You are on page 1of 29

1

March 4, 2019

Quantum Information I & II

H.-S. Sim

• This course is designed for undergraduates and graduates of KAIST Physics.


• Prerequisite: Quantum Mechanics I and II (undergraduate).

Topics
• Qubits and quantum coherence
• Quantum entanglement
• Measurement and operation
• Decoherence
• Information and entropy
• Quantum information devices
• Quantum processing and quantum computing
• Quantum entanglement in many-body systems

References
• Quantum Computation and Quantum Information by MA Nielsen and IL Chuang
• Quantum Mechanics: Fundamentals by K Gottfried and TM Yan

Assignment (spring): Homeworks : Midterm exam : Final exam = 55 : 30 : 15


Homeworks are either to solve problems or to summarize the background and main theme of some articles or
published research papers. The midterm exam will be a usual test solving problem sets. In the final exam,
each student will give a 5-min (approximately) oral presentation (including questions) for a given subject. The
subjects will be selected from the lecture note or the homeworks, and announced a few weeks before the exam.

Note 1: This lecture note is only for the students of my class. It should not be used for the others. This note
could include some figures copied from web without permission. I acknowledge the authors of those figures.

Note 2: Probably, this lecture will be given for two semesters.

No lecture (make-up lectures will be given): Mar. 18, 20, and one or two days in May
2

Preface

Quantum mechanics was invented more than 100 years ago. The invention is a revolution, as the essence of
quantum mechanics is hard to be accepted with the “old” spirit of classical mechanics. Quantum mechanics is in
good agreement with experimental data, and it has been a language that allows us to discover new principles and
phenomena of the Nature in the various research areas of high-energy physics, nuclear physics, AMO (atomic,
molecular, and optical physics), quantum optics, condensed matter physics, quantum chaos, and sometimes
biophysics. Nowadays it is not difficult to find the word “Quantum” in our social life.

Someone says that we are now going to the second revolution of quantum mechanics or the era of quantum
technology. Quantum devices, whose operation principles are quantum-mechanical superposition and entangle-
ment, will form the backbone of the quantum technology. They are estimated to show performance in a level
that “classical” devices cannot reach. Quantum computers will allow us to have computations impractical with
classical machine, quantum communications pursue secure communication immune to hacking, quantum elec-
tron circuits will be designed for new functionality and replace classical electric devices, and quantum mechanical
vibrators lead to a high-precision mechanical detector.
The quantum devices will be applied to solve big problems in science, artificial intelligence, medicine, opti-
mization, material design, etc. For example, they will be useful for physics research. There are many unsolved
problems in many-body quantum effects and non-equilibrium dynamics. To find the solution of such a problem,
a quantum device, called a quantum simulator, will be used. The concept of a quantum simulator was first
considered by R. Feynmann, who recognized that classical-computer-based numerical simulation of a general
many-body system becomes impractical as the size of the Hilbert space exponentially increases with the number
of particles in the system. It is a device with which one “experimentally solves” a target problem in a way
of preparing an initial state, operating its time evolution under the Hamiltonian of the problem, and finally
reading out the result of the evolution.

I hope that this lecture is helpful to understand the information aspect of quantum mechanics and to prepare
the era of quantum technology.

FIG. 1:
3

FIG. 2:

FIG. 3:
4

Contents

1 Quantum Coherence
I. Superposition, entanglement, and coherence 6
A. Density matrix 7
B. Quantum entanglement 13
C. Single qubit 18
D. Single-particle interference and coherence 22
5

Quantum Information
c H.-S. Sim

Part 1
Quantum Coherence
6

Quantum Information
c H.-S. Sim

I. SUPERPOSITION, ENTANGLEMENT, AND COHERENCE

Keywords: Superposition, Density matrix, Purity, Coherence, Interference, Mixed states, Decoherence

Quantum mechanical interference?


• In this chapter, we will focus on the wave nature of quantum mechanics, the mathematical background of
density operators and entanglement, and single-qubit operations.
• We will also see how some decoherence effects reduce interference signal.
7

A. Density matrix

A density matrix (or a density operator) is devised to describe a statistical mixed ensemble of quantum states
(von Neumann 1927). It is a mathematical tool for quantum information. Read Sakurai Chap 3.

Density operator
• Pure states (pure ensembles): All the members of the ensemble are described by a single ket.
• Mixed states: P A certain fraction w1 (weight) of the ensemble is described by |α1 i, another fraction w2
by |α2 i, · · · . i wi = 1. Here |αi i’s are normalized kets.
• Ensemble average [A] of an observable A: It is natural to write
X
[A] ≡ wi hαi |A|αi i.
i

• Introducing the density operator ρ,


X
ρ ≡ wi |αi ihαi |,
i

the ensemble average is written in a simple form,

[A] = Tr(ρA).
P P P P P
P is because Tr(ρA) = Tr( i wi |αi ihαi |A) =
It n i wi hn|αi ihαi |A|ni = n i wi hαi |A|nihn|αi i =
i wi hαi |A|αi i, where {|ni} is an orthonormal basis set. Notice that Tr(· · · ) is independent of the
choice of a basis set; the expectation value of a physical observable has to be independent of the choice.
• Choosing the basis set composed of the eigenstates |ai of A,
X X
[A] = Tr(ρA) = wi hαi |aiha|A|αi i = wi a|hαi |ai|2 .
i,a i,a

Properties of a density operator


• Normalization
X
Trρ = wi hαi |αi i = 1.
i

• Any physical density operator is Hermitian. ∵ (|αi ihαi |)† = |αi ihαi |.
• A 2 × 2 density matrix is parameterized in a basis set as
 
a c + id
ρ=˙ ,
c − id b

where a + b = 1 is imposed by the normalization condition. This density matrix has three independent
real parameters. It is useful for describing a two-level system (a qubit) such as a spin−1/2 particle.
8

Purity
• To see whether an ensemble ρ is pure or mixed, one computes the purity P ≡ Trρ2 .
– For pure ensembles,

0 0
 
 0 
P(ρ) = Trρ = 1 2
∵ ρ=

˙  1 
.

 0 

..
0 .

For any pure states, ρ is written as a diagonal matrix with only one wi = 1.
– By constrast, for all mixed states
1
≤ Trρ2 < 1.
N
Here N is the dimension of the Hilbert space of the system.
• Sometimes, one introduces the complement of purity, the mixedness M(ρ) = 1 − P(ρ).
• The purity of a state is invariant under unitary state transformations ρ → U ρU † . An example of the
unitary operators is the time evolution U (t, t0 ) = e−iH(t−t0 )/~ .
• It is nontrivial to measure the purity of a state in experiments. Basically, the purity is a nonlinear function
of a density matrix, hence it is not well connected to physical observables in general. There have been
experiments that claim measurement of the purity in a specific case. See Homework-1.

Examples: spin-1/2 particles


• Completely polarized particles in |Sz +i
   
1  1 0
ρ = |+ih+|=
˙ 1 0 = .
0 0 0

To represent this state, we choose the basis set of {|±i ≡ |Sz ±i. Notice Trρ2 = 1.
• Completely polarized particles in |Sx ±i
1
± 12
   
1 1 1
ρ = |Sx ±ihSx ± | = ( √ )2 (|+i ± |−i)(h+| ± h−|)=(
˙ √ )2

1 ±1 = 2 .
2 2 ±1 ± 12 1
2

Notice Trρ2 = 1.
• Unpolarized particles: An incoherent mixture of 50% in |Sz +i and 50 % in |Sz −i is described by
 1 
1 1 0
ρ = |+ih+| + |−ih−|= ˙ 2 .
2 2 0 21

~ = 0.
Notice that this state is mixed, Trρ2 < 1. And, [S]
• A mixture of 75% in |Sz +i and 25 % in |Sx +i:
1 1 7 1
     
3 1 0 1 2 2 8 8
ρ=
˙ + 1 1 = 1 1 .
4 0 0 4 2 2 8 8

Check [Sx ] = ~/8, [Sy ] = 0, and [Sz ] = 3~/8.


9
p p
• Consider a pure-state ensemble in ρ1 = |αihα|, |αi = 1/4|+i + 3/4|−i, and another mixed-state
ensemble in ρ2 = (1/4)|+ih+| + (3/4)|−ih−|, where 25 % of the particles are in |+i and 75 % in |−i. The
two ensembles are different. How are they different physically?

Quantum statistical mechanics & Maximally mixed (totally random) states


• von Neumann Entropy S:

S ≡ −kB Tr(ρ ln ρ).

This is a measure of randomness. kB is the Boltzmann constant.


• A pure ensemble is written as (in a basis set which includes the pure state as a basis state)

S=0 ∵ ln 1 = 0.

• A completely random ensemble:


– Its density matrix is (in any basis)
 
1 0
1  1  1
ρrandom =
˙  = I.
 
N
 ..  N
 .
0 1

Here, I is the N × N identity matrix.


– This density matrix does not evolve in time,
∂ρrandom 1
i~ = −[ρrandom , H] = [H, I] = 0.
∂t N
– It has the maximum entropy,
1
S = −kB Tr(ρrandom ln ρrandom ) = −kB N (− ln N ) = kB ln N.
N
Since ρrandom does not evolve in time, its entropy remains maximal.
– U † ρrandom U = ρrandom for any unitary basis transformation operator U .
• A canonical ensemble in thermal equilibrium: The conditions for the equilibrium are
∂ρ
= 0, δS = 0, δ[H] = 0, δ(Trρ) = 0.
∂t
– The first condition leads to [ρ, H] = 0, which means that ρ and H are simultaneously diagonalized.
Using the eigenstates |ki of the Hamiltonian H, ρ is written as a diagonal matrix with elements ρkk ,
 
..
 . 0 
ρ=˙ 
 ρ kk
.

..
0 .

– δS = 0 means
X X
S = −kB ρkk ln ρkk → δS = −kB (δρkk )[1 + ln ρkk ].
k k
10

– The third and fourth conditions are written as


X X
δρkk Ek = 0, δρkk = 0.
k k

– Collecting the conditions, one constructs an equation with Lagrange multipliers


X
δρkk [ln(ρkk ) + 1 + βEk + γ] = 0
k
exp(−βEk ) 1 X
⇒ ρkk = P = exp(−βEk ), Z= exp(−βEk ).
k exp(−βEk ) Z
k
1 −βH
⇒ ρ= e , Z = Tre−βH . (1)
Z
This describes the canonical ensemble. β = 1/(kB T ) and T is temperature. Notice that the thermal
density matrix ρ is an operator determined by the Hamiltonian H (or its excitation spectrum) and
the temperature.
– [A] = Z −1 Tr(e−βH A).
– U = [H] = Z −1 Tr(e−βH H) = −∂(ln Z)/∂β.
– At T = 0, ρ describes a pure ensemble, when there is no degeneracy in the ground state. At T → ∞,
ρ → ρrandom describes a random ensemble.
• Example: A spin-1/2 particle under H = µ ~ at temperature T . B
~ ·B ~ = B0 ẑ.

1 e−β~ω/2
 
0
ρ=
˙ , Z = e−β~ω/2 + eβ~ω/2 .
Z 0 eβ~ω/2
One finds [Sx ] = [Sy ] = 0 and [Sz ] = −(~/2)tanh(β~ω/2).
• Time evolution of the von Neumann entropy: For any mixed state of a system, the value of the von
Neumann entropy does not change under any unitary operation (defined in the Hilbert space) of the state.
The same answer is found for the time evolution operation under the Hamiltonian of the system.
– One remark is on the case that the system of intrest is a part of a larger system. In this case, one
can think about the Hamiltonian of the whole system, and compute the von Neumann entropy by
(i) obtaining the time evolution of the total state of the whole system, (ii) calculating the density
operator of the subsystem of our interest (by tracing out the degree of freedoms of the other part of
the whole system), and then (iii) computing the von Neumann entropy of the density operator of the
subsystem. In this case, the value of the von Neumann entropy increases in time in general.
– An interesting issue related to the above remark is the thermalization of an isolated quantum state
or the eigenstate thermalization hypothesis.

Quantum mechanical superposition versus classical mixture



• Let’s compare |ψ1 i = (|Sz , +i + |Sz , −i)/ 2 and ρ2 = (|Sz , +ihSz , +| + |Sz , −ihSz , −|)/2. The former is a
superposition of |Sz , +i and |Sz , −i, while the latter is a mixture of them.
• They have different density-matrix forms. Their matrix forms in the basis states {|Sz , ±i} are
   
1 1 1 1 1 0
ρ1 = |ψ1 ihψ1 |=
˙ , ρ2 =
˙ .
2 1 1 2 0 1
So, they are different states. How different?
• If two states are different, there are observables whose expectation values are different between the states.
For example, ρ1 and ρ2 show different expectation values of Sx .
• Notice that ρ1 has off-diagonal elements, while ρ2 does not. The physics behind is discussed below.
11

Pure-state decomposition
• We represent a mixed state in terms of projectors onto pure states,
X
ρ ≡ wi |αi ihαi |.
i

Here, the pure states |αi i’s are not necessarily orthogonal to each other. This is called pure-state decom-
position.
• The decomposition is not unique in general. In other words, it is possible to decompose a mixed state
into pure states in many different ways.
• Example: Consider a maximally mixed (totally random) state of a spin-1/2 particle,
 
1 1 0
ρrandom =˙ .
2 0 1

This density matrix is invariant under any unitary transformation for basis changes. It can be decomposed
as
1 1
ρrandom = (|Sz +ihSz + | + |Sz −ihSz − |) = (|Sx +ihSx + | + |Sx −ihSx − |)
2 2
1 1
= (|Sy +ihSy + | + |Sy −ihSy − |) = (|Sn̂ +ihSn̂ + | + |Sn̂ −ihSn̂ − |).
2 2
Here n̂ is a unit vector in the three-dimensional space.
• This means that interpretation of a mixed state, in terms of pure states |αi i and its occurrance probabilities
wi , is not unique, hence, it is nontrivial.
• The non-uniqueness generates notorious difficulty in theoretical and experimental sides.
– Experimentally, one wants to reconstruct the density matrix of a target mixed state by measuring
observables. Then, what is the most efficient way of the reconstruction with the smallest number of
parameters to be determined from the measurements?
– Theoreticaly, one wants to quantify the amount of quantum entanglement in a target mixed state.
Among the possible pure-state decompositions of the mixed state, a specific decomposition provides
a correct value of the quantification. But how do we efficiently find the specific decomposition?
These problems are NP-hard, when a target system has many particles or a large number of degrees of
freedom. These are obstacles in understanding entanglement in a many-body mixed state and multipartite
entanglement.

Note: Computational complexity


• The complexity is measured as lower bounds on time and space resources required to solve computational
problems by the best possible algorithm.
• The complexity can be classified: For a problem of “size” n (here n represents the number of bits, for
example), e.g., whether a particular n-bit number is prime or not,
– Tractable or feasible problems: A class of problems can be solved using resources bounded by a
polynomial in n. Ex) sorting problems, finding the shortest path.
– Hard or intractable problems: Another class requires resources that grow faster than any polynomial
in n, e.g., following an exponential function of n. Ex) searching passwords, prime factorization.
• Decision problems: Problems with a yes or no answer. Ex) “Is a a multiple of b?”
12

– Class P: Decision problems solvable in a polynomial time using a deterministic Turing machine.
– Class NP: Problems solvable in a polynomial time using a indeterministic Turing machine. In other
words, it is a set of the problems which is solvable within a polynomial time when a hint of the
solution is provided. P ⊂ NP. Ex) For a set of n integers, such as {−1, −2, −7, −14, 4, 5, 9, 15}, do
we have a subset whose elements satisfy that their sum is zero? The complexity of this problem is
2n − n − 1. But, if we have a hit to check a subset {−2, −7, 9}, one can find the answer within a
polynomial time. This problem is clearly an NP problem, but it is unclear whether this is in P.
– Class NP hard: Problems at least as hard as the hardest problems in NP. An NP-hard problem does
not have to be an element of NP.
– A big question is whether P = NP.
• The computational power of quantum computers are related to some major open problems in classical
computational complexity theory.

Note: Time evolution of ensembles


P
• Consider a state at time t0 , ρ(t0 ) = i wi |αi , t0 ihαi , t0 |. The Hamiltonian of the system is H. The time
evolution of each |αi , t0 i is |αi , t0 ; ti, i~(∂/∂t)|αi , t0 ; ti = H|αi , t0 ; ti.
• Then, the time evolution of the ensemble is determined by
∂ρ X ∂
i~ = i~ wi |αi , t0 ; tihαi , t0 ; t|
∂t i
∂t
X
= wi [H|αi , t0 ; tihαi , t0 ; t| − |αi , t0 ; tihαi , t0 ; t|H]
i
= −[ρ, H]. (2)

• The form of Eq. (2) is similar to that of the Heisenberg equation of motion. However the Heisenberg
equation of motion does not have the minus sign −.
13

FIG. 4: A composite system.

B. Quantum entanglement

A composite system
• Consider a composite system, composed of two subsystems with degrees of freedom (spin quantum numbers
or coordinates) q1 and q2 , respectively.
• Suppose the total system is in a pure state |Ψtot i. The wave function is Ψtot (q1 , q2 ) = hq1 , q2 |Ψtot i. The
matrix elements of the density matrix ρtot = |Ψtot ihΨtot | is

hq1 q2 |ρtot |q10 q20 i = Ψtot (q1 , q2 )Ψ∗tot (q10 , q20 ).

• For example, when |Ψtot i = |ψi1 ⊗ |φi2 (the subsystem1 of a total system is in |ψi1 and subsystem2 is in
|φi2 ),

ρtot = |Ψtot ihΨtot | = ρ1 ⊗ ρ2 , ρ1 = |ψi1 hψ|1 , ρ2 = |φi2 hφ|2 .

Reduced density matrix


• Consider an observable A1 of subsystem 1. Its expectation value (its ensemble average) is

[A1 ] = Tr(ρtot A1 ) = Tr(ρ1 A1 ).

Here, we introduce the concept of the reduced density matrix ρ1 of subsystem 1 as


X
ρ1 ≡ Tr2 ρtot = hq2 |ρtot |q2 i,
q2
X
hq1 |ρ1 |q10 i = hq1 q2 |ρtot |q10 q2 i.
q2

Tr2 is the trace only over the degree of freedom of subsystem 2. As a result, ρ1 is not a number but a
matrix or an operator.
• This can be easily shown as follows.

hq1 q2 |A1 |q10 q20 i = δ(q2 − q20 )hq1 |A1 |q10 i, ∵ A1 is independent of q2
⇒ [A1 ] = Tr(ρtot A1 )
X X X X
= hq1 q2 |ρtot |q10 q20 ihq10 q20 | A1 |q1 q2 i = hq1 q2 |ρtot |q10 q20 ihq10 |A1 |q1 iδ(q2 − q20 )
q1 q2 q10 q20 q1 q2 q10 q20
| {z }
=I
X
= hq1 |ρ1 |q10 ihq10 |A1 |q1 i
q1 q10

= Tr(ρ1 A1 ).
14

• Similarly,

ρ2 ≡ Tr1 ρtot , [A2 ] = Tr(ρ2 A2 ), A2 ∈ subsystem 2.


P P
• Tr1 (· · · ) = q1 hq1 | · · · |q1 i and Tr2 (· · · ) = q2 hq2 | · · · |q2 i are called partial trace.
• For the observables of subsystem 1, it is enough to study the state ρ1 of subsystem 1. Notice that ρ1 does
not have information of subsystem 2, while ρ has information of the whole system (subsystems 1 and 2).
The information of subsystem 2 in ρtot is lost, when we compute ρ1 from ρ by partially tracing out the
degrees of freedom of subsystem 2.

• Example: |ψi = |Sz , +i1 (|Sz , +i2 + |Sz , −i2 )/ 2 describes particle 1 in |Sz , +i and particle 2 in |Sx , +i.
 
h+ + |ρtot | + +i h+ + |ρtot | + −i h+ + |ρtot | − +i h+ + |ρtot | − −i
 .. .. 
 . . 
ρtot = |ψihψ| =  ..

 .. 
 . . 
h− − |ρtot | + +i h− − |ρtot | − −i
 1 1 
2 2 0 0    1 1 
 1 1 0 0 1 0
=  2 2  = ⊗ 21 21 .
0 0 0 0 0 0 1 2 2 2
0 0 0 0

1 1
   
1 0
ρ1 = Tr2 ρtot = |Sz +ihSz + | = , ρ2 = Tr1 ρtot = |Sx +ihSx + | = 2
1
2
1 .
0 0 2 2

In this example, we have the decomposition√of ρ = ρ1 ⊗ ρ2 since the state |ψi has the direct-product
structure of |Sz , +i1 ⊗ (|Sz , +i2 + |Sz , −i2 )/ 2.
• Example 2: Two noninteracting particles of ρtot = ρ1 ⊗ ρ2 :

Tr1 ρtot = ρ2 , Tr2 ρtot = ρ1 .



• Example 3: |Ψi = (|Sz +i1 |Sz −i2 − |Sz −i1 |Sz +i2 )/ 2. Compute ρ1 and think that the result is physically
reasonable.

Entanglement
• Separable states: A state of a composite system is separable if it is written as a single product of states
of the subsystems,

pure state |Ψtot i = |ϕ1 i|ϕ2 i or mixed state ρtot = ρ1 ⊗ ρ2 ,

where |ϕi i (or ρi ) is a state of subsystem i(= 1, 2).


• Entangled states: A state is an entangled state and has entanglement between the subsystems, if it
cannot be written as a single product of states of the subsystems.
• Examples:
1
|spin singleti = √ (| ↑i1 | ↓i2 − | ↓i1 | ↑i2 ) entangled,
2
| ↑i1 | ↓i2 − | ↑i1 | ↑i2 separable ∵ | ↑i1 (| ↓i2 − ↑i2 )
| ↑i1 | ↑i2 + | ↑i1 | ↓i2 + | ↓i1 | ↑i2 + | ↓i1 | ↓i2 separable ∵ (| ↑i1 + | ↓i1 )(| ↑i2 + | ↓i2 ).
15

• Theorem: When a whole system is in a pure ensemble of an entangled state, its subsystem is in a mixed
state.
– Below we will see an example, instead of a proof for general states.
– Consider a pure state Ψ(q1 , q2 ) written as

Ψ(q1 , q2 ) = c1 u1 (q1 )v1 (q2 ) + c2 u2 (q1 )v2 (q2 ),

where {ui (q1 )} and {vi (q2 )} are orthonormal wave functions of subsystem 1 and 2, respectively,
(namely hui |uj i = δij and hvi |vj i = δij ) and |c1 |2 + |c2 |2 = 1. ρtot = |ΨihΨ|.

– Ψ is entangled if c1 c2 6= 0; when |c1 | = |c2 | = 1/ 2, the state has the same form as the spin singlet.
– In the case of c1 c2 6= 0, ρ1 = Tr2 ρtot is a mixed state. This can be seen below by checking the purity
of ρ1 , Trρ21 = |c1 |4 + |c2 |4 < 1.
– Proof. ρtot = (c1 |u1 i|v1 i + c2 |u2 i|v2 i)(c∗1 hu1 |hv1 | + c∗2 hu2 |hv2 |). Then
X
ρ1 = Tr2 ρtot = (c1 |u1 ihq2 |v1 i + c2 |u2 ihq2 |v2 i)(c∗1 hu1 |hv1 |q2 i + c∗2 hu2 |hv2 |q2 i)
q2
X
= |c1 |2 |u1 ihu1 | + |c2 |2 |u2 ihu2 | + {c1 c∗2 |u1 ihu2 |hq2 |v1 ihv2 |q2 i + h.c.}
q2
| P
{z }
=0∵ q hv2 |q2 ihq2 |v1 i=hv2 |v1 i=0
2

= |c1 |2 |u1 ihu1 | + |c2 |2 |u2 ihu2 |.

ρ21 = |c1 |4 |u1 ihu1 | + |c2 |4 |u2 ihu2 | + |c1 |2 |c2 |2 (|u1 i hu1 |u2 ihu2 | + h.c)
| {z }
=0
= |c1 |4 |u1 ihu1 | + |c2 |4 |u2 ihu2 |.

⇒ Trρ21 = |c1 |4 + |c2 |4 < 1 ∵ |c1 |2 + |c2 |2 = 1, c1 c2 6= 0.

• This theorem is interpreted as follows. For separable pure states, the information of subsystem 1 is inde-
pendent of and uncorrelated from that of subsystem 2. Therefore, when we partially trace out subsystem
2 from the whole system, no information of subsystem 1 is lost and subsystem 1 remains in a pure state.
On the other hand, for pure states having entanglement between subsystems 1 and 2, the information of
subsystem 1 is correlated with that of subsystem 2. For example, in the case of spin singlet, information
correlation happens such that spin 1 is in |Sz +i1 when spin 2 is in |Sz −i2 , while spin 1 is in |Sz −i1 when
spin 2 is in |Sz +i2 . In this case, subsystem 1 becomes in a classical mixture of |Sz +i1 and |Sz −i1 , when
we trace out subsystem 2. Namely, loss of correlation between subsystems in computing a reduced density
matrix of a subsystem leads for the subsystem to be in a mixed state.
• A canonical ensemble of a system is in a mixed state at finite temperature, because information of en-
tanglement between the system and its thermal bath is absent in the thermal state (the reduced density
matrix) of the system.
• As shown above, one computes Trρ21 , ρ1 = Tr2 (|ΨihΨ|), to see whether a pure state |Ψi is entangled or
separable. The purity can be used as an entanglement measure that quantifies the amount of entanglement
in a state. See Homework-1.
– This method is not applicable to see whether a mixed state ρ(q1 q2 ; q10 q20 ) is entangled or separable.
– This question whether a mixed state is separable or entangled is in fact a notoriously difficult problem
(beloning to NP-hard). A mixed state is separable when all the possible pure-state decompositions
of the state are separable — To see whether a mixed state has entanglement, one typically needs to
explore all the possible pure-state decompositions, which requires huge computation resources.
16

Note: Tensor product


• The direct product is sometimes called direct product or Kronecker product.
• The direct product is bilinear (linear in V and also in W ).
• Consider a n-dimensional vector space V with basis {~e1 , ~e2 , · · · , ~en }, and another m-dimensional vector
space W with basis {f~1 , f~2 , · · · , f~m }.

• Then, the direct product V ⊗ W of the two spaces is (nm)-dimensional and has the basis set of {~ei ⊗ f~j }.
Pn Pm Pn Pm Pn Pm
~ = i wi f~i ∈ W . Then, ~v ⊗ w
• ~v = i vi~ei ∈ V , w ~ = ( i vi~ei ) ⊗ ( j wj f~j ) = i=1 j=1 vi wj (~ei ⊗ f~j ).
• For example, consider the following cases of n = 2 and m = 3.

1 0
   
  0   0
  1   0
1 0 1 1
   
~e1 ⊗ f~1 = ⊗  0  =  , ~e1 ⊗ f~3 = ⊗  0  =  ,
0 0 0 0
0 
0 1 0
0 0
  
w1 v1 w1
 
 v1  w2    v1 w2 
  
w3   v w

 1 3

~v ⊗ w
~ = = .
  
 w1

  v2 w1 
 v2  w2    v2 w2 
w3 vw w3

• A : V → V is an operator of the space V . Then, A ⊗ I3×3 is a (2 × 3)-by-(2 × 3) matrix in V ⊗ W ,

A11 0 0 A12 0 0
 

   
  0 A 11 0 0 A 12 0 
A11 A12 A11 I3×3 A12 I3×3  0 0 A11 0 0 A12

A⊗I = ⊗ I3×3 = = .

A21 A22 A21 I3×3 A22 I3×3  A21 0 0 A22 0 0 
 0 A 0 0 A22 0 
21
0 0 A21 0 0 A22

  
A11 I3×3 A12 I3×3 v1 w
~
(A ⊗ I)(~v ⊗ w)
~ = = (A~v ) ⊗ w.
~
A21 I3×3 A22 I3×3 v2 w
~

• Similarly, B : W → W is an operator of the space W . Then,


 
B11 B12 B13  
B 03×3
I ⊗ B = I2×2 ⊗  B21 B22 B23  = .
03×3 B
B31 B32 B33

• (I ⊗ B)(~v ⊗ w)
~ = ~v ⊗ (B w).
~
• (A1 ⊗ I)(A2 ⊗ I) = A1 A2 ⊗ I.
• (A ⊗ I)(I ⊗ B) = A ⊗ B.

 
A11 B A12 B
A⊗B = .
A21 B A22 B
17

• (A ⊗ B)(~v ⊗ w)
~ = (A~v ) ⊗ (B w).
~
• Det(A ⊗ B) = (DetA)m (DetB)n .
• Tr(A ⊗ B) = (TrA)(TrB).
• Example: |px , py , pz i = |px i ⊗ |py i ⊗ |pz i. px |px , py , pz i = px ⊗ I ⊗ Ipx = (px |px i) ⊗ |py i ⊗ |pz i.

• Example: Addition of angular momentum. J~ = L ~ +S ~ = L~ ⊗ I + I ⊗ L.


~ For l = 1 (3-dimensional)
and s = 1/2 (2-dimensional), J = 3 ⊗ 2 (direct product of {|ml i} ⊗ {|ms i}. J~ = 4 ⊕ 2 (direct sum of
~
{|j = 3/2, mj i} and {|j = 1/2, mj i}.

Note: Direct sum


• Consider a n-dimensional vector space V with basis {~e1 , ~e2 , · · · , ~en }, and another m-dimensional vector
space W with basis {f~1 , f~2 , · · · , f~m }.
• Then the direct sum V ⊕ W of the two vector spaces is (n + m)-dimensional and has the basis of
{~e1 , ~e2 , · · · , ~en ; f~1 , f~2 , · · · , f~m }.
• ~v ∈ V and A : V → V is an operator for vectors in V ,
 
v1  
 v2 
~v =  . , A =  n × n .
 
 .. 
vn

~ ∈ W and B : W → W is an operator for vectors in W ,


w
 
w1  
 w2 
~ =  . ,
w B =  m × m .
 
 .. 
wn

 
v1
 . 
 . 
   .   
~v  vn  A On×m
~v ⊕ w
~= =
 w1  ,
 A⊕B = .
w~   Om×n B
 .. 
 . 
wm

Here, On×m is the n × m matrix whose elements are zero.


    
A On×m ~v A~v
(A ⊕ B)~v ⊕ w
~= = = (A~v ) ⊕ (B w).
~
Om×n B w~ Bw~

• (A1 ⊕ B1 )(A2 ⊕ B2 ) = (A1 A2 ) ⊕ (B1 B2 ).


• det(A ⊕ B) = (detA)(detB).
• Tr(A ⊕ B) = (TrA) + (TrB).
18

FIG. 5: Bloch sphere

C. Single qubit

A qubit is a basic unit of quantum information processing. It is represented by a vector in a two-dimensional


complex Hilbert space.

Representation
• The two-dimensional Hilbert space is constituted by two orthonormal basis states {|0i, |1i} in correspon-
dence with two bit values 0 and 1. They are represented by state vectors
   
1 0
|0i=
˙ , |1i=
˙ .
0 1

• Bloch sphere: A pure qubit state is written as |ψi = c0 |0i + c1 |1i, where c0,1 ’s are c-numbers satisfying
|c0 |2 + |c1 |2 = 1. Typically, one chooses c0 = cos θ/2 and c1 = eiφ sin θ/2, 0 ≤ θ ≤ π and 0 ≤ φ < 2π,

cos θ2
 
θ iφ θ
|ψi = cos |0i + e sin |1i= ˙ . (3)
2 2 eiφ sin θ2

The state is visualized by using the surface of a unit sphere, called Bloch sphere. See Fig. 5. θ and φ
behave as the altitudinal-complement angle and the azimuthal angle, respectively.
– |ψ(θ = 0, φ = 0)i = |0i and |ψ(θ = π, φ = 0)i = |1i are on the north pole and on the south pole,
respectively; φ = 0 is conventionally chosen for θ = 0 and π.
√ √
– |ψ(θ = π/2, φ = 0)i = |+i = (|0i + |1i)/ 2 and |ψ(θ = π/2, φ = π)i = |−i = (|0i − |1i)/ 2 are on
the intersection of the x̂ axis and the equator.
√ √
– |ψ(θ = π/2, φ = π/2)i = |ii = (|0i + i|1i)/ 2 and |ψ(θ = π/2, φ = 3π/2)i = | − ii = (|0i − i|1i)/ 2
are on the intersection of the ŷ axis and the equator.
• Stokes parameters: Then, how do we visualize a mixed state in the Bloch P sphere? A general mixed
qubit state is written as a sum of projectors |αi ihαi | with probabilities pi , ρ = i pi |αi ihαi |. Considering
19

the fact that the density matrix is a 2 × 2 hermitian matrix of unit trace, one can represent a mixed state
using the Pauli matrices σi and the so-called Stokes vector Sµ ,
3  
1X 1 S0 + S3 S1 − iS2
ρ= Sµ σµ =
˙ ,
2 µ=0 2 S1 + iS2 S0 − S3
       
0 1 0 −i 1 0 1 0
σ1 = σx =
˙ , σ2 = σy =
˙ , σ3 = σz =
˙ , σ0 = I =
˙ .
1 0 i 0 0 −1 0 1

The representatation is consistent with the fact that a 2 × 2 hermitian matrix can be represented by four
real parameters. The Stoke parameters are expressed as

Sµ = Tr(ρσµ ).

One can prove this, using the identity σi σj = δij σ0 + iijk σk . Sµ ’s are real number.
– For a normalized state ρ, S0 = 1.
– A general (pure or mixed) normalized qubit state is visualized in the Bloch sphere by the Stokes
vector (S1 , S2 , S3 ).
∗ Indeed, the pure state in Eq. (3) is parameterized by (S1 , S2 , S3 ) = (sin θ cos φ, sin θ sin φ, cos θ).
It is on the surface of the Bloch sphere.
∗ For example, |0i is represented by (S1 , S2 , S3 ) = (0, 0, 1), |1i by (0, 0, −1), and |+i by (1, 0, 0).
∗ A mixed state is also represented by the vector (S1 , S2 , S3 ). It lies in the interior of the Bloch
sphere (not on the surface).
∗ The totally random state corresponds to (0, 0, 0).
p
– For a normalized mixed state ρ, the length of the Stokes vector is less than one, S12 + S22 + S32 < 1.
– The square of the length of the Stokes vector is related with the purity,

S12 + S22 + S32 = 2Trρ2 − 1.

Historically, the Stokes parameters were introduced to describe the polarization state of electromagnetic
radiation.

Single-qubit gate operations


• The logic operations of quantum information processing can be done with quantum gates, unitary oper-
ations acting on a target qubit state. The unitary operations are trace-preserving, hence, they do not
change the norm of the target state; S0 = 1 is preserved in the operation.
• Any single-qubit gate operation U can be performed by combining rotations and/or reflections. It is
decomposed into a product of
−iσ3 β −iσ2 γ −iσ3 δ
U = exp(iα) exp( ) exp( ) exp( )
 −iβ/2 2   2 2
cos γ2 − sin γ2 e−iδ/2 0
  
e 0
˙ eiα
=
0 eiβ/2 sin γ2 cos γ2 0 eiδ/2

where α, β, γ, and δ are real. This representation is a general expression of a 2 × 2 matrix belonging to
the U(2) group. See the chapter for the Euler rotation in Sakurai.
• Important gates are shown in Fig. 6.
20

FIG. 6: Symbols and operations of three important single-qubit gates. A copy of Fig.1.5 of Nielsen and Chuang.

• NOT gate: Among classical logic gates, the NOT gate is a nontrivial one. It acts on a classical bit
as 0 → 1 and 1 → 0. The corresponding quantum operation, the quantum NOT gate X, interchanges
between |0i and |1i,
     
0 1 α β
X=σ
˙ x= , X = .
1 0 β α

It is identical to σx . Note that there is the NOT gate,

 
1 1 i
NOT=
˙ √ .
2 i 1

One can achieve the NOT gate (up to an overall phase factor i), by applying the NOT gate twice.
• Hadamard gate: This is an important logic gate. It is represented by
 
1 1 1
H= ˙√ .
2 1 −1

The √Hadamard gate interchanges between |0i ↔ |+i = (|0i + |1i)/ 2 and between |1i ↔ |−i = (|0i −
|1i)/ 2, enabling qubit interference.On theBloch sphere, the Hadamard gate operation is decomposed
1 −1
into two rotations, a rotation (=˙ √12 ) about the ŷ axis by 90◦ and then another rotation (=
˙ −
1 1
 
0 1
i ) about the x̂ axis by 180◦ .
1 0

• Phase-flip gate: This gate does not alter |0i but it changes |1i as |1i → −|1i. It is identical to σz ,
 
1 0
Z=σ
˙ z= .
0 −1

When it acts on | ± ii ≡ (|0i ± i|1i)/ 2, it interchanges |ii and | − ii. It is a special phase rotation gate
P(φ̄) with the phase angle φ̄ = π.

• There are other useful single-qubit gates.


21

– Phase-rotation gate: It is represented by


   −iφ̄/2 
1 0 iφ̄/2 e 0
P(φ̄)=
˙ = e .
0 eiφ̄ 0 eiφ̄/2

It shifts the phase angle φ by φ̄ in the representation of Eq. (3).


– π/2-phase gate: When φ̄ = π/2, the phase-rotation gate is called the π/2-phase gate.
 
1 0
S=˙ .
0 i

When one applies this gate twice, one can get the phase-flip gate Z, S2 = Z.
– π/8 gate: When φ̄ = π/4, the phase-rotation gate is called the π/8 gate.
   −iπ/8 
1 0 iπ/8 e 0
T=˙ =e .
0 eiπ/4 0 eiπ/8

Notice T4 = S2 = Z. One may say that this T gate is better to be called “π/4-phase gate” rather
than “π/8 gate”. Yes, there must be some historical reason.

Note: Universal quantum gates


• A set of universal quantum gates is any set of gates for universal quantum computing. The requirement
is that any unitary operation can be decomposed into a finite sequence of gates from the set.
• Strictly speaking, it is impossible to decompose all unitary operations into a finite sequence of operations
belonging to a given finite set. The requirement means that any unitary operation can be efficiently (with
arbitrary accuracy) approximated by a sequence of gates from the set. This was proved by Solovay and
Kitaev (the Solovay-Kitaev theorem).
• To have a set for universal quantum computing, we need gates for multiple qubits. For example, a set
for two-qubit universal quantum gates is constituted by the Hadamard gate, the π/8 gate, the π/2-phase
gate, and the c-NOT (controlled NOT) two-qubit gate; for a single qubit, a universal set is composed by
the Hadamard gate, the π/8 gate, and the π/2-phase gate. See Nielsen and Chuang.
22

FIG. 7: Double-slit quantum interferomety.

D. Single-particle interference and coherence

Single-particle quantum interference is attributed to superposition, a fundamental principle of quantum me-


chanics. It is a start point towards quantum information. To observe quantum interference (and to perform
more complicated quantum information processing), we need to avoid decoherence and phase averaging.

Superposition and interference


• A particle is in a superposition of two orthogonal states, |ψi = c1 |ϕ1 i+c2 |ϕ2 i. We impose the normalization
|c1 |2 + |c2 |2 = 1. |ci |2 = |hϕi |ψi|2 is the probability of finding the particle in the state |ϕi=1,2 i.
• One asks about the probability of finding the particle at position x. This probability is P (x) = |hx|ψi|2 .
Notice that P (x) 6= |c1 hx|ϕ1 i|2 + |c2 hx|ϕ2 i|2 in general. We have an additional term, P (x) = |c1 hx|ϕ1 i|2 +
|c2 hx|ϕ2 i|2 +2Re[c1 hx|ϕ1 ic∗2 hx|ϕ2 i∗ ]. The last term 2Re[c1 hx|ϕ1 ic∗2 hx|ϕ2 i∗ ] shows the quantum interference
between the two states |ϕ1 i and |ϕ2 i at position x.
• See a corresponding situation of a double-slit interferometer in Fig. 7. A particle is injected from a source
to the double slit setup, and then detected at a screen. The probability of finding the particle on the screen
oscillates as a function of the coordinate x of the screen. This single-particle interference is a manifestation
of the wave-particle duality and quantum mechanical superposition.
• Try to formulate a theory that describes the position dependence of the probability. The following example
provides a hint.

Mach-Zhender interferometer
• The Mach-Zhender interferometer is a simple interferometer. See Fig. 8. Let’s see a quantum mechanical
description.
– Consider a particle. It would be either a boson or a fermion; the identical-particle exchange statistics
is not important because only a single particle is considered. The particle carries momentum k and
charge q.
– The interferometer is on a two-dimensional plane. Magnetic flux Φ is enclosed by the two arms of
the interferometer.
– The particle is injected from the source S1 . At the first 50-50 beam
√ splitter, the particle state becomes
a superposition of |ϕup i and |ϕdown i, |Ψi 7→ (|ϕup i + |ϕdown i)/ 2, where |ϕup i and |ϕdown i are states
of the upper and the lower arms, respectively.
∗ In this example, we consider a particular case. In general, we have |Ψi 7→ c1 |ϕup i + c2 |ϕdown i)
and |c1 |2 + |c2 |2 = 1.
– Then, it gains dynamical phase kLup along the upper arm of length Lup and kLdown along the lower
arm of length Ldown . It also gains the Aharonov-Bohm phase of 2πΦ/Φ0 , where Φ0 = h/q is the
flux quantum. Then, just before arriving√ at the second beam splitter, the particle state becomes
|Ψi → (eiφup |ϕup i + eiφdown |ϕdown i)/ 2. Here, φup − φdown = k(Lup − Ldown ) + 2πΦ/Φ0 .
23

FIG. 8: Mach-Zhender interferomety. Each of the two beam splitters has 50% transmission and 50% reflection probabili-
ties. Φ is the magnetic flux enclosed by the two arms of the interferometer. Lower panel: A sequence of qubit operations
equivalent with the Mach-Zhender interferometry.

– After passing through the second 50-50 beam splitter, the particle is detected at D1 . The detecting
probability is

eiφup + eiφdown 2 1
PD1 (k, L, Φ) = | | = [1 + cos(φup − φdown )]
2  2 
1 2πΦ
= [1 + cos k(Lup − Ldown ) + ]. (4)
2 Φ0

• One can observe an interference pattern as a function of k, ∆L = Lup − Ldown , or Φ.


• Notice that PD1 (k, L, Φ) is zero at certain values of k, L, and Φ. This means that the particle does not
come to the detector D1 . Where does it go?

• The interference processes can be simulated by using qubit operations.


 
1
– The initial state of the particleat the source S1 can be represented by |Ψi=˙ . Note that for
0
 
0
injection from S2 one considers .
1
 
1 1
– The 50-50 beam splitter is then represented by UBS = √12 . The state becomes
−1 1
 
1
|Ψi 7→ UBS , UBS = ZH.
0

The operation UBS is constructed by combining the Hadamard gate H and the phase-flip gate Z. For
1 √
example, the initial state |Ψi=
˙ becomes UBS |Ψi = (|0iMZ − |1iMZ )/ 2. Here |0iMZ means the
0
state propagating along the upper arm inside the interferometer, while |1iMZ along the lower arm.
24
 
eiφup 0
– The phase gain along the arms is described by Uphase = . The state becomes
0 eiφdown
 
1
|Ψi 7→ Uphase UBS .
0

The operation Uphase is described by the phase-rotation


  gate P(φ̄ = φdown − φup ) with rotation
1
angle φ̄ = φdown − φup . The initial state |Ψi=
˙ now becomes Uphase UBS |Ψi = (eiφup |0iMZ −
0

eiφdown |1iMZ )/ 2.
– After the particle passes through the second 50-50 beam splitter, the state becomes
 
1
|Ψi 7→ UBS Uphase UBS .
0
 
1
The qubit state of the particle after the passing the second beam splitter means the particle
0
 
0
state going to the detector D1 , while describes the particle state going to D2 .
1
– Finally, the particle is detected at D1 . Its probability is described by
 
1 2
PD1 = |Q↑ UBS Uphase UBS | .
0
 
1 0
Here Q↑ = | ↑ih↑ |=
˙ describes the projective measurement of the particle at D1 . Note that
0 0
 
0 0
Q↓ = | ↓ih↓ |=
˙ is for the projective measurement at D2 . Compare the result with Eq. (4).
0 1
The qubit operation of the whole process is drawn in Fig. 8.
– Notice that Q↑ is not unitary, while the other operations UBS Uphase UBS are unitary. The projective
measurement Q↑ belongs to a different sort of (non-unitary) operations. More discussion will be
found in the next chapter, Chap. ??.

Complementarity
• Suppose that a particle is a state |Ψi = cup |ϕup i+cdown |ϕdown i while it travels along the arms of the Mach-
Zehnder interferometry. In this case, the first beam splitter is not a 50-50 splitter so that |cup |2 6= 1/2 and
|cup |2 + |cdown |2 = 1. The second beam splitter is a 50-50 splitter.
• The path distinguishability D can be defined as

D = ||cup |2 − |cdown |2 |.

• The interference visibility V can be defined as

max(PD1 ) − min(PD1 )
V= ,
max(PD1 ) + min(PD1 )

where the max (min) is the maximum (minimum) value of the probability PD1 among the values obtained
with varying φup − φdown (in this case, with varying arg(cup c∗down )). The visibility is obtained as V =
2|cup cdown | in this case.
25

• There is an interesting relation, called a complementarity relation, between the path distinguishability
and the interference visibility,

D2 + V 2 = 1.

This is valid for any pure state |Ψi. Its physical meaning is that to have stronger interference signals, one
needs to prepare a more-path-indistinguishable quantum state.
• For general mixed states, the complementarity relation becomes the inequality of

D2 + V 2 ≤ 1. (5)

The equlity holds for pure states.


• To get which-path information of the particle, one adds an ancillary quantum system to the interferometer.
The complementarity relation can be generalized to this situation: The ability of gaining the which-path
information enhances the path distinguishability.
• A good reference for the complementarity is: G. Jaeger, A. Shimony, and L. Vaidman, Two interferometric
complementarities, Phys. Rev. A 51, 54 (1995).

Quantum coherence
• The word “coherence” has been used for describing waves. If a wave shows interference (e.g., when it passes
through Young’s double-slit setup), one could say that the wave is phase coherent. If not, incoherent.
• Quantum coherence is a concept describing the wave nature of quantum mechanical states. If a particle
shows interference (the wave-particle duality), one can say that the particle state is phase coherent.
• Consider √ two orthonormal states |0i and |1i, constituting a basis set. A particle is in |ψ(φ)i = (|0i +
eiφ |1i)/ 2, which is characterized by phase φ. Let’s observe the density matrix ρ = |ψihψ| represented
with the basis set {|0i, |1i},
 
1 1 eiφ
ρ(φ) = |ψ(φ)ihψ(φ)|=˙ .
2 e−iφ 1

– Suppose that one can measure an observable A with tuning the phase φ in experiments. When the
observable shows an interference pattern [A](φ) as a function of φ, ρ is phase coherent.
– For example, suppose that |0i and |1i describe the particle paths (upper and lower paths) of a two-
path interferometer. By changing the length of the lower path, one tunes φ = k(Lup − Ldown ), and
observes interference at a detector with varying φ.
• Next, consider a random ensemble composed of many |ψ(φ)i with different φ’s. Its density matrix is
Z  
1 1 1 0
ρrandom =
˙ dφρ(φ) = .
2π 2 0 1

This ensemble is phase incoherent. No phase information remains in ρrandom . This state does not show
any interference, and it is described simply by classical statistical probabilities.
– This situation happens when there occurs uncertainty in the length Lup or Ldown , for example, by
scattering of the particle by impurities. Such impurity scattering results in phase averaging of φ
hence suppression of the interference.
– It happens when there is uncertainty in the momentum k or the energy due to thermal fluctuations.
This leads to phase averaging and reduces the interference signal.
26

– It also happens when a degree of freedom of the particle couples with environments. For example,
the particle carries a charge which couples with enviromental noises generating random fluctuations
of electric or magnetic fields. If the phase induced by the noise rapidly fluctuates (many times while
the particle passes through the arms of the Mach-Zhender interferometer) between ±2π, the noise
masks the intrinsic phase φ = k(Lup − Ldown ). If we repeat the experiments under the noise, the
resulting ensemble is described by ρrandom . This is called decoherence or dephasing.
– Interactions between particles inside the interferometer can also generate decoherence in the single-
particle interference.
• Notice that the off-diagonal elements of ρ carry information of quantum coherence, the phase factor eiφ .
The off-diagonal elements of the random state ρrandom , a fully incoherent state, are zero indeed.
• Quantum coherence is an essential concept of quantum information processing. To have advantage over
a corresponding classical device, one has to avoid decoherence while operating a quantum information
device.

Aharonov-Bohm effects
• The Aharonov-Bohm effect is a quantum mechanical, topological, and nonlocal effect. In quantum in-
formation processing, it is a useful tool for observing quantum coherence of a charged particle (such as
an electron or a Cooper pair) and for engineering quantum states by chaning a boundary condition of a
target system (such as superconducting flux qubits).
• Quantum ring: A quantum ring is useful for understanding the Aharonov-Bohm effect.
– Consider an electron of mass Me and charge −e(< 0) in a one-dimensional quantum ring of radius
rq on xy plane shown in Fig. 9(a). The magnetic field B ẑ is applied to a disk area of radius r0 < rq
inside the ring so that the ring enclose the magnetic flux Φ = πBr02 . The coordinate along the ring
is represented by φ.
– Its Hamiltonian is
~2 pφ −e ~ 2 ~2 ∂φ Φ 2
H= ( − A) = ( + ) ,
2Me ~ ~ 2Me rq2 i Φe

~ r) = [Br2 /(2r)]φ̂ for r > r0 and A(~


where the vector potential is A(~ ~ r) = [Br/2]φ̂ for r < r0 . Φe = h/e
0
is the flux quantum, the unit flux for the electron. The electron spin is ignored in the Hamiltonian
for simplicity.
– The boundary condition of the eigenstates of H is ψ(φ+2π) = ψ(φ), which comes from the constraint
that the wave function is single-valued. Then, the wave function is

ψn ∝ eimφ , m = 0, ±1, ±2, · · ·


2
~ Φ 2
Hψn = (m + ) ψn .
2Me rq2 Φe

FIG. 9: Aharonov-Bohm effects. (a) Quantum ring enclosing magnetic flux Φ. (b) Two-slit interferometer.
27

The energy depends on the magnetic flux, although there is no magnetic field along the ring. This
is in sharp contrast with a classical particle whose energy is not affected by the magnetic flux. The
flux effect is nonlocal and purely quantum mechanical.
– Notice that (Φ, m) → (Φ + Φe , m − 1) preserves the energy. This can be observed in transport
measurement where the quantum ring couples to two electron reservoirs and electrical current is
measured from one to the other reservoir. The current will be a periodic function of Φ with period
Φe ; here we consider the case that electron interactions are sufficiently weak. The periodic dependence
can be understood from the energy spectrum of the ring as a function of Φ.
• Twist boundary condition: In quantum field theory, one usually considers an action. Let’s look at the
action for the quantum ring.
– The flux term contributes to the action of the electron as
2 Z
−e ~ · d~l = −e rq A
I I
A ~ = −2πrq 1 Br0
~ · dφ Φ
dφ = − ∆φ = θ
∆φ
, θ = −2π
Φ
.
~ ~ Φe 2rq Φe 2π Φe

– This term is not removed by canonical transformation. For example, the gauge transformation
of ψ → eiθφ/(2π) ψ leads to H → −~2 ∂φ2 /(2Me rq2 ), hence, one might consider that the effect of
the θ term becomes absent in the energy spectrum. However, the transformed wave function is
now affected by the twisted boundary condition of ψ(φ + 2π) = ψ(φ)eiθ . The wave function
satisfying the condition has the form of ei(n+θ/(2π))φ , which reproduces the energy spectrum of ~2 (n+
Φ/Φe )2 /(2Me rq2 ).
– Typically, one needs to impose the condition of ∆φ = 2πm, m = 0, ±1, ±2, · · · , when one computes
the partition function. This condition comes from the periodic boundary condition. Then, the action
has the flux term of
−e
Z
~ · d~l = −2π Φ = θm.
A
~ Φe
This term is now independent of “distance”, hence, of any continuous deformation of the coordinate
φ. Therefore, it is topological. Notice that it affects the quantization rule for electron energy.
– This term θm in the action is an example of a topological theta term. Another topological theta
term appears in actions for a quantum spin chain and for quantum Hall effects.

• Topological nature: The topological nature of the Aharonov-Bohm can be seen more easily, by using
the path integral.
– A propagator of a particle from position xi at time ti to position xf at later time tf is written as

K(xf , tf ; xi , ti ) = hxf |U (tf , ti )|xi iΘ(tf − ti ).

– The path integral form of the propagator is written as (see Sakurai for more details)
Z xf
i
K(xf , tf ; x0 , t0 ) = D(x(t)) exp[ S[x(t)]],
x0 ~
Z tf Z t00 Z t00 Z ~x00
m d~x 2 d~x ~ m d~x 2 ~
S[x(t)] = dtLclassical (x, ẋ, t) = dt ( ) + q ·A= dt ( ) + q d~s · A,
ti t0 2 dt dt t0 2 dt x0
~
m d~x 2 d~x ~
Lclassical = ( ) +q · A − V (x, t),
Z xf 2 dt dt Z ∞ Z Z
m
D(x(t)) = limN →∞ ( )N/2 dx1 dx2 · · · dxN −1 .
x0 2πi~∆t −∞
28
R ~x00
– Notice the term q ~ in the action. This term is what we discussed in the above about the
d~s · A
x0
~
R ~x00
topological theta term. This term also implies that a particle gains the phase factor exp[i(q/~) ~x0 d~s ·
~ due to a non-zero vector potential A,
A] ~ while it moves from ~x0 to ~x00 .
– Let’s apply this to the interferometry in Fig. 9(b). A particle has a wave function of |Ψi = c1 |ϕup i +
c2 |ϕdown i) just before it arrives at the detector, when the flux Φ is absent. In the presence of the
flux, the wave function gain additional phase factors,
~ ~
R R
s·A
|Ψi = c1 ei(q/~) upper arm
d~
|ϕup i + c2 ei(q/~) lower arm
s·A
d~
|ϕup i|ϕdown i).

The relative phase is


Z Z I
arg c∗1 c2 + (q/~)[ − ~ = arg c∗1 c2 + (q/~)
]d~s · A ~
d~s · A
lower arm upper arm
I
1 ~ ·∇×A
~
= arg c∗1 c2
+ 2π dS
Φ0
2πΦ
= arg c∗1 c2 + .
Φ0
This shows that the interference is a function of the magnetic flux Φ.
– The magnetic-flux contribution is topological. It is invariant under change of the arms of the inter-
ferometer; the change should not touch the region of nonzero magnetic fields.
• The Aharonov-Bohm effect is helpful for observing quantum interference, since it is topological, namely, it
is invariant under change of particle paths. Similarly, topological quantum computing will be considerably
less sensitive to local errors than the other quantum computing.

Note: Gauge transformation


• The gauge transformation in quantum mechanics is

A~ → A
~ + ∇Λ
~
iqΛ
|α, t0 ; ti → exp( )|α, t0 ; ti (6)
~
iqΛ
ψ(~x, t) → exp( )ψ(~x, t).
~
See Sakurai, for example.
29

Homework-1 (Due 11:59 pm, Mar. 24, 2019): The following problem set for Homework-1 is identical to a home-
work of my previous lecture on Quantum Mechanics for graduate students. For some students, it will be quite
boring to do the same homework. Or some students may want to consider a realistic situation rather than a
purely mathematical problem. So, you can choose either to solve the problems (the total score is 50 pt) or to
provide some discussions after reading one of research papers suggested below (the total score is also 50 pt). Of
course, you can do the both, but the score will be given only for one (no extra/cumulative point will be given)
— of course, you can get some feedback from the T.A. about your answer.

Choice 1 – Problem set: Read any quantum mechanics text book that discusses about quantum entanglement,
or some chapters in the book by Nielsen and Chuang.
• (A exercise problem in Sakurai) Consider a mixed ensemble of spin 1/2 systems. Suppose the ensemble
averages [Sx ], [Sy ], and [Sz ] of the spin operators are all known. Discuss how one may construct the 2 × 2
density matrix that characterizes the ensemble. (10 pt)
• Consider a composite system, system 1 ⊕ system 2. Show that the entropy S(ρ) = −kB Tr(ρ ln ρ) of the
whole system is smaller than the sum of the entropies of its reduced density matrices,

S(ρ) ≤ S(ρ1 ) + S(ρ2 ), ρ1 = Tr2 (ρ), ρ2 = Tr1 (ρ).

When do we have S(ρ) = S(ρ1 ) + S(ρ2 )? (10 pt)


– Hint 1. Prove first the “Klein’s inequality”: Define S(ρ k σ) ≡ Tr(ρ ln ρ) − Tr(ρ ln σ) for density
operators ρ and σ of the same dimension. Then, S(ρ k σ) ≥ 0. The equality holds iff ρ = σ.
– Next, put ρ = ρ and σ = ρ1 ⊗ ρ2 , to show the inequality.
• Consider a system of two particles, particle 1 and particle 2. Each particle has spin 1/2.

– Consider |Ψi = (|+i1 |+i2 +ei(θ1 +θ2 ) |−i1 |−i2 )/ 2, where Sz,i=1,2 |±ii = (±~/2)|±ii . Discuss whether
|Ψi has entanglement or not. Calculate the expectation values of spin Sx operators

hSx,1 i = Tr(Sx,1 |ΨihΨ|),


hSx,2 i = Tr(Sx,2 |ΨihΨ|),
hSx,1 ⊗ Sx,2 i = Tr(Sx,1 ⊗ Sx,2 |ΨihΨ|).

Do hSx,1 i, hSx,2 i, and hSx,1 ⊗ Sx,2 i − hSx,1 ihSx,2 i show interference as a function of θ1 or θ2 ? (20 pt)
– Repeat the above problem for |φi = (|+i1 + eiθ1 |−i1 )(|+i2 + eiθ2 |−i2 )/2. (10 pt)

Choice 2 – reading research papers: Motivating from a question arising in the class, let me suggest to
consider the followings about how to measure the purity of a mixed state in a specfic measurement setting.
• Read the paper by R. Islam et al., Measuring entanglement entropy in a quantum many-body system,
Nature 528, 77 (2015), and some references therein if necessary. And discuss the following points. Note
that you may need some knowledge about the second quantization formalism for bosons, to understand
the paper.
– What is the Hong-Ou-Mandel interference of two identical photons? (20 pt) Discuss the mechanism
of the interference (in any language such as the second quantization or with the first quantized wave
functions) and how to observe the interference.
– Based on Figure 2 of the paper, describe how to observe the purity in the experimental setting of the
paper, with providing some mathematical statement supporting your description. (20 pt)
– Summarize the experimental steps (in Figure 3) and the main results (in Figure 4) of measuring the
purity of a state in the paper. (10 pt) A brief summary is okey.

You might also like