Introduction to quantum computing and quantum information science Iustin Ouatu Pembroke College, University of Oxford, Oxford, United Kingdom I have spent approximately 2 months at the Theoretical Physics Department at the National Insti- tute for Physics and Nuclear Engineering, located in Bucharest, Romania, working with Dr. Radu Ionicioiu, Cambridge PhD (1999) in Quantum Gravity. I was introduced in the world of quantum computing and quantum information by being guided to study from textbooks, scientific articles, lecture courses. The studied material tends to be vast, ranging from ideas of Theoretical Computer Science, abstract mathematics (linear algebra, group theory), to Electrical Engineering and Quantum Mechanics. In the end, I have attended an workshop for Quantum Cryptography and Information held in Bucharest, given by Caltech Postdoc Alexandru Gheorghiu. In these 2 months I realized I want to pursue a PhD in the field of Quantum Information and Quantum Computing and I am truly satisfied that I was able to dig in these fields due to this Rokos STEM Foundation Funded Internship. I would like to say thank you to all those who facilitated me to be funded by this scheme. Above all, I would like to say thank you to the donor, Christopher Rokos, MA in Mathematics, Pembroke College ’89. I. INTRODUCTION, BACKGROUND AND MOTIVATION The world is currently witnessing the second quantum revolution. It all goes back to Heisenberg, Dirac, Schr ö dinger or Einstein, the founding fathers of quantum mechanics, who created the most well-tested scientific theory of all times. Back then, there was mystery and uncertainty about the correctness of such a revolution- ary theory. That was the first quantum revolution. Now, researchers possess the ability to use the precise laws of quantum mechanics to develop quantum technologies. It is all about creatively using superposition, entanglement, non-locality and randomness, the defining features of quantum mechanics, in order to come up with innovations which have a direct impact on economy and society as a whole. Quantum technologies will be integrated along with the existing classical technologies up to the point at which quantum supremacy will prevail. Many areas of the modern world will change. The classical idea of computation will be shifted towards a mix of both classical and quantum computation, a point in time at which quantum computers will coexist with classical computers. This will help to solve essential problems in nowadays world exponentially faster: optimization of a delivery chain or code breaking, for example. In medicine, new drugs will be created due to the superior power of quantum computers over classical ones. There is going to be a revolution in material science. Users will browse on a super-secure quantum internet and will send and receive messages which will be encrypted using quantum security keys. The entire area of sensing/imaging will benefit from the recent developments of quantum technologies: the quantum telescope, the quantum microscope and the quantum radar will be the state-of-the-art machines used in industry.
20
Embed
Introduction to quantum computing and quantum information ... · Introduction to quantum computing and quantum information science Iustin Ouatu Pembroke College, University of Oxford,
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Introduction to quantum computing and quantum information science
Iustin Ouatu
Pembroke College, University of Oxford, Oxford, United Kingdom
I have spent approximately 2 months at the Theoretical Physics Department at the National Insti-
tute for Physics and Nuclear Engineering, located in Bucharest, Romania, working with Dr. Radu
Ionicioiu, Cambridge PhD (1999) in Quantum Gravity. I was introduced in the world of quantum
computing and quantum information by being guided to study from textbooks, scientific articles,
lecture courses. The studied material tends to be vast, ranging from ideas of Theoretical Computer
Science, abstract mathematics (linear algebra, group theory), to Electrical Engineering and Quantum
Mechanics. In the end, I have attended an workshop for Quantum Cryptography and Information
held in Bucharest, given by Caltech Postdoc Alexandru Gheorghiu. In these 2 months I realized I
want to pursue a PhD in the field of Quantum Information and Quantum Computing and I am truly
satisfied that I was able to dig in these fields due to this Rokos STEM Foundation Funded Internship.
I would like to say thank you to all those who facilitated me to be funded by this scheme. Above
all, I would like to say thank you to the donor, Christopher Rokos, MA in Mathematics, Pembroke
College ’89.
I. INTRODUCTION, BACKGROUND AND MOTIVATION
The world is currently witnessing the second quantum revolution. It all goes back to Heisenberg, Dirac, Schr
ö dinger or Einstein, the founding fathers of quantum mechanics, who created the most well-tested scientific
theory of all times. Back then, there was mystery and uncertainty about the correctness of such a revolution-
ary theory. That was the first quantum revolution. Now, researchers possess the ability to use the precise
laws of quantum mechanics to develop quantum technologies. It is all about creatively using superposition,
entanglement, non-locality and randomness, the defining features of quantum mechanics, in order to come up
with innovations which have a direct impact on economy and society as a whole. Quantum technologies will
be integrated along with the existing classical technologies up to the point at which quantum supremacy will
prevail.
Many areas of the modern world will change. The classical idea of computation will be shifted towards
a mix of both classical and quantum computation, a point in time at which quantum computers will coexist
with classical computers. This will help to solve essential problems in nowadays world exponentially faster:
optimization of a delivery chain or code breaking, for example. In medicine, new drugs will be created due
to the superior power of quantum computers over classical ones. There is going to be a revolution in material
science. Users will browse on a super-secure quantum internet and will send and receive messages which
will be encrypted using quantum security keys. The entire area of sensing/imaging will benefit from the recent
developments of quantum technologies: the quantum telescope, the quantum microscope and the quantum radar
will be the state-of-the-art machines used in industry.
2
A lot of interest in quantum technologies can be seen across the world. Australia, Canada, China, EU, Japan,
Singapore, US, they all have very strong quantum research programmes and invest huge amounts of money into
them in order to win the so-called ’quantum race for supremacy’. EU has understood that a developed society
needs to invest in quantum technologies in order to be able to compete at the highest possible level, so it has
launched a 1 billion euros quantum research programme: Quantum Flagship. It is also the case for China, who
have achieved a milestone when they launched the first quantum satellite : Micius - this helped China to become
the first nation to quantum mechanically encrypt a video conference, leading the way to a world of quantum
mechanical communication. They currently invest 10 billion dollars in a big Quantum Research Facility. In
industry, many major companies invest money to become the first to have the first working universal quantum
computer: IBM [1], Google [2], Microsoft [3], Intel [4], Toshiba [5]. ESA [6], NASA [7] are also interested in
quantum technologies.
II. QUANTUM MECHANICS FORMALISM
In this section, a short introduction to the formalism of quantum mechanics is given. The theory of quantum
mechanics is based on postulates.
The first postulate describes the arena in which a quantum system lives and how its current state is de-
scribed. Each quantum system has an associated wavefunction |ψ〉, encapsulating all the information one can
extract about the current state of that system. It is a vector living in a Hilbert space H , that is, a complex vector
space equipped with an inner product. All the vectors are normed to one :
∀ |ψ〉 ∈ H ||ψ|| = 〈ψ|ψ〉 = 1 (1)
The second postulate describes the arena in which composite quantum systems live. Every wavefunction
associated with a composite quantum system lives in a Hilbert space of higher dimensionality than its compo-
nents: H12 = H1 ⊗H2 and dimH12 = dimH1 + dimH2.
The third postulate describes the dynamics of a quantum system: what is the state of a quantum system at
a future moment in time? It evolves according to a unitary matrix U:
t0 −→ t1 (2)
|ψinitial〉 −→ |ψfinal〉 = U |ψinitial〉 (3)
such that UU † = U †U = 1
Finally, the fourth postulate presents how a system responds when a measurement is made on it.
A quantum measurement is described by a set of measurement operators Mm which act on the state space
of the system being measured. If |ψ〉 is measured, the probability that result m occurs is given by :
p(m) = 〈ψ|M †mMm |ψ〉 (4)
3
The state of the system after the measurement is:
Mm |ψ〉√〈ψ|M †mMm |ψ〉
(5)
The measurement operators in the set Mm satisfy the completeness relation (equivalently, the probabilities sum
up to 1): ∑m
M †mMm = 1 (6)
Measurements split up in general measurements, projective measurements and POVM measure-
ments(positive operator-valued measure).
When it comes to projective measurements, one usually only specifies a complete set of orthogonal projec-
tors Pm satisfying∑
m Pm = 1 and PmPm′ = δmm′Pm. This is sufficient to describe measurements which
are repeatable. This formalism is usually studied in undergraduate level quantum mechanics.
However, when one uses a silvered screen to measure the position of a photon, the photon is absorbed and
it is impossible to repeat the measurement again. Thus, a more general approach is needed and the elegant
formalism of general measurements is used.
POVM measurements can be understood as a special case of the general measurements, providing the easiest
way to study the statistics of the measurement when the state of the system after the experiment is not important
to the physicist.
It is known that non-orthogonal states cannot be distinguished (see [8]). A simple application of POVM
might be the ability to distinguish with certainty between non-orthogonal states some of the time, while the
other part of the time gaining no information at all about the states. To specify a POVM, one has to list a set of
positive valued operators Em such that∑
mEm = 1. The probability to obtain result m is then given by :
p(m) = 〈ψ|Em |ψ〉.
The measurement problem is of fundamental importance in quantum mechanics and it is a the core of
quantum information and quantum computation.
Using these postulates, together with electrical engineering, mathematics and computer science, researchers
have built many applications which are based on the beauty of quantum mechanics.
III. QUANTUM COMPUTING
A. Preliminaries
Whenever one thinks of an algorithm, that sequence of operations can be related to how a Turing Machine
works.
Alan Turing (OBE FRS) was a British Computer Scientist, most well known for formalizing the idea of
an algorithm and computation as a whole using a theoretical model called the Turing Machine. The Turing
Machine consists of four elements: a program, a finite state control, a tape, a read-write tape-head.
4
The finite state control is a collection of internal states : q1, q2, q3, ..., qm with m being a fixed constant.
Two important additional states are the halting state: qh and the starting state: qs.
The tape is a vector (1-dimensional object) which extends to infinity, with square-blocks called tape squares.
In each box of the vector, an element from an alphabet (a finite collection of distinct symbols) is written.
The read-write tape-head is a pointer to the current state of the Turing Machine.
The idea is that the execution of the algorithm causes the read-write tape-head to move through the tape,
at the same time changing the current state of the finite state control. The finite state control starts in the state
qs and the tape-head is initially located at the leftmost square of the tape. The program is an ordered list of
program lines and, in broad terms, it causes the tape-head to move. The algorithm terminates when the current
state is the halt-state qh and the output is then the entry at which the tape-head points.
The Turing Machine model of computation completely describes every operation performed on a computer.
Church and Turing independently put forward the so called Church-Turing thesis, an essential statement in
computer science (see [9]):
’The class of functions computable by a Turing machine corresponds exactly to the class of functions which
we would naturally regard as being computable by an algorithm.’
This is an idealized model and while it offers huge insights into theoretical computer science, there is another
way of thinking about computers.
An alternative model which describes the idea of computation is the Circuit model. It is an approach more
feasible for understanding quantum computers as they are physical things and not theoretical, infinite in size,
machines. Using bits, wires and logic gates, every operation on a modern computer can be expressed using a
circuit.
Not all of the well known classical gates can be implemented on a quantum computer. A quantum system
evolves unitarily (see 2), so non-invertible gates are not suitable for quantum computation. Also, the FANOUT
gate violates the no-cloning theorem There are restrictions on the circuits so that one does not get unreasonable
results.
The link between the Turing model of computation and the circuit model of computation is made via circuit
families. A circuit family is uniform if there is an algorithm running on a Turing Machine which generates a
description of that family. In short, it shows an electrical engineer how to connect the wires and logic gates to
perform the algorithm.
It is known and has been shown in computer science that functions computable on uniform circuit families
are those functions computable on Turing Machines.
Quantum computers provide speed-up over classical computers on certain tasks. A method to quantize the
resources used by a computer to perform a task is needed.
Asymptotic notation or essential behaviour of a function shows how that function behaves when relating to
infinity. For example : 10n+ 20ln(n) + constant goes like n an is thus polynomial in n.
Formally, f(n) is O(g(n)) if ∃(n0, c) for which for n > n0 it is the case that f(n) ≤ cg(n).
5
Formally, f(n) is Ω(g(n)) if ∃(n0, c) for which for n > n0 it is the case that f(n) ≥ cg(n).
Formally, f(n) is Θ(g(n)) iff f(n) is both O(g(n)) and Ω(g(n)).
For example, the function 2n is Ω(n3) because for sufficiently large n, n3 ≤ 2n.
In short, if a computer uses resources bounded by a polynomial in n then the problem it solves is said to
be easy, tractable, feasible. If a computer uses resources which grow faster than any polynomial in n, then the
problem it tries to solve is said to be hard, intractable, infeasible.
The above notations lead to complexity classes. Two very important complexity classes are P and NP. P is
a subset of NP. The problems in P are solvable in polynomial times.
The problems in NP are hard for a computer. Given an yes instance of the problem, the computer can easily
verify the validity of it. On contrary, the other direction cannot be efficiently computed. An example is prime
factorization of a number. This requires exponential time to be solved on a computer and most cryptographic
protocols are based on the inability of a modern computer to factorize huge (512 bits) numbers. It is easy to
perform multiplication to see if the given decomposition leads to the required number. However, it is hard to
factorize that number on our own. Quantum computers aim at solving problems from NP. It is not known up
until know which type of problems quantum computers will generally solve. It theoretical computer science it
is agreed that the class of problems solvable by a quantum computer in polynomial time is called BQP. Still,
one cannot exactly relate this complexity class to NP problems.
I found it fascinating to study about theoretical computer science and see how much thinking is behind our
modern-day applications.
B. Quantum gates
Quantum gates are the analogous to the classical gates operating on bits. However, they act on qubits,
the fundamental piece of information in quantum information science. A qubit is a vector with norm 1 living
in a two dimensional complex vector space equipped with an inner product, i.e. a Hilbert space. It can be
represented by any two level system (levels are labelled by |0〉 and |1〉). While two complex numbers need
four real parameters to be fully specified, only two real parameters are needed to fully specify a qubit. A
|ψ〉 = α |0〉+ β |1〉 is constrained by 〈ψ|ψ〉 = 1 and the fact that the overall phase does not matter in quantum
mechanics. That is, the measurement statistics an experimentalist obtains are the same irrespective to the
overall phase. This can be understood by the fact that one measures probabilities (squared absolute values)
and |eiφ|2 = 1. Two degrees of freedom are eliminated and a qubit is represented using two real parameters.
A visualization of a qubit which gives valuable insights how it evolves is given by its representation on the
Bloch’s sphere.
A point on the Bloch’s sphere can be parametrized using two angles (real values):
|ψ〉 = cos (θ
2) |0〉+ eiφ sin (
θ
2) |1〉 (7)
6
FIG. 1: Bloch’s sphere. Any point on the sphere can be parametrized using two real values, θ and φ
Any operation on a qubit must preserve the norm and thus one can only rotate the qubit on the sphere. Only
allowed operations are unitary (2x2 unitary matrices in this case).
Some operations are:
X =
0 1
1 0
Y =
0 −i
i 0
Z =
1 0
0 −1
(8)
H =1√2
1 1
1 −1
S =
1 0
0 i
T =
1 0
0 exp(iπ/4)
Pφ =
1 0
0 eiφ
These are the building blocks of a quantum circuit. Some of these gates form universal sets. They are
universal for quantum computation, that is, any circuit can be simulated to sufficient precision using a finite
number of gates from an universal set. As proved in [10], the set H,Pφ, CNOT is universal for quantum
computation (for the action of CNOT gate, refer to III B 1).
It is important to distinguish rotations about the three axes x, y, z:
Rx(θ) = e−iθX/2 =
cosθ
2−i sin
θ
2
−i sinθ
2cos
θ
2
(9)
Ry(θ) = e−iθY/2 =
cosθ
2− sin
θ
2
sinθ
2cos
θ
2
(10)
7
Rz(θ) = e−iθZ/2 =
e−iθ/2 0
0 eiθ/2
(11)
Using these three rotations, then there exists four real numbers α, β, γ, δ which completely describe any
unitary operation U on a single qubit:
U = eiαRz(β)Ry(γ)Rz(δ) (12)
This leads to another form of the decomposition of any unitary operation on a single qubit:
U = eiαAXBXC (13)
with ABC = 1. This will be important when considering controlled operations (III B 1). These matrices act
on one qubit. For example, the X gate is the flip gate:
X
1
0
=
0
1
(14)
and
X
0
1
=
1
0
(15)
There exist higher dimensionality gates as well, acting on multiple qubits.
1. Controlled operations
A controlled operation is a gate for which its action depends on the value of some control bits. Controlled
operations involve two or more qubits. It is important to distinguish between target and control qubits. The
gate acts on the target qubits, conditioned on the values of the control qubits. The most trivial example is the
CNOT gate. It is depicted in figure 2a. It flips the value of the target qubit iff the value of the control qubit is
set to |1〉. An important application involving CNOT gates is the SWAP gate. It swaps the values of the two
qubits it acts on:
A controlled operation U conditioned on n control qubits is written as : Cn(U). It can be shown ([10]) that
for unitary V with V 2 = U then a C2(U) can be decomposed into:
An important case of C2(U) is the Toffoli gate : C2(X). For this, V =(1− i)(I + iX)
2(see 4).
Putting all this together, the Fredkin swaps the two target qubits iff the control qubit is set to |1〉. Its matrix
8
(a) CNOT-gate (b) SWAP-gate.
FIG. 2: Quantum Gates
FIG. 3: C2(U)-gate, with V 2 = U
representation is:
Fredkin =
1 0 0 0 0 0 0 0
0 1 0 0 0 0 0 0
0 0 1 0 0 0 0 0
0 0 0 1 0 0 0 0
0 0 0 0 1 0 0 0
0 0 0 0 0 0 1 0
0 0 0 0 0 1 0 0
0 0 0 0 0 0 0 1
(16)
and can be pictured as :
It is very well understood the fact that complex n qubit controlled gates can be decomposed into simpler,
elementary operations. For example, for ∀ unitary U2×2, a C2(U) can be simulated with only 14 elementary
operations [10].
It is interesting that a discrete set of gates cannot be used to implement an arbitrary unitary operation exactly,
but only to approximate it.
However, as errors add up linearly at most when in a sequence of gates, the Solovay-Kitaev theorem (as
stated in [11]) gives a powerful result:
”Let G be an instruction set for SU(d), and let a desired accuracy ε > 0 be given. There is a constant c > 0
such that for any U ∈ SU(d) there exists a finite sequence S of gates from G of length O(logc(1
ε
)and such
that d(U, S) < ε.”
9
FIG. 4: C2(U)-gate
A direct application of this is that for a circuit of m CNOT and single-qubit unitary operations a set of
O(m logc(m
ε)) is required to approximate it to accuracy ε.
Single-qubit gates and CNOT’s can implement any two-level unitary operation. Two-level unitary gates are
universal for quantum computation. This is because ∀ d × d unitary matrix U can be written as:
U = V1....Vk (17)
with k 6d(d− 1)
2and Vi being two-level unitary matrices
For a complete, formal description of the behaviour of controlled operations, one can consult [10].
Manipulation of quantum gates is easier when working with their block diagonal forms:
CNOT = diag (I,X) =
I 0
O X
(18)
C(Z) = diag (I, Z) (19)
CNOT = (I ⊗H)C(Z) (I ⊗H)
CNOT = diag (H,H)C(Z)diag (H,H)
CNOT = diag (H,H)diag (I, Z)diag (H,H)
CNOT = diag (HIH,HZH)
CNOT = diag (I,X)
Equipped with all this information, the main algorithms of quantum computing can be understood.
10
C. Quantum Algorithms
There are three classes of quantum algorithms which provide an advantage over classical computation.
The first class is based on Quantum Fourier Transform Algorithm (will be discussed in III C 1). This
algorithm does not provide speed-up when transforming classical data, but can be creatively used and it is the
building block of algorithms which solve problems untracktable (exponentially hard) on normal computers,
thus providing a massive speed-up: Shor’s Algorithm and Discrete Logarithm Problem are example of such
instances.
The second class is based on Quantum Search Algorithm (will be discussed in III C 2). The main idea is
that it implies an oracle, a somehow mysterious concept of a black-box capable of anything in order to help
solving a particular problem. Researchers have the difficult task to come up with a practical implementation
of this black-box and up until now real applications have been developed (i.e. it is possible to think of such
black-boxes).
The third class is based on Quantum Simulation Algorithms (will not be discussed in this report). Here a
quantum computer simulates a quantum system. A recent breakthrough of this kind is the success of the D-
WAVE probabilistic quantum computer to efficiently simulate how a complex quantum system behaves when
placed in a variable magnetic field, giving rise to interesting phase transitions between paramagnetic, anti-
ferromagnetic and spin-glass phases. (for scientifically presented results, see [12]).
1. Quantum Fourier Transform
It is well known that for an input set x0, x1, ..., xN−1, the classical Fourier transform produces an output set
y0, y1, ..., yN−1, with
yk =1√N
N−1∑j=0
xje2πijkN (20)
Quantum mechanically, an input set of states |0〉 , |1〉 , ..., |N − 1〉 is transformed according to:
|j〉 → 1√N
N−1∑k=0
e2πijkN |k〉 (21)
Alternatively,
N−1∑j=0
xj |j〉 →N−1∑k=0
yk |k〉 (22)
This transformation is unitary : 〈j|U †U |j〉 = 1.
The Quantum Fourier Transform (QFT) for a state |000...0〉 with n qubits starting in state |0〉 is then |+〉⊗n,
where |+〉 =|0〉+ |1〉√
2.
The following notation is defined:
0.jljl+1...jm → jl2
+jl+1
22+jl+2
23+ ...+
jm2m−l+1
(23)
11
It can be then shown that the QFT of |j〉 = |j1j2j3...jn〉 is:(|0〉+ e2πi0.jn|1〉√
2
)(|0〉+ e2πi0.jn−1jn|1〉
√2
)...
(|0〉+ e2πi0.j1j2...jn|1〉√
2
)(24)
To implement the QFT in practice, a gate Rk is defined:
Rk =
1 0
0 exp(2πi
2k)
(25)
An efficient algorithm to perform QFT is then:
FIG. 5: QFT
The image shows the need for n + (n − 1) + (n − 2) + ... + 1 =n(n− 1)
2. In addition, one has to add
restoring gates for the qubits, to return them in the original order in which they entered the quantum circuit.
This is done usingn
2SWAP gates, for a total of
3n
2CNOT gates. In total, Θ(n2) gates are used for QFT.
The classical version of Fast Fourier Transform uses Θ(n2n) gates. In conclusion, the QFT is more efficient,
providing polynomial speed!
Apart from this, the QFT is of fundamental importance in different applications which lead to the famous
Shor’s algorithm: phase estimation, order finding (uses phase estimation) and prime factorization (uses
order finding).
These algorithms will not be discussed here, but, at the time of writing, the author is familiar with them and
understands their behaviour.
2. Grover Search
The second class of quantum algorithms which provide speed-up over classical algorithms is the search
algorithms. This offers an alternative approach to searching problem. Supposing a list of N objects indexed
by x ranging from 0 to N − 1, one can define a function f(x) such that f(x) = 0 if x is not a solution to the
problem, while f(x) = 1 if x is a solution to the problem.
The main idea of the search algorithms is the usage of an oracle. This helps identifying the solutions of the
problem. It is a black-box which performs the following unitary operation: |x〉 |q〉 → |x〉 |q ⊕ f(x)〉 To check
if x is a solution, the oracle is applied to the initial state |x〉 |0〉.
12
Sometimes is more convenient to apply the oracle as it follows:
|x〉(|0〉 − |1〉√
2
)→ (−1)f(x) |x〉
(|0〉 − |1〉√
2
)(26)
which can be restated without writing the |−〉 state:
|x〉 → (−1)f(x) |x〉 (27)
The Grover subroutine is summarized below:
1) Start the system as |0〉⊗n
2) Apply Hadamard transform on all qubits : H⊗n
3)
|x〉 → −(−1)δx0 |x〉 (28)
4) Apply Hadamard transform on all qubits: H⊗n
It is schematically depicted in 6.
FIG. 6: Grover subroutine
Steps 2,3,4 have the cumulated effect of:
H⊗n(2 |0〉 〈0| − 1)H⊗n = 2 |ψ〉 〈ψ| − 1 (29)
where
|ψ〉 =1√N
N−1∑x=0
|x〉 (30)
Thus, the Grover subroutine G is:
G = (2 |ψ〉 〈ψ| − 1)O (31)
With this in mind, the search algorithm circuit is as below:
As an addition, the search algorithm can be beautifully understood using a geometric argument.
13
FIG. 7: Search algorithm circuit
Defining two sums, one iterating through the solutions of the problem and the other iterating through the
values which are not solutions of the problem (′
represents the solutions) :
α =1√
N −M
′′∑x
|x〉 (32)
β =1√M
′∑x
|x〉 (33)
the initial state |ψ〉 can be written as:
|ψ〉 =
√N −MN
|α〉+
√M
N|β〉 (34)
Applying the oracle O is equivalent to a reflection about |α〉:
O(a |α〉+ b |β〉) = a |α〉 − b |β〉 (35)
Applying 2 |ψ〉 〈ψ| − 1 represents a rotation about |ψ〉. Two reflections⇐⇒rotation.
The whole idea is that repeated applications of G rotate the initial state closer and closer to |β〉. After k
iterations, the state is:
Gk |ψ〉 = cos2k + 1
2θ |α〉+ sin
2k + 1
2θ |β〉 (36)
where
cosθ
2=
√N −MN
(37)
This is helpful because when, at the end, measuring in the computational basis, one obtains a state which form
|β〉 (i.e. a solution of the problem) with high probability (whp).
To rotate |ψ〉 to withinθ
2≤ π
4of |β〉, one can show that O(
√N
M) Grover iterations are needed. This is a
quadratic speed-up over the classical O(N
M).
The quantum search algorithm is unique and interesting because even without knowing the solution to the
problem, the oracle somehow recognises a solution when it sees one. The tricky part is to come up with practical
implementations of the oracle.
14
IV. QUANTUM INFORMATION
A very short introduction to the complex, abstract field of quantum information science is given in this
chapter. The writer understands some of the key concepts in quantum information, how to mathematically
model the environment and to use the rules of quantum mechanics to create some practical applications.
Quantum information science takes into account the effects of noise, that is the interaction of an open system
with the environment. In practice, systems are considered to be open, as no real application is perfectly closed
and non-interacting with basically anything.
A. How to visualize quantum noise
The mathematical formalism which governs the understanding of noise in open quantum systems is called
quantum operations. A quantum operation is a map which takes as input a quantum state ρ and returns another,
transformed quantum state ρ′:
E(ρ) = ρ′ (38)
There are three equivalent ways to understand how to deal with noise appearing in open quantum systems.
The first idea is to visualize a quantum operation as a CP-map: The open quantum system is embodied in a
bath, the so-called environment. A unitary operation U is applied on the whole construction (environment plus
the system as a whole). To get the transformed state of the quantum system, one simply discards the bath by
tracing over the environment.
A figure which show this is: 8 ([13]): Mathematically:
FIG. 8: A CP-map. Picture taken from Master Thesis Presentation: Weak Values in Quantum Measurements
written by Yutaka Shikano, Dept. of Physics, Tokyo Institute of Technology
E(ρ) = trenv[U(ρ⊗ ρenv)U †
](39)
15
Let|ek〉
be a basis for the environment space. It is sufficient to model the environment as starting in a
pure state |0〉 = |e0〉 :
ρenv = |e0〉 〈e0| (40)
This is because when the experimentalist prepares the states, it usually tries to undo all the previous correlations
between the environment and the system. A product state of the bath and the system is created, making the
above assumption adequate. Equation 39 becomes:
E(ρ) =∑k
〈ek|U[ρ⊗ |e0〉 〈e0|
]U † |ek〉
E(ρ) =∑k
EkρE†k
with
Ek = 〈ek|U |e0〉 (41)
Because the quantum operation ε is trace preserving, it follows that:∑k
E†kEk = 1 (42)
When gaining new information about the process that occurs when the map is applied∑k
E†kEk ≤ 1 (43)
With these, the mathematical formalism of quantum operation is complete.
The second approach is the physical interpretation. It offers a different view on quantum operations. It
implies the principle of implicit measurement : the state of the system does not change when a measurement on
the environment is made. Supposing result k occurs on the environment, the state of the open quantum system
is :
ρk ∼ trenv(|ek〉 〈ek|U(ρ⊗ (|e0〉 〈e0|)U † |ek〉 〈ek|
)= 〈ek|U(ρ⊗ |e0〉 〈e0|)U † |ek〉 =
EkρE†k
Normalizing:
ρk =EkρE
†k
Tr(EkρE†k)
(44)
The action of the noise can be understood as taking the initial state of the system ρ to one of the states in the
setEkρE
†k
Tr(EkρE
†k
) with probability p(k) = Tr(EkρE
†k
).
The third approach is by stating physically-motivated axioms which govern the behaviour of quantum
operations:
16
A1: Tr(E(ρ)) represents the probability that the process represented by E occurs. 0 ≤ Tr(E(ρ)) ≤ 1
A2: E is a convex linear map:
∀pi
E(∑
i
piρi)
=∑i
E(ρi)
A3: E is a completely positive map
E(A) > 0 ∀A
(1⊗ E)(A) > 0 ∀A ∈ RQ1
(any A in the product space of any random system R added to Q1)
There is a theorem which links the axioms to the previously encountered operator-sum form of a map:
The map E satisfies A1, A2, A3 iff E(ρ) =∑
k EkρE†k with
∑iE†iEi ≤ 1.
The proof can be found in any textbook of quantum information and will not be discussed here, see [8] or
[14].
All quantum operations modify the topology of Bloch’s sphere. For example, the bit-flip operation is de-
scribed by: nothing happens with probability p, a bit is flipped (X gate is applied) with probability (1− p):
E0 =√p 1 (45)
E1 =√
1− p X (46)
Its action is depicted in figure 9:
FIG. 9: Geometrical modification of the Bloch sphere
Something more complex is the loss of energy from a quantum system. This is called amplitude damping
and represents a very important quantum channel. Supposing an optical mode with either zero or one pho-
ton a |0〉 + b |1〉, this is coupled to the environment by introducing a beamsplitter in the way of the ’flying
photons’. The superposition mode is coupled to another single optical mode by the unitary transformation
B = exp(θ(a†b − ab†), where b, b† are the operators for the second mode. For the theoretical description of
17
the beamsplitter Hamiltonian, see [15]. Supposing the environment starts in |0〉, the result after the passage
through the beamsplitter is:
B |0〉(a |0〉+ b |1〉) = aB |00〉+ bB |01〉 = (47)
a |00〉+ bBa† |00〉 = (48)
a |00〉+ bBa†B†B |00〉 = (49)
a |00〉+ b(a†cosθ + b†sinθ
)|00〉 = (50)
a |00〉+ b(cosθ |01〉+ sinθ |10〉
)(51)
Here, the second index in the kets represents the system. Applying the partial trace operation with respect to
the first index (discarding the environment), the quantum operation of amplitude damping is obtained:
EAD(ρ) = E0ρE†0 + E1ρE
†1 (52)
with
E0 =
1 0
0√
1− γ
(53)
and
E0 =
0√γ
0 0
(54)
and
γ = sin2θ (55)
The action of this channel can be understood as it follows: E0 leaves |0〉 unchanged and reduces (as cos θ ≤ 1)
the amplitude of |1〉. The environment ’perceives’ the system more likely to be in |0〉 than in |1〉. Here no
energy is dissipated to the environment E1 changes |1〉 to |0〉 : in practice, this is represented by losing a
photon from the system to the environment. Here energy is dissipated. It can be generalized to channels which
do not only leave the state |0〉 invariant alone. It is then called "generalized amplitude damping".
B. How close two quantum states are? How accurate is a quantum channel?
Having in mind the quantum information as quantum states being modified by quantum operations, it is
important to understand the performance of such channels. That is, how well the information is preserved
under the action of quantum channels? Starting from concepts of classical information theory, a very short
introduction to how distances are measured in quantum information theory is given below.
There are many ways to quantify how close are two classical probability distributions. Three of them are:
1)The Kolmogorov distance
18
2)The Kantorovich-Rubinstein distance
3)The bounded-Lipschitz distance
A particularly useful measure is the first one, the Kolmogorov distance, named after a Russian mathemati-
cian with hugely important contributions in probability theory . It is defined as:
D(px, qx
)=
1
2
∑x
|px − qx| (56)
It has the following properties:
1)it is symmetric in its arguments
2)it respects the triangle inequality
D(x, z) ≤ D(x, y) +D(y, z) (57)
1),2) combined give the Kolmogorov distance the property of being a metric. Another useful quantization of
how close together are two probability distributions is the Fidelity:
By definition:
Fidelity(px, qx
)=∑x
√pxqx (58)
This has the physical interpretation that it is the inner product between two vectors (px and qx) lying on the unit
sphere. These vectors lie on the unit sphere because their norm is equal to 1 (the sum of probabilities from a
probability distribution has to sum to 1:∑
x (√px)2).
Moving to the quantum realm, the distance between two quantum states is defined as:
D(ρ, σ) =1
2Tr|ρ− σ| (59)
where |A| =√A†A.
For [σ, ρ] = 0, the two quantum states are diagonal in the same basis, that is they can be written as:
ρ =∑i
ri |i〉 〈i|
σ =∑i
si |i〉 〈i|
If this is the case, the quantum distance becomes a classical distance between two eigenvalues of the states:
D(ρ, σ) =1
2Tr|
∑i
(ri − si
)|i〉 〈i|
= D(ri, si)
Using the Bloch’s sphere representation of a quantum state, the distance between the states can be viewed
as:
ρ =1+ ~r · ~σ
2
σ =1+ ~s · ~σ
2
19
D(ρ, σ) =1
2Tr|ρ− σ| = 1
2tr|(~r − ~s) · ~σ| 1
2
As(~r − ~s
)· ~σ has eigenvalues ±|~r − ~s|,
tr|(~r − ~s) · σ| = 2|~r − ~s| (60)
Thus,
D(ρ, σ) =1
2|~r − ~s| (61)
Also, the trace distance is preserved under unitary operations:
D(ρ, σ) = D(UρU †, UσU †) (62)
It can be shown that it is a metric on the space of density operators:
1) D(ρ, σ) = 0 iff ρ = σ,
2) D(·, ·) is symmetric in its arguments and
3) it respects the triangle identity: D(ρ, τ) ≤ D(ρ, σ) +D(σ, τ).
The above can lead to the fact that trace preserving quantum operations are contractive, that is:
D(E(ρ), E(σ)
)≤ D(ρ, σ) (63)
For completeness, one can also list the fact that the trace distance possesses strong convexity and thus is
jointly convex in its inputs.
V. CONCLUSION AND ACKNOWLEDGEMENTS
Throughout this report, I have extensively used ideas from the following books: [8], [14]. As I found out
from more than one researcher in quantum computing/information, these are the two main books one starts
with when studying about these topics. I am glad I have managed to go through many topics presented in these
two books. None of the results cited above are my own and more in depth-references can be found at the end
of every chapter in [8].
I have given a short introduction to the ideas I have played around with during my internship at the Depart-
ment of Theoretical Physics of the National Institute of Physics and Nuclear Engineering (NIPNE), Bucharest,
Romania. As this is intended to be a short report, it is somehow understood that the size of this document has
not to be too big. Thus, I only cited some results, omitting steps and not giving complete proofs. I hope that a
vague idea of my work and how I spent the 2 months can be seen from this report and is clear to the reader.
I have truly enjoyed studying about quantum computing and quantum information. I have seen what it
means to work in a theoretical physics department and I have benefited from insights given by researchers who
published in international journals. This is definitely going to be useful for me in the long run.
I would like to say thank you to all the persons who have facilitated me to obtain this fund. I would like to
express my gratitude to the donor, Christopher Charles Rokos, MA in Mathematics, Pembroke College ’89.
20
Bibliography
[1] Y. Wang, Y. Li, Z. Yin, and B. Zeng. 16-qubit IBM universal quantum computer can be fully entangled :
arxiv:1801.03782v2. 2
[2] N. Rubin, J. McClean, and R. Babbush. Application of fermionic marginal constraints to hybrid quantum algorithms.