Quantum Computation and Quantum Information

See discussions, stats, and author profiles for this publication at: https://www.researchgate.
net/publication/220594564
Quantum computation and quantum information
Article in International Journal of Parallel Emergent and Distributed Systems · February 2006

DOI: 10.1080/17445760500355678 · Source: DBLP
CITATIONS READS
34 6,537
2 authors, including:
Marius Nagy
Prince Mohammad University
31 PUBLICATIONS 227 CITATIONS
SEE PROFILE
All content following this page was uploaded by Marius Nagy on 27 May 2014.
The user has requested enhancement of the downloaded file.

The International Journal of Parallel, Emergent and Distributed Systems,
Vol. 21, No. 1, February 2006, 1–59
Quantum computation and quantum information*

MARIUS NAGY* and SELIM G. AKL†
School of Computing, Queen’s University, Kingston, Ont., Canada K7L 3N6
(Received 13 May 2005; in final form 8 July 2005)
The paper is intended to be a survey of all the important aspects and results that have shaped the field of
quantum computation and quantum information. The reader is first familiarized with those features and
principles of quantum mechanics providing a more efficient and secure information processing. Their
applications to the general theory of information, cryptography, algorithms, computational complexity and
error-correction are then discussed. Prospects for building a practical quantum computer are also analyzed.
Keywords: Quantum mechanics; Qubit; Superposition; Interference; Measurement; Entanglement;
Quantum gates; Information theory; Cryptography; Quantum computer; Algorithms; Complexity;
Error-correction; Experiments
1. Introduction and overview
Quantum computation and quantum information can be defined as the study of information
processing tasks accomplished using quantum mechanical systems. One of the most striking
aspects of it is the complete uncertainty about its future. It could rise to meet the expectations
of enthusiasts of the field, causing computer science to be reconsidered (and effectively
rewritten) in the new quantum light. In such a case, quantum computers and practical
applications of quantum information theory will substantially impact our everyday life. But it
may also happen that quantum mechanics will one day be disproved or the formidable
obstacles towards making quantum computers a viable technology will prove
insurmountable. This will leave quantum computation and quantum information as abstract
mathematical curiosities, without substance.
1.1 Origins of quantum computing
Most people involved in the field associate the birth of quantum computation and quantum
information with a talk Richard Feynman gave at MIT in 1981 (see [80] for the journal
*Corresponding author. Email: [email protected]

*This research was supported by the Natural Sciences and Engineering Research Council of Canada.
†
Email: [email protected]
The International Journal of Parallel, Emergent and Distributed Systems

ISSN 1744-5760 print/ISSN 1744-5779 online q 2006 Taylor & Francis
http://www.tandf.co.uk/journals
DOI: 10.1080/17445760500355678
2 M. Nagy and S. G. Akl
version). In his talk, he pointed out the difficulty of simulating quantum systems using
classical computers. In part, this is due to the number of variables a computer must keep track
of, when simulating the behavior of a quantum system. This number grows exponentially
with the size of the system being modeled. Other reasons include strange effects that are
specific to quantum mechanics and cannot be accurately simulated through classical means.
Consequently, Feynman conjectured that a machine built in such a way as to allow it to make
use of quantum effects would be able to efficiently simulate quantum systems. And so, the
idea of a quantum computer was born.
Although Feynman’s motivation was the concept of a “universal quantum simulator”, a
machine capable of imitating any quantum system (including the physical world), from a
computer science perspective his observations led to speculation that perhaps computation in
general could be done more efficiently if it made use of these quantum effects. In a later paper
[81], Feynman analyzes the physical limitations of computers due to the laws of physics and
tries to exhibit the principles on which a quantum mechanical computer could be built.
As researchers began to explore ways to harness quantum effects in order to speed-up
computation or find useful applications, the field of quantum computation and quantum
information gradually came into its own. In their effort to develop tools that would allow
them to gain a better understanding of quantum phenomena, physicists have also contributed
to the progress of the field.
Quantum computation and quantum information is a multi-disciplinary area, so it comes
as no surprise that people involved in it may have fairly different backgrounds
(mathematicians, statisticians, computer scientists, physicists, cryptographers, information
theorists, etc.). The power of this novel computational paradigm comes from its foundation:
quantum mechanics. Quantum mechanics, physicists say, is the most complete and accurate
description of the world we currently have. It is a mathematical theory consisting of rules and
principles defining a framework used to develop physical theories. In this sense (as a
mathematical theory) quantum mechanics is not difficult to understand. Familiarity with the
tools linear algebra provides for manipulating state vectors in Hilbert spaces is all a computer
scientist needs in order to tackle quantum information processing tasks. The difficulty
appears when we try to apply quantum mechanics to understanding the structure of some
complicated molecules or grasping the nature of the forces responsible for keeping the
particles in atomic nuclei together.
1.2 Interpretations of quantum mechanics

From a broader perspective, the difficulties associated with quantum mechanics lie in finding
a correspondence between its predictions and what happens in the physical reality. In other
words, it is a matter of the interpretation given to quantum mechanics. It is in this context
that we must understand Neils Bohr’s statement that “anyone who thinks he understands
quantum theory doesn’t really understand it”. This sentence expresses the fact that we cannot
use our everyday life experience and common sense intuitions to gain insights into quantum
phenomena.
1.2.1 The Copenhagen interpretation. Neils Bohr, one of the founder of quantum theory
and apparently a very determined and strongly willed person, is credited to have imposed
what we now call the Copenhagen interpretation, which became the standard view among
Quantum computation and information 3
many physicists. According to this view, it does not make sense to ascribe intrinsic properties
(such as position or velocity) to isolated quantum entities (such as electrons, photons or other
elementary particles). The properties of quantum systems only make sense in light of the
measurements we make. Taken to its limit, the Copenhagen interpretation denies the reality
of an individual photon or electron until it has been observed.
Let us see how different interpretations of quantum mechanics can explain the result of the
double-slit experiment. It was first conducted by Thomas Young in 1801 and demonstrated
that light behaves like waves. In his experiment, Young projected light onto a screen through
a barrier pierced with two closely spaced slits (see figure 1). What he observed on the screen
was an interference pattern, the hallmark of waves. The importance of modern-day versions
of Young’s experiment is best illustrated by Richard Feynman in his Lectures [82]. He
believed that the result of the double-slit experiment was the fundamental mystery of
quantum mechanics.
If Young performed his experiment using simple screens and candlelight, the tremendous
advances in technology allow us today to repeat the experiment with very weak light, that is,
light produced as one photon at a time. Thus, it is very unlikely that several photons would be
found within the experimental apparatus at the same time. Surprisingly (and against our
intuitions), given that enough time elapses as to allow the photons, arriving one at a time, to
accumulate on the screen, the same interference pattern will appear. The obvious question is:
what was each photon interfering with, if it was alone in the experimental apparatus?
According to the Copenhagen interpretation, the only possible answer can be: with itself.
In the absence of any observations, it does not make sense to associate a specific route to the
photon in its way from the light source to the screen. In a sense, each particle went not
through one slit, but rather through both slits, and as it appeared on the other side, it interfered
with itself. As we will see in the next section, this behavior is a manifestation of the quantum
principle of superposition of states, a principle without which quantum computation and
quantum information would be unconceivable. The duality between particles and waves has
also been demonstrated for other quanta that can be localized (electrons, neutrons, atoms)
and even for larger entities, like complex molecules composed of tens of atoms.
Figure 1. Young’s double-slit experiment.

1.2.2 Many worlds interpretation. Although, the Copenhagen interpretation is over half a
century old, there still are physicists today who embrace it. Suffices to name Anton Zeilinger,
who was involved in the experimental demonstration of superdense coding and quantum
teleportation, two of the most important applications of quantum information theory.
However, tentatives of applying quantum mechanics to the universe as a whole led to
renewed interest in the many-universes interpretation, initially proposed by Hugh Everett III
in 1957. His interpretation essentially differs from the Copenhagen interpretation because it
removes completely the measurement problem, by eliminating any references to observers.
According to the modern version of the many-worlds interpretation, our universe is
embedded into an infinitely larger and more complex structure, called the multiverse, which
we can approximate to a system of parallel universes [42,63,64]. Each time there is a decision
at the quantum level, such as the probability of a radioactive atom to decay or a photon
passing through a polarizing filter, the assemblage of universes differentiates along different
paths.
How does this interpretation explain the interference pattern created by single particles?
Since there is a choice regarding which route the particle will take, at that point the universe
will split into two distinct universes. In one of them, the particle goes through the first slit,
while its copy from the second universe will traverse the barrier through the second split.
On the other side of the barrier, the two universes will interfere with each other creating a
series of fringes on the screen. So, according to the many-universes interpretation, the
particle is not interfering with itself, but rather with its copy from a parallel universe.
Although it sounds like a rather crazy idea, the multi-verse structure offers sometimes a more
intuitive explanation to some quantum phenomena than the classical Copenhagen
interpretation. David Deutsch, who may be regarded as the strongest advocate of the
many-worlds interpretation (and also had some remarkable contributions to the development
of quantum complexity theory), even proposed an experiment that would test the existence of
multiple universes, and therefore be able to distinguish between the different interpretations
of quantum theory [60].
1.3 Overview
The remaining sections of this paper try to address the fundamental concepts and questions
underlying quantum computation and quantum information. Is indeed this novel approach to
performing computation more powerful than the one in use today? How much more
powerful? What are those information processing tasks that prove its superiority? Can
quantum effects only speed up computation or are there tasks that can only be performed
through quantum means? Can we identify the conceptual elements responsible for the power
exhibited by a quantum computer? And, finally, is it possible to build such a machine that
would exploit quantum effects in order to achieve unprecedented computational power?
The next section describes the fundamental principles of quantum mechanics on which
quantum computation and quantum information is based. Section 3 is constructed in analogy
with classical information theory and discusses two simple (yet important) quantum
information processing tasks: superdense coding and quantum teleportation. Section 4 is
concerned with the most promising practical application of quantum information (at least in
the short term): quantum cryptography. The most important quantum algorithms conceived
up to now are presented in Section 5. Section 6 introduces the main quantum complexity
classes and their relations with classical ones. Any computing technology, regardless of the
physical implementation, is subject to various types of errors, and quantum computing makes
no exception. Section 7 analyzes possible solutions to this problem. Section 8 is a review of
physical realizations of quantum bits and logical operations performed on them (quantum
gates). Conclusions and final remarks are presented in Section 9.
2. Fundamentals of quantum information and quantum computation
The field of quantum information and quantum computation is based on the postulates
governing quantum mechanics. The aim of this section is to provide a description of these
postulates and mathematical formalisms required to work with them to the extent needed for
quantum computing. Good introductions to quantum mechanics for computing scientists can
be found in [27,98,139,146,155], but for detailed expositions of the field one should see [82].
Quantum mechanics takes place in the framework provided by linear algebra. We can
associate to any isolated physical system a complex vector space with an inner product
defined on it, known as the state space of the system. Mathematically, such a vector space
with an inner product is called a Hilbert space. At any given point in time, the system is
completely described by its state vector, which must be a unit vector in the system’s state
space.
Quantum state spaces and the transformations acting on them are traditionally described in
terms of vectors and matrices using the compact bra/ket notation introduced by Dirac [70].
According to his conventional notation, kets like jxl are simply column vectors, typically
used to describe quantum states. Similarly, the matching brakxj is a row vector denoting the
conjugate transpose of jxl.
2.1 The qubit

At an abstract level, the simplest quantum mechanical system is the quantum bit, or qubit.
A qubit is a unit vector in a two-dimensional state space, for which a particular orthonormal
basis, denoted by {j0l, j1l} has been fixed. The two basis vectors j0l and j1l correspond to the
possible values a classical bit can take. However, unlike classical bits, a qubit can also take
many other values. In general, an arbitrary qubit jCl can be written as a linear combination of
the computational basis states:
jCl ¼ aj0l þ bj1l; ð1Þ
where a and b are complex numbers such that jaj2 þ jbj2 ¼ 1. This is the fundamental
difference distinguishing quantum bits from classical ones and is a direct application of the
quantum principle of superposition of states. The qubit jCl in equation (1) is in a
superposition of j0l and j1l, a state in which it is not possible to say that the qubit is definitely
in the state j0l, or definitely in the state j1l. After all, what better intuition about the
superposition principle than the idea (quite old and widely accepted now) that each particle is
also a wave?
For a single qubit, there is a very intuitive graphical representation of its state as a point on
the unit three-dimensional sphere (also called the Bloch sphere). Figure 2 depicts four
possible states of a qubit using the Bloch sphere representation. Note that the states
Figure 2. The Bloch sphere representation of a qubit.
corresponding to the points on the equatorial circle have all equal contributions of 0-ness and
1-ness.
pffiffiffi What distinguishes
pffiffiffi them is the phase. For example, the two states displayed above,
1= 2ðj0l þ j1lÞ and 1= 2ðj0l 2 j1lÞ are the same up to a relative phase shift of p, because
the j0l amplitudes are identical and the j1l amplitudes differ only by a relative phase factor of
e ip ¼ 2 1.
We have described qubits as mathematical objects, but there are real physical systems
which may be described in terms of qubits. Possible physical realizations of a qubit include
two different polarizations of a photon, the alignment of a nuclear spin in a uniform magnetic
field or two electronic levels in an atom. There will be more to say about this in section 8.
2.2 Measurements
We now turn our attention on the amount of information that can be stored in a qubit and,
respectively, retrieved from a qubit. Since any point on the Bloch sphere can be characterized
by a pair of real-valued parameters taking continuous values, it follows that, theoretically, a
qubit could hold an infinite amount of information. However, we cannot extract more
information from such a qubit than we are able to from a classical bit. The reason is that we
have to measure the qubit in order to determine which state it is in. And another of the
fundamental postulates of quantum mechanics, the one regarding measurements, restricts us
in the amount of information that can be gained about a quantum state through measurement.
According to this postulate, when we measure a qubit jCl ¼ aj0l þ bj1l with respect to the
standard basis for quantum computation {j0l, j1l}, we get either the result 0 with probability
jaj2, or the result 1 with probability jbj2. The condition that the probabilities must sum to one
corresponds geometrically to the requirement that the qubit state be normalized to length 1,
that is the inner product kCjCl equals 1.
Furthermore, measurement alters the state of a qubit, collapsing it from its superposition of
j0l and j1l to the specific state consistent with the measurement result. For example, if we
observe jCl to be in state j0l through measurement, then the post-measurement state of the
qubit will be j0l, and any subsequent measurements (in the same basis) will yield 0 with
probability 1. In general, measurement of a state transforms the state into one of the
measuring device’s associated basis vectors. The probability that the state is measured
as basis vector jul is the square of the norm of the amplitude of the component of the original
state in the direction of the basis vector jul.
2.3 Uncertainty principle

As noted by Julian Brown (in [42] page 120), Berthiaume and Brassard connect Heisenberg’s
uncertainty principle with the idea that any measurement yielding information in the classical
form induces an irrevocable destruction of some remaining information, making further
measurements less informative, if not completely useless [28]. Our objection to this point of
view is that while it is perfectly true that, in general, measurement disturbs the quantum state
of the observed system, this is not what Heisenberg’s uncertainty principle is trying to
quantify. Heisenberg showed that it is not possible, according to quantum mechanics, to
simultaneously know both the position and momentum of a particle with arbitrary accuracy.
Position and momentum form a pair of canonically conjugate quantum variables and the
more we know about one of them, the more we are restricted in the precision with which the
other one can be measured.
Later, Heisenberg’s uncertainty principle was generalized to any two quantities
corresponding to non-commuting operators [157]. An equivalent formulation of this general
uncertainty relation is developed in [146] based on expressing projective measurement and
standard deviations in terms of observables. Of course, there are many other pairs of
conjugate variables (beside position and momentum), and we only mention here energy and
time (who actually needs a special treatment, as given in [167]), horizontal and vertical
polarizations or right-handed and left-handed circular polarizations. It is also important to
stress that uncertainty relations of any kind do not express the incapacity of the measurement
device to provide accurate observations of the variables, due to their imperfections. These
relations are mathematical results derived from the principles governing quantum mechanics
and therefore hold regardless of the performances of the measuring devices used.
2.4 No-clonability
Naturally, measurements in bases other than the computational basis are always possible, but
this will not help us in determining a and b from a single measurement. One might think of
solving this problem by making multiple copies of the initial qubit jCl and then measure
each of the copies in order to obtain an estimation of a and b. In fact, it turns out to be
impossible to make a copy of an unknown quantum state. The no-cloning theorem, one of the
earliest results of quantum computation and quantum information [184], states that quantum
mechanics prevents us from building a quantum cloning device capable of copying non-
orthogonal quantum states. The ability to clone orthogonal quantum states translates into the
ability to copy classical information, since the different states of classical information can be
thought of merely as orthogonal quantum states. So it seems that quantum mechanics places
severe limitations on the accessibility of quantum information, but sometimes this can be
used to our advantage, as we will see in the case of quantum cryptography.
2.5 Superposition and interference
The concepts of superposition, interference and measurement can be very well illustrated in
the case of Young’s two-slit experiment. The experimental setup provides the particle with
a particular kind of superposition. If we ascribe state j0l to the particle when it passes through
slit A and state j1l when it passes through slit B,
pffiffithen,
ffi effectively,
pffiffiffi the particle’s behavior can
be described by the superposition of states 1= 2j0l þ 1= 2j1l; which is a combination of
“particle goes through slit A” with “particle goes through slit B”. In the case when we choose
to observe the particle as it goes through the experimental apparatus (that is, to measure its
state), the wave function describing it will collapse into one of the two possible outcomes and
the particle will be detected passing through slit A or B with equal probability. In either case,
the superposition is destroyed and with it any chance of interference. But if the particle is not
observed until the end, as it collects on the screen, then the superposition holds through to the
end, enabling the interference phenomenon witnessed on the screen.
2.6 Quantum registers
Let us examine now more complex quantum systems, composed of multiple qubits.
In classical physics, individual two-dimensional state spaces of n particles combine through
the Cartesian product to form a vector space of 2n dimensions, representing the state space of
the ensemble of n particles. However, this is not how a quantum system can be described in
terms of its components. Quantum states combine through the tensor product to give a
resulting state space of 2n dimensions, for a system of n qubits. It is this exponential growth
of the state space with the number of particles that quantum computers try to exploit in their
attempt to achieve exponential speed-up of computation over classical computers.
For a system of two qubits, each with basis {j0l, j1l}, the resulting state space is the set of
normalized vectors in the four-dimensional space spanned by basis vectors {j0l^j0l,
j0l^j1l, j1l^j0l, j1l^j1l}, where jxl^jyl denotes the tensor product between column
vectors jxl and jyl. It is customary to write the basis in the more compact notation {j00l, j01l,
j10l, j11l}. This generalizes in the obvious way to an n-qubit system with 2n basis vectors.
2.7 Entanglement
Similar to single qubits, multiple-qubit systems can also be in a superposition state.
The vector
1
jCl ¼ ðj00l þ j01l þ j10l þ j11lÞ ð2Þ
2
describes a superposition state of a two-qubit system in which all four components
(corresponding to the four basis vectors) have equal amplitudes. What about the two qubits
composing the system? Can we characterize their states individually? If we rewrite equation
(2) in order to express jCl as the tensor product

1 1 1 1
jCl ¼ pffiffiffi j0l þ pffiffiffi j1l ^ pffiffiffi j0l þ pffiffiffi j1l ð3Þ
2 2 2 2
then we can legitimately assert that each of the component qubits is also in a superposition
state, perfectly balanced between j0l and j1l. Now let us drop the two middle terms in
equation (2) and consider the superposition state described by
1 1
jFl ¼ pffiffiffi j00l þ pffiffiffi j11l ð4Þ
2 2
In this case, it is no longer possible to find complex numbers a, b, g and d such that
1 1
ðaj0l þ bj1lÞ^ðgj0l þ dj1lÞ ¼ pffiffiffi j00l þ pffiffiffi j11l ð5Þ
2 2
The state of the system cannot be decomposed into a product of the states of the
constituents. Even though the state of the system is well defined (through the state vector
jFl), neither of the two component qubits is in a well-defined state. This is again in contrast
to classical systems, whose states can always be broken down into the individual states of
their components. Furthermore, if we try to measure the two qubits, the superposition will
collapse into one of the two basis vectors contributing to the superposition and the outcomes
of the two measurements will always coincide. In other words, if one of the qubits is found to
be in state j0l, then the second one will necessarily be in the same state, while a state j1l
assumed after measurement will be shared by both qubits. Therefore, we say that the two
qubits are entangled and jFl describes an entangled state of the system.
Entanglement defines the strong correlations exhibited by two or more particles when they
are measured and which cannot be explained by classical means. This does not imply that
entangled particles will always be observed in the same state, as entangled states like
1 1
pffiffiffi j01l ^ pffiffiffi j10l ð6Þ
2 2
prove it. States like these or the one in equation (4) are known as Bell states or EPR pairs after
some of the people (see below) who pointed out their strange properties.
In some sense, we can say that superposition encompasses entanglement, since
entanglement can be viewed as a special case of superposition. It is also interesting to make
an analogy between entanglement and the concept of primality from number theory. Indeed,
an entangled state of the system corresponds to a prime number, since it cannot be factored or
decomposed as a product of subsystem states.
2.7.1 EPR thought experiment. Entanglement is by far the most counterintuitive concept in
quantum mechanics that we have met. It can give rise to strange phenomena that have fueled
endless debates between the advocates of the theory of quantum mechanics and those trying
to disprove it (the arguments exchanged by Bohr and Einstein on this subject have become
history [2]). Einstein was deeply dissatisfied with the fact that quantum mechanics allowed
correlations between entangled particles to manifest themselves instantaneously over
arbitrary large distances.
According to quantum mechanics, if the two particles in an EPR pair are arbitrarily far
apart and we measure one of them, the other will instantly acquire a definite state. This
seemingly instant influence over possibly infinite distances, unmediated by any
communication or propagation medium is called non-locality, but Einstein termed it
“spooky action at a distance” ([2] page 122). He believed that such phenomenon is due to
some hidden variable in the state of the system that we simply do not have knowledge of.
Supporters of this local hidden variable theory assume that each particle has some internal
state (for the moment hidden from us) that completely determines what the result of any
given measurement will be. The strong correlation exhibited, e.g. by the EPR pair in equation
(4) has in their view a simple explanation: the particles are either both in state j0l or both in
state j1l, we just do not happen to know which.
Einstein presented his viewpoint in a 1935 paper ([74]) written together with Podolsky
and Rosen (hence the EPR acronym). They imagined a thought experiment that would either
show quantum mechanics as being an incomplete theory of Nature or, basically, contradict
Heisenberg’s uncertainty principle. Bohr’s reply came only a few months later, [33],
pointing out the flaw in Einstein’s argument: two entangled particles form an inseparable
quantum system and any measurement carried out on one of them will inevitably influence
the circumstances of the other. But the dispute between adepts of realistic local theories and
supporters of quantum mechanics (and implicitly entanglement) was far from being over.
2.7.2 Bell’s inequality. John Bell set up a new battlefield in 1964 ([12]), in the form
of a mathematical expression that clearly separates classical from quantum correlations.
This expression, which takes the form of an inequality known as Bell’s inequality quantifies
the limit on the strength of correlations we can expect to see if hidden variables are involved.
On the other hand, quantum mechanics ought to violate this inequality. Bell’s inequality was
only the first in a larger set of inequalities of this kind. Another example is the CHSH
inequality, named after the initials of its four discoverers [53].
Bell’s result prompted a series of experiments arduously aimed at demonstrating that the
laws of Nature do indeed violate Bell’s inequality. In the most famous of them, the
correlations in polarizations between entangled photons was measured [3]. Alain Aspect and
his colleagues claimed that the results confirmed the violation of Bell’s inequality, thus
proving the existence of entanglement in Nature and the validity of quantum mechanics.
Later experiments demonstrated quantum-correlations over larger distances [170,175]. But
there are voices who doubt the results of these experiments, accusing a lack of justification
for all the assumptions on which the experiments are based [134,172 –174]. They consider
the attitude of the experimenters to be biased and the whole demonstration of quantum
entanglement as a circular argument: the assumptions needed to back quantum mechanics are
unlikely to be true unless quantum mechanics is true.
Any experiment carried out to prove that quantum mechanics violates the Bell inequalities
requires an extended run of measurements. The frequencies with which pairs of
measurements agree with one another are then calculated. It is claimed that the EPR
effect has been convincingly demonstrated with statistically significant results. But it is
exactly the statistical nature of the argument that has been pointed out as a weakness. Critics
of these experiments consider that their statistical results have been biased by resorting to
convenient assumptions, insufficiently founded.
2.7.3 3-particle entanglement. Under these circumstances, conceiving a new kind of

experiment that would allow the clash between quantum and classical reality to be decided in
one measurement seemed an important achievement. The GHZ experiment [92] is basically
an extension of the EPR experiment with three correlated particles instead of two. The three-
particle entanglement in the GHZ proposal provides the means to prove the contradiction
without the cumbersome use of inequalities, in a much more direct and non-statistical way, as
compared with Bell’s original theorem.
In 1999, a team of researchers at MIT produced a GHZ state using nuclear spins instead
of photon polarizations and the techniques of nuclear magnetic resonance (NMR)
spectroscopy to manipulate their sample [144]. Although, they were able to measure the
weird quantum correlations exhibited by the GHZ state, the NMR techniques used
prevented them from testing the non-local aspects of the GHZ experiment. The three qubits,
embodied as nuclear spins within a molecule, were far too close together to allow for a
delayed-choice experiment of the kind Aspect performed on EPR pairs. There is also a very
recent report about the successful creation of a 3-particle entangled GHZ state between
trapped ions [55], as part of the effort to develop a scalable technology for building a
quantum computer.
2.7.4 Entanglement as a physical resource. Today, entanglement is no longer regarded as

merely a quantum curiosity. It has become the essential feature of quantum mechanics and
people are seeing it as a physical resource that can be spent in order to solve information-
processing tasks in new ways. Some researchers have even begun to quantify
entanglement by introducing a standard unit of entanglement in the form of a
maximally entangled pair, which is said to carry one e-bit of entanglement [145]. By
contrast, incompletely entangled pairs carry less than one e-bit. They developed ways to
weigh entanglement using entanglement distillation and dilution. Therefore, the question
of whether this “resource” really exists in Nature and can be exploited in the ways we
currently believe it can, is of crucial importance for the future of quantum
computation and quantum information†. Superdense coding, teleporting a quantum
state or finding the prime factors of a composite integer in polynomial time on a quantum
computer are only a few of the most important applications of entanglement. The current
potential quantum computation and quantum information are credited with will certainly
be affected if some day entanglement will be disproved.
2.8 Quantum evolution

Let us assume for now that we can harness entanglement to increase the efficiency of
computation and information-processing tasks in general. What are the “circuits”
composing a hypothetical quantum computer? The answer to this question is in strong
relation with the way an isolated quantum system evolves over time. We already saw what
happens when we try to measure such a quantum system. If, for example, we are trying to
read the content of a quantum memory register, the system will undergo a sudden,
unpredictable jump into one of the classical bit string configurations composing the
original superposition. In other words, there will be a discontinuity in the evolution of the
quantum memory register. But, if we leave the register unobserved, the system will
undergo a smooth, continuous evolution governed by Schrödinger’s equation, a
deterministic differential equation which enables us to predict the future or uncover the
past evolution of the memory register. Consequently, any quantum computation is
reversible and therefore quantum gates (the quantum analog of classical gates) must
always have as many outputs as they have inputs, in order to avoid any loss of information
that would prevent the computation to be undone.
†
Text in “bold” is meant to highlight important open problems or areas of intense research.
2.9 Quantum gates
A quantum NOT gate acting on a single qubit will evolve the initial state aj0l þ bj1l into the
final state aj1l þ bj0l, in which the roles of j0l and j1l have been interchanged. Because
every quantum gate acts linearly, the transformation is fully specified by its effect on the
basis vectors. Hence, there is a very convenient representation of a quantum gate in matrix
form. The matrix X representing the quantum NOT gate is then defined as follows:
" #
0 1
X¼ :
1 0
The first column represents the effect of applying the NOT gate to state j0l, while the
second column is the result of applying the NOT gate to state j1l. We can now describe the
operation of the quantum NOT gate, acting on an arbitrary qubit state, through the following
equation:
" # " #
a b
X· ¼ :
b a
Other examples of single qubit gates are the Z gate:

" #
1 0
Z¼ ;
0 21
which leaves j0l unchanged, but introduces a phase shift by flipping the sign of j1l, and the
Hadamard gate:
" #
1 1 1
H ¼ pffiffiffi ;
2 1 21
which is one of the most useful quantum gates, because it creates superpositions of j0l
and j1l.
Although, there are an infinite number of single qubit gates, no two by two matrix is a
legitimate representation of a quantum gate. Schrödinger’s equation states that the dynamics
of a quantum system must take states to states in a way that preserves orthogonality. In other
words, the normalization condition jaj2 þ jbj2 ¼ 1 for the initial state aj0l þ bj1l must also
be true for the quantum state after the gate has acted. This translates into the requirement that
the matrix U describing the single qubit gate be unitary, that is U*·U ¼ I where U* is the
conjugate transpose of U. Single qubit gates can be conveniently visualized as rotations of
the arrow representing the qubit state on the surface of the Bloch sphere.
Quantum gates on multiple qubits can also be defined. Figure 3 depicts a controlled-NOT
gate, an instance of the more abstract controlled-U gate, where U ¼ X. The target bit jyl is
flipped if and only if the control bit jxl is set to 1.
Multiple qubit gates must also satisfy the requirement that probability be conserved, so
they too must be unitary transformations. Since any unitary matrix is invertible and the
inverse is also a unitary matrix, it follows that a quantum gate can always be inverted by
another quantum gate. The set of all 1-qubit rotations (gates) together with the controlled-
NOT gate is universal for quantum computation. But finite universal sets of gates exist as
well. Two researchers working independently have shown that any imaginable quantum
Figure 3. Controlled-NOT quantum gate.
computation can be performed by connecting together multiple copies of a certain 2-qubit

gate [4,72]. Such universal quantum gates are similar to the NAND gate in classical
computation.
2.10 Density operators
The purpose of this section was to introduce the postulates of quantum mechanics on which
quantum computation and quantum information processing is based. These postulates were
formulated using the language of state vectors. However, there is an alternate formulation,
that is mathematically equivalent to the state vector approach, which proves to be much more
convenient in some situations, notably as a tool for the description of individual subsystems
of a composite quantum system. This tool, known as the density operator or density matrix
provides a convenient means for characterizing quantum systems whose state is not
completely known.
Suppose that a quantum system is in one of a number of states jcil, with respective
probabilities pi. Then, by use of outer products, we can describe the state of the system
through the following density operator (matrix):
X
r¼ pi jci lkci j: ð7Þ
i
As the states jcil are known exactly (they can be represented by a state vector), they are
called pure states, as opposed to the state described by r, which is a mixed state. There is a
simple criterion to determine whether a state is pure or mixed, using the trace of a matrix.
A pure state satisfies tr(r 2) ¼ 1, while for a mixed state, tr(r 2) is always less than 1.
Density operators may be regarded as a more general approach than state vectors. This is
clearly seen in the analysis of composite quantum systems, where the use of reduced density
operators is virtually indispensable. Here is a significant pexample.
ffiffiffi According to the
definition, the density matrix for the Bell state ðj00l þ j11lÞ= 2 is:

j00l þ j11l k00j þ k11j j00lk00j þ j11lk00j þ j00lk11j þ j11lk11j
r¼ pffiffiffi pffiffiffi ¼ :
2 2 2
If we want a description of the state of the first qubit, we can find its reduced density
operator by tracing out the second qubit from the matrix representing the joint system:
j0lk0j þ j1lk1j I
r1 ¼ tr2 ðrÞ ¼ ¼ :
2 2
This result tells us that the first qubit is in a mixed state, since
2 !
I 1
tr ¼ , 1:
2 2
Naturally, a similar result can be obtained for the second qubit. Notice again the paradox
introduced by entanglement. Although the joint system is in a pure state (it is known
completely), any of the composing subsystems are in mixed states, apparently suggesting that
we do not have complete knowledge about it.
From this example we can see that density operators are able to provide a description for
the state of a quantum subsystem, even when no state vector can be associated with that
subsystem. Yet, in our opinion, density operators fail to capture the essence of entanglement,
since a qubit that is either in state j0l or in state j1l with equal probability will yield the same
density matrix
j0lk0j þ j1lk1j
r¼ ;
2
as the entangled qubit above.
3. Quantum information theory
Quantum information theory is the field concerned with the study of elementary information
processing tasks achieved using quantum mechanics. In a wide context, quantum information
theory seeks to identify those resources, separating the quantum from the classical world,
which create new possibilities for processing information. More concrete, and in analogy
with the classical field, work on quantum information theory can be characterized by the
following fundamental goals:
. identify elementary classes of static resources in quantum mechanics (e.g. bits, qubits,
entanglement)
. identify elementary classes of dynamical processes in quantum mechanics (e.g. classical
or quantum information transmission, noise in a classical or quantum communications
channel)
. quantify the resources required to perform elementary dynamical processes.
Since classical physics arises as a special case of quantum physics, all the static and
dynamic elements of classical information theory are also present in quantum information
theory. However, the latter is broader in scope, for it also includes additional static and
dynamic elements, that are specific to quantum mechanics.
3.1 Classical information through quantum channels

The most fundamental results in classical information theory are Shannon’s noiseless
channel coding theorem and noisy channel coding theorem [161]. It is natural for
quantum information theory to look at these two problems in a more general context. In a
first step, only the storage medium is changed, so that classical information is transmitted
using quantum states as the vehicle. It turns out that using qubits to compress classical
information does not result in any significant saving in the amount of communication
required to transmit information over a noiseless channel [146]. In the more realistic case,
where the communications channel is affected by noise, the channel’s true capacity is
quantified exactly by Shannon’s noisy channel coding theorem. This result proved difficult
to replicate for a quantum channel, due to the huge variety of noise models allowed by
the continuous space in which quantum evolution takes place. The Holevo– Schumacher –
Westmoreland theorem [103,160] provides only a lower bound on the capacity of such a
channel. It is still a major open problem of quantum information theory to determine
whether or not encoding using entangled states can be used to raise the capacity
beyond this lower bound.
3.2 Quantum information through quantum channels
The analogy with Shannon’s coding theorems can be taken further, by considering
quantum states themselves as the static resource involved in compression and
decompression manipulations. Compressing the output produced by a quantum
information source is still possible, but the process may no longer be error-free.
The quantum states being produced by the source may be slightly distorted by the
compression – decompression procedure. The average distortion introduced by a
compression scheme is quantified by a fidelity measure, analogous to the probability of
doing the decompression correctly. Schumacher’s quantum noiseless channel coding
theorem [159] quantifies the resources required to perform quantum data compression,
with the restriction that it be possible to recover the source with fidelity close to 1, in the
limit of large block lengths.
Shannon’s noiseless channel coding theorem tells us that the number of bits of
information necessary to represent, on average, each use of a classical source of
information is given by a function of the source probability distribution, called the
Shannon entropy. Similarly, Schumacher’s theorem introduces a new entropic quantity,
namely the von Neumann entropy, as the limit to which a quantum source may be
compressed. This new entropy can be interpreted as a generalization of Shannon’s entropy
to quantum states. The von Neumann entropy agrees with Shannon’s entropy if the states
produced by the quantum source are orthogonal, but in general it is strictly smaller than
Shannon’s entropy. This decrease in resources required to compress a quantum source is
possible exactly by exploiting the inherent redundancy arising in non-orthogonal quantum
states. A fully satisfactory analog of Shannon’s noisy channel coding theorem for
encoding and decoding quantum states traveling through a noisy quantum channel
has not yet been found. The capacity has been established, however, for some specific
channels, like the quantum erasure channel [23].
Entanglement is probably the most bizzare elementary static resource of quantum
mechanics. Its properties, which are yet to be well understood, are essentially different
from those of the resources most familiar from classical information theory. But these
strange properties of entanglement are also responsible for creating novel and surprising
possibilities of accomplishing information processing tasks. In the following we give two
examples of simple, yet unexpected applications of elementary quantum mechanics to the
problem of transmitting information between two parties, conventionally known as Alice
and Bob, who are a long way away from one another.
3.3 Superdense coding
In the first application, known as superdense coding [25], entanglement allows Alice to
communicate to Bob two bits of classical information by sending him only one qubit. To do
this though, pthey
ffiffiffi must previously share a pair of qubits in the entangled state
ðj00l þ k11jÞ= 2. Alice is initially in possession of the first qubit, while Bob has possession
of the second one. There are four different messages Alice can transmit over to Bob using
two bits of information: “00”, “01”, “10” or “11”. For each of these four bit strings, Alice
applies a specific quantum gate to her qubit and then sends it over to Bob (see figure 4). Now
Bob has hold of both qubits and is free to perform any kind of measurement on them.
Because the two qubits were entangled, any manipulation of the first qubit done by Alice
has inevitably affected the state of the ensemble, such thatpthe
ffiffiffi final state describing
pffiffiffi the qubits
is one of the four Bell states or EPR pairs: ðj00l ^ k11jÞ= 2, ðj01l ^ k10jÞ= 2. The catch is
that the Bell states form an orthonormal basis and can therefore be distinguished by an
appropriate quantum measurement. All that Bob has to do in order to determine which of the
four possible bit strings Alice sent is to perform a joint measurement on the two qubits in the
Bell basis.
In terms of the three goals of information theory stated at the beginning of this section, we
can identify the two qubits and the entanglement between them as the static resources
involved. As dynamical processes we can name the entanglement transformation performed
by Alice, the transmission of the qubit from Alice to Bob and the measurement performed by
Bob. What are the physical resources that have to be spent in order to complete the task? The
answer to this question depends ultimately on the physical realization of the joint
measurement. Theoretically, since the state of the 2-qubit ensemble is already perfectly
aligned with one of the measurement’s projectors, the act of measurement should not change
this state. The EPR pair should retain its state throughout the measurement process. As one
would expect, such a joint measurement is very difficult to implement in practice. Therefore,
assuming that Bob’s measurement collapses the entangled state, the resources consumed in
the process are one qubit worth of communication and the entanglement relating the two
qubits. Transmitting two bits of classical information through one qubit is only possible by
spending one e-bit of entanglement. Of course, the protocol requires two qubits, but Alice
Figure 4. Superdense coding.

need never interact with the second qubit. The initial, entangled state of the ensemble is a
fixed state, so there is no need for Alice to have sent Bob any qubits in order to prepare this
state. Instead, some third party (an “entanglement provider”) may prepare the entangled state
ahead of time, sending one of the qubits to Alice and the other to Bob.
3.3.1 Experimental demonstration. The superdense coding protocol has received partial
verification in the laboratory. An experiment performed at the University of Innsbruck in
Austria [135] implemented a variant of superdense coding using entangled photons and the
techniques of two-particle interferometry, which involve interference between pairs of
particles. The biggest problem was the joint measurement that Bob has to perform to
distinguish between the four Bell states. In the Innsbruck experiment, two of the possibilities
cannot be distinguished from one another, so Bob can only read three out of four possible
states for each photon Alice sends him. Therefore, although this experiment failed to achieve
the theoretical two classical bits per qubit, it nevertheless still managed more than one bit
(more precisely, a qubit was carrying one trit of information).
3.4 Quantum teleportation
Superdense coding is a protocol for transmitting classical information through a quantum

channel. We now move to the second example, quantum teleportation [20], in which Alice is
able to send Bob an unknown quantum state, using only a classical communications channel.
Furthermore, the transfer of hidden quantum information implied by the process appears to
happen without that state having to move through the intervening space, effectively
teleporting the quantum state to Bob. Again, this is only possible if Alice and Bob share an
EPR pair. So, initially, Alice has in her possession two qubits: the state to be teleported
jcl ¼ aj0l þ bj1l,pwhere
ffiffiffi a and b are unknown amplitudes, and her half of the entangled
pair ðj00l þ k11jÞ= 2, while Bob is in possession of the other half. To achieve teleportation,
Alice interacts the qubit jcl with her half of the EPR pair and then measures the two qubits in
her possession, obtaining one of four possible classical results: 00, 01, 10 or 11.
In the first step, Alice applies a controlled-NOT gate to her qubits (with jcl acting as the
control qubit) changing the initial state of the 3-qubit system
1
jC0 l ¼ pffiffiffi aj0lðj00l þ j11lÞ þ bj1ljiðj00l þ j11lÞ
2
to
1
jC1 l ¼ pffiffiffi aj0lðj00l þ j11lÞ þ bj1lðj10l þ j01lÞ :
2
In the expressions above, the first two qubits belong to Alice and the last one to Bob. The
EPR pair consists of the last two qubits.
Notice how this operation has transformed the initial entanglement into another form. No
two qubits in jC1l form a maximally entangled pair, but the entanglement now involves all
three qubits. We can say that the initial entanglement has diffused to the first qubit as well.
In the last step before measurement, Alice sends her first qubit through a Hadamard gate,
determining the system to evolve into the state
1
jC2 l ¼ aðj0l þ j1lÞðj00l þ j11lÞ þ bðj0l 2 j1lÞðj10l þ j01lÞ :
2
Since jC2l is a superposition of all possible 23 ¼ 8 classical states for the 3-qubit system,
we can intuitively assert that, in some sense, the entanglement dilution process initiated in the
first step is now complete. Does this mean that the teleportation was achieved? By regrouping
the terms composing the jC2l state, we can see that Bob’s qubit can only be in one of the
following four states: aj0l ^ bj1l, aj1l ^ bj0l, corresponding to the four possible outcomes
for Alice’s measurement. Tracing out Alice’s system, the reduced density operator of Bob’s
system can be shown to be
j0lk0j þ j1lk1j I
r Bob ¼ ¼ :
2 2
This state has no dependence upon the state jcl being teleported. Consequently, any
measurement Bob might perform on his qubit will contain no information about jcl. The
situation remains unchanged for Bob, even after Alice has completed her measurements†.
It is only when Bob is informed of the outcome of Alice’s measurement that he knows what
quantum transformation (rotation) to apply to his qubit in order to recover the state jcl
accurately, thus completing the teleportation.
Because Alice must transmit her measurement result to Bob over a classical
communications channel, which is by definition limited to the speed of light, quantum
teleportation cannot be used to transmit quantum states faster than light. The use of the
reduced density operator above makes this argument mathematically rigorous. On the other
hand, faster than light communication would have generated a series of paradoxes derived
from Einstein’s theory of relativity (like the possibility to send information backwards in
time, for instance).
A superficial glance at quantum teleportation may lead one to believe that the process
creates a copy of the quantum state being teleported, thus violating the no-cloning theorem
imposed by quantum mechanics. This violation is only apparent though, since after the
teleportation process only Bob’s qubit is left in the state jcl, while the original data qubit
ends up in one of the computational basis states j0l or j1l, following the measurement
performed by Alice on the first qubit.
The information-theoretic static resources making quantum teleportation possible are
three qubits (two of them entangled in an EPR pair) and two classical bits. The dynamic
processes acting on these resources are the various quantum transformations (gates) to which
the three qubits are subjected, Alice’s measurement of her two qubits and the two classical
bits worth of communication between Alice and Bob. As opposed to superdense coding, here
the measurements on Alice’s two qubits are carried out independently of one another, in the
computational basis. Therefore, we can affirm with certainty that teleporting an unknown
quantum state consumes one e-bit of entanglement, since there is no entanglement left in the
state of the 3-qubit system at the end of the process. From a broader perspective, quantum
teleportation emphasizes the interchangeability of different resources in quantum mechanics,
†
Note how the knowledge about Bob’s subsystem (qubit) is relative to the two observers, Alice and Bob.
showing that one e-bit of entanglement together with two classical bits of communication is a
resource at least the equal of one qubit of communication.
3.4.1 Experimental demonstrations. Quantum teleportation is not only a neat trick, but has
applications of practical importance, like designing quantum gates that are resistant to the
effects of noise. Practical verifications of quantum teleportation have been experimentally
realized in various different forms by several teams [11,34,35,86,147]. Zeilinger and his
colleagues at Innsbruck [35], for example, used a variation of their dense coding scheme to
teleport the polarization state of a photon. It may be worthwhile to also mention the original
idea on which Francesco De Martini and colleagues at the University of Rome [35] built their
experiment. Instead of measuring two photons, they chose to measure two different aspects
of one photon. One aspect was the polarization and the other was the choice between
traveling along one of two different routes. As a matter of fact, the entanglement in the initial
EPR pair was also in the choice of paths. An interesting novelty was brought by the most
recent experiment [11], in which electrons (and not photons) were used to embody the
teleported quantum state.
In both examples that we have described in this section, entanglement plays a crucial role
in the successful completion of the respective information processing tasks. The importance
of entanglement for quantum computation and quantum information has motivated the
development of a fascinating area of study about the creation and transformation of
entanglement.
4. Quantum cryptography
If one day, the impressive potential arising from the application of quantum mechanical
principles to computer science will be achieved and building a practical quantum computer
will no longer be a distant goal, perhaps the most spectacular impact will be seen on
cryptography. The consequences on the security of cryptographic techniques and protocols
can be both destructive and constructive. It is the aim of this section to show these effects,
together with the current status of quantum cryptography.
4.1 Private-key systems
Historically, the first methods used to exchange secret information belong to the class of
private-key systems. The two parties wishing to communicate have to meet and agree on a set
of secret keys, subsequently used to encrypt and decrypt any messages between them. The
one-time pad cryptosystem is an example from this category. Its security rests on several
assumptions: each key is used only once, nobody is eavesdropping while the keys are
randomly generated and agreed upon and finally, each of the two participants in the protocol
is able to securely store its copy of the pad containing the secret keys. If these requirements
are met, then the system is guaranteed to be totally foolproof. However, there is one major
drawback all private-key systems share: the distribution of keys. While meeting face to face
in order to distribute keys is not unconceivable for inter-governmental communications, it is
certainly out of the question for commercial transactions over the Internet, for example.
4.2 Public-key systems
Public-key systems were invented exactly to address the problem of securely distributing
cryptographic keys, without the communicating parties having to meet. In contrast to the
symmetry of private-key systems, in which the same key is used for encryption and
decryption, public-key systems use a pair of asymmetric keys. One is public, and therefore
can be widely distributed, and the other is private and must be kept secret. A message
encrypted using the public key can only be decrypted using the associated private key. This is
achieved using the mathematical concept of a trapdoor function. A trapdoor function is some
mathematical procedure that is easy to compute, yet very hard to invert (for this reason they
are sometimes called one-way functions), unless you have access to a special key that can
unlock the trapdoor. Computing a trapdoor function corresponds to encrypting a message
using the public key. To break such a code, one is forced to invert the trapdoor function
without knowledge of the private key, while the intended recipient of the encoded message
can use his or her private key in order to easily decipher the encrypted message.
4.2.1 Diffie –Hellman algorithm. The first practical public-key algorithm was devised by
Diffie and Hellman in 1976 [69]. It was a continuation of Ralph Merkle’s work on this track
[138]. The Diffie – Hellman procedure does not completely depart from the symmetry of
private-key systems. Alice and Bob, the prototypical participants in any cryptographic
protocol, exchange their public keys in order to construct (based also on their secret keys) a
common session key, which they can use to encode and decode messages between them, just
like in a normal private-key system. From this point of view, one can look at the Diffie –
Hellman procedure as a safe way of distributing private keys using only public
communications channels. Reconstructing one of the secret keys, given knowledge of the
corresponding public key, amounts to solving a discrete logarithm problem, which is very
hard to calculate if the numbers involved are all sufficiently large. The security of the Diffie –
Hellman algorithm depends on this fact.
4.2.2 RSA algorithm. But the most successful public-key system in use today is based on
the RSA algorithm developed in 1977 [156]. Unlike the Diffie – Hellman algorithm, it is a
genuine public-key system, in the sense that anyone can send an encrypted message to Alice
using only her public key. Alice will then use her private (secret) key in order to decode the
message. The RSA algorithm is also more versatile, offering the possibility to digitally sign a
document due to the interchangeability between the public and private keys in this algorithm.
Its security ultimately rests on the difficulty of factoring large numbers in a reasonable
amount of time, a problem thought to be intractable, although nobody was able to prove
that it is not in P. On the other hand, nobody was able to prove it as an NP-complete
problem either.
The lack of a polynomial time solution to the factoring problem means that RSA
encryption codes are safe for the time being, provided long enough keys are used. More
precisely, a 129-digit number used as the public key to encrypt a message by applying the
RSA algorithm was already factored in 1994. However, this result was possible only after a
gargantuan computational effort that lasted about eight months and involved some 1600
computers distributed over the Internet. Improvements in factoring technology made possible
a much quicker factorization of a 130-digit RSA key two years later. Replacing the quadratic
sieve with the number field sieve as the factorization method sped-up the computation
approximately by a factor of 5. The number field sieve [126] is currently the best-known
classical algorithm for factoring large integers. Its running time, while still super-
polynomial, is sub-exponential in the size (number of digits) of the number being factored.
4.3 Cracking RSA codes

There are two useful remarks we can make here. First, the power of parallelism should not be
underestimated. Lenstra, who was the first together with several colleagues to factor the ninth
9
Fermat number 22 þ 1 (155 decimal digits) and was involved in cracking both RSA-129 and
RSA-130, thinks that RSA codes depending on 512-bit numbers (corresponding to 155
decimal digits) are within the reach of cypherpunks, if they could get hundreds of thousands
of machines computing for them (see [42] page 167).
Secondly, since the security of RSA codes is based upon unproven assumptions,
advancements in factoring algorithms are always possible. Coupled with technology
improvements that allow ever faster computers to be built, this will force people to use longer
and longer keys in order to keep the RSA systems secure. However, unless a polynomial time
algorithm is discovered, factoring a 2000-digit number would take more than the entire life
of the Universe, assuming that every particle in the Universe was a classical computer
working at full speed (see [181] page 129). Even messages encrypted using a 400-digit key
are well beyond the decryption capabilities of modern computers.
The cryptographic landscape, as it is perceived today, would however change dramatically
if large-scale quantum computers could be built. This is because the algorithms discovered
by Peter Shor in 1994 for solving the discrete logarithm problem and factoring numbers
require only polynomial time on a quantum computer. Recall that the security of Diffie-
Hellman procedure and RSA algorithm respectively, rely exactly on the presumptive
intractability of these two problems on a classical machine. Given the current spread of
public-key systems (especially RSA) and the increasing role they are expected to play as
electronic commerce over the Internet will grow, the feasibility of a quantum computing
device capable of executing Shor’s algorithms would deliver a devastating blow to public-key
cryptography. We will sketch the main features of Shor’s factorization algorithm in the next
section, in the more general context of quantum algorithms.
It appears that computing devices with quantum mechanical capabilities would render
public-key systems useless, bringing the problem of distributing cryptographic keys through
insecure communications channels back into attention. It turns out that the principles of
quantum mechanics, which reactivated the problem in the first place, are also able to solve it.
The answer comes from quantum cryptography, a discipline that has flourished initially
independently of quantum computing, but now as an integral part of it, since both depend on
the same technology and the same conceptual foundations.
4.4 Quantum key distribution

The birth of quantum cryptography may be assimilated with Stephen Wiesner’s ideas in the
late 1960s to use quantum mechanics in order to produce unforgeable money [180]. Although
not very practical, his ideas stimulated other researchers and eventually Bennett and Brassard
[18] were able to develop a protocol (generally referred to as BB84) for implementing a
system of quantum key distribution. Their method employed quantum mechanical principles
to exchange secret keys securely over public channels. So, unlike the public-key systems in
use today, the unbreakability of quantum key distribution seems to be guaranteed by the very
laws of physics (quantum mechanics, in this case).
The quantum key distribution scheme is amenable to any kind of physical realization, but
photon polarizations offer a convenient way to explain and implement it. Suppose Alice and
Bob wish to construct a key consisting of a random sequence of bits, known only to them,
that they will subsequently use to encrypt and decrypt messages between them, using any
private-key algorithm. Alice chooses a random set of bits that she encodes in either the
rectilinear polarization (horizontal/vertical) or the diagonal polarization (458/1358).
The choice of polarization orientation must be random too (see figure 5). Alice then sends
the stream of photons (each carrying a bit of information) to Bob over an open quantum
communication channel. Upon receipt of the photons, it is Bob’s turn to choose an orientation
for the device used to measure the direction of polarization of the incoming photons.
According to the choices made between rectilinear and diagonal polarization measurement
for each photon, Bob will extract a certain set of bits from the observed photons. In the last
step of their protocol, Alice and Bob enter into a public communication. Alice divulges the
polarizer orientations she used to encode the bits, while Bob informs Alice about the
measurement basis he used to decode each photon received. Those bits for which both have
chosen the same orientation will form their shared secret key. The others will simply be
discarded. Note that the exact sequence of bits exchanged between Alice and Bob is
irrelevant, as in the end they, and only they, come to learn the identity of a common subset of
the bits, without having to reveal them to each other or to the outside world.
The above scenario is only possible if the encoding, decoding and transmission steps are
error-free, and perhaps more important, there is no eavesdropping. If Eve, the prototypical
eavesdropper, intercepts the photons on their way from Alice to Bob and wishes to gain some
information about the secret key, she is forced to measure them, just as Bob would do.
Unfortunately for her, the principle of quantum mechanics regarding measurements is not on
her side and the quantum state of those photons for which Eve chooses an incorrect
measurement basis will be inevitably disturbed. After discarding the irrelevant bits, Alice and
Bob can randomly test some of the remaining bits to reveal the presence of any potential
eavesdropper. For each bit tested, the probability of that test revealing Eve’s presence is 1/4.
Thus, the probability of detecting eavesdropping can be made arbitrarily close to 1, by testing
a sufficiently large number of bits. Since the tested bits can no longer be part of the key, there
is a trade-off here between the desired level of security and the amount of quantum and
classical communication required between Alice and Bob. Bennett and Brassard have
addressed this issue by proposing alternate methods of testing, like parity checking, which
leaves open the possibility for the set of bits involved in the test to still be included in the
private key. Quantum key distribution is a good example of a case where quantum
Figure 5. Quantum key distribution in the absence of eavesdropping.

mechanical postulates seeming very restrictive and imposing severe limitations (like
measurement and no-clonability, in this case) can actually be used in a constructive way and
have important practical applications.
4.4.1 Experimental demonstrations. The first practical demonstration of their quantum

key distribution protocol was performed by Bennett and Brassard in 1989 [19] over a
distance of 32 cm. In spite of the insignificant distance, the experiment was very important as
it showed that quantum cryptography is viable, but it also pointed out some difficulties.
Trying to hide her presence behind the noise in the quantum channel and the imperfections of
Bob’s photon detectors, a clever eavesdropper could choose to measure only some of the
passing photons, according to the idea that partial information about the key is better than no
information. To cope with such low levels of eavesdropping, Bennett et al. [22] have
proposed the method of privacy amplification, a mathematical technique based on the
principle of hashing functions that magnifies Eve’s uncertainty over the final form of the key.
There is also an efficient way of applying privacy amplification to the problem of reducing
string oblivious transfer to bit oblivious transfer [39].
The ensuing rapid progress in quantum cryptography led to several experimental
demonstrations of quantum key distribution over tens of kilometers of fiber-optic cable
[84,105,106,108,133,142]. Townsend and Marand at British Telecom [133] have even
managed to convey their cryptographic signals alongside simulated telephone traffic carried
on the same fiber-optic cables. Taking things one step closer to reality, Townsend also
demonstrated how quantum keys could be securely distributed to many different receivers
over a passive fiber-optic network. Also notable is the error-cancellation technique used by
the team from the University of Geneva [142].
However, using optical fibers as the transmission medium for photons has its limitations.
Because the error rate increases with the length of the optical cable, quantum key distribution
beyond 100 km is inconceivable today†. One proposed solution to the range limitation is to
establish a chain of keys between Alice and Bob, but this may leave the signals vulnerable at
the repeater stations. The distance limitation of cryptography could be lifted completely if
genuine quantum repeater stations based on teleporting quantum states could eventually be
built.
Finally, another exciting possibility currently under exploration is to perform quantum key
distribution directly through the atmosphere using lasers and satellites, with no need for
special optical fibers. The main problem in this case is to combat the disturbances caused by
atmospheric turbulences and background radiation. The hope is that from a large number of
raw bits exchanged, a reasonable number of error-free key bits could be distilled. Although
free-space experiments have got further to go before they catch up with the fiber-optic
demonstrations, progress along this path has been achieved too. From 30 cm initially
demonstrated by Bennett and collaborators in 1991 [17], gradual improvements have made
possible free-space quantum key distribution over 10 km during daylight and at night [109].
Thus, the remarkable experimental progress shown in quantum key distribution imposes
quantum cryptography as the most promising practical application of quantum information
theory, at least in the short term. Even at the current stage of development, quantum key
†
Note that conventional repeaters cannot be used due to the no-cloning theorem.
distribution schemes using fiber-optic technology are sufficiently advanced to “wire the
financial district of any major city” ([182] page 155). There is reason to believe that it will not
take long before such technologies will become commercially viable.
4.5 Quantum solutions to discreet decision problems

So far we have seen how quantum key distribution in conjunction with a guaranteed secure
classical cryptosystem (like the one-time pad, for instance) make practical classical
communications as secure as our current knowledge of physics allows. An important question
is whether quantum cryptography can have other applications beside quantum key distribution.
Classical cryptography is able to offer solutions to a wide variety of situations classified as
discreet decision problems. In these situations, discretion is vital to achieving agreements.
Examples include negotiating arms treaties, forming business partnerships and organizing
mergers. These applications are amenable to classical cryptographic solutions involving public-
key systems, but as we have seen, they are based on unproven assumptions about the difficulty of
factoring large numbers and other related problems. What quantum cryptographers were looking
for was a totally secure system, guaranteed by the laws of physics.
Research in classical cryptography has shown that solutions to such discreet decision
problems can be constructed from basic cryptographic building blocks, like the notion of
oblivious transfer [154] or 1-out-of-2 oblivious transfer, an idea foreseen in a quantum guise
by Stephen Wiesner [180]. More recently, Claude Crépeau and Joe Kilian have demonstrated
that oblivious transfer can be used as a building block for solving two-party problems
requiring discretion [58]. In turn, to provide totally secure quantum oblivious transfer, one
would need a secure form of bit commitment. Consequently, much of the research effort in
quantum cryptography in the early 1990’s was devoted to finding a protocol for quantum bit
commitment that is absolutely and provably secure. That result (known as BCJL after the
authors’ names) was reported in 1993 [38] and became the foundation for numerous
applications in quantum cryptography, pertaining to discreet decision making.
The surprise came in 1995 when Dominic Mayers discovered how Alice could cheat in the
BCJL bit commitment protocol by using EPR particles [136]. Furthermore, he proved [137] that
it would be possible for Alice to cheat in any protocol for quantum bit commitment. An intuitive
explanation is that the information she sends Bob describing the safe for her bit must give nothing
away about the committed bit. Consequently, regardless of the particular bit commitment
scheme employed, the quantum states of the safe containing either 0 or 1 must be very similar (if
not identical) since otherwise Bob would be able to discern the difference and gain knowledge
about the committed bit prematurely. But the very fact that the two states are virtually the same
gives Alice the possibility to keep her options open and postpone her commitment for later on.
Although in their 1996 review paper of quantum cryptography, Brassard and Crépeau [37]
argued that for the time being the practical implications of the flaw discovered in the quantum bit
commitment protocol are minimal, the weakness definitely affected the entire edifice of quantum
cryptography built upon quantum bit commitment.
4.6 Entanglement-based cryptography
However, analog to the two-fold influence of quantum mechanics upon cryptography,

entanglement can also be used in a beneficial way, enhancing methods of key distribution.
Inspired by EPR experiments designed to test Bell’s inequality, Artur Ekert thought of a way
of using entangled pairs for distributing cryptographic keys by quantum means [75]. In his
scheme, Alice and Bob receive entangled particles from a central source and perform
independent measurements upon them. For each measurement, the orientation is chosen at
random among three possibilities. The presence of any potential eavesdropper can be
revealed when Alice and Bob publicly confront the results they got for the measurements in
which they adopted different orientations. If the original EPR pairs were untampered with,
then the strength of their correlations must exceed anything that is possible classically. These
correlations will be disrupted if someone attempts to make measurements on the particles,
before they arrive at the legitimate receiver.
So the security of Ekert’s quantum key distribution scheme rests exactly on the dismissal
of the hidden-variable theory. Eve’s only hope is that entangled particles might carry hidden
information about how they will behave in subsequent measurements. But she cannot elicit
any information from the transiting particles simply because there is no information encoded
there. The information about the secret key has yet to come into being, once Alice and Bob
perform their measurements.
Ekert’s entanglement-based scheme also offers a couple of potential advantages over the
original single photon protocol invented by Bennett and Brassard. The first refers to the
possibility of storing cryptographic keys securely, while the second involves the issue of
privacy amplification. The limitations of the classical privacy amplification based on hashing
algorithms are overcome in the quantum privacy amplification technique, developed in 1996
by a group of researchers including Deutsch et al. [66]. The quantum procedure, which is
applicable only to entanglement-based quantum cryptography, can be repeatedly applied to
impurely entangled particles to cleanse them of any signs of tampering by Eve. The
entanglement purification process is actually an extension of previously published work [21]
and needs only some simple quantum logic.
However, these advantages of entanglement-based cryptography are rather theoretical at
the moment because storing entangled particles is only possible for a fraction of a second as
yet, and entanglement purification depends on quantum computational hardware that,
although simple, has yet to be built. This, of course, assuming that entanglement is indeed a
real physical resource that can be harnessed for our computation or communication purposes.
Although Ekert and Rarity [77] devised a plan for implementing a practical method of
entanglement-based cryptography and Nicholas Gisin’s team in Geneva reported an
experimental demonstration of quantum-correlations over more than 10 km [175], these
results are nowhere near the remarkable progress achieved by quantum key distributions
using the original BB84 protocol, which is well within the capabilities of current technology.
Work has also been carried out to develop efficient cryptographic protocols based on noisy
channels [57] and, more recently, to comparatively assess the potentials of classical and
quantum protocols for key agreement in various situations [88 –90].
5. Quantum algorithms
Quantum computing is a fundamentally novel paradigm with respect to both the “hardware” and
the “software” employed. The hardware in a quantum computer concerns the physical
realization of the qubits organized into a memory register and the quantum logic gates acting on
them. Designing a quantum algorithm to solve a certain problem boils down to the choice of the
quantum gates, that is unitary transformations, which, when chained together in a proper way,
evolve the initial state of the quantum memory register into the desired final state. Thus, we can
say that the LOAD–RUN–READ operational cycle characteristic to a classical computer is
replaced by a PREPARE–EVOLVE–MEASURE cycle in a quantum computer.
Quantum parallelism: The “programming techniques” employed to achieve the desired
evolution are essentially different from their classical counterparts. This is due to the
necessity of playing by the rules imposed by quantum mechanics, which severely restrict the
applicability of classical algorithmic design methods. The main feature of a quantum
computer, used in any quantum algorithm, is quantum parallelism. Quantum parallelism
refers to the capability of a quantum computer to evaluate a function f(x) for exponentially
many different values of x in the time it takes a classical computer to evaluate the function for
just one value. This is possible by loading the memory register with a superposition of all
possible input values x and then apply the appropriate unitary transformation that will evolve
this state into a superposition of all function values f(x). The enormous potential of a quantum
computer to outperform a classical machine lies in the massive parallelism it offers “within a
single piece of hardware”, as Berthiaume and Brassard put it [28].
Extracting the answer: However, a direct observation of the quantum memory register will not
yield more information than is possible to obtain using a classical computer. Any measurement
attempt will collapse the superposition and reveal only one of the answers, without even knowing
which one beforehand. Therefore, quantum parallelism alone does not justify all the trouble of
going into quantum computing. Additional techniques seem necessary in order to exploit
quantum parallelism and make a quantum computer useful. Fortunately, quantum mechanics has
also the resources to achieve this task: interference and entanglement. The heart of any quantum
algorithm is the way in which it manipulates entanglement and interference between various
computational paths, so that the final measurement will yield desired results with high
probability. Generally, the design of a quantum algorithm is focused on how interference can
constructively recombine different alternatives in a superposition to strengthen the amplitude
of solutions, while non-solutions will interfere destructively, canceling each other. According
to the specific technique employed to achieve this sort of manipulation as well as their area of
applicability, we can identify three main classes of quantum algorithms which provide
an advantage over known classical algorithms: quantum algorithms based upon some kind of
Fourier transform, quantum search algorithms and quantum simulations.
5.1 Algorithms using quantum Fourier transforms

The discrete Fourier transform can be brought about in a quantum mechanical context by
defining a linear transformation on n qubits whose action on the computational basis states
jjl, 0 # j # 2n 2 1, is described below:
n
QFT 1 X 2pijk=2n
2 21
jjl ! pffiffiffiffiffi e jkl: ð8Þ
2n j¼0
Expressing the effect of applying this Quantum Fourier Transform on a superposition of

the computational basis states, in the form
n
2X 2n 21
QFT X
21
xj jjl ! yk jkl; ð9Þ
j¼0 k¼0
corresponds to a vector notation for the usual discrete Fourier transform, in which the
complex numbers xj are transformed into the complex numbers yk.
5.1.1 Deutsch’s algorithm. Under various forms, quantum Fourier transforms play an
essential role in many algorithms, bringing about the necessary interference in order to find
common properties of all the values in a superposition without having to reveal any of the
individual values explicitly. The first algorithm to exploit this idea was devised by Deutsch as an
example of how quantum parallelism backed by interference can “beat” a classical computer
[61]. Given a function f:{0,1} ! {0,1} Deutsch presented a quantum algorithm able to compute
f(0) % f(1) in a single evaluation of the function f. Beside computing f(0) and f(1) in quantum
parallel using a superposition of the two inputs, Deutsch was able to encode the value of the
function f(x) in a phase factor (21) f(x). This, in turn, enabled a quantum mechanical interference
between the phase factors to reveal the desired joint property of f(0) and f(1). It is important to
note that Deutsch’s algorithm can only answer the question whether f(0) equals f(1) or not,
without giving any information about either function value individually.
5.1.2 The Deutsch – Jozsa algorithm. Deutsch and Jozsa [67] generalized this problem to
the n-bit case, allowing the domain of f to be composed of all integers in the interval
[0,2n 2 1] and determining whether f is constant for all values of x or perfectly balanced
between 0 and 1 in just one evaluation of the function f. The quantum circuit implementing
the general Deutsch –Jozsa algorithm is depicted in figure 6. The Walsh – Hadamard
transform H ^n is a generalization of the single-qubit Hadamard gate H and corresponds to n
Hadamard gates acting in parallel on n qubits. When applied to n qubits, all prepared in the
j0l state, the Walsh –Hadamard gate creates an equally weighted superposition of all integers
in the range [0,2n 2 1]. This makes it widely used in the initial step of virtually any quantum
algorithm. The papplication
ffiffiffi of the Hadamard gate on the single qubit j1l gives the
superposition 1= 2ðj0l 2 j1lÞ, responsible for encoding the results of the function evaluation
in the amplitude of the corresponding term in the superposition state jCl:
X ð21Þf ðx1 ...xn Þ jx1 . . .xn l j0l 2 j1l
jCl ¼ pffiffiffiffiffi pffiffiffi : ð10Þ
x1 ...xn 2n 2
In the final step of the computation, another Walsh – Hadamard gate acts on the first n
qubits interfering the terms in the superposition:
X X ð21Þx1 z1 þ· · ·þxn zn þf ðx1 ...xn Þ jz1 . . .zn l j0l 2 j1l

0
jC l ¼ pffiffiffi : ð11Þ
z1 ...zn x1 ...xn
2n 2
Figure 6. Quantum circuit implementing the Deutsch–Jozsa algorithm.

A closer look to the amplitude for the state j0l^n, namely

X ð21Þf ðxÞ
x[{0;1}n
2n
shows that it is possible to discern whether f is constant or balanced by measuring the first n
qubits. If f is constant, then the above amplitude is þ 1 or 2 1 depending on the constant value
f(x) takes. However, in either case, all the other amplitudes must be zero because jC0 l is of
unit length, by definition. Therefore, if the measured qubits are all 0, we know that f must be
constant. On the other hand, when f is balanced, the positive and negative contributions to the
amplitude for j0l^n cancel each other, giving an amplitude of zero. This means that at least
one of the measured qubits must yield a result of 1.
The original algorithms for Deutsch’s problem and the Deutsch– Jozsa problem have been
substantially improved subsequently by Cleve et al. [54]. The original algorithm of Deutsch
only worked probabilistically, giving a meaningful answer only half the time.
The presentation above is based on the improved versions of these algorithms.
5.1.3 Instances of quantum Fourier transforms. Other researchers tried to extend the
work of Deutsch and Jozsa by developing variants of their balanced vs. constant problem.
Berthiaume and Brassard, for instance, studied the original Deutsch –Jozsa problem in an
oracle setting [28]. Dan Simon developed a variant that showed how a quantum computer
could efficiently decide whether a function is 1-to-1 or 2-to-1 [165]. Although only periodic
2-to-1 functions were allowed, the interesting thing about Simon’s algorithm was that for
such functions it was able to pluck out their periodicity by exploiting a simple kind of Fourier
transform. But the first paper to explicitly show how Fourier transforms can be implemented
on a quantum computer belonged to Bernstein and Vazirani. In their paper about quantum
complexity theory [26] they also showed how to sample from the Fourier spectrum of a
Boolean function on n bits in polynomial time on a quantum Turing Machine (QTM).
Nevertheless, as it was later realized, the essence of the Fourier transform idea was actually
already hidden within the Deutsch –Jozsa quantum algorithm solving the constant vs.
balanced problem. Indeed, the Walsh– Hadamard transformation applied in the last step of
the algorithm can be intuitively assimilated with a quantum Fourier transformation which
generates an interference between all computational paths.
5.1.4 Shor’s factorization algorithm. All these developments progressively enlarged the
scope of what was possible using quantum algorithms, paving the way for the most
spectacular result as yet, namely Peter Shor’s algorithm for factoring large integers and
computing discrete logarithms efficiently on a quantum computer [164]. Of great inspiration
to Shor was the work of Dan Simon [165] and Bernstein and Vazirani [26]. As a number
theorist, Peter Shor was well aware of the relation between factoring a number n and
calculating the orders (or periods) of numbers modulo n. He was hoping to build upon
Simon’s paper about finding periodicities in 2-to-1 functions and devise an efficient method
to compute the orders for functions of the form fx,n(a) ¼ x a mod n. In classical complexity
theory it has been long known that finding such orders when the modulus n gets very large is
as hard as factoring n. In the same paper that offered a polynomial-time algorithm for
primality testing, Gary Miller showed that the problem of finding orders is computationally
equivalent to the problem of factoring integers [140]. But Shor was counting on the quantum
Fourier transform to find orders efficiently using quantum mechanics. Once an even number
period r is found for x a mod n by varying the value of x, x r/2 2 1 and x r/2 þ 1, respectively,
have a good chance of sharing a common divisor with n. A sketch of Shor’s quantum
algorithm for factoring integers is given below.
1. Set up a quantum memory register partitioned into Register 1 and 2. Load Register 1 with
a superposition of all possible integers by means of a Walsh – Hadamard gate. Load
Register 2 with zeros.
2. Pick a random integer x which is coprime with n and evaluate the function x a mod n, in
quantum parallel, for all terms in the superposition, such that each number a in Register 1
is entangled with the corresponding result, placed in Register 2.
3. Observe the state of Register 2 to be some integer k. Because of the entanglement between
the two registers, this measurement will also have the effect of projecting out the state of
Register 1 to be a superposition of just those values of a such that x a mod n ¼ k.
4. Compute the Fourier transform of the post-measurement state of Register 1.
5. Measure the state of Register 1 to sample from the Fourier transform and obtain some
multiple of 2q/r, where q is the number of qubits in Register 1. Use a continued fraction
technique to determine the period r.
6. Obtain the factors of n by computing gcd(x r/2 2 1, n) and gcd(x r/2 þ 1, n).
Analysis: Note that Shor’s algorithm is probabilistic. Several things can go wrong: the
period of f(a) ¼ x a mod n is odd, the value sampled from the Fourier transform is not closed
enough to a multiple of 2q/r, the continued-fraction expansion yields a factor of r and not the
period itself, x r/2 2 1 or x r/2 þ 1 is a multiple of n. Nevertheless, Shor showed that few
repetitions of this algorithm reveal a factor of n with high probability, thus providing a
bounded probability polynomial time algorithm for factoring numbers on a quantum
computer. A condition for keeping the complexity of the algorithm at a polynomial level was
to devise an efficient quantum implementation for the Fourier transform in step 4. In his
paper, Shor describes a way of constructing the quantum Fourier transform with base 2m
using only m(m þ 1)/2 gates. Subsequent work by Adriano Barenco and colleagues at
Oxford [9] showed that the circuit required for computing the QFT can be simplified by
making certain approximations, thus even becoming computationally less demanding than
the procedure for calculating the powers stored in the second register in step 2.
There is also an important observation that has to be made about step 3 in the algorithm
described above. The possibility of a measurement made on the contents of the second
register before carrying out the Fourier transform was actually hinted at in a later paper [8].
In Shor’s original paper, he suggested making the measurement only after the Fourier
transform on the first register. This may look quite intriguing from a classical computational
point of view, since the values of the function f(a) ¼ x a mod n computed in Register 2 seem
not to be referred to again. This apparent redundancy illustrates very well the differences
between conventional and quantum mechanical programming techniques and highlights
some of the subtleties of working with entanglement in superpositions. The entanglement
between Register 1 and 2 ensures that only those amplitudes corresponding to numbers a
having the same image f(a) will be able to interfere when the Fourier transform on Register 1
is invoked. The result will be a superposition of Fourier transforms of a set of functions
having the same period r, but which do not interfere with each other. Therefore, the
measurement in step 3 can be skipped entirely. More generally, Bernstein and Vazirani [26]
showed that measurements in the middle of an algorithm can always be avoided.
Thus, Shor’s quantum algorithm for factoring integers relies on quantum parallelism to
create a superposition of values of the periodic function fx,n(a), relies on entanglement and
the Quantum Fourier Transform to create the desired interference effect between solutions
(integer multiples of 1/r) and non-solutions (numbers that are not integer multiples of 1/r)
and finally relies on measurement to project out a multiple of the inverse of the sought-after
period r. The rest are techniques from classical number theory. Another useful application of
the quantum fast Fourier transform was found by Abrams and Lloyd [1]. They managed to
devise a new polynomial time quantum algorithm that finds eigenvalues and eigenvectors of
certain matrices for which all known classical algorithms require exponential time.
Generalization: The usefulness of quantum Fourier transforms has prompted the
development of a generalized theory of quantum Fourier transforms involving some
technical ideas from the theory of finite groups. This line of thought culminated in Kitaev’s
discovery of a method to solve the Abelian stabilizer problem [118] and the generalization to
the hidden subgroup problem [98]. The Deutsch –Jozsa algorithm, Shor’s algorithms and
related exponentially fast quantum algorithms can all be viewed as special cases of this
algorithm. It would be interesting to know whether other problems of practical
importance can be accommodated to fit this general framework nicely.
Recently, steps have been taken to develop new techniques that are not based on quantum
Fourier transforms, but still provide an exponential algorithmic speed-up relative to a
classical computer. Childs et al. [47] show how the quantum walk model proposed in [79]
can be exploited to construct an oracular (i.e. black box) problem in which a graph whose
structure is provided in the form of an oracle can be traversed exponentially faster than is
possible by any classical algorithm.
5.2 Quantum search algorithms

Another methodology to construct useful quantum algorithms is based on the idea of
transforming a superposition quantum state in such a way as to amplify the values of interest
at the expense of the amplitudes of the other terms in the superposition. This will give the
solutions a higher probability of being measured in the end. The most representative
exponent of this class is Grover’s unstructured search algorithm [94].
5.2.1 Grover’s search algorithm. Formally, the unstructured search problem can be defined
as finding some x in a set of possible solutions such that a certain statement P(x) is true.
In addition, no assumption is to be used about the structure of the search space and the
statement P. As expected, Grover’s algorithm relies on the same quantum mechanical
principles that give quantum algorithms in general the upper hand over corresponding
classical algorithms. We refer to quantum parallelism, entanglement and interference, as can
be seen from the description of Grover’s algorithm given below.
1. Prepare two quantum registers, the first containing a superposition of all possible input
values xi [ [0. . .2n 2 1] and the second one set to 0.
2. Compute P(xi) in quantum parallel for all the values in the first register, creating a
superposition of results in the second one, which is now entangled with the first.
pffiffiffiffiffi
3. Repeat approximately ðp=4 2n Þ times
3.1. Change the sign of the amplitude for the state xj such that P(xj) ¼ 1.
3.2. Invert all the amplitudes about the average. This will increase the amplitude of the
target state, while all the other amplitudes will be diminished imperceptibly.
4. Read the result. If Register 2 is 1 (the value representing True), Register 1 will contain the
sought-after value x.
Analysis: Clearly, the most important step in Grover’s algorithm is the amplitude
amplification operation performed in step 3. At the first glance, it might look surprising how
we can change the sign only for the target state without knowing it beforehand. However, if
we evaluate the predicate P by means of a gate array UP performing the transformation
jx,bl ! jx,b%P(x)l, we can apply UP to the superposition
n
1 2X 21
pffiffiffiffiffi jxl
2n x¼0
pffiffiffi
and choose b ¼ 1= 2ðj0l 2 j1lÞ. This way, we will end up in a state where the sign of all x
with P(x) ¼ 1 has been changed and b is unchanged. Grover also showed that the inversion
about the average can be accomplished efficiently by decomposing it into O(n) elementary
quantum gates with the aid of the Walsh –Hadamard transform.
Grover’s algorithm is another good opportunity to point out the fundamental differences
between classical and quantum programming techniques. Many classical algorithms rely on
indefinite repetitions of the same procedure to keep improving the results. In contrast, repeating a
quantum procedure may improve results only up to a certain point, after which the results will get
worse again. The reason is that quantum procedures are unitary transformations, which can be
interpreted as rotations of the vector state in a complex space. Thus, the repeated application of a
quantum transform may rotate the initial state closer and closer to the target state, eventually
getting past it farther and farther away. Therefore, the optimal number of iterations of a quantum
procedure is an important design issue. Boyer et al. [36], who provide a detailed pffiffiffiffiffianalysis of
Grover’s algorithm, show that for a single solution x0 such that P(xp0)ffiffiffiffiisffi true, p=8ð 2n Þ iterations
pffiffiffiffiffi p=8ð 2 Þ iterations, the failure rate
of step 3 above will bring the failure rate to 0.5. After another n
n
will drop to 1/2 . If we keep iterating for another p=4ð 2n Þ times though, the final measurement
is almost guaranteed to give us a non-solution.
The number of repetitions of step 3 also determines the complexity of Grover’s algorithm.
It follows that on a quantum computer, the unstructured search problem can be solved with
pffiffiffiffi
bounded probability of error within Oð N Þ evaluations of P, where N is the size of the search
space. This represents only a quadratic speed-up relative to the best possible classical algorithm,
which is far less impressive than the exponential speed-up achieved by Shor’s factoring
algorithm, for example. Still, the importance of quantum search algorithms is justified by the fact
that searching heuristics have a wider range of applications than problems solved using the
quantum Fourier transform, and if adaptations of the quantum search algorithm are also taken
into account, the range of problems that can benefit from them is even broader.
Extensions: Biron et al. [31] showed how Grover’s search technique can be used with pffiffiffiffi
arbitrary initial amplitude distributions, while still maintaining the overall Oð N Þ
complexity. This means that Grover’s algorithm can be used as a subroutine in other
quantum computations. Grover’s algorithm has also been combined with Shor’s algorithm in
order to perform quantum counting, that is to determine the number of solutions and the
optimal number of iterations [40]. Thus,ffi if the search space contains S solutions, then a
pffiffiffiffiffiffiffiffi
quantum computer takes only Oð N=SÞ steps to find one of them. Grover himself extended
his algorithm in two ways: to show that certain search problems that classically run in
O(log N) can be solved in constant time on a quantum computer, and to achieve quadratic
speed-up for other non-search problems such as computing the mean and median of a
function [95]. The intrinsic robustness of Grover’s quantum search algorithm in the presence
of noise is evaluated in [149].
5.2.2 Heuristic quantum search algorithms. For completely unstructured searches,

Grover’s algorithm is optimal [16,36,186]. However, there are also search problems, like
constraint satisfaction problems such as 3-SAT or graph colorability, in which information
about the search space and statement P can be exploited for heuristic algorithms that yield
efficient solutions for some problem instances. The techniques of Grover’s search algorithm
can be used to construct quantum analogs with quadratic speed-up to classical heuristic
searches [40,46]. But there is also hope that for certain structured problems a speed-up
greater than quadratic is possible by using specific properties of quantum computation to
exploit problem structure and not just merely construct quantum implementations of the
classical algorithms. The work of Tad Hogg is representative for this approach [99 –102].
He has developed heuristic quantum search algorithms that exploit the inherent structure of
constraint satisfaction problems in a distinctly non-classical way, which uses unique
properties of quantum computation, like interference.
Constraint satisfaction problems: Solutions to a generic constraint satisfaction problem lie
in the space of assignments of m different values to n variables such that a set of l constraints
are satisfied. Such a search space harbors a natural structure taking the form of a lattice given
by set containment. Figure 7 shows the lattice structure formed in the assignment space
spanned by two variables taking two possible values: 0 and 1. The sets in this lattice can be
Figure 7. Lattice of variable assignments.

put in one-to-one correspondence with the standard basis vectors for a four-qubit quantum
state space, such that a 1 in the binary sequence representing a basis vector corresponds to
inclusion, while a 0 corresponds to exclusion of the respective element (variable
assignment). The result will be a lattice of variable assignments in ket form, depicted in
figure 8.
Moving the amplitude up: Hogg’s idea is to start with a single initial amplitude
concentrated in the j0. . .0l state and then iteratively move amplitude up the lattice, from sets
to supersets and away from “bad sets” (i.e. sets that violate the constraints). Note the
interesting particularity of this algorithm, to start differently from all the quantum algorithms
presented so far (Deutsch –Jozsa, Shor’s, Grover’s) which all begin by computing a function
on a superposition of all the input values at once.
Hogg devised two methods for moving the amplitude up the lattice [99,100] by
constructing a unitary matrix which maximizes the movement of amplitude from a set to its
supersets. For moving the amplitude away from bad sets, Hogg suggests a general technique
based on adjusting the phases of the bad sets and using interference to decrease the amplitude
of sets that have bad subsets (due to cancellations), while the amplitude of good subsets will
add to increase the amplitude of the encompassing superset. Depending on the particular
problem to be solved, the choice of different policies for manipulating the phase changes will
result in different cancellations obtained. The technique exploits the property that if a state
violates a constraint, then so do all states above it in the lattice.
Efficiency: The problem with Hogg’s algorithms is one shared by heuristic algorithms in
general. The use of problem structure is complicated enough not to allow for an accurate
estimation of the probability of obtaining a solution from a single execution of the algorithm.
Therefore, it is difficult to analyze the efficiency of Hogg’s quantum search heuristics.
The efficiency of classical heuristic algorithms is estimated by empirically testing the
algorithm. Since a practical quantum computer on which Hogg’s algorithms could be tested
was not yet built (nor will it be in the near future), all we can do is simulate his quantum
algorithms on a classical computer. Unfortunately, this incurs an exponential slowdown, thus
making it feasible only on small cases. From the few small experiments that have been done,
the guess is that Hogg’s algorithms are more efficient than Grover’s algorithm applied to
structured search problems, though the speed-up is likely to be only polynomial. Even so, for
Figure 8. Lattice of variable assignments in ket form.

the computationally difficult problems to which they can be applied, a small polynomial
speed-up on average is of significant practical interest.
5.3 Quantum simulations
If practical quantum computers will ever become a reality, then a class of tasks at which they
could naturally outperform any classical machine is simulating quantum mechanical systems
occurring in Nature. As the size (number of constituents) of a quantum system increases, the
number of variables required to describe the state of the system grows exponentially. So, in
order to store the quantum state of a system with n distinct components, a classical computer
would need some c n bits of memory, with the constant c depending upon the system being
simulated and the desired accuracy of the simulation. Furthermore, calculating its evolution
over time would require the manipulation of a huge matrix, involving c n £ c n bits. On the
other hand, a machine that worked by quantum means would intrinsically make a much more
efficient simulator, requiring only a linear number of qubits. This is also the reason for which
efficient simulations of a quantum computer on a classical machine are not known to exist.
Feynman was the first to hint at the idea that rich and complex dynamical evolutions of some
quantum systems could be simulated resorting only to simple local interactions [80]. Feynman’s
idea was refined by Lloyd, who showed in 1996 that the logical operations available on a
quantum computer could be marshaled to simulate the behavior of virtually any quantum system
whose dynamics is determined by local interactions [129]. However, the problem of obtaining
the desired information about the simulated quantum system from the information hidden in the
wavefunction characterizing its state still remains. Thus, a crucial step in making quantum
simulations useful is the development of systematic means by which desired answers can be
efficiently extracted. As we have already seen, direct observation (measurement) is of little help
due to the irretrievable loss of information incurred by the collapse of the wavefunction.
The impact of obtaining faster and more accurate simulations of systems in which
quantum phenomena are important may materialize in significant advances in those fields.
We enumerate here some of the areas that would benefit from such efficient simulations:
molecular chemistry with direct applications in pharmaceutical industry, studying the
behavior of liquids and gases, gaining insights about the nature of forces governing the
behavior of nuclear particles, verifying the strength of various theories trying to explain
superconductivity, especially at “high” temperatures. As Lloyd has noted, it might even be
possible for quantum simulations to take advantage of a greater range of quantum hardware.
Thus, for instance, decoherence, instead of being a liability and with effort spent trying to
neutralize its effects, could be turned to our advantage by using it to replicate the interaction
of the simulated system with its environment. Finally, quantum simulation algorithms could
also be employed as a general method to obtain insight into other quantum algorithms.
Shape of quantum circuits: We are going to conclude this review of quantum algorithms
with a general observation about the shape of the quantum circuits implementing them. As it
became apparent with the unifying work of Richard Cleve, Artur Ekert and colleagues [54],
quantum algorithms tend to have a rather similar structure. The quantum circuits used to
describe their operation usually have a specific (problem dependent) unitary transformation
in the middle, sandwiched between Hadamard or Fourier transforms. It would be interesting
to know whether more or less all quantum circuits will take this form. A positive answer
to this question may induce the feeling that this structure offers only a rather limited range of
opportunities.
6. Quantum complexity
Similar to classical complexity theory, quantum complexity theory is concerned with

classifying the difficulty of various computational problems, grouping them into complexity
classes according to the spatial (memory) and time resources needed by a quantum computer
to solve those problems. Naturally, in this endeavor, of great interest is the comparison of the
relative powers of the classical model of computation (represented by the Universal Turing
Machine, for example) and the computational model based on quantum mechanics.
In the previous section, several examples of quantum algorithms were presented which
perform better than any known classical algorithm confronted with the same computational
problem. But are quantum computers strictly more powerful than classical ones? Are
there computational tasks which can be efficiently solved quantum-mechanically and
yet no Turing machine is able to tackle them efficiently? The answers to these questions is
not known, despite the suspicions fostered by examples such as factoring, which suggest that
the above questions can be answered in the affirmative. As a matter of fact, the initial
excitement caused by an eventual positive answer has been tempered lately by some
complexity theorists who now ask: “How long until we can prove in the general case that a
tight upper bound for quantum performance speedups is quadratic?” [83]. This section
is intended to offer a map containing the most important milestones in the development of
quantum complexity theory.
6.1 Reversibility
The first step in evaluating the computational power of a quantum computer was to determine
whether we can simulate a classical logic circuit using a quantum circuit. Bennett showed
that any classical circuit can be converted into an equivalent reversible circuit, and moreover,
that this conversion can be done efficiently [15]. One way to achieve this is to make use of a
reversible gate known as the Toffoli gate, which is universal for classical computation and
can be used as a basic building block in constructing reversible circuit models of computation
[85]. Since the Toffoli gate is reversible (its inverse is itself), it can also be implemented as a
quantum logic gate, depicted in figure 9. Following its operation rules, the Toffoli gate can be
seen as a controlled-controlled-NOT gate. Thus, we can immediately conclude that a
quantum computer is at least as powerful as a classical deterministic computer. Furthermore,
quantum computers are also capable of performing any computation which a classical
probabilistic computer may do by using a Hadamard gate to produce random fair coin tosses.
6.2 Quantum Turing Machines
In order to be able to prove theorems about further capabilities of quantum computers,

researchers in quantum complexity theory tried to capture the essential features and
principles underlying quantum computation into an abstract model, free of any particular
implementation details. Since the Turing machine has proved to be the most influential
among all models of classical computation, much of the research effort was focused on
defining a quantum analog of the Turing machine.
Charles Bennett had already shown in 1973 that a reversible Turing machine was a
theoretical possibility [15]. Then, Paul Benioff realized the similarity between the operation
of a reversible Turing machine and the dynamical evolution of an isolated quantum system,
governed by the unitary evolution of the Hamiltonian in Schrödinger’s equation. Thus, he
devised a hypothetical quantum machine (system) whose evolution over time mimicked the
actions of a classical reversible Turing machine [13]. However, Benioff’s quantum
mechanical implementation of a Turing machine did not incorporate any of the features
responsible for the computational power attributed to quantum computers, such as
superposition, interference and entanglement. For this reason, Benioff’s design can do no
more than a classical reversible Turing machine.
Although Feynman began by trying to simulate quantum systems on classical machines
[80] (which is the opposite of what Benioff did), he also investigated the feasibility of a
quantum mechanical computer based on a model of reversible computation developed by
Toffoli and Fredkin [85]. His conclusion was that “the laws of physics present no barrier to
reducing the size of computers until bits are the size of atoms, and quantum behavior holds
dominant sway” [81]. Despite the fact that Feynman’s model was an advance over Benioff’s
machine, it appears that its capabilities are also limited to mimic the operations of a general
purpose computer, with no references to simulating quantum mechanics being made.
The first true QTM capable of exploiting genuine quantum mechanical effects was devised
by Deutsch in 1985 [61]. It called for a quantum mechanical processor responsible for
controlling the read, write and shift operations performed by the head through quantum
mechanical interactions. For each internal state in which the head can exist, the action of the
QTM is specified by a set of quantum control rules taking the form of a Hamiltonian operator
that specify the probability amplitude for the occurrence of all allowed state-to-state
transitions. Deutsch’s quantum computer also makes use of an infinitely long tape, on which
both the program and the input data would be stored. Defining such an abstract QTM was a
key step in making it possible to study the computational power of quantum computers. It
gave researchers in the field an important tool to address fundamental issues like universality,
computability, provability and complexity.
6.3 Universality
The question of universality concerns whether a given machine can simulate all others and
how efficiently can it do this. Deutsch showed that it was possible to devise a Universal
QTM, but the simulation overhead was exponential in the running time of the simulated
Turing Machine in the worst case. The efficiency of Deutsch’s universal simulator for QTMs
Figure 9. Quantum Toffoli (controlled–controlled-NOT) gate.

was improved by Bernstein and Vazirani [26] who were able to prove the existence of a
Universal QTM whose simulation overhead is polynomially bounded.
But quantum information theory and universality are also connected in a subtle and therefore
unexpected way. The example presented in [143] shows that, with respect to measurements, only
a parallel approach can succeed in distinguishing between entangled quantum states. This result
implies that no finite machine exists capable of simulating all possible computations. In the
context invoked, the notion of a Universal Computer is consequently a myth.
6.4 Computability
In terms of computability, we have already seen that anything computable by a classical

computer is also computable by a quantum computer. But is the inverse of this statement also
true? It is not difficult to envisage a classical Turing machine that simulates an arbitrary
quantum circuit, if one does not care about efficiency. The simulation in [14] requires space,
and therefore time, exponential in the number of qubits in the quantum circuit. Bernstein and
Vazirani [26] have given a simulation that takes polynomial space, but exponential time.
The lack of an efficient classical simulation of a quantum computer induced the idea that a
quantum computing machine may be inherently faster. However, no one has been able to
prove that such an efficient simulation does not exist.
So when it comes to the continuous evolution, described by Schrödinger’s equation, of a
quantum system, it seems that a quantum computer is no more powerful (with respect to
computability) than a classical Turing machine. Nonetheless, a measurement operation
causes the system to undergo a sudden, discontinuous transformation in which a superposed
state will collapse onto one of the basis states. The outcome of the measurement depends on
the probabilities given by the amplitudes of the components in the wave function. This
intrinsic non-determinism in the behavior of any observed quantum system can only be faked
by a classical machine through the use of a pseudo-random number generator. This
observation led Deutsch to the conclusion that the notion of computability depends not on
mathematical ideas about the nature of algorithms, but on the computational capabilities of
physical systems. His point of view is reinforced by Rolf Landauer who stresses that
information is a physical quantity: “Information is inevitably represented by real physical
entities and is therefore tied to the laws of physics” (in [42] p. 116). Consequently, Deutsch
proposed the reformulation of the Church-Turing thesis in physical terms. In his book
“The Fabric of Reality” he gives the following articulation of what he called the Turing
principle: “There exists an abstract universal computer whose repertoire includes any
computation that any physical possible object can perform” ([63] p. 132).
There are also some theoretical results that seem to hint at the idea that quantum devices may
be strictly more powerful than classical machines [44,45]. These results refer to computations
performed through quantum means that go beyond Turing’s barrier. Calude and Pavlov show
how a mathematical quantum device, which encodes the whole data into an infinite
superposition, can be constructed to solve the Halting Problem [45]. Although their “halting
machine” may not be of practical interest, the theoretical breakthrough is certainly important.
6.5 Provability
Long before quantum mechanics was suspected of enhancing the computational capabilities
of classical Turing machines, Gödel showed that truth and provability are distinct concepts in
any sufficiently strong formal system. Physics (and quantum mechanics in particular) makes
no exception, since as a mathematical science it is treated formally. The introduction of
quantum computers brings another interesting differentiation between the ability to prove
and the ability to provide the proof trace. In principle, a QTM could be used to create some
proof that relied upon quantum mechanical interference among all the computational paths
pursued in parallel in a superposition. We can measure the final state of the QTM in order to
find an answer to our conjecture, but there is no way to record the intermediate steps of the
proof without invariably disrupting its future course.
6.6 Complexity
Unlike computability, which tries to clearly distinguish between problems that computers
can and cannot do, complexity focuses on the efficiency with which an answer to a solvable
problem is computed. In other words, complexity quantifies how the memory and time
resources scale with problem size. From the complexity point of view, the most interesting
open question is whether a QTM can make all NP problems tractable, since many of
these problems have very practical applications, but are computationally hard to solve.
QTMs can be thought of as quantum mechanical generalizations of probabilistic Turing
machines (PTM). The key difference however is that in a PTM only one particular
computational trajectory is followed (even if it is non-deterministic), while in a QTM all
possible computational trajectories are pursued simultaneously, leading to a superposition of
the achievable states and allowing complex phenomena, like quantum mechanical
interference, to take place. To support this behavior, each cell of the QTM’s tape must be
able to encode a blend of 0 and 1 simultaneously. Furthermore, the tape as a whole can exist
in highly non-classical states, with different parts of it becoming entangled together as the
computational process progresses. The head of the machine can also be in a spatial
superposition state, because in order to be able to follow multiple computational paths
simultaneously, the head must have the capacity to be at several locations at the same time.
These capabilities allowed the development of a new technique, which Deutsch called
quantum parallelism, and which proved to be useful in computing some joint properties of all
the outputs of a function faster than a classical Turing machine [61]. However, with respect
to mere function calculation, Deutsch proved that QTMs have the same complexity class as
TMs. Jozsa analyzed the power of quantum parallelism, giving a mathematical
characterization for the classes of joint properties that can and cannot be computed by
quantum parallelism [112].
6.6.1 QTM vs. DTM. Deutsch and Jozsa exhibited for the first time an example in which the
QTM is exponentially faster than a deterministic Turing machine (DTM) [67]. Their
quantum algorithm distinguishing between constant and balanced functions needs only one
evaluation of the function, simultaneously on all possible inputs, while the classical machine
has no other choice but to compute all function values sequentially and then counting the two
possible outputs to see whether they balance. Despite its impressive speedup, the Deutsch–
Jozsa algorithm is of no practical importance, having no known applications. Furthermore,
the problem is also easy for a PTM, which can solve it very quickly with high probability.
Therefore, the race was now on to find a problem for which a QTM could beat both a DTM
and a PTM.
6.6.2 Quantum complexity classes. Deutsch and Jozsa were also the first to propose the
use of quantum complexity classes to capture the difficulty of solving specific problems
on quantum models of computation. Thus, in analogy with the classical classes P, Zero
error Probability in Polynomial time (ZPP) and Bounded error Probability in Polynomial
time (BPP) we have the quantum classes QP, ZQP and BQP. These mean that a problem can be
solved with certainty in worst-case polynomial time, with certainty in average-case
polynomial time, and with probability greater than 2/3 in worst-case polynomial time,
respectively, by a QTM. The work of Yao [185], who showed that complexity theory
for quantum circuits matches that of QTMs, legitimizes the study of quantum circuits,
which are simpler to design and analyze than QTMs. Thus, the running time of a
quantum algorithm is usually expressed as a function of the number of elementary operations,
that is, elementary unitary transformations (quantum gates) that have to be applied to the initial
state in order to evolve it into the final state from which the answer can be extracted through
measurement.
6.6.3 QTM vs. PTM. The first hint that QTMs might be more powerful than PTMs was
given by Bernstein and Vazirani, who showed how to sample from the Fourier spectrum of
any Boolean function on n bits in polynomial time on a QTM [26]. No algorithm was known
to replicate this result on a PTM. Then, Berthiaume and Brassard were able to construct an
oracle, relative to which a decision problem exists that could be solved with certainty in
polynomial time in the worst case on a quantum computer, but could not be solved classically
in probabilistic expected polynomial time, if errors were not tolerated [28]. In the same
paper, they also show that there is a decision problem solvable in exponential time on a QTM
and in double exponential time on all but finitely many instances on any DTM. These two
results, besides being a victory of quantum computers over classical machines (deterministic
or probabilistic) also prove that the power of quantum computation cannot simply be
ascribed to the indeterminism inherent in quantum theory.
Further evidence that QTMs are more powerful than PTMs was brought by Bernstein and
Vazirani [26] who showed that there exist oracles under which there are problems belonging to
BQP but not BPP. Moreover, Simon managed to prove the stronger result that there exists an
oracle relative to which BQP cannot even be simulated by a PTM allowed to run for an
exponential number of steps [165].
Unfortunately, all these results share the same drawbacks as the Deutsch– Jozsa algorithm.
In the first place, they are relativized results, so they do not break any major ground in terms of
computational complexity. Note, in this context, that the quantum circuitry responsible for
computing the function f in the Deutsch –Jozsa algorithm can also be assimilated with a black
box or oracle. Secondly, they are generally contrived problems, defined with a specific
theoretical purpose in mind, and do not offer practical applications. The development of
Shor’s algorithms for factoring integers and computing discrete logarithms was important
especially from this point of view. Although they were not the first quantum algorithms
achieving an exponential speed-up relative to the corresponding best known classical
algorithms, they certainly have the potential to deliver a devastating blow to the security of
currently used cryptographic codes. As for the first observation above, despite the fact that
they do not rely on the existence of any oracle, they still fail to fulfill the dream of solving all
NP problems efficiently. The reason is that neither factoring nor computing discrete
logarithms is known to be NP-complete, in spite of the general belief that they are not in P.
6.6.4 Quantum vs. classical complexity classes. The relative power of quantum computers
with respect to classical ones can be couched in the relationships between classical and
quantum complexity classes. The lack of a precise answer to the alleged superiority of
quantum computation is also reflected in the difficulty of placing quantum complexity
classes among classical ones. Few such universal (unrelativized) results have been proven so
far. Naturally, the class BQP attracts most interest, being considered, even more so than QP,
the class of all computational problems which can be solved efficiently on a quantum
computer. Shor’s algorithms, for instance, belong to BQP, while it is not known whether they
are also in BPP or not. So, exactly where BQP fits with respect to P, BPP, NP and PSPACE is
as yet unknown. What is known is that BPP # BQP (that is, quantum computers can
efficiently solve all the problems that are tractable for a PTM) and BQP # PSPACE (there
are no problems outside of PSPACE which quantum computers can solve efficiently) [26].
Consequently, from P # BPP # BQP # PSPACE we can see that BQP lies somewhere
between P and PSPACE (see figure 10). Thus, we know for sure that BQP contains all of P
and BPP, but whether it also contains some problems in PSPACE that are not in NP, for
example, remains an open question.
The difficulty to settle these issues comes in part from unsolved classical complexity
problems, like the question whether PSPACE is strictly bigger than P. Note the important
interdependence between quantum complexity issues and classical complexity theory.
If, some day, quantum computers will be proved to be strictly more powerful than
classical computers (i.e. BPP , BQP) then it will follow that P is not equal to PSPACE.
Many researchers have unsuccessfully attempted to prove this latter result, which is an
indication that it may be quite non-trivial to prove the strict superiority of quantum
computers over classical ones, despite some evidence in favor of this proposition.
6.6.5 Quantum speed-up. The important change in the attitude of quantum complexity
theorists relative to the speed-up gained by quantum computers when dealing with
Figure 10. Relationships between quantum and classical complexity classes.

NP-complete problems is also of great relevance to the task of identifying those problems
that are best suited for enabling a quick quantum solution. The discovery of fast quantum
algorithms for factoring and computing discrete logarithms has raised great hopes for the
development of an efficient way to tackle all NP problems through quantum means. One
approach pursued by some researchers exploits the special structure that NP-complete
problems share, allowing complete solutions to be assembled out of smaller, partial solutions
[183]. Grover, together with Cerf and Williams, employed the amplitude amplification
technique from his unstructured quantum search algorithm to perform a search in a tree
structure [46]. The algorithm works by nesting a quantum search in the middle of the tree
with another in the fringe, in order to bias the search in the fringe of the tree in favor of only
the extensions of the partial solutions. For an NP-complete problem that takes N x steps
classically,
p ffiffiffiffiffiffi this quantum-mechanical approach will reduce the number of steps to roughly
N x , where x , 1 depends on the degree of constrainedness of the problem. While still far
from an exponential speed-up, this result may nevertheless be of practical interest.
Little by little, people realized that in many cases an exponential increase in efficiency, due
to the use of quantum techniques of computation, is not possible. Beals et al. [10] proved
bounds on the efficiency of quantum computers relative to classical deterministic computers
for several computational problems. In many cases, if the quantum machine takes N steps,
then the classical one takes at most O(N 6) steps. Finally, Bennett et al. [16] have concluded
that treating the circuitry relating to an NP-complete problem as an oracle and trying to use
some form of quantum parallelism to search in parallel through all the possible solutions to
the problem will not improve on the speed of Grover’s algorithm. Although this result does
not prove that NP cannot be contained in BQP, it does establish the fact that there is no
intrinsic property of quantum computation that will function like a black box to solve
NP-complete problems. And since Grover’s search algorithm was proved to be optimal, the
current belief is that the quadratic improvement may be the best we can get out of a
quantum computer in these kinds of tasks [158].
We should not forget, however, that Grover’s algorithm is optimal for an unstructured
search space and that deeper structures may exist, which can be exploited easier using the set
of tools provided by quantum computation, to yield a polynomial running time solution.
This observation can be made compatible with the conjectured quadratic tight upper bound
on the speed-up provided by quantum computers on NP-complete problems, only if we
assume that problems like factoring do not belong to this category. There are two interesting
implications of this hypothesis. First, it should reinforce the justification for the efforts
devoted to showing that factoring is tractable even through classical means of computation,
for otherwise where would this problem stand? In the second place, a question arises as to
whether similar complex or well-disguised structures exist in other problems, that
prevented an efficient classical solution to be discovered so far, and yet clever quantum
algorithms may be able to exploit them in order to produce a quick solution.
7. Quantum error correction
As the field of quantum computation and quantum information was developing, a huge gap
opened up between theory and experiment. While theorists were racing ahead to devise novel
applications for idealized quantum machines endowed with thousands of gates, experiment
has barely got beyond the stage of building a single quantum gate. The extreme fragility of
quantum states makes it very difficult to maintain their coherence in order to perform useful
computations.
A challenging task. The task of running a quantum computer at a reasonable degree of
accuracy is much more challenging than in the case of a digital computer. Digital information
is much easier to protect against errors than the analog information implied by the continuous
variables describing an arbitrary quantum state. The laws of quantum mechanics prevent, in
general, a direct application of the classical error-correction techniques. We cannot inspect
(measure) at leisure the state of a quantum memory register to check whether an ongoing
computation is not off track without the risk of altering the intended course of the
computation. Moreover, because of the no-cloning theorem, quantum information cannot be
amplified in the same way digital signals can. So, it is no surprise that many researchers
initially doubted the feasibility of quantum computers [125,177]. Since then, research efforts
have managed to demonstrate that dealing with errors in quantum states is possible, at least
from the theoretical point of view, making quantum error correction one of the most active
research areas in quantum computation and quantum information.
7.1 Quantum errors
The decoherence problems responsible for the alteration of quantum states occur because
keeping a quantum system isolated from its environment is virtually impossible. Therefore,
over time, the quantum system tends to couple with the environment, causing quantum
information to leak out. The simplest idea to avoid such a scenario is to perform the useful
quantum computation before major external errors may occur. Errors that describe the result
of a quantum system coupling to the environment are termed external, as opposed to internal
errors, which arise when the input state is not prepared exactly as we intended or when the
architecture of the quantum computer is not exactly correct. Fortunately, in the case of internal
errors, Wojciech Zurek found that input errors do not grow with time and that architectural
errors determine an error in the computation, which is proportional to the square of the size of
the error [187]. Both of these types of internal errors grow much more slowly in a quantum
computer than in a classical computer, so the real concern remains eluding the external errors.
The two processes mainly responsible for inducing external errors are dissipation and
decoherence. A dissipative error process usually causes a qubit to flip by losing energy to its
environment. The loss of coherence, on the other hand, is a much more subtle process.
It affects the phase of a qubit, a pure quantum mechanical (that is, non-classical) feature,
undetectable through direct measurement. The entanglement between the state of a quantum
memory register and the state of its environment tends to randomize the relative phases of the
possible states of the memory register. The immediate consequence is the annihilation of the
interference effects, a powerful tool used in any genuine quantum computation.
Decoherence† is very problematic because of the speed with which it occurs, allowing
little time for a useful quantum evolution before classical behavior effects take over. Joos has
estimated the coherence time of different-sized systems under various environments [111].
His analysis shows that coherence time is most affected by temperature and interactions with
surrounding gas particles. DiVincenzo also tried to give an estimation for the maximal
†
The term “decoherence” is sometimes used to denote the alteration of a quantum state in general. Hopefully, the
context in which it is used can always eliminate any possible ambiguity.
number of computational steps that can be performed without losing coherence, but from the
point of view of the possible “materials” used for the physical realization of a qubit [71].
Trying to choose the best suited materials for the components of a quantum computer that
is to be operated at low temperatures and in vacuum represents the passive alternative
in managing decoherence. Although this may be enough for certain tasks, a general-purpose
quantum computer will necessarily require a more active approach in undoing whatever
errors may arise during the computation. Correcting quantum errors certainly requires much
more ingenuity than fixing classical bits, but the basic idea of using redundancy is still useful.
7.2 Error correction via symmetrization

The technique called error correction via symmetrization [6,29] is yet another example of
how the duality of quantum-mechanical laws can be exploited for the benefit of quantum
computation. Although the measurement postulate severely restricts us in recycling
techniques from classical error correction, it can still offer conceptually new ways of
achieving error correction that are simply unavailable to classical computers. Error
correction via symmetrization relies on the projective effect of measurements to do the job.
The technique uses n quantum computers, each performing the same computation. Provided
no errors occur, the joint state of the n computers is a symmetric one, lying somewhere in the
small symmetric subspace of the entire possible Hilbert space. Devising a clever
measurement that projects the joint state back into the symmetric subspace should be able to
undo possible errors, without even knowing what the error is.
To achieve this, the n quantum computers need to be carefully entangled with a set of
ancilla qubits placed in a superposition representing all possible permutations of n objects.
In this way, the computation can be performed over all permutations of the computers
simultaneously. Then, by measuring the ancilla qubits, the joint state of the n computers can
be projected back into just the symmetric computational subspace, without the errors being
measured explicitly. Peres has shown that this technique is most appropriate for correcting
several qubits that are slightly wrong, rather than correcting a single qubit that is terribly
wrong [150]. Error correction via symmetrization can be applied repeatedly, at regular time
intervals, to avoid the accumulation of large errors and continually project the computation
back into its symmetric subspace. Although the symmetrization algorithm is cumbersome
and unattractive from the practical point of view, due to its low efficiency, it has nevertheless
the important theoretical merit of proving that quantum error correction is, in principle,
possible. Later on, it was even shown how the algorithm could be employed to improve
frequency standards and the accuracy of atomic clocks [104].
7.3 Error-correcting codes

However, sudden large errors, like those induced by spontaneous emission, for example,
require a different error-correction strategy based on the use of quantum error-correcting
codes. Like in the classical case, the information contained in a qubit is spread out over
several qubits so that damage to any one of them will not influence the outcome of the
computation. In the quantum case, though, the encoding of the logical qubit is achieved
through the use of specific resources, by entangling the logical qubit with several ancilla
qubits. In this way, the information in the state of the qubit to be protected is spread among
the correlations characterizing an entangled state. Paradoxically enough, entanglement with

the environment can be fought back using quantum error-correcting codes based on
entanglement [152].
Peter Shor’s second major contribution to the advancement of quantum computation was
the creation in 1995 of an algorithm that could correct any kind of error (amplitude and/or
phase errors) affecting a single qubit in a 9-qubit code [162]. In a different approach, Steane
studied the interference properties of multiple particle entangled states and managed to
devise a shorter, 7-qubit code [168]. The number of qubits necessary for a perfect recovery
from a single error was later squeezed down to a minimum of five [24,124].
Naturally, in order to cope with more than one error at a time, it is necessary to use larger
and more elaborate codes. The construction of all these codes is based on the surprising, yet
beautiful idea of digitizing the errors. How can quantum errors be digitized when, as the
variables they affect, they form a continuum? The answer lies in the linear nature of quantum
mechanics. Any possible error affecting a single qubit can be expressed as a linear
combination of no errors (I), bit flip errors (X), phase errors (Z) and bit flip phase errors (Y).
P
Generalizing to the case of a quantum register, an error can be written as i ei Ei for some
error operators Ei and coefficients ei. The error operators can be tensor products of the single-
bit error transformations or more general multibit transformations. An error correcting code
that can undo the effect of any error belonging to a set of correctable errors Ei will embed n
data qubits (logical qubits) in n þ k code qubits (physical qubits). The joint state of the
ensemble of code qubits is subject to an arbitrary error, mathematically expressed as a linear
combination of the correctable error operators Ei.
To recover the original encoded state, a syndrome extraction operator has to be applied that
uses some ancilla qubits to create a superposition of the error indices i corresponding to those
correctable error operators Ei that have transformed the encoded state. Measuring only the
ancilla qubits will collapse the superposition of errors, yielding only one index k. But because
the ancilla qubits were entangled with the code qubits through the application of the
syndrome extraction operator, the side effect of the measurement is that the corruption
caused by all error transformations will be undone, save for the one corresponding to index k.
Consequently, only one inverse error transformation is required in order to complete the
recovery process. In essence, knowing how to deal with a set of fundamental error
transformations allows us to tackle any linear combination of them by projecting it to one of
the basis components. This process is referred to as digitizing or discretizing the errors.
The book of Nielsen and Chuang [146] offers a detailed treatment of quantum codes,
explaining how ideas from classical linear codes can be used to construct large classes of
quantum codes, as the Calderbank – Shor – Steane (CSS) codes [43,169], or the stabilizer
codes (also known as additive quantum codes), which are even more general than the CSS
codes and are based on the stabilizer formalism developed by Gottesman [91].
7.3.1 Scalability. The major drawback in using large and intricate quantum codes is that
the corrective circuit itself is as much prone to errors as the quantum circuit responsible
for the main computation. The more errors we are attempting to rectify, the more the
complexity and length of the recovery procedure will increase (see [76] for some theoretical
bounds on the relationship between the number of data qubits, the total number of entangled
qubits and the maximal number of errors that can be tolerated). Thus, we can only increase
the size of the error correction codes up to a certain cut-off point, past which no further gains
in accuracy can be made.
7.3.2 Concatenated codes. One attempt to overcome this limitation are the concatenated
codes. If a certain code uses n physical qubits to encode one logical qubit, a concatenated
version of that code is obtained by further encoding each of the n qubits in another block of n.
This hierarchical structure (tree) can be further expanded to accommodate as many levels as
desired. An important theoretical result was proved for such concatenated codes.
The threshold theorem states that by adding more levels of concatenation, the overall
chance for an error can be made arbitrarily small, provided that the probability of an
individual error is kept below a certain critical threshold [153]. What this result is saying is
that, in principle, we can reliably perform an arbitrarily long quantum computation.
The threshold level depends on many factors, like the specifics of the code used, the type of
errors and whether errors occur more frequently in qubit storage or in gate processing.
Because the process of error recovery within each tier of the hierarchy remains almost as
simple as that of the original n-qubit code, there is only a small overhead in the size of the
circuit necessary to ensure reliability. Of course, the high cost of using concatenated codes
lies in the exponential increase in the number of qubits with the number of levels added.
7.4 Fault-tolerance
Fault-tolerant quantum computation is another approach in trying to cope with the effects of noise
both in the main computational circuit and the recovery circuitry. The operation of each quantum
gate on the encoded data has to be carefully re-designed into a procedure for performing an
encoded gate on the encoded state in such a way as to prevent error propagation and error
accumulation. This will ensure that error correction will be effective at removing the errors. Since
the fault-tolerant gates form a discrete set, part of the problem is to simulate any possible idealized
quantum gates using only fault-tolerant procedures. For example, single-bit phase changes
required for performing the Fourier transform in Shor’s factoring algorithm can be approximated
using combinations of fault-tolerant fixed-angle versions. Following this line of thought it is
possible to perform a universal set of logical operations (the Hadamard, phase, controlled-NOTand
p/8 gates are one choice, but there are others as well) using only fault-tolerant procedures.
The second aspect of fault-tolerance refers to the possibility of introducing errors on the
encoded qubits by the error-corrective process itself. Peter Shor outlined a method of fault-
tolerant recovery which uses extra ancilla qubits and some extra circuitry to double-check the
diagnosis of errors [163]. On the other hand, John Preskill has identified five fault-tolerance
criteria which, when met, ensure that failures during the procedure for error correction do not
propagate to cause too many errors in the encoded data [153]. Fault-tolerance principles and
techniques can be successfully combined with concatenation codes to make possible a more
effective error correction in quantum computing and achieve the arbitrary level of accuracy
guaranteed by the threshold theorem.
7.5 Topological quantum computing
The final possibility that we mention here to fight decoherence goes by the name of topological
quantum computing. This method is the most speculative but potentially the most robust. Its aim
is to achieve fault-tolerance at the very level of the physical processes responsible for the
operation of a quantum gate. Topological quantum computing takes advantage of the non-local
characteristics of some quantum interactions by encoding quantum information in a more global
fashion. In this way, local disturbances will not affect the quality of the computation. A quantum
interaction having the desired properties and which, therefore, could be used as a building block
in the implementation of topological quantum gates is the Aharonov–Bohm effect [152]. It has
even been showed that it is possible to construct universal quantum gates using only Aharonov–
Bohm interactions [119].
The remarkable advances witnessed in quantum error correction have been acknowledged
even by the most hardened skeptics and have transformed the prospects for making quantum
computing a practical reality. However, the question of whether the theoretical results
outlined in this section will eventually be brought to life by a physically realizable
implementation is still to be decided by the years to come.
8. Physical embodiments of a quantum computer
The first models for a quantum computer (Benioff’s quantum upgrade of a classical reversible
Turing machine [13], Feynman’s quantum mechanical computer [81] and Deutsch’s Universal
QTM [61]) were abstract, theoretical tools designed to fathom the possibilities and limitations of
quantum information processing, without any intention to give them a physical interpretation.
An important step to move out of the “designer Hamiltonians” era towards a more practical approach
was Deutsch’s introduction of quantum logic gates and circuits (or networks) [62]. Addressing the
universality issue, researchers have discovered that single-qubit rotations supplemented with a two-
qubit logic gate (such as controlled-NOT) are enough to perform any possible quantum computation
[5]. But the surprising result related to this matter was the finding that almost any two-bit quantum
gate is good enough to build a universal quantum computer [65,128]. This computation friendly
feature of quantum mechanics (the fact that the laws of physics support computational universality)
was good news for experimentalists eager to build a quantum computer, because it meant that
computation could be built out of almost any kind of interaction or physical process.
DiVincenzo promulgates the following five criteria necessary for a practical quantum computer:
. a scalable physical system with well-characterized qubits;

. the ability to initialize the qubit state;
. decoherence times much longer than the quantum gate operation time;
. a universal set of quantum gates;
. the ability to measure specific qubits.
It is worth noting that some powerful quantum calculations can be performed without any two-
bit quantum gates at all. Two Carnegie-Mellon scientists showed that the quantum Fourier
transform involved in Shor’s factoring algorithm can be implemented using only single-bit
quantum gates [93]. The qubit interactions normally required to carry on the computation were
simulated in a semi-classical way: according to the results of measurements performed on certain
qubits in the calculation, the phases of others are adjusted using single-bit gates. However, the
modular exponentiation necessary to complete Shor’s algorithm still requires the use of some two-
bit quantum gates.
8.1 The quantum molecular switch
The first blueprint for a practical quantum computing machine was devised by Lloyd [127],
who was building upon the work of a group of German physicists regarding the realization
of a molecular quantum switch [171]. Lloyd extended their model to include quantum
superpositions, conditional logic and measurement techniques. His proposal was to use an
array of weakly interacting quantum states as the basis for a quantum memory register.
A concrete hardware platform for his design could be a heteropolymer. Each different atom
in such a molecule could play the role of a qubit, with the ground state and a metastable
excited state implementing the necessary binary logic.
The software for such a polymer machine would consist of sequences of laser light pulses
carefully tuned on the resonance frequencies (the difference between the energy of the
ground state j0l and the energy of the excited state j1l) characterizing each different atom in
the molecular chain. By varying the length of the pulse we can put the responding molecular
unit into any kind of superposition of ground and excited states. In particular, if a laser pulse
that flips the state of a certain qubit is called a p-pulse (because it rotates the state by 1808),
then by applying a p/2-pulse (a burst of light that lasts only half the time of p-pulses) we will
obtain an equally weighted superposition of j0l and j1l. This is how a Hadamard gate or a
square-root-of-NOT gate could be implemented.
Furthermore, Lloyd showed how the interatomic forces between three neighboring
molecular units could detune the resonance frequency of the middle atom (or unit) sufficient
enough to allow the implementation of a three-bit quantum gate. Finally, an extra, short-lived
excited state could be used to measure the state of an atom, by detecting if a photon with a
certain characteristic energy is emitted during the process. Similar schemes for manipulating
qubits can be applied to virtually any system in which there are local interactions.
8.2 Ion traps
The ion trap scheme imagined by Cirac and Zoller [50] was considered at some point the
favorite runner in the quantum race. According to their design, a quantum memory register
would be physically realized by using “fences” of electromagnetic fields to trap a number of
ions within the central region of an evacuated chamber. Each imprisoned ion embodies a
qubit, with the ground state representing j0l and a metastable state representing j1l.
Transitions between the two internal energy levels are obtained through optical means, by
shining a pulse of light from a laser beam of the appropriate frequency onto the target ion.
Lasers are also employed to cool down the ions into the ground state, and thus initialize them
for the computation. The cooling system, known as optical molasses, can also be used to
improve the accuracy of atomic clocks and create the conditions for novel states of matter
known as Bose –Einstein condensates (see [42] p. 242 –243).
Being charged particles, the ions are strongly coupled together by the combination of the
electric repulsion between them and the squeezing effect induced by the electric fields of the
trap. These antagonist forces create vibrational waves traveling among the trapped ions.
Cirac and Zoller showed how the quantized collective motion of the ions inside the trap (their
vibrations) can serve as a mechanism for implementing the necessary conditional logic.
The ion trap system potentially offers enormous flexibility, allowing the qubits interacting in
a quantum gate to be non-adjacent, due to the fact that the vibrations influence the whole
system. The measurement process follows the procedure outlined by Lloyd for the polymer
machine. The absence or presence of a fluoresced photon tells us which state the atom was in.
The scheme requires an extra energy level that couples strongly to the lowest energy state.
The first quantum logic gate built on ion trap technology was produced in 1995 [141].
The experimenters managed to achieve a 90% success rate on the proper operation
of a controlled-NOT gate. The original design of Cirac and Zoller was somewhat simplified,
in that they used a single beryllium ion to encode both qubits. The ion’s hyperfine internal
energy levels (determined by the interaction between the spin of the single electron in the
outermost shell and the ion’s nucleus) were chosen to incarnate one qubit, while two
vibrational energy levels of the ion as a whole were selected to represent the basis states of
the second qubit. Thus, we can say that the one-bit-per-atom limit was surpassed in this case.
The solution adopted avoided the technological problem of building small enough lasers to
address each qubit individually. Unfortunately, this becomes a serious issue for any attempt
to scale up the ion trap design. Also, the measured decoherence time safely allowed the
completion of the controlled-NOT operation, but seemed insufficient for an extended
computation.
The idea of using the electronic and motional states of a single ion (this time of the element
Calcium) to encode a 2-qubit state was later exploited to implement the Deutsch –Jozsa
algorithm on an ion-trap quantum computer [96]. And if scalable solutions to the ion-trap
architecture will ever be found, then it will be possible, in principle, to factor a number with
385 bits using “just” 1926 ytterbium atoms and 30 billion laser pulses [107]. Therefore, the
current research effort is focused on developing architectures for a large-scale ion-trap
quantum computer [51,116].
8.3 Flying qubits

Another proposal, which goes by the name of flying qubit-based quantum computers, uses the
cavity QED (quantum electrodynamics) technology to emulate the functionality of a
controlled-NOT quantum gate. Quantum information is encoded in the polarization states of
photons and, to facilitate their interaction, they are placed inside a small cavity with highly
reflecting walls together with a drifting cesium atom. The spacing between the mirrors in the
cavity can be adjusted to resonate with a particular transition between two energy levels of
the cesium atom and the target and control photons. The control qubit is initially prepared in
a circularly polarized state, either j þ l (left or counterclockwise) or j 2 l (right or
clockwise), while the target qubit is linearly polarized, meaning that it can be described by an
equal superposition of a left and a right circularly polarized state, such as
1 1
pffiffiffi j þ l þ pffiffiffi j 2 l
2 2
It turns out that the j þ l component of the target qubit is phase-shifted only if the control
qubit has excited the cesium atom, which in turn happens only when the control photons are
circularly polarized in a parallel direction to the atom’s spin. This conditional phase-shift is
the basic building block to construct any quantum logic circuit, when supplemented with
one-bit rotations.
Although, quantum-phase gates based on cavity QED have been successfully realized
experimentally [59,176], it is a very challenging endeavor to extend this technology to
complicated quantum circuits. An alternative would be to use beam splitters, phase shifters,
single-photon sources and photo-detectors in an all-optical attempt to build a quantum

computer [120,148,151]. Better still, the cavity QED and ion trap technologies could be
combined to scale up the atom trap technology [52]. Since they are traveling very fast,
photons could be used to transfer quantum information between distant trapped atoms
(perhaps through fiber optics) with each of the multi-bit ion traps responsible for storing
information and local processing. The cavity QED interactions would provide the necessary
methods for exchanging quantum information between the two different carriers. The same
goal could be achieved by using entanglement between a trapped atom and a photon [32].
8.4 NMR quantum computing

The most advanced quantum computer prototypes that seemed to have been built so far are
based on NMR spectroscopy, the same technique that is known in the medical setting as
magnetic resonance imaging (MRI) [56,87]. The method manipulates the nuclear spin states
of the atoms in a molecule to represent the qubits. When the sample is subjected to a very
powerful and uniform external magnetic field, the spins of the nuclei tend to occupy a lower
energy state, where they point with the field. Electromagnetic pulses in the radio frequency
range are employed to flip spins completely or at intermediate orientations. The energy levels
associated with the spin states of a nucleus are slightly altered by the closest electron orbits of
neighboring atoms (effect known as the chemical shift) and also by the nuclear magnetic field
of neighboring nuclei (spin – spin coupling). Thus, the resonance peaks in the NMR spectrum
of a particular type of nucleus may be shifted to reflect the chemical composition of the
sample. Analyzing these subtle differences in the spectra we may be able to identify the
molecular structure of the sample. But for NMR quantum computing, chemical shifts offer a
way to address each qubit in a molecule individually, while spin – spin coupling is the
solution to achieve conditional logic.
What distinguishes an NMR quantum computer from the other approaches is its massive
parallelism and ultra-high redundancy. Each molecule in the test tube can be considered as an
individual processing unit with a local quantum memory register. Each “programming” radio
frequency pulse is tuned to resonate with the same part in each molecule simultaneously.
Thus, the quantum computation is carried out on a vast number of “processors”, in parallel.
It is true that the majority of them are affected by thermal noise and decoherence, but we do
not care about this, as long as the NMR signals generated by random spin orientations
average out to zero.
However, even after the removal of the random background, what we are left with is a
thermal distribution of states rather than a pure quantum state. So, in order to properly
initialize our computation, we are faced with the problem of selecting an ensemble of
molecules that share the same quantum state. One possible way to achieve this is to sacrifice
some nuclear spins on each molecule to check for quantum purity [87]. Alternatively, the
spatial averaging method [56] does not expend precious qubits, but it is very sensitive to
inaccuracies. To measure a qubit, we must first apply certain photon pulses and then interpret
the absorption spectrum obtained by processing the signal collected by the coils surrounding
the apparatus.
The NMR quantum computing technology was able to support the implementation of
complex algorithms. Grover’s search algorithm was run inside a chloroform sample in the
form of a quantum database query experiment [48]. The Deutsch –Jozsa algorithm was
successfully attempted by two independent groups [49,110]. In 1999, the first
implementation of the quantum Fourier transform was demonstrated. Although there were
doubts cast on whether the NMR experiments were capable of producing truly entangled
states [41], more people tried to produce working experimental realizations of increasingly
complicated quantum computations. In December 2001, scientists at IBM’s Almaden
Research Center claimed to have performed the world’s most complicated quantum-
computer calculation at that time [179]. In their 7-qubit experiment, they controlled billions
of custom-designed molecules to implement the simplest meaningful instance of Shor’s
algorithm for factoring the number 15. Considering the unprecedented control required over
the seven spins during the calculation, this result is quite an achievement.
Unfortunately, as with the other designs for a practical quantum computer, huge obstacles
have to be surpassed in order to make NMR a scalable technology. First, the size of the
quantum memory register is restricted by the number of nuclear spins, and hence, atoms in a
single molecule. Then, as more qubits are added to the system, the strength of the NMR
signal decreases exponentially as the number of possible spin states for the whole molecule
increases exponentially. In other words, the number of representative molecules in the initial,
sought-after “pure” state (with all spins pointing with the field) decreases exponentially.
Therefore, there is a trade-off between the computational power gained and the weakening of
the output signal (which has to be maintained above noise level). In recent years, there have
been a large number of experiments implementing various quantum algorithms on a small
number of qubits [73,117,130178].
8.5 Solid state approaches
All proposed technologies we have discussed so far are more or less extravagant and
represent a radical departure from the semiconductor technology on which today’s computer
industry is based. There are also some proposals that are challenging this point of view and
try to upgrade the semiconductor technology to the quantum level.
The Australian physicist Bruce Kane designed a silicon-based quantum computer in which
qubits are hosted by the nuclear spins of phosphorus atoms [115]. Each qubit can be addressed
individually by applying a tiny voltage to a metal strip or gate placed above each phosphorus
atom. The qubits can be made to interact indirectly via a coupling mediated by electron spins.
Conditional interaction is achieved through another type of gate, controlling two adjacent
nuclear spins. Measuring the spin state of a phosphorus nucleus involves the detection of a small
current due to the electron migration between phosphorus ion donors, which in turn depends on
the various states of the nuclear and electron spins and the strength of the gates.
Quantum dots. In the same class of semiconductor devices that may one day become the
key components of a quantum computer are the quantum dots. These are tiny islands of
semiconducting material on a chip, typically measuring a few hundred atoms across and
usually surrounded by an electrical insulator. By applying a negative voltage around these
blobs it is possible to squeeze out some of the freely moving electrons, such that only one
electron is left inside. The energy levels of this electron are discrete, so they can be used as
representations for 0 and 1. Single-qubit rotations would be straightforward to implement
using laser light pulses of various lengths tuned to the frequency corresponding to the energy
difference between the ground state and the first excited state of the electron. Conditional
logic could be achieved if two closely spaced quantum dots were subjected to an electric
field. This will make the resonance frequency of one dot (the target qubit) sensitive to the
state of the other dot [7]. Provided the resonance frequencies differ sufficiently, an
appropriately tuned p-pulse could selectively flip the target qubit state, subject to the state of
the control qubit, thus realizing a controlled-NOT quantum gate.
These solid state approaches are much more scalable with respect to the number of qubits,
but they suffer more from decoherence. Consequently, they do not allow, at this stage, any but
some very simple quantum calculations to be completed before the loss of coherence, which
is just another side of the scaling problem. Moreover, quantum dots and related devices need
special manufacturing techniques and liquid helium temperatures to function properly. On the
other hand, research into nanoscale technology may also have a huge impact on conventional
silicon chip manufacturing, if the concept of single-electron transistors will reach the point
of industry production. The colossal reductions in size and amount of power consumed by
conventional memory and logic devices would allow the current trends in miniaturization
to be sustained.
However, for those interested in seeing quantum dots becoming a serious candidate for
doing real quantum computing experiments, decoherence remains the main concern. Loss
and DiVincenzo proposed a more robust way of exploiting quantum dots, by storing the
quantum information in the spin orientation (rather than the energy level) of each electron in
a quantum dot [131]. They showed how quantum tunneling, achieved by raising the voltage
of a metal gate for a certain period of time, could be manipulated to simulate the functionality
of a quantum XOR (equivalent to a controlled-NOT) gate. Hellberg [97] also addresses
decoherence and the difficulty of generating local magnetic fields for single-qubit rotations.
Some hopes seem to be raised by the idea of using superconducting quantum dots,
incorporating Josephson junctions to implement qubits [132]. An all silicon design for a
quantum computer claims to take advantage of NMR successful aspects (ensemble
measurement, radio frequency control, long decoherence times), but at the same time allows
for more qubits and improved initialization [123]. A more recent proposal suggests spin-pair
encoded qubits in silicon [166].
With the large number of research papers reporting new experimental realizations every
month, it is difficult to predict which (if any) of the enumerated approaches will ever evolve
into a viable technology for building a practical and useful quantum computer. It may also
happen that a novel, currently unknown technology will emerge, transforming quantum
computing into a practical reality.
9. Final remarks
Since the early 1980s, when the field began to materialize, quantum computation and
quantum information has come a long way to impose itself as a stand-alone multidisciplinary
science. Indeed, today we can rightfully acknowledge the existence of a quantum computer
science with deep ramifications into information theory, algorithms and complexity,
cryptography, networking and communications, fault tolerance and error correction.
Quantum mechanics offers some really powerful tools that researchers have tried to exploit in
order to make information processing more efficient and secure. Superpositions, quantum
parallelism, interference and entanglement are ultimately responsible for the computational
power attributed to a quantum mechanical machine. Similarly, the measurement principle and
no-clonability theorem form the basis on which quantum cryptographic protocols were
developed. Some quantum effects, especially entanglement and non-locality, seem to go so
much against our common sense that even today there are voices who doubt their existence as
“elements of reality” (to paraphrase Einstein) and have expressed reservations with respect to the
experiments trying to “prove” them.
This ambiguous status of entanglement and the difficulty experimenters face in creating
multibit entanglement prompted some researchers to analyze more rigorously the implications
of entanglement for quantum computing. While proving that multi-partite entanglement is
necessary for quantum algorithms operating on pure states in order to offer an exponential speed-
up, Jozsa and Linden also argue that it is misleading to view entanglement as a key resource for
quantum computational power [114]. Other researchers seem to have reached similar
conclusions. Biham et al. show that quantum computing without entanglement still provides
some advantage over classical computing, but that, at the moment, entanglement is necessary for
all practical purposes [30].
Counterfactual quantum computing. However, seemingly absurd phenomena may occur
even in the absence of entanglement. An example is counterfactual quantum mechanics.
Counterfactuals, things that might have happened, although they did not in fact happen, have
no physical consequences in classical physics. However, the potential occurrence of a
quantum event can change the probabilities of obtaining certain experimental outcomes.
Interaction-free measurements provide one possible context for quantum counterfactuals. An
interaction-free measurement can be characterized as a kind of non-disturbing quantum
measurement in which no energy is exchanged between the probe and the object [122].
Consequently, the result is rather inferred than obtained through direct inspection [68].
Elitzur and Vaidman [78] describe how a Mach – Zehnder interferometer can be used to
perform an interaction-free measurement in the dramatic context of detecting an
ultrasensitive bomb. The Mach –Zehnder interferometer (depicted in figure 11) is an optical
device composed of beam splitters, mirrors and photon detectors carefully placed to bring
about quantum interference when a photon travels through the apparatus. Thus, when a
photon enters the first beam splitter horizontally, it will always emerge from the horizontal
port of the second beam splitter, provided the two arms of the interferometer have equal
lengths. As in the case of Young’s two-slit experiment, the reason is self-interference. Based
on this phenomenon, Elitzur and Vaidman show that, if a bomb that will explode when hit by
a single photon is placed in one arm of the interferometer, we still have a 25% chance of
detecting it without causing it to explode. Furthermore, it is actually sufficient to couple the
bomb to a sensor that performs a quantum nondemolition measurement on the photon,
registering its passage but allowing it to continue to the final beam splitter. There are
schemes through which the efficiency of this interaction-free measurement can be made
arbitrarily close to one [121].
Jozsa gave a computational spin to the interaction-free bomb detection thought experiment
[113]. He replaced the bomb with a quantum computer capable of answering a certain
decision problem, such as deciding whether a given number is prime, for example.
The computer performs its computation when a photon is detected passing through that arm,
otherwise it sits idle. Under these conditions, a quantum counterfactual result can be obtained
again. In 25% of the cases when the tested number is prime, we can actually reach this
conclusion without having to run the computer. So the mere fact that the computer is capable
of producing the solution allows us to infer the correct answer without actually running the
computer.
Exotic as it is, Deutsch’s many worlds interpretation may be the most “intuitive” way to
explain quantum counterfactuals. According to the multiverse interpretation, the computer
did not run in our universe, but it did run in some other, parallel universe, therefore making
the condition that the quantum computer has the potential to decide the question very
important. We also note that there is a subtle quantum information exchange between the
possible universes, which determines whether the interference phenomenon at the final beam
splitter will take place or not. In his book “The Fabric of Reality” Deutsch strongly advocates
the “many worlds” interpretation, bringing a powerful argument. When a 250-digit number is
factored using Shor’s algorithm, the number of interfering universes is of the order of 10500.
Since there are only about 1080 atoms in the entire visible universe, he argues that “physical
reality would not even remotely contain the resources required to factorize such a large
number” ([63] p. 217). As we have mentioned in the opening section, the differences between
the various interpretations of quantum mechanics are a matter of taste to physicists and
ultimately involve philosophical notions like consciousness, which are beyond the scope of
this paper.
Quantum mechanics is considered the most accurate description of the Universe we
currently have, but in the light of possible future discoveries, we may have to adjust this
theory some day. It is not clear now how the theoretical foundations of quantum computation
and quantum information will be affected in such a case, but the optimistic view is that even
in the worst case, the novel physical theory that will emerge may give rise to a new
computational paradigm, maybe even more powerful than quantum computing.
Although, from the computational complexity perspective, the question whether a
quantum computer is ultimately more powerful than a classical machine was not given a clear
and definitive answer yet (see the discussion in Section 6), it would still make a big difference
if a computing machine based on quantum principles could be built. The possibility of
breaking today’s public key cryptographic codes is certainly the most appealing, but even
obtaining a quadratic speed-up for the hard NP problems would be a significant improvement
of practical interest. The previous section has made us aware of the great challenges and
obstacles towards building a practical quantum computer. Fortunately, in cryptography, the
experiments designed to implement quantum protocols seem to be much more advanced.
They could become the most visible touch of quantum technology in the short run.
Figure 11. A Mach–Zehnder interferometer (BS, beam splitter; M, mirror).

But even if large-scale quantum computing machines will prove unfeasible, a quantum
computer with only a relatively small number of qubits might still be of some use. Such a
device could be the simulator Feynman first envisaged, a machine used to effectively emulate
other quantum systems. Also, a small-scale quantum computer may eventually become a
useful tool of experimental physics. The ability of creating and manipulating just a handful
of qubits can allow physicists to run interesting tests on certain predictions of quantum
theory, thus helping basic physics research at the very least.
Whatever the future of quantum information processing is, the most fundamental idea we
can learn from it is that information is intrinsically physical and the power of a certain
computational paradigm ultimately depends on the characteristics of the physical support
used to embody information.
References
[1] Abrams, D.S. and Lloyd, S., A quantum algorithm providing exponential speed increase for finding
eigenvalues and eigenvectors, http://xxx.lanl.gov/abs/quant-ph/9807070, July 1998.
[2] Aczel, A.D., 2002, Entanglement (Vancouver: Raincoast Books).
[3] Aspect, A., Dalibard, J. and Roger, G., 1982, Experimental test of Bell’s inequalities using time-varying
analyzers, Physical Review Letters, 49, 1804–1807.
[4] Barenco, A., 1995, A universal two-bit gate for quantum computation, Proceedings of the Royal Society of
London A, 449, 679–683.
[5] Barenco, A., Bennett, C.H., Cleve, R., DiVincenzo, D.P., Margolus, N., Shor, P., Sleator, T., Smolin, J.A.
and Weinfurter, H., 1995, Elementary gates for quantum computation, Physical Review A, 52, 3457– 3467
http://arxiv.org/abs/quantph/9503016.
[6] Barenco, A., Berthiaume, A., Deutsch, D., Ekert, A., Jozsa, R. and Macchiavello, C., Stabilization of quantum
computations by symmetrization, http://xxx.lanl.gov/abs/quant-ph/9604028, April 1996.
[7] Barenco, A., Deutsch, D. and Ekert, A., 1995, Conditional quantum dynamics and logic gates, Physical
Review Letters, 74(20), 4083–4086 http://arxiv.org/abs/quantph/9503017.
[8] Barenco, A. and Ekert, A., 1995, Quantum computation, Acta Physica Slovaca, 45, 205 –216.
[9] Barenco, A., Ekert, A., Suominen, K-A. and Törmä, P., 1996, Approximate quantum Fourier transform and
decoherence, Physical Review A, 54(1), 139–146.
[10] Beals, R., Burhman, H., Cleve, R., Mosca, M. and de Wolf, R., 1998, Quantum lower bounds by polynomials.
Proceeding of the 39th Annual Symposium on Foundations of Computer Science, Palo Alto, CA, November
8–11, FOCS’98, http://arxiv.org/abs/quant-ph/9802049, pp. 352–361.
[11] Beenakker, C.W.J. and Kindermann, M., 2004, Quantum teleportation by particle-hole annihilation in the
Fermi sea, Physical Review Letters, 92, 056801.
[12] Bell, J., 1964, On the Einstein–Podolsky–Rosen paradox, Physics, 1, 195–200.
[13] Benioff, P., 1980, The computer as a physical system: a microscopic quantum mechanical Hamiltonian model
of computers as represented by Turing machines, Journal of Statistical Physics, 22, 563 –591.
[14] Benjamin, E., Huang, K., Kamil, A. and Kittiyachavalit, J., Quantum computability and complexity and the
limits of quantum computation, http://www.cs.berkeley.edu/kamil/quantum/qc4.pdf, December 2003.
[15] Bennett, C.H., 1973, Logical reversibility of computation, IBM Journal of Research and Development, 17(6),
525–532.
[16] Bennett, C.H., Bernstein, E., Brassard, G. and Vazirani, U.V., 1997, Strengths and weakness of quantum
computing, Special issue on Quantum Computation of the SIAM Journal on Computing, 26(5), 1510– 1523
http://xxx.lanl.gov/abs/quantph/9701001.
[17] Bennett, C.H., Bessette, F. and Brassard, G., 1991, Experimental quantum cryptography. Lecture Notes in
Computer Science, (Berlin: Springer-Verlag), 473, pp. 253 –265.
[18] Bennett, C.H. and Brassard, G., 1984, Quantum cryptography: public key distribution and coin tossing.
Proceedings of IEEE International Conference on Computers, Systems and Signal Processing, (New York:
Bangalore, India, December 1984 IEEE), pp. 175– 179.
[19] Bennett, C.H. and Brassard, G., 1989, The dawn of a new era for quantum cryptography: the experimental
prototype is working!, SIGACT News, 20, 78 –82.
[20] Bennett, C.H., Brassard, G., Crépeau, C., Jozsa, R., Peres, A. and Wootters, W., 1993, Teleporting and
unknown quantum state via dual classical and EPR channels, Physical Review Letters, 70, 1895– 1899.
[21] Bennett, C.H., Brassard, G., Popescu, S., Schumacher, B., Smolin, J.A. and Wootters, W.K., 1996, Purification
of noisy entanglement and faithful teleportation via noisy channels, Physical Review Letters, 76, 722 –725
http://arxiv.org/abs/quantph/9511027.
[22] Bennett, C.H., Brassard, G. and Robert, J.-M., 1988, Privacy amplification by public discussion, SIAM Journal
on Computing, 17(2), 210–229.
[23] Bennett, C.H., DiVincenzo, D.P. and Smolin, J.A., 1997, Capacities of quantum erasure channels, Physical
Review Letters, 78(16), 3217–3220 http://arxiv.org/abs/quantph/9701015.
[24] Bennett, C.H., DiVincenzo, D.P., Smolin, J.A. and Wootters, W.K., 1996, Mixed state entanglement and
quantum error correction, Physical Review A, 54, 3824– 3851 http://arxiv.org/abs/quant-ph/9604024.
[25] Bennett, C.H. and Wiesner, S.J., 1992, Communication via one- and two-particle operators on Einstein–
Podolsky–Rosen states, Physical Review Letters, 69(20), 2881–2884.
[26] Bernstein, E. and Vazirani, U., 1997, Quantum complexity theory, Special issue on Quantum Computation of
the SIAM Journal on Computing, 26(5), 1411–1473 http://arxiv.org/abs/quant-ph/9701001.
[27] Berthiaume, A., 1997, Quantum computation. In: L.A. Hemaspaandra and A.L. Selman (Eds.) Complexity
Theory Retrospective II (New York: Springer-Verlag), pp. 23 –51.
[28] Berthiaume, A. and Brassard, G., 1994, Oracle quantum computing, Journal of Modern Optics, 41(12),
2521–2535.
[29] Berthiaume, A., Deutsch, D. and Jozsa, R., 1994, The stabilization of quantum computation. Proceedings of
the Workshop on Physics and Computation: PhysComp ’94 (Los Alamitos, CA IEEE Computer Society
Press), pp. 60– 62.
[30] Biham, E., Brassard, G., Kenigsberg, D. and Mor, T., 2003, http://arxiv.org/abs/quant-ph/0306182 Quantum
computing without entanglement.
[31] Biron, David, et al., 1998, http://xxx.lanl.gov/abs/quant-ph/9801066 Generalized Grover search algorithm for
arbitrary initial amplitude distribution.
[32] Blinov, B.B., Moehring, D.L., Duan, L.-M. and Monroe, C., 2004, Observation of entanglement between
a single trapped atom and a single photon, Nature, 428, 153 –157.
[33] Bohr, N., 1935, Can quantum-mechanical description of physical reality be considered complete?, Physical
Review, 48, 696– 702.
[34] Boschi, D., Branca, S., De Martini, F., Hardly, L. and Popescu, S., 1998, Experimental realization of
teleporting an unknown pure quantum state via dual classification and Einstein– Podolsky– Rosen channels,
Physical Review Letters, 80, 1121–1125 http://arxiv.org/abs/quant-ph/9710013.
[35] Bouwmeester, D., Pan, J.W., Mattle, K., Eibl, M., Weinfurter, H. and Zeilinger, A., 1997, Experimental
quantum teleportation, Nature, 390(6660), 575–579.
[36] Boyer, M., Brassard, G., Hoeyer, P. and Tapp, A., 1996, Tight bounds on quantum searching. Proceedings of
the Workshop on Physics and Computation PhysComp ’96 Los Alamitos, CAhttp://xxx.lanl.gov/abs/quant-ph/
9605034 (IEEE Computer Society Press), pp. 36–43.
[37] Brassard, G. and Crépeau, C., 1996, 25 years of quantum cryptography, SIGACT News, 27(3), 13–24.
[38] Brassard, G., Crépeau, C., Jozsa, R. and Langlois, D., 1993, A quantum bit commitment scheme provably
unbreakable by both parties. Proceedings of the 34th Annual IEEE Symposium on Foundations of Computer
Science (IEEE press), pp. 362–371.
[39] Brassard, G., Claude, C. and Wolf, S., 2003, Oblivious transfers and privacy amplification, Journal of
Cryptology, 16(4), 219 –237.
[40] Brassard, G., Hoyer, P. and Tapp, A., 1998, http://xxx.lanl.gov/abs/quant-ph/9805082 Quantum counting.
[41] Braunstein, S.L., Caves, C.M., Jozsa, R., Linden, N., Popescu, S. and Schack, R., 1999, Separability of very
noisy mixed states and implications for NMR quantum computing, Physical Review Letters, 83, 1054–1057
http://xxx.lanl.gov/abs/quant-ph/9811018.
[42] 2001, The quest for the quantum computer, Touchstone ed. (New York: Simon & Schuster).
[43] Calderbank, A.R. and Shor, P.W., 1996, Good quantum error-correcting codes exist, Physical Review A, 54(2),
1098–1106 http://arxiv.org/abs/quant-ph/9512032.
[44] Calude, C.S., Dinneen, M.J. and Svozil, K., 2000, Reflections on quantum computing, Complexity, 6, 35 –37.
[45] Calude, C.S. and Pavlov, B., 2002, Coins, quantum measurements, and Turing’s barrier, Quantum Information
Processing, 1(1–2), 107–127.
[46] Cerf, N.J., Grover, L.K. and Williams, C.P., 2000, Nested quantum search and NP-complete problems,
Physical Review A, 61, 032303 http://xxx.lanl.gov/abs/quantph/9806078.
[47] Childs, A.M., Cleve, R., Deotto, E., Farhi, E., Gutmann, S. and Spielman, D.A., 2003, Exponential
algorithmic speedup by quantum walk. Proceedings of the 35th Annual ACM Symposium on the Theory of
Computing, http://arxiv.org/abs/quant-ph/0209131, pp. 59–68.
[48] Chuang, I.L., Gershenfeld, N. and Kubinec, M., 1998, Experimental implementation of fast quantum
searching, Physical Review Letters, 80, 3408–3412.
[49] Chuang, I.L., Vandersypen, L.M.K., Zhou, X., Leung, D.W. and Lloyd, S., 1998, http://xxx.lanl.gov/abs/
quantph/9801037 Experimental realization of a quantum algorithm.
[50] Cirac, I. and Zoller, P., 1995, Quantum computations with cold trapped ions, Physical Review Letters, 74,
4091–4094.
[51] Cirac, I. and Zoller, P., 2000, A scalable quantum computer with ions in an array of microtaps, Nature, 404,
579–581.
[52] Cirac, I., Zoller, P., Kimble, H.J. and Mabuchi, H., 1997, Quantum state transfer and entanglement distribution
among distant nodes in a quantum network, Physical Review Letters, 78(16), 3221–3224 http://arxiv.org/abs/
quant-ph/9611017.
[53] Clauser, J.F., Horne, M.A., Shimony, A. and Holt, A., 1969, Proposed experiment to test local hidden-variable
theories, Physical Review Letters, 49, 1804–1807.
[54] Cleve, R., Ekert, A., Macchiavello, C. and Mosca, M., 1998, Quantum algorithms revisited, Proceedings of the
Royal Society of London A, 454, 339–354.
[55] Cory, D.G. and Havel, T.F., 2004, Ion entanglement in quantum information processing, Science, 304(5676),
1456–1457.
[56] Cory, D.G., Price, M.D. and Havel, T.F., 1997, http://xxx.lanl.gov/abs/quant-ph/9709001 Nuclear magnetic
resonance spectroscopy: an experimentally accessible paradigm for quantum computing.
[57] Crépeau, C., 1997, Efficient cryptographic protocols based on noisy channels, Lecture Notes in Computer
Science, 1233, 306 –317.
[58] Crépeau, C. and Kilian, J., 1988, Achieving oblivious transfer using weakened security assumptions.
Proceedings of the 29th Annual IEEE Symposium on Foundations of Computer Science, (IEEE press),
pp. 42– 52.
[59] Davidovich, L., et al., 1993, Quantum switches and nonlocal microwave fields, Physical Review Letters,
71(15), 2360–2363.
[60] Deutsch, D., 1985, Quantum theory as a universal physical theory, International Journal of Theoretical
Physics, 24(1), 1– 41.
[61] Deutsch, D., 1985, Quantum theory, the Church–Turing principle, and the Universal Quantum Computer,
Proceedings of the Royal Society of London A, 400, 97–117.
[62] Deutsch, D., 1989, Quantum computational networks, Proceedings of the Royal Society of London A, 425,
73–90.
[63] Deutsh, D., 1998, The Fabric of Reality, (Harmondsworth: Penguin books).
[64] Deutsch, D., 2001, http://xxx/lanl.gov/abs/quant-ph/0104033 The structure of the multiverse.
[65] Deutsch, D., Barenco, A. and Ekert, A., 1995, Universality in quantum computation, Proceedings of the Royal
Society of London A, 449(1937), 669–677.
[66] Deutsch, D., Ekert, A., Jozsa, R., Macchiavello, C., Popescu, S. and Sanpera, A., 1996, Quantum privacy
amplification and the security of quantum cryptography over noisy channels, Physical Review Letters, 77,
2818–2821 http://arxiv.org/abs/quantph/9604039.
[67] Deutsch, David and Jozsa, Richard, 1992, Rapid solution of problems by quantum computation, Proceedings
of the Royal Society of London A, 439, 553–558.
[68] Dicke, R., 1981, Interaction-free quantum measurement: A paradox?, American Journal of Physics, 49,
925–930.
[69] Diffie, W. and Hellman, M., 1976, New directions in cryptography, IEEE Transactions on Information Theory,
22(6), 644–654.
[70] Dirac, P., 1958, The Principles of Quantum Mechanics, 4th ed. (Oxford: Oxford University Press).
[71] DiVincenzo, D., 1995, Quantum computation, Science, 270, 255–261.
[72] DiVincenzo, D., 1995, Two-bit gates are universal for quantum computation, Physical Review A, 51,
1015–1022.
[73] Du, J., et al., 2003, Experimental implementation of the quantum random-walk algorithm, Physical Review A,
67, 042316.
[74] Einstein, A., Podolsky, B. and Rosen, N., 1935, Can quantum-mechanical description of physical reality be
considered complete?, Physical Review, 47, 777– 780.
[75] Ekert, A., 1991, Quantum cryptography based on Bell’s theorem, Physical Review Letters, 67, 661– 663.
[76] Ekert, A. and Macchiavello, C., 1996, Quantum error correction for communication, Physical Review Letters,
77, 2585–2588.
[77] Ekert, A., Rarity, J., Tapster, P. and Palma, G., 1992, Practical quantum cryptography based on two-photon
interferometry, Physical Review Letters, 69, 1293– 1295.
[78] Elitzur, A. and Vaidman, L., 1993, Quantum-mechanical interaction-free measurements, Foundations of
Physics, 23, 987–997.
[79] Farhi, E. and Gutmann, S., 1998, Quantum computation and decision trees, Physical Review A, 58(2),
915–928 http://arxiv.org/abs/quant-ph/9706062.
[80] Feynman, R., 1982, Simulating physics with computers, International Journal of Theoretical Physics,
21(6&7), 467 –488.
[81] Feynman, R., 1986, Quantum mechanical computers, Foundations of Physics, 16(6), 507 –531.
[82] Feynman, R., Leighton, R.B. and Sands, M., 1965, The Feynman Lectures on Physics, (Reading, MA:
Addison-Wesley).
[83] Fortnow, L., 2003, One complexity theorist’s view of quantum computing, Theoretical Computer Science,
292(3), 597–610.
[84] Franson, J. and Ilves, H., 1995, Quantum cryptography using optical fibers, Applied Optics, 33, 2949– 2954.
[85] Fredkin, E. and Toffoli, T., 1982, Conservative logic, International Journal of Theoretical Physics, 21(3/4),
219–253.
[86] Furusawa, A., Sørensen, J.L., Braunstein, S.L., Fuchs, C.A., Kimble, H.J. and Polzik, S., 1998, Unconditional
quantum teleportation, Science, 282, 706 –709.
[87] Gershenfeld, N. and Chuang, I.L., 1998, Quantum computing with molecules, Scientific American, 66– 71.
[88] Gisin, N., Renner, R. and Wolf, S., 2002, Linking classical and quantum key agreement: Is there a classical
analog to bound entanglement?, Algorithmica, 34(4), 389–412.
[89] Gisin, N. and Wolf, S., 1999, Quantum cryptography on noisy channels: quantum versus classical key-
agreement protocols, Physical Review Letters, 83, 4200–4203 http://arxiv.org/abs/quant-ph/9902048.
[90] Gisin, N. and Wolf, S., Linking classical and quantum key agreement: Is there “bound information”?, http://
arxiv.org/abs/quant-ph/0005042, May 2000.
[91] Gottesman, D., 1996, Class of quantum error-correcting codes saturating the quantum hamming bound,
Physical Review A, 54, 1862–1868 http://arxiv.org/abs/quant-ph/9604038.
[92] Greenberger, D.M., Horne, M.A., Shimony, A. and Zeilinger, A., 1990, Bell’s theorem without inequalities,
American Journal of Physics, 58(12), 1131–1143.
[93] Griffiths, R. and Niu, C-S., 1996, Semiclassical Fourier transform for quantum computation, Physical Review
Letters, 76, 3228–3231.
[94] Grover, L.K., 1996, A fast quantum mechanical algorithm for database search. Proceedings of the 28th Annual
ACM Symposium on the Theory of Computing, 24, Philadelphia, Pennsylvania, 22–24 May 1996,
pp. 212 –219.
[95] Grover, L.K., 1998, A framework for fast quantum mechanical algorithms. Proceedings of the 30th Annual
ACM Symposium on the Theory of Computing, http://xxx.lanl.gov/abs/quant-ph/9711043, pp. 53–62.
[96] Gulde, S., 2003, Implementation of the Deutsch –Jozsa algorithm on an ion-trap quantum computer, Nature,
421, 48 –50.
[97] Hellberg, C.S., Robust quantum computation with quantum dots, http://arxiv.org/abs/quant-ph/0304150, April
23, 2003.
[98] Hirvensalo, M., 2001, Quantum Computing (Berlin: Springer-Verlag).
[99] Hogg, T., 1996, Quantum computing and phase transitions in combinatorial search, Journal of Artificial
Intelligence Research, 4, 91 –128 http://xxx.lanl.gov/abs/quant-ph/9508012.
[100] Hogg, T., 1998, Highly structured searches with quantum computers, Physical Review Letters, 80(11),
2473–2476.
[101] Hogg, T., 2000, Quantum search heuristics, Physical Review A, 61, 052311.
[102] Hogg, T. and Portnov, D., 2000, Quantum optimization, Information Sciences, 128, 181 –197.
[103] Holevo, A.S., 1998, The capacity of the quantum channel with general signal states, IEEE Transactions on
Information Theory, 44(1), 269–273.
[104] Huelga, S.F., Macchiavello, C., Pellizzari, T., Ekert, A.K., Plenio, M.B. and Cirac, J.I., 1997, Improvement of
frequency standards with quantum entanglement, Physical Review Letters, 79(20), 3865–3868.
[105] Hughes, R., et al., 1995, Quantum cryptography, Contemporary Physics, 36(3), 149 –163 http://arxiv.org/abs/
quant-ph/9504002.
[106] Hughes, R., et al., 1997, Secure communications using quantum cryptography, Proceedings of SPIE, 3076,
2–11.
[107] Hughes, R.J., et al. Decoherence bounds on quantum computation with trapped ions, http://xxx.lanl.gov/abs/
quant-ph/9604026, April 1996.
[108] Hughes, R.J., Morgan, G.L. and Peterson, C.G., 2000, Quantum key distribution over a 48 km optical fibre
network, Journal of Modern Optics, 47(2/3), 533–547.
[109] Hughes, RJ., Nordholt, J.E., Derkacs, D. and Peterson, C.G., 2002, Practical free-space quantum key
distribution over 10 km on daylight and at night, New Journal of Physics, 4(43), 43.1–43.14.
[110] Jones, J.A. and Mosca, M., 1998, Implementation of a quantum algorithms to solve Deutsch’s problem on a
nuclear magnetic resonance quantum computer, Journal of Chemical Physics, 109, 1648–1653 http://xxx.
lanl.gov/abs/quant-ph/9801027.
[111] Joos, E. and Zeh, D., 1985, The emergence of classical properties through interaction with the environment,
Zeitschrift für physic B, 59, 223–243.
[112] Jozsa, R., 1991, Characterizing classes of functions computable by quantum parallelism, Proceedings of the
Royal Society of London A, 435, 563–574.
[113] Jozsa, R., Quantum effects in algorithms, http://xxx.lanl.gov/abs/quant-ph/9805086, May 29, 1998.
[114] Jozsa, R. and Linden, N., On the role of entanglement in quantum computational speed-up, http://aexiv.org/
abs/quant-ph/0201143, 2002.
[115] Kane, B.E., 1998, A silicon-based nuclear spin quantum computer, Nature, 393, 133 –137.
[116] Kielpinski, D., Monroe, C. and Wineland, D.J., 2002, Architecture for a large-scale ion-trap quantum
computer, Nature, 417, 709 –711.
[117] Kim, J., Lee, J.-S. and Lee, S., 2002, Experimental Realization of a target-accepting quantum search by NMR,
Physical Review A, 65, 054301.
[118] Kitaev, A.Y., Quantum measurements and the Abelian stabilizer problem, http://arxiv.org/abs/quant-ph/
9511026, November 1995.
[119] Kitaev, A.Y., Fault-tolerant quantum computation by anyons, http://xxx.lanl.gov/abs/quant-ph/9707021,
July 1997.
[120] Knill, E.H., Laflamme, R. and Milburn, G.J., 2001, A scheme for efficient quantum computation with linear
optics, Nature, 409, 46–52.
[121] Kwiat, P.G., et al., 1999, High-efficiency quantum interrogation measurements via the quantum Zeno effect,
Physical Review Letters, 83(23), 4725–4728.
[122] Kwiat, P.G., Weinfurter, H., Herzog, T. and Zeilinger, A., 1995, Interaction-free measurement, Physical
Review Letters, 74, 4763–4766.
[123] Ladd, T.D., et al. An all silicon quantum computer, http://arxiv.org/abs/quant-ph/0109039, September 7, 2001
[124] Laflamme, R., Miquel, C., Paz, J.P. and Zurek, W.H., Perfect quantum error correction code, http://arxiv.org/
abs/quant-ph/9602019, February 1996.
[125] Landauer, R., 1995, Is quantum mechanics useful?, Philosophical Transaction of the Royal Society of London
A, 353(1703), 367 –376.
[126] Lenstra, A.K., Lenstra, H.W. Jr (Eds.), 1993, The development of the Number Field Sieve (New York:
Springer-Verlag).
[127] Lloyd, S. and potentially, A., 1993, A potentially realizable quantum computer, Science, 261, 1569–1571.
[128] Lloyd, S., 1995, Almost any quantum logic gate is universal, Physical Review Letters, 75(2), 346–349.
[129] Lloyd, S., 1996, Universal quantum simulators, Science. 273(5278), 1073–1078.
[130] Long, GuiLu and Xiao, Li, Experimental realization of a fetching algorithm in a 7 qubit NMR quantum
computer, http: //arxiv.org/abs/quant-ph/0207079, July 15, 2002.
[131] Loss, D. and DiVincenzo, D.P., 1998, Quantum computation with quantum dots, Physical Review A, 57(1),
120–126 http: //arxiv.org/abs/cond-mat/9701055.
[132] Makhlin, Y., Scöhn, G. and Shnirman, A., 1999, Josephson-junction qubits with controlled couplings, Nature,
396, 305 –307.
[133] Marand, C. and Townsend, P., 1995, Quantum key distribution over distances as long as 30 km, Optics Letters,
20(16), 1695–1697.
[134] Marshall, T.W., Santos, E. and Selleri, F., 1983, Local realism has not been refuted by atomic cascade
experiments, Physics Letters A, 98, 5–9.
[135] Mattle, K., Weinfurter, H., Kwiat, P.G. and Zeilinger, A., 1996, Dense coding in experiment quantum
communication, Physical Review Letters, 76(25), 4656–4659.
[136] Mayers, D., The trouble with quantum bit commitment, http: //xxx.lanl.gov/abs/quant-ph/9603015, March
1996.
[137] Mayers, D., 1997, Unconditionally secure quantum bit commitment is impossible, Physical Review Letters,
78, 3414–3417.
[138] Merkle, R., 1978, Secure communications over insecure channels, Communications of the ACM, 21,
294–299.
[139] Mermin, N.D., From Cbits to Qbits: Teaching computer scientists quantum mechanics, http:
//arxiv.org/abs/quant-ph/0207118, July 2002.
[140] Miller, G.L., 1976, Riemann’s hypothesis and tests for primality, Journal of Computer and Systems Sciences,
13(3), 300–317.
[141] Monroe, C., et al., 1995, Demonstration of a fundamental quantum logic gate, Physical Review Letters,
75(25), 4714–4717.
[142] Muller, A., Zbinden, H. and Gisin, N., 1996, Quantum cryptography over 23 km in installed under-lake
telecom fibre, Europhysics Letters, 33, 335 –339.
[143] Nagy, M. and Akl, S.G., On the importance of parallelism for quantum computation and the concept of a
universal computer, Proceedings of the 4th International Conference on Unconventional Computation, UC
2005, Sevilla, Spain, October 2005, LNCS 3699, 176–190.
[144] Nelson, R.J., Gory, D.G. and Lloyd, S., 2000, Experimental demonstration of (Greenberger–Horne–Zeilinger
correlations using nuclear magnetic resonance, Physical Review A, 61, 022106 5 pages.
[145] Nielsen, M.A., 2003, Simple rules for a complex quantum world, Scientific American, 13(1), 24 –33 Special
edition: The edge of physics.
[146] Nielsen, M.A. and Chuang, I.L., 2000, Quantum Computation and Quantum Information (Cambridge:
Cambridge University Press).
[147] Nielsen, M.A., Knill, E.H. and Laflamme, R., 1998, Complete quantum teleportation using nuclear magnetic
resonance, Nature, 396(6706), 52– 55 http: //xxx.lanl.gov/abs/quant-ph/9811020.
[148] O’Brien, J.L., 2003, Demonstration of an all-optical quantum controlled-NOT gate, Nature, 426, 264–267.
[149] Pablo-Norman, B. and Ruiz-Altaba, M., 2000, Noise in grover’s quantum search algorithm, Physical
Review A, 61, 012301 5 pages.
[150] Peres, A., Error symmetrization in quantum computers, http: //xxx.lanl.gov/abs/quant-ph/9605009, May
1996.
[151] Pittman, T.B., Fitch, M.J., Jacobs, B.C. and Franson, D., 2003, Experimental controlled-NOT logic gate for
single Photons in the coincidence basis, Physical Review A, 68, 032316.
[152] Preskill, J., 1998, Fault-tolerant quantum computation. In: Hoi-Kwong Lo, S. Popescu and T. Spiller (Eds.)
Introduction to quantum computation and information (Singapore: http: //xxx.lanl.gov/abs/quant-ph/9712048
World Scientific), pp. 213– 269.
[153] Preskill, J., 1998, Reliable quantum computers, Proceedings of the Royal Society of London A, 454, 385 –440.
[154] Rabin, M.O., 1981, Technical Report TR-81, Aiken Computation Laboratory, Harvard University How to
exchange secrets by oblivious transfer.
[155] Rieffel, E. and Polak, W., 2000, An introduction to quantum computing for non-physicists, ACM Computing
Surveys, 32(3), 300–335.
[156] Rivest, R.L., Shamir, A. and Adleman, L.M., 1978, A method of obtaining digital signatures and public-key
cryptosystems, Communication of the ACM, 21(2), 120–126.
[157] Robertson, H.P., 1929, The uncertainty principle, Physical Review, 34, 163–164.
[158] Robinson, S., 2003, Emerging insights on limitations of quantum computing shape quest for fast algorithms,
SIAM News, 36(1).
[159] Schumacher, B., 1995, Quantum coding, Physical Review A, 51, 2738–2747.
[160] Schumacher, B. and Westmoreland, D., 1997, Sending classical information via noisy quantum channels,
Physical Review A, 56(1), 131–138.
[161] Shannon, C.E. and Weaver, W., 1949, The Mathematical Theory of Communication (Urbana: University of
Illinois Press).
[162] Shor, P.W., 1995, Scheme for reducing decoherence in quantum computer memory, Physical Review A, 52,
2493–2496.
[163] Shor, P.W., 1996, Fault-tolerant, quantum computation. Proceedings of the 37th Annual Symposium on
Foundations of Computer Science Los Alamitos, CA, 1996 (IEEE Computer Society Press), pp. 56–65.
[164] Shor, P.W., 1997, Polynomial-time algorithms for prime factorization and discrete logarithms on a quantum
computer, Special issue on Quantum Computation of the SIAM Journal on Computing, 26(5), 1484–1509.
[165] Simon, D.R., 1997, On the power of quantum computation, Special issue on Quantum Computation of the
SIAM Journal on Computing, 26(5), 1474–1483.
[166] Skinner, A.J., Davenport, M.E. and Kane, B.E., 2003, Hydrogenic spin quantum computing in silicon:
A digital approach, Physical Review Letters, 90, 087901.
[167] Smith, W.D., 1999, The energy-time uncertainty principle, http: //external.nj.nec.com/homepages/wds/e-
tuncert2TR.ps, typeset 811, September 21.
[168] Steane, A.M., 1996, Error correcting codes in quantum theory, Physical Review Letters, 77(5), 793 –797.
[169] Steane, A.M., 1996, Multiple particle interference and quantum error correction, Proceedings of the Royal
Society of London A, 452, 2551– 2576.
[170] Tapster, P.R., Rarity, J.C. and Owens, C.M., 1994, Violation of Bell’s inequality over 4 km of optical fibre,
Physical Review Letters, 73, 1923–1926.
[171] Teich, W., Obermayer, K. and Mahler, G., 1988, Structural basis for multistationary quantum systems ii:
Effective few-particle dynamics, Physical Review B, 37(14), 8111–8120.
[172] Thompson, C.H., The Chaotic Ball: An intuitive model for EPR experiments, http://xxx.lanl.gov/abs/
quant-ph/96l1037, November 1996.
[173] Thompson, C.H., Timing, “accidentals” and other artifacts in EPR experiments, http://xxx.lanl.gov/ahs/
qnant-ph/9711044, November 1997.
[174] Thompson, C.H., 1999, The tangled methods of quantum entanglement, experiments, Accountability in
Research, 6(4), 311–332.
[175] Tittel, W., Brendel, J., Gisin, B., Herzog, T., Zbinden, H. and Gisin, N., 1998, Experimental demonstration of
quantum-correlations over more than 10 kilometers, Physical Review A, 57, 3229–3232.
[176] Turchette, Q., et al., 1995, Measurement of conditional phase shifts for quantum logic, Physical Review
Letters, 75(25), 4710– 4713.
[177] Unruh, W.G., 1995, Maintaining coherence in quantum computers, Physical Review A, 51, 992 –997
http://arxiv.org/abs/hep-th/9406058.
[178] Vandersypen, L.M.K., Steffen, M., Breyta, G., Yannoni, C.S., Cleve, R. and Chuang, I.I., 2000, Experimental
realization of Shor’s quantum factoring algorithm with an NMR quantum computer, Physical Review Letters,
85(25), 5452–5455.
[179] Vandersypen, L.M.K., Steffen, M., Breyta, G., Yannoni, C.S., Sherwood, M.H. and Chuang, I.I., 2001,
Experimental realization of Shor’s quantum factoring algorithm using nuclear magnetic resonance, Nature,
414, 829 –938.
[180] Wiesner, S., 1983, Conjugate coding, SIGACT News, 15, 78– 88.
[181] Williams, C.P. and Clearwater, S.H., 1998, Explorations in quantum computing (New York: Springer-Verlag).
[182] Williams, C.P. and Clearwater, S.H., 2000, Ultimate zero and one: computing at the quantum frontier
(New York: Springer-Verlag).
[183] Williams, C.P. and Hogg, T., 1994, Exploiting the deep structure of constraint problems, Artificial Intelligence
Journal, 70, 73–117.
[184] Wootters, W.K. and Zurek, W.H., 1982, A single quantum cannot be cloned, Nature, 299, 802–803.
[185] Yao, A.C., 1993, Quantum circuit complexity. Los Alamitos, CA, 1993 Proceedings of the 34th Annual
Symposium on Foundations of Computer Science (IEEE Computer Society Press), pp. 352– 360.
[186] Zalka, C., 1999, Grover’s quantum searching algorithm is optimal, Physical Review A, 60, 2746– 2751
http://xxx.lanl.gov/abs/quaut-ph/9711070.
[187] Zurek, W.H., 1984, Reversibility and stability of information processing systems, Physical Review Letters, 53,
391–394.
View publication stats

Quantum Computation and Quantum Information

Uploaded by

Copyright:

Available Formats

Quantum Computation and Quantum Information

Uploaded by

Document Information

Original Title

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 Computation and Quantum Information

Uploaded by

Copyright:

Available Formats

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

Quantum computation and quantum information

Article in International Journal of Parallel Emergent and Distributed Systems · February 2006

The user has requested enhancement of the downloaded file.

Quantum computation and quantum information*

(Received 13 May 2005; in final form 8 July 2005)

1. Introduction and overview

1.1 Origins of quantum computing

*Corresponding author. Email: [email protected]

The International Journal of Parallel, Emergent and Distributed Systems

1.2 Interpretations of quantum mechanics

Figure 1. Young’s double-slit experiment.

2. Fundamentals of quantum information and quantum computation

2.1 The qubit

jCl ¼ aj0l þ bj1l; ð1Þ

Figure 2. The Bloch sphere representation of a qubit.

2.3 Uncertainty principle

2.5 Superposition and interference

2.6 Quantum registers

2.7.3 3-particle entanglement. Under these circumstances, conceiving a new kind of

2.7.4 Entanglement as a physical resource. Today, entanglement is no longer regarded as

2.8 Quantum evolution

2.9 Quantum gates

Other examples of single qubit gates are the Z gate:

Figure 3. Controlled-NOT quantum gate.

computation can be performed by connecting together multiple copies of a certain 2-qubit

2.10 Density operators

3. Quantum information theory

3.1 Classical information through quantum channels

3.2 Quantum information through quantum channels

3.3 Superdense coding

Figure 4. Superdense coding.

3.4 Quantum teleportation

Superdense coding is a protocol for transmitting classical information through a quantum

4.1 Private-key systems

4.2 Public-key systems

4.3 Cracking RSA codes

4.4 Quantum key distribution

Figure 5. Quantum key distribution in the absence of eavesdropping.

4.4.1 Experimental demonstrations. The first practical demonstration of their quantum

4.5 Quantum solutions to discreet decision problems

4.6 Entanglement-based cryptography

However, analog to the two-fold influence of quantum mechanics upon cryptography,

5.1 Algorithms using quantum Fourier transforms

Expressing the effect of applying this Quantum Fourier Transform on a superposition of

X X ð21Þx1 z1 þ· · ·þxn zn þf ðx1 ...xn Þ jz1 . . .zn l j0l 2 j1l

Figure 6. Quantum circuit implementing the Deutsch–Jozsa algorithm.

A closer look to the amplitude for the state j0l^n, namely

5.2 Quantum search algorithms

5.2.2 Heuristic quantum search algorithms. For completely unstructured searches,

Figure 7. Lattice of variable assignments.

Figure 8. Lattice of variable assignments in ket form.

5.3 Quantum simulations

Similar to classical complexity theory, quantum complexity theory is concerned with

6.2 Quantum Turing Machines

In order to be able to prove theorems about further capabilities of quantum computers,

Figure 9. Quantum Toffoli (controlled–controlled-NOT) gate.

In terms of computability, we have already seen that anything computable by a classical

Figure 10. Relationships between quantum and classical complexity classes.

X X ð21Þx1 z1 þ· · ·þxn zn þf ðx1 ...xn Þ jz1 . . .zn l j0l 2 j1l