Information and Computation 165, 174–182 (2001)

doi:10.1006/inco.2000.2911, available online at http://www.idealibrary.com on

Regular Languages Accepted by Quantum Automata

Alberto Bertoni and Marco Carpentieri

Dipartimento di Scienze dell’Informazione, Università di Milano, Via Comelico 39, 20135 Milan, Italy

Received January 21, 1999

In this paper we analyze some features of the behaviour of quantum automata. In particular we
prove that the class of languages recognized by quantum automata with isolated cut point is the class
of reversible regular languages. As a more general result, we give a bound on the inverse error that
implies the regularity of the language accepted by a quantum automaton. °C 2001 Academic Press

1. INTRODUCTION

Even if the present technology does consent to realize only very simple devices based on the prin-
ciples of quantum mechanics, many authors considered it worth asking whether a theoretical model
of quantum computation could offer any substantial benefits over the correspondent theoretical model
based on the assumptions of classical physics. This question has recently received considerable attention
because of the growing belief that quantum mechanical processes might be able to perform computation
that traditional computing machines can only perform inefficiently. For an extensive bibliography and
illustration of the main results in the area the reader is referred to [3, 6, 19, 21, 35, 36, 39].
In 1982, Benioff [2] first considered that devices computing according to the principles of quantum
mechanics could be at least as powerful as classical computers. The question whether the computational
power of quantum mechanical processes might be beyond that of traditional computation models was
raised by Feynmann [22] who gave arguments as to why quantum mechanics might be computationally
expensive to simulate on a classical computer. In 1985, Deutsch [16] re-examined the Church Turing prin-
ciple, on which the current computational complexity theory is founded, and he proposed a precise model
of a quantum physical computer, thus defining quantum Turing machines. Then, Deutsch [17] defined
quantum networks and investigated some of their properties. Bernstein and Vazirani [6] gave the founda-
tions of the quantum theory of computational complexity and described an efficient universal quantum
computer that simulates a large class of quantum Turing machines. Yao [39] introduced the quan-
tum complexity theory in terms of quantum networks and showed the existence of an efficient quantum
simulator for each quantum Turing machine.
Several authors offered evidence that the quantum model of computation may have significantly
more complexity theoretic power than traditional Turing machines [6, 8, 9, 18, 22, 23, 35, 36] and [36].
Berthiaume and Brassard [8, 9] and Deutsch and Jozsa [18] introduced problems that quantum computers
can quickly solve exactly, while classical computers can only solve quickly with a bounded probability
of error. Bernstein and Vazirani [6] proposed an oracle problem that can be solved in polynomial time
by quantum computation, but requires superpolynomial time on a classical machine. This result was
improved by Simon [36] who gave a simpler construction of an oracle problem that takes polynomial
time by quantum computation, but exponential time on a classical computer. Simon’s algorithm inspired
the work of Shor [35] that presented quantum polynomial time algorithms for the discrete logarithm
and integer factoring problems that, as it is well known, are unlikely to be solvable in polynomial time
by classical computation. Indeed, the integer factoring is so widely believed hard that the RSA public
cryptosystem [34] is based on the assumption of its hardness.
Although some suggestions have been made to design quantum computers [12, 13, 20, 29, 30, 37,
38], there are substantial difficulties in building any of these because of the destabilizing effects of
the environmental interaction that is a major experimental (and theoretical) obstacle. Such difficulties
become very serious as the computation time and the size of the computer grow so that it is conceivable
to build only small or very simple quantum machines.
174

0890-5401/01 $35.00
Copyright ° C 2001 by Academic Press

All rights of reproduction in any form reserved.

 QUANTUM AUTOMATA 175

The problems of the destabilizing effects of the interaction with an environment suggest the study of
quantum devices, simpler than quantum machines, such as those corresponding to classical automata,
that can be experimentally useful to better understand and possibly control quantum phenomena. Quan-
tum automata were first introduced in [15] and [28], but with different definitions. More particularly,
Critchfield and Moore [15] defined such computation models as the correspondent probabilistic ones,
while Kondacs and Watruos considered an automaton in which measurements are performed at every
computation step and the control states are classified in nonhalting, accepting, and rejecting states. Here,
we shall consider as basic definition that of Critchfield and Moore [15].
A finite control state Quantum Automaton can be viewed as a particular quantum Turing machine,
where the head moves only to the right reading and writing the same symbol. The states of a quantum
automaton QA with m control states {1, . . . , m} can be described as unit length m-dimensional complex
(column) vectors whose kth component represents the amplitude of the control state k (1 ≤ k ≤ m).
We recall that an observation of the state υ = (v1 , . . . , vm )T ∈ Cm produces the control state k with
probability |vk |2 . The possible input messages are words over a finite alphabet 6; the input symbol σ ∈ 6
causes a change of state according to a unitary transformation M(σ ) : Cm → Cm such that M(σ )(υ) =
M(σ )υ. Fixed in initial state α, a word σ1 · · · σn ∈ 6 ∗ determines a new state υ 0 = M(σ1 ) · · · M(σn )α.
The probabilistic event realized by QA is defined by the probability PQA (σ1 · · · σn ) that the control state
observed from υ 0 belongs to a preassigned set F of final control states. Given a cut point λ ∈ [0, 1), the
behaviour of the quantum automaton QA can be defined by the language L QA,λ containing the input
words σ1 · · · σn for which p(σ1 · · · σn ) > λ. An important notion associated to the automaton QA with
cut point λ is the error function ²QA,λ : N → [0, 1] that represents the difference between the minimum
probability of an accepted word of length at most n and the maximum probability of a rejected word
−1
of length at most n. The inverse error ²QA,λ (n), for ²QA,λ (n) 6= 0, is an estimation of the number of
repetitions of an experiment to decide the correct membership of a word of length at most n with high
confidence. If there is ² > 0 for which ²QA,λ (n) ≥ ², we say that the cut point λ is isolated (notice that
this definition is slightly different from that introduced by Rabin [33]).
In general, quantum automata can accept nonregular languages. In fact, we are able to exhibit a
quantum automaton accepting a nonregular language with inverse error polynomially bounded. Nev-
ertheless, in this paper we consider only quantum automata accepting regular languages. We prove
that the class of languages recognized by quantum automata with isolated cut point is a subclass of
the regular languages and more precisely that of the reversible regular languages [32] (this result was
independently and simultaneously found by Brodsky and Pippenger [11] in “Characterization of 1-way
Quantum Finite Automata,” http://xxx.lanl.gov/abs/quant-ph/9903014, even if for a slightly less general
definition, equivalent to that for probabilistic automata in Rabin’s sense [33]). As a more general result,
we give a bound on the polynomial growth of the inverse error that implies the regularity of the language
accepted by a quantum automaton.

2. PRELIMINARIES

In this section we review the basic concepts used in the rest of the paper. For a more exhaustive
illustration of the topics presented here the reader is referred to [14, 24–26, 31].

2.1. Metric Spaces

A metric space hD, di consists of a set D of elements and a distance d, i.e., a single-valued nonnegative
real function d : D × D → R satisfying the following properties

1. d(x, y) = 0 iff x = y,
2. d(x, y) = d(y, x),
3. d(x, z) ≤ d(x, y) + d(y, z),

for all x, y, z ∈ D. A sequence {xn }n∈N of elements in D converges to x ∈ R iff

lim d(xn , x) = 0,
n→∞
176 BERTONI AND CARPENTIERI

while {xn }n∈N is a Cauchy sequence if, given any ² > 0, there is an integer N² such that d(xn , xn 0 ) < ²
for all n, n 0 > N² . It is well known that the convergence of {xn }n∈N implies that {xn }n∈N is a Cauchy
sequence. A metric space is said to be compact when from each sequence {xn }n∈N of elements in D we
can extract a convergent subsequence {xn k }k∈N . The following theorem holds.
THEOREM 1. Let δ > 0. If < D, d > is compact, a sequence {xn }n∈N of infinite distinct elements in
D such that δ(x n , xn 0 ) ≥ δ for n 6= n 0 and n, n 0 ∈ N does not exist.

2.2. Quantum Automata

Let h6 ∗ , ·, 1i be the free monoid generated by a finite alphabet 6 consisting of the words over 6
along with concatenation product and the empty word 1. We denote the length of the word w ∈ 6 ∗ by
|w|, while by 6 ≤n we mean the set of the words in 6 ∗ of length at most n.
DEFINITION 1. A quantum automaton QA with m control states over 6 is a system

QA = hα, {M(σ ), σ ∈ 6}, Fi,

where α is a (column) vector in Cm such that kαk = 1, M(σ ) defines a unitary transformation M(σ ) :
Cm → Cm , and F ⊆ {1, . . . , m}.
By M(σ1 · · · σl ) : Cm → Cm (σ1 · · · σl ∈ 6 ∗ ) we mean the transformation

Y
l−1
M(σ1 · · · σl ) = M(σl− j ).
j=0

The stochastic event generated by QA is the function

PQA : 6 ∗ → [0, 1]

defined by
X
PQA (w) = |(M(w)α)k |2 .
k∈F

The language L QA,λ accepted by QA with cut point λ ∈ [0, 1) is

L QA,λ = {w : PQA (w) > λ}.

Given a quantum automaton QA and λ ∈ [0, 1), the error function ²QA,λ : N → [0, 1] is defined by

 min |PQA (w) − λ| if L QA,λ ∩ 6 ≤n = ∅ or L cQA,λ ∩ 6 ≤n = ∅
 w:|w|≤n
²QA,λ (n) =

 w∈Lmin∩6 ≤n PQA (w) − w0 ∈Lmax PQA (w0 ) otherwise.
QA,λ
c
∩6 ≤n
QA,λ

Moreover, when there exists ² > 0 such that ²QA,λ (n) ≥ ² for every n ∈ N, then λ is said to be isolated
with respect to QA. Notice that the definition of isolated cut point introduced by Rabin [33] implies the
definition given in this paper.

3. QUANTUM AUTOMATA ACCEPTING REGULAR LANGUAGES

This section is dedicated to find conditions on the error function ²QA,λ implying that the language
L QA,λ accepted by a quantum automaton QA with m control states over 6 is regular. First, we prove
the following lemma.
 QUANTUM AUTOMATA 177

LEMMA 1. Let w ∈ L QA,λ and w0 6∈ L QA,λ be words in 6 ∗ of length at most n. Then it holds

1
kM(w)α − M(w 0 )αk > ²QA,λ (n).
2m

Proof. Set M(w)α = (v1 , . . . , vm )T and M(w0 )α = (v10 , . . . , vm0 )T . We know that
X X
|vk |2 − |vk0 |2 ≥ ²QA,λ (n)
k∈F k∈F

and, consequently, there is an index k (1 ≤ k ≤ m) such that

1
kvk |2 − |vk0 |2 | ≥ ²QA,λ (n). (1)
m

The proof follows by observing that

1
²QA,λ (n) ≤ kvk |2 − |v0k |2 | (by (1))
m
= (|vk | + |v0k |)kvk | − |v0k k
< 2kvk | − |v0k k (since |vk |, |v0k | ≤ 1 and ²QA,λ (n) 6= 0)
≤ 2|vk − v0k |
v
uX
u m
≤ 2t |v j − v0j |2
j=1

= 2kM(w)α − M(w 0 )αk. j

Suppose that L QA,λ 6= ∅ and L QA,λ 6= 6 ∗ . Then, there exists a finite n ∈ N such that L QA,λ ∩ 6 ≤n 6= ∅
and L cQA,λ ∩ 6 ≤n 6= ∅. For a fixed such n ∈ N, our aim is to construct a deterministic finite state
automaton DA that recognizes a language L DA such that

L QA,λ ∩ 6 ≤n = L DA ∩ 6 ≤n .

In this regard, some preliminary definitions are useful.

DEFINITION 2. Given a quantum automaton QA, we say that two words w, w0 ∈ 6 ∗ are ²-connected
within length l if there is a sequence w1 , . . . , wr of words such that

1. w1 = w, wr = w0 , |wk | ≤ l (1 ≤ k ≤ r ),
2. kM(wk+1 )α − M(wk )αk ≤ ² (1 ≤ k ≤ r − 1).

In the case in which w1 = w, wr = w0 and Property 2 holds we shall simply say that w and w0 are
²-connected by the sequence w1 , . . . , wr .

DEFINITION 3. By R we mean the binary relation on 6 ≤n such that wRw0 iff w, w0 are 1
²
2m QA,λ
(2n)-
connected within length n.

It is easy to verify that R is an equivalence relation; denote by [w]R the equivalence class containing
w. Consider the following total order in 6 ≤2n :

• w ≤ w0 iff |w| < |w0 | or |w| = |w 0 | and w is less than or equal to w0 in the lexicographical
order.
178 BERTONI AND CARPENTIERI

It is convenient to consider as representative of the class [w]R the minimum element min[w]R with respect
to the total order ≤. We are now able to define the following deterministic automaton DA over 6,

DA = hQ, q0 , δ, Fi,

where
• the set Q of control states consists of the representatives of the equivalence classes defined by
R; i.e., Q = {min[w]R : w ∈ 6 ≤n },
• the initial control state is q0 = 1,
• the transition function δ : Q × 6 → Q is defined by δ(q, σ ) = min[qσ ]R when |qσ | ≤ n and
δ(q, σ ) = 1 when |qσ | > n,
• the set F of final control states is F = {q : q ∈ L QA,λ }.
The main feature of the automaton DA is stated by the next lemma.
LEMMA 2. If |w| ≤ n then w and δ(1, w) are 1
²
2m QA,λ
(2n)-connected within length n + |w|.
Proof. We prove the lemma by induction on the length |w| of a word w ∈ 6 ≤n . If |w| = 0, i.e.,
w = 1, the proof is immediate. Suppose that |wσ | ≤ n and that w is 2m
1
²QA,λ (2n)-connected to δ(1, w)
within length n + |w| by the sequence w1 , . . . , wr , where w = w1 and δ(1, w) = wr . Then wσ is
1
²
2m QA,λ
(2n)-connected to δ(1, w)σ by w1 σ, . . . , wr σ since

kM(wk σ )α − M(wk+1 σ )αk = kM(σ )(M(wk ) − M(wk+1 ))αk

= kM(wk )α − M(wk+1 )αk (since M(σ ) preserves length)
1
≤ ²QA,λ (2n)
2m

for 1 ≤ k < r. Hence wσ and δ(1, w)σ are 2m 1

²QA,λ (2n)-connected within n +|wσ |. Moreover, δ(1, w)σ
and δ(1, wσ ) are 2m ²QA,λ (2n)-connected within n since δ(1, wσ ) ∈ [δ(1, w)σ ]R . We conclude that wσ
1
1
is 2m ²QA,λ (2n)-connected to δ(1, wσ ) within n + |wσ |. j
We are now ready to prove the main result.
LEMMA 3. It results that

L QA,λ ∩ 6 ≤n = L DA ∩ 6 ≤n .

Proof. Suppose there is w ∈ 6 ≤n such that w ∈ L QA,λ but w 6∈ L DA . By Lemma 2 w and δ(1, w) are
1
²
2m QA,λ
(2n)-connected within length 2n by a sequence w1 , . . . , wr . We have that w1 = w ∈ L QA,λ but
wr = δ(1, w) 6∈ L QA,λ since w 6∈ L DA . This implies that there is k (1 ≤ k < r ) such that wk ∈ L QA,λ ,
wk+1 6∈ L QA,λ , and |wk |, |wk+1 | ≤ 2n. But then

1
kM(wk )α − M(wk+1 )αk ≤ ²QA,λ (2n),
2m

which contradicts Lemma 1. The proof is similar in the case in which w 6∈ 6 ≤n ∩ L QA,λ and w ∈
6 ≤n ∩ L DA . j
P
DEFINITION 4. Let USm = {υ ∈ Cm : mj=1 |v j |2 = 1} be the unitary sphere in Cm . An ²-net for
USm is a subset ξ of points in USm such that for every element υ ∈ USm there is an element υ 0 ∈ ξ
for which kυ − υ 0 k ≤ ². A large upper bound on the minimum cardinality N (m, ²) of an ²-net for the
unitary sphere USm is

(2m)m−1
N (m, ²) < . (2)
² 2m−1
 QUANTUM AUTOMATA 179

DEFINITION 5. Given a language L ⊆ 6 ∗ , let

C L (n) = min{|A| : L A ∩ 6 ≤n = L ∩ 6 ≤n },
A

where for a deterministic finite state automaton A |A| is the number of control states and L A is the
language recognized.
If L is regular, then the number of control states of the minimum deterministic finite state automaton
for L is an upper bound for C L (n). Conversely, if L is not regular a linear lower bound for C L (n) was
stated by Karp [27] in 1967.
Fact 1. If L is not regular, then C L (n) ≥ n+3
2
for infinite n.
We are able to give the following bound on C L QA,λ (n).
THEOREM 2. It holds that
23m−1 m m
C L QA,λ (n) ≤ ¡ ¢2m−1 .
1
²
2m QA,λ
(2n)

Proof. By Lemma 3 C L QA,λ (n) is at most the number of control states of the automaton DA, i.e., the
1/2m ²QA,λ (2n)
number of equivalence classes of the relation R. Such a number is at most N (m, 2
) and the
result follows by 2. j
We can conclude the section with a condition on ²QA,λ (n) implying that the language L QA,λ is regular.
THEOREM 3. If for each n ∈ N ²QA,λ (n) = Ä(n −1/(2m−1) ), where m is the number of control states
of QA, then L QA,λ is regular.
Proof. It is an immediate consequence of Fact 1 and of Theorem 2. j

Remark. We are able to exhibit a quantum automaton accepting the nonregular language {w : #a (w) 6=
#b (w)}, where {a, b} is the input alphabet and by #x (w) we mean the number of occurrences of symbol
x in word w. The automaton has inverse error polynomially bounded and accepts the language with
cut point 0. It is well known that stochastic automata [31] with cut point 0 can accept only regular
languages.

4. LANGUAGES ACCEPTED WITH ISOLATED CUT POINT

Theorem 3 implies that the language L QA,λ accepted by a quantum automaton QA with isolated cut
point is regular. In this section we show that L QA,λ is regular reversible. First, we prove the following
lemma.
LEMMA 4. If M is any unitary matrix of order m over the complex field and Im is the identity matrix
over Cm , then for any ² > 0 there exists ν ∈ N such that it holds that

kM ν − Im k ≤ ².

Proof. Consider the linear space of matrices of order m over the complex field with norm kMk =
supυ∈USm kMυk. Each unitary matrix M is such that kMk = 1; moreover, the set of unitary matrices
along with the distance d(M, M 0 ) = kM − M 0 k is a compact metric space. Then, from the sequence
{M n }n∈N we can extract a Cauchy sequence {M n k }k∈N ; i.e., for each ² > 0 there exists ν² ∈ N such that
n k1 , n k2 > ν² implies kM n k2 − M n k1 k ≤ ². Fix n k2 > n k1 > ν² and set ν = n k2 − n k1 . We have

kM ν − Im k = kM −n k1 M n k1 +ν − M −n k1 M n k1 k
= kM −n k1 (M n k1 +ν − M n k1 )k
= kM n k1 +ν − M n k1 k (since M −n k1 preserves length)
≤ ². j
180 BERTONI AND CARPENTIERI

Let L QA,λ be accepted by a quantum automaton QA with isolated cut point; i.e., there is ² > 0 such
that ²QA,λ (n) > ² for every n ∈ N and let R be the binary relation on 6 ∗ such that wRw 0 iff w and w0
²
are 2m -connected by a sequence of words w1 , . . . , wr ∈ 6 ∗ . The next lemmas state the main properties
satisfied by R.

LEMMA 5. The relation R is an equivalence relation on 6 ∗ of finite index.

Proof. The unitary sphere USm along with the distance d : USm × USm → R defined by d(υ, υ 0 ) =
kυ − −υ 0 k is a compact metric space. To check that R is an equivalence relation on 6 ∗ is straight-
forward. Moreover, if R were not of finite index, we could find infinite points υ1 , . . . , υr , . . . in
²
USm such that d(υ j , υ j 0 ) > 2m for any distinct indices j, j 0 ∈ N and the metric space would not be
compact. j

LEMMA 6. For each σ ∈ 6 and w, w0 ∈ 6 ∗ , wRw0 if and only if wσ Rw0 σ.

Proof. If wRw 0 , there are w1 , . . . , wr ∈ 6 ∗ such that w = w1 , w 0 = wr , and kM(wk )α −

²
M(wk+1 )αk ≤ 2m , for 1 ≤ k < r. If σ ∈ 6 we have

kM(wk σ )α − M(wk+1 σ )αk = kM(σ )(M(wk ) − M(wk+1 ))αk

= kM(wk )α − M(wk+1 )αk (since M(σ ) preserves length)
²
≤ (1 ≤ k < r ).
2m

Therefore, it follows that wσ Rw0 σ.

Suppose now that wσ Rw0 σ for some σ ∈ 6. Thus, there are w1 , . . . , wr 0 ∈ 6 ∗ such that w1 = wσ,
wr 0 = w0 σ , and

²
kM(wk )α − M(wk+1 )αk ≤ (1 ≤ k < r 0 ). (3)
2m

By Lemma 4, we can find a positive integer ν such that

²
kM ν (σ ) − Im k ≤ . (4)
2m

Set w10 = w, wr0 0 +2 = w 0 and wk0 = wk−1 σ ν−1 , for 2 ≤ k ≤ r 0 + 1. The proof follows by noticing that
w10 = w, wr0 0 +2 = w0 , and w10 , . . . , wr0 0 +2 are such that

kM(w10 )α − M(w20 )αk = kM(w)α − M(wσ ν )αk

= k(Im − M ν (σ ))M(w)αk
≤ kIm − M ν (σ ))k kM(w)αk
= kIm − M ν (σ )k (since kM(w)αk = 1)
²
≤ (by (4)),
2m

kM(wk0 )α − M(wk+1
0
)αk = kM(wk−1 σ ν−1 )α − M(wk σ ν−1 )αk
= kM ν−1 (σ )(M(wk−1 ) − M(wk ))αk
= kM(wk−1 )α − M(wk )αk (since M ν−1 (σ ) preserves length)
²
≤ (by (3)),
2m
 QUANTUM AUTOMATA 181

for 2 ≤ k ≤ r 0 ,

kM(wr0 0 +1 )α − M(wr0 0 +2 )αk = kM(w0 σ ν )α − M(w 0 )αk

= k(M ν (σ ) − Im )M(w0 )αk
≤ kM ν (σ ) − Im k kM(w0 )αk
= kM ν (σ ) − Im k (since kM(w0 )αk = 1)
²
≤ (by (4)). j
2m

We are now able to introduce the following deterministic automaton DA over 6,

DA = hQ, q0 , δ, Fi,

where
• the set Q of control states consists of the representatives of the equivalence classes defined by
R, i.e., Q = {min[w]R : w ∈ 6 ≤n },
• the initial control state is q0 = 1,
• the transition function δ : Q × 6 → Q is defined by δ(q, σ ) = min[qσ ]R ,
• the set F of final control states is F = {q : q ∈ L QA,λ }.
By arguments similar to those used to prove Lemmas 1–3 it can be directly proved that DA recognizes
L QA,λ . Moreover, the following lemma holds.
LEMMA 7. The automaton DA is reversible.
Proof. Suppose that δ(w, σ ) = δ(w0 , σ ) for some σ ∈ 6, where w, w0 ∈ 6 ∗ are representatives
of two distinct equivalence classes of R. Then, wσ Rw 0 σ would imply w R6 w0 and Lemma 6 would be
contradicted. Thus, for each σ ∈ 6, the finite transformation individuated by δ is injective and DA is
reversible. j
The main result of this section is now straightforward.
THEOREM 4. The class of languages accepted by quantum automata with isolated cut point is the
class of the reversible regular languages.

REFERENCES

1. Barenco, A., Bennet, 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 Phys. Rev. A 52, 3457–3467.
2. Benioff, P. (1982), Quantum mechanical hamiltonian models of turing machines, J. Statist. Phys. 29, 515–546.
3. Bennet, C. H. (1995), Quantum computation and information, Physics Today 48(10), 24–30.
4. Bennet, C. H. (1973), Logical reversibility of computation, IBM J. Res. Develop. 17, 525–532.
5. Bennet, C. H., Bernstein, E., Brassard, G., and Vazirani, U. (1994), Strengths and weaknesses of quantum computing, SIAM
J. Comput.
6. Bernstein, E., and Vazirani, U. (1993), Quantum complexity theory in “Proc. 25th ACM Symp. on Theory of Computation,”
pp. 11–20.
7. Berstel, J., and Retenauer, C., “Rational Series and Their Languages,” Springer-Verlag, New York, 1988.
8. Berthiaume, A., and Brassard, G. (1992), The quantum challenge to structural complexity theory, in “Proc. 7th IEEE Con-
ference on Structure in Complexity Theory.”
9. Berthiaume, A., and Brassard, G. (1992), Oracle quantum computing, in “Proc. Physics of Computation.”
10. Bertoni, A., and Carpentieri, M., Analogies and differences between quantum and stochastic automata, Theoret. Comput.
Sci., to appear.
11. Brodsky, A., Pippenger, N., Characterization of 1-way quantum finite automata, http://xxx.lanl.gov/abs/quant-ph/9903014.
12. Chuang, L., and Yamamoto, Y. (1995), A simple quantum computer, Phys. Rev. A 52, 3489–3496.
13. Cirach, J. I., and Zoller, P. (1995), Quantum computation with cold trapped ions, Phys. Rev. Lett. 74, 4091–4094.
14. Cohen, D. W. (1989), “An Introduction to Hilbert Space and Quantum Logic,” Springer-Verlag, New York.
182 BERTONI AND CARPENTIERI

15. Crutchfield, J. P., and Moore, C., Quantum automata and quantum grammars, http://xxx.lanl.gov/abs/quant-ph/9707031.
16. Deutsch, D. (1985), Quantum theory, the Church Turing principle and the universal quantum computer, Proc. Roy. Soc.
London A 400, 73–90.
17. Deutsch, D. (1889), Quantum computational networks, Proc. Roy. Soc. London A 425, 73–90.
18. Deutsch, D., and Jozsa, R. (1992), Rapid solutions of problems by quantum computation, Proc. Roy. Soc. London A 439,
553–555.
19. DiVincenzo, D. P. (1995), Quantum computation, Science 269, 256–261.
20. DiVincenzo, D. P. (1995), Two-bit gates are universal for quantum computation, Phys. Rev. A 51, 1015–1022.
21. Ekert, A., and Jozsa, R. (1996), Quantum computation and Shor’s factoring algorithm, Rev. Modern Phys. 68, 733–754.
22. Feynman, R. (1982), Simulating physics with computers, Internat. J. Theoret. Phys. 21, 467–488.
23. Feynman, R. (1986), Quantum mechanical computers, Foundations Phys. 16, 507–531.
24. Fomin, S. V., and Kolmogorov, A. N. (1975), “Introductory Real Analysis,” Dover, New York, 1975.
25. Hopcroft, J. E., and Ullman, J. D. (1979), “Introduction to Automata Theory, Languages and Computation,” Addison-Wesley,
Reading, MA.
26. Hughes (1989), “The Structure and Interpretation of Quantum Mechanics,” Harvard University Press, New Haven, CT.
27. Karp, R. M. (1967), Some bounds on the storage-requirement of sequential machines and Turing machines, J. Assoc. Comput.
Mach. 14, 478–489.
28. Kondacs, A., and Watrous, J. (1997), On the power of quantum finite state automata, in “Proc. 38th FOCS,” pp. 66–75.
29. Lloyd, S. (1993), A potentially realizable quantum computer, Science 261, 1659–1571.
30. Lloyd, S. (1994), “Envisioning a quantum supercomputer,” Science 263, 695.
31. Paz, A. (1971), “Introduction to Probabilistic Automata,” Academic Press, New York/London.
32. Pin, J. E. (1987), On the languages recognized by finite reversible automata, in “14th ICALP,” Lecture Notes in Computer
Science, pp. 237–249, Springer-Verlag, Berlin/New York.
33. Rabin, M. O. (1964), Probabilistic Automata, in “Sequential Machines,” Addison-Wesley, Reading, MA.
34. Rivest, R. L., Shamir, A., and Adelman, L. (1978), A method of obtaining digital signatures and public-key cryptosystems,
Comm. Assoc. Comput. Mach. 21, 120–126.
35. Shor, P. (1994), Algorithms for quantum computation: Discrete log and factoring, in “Proc. 35th IEEE Symp. on Foundations
of Computer Science.”
36. Simon, D. (1994), On the power of quantum computation, in “Proc. of 35th IEEE Symp. on Foundations of Computer
Science.”
37. Sleator, T., and Weinfurther, H. (1995), Realizable quantum logic gates, Phys. Rev. Lett. 74, 4087–4090.
38. Teich, G., Obermayer, K., and Mahler, G. (1988), Structural basis of multistationary quantum systems II: Effective few
particle dynamics, Phys. Rev. B 37, 173–192.
39. Yao, A. (1993), Quantum circuit complexity, in “Proc. 34th IEEE Symp. on Foundations of Computer Science.”