 3-vector n̂, |ψiAB has a Schmidt
decomposition of the form
1
|ψiAB = √ (| ↑n̂ iA | ↑n̂0 iB + | ↓n̂ iA | ↓n̂0 iB ) . (2.108)
2
We see that by measuring qubit B in a suitable basis, we can realize any
interpretation of ρA as an ensemble of two pure states.
This property suggests a mechanism for faster-than-light communica-
tion. Many copies of |ψiAB are prepared. Alice takes all of the A qubits
to the Andromeda galaxy and Bob keeps all of the B qubits on earth.
When Bob wants to send a one-bit message to Alice, he chooses to mea-
sure either σ 1 or σ 3 for all his spins, thus preparing Alice’s spins in either
the {| ↑z iA , | ↓z iA } or {| ↑x iA , | ↓x iA } ensembles. (U is real in this case, so

U = U ∗ and n̂ = n̂0 .) To read the message, Alice immediately measures
her spins to see which ensemble has been prepared.
This scheme has a flaw. Though the two preparation methods are
surely different, both ensembles are described by precisely the same den-
sity matrix ρA . Thus, there is no conceivable measurement Alice can
make that will distinguish the two ensembles, and no way for Alice to tell
what action Bob performed. The “message” is unreadable.
Why, then, do we confidently state that “the two preparation methods
are surely different?” To qualm any doubts about that, imagine that Bob
either (1) measures all of his spins along the ẑ-axis, or (2) measures
all of his spins along the x̂-axis, and then calls Alice on the intergalactic
telephone. He does not tell Alice whether he did (1) or (2), but he does
tell her the results of all his measurements: “the first spin was up, the
second was down,” etc. Now Alice performs either (1) or (2) on her
spins. If both Alice and Bob measured along the same axis, Alice will
find that every single one of her measurement outcomes agrees with what
Bob found. But if Alice and Bob measured along different (orthogonal)
axes, then Alice will find no correlation between her results and Bob’s.
About half of her measurements agree with Bob’s and about half disagree.
If Bob promises to do either (1) or (2), and assuming no preparation or
measurement errors, then Alice will know that Bob’s action was different
than hers (even though Bob never told her this information) as soon as
one of her measurements disagrees with what Bob found. If all their
measurements agree, then if many spins are measured, Alice will have
very high statistical confidence that she and Bob measured along the
same axis. (Even with occasional measurement errors, the statistical test
will still be highly reliable if the error rate is low enough.) So Alice does
have a way to distinguish Bob’s two preparation methods, but in this case
there is certainly no faster-than-light communication, because Alice had
to receive Bob’s phone call before she could perform her test.
2.5.4 Quantum erasure

We had said that the density matrix ρA = 21 I describes a spin in an
incoherent mixture of the pure states | ↑z iA and | ↓z iA . This was to be
distinguished from coherent superpositions of these states, such as
1
| ↑x , ↓x i = √ (| ↑z i ± | ↓z i) ; (2.109)
2
in the case of a coherent superposition, the relative phase of the two states
has observable consequences (distinguishes | ↑x i from | ↓x i). In the case
of an incoherent mixture, the relative phase is completely unobservable.
The superposition becomes incoherent if spin A becomes entangled with

another spin B, and spin B is inaccessible.
Heuristically, the states | ↑z iA and | ↓z iA can interfere (the relative
phase of these states can be observed) only if we have no information
about whether the spin state is | ↑z iA or | ↓z iA . More than that, in-
terference can occur only if there is in principle no possible way to find
out whether the spin is up or down along the z-axis. Entangling spin A
with spin B destroys interference, (causes spin A to decohere) because it
is possible in principle for us to determine if spin A is up or down along
ẑ by performing a suitable measurement of spin B.
But we have now seen that the statement that entanglement causes
decoherence requires a qualification. Suppose that Bob measures spin B
along the x̂-axis, obtaining either the result | ↑x iB or | ↓x iB , and that he
sends his measurement result to Alice. Now Alice’s spin is a pure state
(either | ↑x iA or | ↓x iA ) and in fact a coherent superposition of | ↑z iA and
| ↓z iA . We have managed to recover the purity of Alice’s spin before the
jaws of decoherence could close!
Suppose that Bob allows his spin to pass through a Stern–Gerlach appa-
ratus oriented along the ẑ-axis. Well, of course, Alice’s spin can’t behave
like a coherent superposition of | ↑z iA and | ↓z iA ; all Bob has to do is
look to see which way his spin moved, and he will know whether Al-
ice’s spin is up or down along ẑ. But suppose that Bob does not look.
Instead, he carefully refocuses the two beams without maintaining any
record of whether his spin moved up or down, and then allows the spin to
pass through a second Stern–Gerlach apparatus oriented along the x̂-axis.
This time he looks, and communicates the result of his σ 1 measurement
to Alice. Now the coherence of Alice’s spin has been restored!
This situation has been called a quantum eraser. Entangling the two
spins creates a “measurement situation” in which the coherence of | ↑z iA
and | ↓z iA is lost because we can find out if spin A is up or down along ẑ by
observing spin B. But when we measure spin B along x̂, this information
is “erased.” Whether the result is | ↑x iB or | ↓x iB does not tell us anything
about whether spin A is up or down along ẑ, because Bob has been careful
not to retain the “which way” information that he might have acquired
by looking at the first Stern–Gerlach apparatus. Therefore, it is possible
again for spin A to behave like a coherent superposition of | ↑z iA and
| ↓z iA (and it does, after Alice hears about Bob’s result).
We can best understand the quantum eraser from the ensemble view-
point. Alice has many spins selected from an ensemble described by
ρA = 12 I, and there is no way for her to observe interference between
| ↑z iA and | ↓z iA . When Bob makes his measurement along x̂, a partic-
ular preparation of the ensemble is realized. However, this has no effect
that Alice can perceive – her spin is still described by ρA = 12 I as before.
But, when Alice receives Bob’s phone call, she can select a subensemble
of her spins that are all in the pure state | ↑x iA . The information that
Bob sends allows Alice to distill purity from a maximally mixed state.
Another wrinkle on the quantum eraser is sometimes called delayed
choice. This just means that the situation we have described is really
completely symmetric between Alice and Bob, so it can’t make any dif-
ference who measures first. (Indeed, if Alice’s and Bob’s measurements
are spacelike separated events, there is no invariant meaning to which
came first; it depends on the frame of reference of the observer.) Alice
could measure all of her spins today (say along x̂) before Bob has made
his mind up how he will measure his spins. Next week, Bob can decide
to “prepare” Alice’s spins in the states | ↑n̂ iA and | ↓n̂ iA (that is the
“delayed choice”). He then tells Alice which were the | ↑n̂ iA spins, and
she can check her measurement record to verify that
hσ1 in̂ = n̂ · x̂ . (2.110)
The results are the same, irrespective of whether Bob “prepares” the spins
before or after Alice measures them.
We have claimed that the density matrix ρA provides a complete phys-
ical description of the state of subsystem A, because it characterizes all
possible measurements that can be performed on A. One might object
that the quantum eraser phenomenon demonstrates otherwise. Since the
information received from Bob enables Alice to recover a pure state from
the mixture, how can we hold that everything Alice can know about A is
encoded in ρA ?
I prefer to say that quantum erasure illustrates the principle that “in-
formation is physical.” The state ρA of system A is not the same thing
as ρA accompanied by the information that Alice has received from Bob.
This information (which attaches labels to the subensembles) changes the
physical description. That is, we should include Alice’s “state of knowl-
edge” in our description of her system. An ensemble of spins for which
Alice has no information about whether each spin is up or down is a dif-
ferent physical state than an ensemble in which Alice knows which spins
are up and which are down. This “state of knowledge” need not really
be the state of a human mind; any (inanimate) record that labels the
subensemble will suffice.
2.5.5 The HJW theorem

So far, we have considered the quantum eraser only in the context of
a single qubit, described by an ensemble of equally probable mutually
orthogonal states, (i.e., ρA = 12 I). The discussion can be considerably
generalized.
We have already seen that a mixed state of any quantum system can
be realized as an ensemble of pure states in an infinite number of different
ways. For a density matrix ρA , consider one such realization:
X X
ρA = pi |ϕi ihϕi |, pi = 1. (2.111)
i
Here the states {|ϕi iA } are all normalized vectors, but we do not assume
that they are mutually orthogonal. Nevertheless, ρA can be realized as
an ensemble, in which each pure state |ϕi ihϕi | occurs with probability pi .
For any such ρA , we can construct a “purification” of ρA , a bipartite
pure state |Φ1 iAB that yields ρA when we perform a partial trace over
HB . One such purification is of the form
X√
|Φ1 iAB = pi |ϕi iA ⊗ |αi iB , (2.112)
i
where the vectors |αi iB ∈ HB are mutually orthogonal and normalized,

hαi |αj i = δij . (2.113)
Clearly, then,
trB (|Φ1 ihΦ1 |) = ρA . (2.114)
Furthermore, we can imagine performing an orthogonal measurement in
system B that projects onto the |αi iB basis. (The |αi iB ’s might not span
HB , but in the state |Φ1 iAB , measurement outcomes orthogonal to all
the |αi iB ’s never occur.) The outcome |αi iB will occur with probability
pi , and will prepare the pure state |ϕi ihϕi | of system A. Thus, given
the purification |Φ1 iAB of ρA , there is a measurement we can perform in
system B that realizes the |ϕi iA ensemble interpretation of ρA . When
the measurement outcome in B is known, we have successfully extracted
one of the pure states |ϕi iA from the mixture ρA .
What we have just described is a generalization of preparing | ↑z iA by
measuring spin B along ẑ (in our discussion of two entangled qubits). But
to generalize the notion of a quantum eraser, we wish to see that in the
state |Φ1 iAB , we can realize a different ensemble interpretation of ρA by
performing a different measurement of B. So let
X
ρA = qµ |ψµ ihψµ |, (2.115)
µ
be another realization of the same density matrix ρA as an ensemble of

pure states. For this ensemble as well, there is a corresponding purification
X√
|Φ2 iAB = qµ |ψµ iA ⊗ |βµ iB , (2.116)
µ
where again the {|βµ iB ’s} are orthonormal vectors in HB . So in the state
|Φ2 iAB , we can realize the ensemble by performing a measurement in HB
that projects onto the {|βµ iB } basis.
Now, how are |Φ1 iAB and |Φ2 iAB related? In fact, we can easily show
that
|Φ1 iAB = (I A ⊗ U B ) |Φ2 iAB ; (2.117)
the two states differ by a unitary change of basis acting in HB alone, or
X√
|Φ1 iAB = qµ |ψµ iA ⊗ |γµ iB , (2.118)
µ
where
|γµ iB = U B |βµ iB , (2.119)
is yet another orthonormal basis for HB . We see, then, that there is a sin-
gle purification |Φ1 iAB of ρA , such that we can realize either the {|ϕi iA }
ensemble or {|ψµ iA } ensemble by choosing to measure the appropriate
observable in system B!
Similarly, we may consider many ensembles that all realize ρA , where
the maximum number of pure states appearing in any of the ensembles
is d. Then we may choose a Hilbert space HB of dimension d, and a
pure state |ΦiAB ∈ HA ⊗ HB , such that any one of the ensembles can
be realized by measuring a suitable observable of B. This is the HJW
theorem (for Hughston, Jozsa, and Wootters); it expresses the quantum
eraser phenomenon in its most general form.
In fact, the HJW theorem is an easy corollary to the Schmidt decom-
position. Both |Φ1 iAB and |Φ2 iAB have Schmidt decompositions, and
because both yield the same ρA when we take the partial trace over B,
these decompositions must have the form
Xp
|Φ1 iAB = λk |kiA ⊗ |k10 iB ,
k
Xp
|Φ2 iAB = λk |kiA ⊗ |k20 iB , (2.120)
k
where the λk ’s are the eigenvalues of ρA and the |kiA ’s are the correspond-
ing eigenvectors. But since {|k10 iB } and {|k20 iB } are both orthonormal
bases for HB , there is a unitary U B such that
|k10 iB = U B |k20 iB , (2.121)
from which eq.(2.117) immediately follows.
In the ensemble of pure states described by Eq. (2.111), we would say
that the pure states |ϕi iA are mixed incoherently — an observer in sys-
tem A cannot detect the relative phases of these states. Heuristically, the
reason that these states cannot interfere is that it is possible in principle

to find out which representative of the ensemble is actually realized by
performing a measurement in system B, a projection onto the orthonor-
mal basis {|αi iB }. However, by projecting onto the {|γµ iB } basis instead,
and relaying the information about the measurement outcome to system
A, we can extract one of the pure states |ψµ iA from the ensemble, even
though this state may be a coherent superposition of the |ϕi iA ’s. In effect,
measuring B in the {|γµ iB } basis “erases” the “which way” information
(whether the state of A is |ϕi iA or |ϕj iA ). In this sense, the HJW theorem
characterizes the general quantum eraser. The moral, once again, is that
information is physical — the information acquired by measuring system
B, when relayed to A, changes the physical description of a state of A.
2.6 How far apart are two quantum states?

2.6.1 Fidelity and Uhlmann’s theorem
The distinguishability of two pure states |ψi and |ϕi is quantified by the
deviation from 1 of their overlap |hϕ|ψi|2 , also called fidelity. For two
density operators ρ and σ the fidelity is defined by
q 2
1 1
F (ρ, σ) ≡ tr ρ 2 σρ 2 . (2.122)
(Some authors use the name “fidelity” for the square root of this quantity.)
The fidelity is nonnegative, vanishes if ρ and σ have support on mutually
orthogonal subspaces, and attains its maximum value 1 if and only if the
two states are identical. If ρ = |ψihψ| is a pure state, then the fidelity is
F (ρ, σ) = hψ|σ|ψi. (2.123)
Suppose we perform an orthogonal measurement on ρ with the two out-

comes: “Yes” if the state if |ψi, “No” if the state is orthogonal to |ψi.
Then the fidelity is the probability that the outcome is “Yes.”
We may also express the fidelity in terms of the L1 norm,
1 1 2
F (ρ, σ) = σ 2 ρ 2 , (2.124)

1
where p
kAk1 = tr A† A. (2.125)
The L1 norm is also sometimes called the trace norm. (For Hermitian
A, kAk1 is just the sum of the absolute values of its eigenvalues.) The
fidelity F (ρ, σ) is actually symmetric in its two arguments, although the
symmetry is not manifest in eq. (2.122). To verify the symmetry, note

that for any Hermitian A and B, the L1 norm obeys
kABk1 = kBAk1 . (2.126)
This holds because BAAB and ABBA have the same eigenvalues —
if |ψi is an eigenstate of ABBA with eigenvalue λ, the BA|ψi is an
eigenstate of BAAB with eigenvalue λ.
It is useful to know how the fidelity of two density operators is related
to the overlap of their purifications. We say that |ΦiAB is a purification
of the density operator ρA if
ρA = trB (|ΦihΦ|) . (2.127)
If X
ρ= pi |iihi| (2.128)
i
(where |iiA is an orthonormal basis for system A), then a particular pu-
rification of ρ has the form
X√
|Φρ i = pi |iiA ⊗ |iiB (2.129)
i
where |iiB is an orthonormal basis for system B. According to the HJW

theorem, a general purification has the form
|Φρ (V )i = I ⊗ V |Φρ i (2.130)
where V is unitary, which may also be written

1
|Φρ (V )i = ρ 2 ⊗ V |Φ̃i, (2.131)
where |Φ̃i is the unconventionally normalized maximally entangled state

X
|Φ̃iAB = |iiA ⊗ |iiB . (2.132)
i
If ρ and σ are two density operators on A, the inner product of their

purifications on AB can be expressed as
1 1
hΦσ (W )|Φρ (V )i = hΦ̃|σ 2 ρ 2 ⊗ W † V |Φ̃i. (2.133)
Noting that
X X
U ⊗ I|Φ̃i = |ji ⊗ |iiUji = |ji ⊗ |iiUijT = I ⊗ U T |Φ̃i, (2.134)
i,j i,j
we have
1 1
1 1
hΦσ (W )|Φρ (V )i = hΦ̃|σ 2 ρ 2 U ⊗ I|Φ̃i = tr σ 2 ρ 2 U , (2.135)
T
where U = W † V .
Now we may use the polar decomposition
p
0
A = U A† A, (2.136)
where U 0 is unitary, to rewrite the inner product as

q X
1 1
0
hΦσ (W )|Φρ (V )i = tr U U ρ 2 σρ 2 = λa ha|U U 0 |ai, (2.137)
a
q
1 1
where {λa } are the nonnegative eigenvalues of ρ 2 σρ 2 and {|ai} are
the corresponding eigenvectors. It is now evident that the inner product
has the largest possible absolute value when we choose U = U 0−1 , and
hence we conclude
q 2
1 1
F (ρ, σ) = tr ρ 2 σρ 2 = max |hΦσ (W )|Φρ (V )i|2 . (2.138)
V ,W
The fidelity of two density operators is the maximal possible overlap of

their purifications, a result called Uhlmann’s theorem.
One corollary of Uhlmann’s theorem is the monotonicity of fidelity:
F (ρAB , ρAB ) ≤ F (ρA , ρA ), (2.139)
which says that tracing out a subsystem cannot decrease the fidelity of
two density operators. Monotonicity means, unsurprisingly, that throwing
away a subsystem does not make two quantum states easier to distinguish.
It follows from Uhlmann’s theorem because any state purifying ρAB also
provides a purification of A; therefore the optimal overlap of the purifi-
cations of ρAB and σ AB is surely achievable by purifications of ρA and
σA.
2.6.2 Relations among distance measures

There are other possible ways besides fidelity for quantifying the difference
between quantum states ρ and σ, such as the distance between the states
using the L1 or L2 norm,
kρ − σk1 or kρ − σk2 , (2.140)

where the L2 norm of an operator is defined by

p
kAk2 = trA† A. (2.141)
(For Hermitian A, kAk2 is the square root of the sum of the squares
of its eigenvalues.) The L1 distance is a particularly natural measure
of state distinguishability, because (as shown in Exercise 2.5) it can be
interpreted as the distance between the corresponding probability dis-
tributions achieved by the optimal measurement for distinguishing the
states. Although the fidelity, L1 distance, and L2 distance are not simply
related to one another in general, there are useful inequalities relating
these different measures. √
If {|λi |, i = 0, 1, 2, . . . d−1} denotes the eigenvalues of A† A, then
v
d−1
X
ud−1
uX
kAk1 = |λi |; kAk2 = t |λi |2 . (2.142)
i=0 i=0
If we regard kAk1 as the inner product of the two vectors (1, 1, 1, . . . , 1)

and (|λ0 |, |λ1 |, . . . , |λd−1 |), then from the Cauchy-Schwarz inequality we
find √
kAk1 ≤ d kAk2 . (2.143)
√
Because of the factor of d on the right hand side, for a high-dimensional
system density operators which are close together in the L2 norm might
not be close in the L1 norm.
There is, however, a dimension-independent inequality relating the L1
distance between ρ and σ and the L2 distance between their square roots:
√ √
k ρ − σk22 ≤ kρ − σk1 . (2.144)
To derive this inequality, we first expand the difference of square roots in

its basis of eigenvectors,
√ √ X
ρ− σ = λi |iihi|, (2.145)
i
and note that the absolute value of this difference may be written as
√ √ X √ √ √ √
| ρ − σ| = |λi | |iihi| = ρ− σ U =U ρ − σ , (2.146)
i
where X
U= sign(λi )|iihi|. (2.147)
i
Using
1 √ √ √ √ 1 √ √ √ √
ρ−σ = ρ− σ ρ+ σ + ρ+ σ ρ − σ (2.148)
2 2
and the cyclicity of the trace, we find
√ √ √ √ X √ √
tr (ρ − σ) U = tr| ρ − σ| ρ + σ = |λi | hi| ρ + σ|ii
i
X √ √ X √ √
≥ |λi | hi| ρ − σ|ii = |λi |2 = k ρ − σk22 .
i i
(2.149)
Finally, using
kρ − σk1 = tr|ρ − σ| ≥ tr (ρ − σ) U , (2.150)
which is true for any unitary U , we obtain eq.(2.144).

This L2 distance between square roots can be related to fidelity. First
we note that
√ √ √ √
ρ − σ 2 = tr √ρ − σ 2 = 2 − 2 tr √ρ σ , (2.151)
2
√
since tr ρ = tr σ = 1. From the polar√ decomposition A = U A† A
†
(where U is unitary), we see that tr A A ≥ |tr A|, and therefore
√ √
q
p 1 1
F (ρ, σ) = tr ρ 2 σρ 2 ≥ tr ρ σ (2.152)
and hence
p 1 √ √ 2 1
F (ρ, σ) ≥ 1 − ρ − σ 2 ≥ 1 − kρ − σk1 . (2.153)
2 2
Eq.(2.153) tells us that if ρ and σ are close to one another in the L1 norm,
then their fidelity is close to one.
The L1 distance also provides an upper bound on fidelity:
1
F (ρ, σ) ≤ 1 − kρ − σk21 . (2.154)
4
To derive eq.(2.154) we first show that it holds with equality for pure
states. Any two vectors |ψi and |ϕi lie in some two-dimensional subspace,
and by choosing a basis and phase conventions appropriately we may write

cos θ/2 sin θ/2
|ψi = , |ϕi = , (2.155)
sin θ/2 cos θ/2
2.7 Summary 41
for some angle θ; then

cos θ 0
|ψihψ| − |ϕihϕ| = . (2.156)
0 − cos θ
and
|hϕ|ψi|2 = sin2 θ. (2.157)
Therefore,
k|ψihψ| − |ϕihϕ|k21 = (2 cos θ)2 = 4 (1 − F (|ψihψ|, |ϕihϕ|)) . (2.158)
Next, note that L1 distance, like fidelity, is monotonic:
kρAB − σ AB k1 ≥ kρA − σ A k1 . (2.159)
This follows because the L1 distance is the optimal distance between

probability distributions when we measure the two states, and the optimal
measurement for distinguishing ρA and σ A is also a possible measurement
for ρAB and σ AB , one that happens to act trivially on B.
Finally, we invoke Uhlmann’s theorem. If ρAB and σ AB are the purifi-
cations of ρA and σ A with the largest possible overlap, then
1
F (ρA , σ A ) = F (ρAB , σ AB ) = 1 −kρ − σ AB k21
4 AB
1
≤ 1 − kρA − σ A k21 , (2.160)
4
where the first equality uses Uhlmann’s theorem, the second uses
eq.(2.158), and the final inequality uses monotonicity. By combining
eq.(2.160) with eq.(2.153) we have
p 1 p
1− F (ρ, σ) ≤ kρ − σk1 ≤ 1 − F (ρ, σ); (2.161)
2
hence ρ and σ are close to one another in the L1 norm if and only if their
fidelity is close to one.
2.7 Summary
Axioms. The arena of quantum mechanics is a Hilbert space H. The
fundamental assumptions are:
(1) A state is a ray in H.
(2) An observable is a self-adjoint operator on H.
(3) A measurement is an orthogonal projection.
(4) Time evolution is unitary.
(5) A composite system AB is described by the tensor product HA ⊗ HB .
Density operator. But if we confine our attention to only a portion

of a larger quantum system, assumptions (1)-(4) need not be satisfied. In
particular, a quantum state is described not by a ray, but by a density
operator ρ, a nonnegative operator with unit trace. The density operator
is pure (and the state can be described by a ray) if ρ2 = ρ; otherwise, the
state is mixed. An observable M has expectation value tr(M ρ) in this
state.
Qubit. A quantum system with a two-dimensional Hilbert space is
called a qubit. The general density matrix of a qubit is
1
ρ(P~ ) = (I + P~ · ~
σ) (2.162)
2
where P~ is a three-component vector of length |P~ | ≤ 1. Pure states have

|P~ | = 1.
Schmidt decomposition. For any pure state |ψiAB of a bipartite
system, there are orthonormal bases {|iiA } for HA and {|i0 iB } for HB
such that X√
|ψiAB = pi |iiA ⊗ |i0 iB ; (2.163)
i
Eq.(2.163) is called the Schmidt decomposition of |ψiAB . In a bipartite

pure state, subsystems A and B separately are described by density op-
erators ρA and ρB ; it follows from eq.(2.163) that ρA and ρB have the
same nonvanishing eigenvalues (the pi ’s). The number of nonvanishing
eigenvalues is called the Schmidt number of |ψiAB . A bipartite pure state
is said to be entangled if its Schmidt number is greater than one.
Ensembles. The density operators on a Hilbert space form a convex
set, and the pure states are the extremal points of the set. A mixed
state of a system A can be prepared as an ensemble of pure states in
many different ways, all of which are experimentally indistinguishable if
we observe system A alone. Given any mixed state ρA of system A,
any preparation of ρA as an ensemble of pure states can be realized in
principle by performing a measurement in another system B with which A
is entangled. In fact given many such preparations of ρA , there is a single
entangled state of A and B such that any one of these preparations can
be realized by measuring a suitable observable in B (the HJW theorem).
By measuring in system B and reporting the measurement outcome to
system A, we can extract from the mixture a pure state chosen from one
of the ensembles.
2.8 Exercises 43
1 1 2
Fidelity. The fidelity F (ρ, σ) = σ 2 ρ 2 , quantifies the distinguisha-

1
bility of two density operators — it is the maximum overlap achieved by
their purifications (Uhlmann’s theorem). The fidelity F (ρ, σ) is close to
one if and only if the L1 distance kρ − σk1 is small.
Further important ideas are developed in the Exercises.
2.8 Exercises
2.1 Fidelity of measurement
a) For two states |ψ1 i and |ψ2 i in an N -dimensional Hilbert space,

define the relative angle θ between the states by
|hψ2 |ψ1 i| ≡ cos θ , (2.164)
where 0 ≤ θ ≤ π/2. Suppose that the two states are selected

at random. Find the probability distribution p(θ)dθ for the
relative angle. Hint: We can choose a basis such that
|ψ1 i = (1, ~0)

|ψ2 i = (eiϕ cos θ, ψ2⊥ ) . (2.165)
(2.166)
“Selected at random” means that the probability distribution

for the normalized vector |ψ2 i is uniform on the (real) (2N −1)-
sphere (this is the unique distribution that is invariant under
arbitary unitary transformations). Note that, for fixed θ, eiϕ
parametrizes a circle of radius cos θ, and |ψ2⊥ i is a vector that
lies on a 2N − 3 sphere of radius sin θ.
b) A density operator ρ is said to approximate a pure state |ψi with
fidelity
F = hψ|ρ|ψi . (2.167)
Imagine that a state |ψ1 i in an N -dimensional Hilbert space is
selected at random, and we guess at random that the state is
|ψ2 i. On the average, what will be the fidelity of our guess?
c) When we measure, we collect information and cause a distur-
bance – an unknown state is replaced by a different state that
is known. Suppose that the state |ψi is selected at random,
and then an orthogonal measurement is performed, projecting
onto an orthonormal basis {|E a i}. After the measurement, the
state (averaged over all possible outcomes) is described by the

density matrix X
ρ= E a |ψihψ|E a , (2.168)
a
where E a = |E a ihE a |; this ρ approximates |ψi with fidelity

X
F = (hψ|E a |ψi)2 . (2.169)
a
Evaluate F , averaged over the choice of |ψi. Hint: Use Bayes’s

rule and the result from (a) to find the probability distribution
for the angle θ between the state
|ψi and
the projected state
E a |ψi/kE a |ψik. Then evaluate cos2 θ in this distribution.
Remark: The improvement in F in the answer to (c) compared to
the answer to (b) is a crude measure of how much we learned by
making the measurement.
2.2 Measurement without disturbance?
Charlie prepares the system A in one of two nonorthogonal states,
|ϕiA or |ϕ̃iA , and he challenges Alice to collect some information
about which state he prepared without in any way disturbing the
state. Alice has an idea about how to meet the challenge.
Alice intends to prepare a second “ancillary” system B in the state
|βiB , and then apply to the composite system AB a unitary trans-
formation U that acts according to
U : |ϕiA ⊗ |βiB → |ϕiA ⊗ |β 0 iB

|ϕ̃iA ⊗ |βiB → |ϕ̃iA ⊗ |β̃ 0 iB , (2.170)
which does indeed leave the state of system A undisturbed. Then

she plans to perform a measurement on system B that is designed
to distinguish the states |β 0 iB and |β̃ 0 iB .
a) What do you think of Alice’s idea? Hint: What does the uni-
tarity of U tell you about how the states |β 0 iB and |β̃ 0 iB are
related to one another?
b) Would you feel differently if the states |ϕiA and |ϕ̃iA were or-
thogonal?
2.3 Quantum bit commitment

The Yankees are playing the Dodgers in the World Series. Alice is
sure that she knows who will win. Alice doesn’t like Bob, so she
2.8 Exercises 45
doesn’t want to tell him who the winner will be. But after the Series
is over, Alice wants to be able to convince Bob that she knew the
outcome all along. What to do?
Bob suggests that Alice write down her choice (0 if the Yankees will
win, 1 if the Dodgers will win) on a piece of paper, and lock the
paper in a box. She is to give the box to Bob, but she will keep the
key for herself. Then, when she is ready to reveal her choice, she
will send the key to Bob, allowing him to open the box and read
the paper.
Alice rejects this proposal. She doesn’t trust Bob, and she knows
that he is a notorious safe cracker. Who’s to say whether he will be
able to open the box and sneak a look, even if he doesn’t have the
key?
Instead, Alice proposes to certify her honesty in another way, using
quantum information. To commit to a value a ∈ {0, 1} of her bit, she
prepares one of two distinguishable density operators (ρ0 or ρ1 ) of
the bipartite system AB, sends system B to Bob, and keeps system
A for herself. Later, to unveil her bit, Alice sends system A to Bob,
and he performs a measurement to determine whether the state of
AB is ρ0 or ρ1 . This protocol is called quantum bit commitment.
We say that the protocol is binding if, after commitment, Alice is
unable to change the value of her bit. We say that the protocol is
concealing if, after commitment and before unveiling, Bob is unable
to discern the value of the bit. The protocol is secure if it is both
binding and concealing.
Show that if a quantum bit commitment protocol is concealing, then
it is not binding. Thus quantum bit commitment is insecure.
Hint: First argue that without loss of generality, we may assume
that the states ρ0 and ρ1 are both pure. Then apply the HJW
Theorem.
Remark: Note that the conclusion that quantum bit commitment
is insecure still applies even if the shared bipartite state (ρ0 or ρ1 ) is
prepared during many rounds of quantum communication between
Alice and Bob, where in each round one party performs a quantum
operation on his/her local system and on a shared message system,
and then sends the message system to the other party.
2.4 Completeness of subsystem correlations
Consider a bipartite system AB. Suppose that many copies of the
(not necessarily pure) state ρAB have been prepared. An observer
Alice with access only to subsystem A can measure the expectation
value of any observable of the form M A ⊗I B , while an observer Bob

with access only to subsystem B can measure the expectation value
of any observable of the form I A ⊗ N B . Neither of these observers,
working alone, can expect to gain enough information to determine
the joint state ρAB .
But now suppose that Alice and Bob can communicate, exchang-
ing (classical) information about how their measurement outcomes
are correlated. Thereby, they can jointly determine the expecta-
tion value of any observable of the form M A ⊗ N B (an observable
whose eigenstates are separable direct products states of the form
|ϕiA ⊗ |χiB ).
The point of this exercise is to show that if Alice and Bob have
complete knowledge of the nature of the correlations between sub-
systems A and B (know the expectation values of any tensor product
observable M A ⊗N B ), then in fact they know everything about the
bipartite state ρAB – there will be no surprises when they measure
entangled observables, those whose eigenstates are entangled states.
a) Let {M a , a = 1, 2, . . . , d2 } denote a set of d2 linearly inde-

pendent self-adjoint operators acting on a Hilbert space H of
dimension d. Show that if ρ is a density operator acting on H,
and tr (ρM a ) is known for each a, then hϕ|ρ|ϕi is known for
any vector |ϕi in H.
b) Show that if hϕ|ρ|ϕi is known for each vector |ϕi, then ρ is
completely known.
c) Show that if {M a } denotes a basis for self-adjoint operators on
HA , and {N b } denotes a basis for self-adjoint operators on
HB , then {M a ⊗ N b } is a basis for the self-adjoint operators
on HA ⊗ HB .
Remark: It follows from (c) alone that the correlations of the “lo-
cal” observables determine the expectation values of all the observ-
ables. Parts (a) and (b) serve to establish that ρ is completely
determined by the expectation values of a complete set of observ-
ables.
d) State and prove the result corresponding to (c) that applies to

a n-part system with Hilbert space H1 ⊗ H2 ⊗ · · · ⊗ Hn .
e) Discuss how the world would be different if quantum states
resided in a real Hilbert space rather than a complex Hilbert
space. Consider, in particular, whether (c) is true for symmet-
ric operators acting on a real vector space.
2.8 Exercises 47
2.5 Optimal measurement distinguishing two quantum states

Consider two quantum states described by density operators ρ and
σ in an d-dimensional Hilbert space, and consider the complete
orthogonal measurement {E a , a = 0, 1, 2, . . . d−1}, where the E a ’s
are one-dimensional projectors satisfying
d−1
X
E a = I. (2.171)
a=0
When the measurement is performed, outcome a occurs with proba-

bility pa = tr ρE a if the state is ρ and with probability qa = tr σE a
if the state is σ.
The L1 distance between the two probability distributions is defined
as
d−1
X
d(p, q) ≡ kp − qk1 ≡ |pa − qa | ; (2.172)
a=0
this distance is zero if the two distributions are identical, and attains
its maximum value two if the two distributions have support on
disjoint sets.
a) Show that
d−1
X
d(p, q) ≤ |λi | = kρ − σk1 ≡ d(ρ, σ), (2.173)
i=0
where the λi ’s are the eigenvalues of the Hermitian operator

ρ − σ. Hint: Working in the basis in which ρ − σ is diagonal,
find an expression for |pa − qa |, and then find an upper bound
on |pa − qa |. Finally, use the completeness property eq.(2.171)
to bound d(p, q).
b) Find a choice for the orthogonal projector {E a } that saturates
the upper bound eq.(2.173).
c) If the states ρ = |ψihψ| and σ = |ϕihϕ| are pure, show that
d(p, q) ≤ 2 k|ψi − |ϕik (2.174)
where k · k denotes the Hilbert space norm.
2.6 What probability distributions are consistent with a mixed

state?
A density operator ρ, expressed in the orthonormal basis {|αi i} that

diagonalizes it, is
X
ρ= pi |αi ihαi |. (2.175)
i
We would like to realize this density operator as an ensemble of pure

states {|ϕµ i}, where |ϕµ i is prepared with a specified probability qµ .
This preparation is possible if the |ϕµ i’s can be chosen so that
X
ρ= qµ |ϕµ ihϕµ |. (2.176)
µ
We say that a probability vector q (a vector whose components are

nonnegative real numbers that sum to 1) is majorized by a proba-
bility vector p (denoted q ≺ p), if there exists a doubly stochastic
matrix D such that X
qµ = Dµi pi . (2.177)
i
A matrix is doubly
P stochastic
P if its entries are nonnegative real num-
bers such that µ Dµi = i Dµi = 1. That the columns sum to one
assures that D maps probability vectors to probability vectors (i.e.,
is stochastic). That the rows sum to one assures that D maps the
uniform distribution to itself. Applied repeatedly, D takes any input
distribution closer and closer to the uniform distribution (unless D
is a permutation, with one nonzero entry in each row and column).
Thus we can view majorization as a partial order on probability
vectors such that q ≺ p means that q is more nearly uniform than p
(or equally close to uniform, in the case where D is a permutation).
Show that normalized pure states {|ϕµ i} exist such that eq.(2.176)
is satisfied if and only if q ≺ p, where p is the vector of eigenvalues
of ρ.
Hint: Recall that, according to the HJW Theorem, if eq.(2.175)
and eq.(2.176) are both satisfied then there is a unitary matrix Vµi
such that X√
√
qµ |ϕµ i = pi Vµi |αi i. (2.178)
i
You may also use (but need not prove) Horn’s Lemma: if q ≺ p,
then there exists a unitary (in fact, orthogonal) matrix Uµi such
that q = Dp and Dµi = |Uµi |2 .
2.7 Alice does Bob a favor

2.8 Exercises 49
Alice, in Anaheim, and Bob, in Boston, share a bipartite pure state

|Ψi, which can be expressed in the Schmidt form
X√
|Ψi = pi |αi i ⊗ |βi i, (2.179)
i
where {|αi i} is an orthonormal basis for Alice’s system A, {|βi i}

is an orthonormal basis for Bob’s system B, and the {pi } are non-
negative real numbers summing to 1. Bob is supposed to perform a
complete orthogonal local measurement on B, characterized by the
set of projectors {E Ba } — if the measurement outcome is a, then
Bob’s measurement prepares the state
I ⊗ EB

a |Ψi
|Ψi 7→ |Ψa i = . (2.180)
hΨ| I ⊗ E B

1/2
a |Ψi
|Ψa i can also be expressed in the Schmidt form if we choose ap-

propriate orthonormal bases for A and B that depend on the mea-
surement outcome. The new Schmidt basis elements can be written
as
0 0
|αa,i i = UA
a |αi i, |βa,i i = UB
a |βi i, (2.181)
where U A B
a , U a are unitary.
Unfortunately, Bob’s measurement apparatus is broken, though he
still has the ability to perform local unitary transformations on B.
Show that Alice can help Bob out by performing a measurement
that is “locally equivalent” to Bob’s. That is, there are orthogonal
projectors {E A A B
a } and unitary transformations V a , V a such that
!
A

E a ⊗ I |Ψi
|Ψa i = V Aa ⊗Va
B
(2.182)
hΨ| E A

1/2
a ⊗ I |Ψi
for each a, and furthermore, both Alice’s measurement and Bob’s

yield outcome a with the same probability. This means that instead
of Bob doing the measurement, the same effect can be achieved if
Alice measures instead, tells Bob the outcome, and both Alice and
Bob perform the appropriate unitary transformations. Construct
EA A B
a (this is most conveniently done by expressing both E a and E a
in the Schmidt bases for |Ψi) and express V A B
a and V a in terms of
A B
U a and U a .
Remark: This result shows that for any protocol involving local
operations and “two-way” classical communication (2-LOCC) that
transforms an initial bipartite pure state to a final bipartite pure
state, the same transformation can be achieved by a “one-way” (1-

LOCC) protocol in which all classical communication is from Alice
to Bob (the Lo-Popescu Theorem). In a two-way LOCC protocol,
Alice and Bob take turns manipulating the state for some finite
(but arbitrarily large) number of rounds. In each round, one party
or the other performs a measurement on her/his local system and
broadcasts the outcome to the other party. Either party might use
a local “ancilla” system in performing the measurement, but we
may include all ancillas used during the protocol in the bipartite
pure state |Ψi. Though a party might discard information about
the measurement outcome, or fail to broadcast the information to
the other party, we are entitled to imagine that the complete in-
formation about the outcomes is known to both parties at each
step (incomplete information is just equivalent to the special case
in which the parties choose not to use all the information). Thus
the state is pure after each step.
The solution to the exercise shows that a round of a 2-LOCC pro-
tocol in which Bob measures can be simulated by an operation per-
formed by Alice and a local unitary applied by Bob. Thus, we can
allow Alice to perform all the measurements herself. When she is
through she sends all the outcomes to Bob, and he can apply the
necessary product of unitary transformations to complete the pro-
tocol.
2.8 The power of noncontextuality

We may regard a quantum state as an assignment of probabilities
to projection operators. That is, according to Born’s rule, if ρ is a
density operator and E is a projector, then p(E) = tr (ρE) is the
probability that the outcome E occurs, if E is one of a complete
set of orthogonal projectors associated with a particular quantum
measurement. A notable feature of this rule is that the assignment of
a probability p(E) to E is noncontextual. This means that, while the
probability p(E) depends on the state ρ, it does not depend on how
we choose the rest of the projectors that complete the orthogonal
set containing E.
In a hidden variable theory, the probabilistic description of quantum
measurement is derived from a more fundamental and complete de-
terministic description. The outcome of a measurement could be
perfectly predicted if the values of the hidden variables were pre-
cisely known — then the probability p(E) could take only the values
0 and 1. The standard probabilistic predictions of quantum theory
arise when we average over the unknown values of the hidden vari-
2.8 Exercises 51
ables. The purpose of this exercise is to show that such deterministic

assignments conflict with noncontextuality. Thus a hidden variable
theory, if it is to agree with the predictions of quantum theory after
averaging, must be contextual.
Let {I, X, Y , Z} denote the single-qubit observables

1 0 0 1
I = , X = ,
0 1 1 0

0 −i 1 0
Y = , Z = , (2.183)
i 0 0 −1
and consider the nine two-qubit observables:
X ⊗I I ⊗X X ⊗X
I ⊗Y Y ⊗I Y ⊗Y . (2.184)
X ⊗Y Y ⊗X Z ⊗Z
The three observables in each row and in each column are mutu-
ally commuting, and so can be simultaneously diagonalized. In fact
the simultaneous eigenstates of any two operators in a row or col-
umn (the third operator is not independent of the other two) are
a complete basis for the four-dimensional Hilbert space of the two
qubits. Thus we can regard the array eq.(2.184) as a way of present-
ing six different ways to choose a complete set of one-dimensional
projectors for two qubits.
Each of these observables has eigenvalues ±1, so that in a determin-
istic and noncontextual model of measurement (for a fixed value of
the hidden variables), each can be assigned a definite value, either
+1 or −1.
a) Any noncontextual deterministic assignment has to be consis-

tent with the multiplicative structure of the observables. For
example, the product of the three observables in the top row is
the identity I ⊗ I. Therefore, if we assign a value ±1 to each
operator, the number of −1’s assigned to the first row must be
even. Compute the product of the three observables in each
row and each column to find the corresponding constraints.
b) Show that there is no way to satisfy all six constraints simulta-
neously.
Thus a deterministic and noncontextual assignment does not exist.
2.9 Schmidt-decomposable states

We have seen that any vector in a bipartite Hilbert space HAB =

HA ⊗ HB can be expressed in the Schmidt form: Given the vector
|ψiAB ∈ HA ⊗ HB , where HA and HB are both d-dimensional, we
can choose orthonormal bases {|iiA } for HA and {|iiB } for HB so
that
d−1 p
X
|ψiAB = λi |iiA ⊗ |iiB , (2.185)
i=0
where the λi ’s are real and nonnegative. (We’re not assuming here
that the vector has unit norm, so the sum of the λi ’s is not con-
strained.) Eq.(2.185) is called the Schmidt decomposition of the
vector |ψiAB . Of course, the bases in which the vector has the
Schmidt form depend on which vector |ψiAB is being decomposed.
A unitary transformation acting on HAB is called a local unitary
if it is a tensor product U A ⊗ U B , where U A , U B are unitary
transformations acting on HA , HB respectively. The word “local”
is used because if the two parts A and B of the system are widely
separated from one another, so that Alice can access only part A
and Bob can access only part B, then Alice and Bob can apply this
transformation by each acting locally on her or his part.
a) Now suppose that Alice and Bob choose standard fixed bases
{|iiA } and {|iiB } for their respective Hilbert spaces, and are
presented with a vector |ψAB i that is not necessarily in the
Schmidt form when expressed in the standard bases. Show
that there is a local unitary U A ⊗ U B that Alice and Bob can
apply so that the resulting vector
|ψi0AB = U A ⊗ U B |ψiAB (2.186)
does have the form eq.(2.185) when expressed in the standard

bases.
b) Let’s verify that the result of (a) makes sense from the point
of view of parameter counting. For a generic vector in the
Schmidt form, all λi ’s are nonvanishing and no two λi ’s are
equal. Consider the orbit that is generated by letting arbitrary
local unitaries act on one fixed generic vector in the Schmidt
form. What is the dimension of the orbit, that is, how many
real parameters are needed to specify one particular vector
on the orbit? (Hint: To do the counting, consider the local
unitaries that differ infinitesimally from the identity I A ⊗ I B .
Choose a basis for these, and count the number of independent
linear combinations of the basis elements that annihilate the
2.8 Exercises 53
Schmidt-decomposed vector.) Compare the dimension of the

orbit to the (real) dimension of HAB , and check the consistency
with the number of free parameters in eq.(2.185).
A vector |ψiA1 ...Ar in a Hilbert space HA1 ⊗ · · · ⊗ HAr with r parts

is said to be Schmidt decomposable if it is possible to choose or-
thonormal bases for HA1 , . . . HAr such that vector can be expressed
as Xp
|ψiA1 ...Ar = λi |iiA1 ⊗ |iiA2 ⊗ · · · ⊗ |iiAr . (2.187)
i
Though every vector in a bipartite Hilbert space is Schmidt decom-
posable, this isn’t true for vectors in Hilbert spaces with three or
more parts.
c) Consider a generic Schmidt-decomposable vector in the tripartite
Hilbert space of three qubits. Find the dimension of the orbit
generated by local unitaries acting on this vector.
d) By considering the number of free parameters in the Schmidt
form eq.(2.187), and the result of (c), find the (real) dimension
of the space of Schmidt-decomposable vectors for three qubits.
What is the real codimension of this space in the three-qubit
Hilbert space C8 ?

Quantum States and Ensembles2

Uploaded by

Copyright:

Available Formats

Quantum States and Ensembles2

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Quantum States and Ensembles2

Uploaded by

Copyright:

Available Formats

Lecture Notes for Ph219/CS219:

Updated July 2015

2 Foundations I: States and Ensembles 3

2.1 Axioms of quantum mechanics

Axiom 1. States. A state is a complete description of a physical

What is a Hilbert space?

a) It is a vector space over the complex numbers C. Vectors will be

ii) Linearity: hϕ|(a|ψ1 i + b|ψ2 i) = ahϕ|ψ1 i + bhϕ|ψ2 i.

(Completeness is an important proviso in infinite-dimensional function

Axiom 2. Observables. An observable is a property of a physical

An operator is a linear map taking vectors to vectors,

A : |ψi 7→ A|ψi, A (a|ψi + b|ϕi) = aA|ψi + bA|ϕi. (2.2)

(We will often denote operators by boldface letters.) The adjoint A† of

hϕ|Aψi = hA† ϕ|ψi, (2.3)

only if A and B commute. Note that AB + BA and i(AB − BA) are

Here each an is an eigenvalue of A, and E n is the corresponding orthog-

The orthogonal projector onto the one-dimensional space spanned by the

where {|ni} is the orthonormal basis of eigenstates of A, with A|ni =

Axiom 3. Measurement. A measurement is a process in which in-

Prob(an ) = kE n |ψik2 = hψ|E n |ψi ; (2.7)

if the outcome an is attained, then the (normalized) quantum state

If the measurement is immediately repeated, then according to this rule

Axiom 4. Dynamics. Dynamics describes how a state evolves over

In the Schrödinger picture of dynamics, if the initial state at time t is

|ψ(t0 )i = U (t0 , t)|ψ(t)i|ψ(t)i, (2.10)

where U (t0 , t) is the unitary time evolution operator. Infinitesimal time

|ψ(t + dt)i = (I − iH(t)dt)|ψ(t)i. (2.12)

Thus the operator U (t + dt, t) ≡ I − iH(t)dt is unitary; because H is

Axiom 5. Composite Systems. If the Hilbert space of system A

What is a tensor product of Hilbert spaces? If {|iiA } denotes an or-

AB hi, µ|j, νiAB = δij δµν . (2.13)

2.2 The Qubit

The smallest nontrivial Hilbert space is two-dimensional. We may de-

a|0i + b|1i, (2.15)

is commutative, and therefore the time evolution operator eitH commutes

U (R)e−itH = e−itH U (R) ; (2.20)

expanding to linear order in t we obtain

U (R)H = HU (R). (2.21)

For a continuous symmetry, we can choose R infinitesimally close to the

U = I − iεQ + O(ε2 ), (2.22)

where Q is an operator determined by T . From the unitarity of U (to

the observable Q commutes with the Hamiltonian.

determined how finite transformations are represented, for these can be

U (R)J k U (R)† = Rk` J ` . (2.35)

Thus, if a state |mi is an eigenstate of J 3

J 3 |mi = m|mi, (2.36)

then U (R)|mi is an eigenstate of RJ 3 with the same eigenvalue:

RJ 3 (U (R)|mi) = U (R)J 3 U(R)† U (R)|mi

Therefore, we can construct eigenstates of angular momentum along the

and applying it to 10 , the J 3 eigenstate with eigenvalue 1, we obtain

2.2.2 Photon polarizations

We will not present here a detailed discussion of the unitary represen-

|xi → cos θ|xi + sin θ|yi

This two-dimensional representation is actually reducible. The matrix

has the eigenstates

with eigenvalues ±1. Because the eigenvalues are ±1 (not ± 21 ) we say

2.3 The density operator

1. States are not rays.

To arrive at the laws obeyed by open quantum systems, we must recall