0% found this document useful (0 votes)
81 views25 pages

Simulations of Quantum Turing Machines by Quantum Multi-Counter Machines

This document proposes simulating quantum Turing machines (QTMs) using quantum multi-stack machines (QMSMs) and quantum multi-counter machines (QMCMs). It defines QMSMs and QMCMs, provides well-formedness conditions for their unitary evolution, and shows how to efficiently simulate QTMs using these models with only a polynomial time slowdown. The main goals are to establish QMSMs and QMCMs as quantum computing devices and use them to simulate QTMs in a time-efficient manner similar to how classical machines can simulate Turing machines.

Uploaded by

Gabriel Senno
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
81 views25 pages

Simulations of Quantum Turing Machines by Quantum Multi-Counter Machines

This document proposes simulating quantum Turing machines (QTMs) using quantum multi-stack machines (QMSMs) and quantum multi-counter machines (QMCMs). It defines QMSMs and QMCMs, provides well-formedness conditions for their unitary evolution, and shows how to efficiently simulate QTMs using these models with only a polynomial time slowdown. The main goals are to establish QMSMs and QMCMs as quantum computing devices and use them to simulate QTMs in a time-efficient manner similar to how classical machines can simulate Turing machines.

Uploaded by

Gabriel Senno
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 25

Simulations of Quantum Turing Machines by

Quantum Multi-Stack Machines

Daowen Qiu∗
arXiv:quant-ph/0501176v2 6 Jun 2005

Department of Computer Science, Zhongshan University, Guangzhou, 510275, P.R. China

Abstract: As was well known, in classical computation, Turing machines, circuits,


multi-stack machines, and multi-counter machines are equivalent, that is, they can simulate
each other in polynomial time. In quantum computation, Yao [11] first proved that for
any quantum Turing machines M , there exists quantum Boolean circuit (n, t)-simulating
M , where n denotes the length of input strings, and t is the number of move steps before
machine stopping. However, the simulations of quantum Turing machines by quantum multi-
stack machines and quantum multi-counter machines have not been considered, and quantum
multi-stack machines have not been established, either. Though quantum counter machines
were dealt with by Kravtsev [6] and Yamasaki et al. [10], in which the machines count with
0, ±1 only, we sense that it is difficult to simulate quantum Turing machines in terms of
this fashion of quantum computing devices, and we therefore prove that the quantum multi-
counter machines allowed to count with 0, ±1, ±2, . . . , ±n for some n > 1 can efficiently
simulate quantum Turing machines.

So, our mail goals are to establish quantum multi-stack machines and quantum multi-
counter machines with counts 0, ±1, ±2, . . . , ±n and n > 1, and particularly to simulate
quantum Turing machines by these quantum computing devices. The major technical con-
tributions of this article are stated as follows:

(i) We define quantum multi-stack machines (abbr. QMSMs) by generalizing a kind of


quantum pushdown automata (abbr. QPDAs) from one-stack to multi-stack, and the well-
formedness (abbr. W-F) conditions for characterizing the unitary evolution of the QMSMs
are presented.

(ii) By means of QMSMs we define quantum multi-counter machines (abbr. QMCMs)



Email-address: [email protected] (D. Qiu).

1
whose state transition functions are different from the quantum counter automata (abbr.
QCAs) in the literature; as well, the W-F conditions are given for these defined QMCMs.

(iii) To simulate quantum Turing machines (abbr. QTMs), we deal with a number of
simulations between QMCMs with different counters and different counts. Therefore, we
show that any QMCM allowed to count with ±n for n > 1 can be simulated by another
QMCM that counts with 0, ±1 only.

(iv) In particular, we demonstrate the efficient simulations of QTMs in terms of QMSMs,


and the simulation of QMCMs by QMSMs with the same time complexity. Therefore, we
show that QTMs can be simulated by QMSMs as well. To conclude, a number of issues are
proposed for further considerations.

Keywords: Quantum Computation; Quantum Turing Machines; Quantum Multi-Counter


Machines; Quantum Multi-Stack Machines

1. Introduction

1.1. Motivation and purpose

Quantum computing is an intriguing and promising research field, which touches on quantum
physics, computer science, and mathematics [4]. To a certain extent, this intensive attention
given by the research community originated from Shor’s findings of quantum algorithms for
factoring prime integers in polynomial time and Grover’s algorithm for searching through a
database which could also be sped up on a quantum computer [4].

Let us briefly recall the work of pioneers in this area. (Due to limited space, the detailed
background and related references are referred to [4].) In 1980, Benioff first considered that
the computing devices in terms of the principles of quantum mechanics could be at least
as powerful as classical computers. Then Feynman pointed out that there appears to be
no efficient way of simulating a quantum mechanical system on a classical computer, and
suggested that a computer based on quantum physical principles might be able to carry out
the simulation efficiently. In 1985 Deutsch re-examined the Church-Turing Principle and
defined QTMs. Subsequently, Deutsch considered quantum network models.

Quantum computation from a complexity theoretical viewpoint was studied systematically


by Bernstein and Vazirani [1] and they described an efficient universal QTM that can simulate

2
a large class of QTMs. Notably, in 1993 Yao [11] demonstrated the equivalence between QTMs
and quantum circuits. More exactly, Yao [11] showed that any given QTM, there exists a
quantum Boolean circuit (n, t)-simulating this QTM with polynomial time slowdown, where
n denotes the length of input strings, and t is the number of move steps before machine
stopping.

In the theory of classical computation [5], both 2-stack machines, as a generalization of


pushdown automata, and 2-counter machines can efficiently simulate Turing machines [7,2,5].
However, as the authors are aware, the simulations of QTMs in terms of QMSMs and QMCMs
still have not been considered. Since Turing machines, circuits, multi-stack machines, and
multi-counter machines are equivalent in classical computation, we naturally hope to clarify
their computing power in quantum computers. Therefore, our focuses in this article are to
introduce QMSMs and QMCMs that are somewhat different from the QCAs in the literature
[6,10], and particularly, to simulate QTMs by virtue of these two quantum computing devices.

Indeed, in quantum computing devices, the unitarity of evolution operators is generally


characterized by the W-F conditions of the local transition function of the quantum models
under consideration. Bernstein and Vazirani [1] gave the W-F conditions for the QTMs whose
read/write heads are not allowed to be stationary in each move. In QTMs whose read/write
heads are allowed to be stationary (called generalized QTMs, as in [1]), the first sufficient
conditions for preserving the unitarity of time evolution were given by Hirvensalo, and then
Yamasaki [12] gave the simple W-F conditions for multiple-tape stationary-head-move QTMs,
and Ozawa and Nishimura further presented the W-F conditions for the general QTMs. For
the details, see [4, p. 173]. Golovkins [3] defined a kind of QPDAs and gave the corresponding
W-F conditions; Yamasaki et. al. [10] defined quantum 2-counter automata and presented
the corresponding W-F conditions, as well.

We see that those aforementioned W-F conditions given by these authors for correspond-
ing quantum computing devices are quite complicated. Therefore, based on the QPDAs
proposed in [9] where QPDAs in [9] and [8] are shown to be equivalent, we would like to de-
fine QMSMs that generalize the QPDAs in [9], and further define QMCMs. As well, we will
give the W-F conditions for these defined devices. Notably, these W-F conditions are more
succinct than those mentioned above. In particular, motivated by Yao’s work [11] concerning
the (n, t)-simulations of QTMs by quantum circuits, we will use QMSMs to (n, t)-simulate
QTMs, where n denotes that the length of input strings are not beyond n, and t represents
that the number of move steps of QTMs (time complexity) is not bigger that t for those input

3
strings.

1.2. Main results

According to the above analysis, we state the main contributions in this article. In Section
2, we define QMSMs by generalizing QPDAs in [9] from one-stack to muti-stack, and present
the corresponding W-F conditions (Theorem 1) for the defined quantum devices.

In Section 3, by means of QMSMs we define QMCMs that are somewhat different from
the QCAs by Kravtsev [6] and Yamasaki et al. [10]; as well, the W-F conditions (Theorem 2)
are given for the defined QMCMs. It is worth indicating that the state transition functions
in QCAs defined by Kravtsev [6] and Yamasaki et al. [10] have local property, since they
are defined on Q × {0, 1} × (Σ ∪ {#, $}) × Q × {0, 1}, but their W-F conditions are quite
complicated, while in QMCMs defined in this article, the state transition functions are on
Q × Nk × (Σ ∪ {#, $}) × Q × Nk , and consequently, the corresponding W-F conditions are
more succinct (see Theorem 2).

To simulate QTMs, we deal with a number of properties regarding simulations between


QMCMs with different counters and different counts (Lemmas 1 and 2). We show that
QMCMs allowed to count with 0, ±1, ±2, . . . , ±n can be simulated by QMCMs that are able
to count with 0, ±1 only but need more counters.

In particular, in Section 4, we present the simulations of QTMs in terms of QMCMs


with polynomial time slowdown. More specifically, we prove that any QTM M1 , there exists
QMSM M2 (n, t)-simulating M1 , where n denotes the length of input strings not bigger than
n, and t represents that the number of move steps of QTMs is not bigger that t for those
input strings. Also, we show that QMCMs can be simulated by QMSMs with the same time
complexity, and by this result it then follows the efficient simulations of QTMs by QMSMs.

Due to the page limit, these detailed proofs of Theorems and Lemmas are put in Appen-
dices. In this extended abstract, notations will be explained when they first appear.

2. Quantum multi-stack machines

QPDAs were considered by the authors in [8,3,9]. Here we will define quantum k-stack
machines by generalizing the QPDAs in [9] from one stack to k stacks.

4
Definition 1. A quasi-quantum two-stack machine is defined as M = (Q, Σ, Γ, δ, Z0 , q0 , qa , qr )
where Q is the set of states, Σ is the input alphabet, Γ is the stack alphabet, Z0 ∈ Γ denotes
the most bottom symbol that is not allowed to be popped, q0 ∈ Q is the initial state, and
qa , qr ∈ Q are respectively the accepting and rejecting states, and transition function δ is
defined as follows:

δ : Q × Γ∗ × Γ∗ × (Σ ∪ {#, $}) × Q × Γ∗ × Γ∗ → C
′ ′ ′
where Γ∗ denotes the set of all strings over Γ, and δ(q, γ1 , γ2 , σ, q , γ1 , γ2 ) 6= 0 if and only if
′ ′ ′
(i) γ1 = γ1 or Xγ1 = γ1 or γ1 = Xγ1 for some X ∈ Γ\{Z0 }; and
′ ′ ′
(ii) γ2 = γ2 or Y γ2 = γ2 or γ2 = Y γ2 for some Y ∈ Γ\{Z0 }.
′ ′
In addition, if γ1 = Z0 , then γ1 = Z0 or γ1 = ZZ0 for some Z ∈ Γ with Z 6= Z0 ; similar
restriction is imposed on γ2 .

A configuration of the machine is described by |qi|γ1 i|γ2 i, where q is the current control
state, γ1 and γ2 represent the current strings of stack symbols in two stacks, respectively,
and we identify the leftmost symbol of γi with the top stack symbol of stack i for i = 1, 2.
Therefore, the rightmost symbol of γi is Z0 . We denote by CM as the set of all configurations
of M , that is,
CM = {|qi|γ1 Z0 i|γ2 Z0 i : q ∈ Q, γi ∈ Γ∗ , i = 1, 2}.

Let HX represent the Hilbert space whose orthonormal basis is the set X, that is, HX = l2 (X).
Therefore, HQ ⊗ HΓ∗ ⊗ HΓ∗ is a Hilbert space whose orthonormal basis is CM , that is,
l2 (CM ) = HQ ⊗ HΓ∗ ⊗ HΓ∗ . In addition, we assume that there are endmarkers # and $
representing the leftmost and rightmost symbols for any input string x ∈ Σ∗ . Therefore, any
input string x ∈ Σ∗ is put on the input tape in the form of #x$, and the read head of M
begins with # and ends after reading $.
′ ′ ′
Intuitively, δ(q, γ1 , γ2 , σ, q , γ1 , γ2 ) denotes the amplitude of the machine evolving into
′ ′ ′
configuration |q i|γ1 i|γ2 i from the current one |qi|γ1 i|γ2 i after reading input σ.

For any σ ∈ Σ ∪ {#, $} we defined the time evolution operators Uσ and Uσ from HQ ⊗
HΓ∗ ⊗ HΓ∗ to HQ ⊗ HΓ∗ ⊗ HΓ∗ as follows:
X ′ ′ ′ ′ ′ ′
Uσ (|qi|γ1 i|γ2 i) = δ(q, γ1 , γ2 , σ, q , γ1 , γ2 )|q i|γ1 i|γ2 i, (1)
′ ′ ′
q ,γ1 ,γ2

′ X ′ ′ ′ ′ ′ ′
Uσ (|qi|γ1 i|γ2 i) = δ∗ (q , γ1 , γ2 , σ, q, γ1 , γ2 )|q i|γ1 i|γ2 i, (2)
′ ′ ′
q ,γ1 ,γ2

5

where δ∗ denotes the conjugate complex number δ. By linearity Uσ and Uσ can be extended
to HQ ⊗ HΓ∗ ⊗ HΓ∗ .
′ ′
Remark 1. Uσ is the adjoint operator of Uσ . Indeed, for any (qi , γi1 , γi2 ) ∈ Q × Γ∗ × Γ∗ ,
by means of Eqs. (1,2) we have

hUσ |q1 i|γ11 i|γ12 i, Uσ |q2 i|γ21 i|γ22 ii


X
= δ(q1 , γ11 , γ12 , σ, q, γ1 , γ2 ) × δ∗ (q2 , γ21 , γ22 , σ, q, γ1 , γ2 )
q,γ1 ,γ2
D ′
E
= |q1 i|γ11 i|γ12 i, Uσ Uσ |q2 i|γ21 i|γ22 i . (3)

Definition 2. Let M be a quasi-quantum two-stack machine with input alphabet Σ. If


Uσ is unitary for any σ ∈ Σ ∪ {#, $}, then M is called a quantum two-stack machine.

Now we give the well-formedness conditions for justifying the unitarity of Uσ for any
σ ∈ Σ ∪ {#, $}, that are described by the following theorem.

Theorem 1. Let M be a quasi-quantum two-stack machine with input alphabet Σ. Then


for any σ ∈ Σ ∪ {#, $}, linear operator Uσ is unitary if and only if δ satisfies the following
well-formedness conditions:

(I) For any σ ∈ Σ ∪ {#, $},


X ′ ′ ′ ′ ′ ′
δ(q1 , γ11 , γ12 , σ, q , γ1 , γ2 ) × δ∗ (q2 , γ21 , γ22 , σ, q , γ1 , γ2 )
′ ′
q ′ ,γ1 ,γ2

 1, if (q1 , γ11 , γ12 ) = (q2 , γ21 , γ22 ),
= (4)
 0, otherwise.

(II) For any σ ∈ Σ ∪ {#, $},


X ′ ′ ′ ′ ′ ′
δ(q , γ1 , γ2 , σ, q1 , γ11 , γ12 ) × δ∗ (q , γ1 , γ2 , σ, q2 , γ21 , γ22 )
′ ′ ′
q ,γ1 ,γ2

 1, if (q1 , γ11 , γ12 ) = (q2 , γ21 , γ22 ),
= (5)
 0, otherwise.

Proof. Is is similar to Theorem 5 in [9], and the details are referred to Appendix I. 2

3. Quantum multi-counter machines

QCAs were first considered by Kravtsev [6], and further developed by Yamasaki et al. [10].
In this section, we introduce a different definition of quantum k-counter machines, and then

6
deal with some simulations between QMCMs with different counters and with different counts
in each move.

Definition 3. A quasi-quantum k-counter machine is defined as M = (Q, Σ, δ, q0 , qa , qr )


where Q is a set of states with initial state q0 ∈ Q and states qa , qr ∈ Q representing
accepting and rejecting states, respectively, Σ is an input alphabet, and transition function
δ is a mapping from

Q × Nk × (Σ ∪ {#, $}) × Q × Nk
to C, where N denotes the set of all nonnegative integer and #, $ represent two endmarkers
that begins with # and ends with $, and δ satisfies that
′ ′ ′ ′
δ(q, n1 , n2 , . . . , nk , σ, q , n1 , n2 , . . . , nk ) 6= 0 (6)

only if |ni − ni | ≤ 1 for i = 1, 2, . . . , k. Furthermore, let |qi|n1 i|n2 i . . . |nk i represent a
configuration of M , where q ∈ Q, ni ∈ N for i = 1, 2, . . ., and let the set

CM = {|qi|n1 i|n2 i . . . |nk i : q ∈ Q, ni ∈ N, i = 1, 2, . . . , k}


be an orthonormal basis for the space HCM = l2 (CM ). For any σ ∈ Σ, linear operator Vσ on
HCM is defined as follows:
X ′ ′ ′ ′ ′ ′ ′ ′
Vσ |qi|n1 i|n2 i . . . |nk i = δ(q, n1 , n2 , . . . , nk , σ, q , n1 , n2 , . . . , nk )|q i|n1 i|n2 i . . . |nk i
′ ′ ′
q ′ ,n1 ,n2 ,...,nk
(7)
and Vσ is extended to HCM by linearity.

Definition 4. We say that the quasi-quantum counter machine M = (Q, Σ, δ, q0 , qa , qr )


defined above is a quantum k-counter machine, if Vσ is unitary for any σ ∈ (Σ ∪ {#, $}).

Also, we define linear operator Vσ on HCM as follows:
′ X ′ ′ ′ ′ ′ ′ ′ ′
Vσ |qi|n1 i|n2 i . . . |nk i = δ∗ (q , n1 , n2 , . . . , nk , σ, q, n1 , n2 , . . . , nk )|q i|n1 i|n2 i . . . |nk i
′ ′ ′
q ′ ,n1 ,n2 ,...,nk
(8)

Remark 2. Clearly Vσ is an adjoint operator of Vσ , which can be checked in terms of
the process of Remark 1, and the details are therefore omitted here.

Now we give the W-F conditions for characterizing the unitarity of Vσ , that are described
in the following theorem. For the sake of simplicity, we deal with the case of k = 2, and the
other cases are exactly similar.

Theorem 2. Let M be a quasi-quantum two-counter machine with input alphabet Σ.

7
Then for any σ ∈ Σ ∪ {#, $}, Vσ defined as Eq. (7) is unitary if and only if δ satisfies the
following W-F conditions:

(I) For any σ ∈ Σ ∪ {#, $},


X ′ ′ ′ ′
δ(q1 , n11 , n12 , σ, p, n1 , n2 ) × δ∗ (q2 , n21 , n22 , σ, p, n1 , n2 )
′ ′
p,n1 ,n2

 1, if (q1 , n11 , n12 ) = (q2 , n21 , n22 ),
= (9)
 0, otherwise.

(II) For any σ ∈ Σ ∪ {#, $},


X ′ ′ ′ ′
δ(p, n1 , n2 , σ, q1 , n11 , n12 ) × δ∗ (p, n1 , n2 , σ, q2 , n21 , n22 )
′ ′
p,n1 ,n2

 1, if (q1 , n11 , n12 ) = (q2 , n21 , n22 ),
= (10)
 0, otherwise.

Proof. It is similar to Theorem 1 above, and the details are presented in Appendix II. 2

In order to simulate QTMs by QMSMs, we need some related lemmas and definitions. In
general, quantum counter machines are allowed to count by ±1 and 0 only. Here we would
like to deal with the quantum machines with count beyond such a bound, and show that they
are indeed equivalent.

Definition 5. A quasi-quantum k-counter machine M = (Q, Σ, δ, q0 , qa , qr ) is called to


count with ±r for r ≥ 1, if its k’s counters are allowed to change with numbers 0, ±1, or ±r
′ ′
at each step. In this case, if |ni − ni | ≤ 1 or |ni − ni | = r for i = 1, 2, . . . , k, then
′ ′ ′ ′
δ(q, n1 , n2 , . . . , nk , σ, q , n1 , n2 , . . . , nk ) 6= 0
may hold; otherwise it is 0. We say that the quasi-quantum k-counter machine M is quantum
if for any σ ∈ Σ ∪ {#, $}, Vσ is a unitary operator on l2 (CM ), where

CM = {|qi|n1 i|n2 i . . . |nk i : q ∈ Q, ni ∈ N, i = 1, 2, . . . , k}.

It is ready to obtain that Theorem 2 also holds for quantum k-counter machines with
count ±r for r ≥ 1.

Theorem 3. Let M = (Q, Σ, δ, q0 , qa , qr ) be a quasi-quantum k-counter machine that is


allowed to count with a certain ±r for r ≥ 1. Then for any σ ∈ Σ ∪ {#, $}, Vσ defined as Eq.
(7) is unitary if and only if δ satisfies Eqs. (9,10).

8
Proof. Similar to Theorem 2. 2

Definition 6. Let M1 and M2 be quantum k1 -counter machine M1 and quantum k2 -


counter machine M2 , respectively, and, M1 and M2 have the same input alphabet Σ. For
(1) (2)
any σ ∈ Σ ∪ {#, $}, Vσ and Vσ defined as Eq. (7) represent the evolution operators in M1
and M2 , respectively. We say that M1 can simulate M2 , if for any string σ1 σ2 . . . σn ∈ Σ∗ ,
X (1) (1) (1)
hik1 | . . . hi1 |hqa(1) |V$ Vσ(1)
n
Vσ(1)
n−1
. . . Vσ(1)
1
V# |q0 i|0i . . . |0i
i1 ,i2 ,...,ik1 ≥0
X (2) (2) (2)
= hik1 | . . . hi1 |hqa(2) |V$ Vσ(2)
n
Vσ(2)
n−1
. . . Vσ(2)
1
V# |q0 i|0i . . . |0i, (11)
j1 ,j2 ,...,jk1 ≥0

(i) (i)
where q0 and qa denote the initial and accepting states of Mi , respectively, i = 1, 2.

For convenience, for any quantum k-counter machine M = (Q, Σ, δ, q0 , qa , qr ), we define


M
the accepting probability Paccept (σ1 σ2 . . . σn ) for inputting σ1 σ2 . . . σn as:

M
Paccept (σ1 σ2 . . . σn )
X
= hik | . . . hi1 |hqa |V$M VσM V M . . . VσM
n σn−1 1
V#M |q0 i|0i . . . |0i, (12)
i1 ,i2 ,...,ik ≥0

where VσM is unitary operator on l2 (CM ) for any σ ∈ Σ ∪ {#, $}.

The detailed proofs of the following two lemmas are given in Appendix III.

Lemma 1. For any quantum k-counter machine M1 that is allowed to count with ±r
for r ≥ 1, there exists quantum 2k-counter machine M2 simulating M1 with the same time
complexity, where M2 is allowed to count with 0, ±1, and ±(r − 1).

Lemma 2. For any quantum k-counter machine M1 that is allowed to count with
0, ±1, ±2, . . . , ±r, then there exists a quantum kr-counter machine M2 simulating M1 with
the same time complexity, where M2 is allowed to count with 0, ±1 only.

4. Simulations of quantum Turing machines

4.1. Simulations of quantum Turing machines in terms of quantum multi-counter ma-


chines

To simulate QTMs in terms of QMCMs, we give the definition of QTMs in terms of Bernstein
and Vazirani [1], in which the read-write head will move either to the right or to the left at
each step. Indeed, generalized QTMs can also be simulated by QMCMs, but the discussion

9
regarding unitarity is much more complicated. For the sake of simplicity, we here consider
the former QTMs.

Definition 7. A QTM is defined by M = (Σ, Q, δ, B, q0 , qa , qr ), where Σ is a finite input


alphabet, B is an identified blank symbol, Q is a finite set of states with an identified initial
state q0 and final state qa , qr 6= q0 , where qa and qr represent accepting and rejecting states,
respectively, and the quantum transition function δ is defined as

δ : Q × Σ × Σ × Q × {L, R} → C.

The QTM has a two-way infinite tape of cells indexed by Z and a single read-write tape head
that moves along the tape. A configuration of this machine is described by the form |qi|τ i|ii,
where q denotes the current state, τ ∈ ΣZ describes the tape symbols, and i ∈ Z represents
the current position of tape head. Naturally, a configuration containing initial or final state is
called an initial or final configuration. Let CM denote the set of all configurations in M , and
therefore HCM = l2 (CM ), that is a Hilbert space whose orthonormal basis can be equivalently
viewed as CM . Then the evolution operator UM on l2 (CM ) can be defined in terms of δ: for
any configuration |ci ∈ CM ,
X ′ ′
UM |ci = a(c, c )|c i, (13)

|c i∈CM
′ ′
where a(c, c ) is the amplitude of configuration |ci evolving into |c i in terms of the transition
function δ. UM is a unitary operator on l2 (CM ).

As in [1], we define that QTM halts with running time T on input x if after the T ’s
step moves beginning with its initial configuration, the superposition contains only final
configurations, and at any time less than T the superposition contains no final configuration.
Therefore, we assume that the QTM satisfies this requirement.

Definition 8. For nonnegative integer n, T , let M1 = (Q1 , Σ1 , δ1 , B1 , q10 , q1a , q1r ) be a


quantum Turing machine with initial state q10 , and let M2 be a quantum k-counter machine
with initial state q20 and the same input alphabet Σ1 as M1 . We say that M2 (n, T )-
simulates quantum Turing machine M1 with polynomial time O(n, T ) slowdown, if there
exist some tape symbols added in M2 , say B2 , B3 , . . . , Bm such that for any input x =
σ1 σ2 . . . σl ∈ Σ∗2 (l ≤ n), if the computation of M1 ends with t steps (t ≤ T ), then there
is nonnegative integers kl1 , kl2 , . . . , klm1 and ks1 , ks2 , . . . , ksm2 that are related to l and t,
Pm1 Pm2
satisfying i=1 kli + i=1 ksi ≤ O(n, T ), and

PaM1 (x) = PaM2 (x), (14)

10
where
X 2
PaM1 (x) = t

hi|hτ |hq1a |UM 1
|q 10 i|τ 0 i|0i (15)
−T ≤i≤T,τ ∈Σ[−T,T ]Z

j+1 , if j ∈ [0, l − 1]Z ,
 σ
where τ0 is defined as: τ0 (j) = and
 B ,
1 j ∈ [−T, −1]Z ∪ [l, T ]Z ,

hnk | . . . hn1 |hq2a |V$ V ksm2 . . . V ks1
X
PaM2 (x) = Bsm2 Bs1
n1 ,n2 ,...,nk ≥0
2
klm k
Vσl Vσl−1 . . . Vσ1 VBl 1 . . . VBll1 V# |q0 i|0i . . . |0i . (16)
m 1 1

The main result of this subsection is as follows:

Theorem 4. For any QTM M1 = (Q1 , Σ1 , δ1 , B1 , q10 , q1a , q1r ) with initial state q10 and
accepting and rejecting states q1a , q1r , and for any nonnegative integer n, t with n ≤ t+1, there
exists a quantum (2t+2)-counter machine M2 that (n, t)-simulates M1 with most slowdown
O(n + t).

Proof. The details are referred to Appendix IV, but we outline the basic idea as follows:
We add three assistant input symbols B2 , B3 , B4 in M2 . For any input string σ1 σ2 . . . σk ∈ Σ∗1
with k ≤ n, we take certain integer numbers l2 (k), l3 (k), l4 (k) that are related to k, and the
l (k) l (k) l (k)
input string σ1 σ2 . . . σk is put on the tape of M2 in the form #σ1 σ2 . . . σk B22 B33 B44 $.
l (k)
After M2 finishes the reading of B33 , the counters from t + 1 to t + k have numbers
corresponding to σ1 , σ2 , . . . , σk , respectively, those from 1 to t are numbers corresponding to
blank B1 in M1 , and the last counter 2t + 2 is always used to simulate the position of the
l (k)
read-write head of M1 . Assistant symbols B44 are read one by one in the process of M1
computing σ1 σ2 . . . σk . When M1 ends, M2 will read $ and stop. In order to preserve the
unitarity of M2 , some additional definitions for the transition function δ2 of M2 are necessary.
2

4.2. Simulations of quantum Turing machines in terms of quantum multi-stack ma-


chines

QTMs can be also (n, t)-simulated by quantum multi-stack machine, since quantum k-counter
machine can be simulated by quantum multi-stack machine in terms of the following Theorem
5.

11
Definition 9. We say that quantum k-stack machine M2 = (Q2 , Σ2 , Γ2 , δ2 , Z0 , q20 , q2a , q2r )
simulates quantum k-counter machine M1 = (Q1 , Σ1 , δ1 , q10 , q1a , q1r ) that has the same input
alphabet Σ1 = Σ2 with the same time complexity, if for any input string x = σ1 σ2 . . . σn ∈ Σ∗1 ,
we have
M1 M2
Paccept (x) = Paccept (x) (17)

where
X
M1
Paccept (x) = |hγk |hγk−1 | . . . hγ1 |hq1a | U$ Uσn . . . Uσ1 U# |q10 i|0i . . . |0i|2 . (18)
γ1 ,γ2 ,...,γk

As well, the (n, t)-simulations of QTMs in terms of quantum k-stack machine can be
similarly defined as Definition 8, and we leave out the details here.

Theorem 5. For any given quantum k-counter machine M1 = (Q1 , Σ1 , δ1 , q10 , q1a , q1r ),
there exists quantum k-stack machine M2 that simulates M1 with the same time complexity.

Proof. Let M2 = (Q2 , Σ2 , δ2 , q20 , q2a , q2r ) where Q2 = Q1 , Σ2 = {Z0 , X}, q20 = q10 ,
q2a = q1a , q2r = q1r . Define mapping m : N → Σ∗2 as follows:

m(n) = Z0 X |n|
where we denote X 0 = ǫ, that is, Z0 X 0 = Z0 . We define δ2 as follows:
′ ′ ′
δ2 (q, Z0 X l1 , Z0 X l2 , . . . , Z0 X lk , σ, p, Z0 X l1 , Z0 X l2 , . . . , Z0 X lk )
′ ′ ′
= δ1 (q, l1 , l2 , . . . , lk , σ, p, l1 , l2 , . . . , lk )


for any q, p ∈ Q1 , σ ∈ Σ1 ∪ {#, $}, and li , li ∈ N. Then it is easy to check that δ2 satisfies
the W-F conditions Eqs. (4,5), and that M2 simulates M1 step by step. This completes the
proof. 2

Corollary 1. For any n, t ∈ N, and any QTM M1 , there exists QMSM M2 that simulates
M1 with slowdown O(n + t).

5. Concluding remarks

The unitary evolution of quantum physics requires that quantum computation should be nec-
essarily time reversible (unitary). This makes some simulations between quantum computing
devices quite complicated. Indeed, the unitarity is reflected by the W-F conditions. The W-F
conditions for these QMSMs and QMCMs defined in this paper are more succinct than the

12
W-F conditions for QCAs introduced by Yamasaki et al. [10], but we note that the transition
functions in our quantum devices employ the whole property of the symbols in the stacks or
counters at each move. An issue worthy of further consideration is to give also succinct W-F
conditions but yet more local transition functions for characterizing the unitarity of these
QMSMs and QMCMs defined in this paper. Moreover, the relationships between QMCMs in
the paper and QCAs by Yamasaki et al. [10] still need to be further clarified. Finally, how
to improve the (n, t)-simulations of QTMs by QMCMs and QMSMs towards more general
simulations and how to decrease the number of counters of QMCMs for simulating QTMs
are also worth studying.

Acknowledgements

This research is supported by the National Natural Science Foundation (No. 90303024), and
the Natural Science Foundation of Guangdong Province (No. 020146, 031541) of China.

References

[1] E. Bernstein and U. Vazirani, Quantum complexity theory, SIAM J. Comput. 26


(1997) 1411-1473.

[2] P.C. Fischer, Turing machine with restricted memory access, Information and Con-
trol 9 (4) (1966) 364-379.

[3] M. Golovkins, Quantum Pushdown Automata, in: Proc. 27th Conf. on Current
Trends in Theory and Practice of Informatics, Milovy, Lecture Notes in Computer
Science, Vol. 1963, Spring-Verlag, Berlin, 2000, pp. 336-346.

[4] J. Gruska, Quantum Computing, McGraw-Hill, London, 1999.

[5] J.E. Hopcroft and J.D. Ullman, Introduction to Automata Theory, Languages, and
Computation, Addision-Wesley, New York, 1979.

[6] M. Kravtsev, Quantum finite one-counter automata, in: SOFSEM’99, Lecture Notes
in Computer Science, Vol.1725, Springer-Verlag, Berlin, 1999, pp.431-440.

[7] M.L. Minsky, Recursive unsolvability of Post’s problem of ’tag’ and other topics in
the theory of Turing machines, Annals of Mathematics 74 (3) (1961) 437-455.

13
[8] C. Moore and J.P. Crutchfield, Quantum automata and quantum grammars, The-
oret. Comput. Sci. 237 (2000) 275-306. Also quant-ph/9707031, 1997.

[9] D.W. Qiu and M.S. Ying, Characterization of quantum automata, Theoret. Comput.
Sci. 312 (2004) 479-489.

[10] T. Yamasaki, H. Kobayashi, H. Imai, Quantum versus deterministic counter au-


tomata, Theoret. Comput. Sci. to appear.

[11] A.C. Yao, Quantum circuit complexity, in: Proc. 34th IEEE Symp. on Foundations
of Computer science, 1993, pp. 352-361.

[12] T. Yamakami, A Foundation of Programming a Multi-Tape Quantum Turing Ma-


chine, in: MFCS’99, Lecture Notes in Computer Science, Vol.1672, Springer-Verlag,
Berlin, 1999, pp. 430-441.

14
Appendix I. The proof of Theorem 1

First we show that if δ satisfies the well-formedness conditions (I) and (II) above, then for
any σ ∈ Σ ∪ {#, $}, Uσ is unitary. For any |qi , i|γi1 i|γi2 i ∈ CM , i = 1, 2, by condition (I) we
have

hUσ |q1 i|γ11 i|γ12 i, Uσ |q2 i|γ21 i|γ22 ii


*
X ′ ′ ′ ′
= δ(q1 , γ11 , γ12 , σ, p1 , γ11 , γ12 )|p1 i|γ11 i|γ12 i,
′ ′
p1 ,γ11 ,γ12
+
X ′ ′ ′ ′
δ(q2 , γ21 , γ22 , σ, p2 , γ21 , γ22 )|p2 i|γ21 i|γ22 i
′ ′
p2 ,γ21 ,γ22
X
= δ(q1 , γ11 , γ12 , σ, q, γ1 , γ2 ) × δ∗ (q2 , γ21 , γ22 , σ, q, γ1 , γ2 )
q,γ1 ,γ2

 1, if (q1 , γ11 , γ12 ) = (q2 , γ21 , γ22 ),
=
 0, otherwise.

Furthermore, for any σ ∈ Σ ∪ {#, $}, linear operator Uσ on HQ ⊗ HΓ∗ ⊗ HΓ∗ is defined as
Eq. (2). Then for any |qi|γ1 i|γ2 i ∈ CM , with condition (II) (Eq. (4)) we have


Uσ Uσ |qi|γ1 i|γ2 i
X ′ ′ ′ ′ ′ ′
= δ∗ (q , γ1 , γ2 , σ, q, γ1 , γ2 )Uγ |q i|γ1 i|γ2 i
′ ′
q ′ ,γ1 ,γ2
X ′ ′ ′ X ′ ′ ′ ′′ ′′ ′′ ′′ ′′ ′′
= δ∗ (q , γ1 , γ2 , σ, q, γ1 , γ2 ) × δ(q , γ1 , γ2 , σ, q , γ1 , γ2 )|q i|γ1 i|γ2 i
′ ′ ′′ ′′
q ′ ,γ1 ,γ2 q ′′ ,γ1 ,γ2
= |qi|γ1 i|γ2 i,

and similarly,

Uσ Uσ |qi|γ1 i|γ2 i = |qi|γ1 i|γ2 i.
′ ′ ′ ′
So, we have verified that Uσ Uσ = Uσ Uσ = I, and therefore, Uσ = Uσ−1 . Therefore Uσ is also
surjective and Uσ has been shown to be unitary. On the other hand, if Uσ is unitary, then
′ ′ ′
Uσ Uσ = Uσ Uσ = I for any σ ∈ Σ ∪ {#, $}, since Uσ is the adjoint operator of Uσ . We need
to demonstrate that the well-formedness conditions (I) and (II) hold. Indeed, the unitarity
of Uσ implies that for any |qi , i|γi1 i|γi2 i ∈ CM , i = 1, 2,
X ′ ′ ′ ′ ′ ′
δ(q1 , γ11 , γ12 , σ, q , γ1 , γ2 ) × δ∗ (q2 , γ21 , γ22 , σ, q , γ1 , γ2 )
′ ′ ′
q ,γ1 ,γ2
= hUσ |q1 i|γ11 i|γ12 i, Uσ |q2 i|γ21 i|γ122 ii

15
= h|q1 i|γ11 i|γ12 i, |q2 i|γ21 i|γ122 ii

 1, if (q1 , γ11 , γ12 ) = (q2 , γ21 , γ22 ),
=
 0, otherwise.

′ ′ ′ ′
Furthermore, since Uσ Uσ = Uσ Uσ = I, Uσ is also unitary. Therefore, the unitarity of Uσ
implies that the condition (II) holds true, as well. 2

Appendix II. The proof of Theorem 2

If δ satisfies the conditions (I) and (II) given by Eqs. (9,10), then for any configurations
|qi|n1 i|n2 i, in light of condition (I) we have
 
′ ′  X ′ ′ ′ ′ ′ ′ 
Vσ Vσ |qi|n1 i|n2 i = Vσ  δ(q, n1 , n2 , σ, q , n1 , n2 )|q i|n1 i|n2 i
′ ′
q ′ ,n1 ,n2
X ′ ′ ′ X ′ ′ ′
= δ(q, n1 , n2 , σ, q , n1 , n2 ) δ∗ (p, m1 , m2 , σ, q , n1 , n2 )|pi|m1 i|m2 i

q ′ ,n1 ,n2
′ p,m1 ,m2
X ′ ′ ′ ′ ′ ′
= δ(q, n1 , n2 , σ, q , n1 , n2 )δ∗ (q, n1 , n2 , σ, q , n1 , n2 )|qi|n1 i|n2 i
′ ′ ′
q ,n1 ,n2
= |qi|n1 i|n2 i,

and with condition (II) we have


 
′ X ′ ′ ′ ′ ′ ′ 
δ∗ (q , n1 , n2 , σ, q, n1 , n2 )|q i|n1 i|n2 i

Vσ Vσ |qi|n1 i|n2 i = Vσ 
′ ′
q ′ ,n1 ,n2
X ′ ′ ′ X ′ ′ ′
= δ∗ (q , n1 , n2 , σ, q, n1 , n2 ) δ(q , n1 , n2 , σ, p, m1 , m2 )|pi|m1 i|m2 i

q ′ ,n1 ,n2
′ p,m1 ,m2
X ′ ′ ′ ′ ′ ′
= δ∗ (q , n1 , n2 , σ, q, n1 , n2 )δ(q , n1 , n2 , σ, q, n1 , n2 , )|qi|n1 i|n2 i
′ ′ ′
q ,n1 ,n2
= |qi|n1 i|n2 i.
′ ′
Therefore, Vσ Vσ = Vσ Vσ = I for any σ ∈ Σ ∪ {#, $}, and Vσ is therefore unitary.

On the other hand, if Vσ is unitary, then for any (qi , ni1 , ni2 ) ∈ Q × N × N, i = 1, 2, we
have

h|q1 i|n11 i|n12 i, |q2 i|n21 i|n22 ii

= hVσ |q1 i|n11 i|n12 i, Vσ |q2 i|n21 i|n22 ii


X ′ ′ ′ ′
= δ(q1 , n11 , n12 , σ, p, n1 , n2 ) × δ∗ (q2 , n21 , n22 , σ, p, n1 , n2 )
′ ′
p,n1 ,n2

16

 1, if (q1 , n11 , n12 ) = (q2 , n21 , n22 ),
=
 0, otherwise.

As well, the unitarity of Vσ implies that condition (II) holds. 2

Appendix III. The proofs of Lemmas 1 and 2

The proof of Lemma 1:

We here consider the case of k = 1 without loss of generality, since it is similar to show
the general situation. Suppose M1 = (Q1 , Σ1 , δ1 , q10 , q1a , q1r ). We define a desired quantum
two-counter machine M2 simulating M1 . A basic idea is that one of counters in M2 simulates
the changes of d1 ∈ {−1, 0, 1} in M1 and the other counter of M2 simulates the changes of
d2 ∈ {−r, 0, r} in M1 . More formally, we define M2 = (Q2 , Σ2 , δ2 , q20 , q2a , q2r ) as follows:
Q2 = Q1 , Σ2 = Σ1 , q10 = q20 , q2a = q1a , q2r = q1r , and

δ2 : Q2 × [0, r − 1]Z × N × Σ × Q2 × [0, r − 1]Z × N → C


where [0, r − 1]Z denotes the set {0, 1, 2, . . . , r − 1}. Specifically, δ2 is defined in the following
way: If
δ1 (q, n1 + k1 r, σ, p, n2 + k2 r) = c (19)

where 0 ≤ n1 , n2 < r, k1 , k2 ≥ 0, and c ∈ C denotes its amplitude, then

δ2 (q, n1 , k1 , σ, p, n2 , k2 ) = c. (20)

For example, if δ1 (q, kr + r − 1, σ, p, (k + 1)r) = c, then δ2 (q, r − 1, k, σ, p, 0, k + 1) = c; if


δ1 (q, kr, σ, p, kr − 1) = c, then δ2 (q, 0, k, σ, p, r − 1, k − 1) = c.

In terms of the definition of δ2 as Eqs. (19,20), we have also defined a linear operator VσM2
by Eq. (7) on space l2 (CM2 ) = HQ2 ⊗ H[0,r−1]Z ⊗ HN , where HX , as above, is identified with
a Hilbert space whose orthonormal basis is the set X, and CM2 is the set of configurations
of M2 , as follows:

CM2 = {|qi|ii|ji : q ∈ Q2 , 0 ≤ i ≤ r − 1, j ≥ 0}.


By means of Eqs. (19,20), we will show that δ1 satisfying Eqs. (9,10) implies that δ2 also
satisfies Eqs. (9,10), and therefore, by Theorem 3 VσM2 is a unitary operator on l2 (CM2 ).
Furthermore, for configuration |qi|ii|ji with i ≥ r, we may define

17
VσM2 |qi|ii|ji = |qi|ii|ji,
then VσM2 is exactly extended to be a unitary operator on HQ2 ⊗ HN ⊗ HN .

More precisely, we show that VσM2 is a unitary operator on l2 (CM2 ) = HQ2 ⊗H[0,r−1]Z ⊗HN .
For any (qi , ni1 , ni2 ) ∈ Q2 × [0, r − 1]Z × N, i = 1, 2,
X
δ2 (q1 , n11 , n12 , σ, p, n1 , n2 ) × δ2∗ (q2 , n21 , n22 , σ, p, n1 , n2 )
p,n1 ,n2
X
= δ1 (q1 , n11 + n12 r, σ, p, n1 + n2 r) × δ1∗ (q2 , n21 + n22 r, σ, p, n1 + n2 r)
p,n1 ,n2

 1, if (q1 , n11 , n12 ) = (q2 , n21 , n22 ),
= (21)
 0, otherwise.

Similarly, we have that for any (pi , ni1 , ni2 ) ∈ Q × [0, r − 1]Z × N, i = 1, 2,
X
δ2 (q, n1 , n2 , σ, p1 , n11 , n12 ) × δ2∗ (q, n1 , n2 , σ, p, n21 , n22 )
q,n1 ,n2
X
= δ1 (q, n1 + n2 r, σ, p1 , n11 + n12 r) × δ1∗ (q, n1 + n2 r, σ, p2 , n21 + n22 r)
q,n1 ,n2

 1, if (p1 , n11 , n12 ) = (p2 , n21 , n22 ),
= (22)
 0, otherwise.

Therefore, by Theorem 3 VσM2 is unitary for any σ ∈ Σ2 . The remainder is to show that
M1 M2
Paccept (x) = Paccept (x) for any x ∈ Σ∗1 , which follows from Eqs. (19,20) and the definition of
M
Paccept (x) described by Eq. (12). 2

The proof of Lemma 2:

We prove the case of k = 1 without loss of generality. Let M1 = (Q1 , Σ1 , δ1 , q10 , q1a , q1r ) be a
quantum 1-counter machine. Then quantum r-counter machine M2 = (Q2 , Σ2 , δ2 , q20 , q2a , q2r )
is defined as: Q2 = Q1 , Σ2 = Σ1 , q20 = q10 , q2a = q1a , q2r = q1r , and δ2 is defined as follows:
If
δ1 (q, k1 r + i1 , σ, p, k2 r + i2 ) = c, (23)

where 0 ≤ i1 , i2 ≤ r − 1 and k1 , k2 ≥ 0, then we define

δ2 (q, (0, 0, . . . , 0, 1, 0, . . . , 0, k1 ), σ, p, (0, 0, . . . , 0, 1, 0, . . . , 0, k2 )) = c, (24)


| {z } | {z }
i1 i2

where δ2 satisfies that if k2 − k1 = 1 then i2 ≤ i1 ; if k2 − k1 = −1, then i2 ≥ i1 ;


0, 0, . . . , 1 denotes that the i1 th number is 1, the jth number is 0 for 1 ≤ j ≤ i1 − 1,
| {z }
i1

18
and, (0, 0, . . . , 0, 1, 0, . . . , 0, k1 ) represents that the rth number is k1 . Therefore the definition
| {z }
i1
by Eq. (24) describes equally the amplitude that in M1 the current number in the counter is
k1 r + i1 and reading σ leads to the number becoming k2 r + i2 . As well, it is easy to see that if
in M1 the count changes with numbers 0, ±1, ±2, . . . , ±r, then in M2 the count changes with
0, ±1. Furthermore, we show that the unitarity of M1 leads to the evolution operator VσM2
defined by Eq. (7) in M2 being unitary for any σ ∈ Σ2 ∪ {#, $}. Indeed, δ2 is a mapping on
set
n Pr−1 o
(q, n1 , n2 , . . . , nr ) : q ∈ Q, ni ∈ N, i = 1, 2, . . . , r, i=1 ni ≤ 1
and therefore, Vσ is a linear operator on l2 (CM2 ) where
n Pr−1 o
CM2 = |qi|n1 i|n2 i . . . |nr i : q ∈ Q, ni ∈ N, i = 1, 2, . . . , r, j=1 nj ≤ 1 ,
that is, l2 (CM2 ) = HQ ⊗ (H{0,1} )⊗(r−1) ⊗ HN .

For any |qi i|ni1 , ni2 , . . . , nir i ∈ CM2 , i = 1, 2, then


X ′ ′ ′
δ2 (q1 , n11 , n12 , . . . , n1r , σ, p, n1 , n2 , . . . , nr )
′ ′ Pr−1 ′
p,n1 ,n2 ,...,n′r , i=1
ni ≤1
′ ′ ′
×δ∗ (q2 , n21 , n22 , . . . , n2r , σ, p, n1 , n2 , . . . , nr )
X r
X r
X r
X r
X
′ ′
= δ1 (q1 , n1i i, σ, p, ni i) × δ1∗ (q2 , n2i i, σ, p, ni i)
′ ′ Pr−1 ′ i=1 i=1 i=1 i=1
p,n1 ,n2 ,...,n′r , i=1
ni ≤1

X r
X r
X
= δ1 (q1 , n1i i, σ, p, n) × δ1∗ (q2 , n2j j, σ, p, n)
p,n i=1 j=1
 Pr Pr Pr−1
 1, if (q1 , = (q2 , with nij ≤ 1 for i = 1, 2,
i=1 n1i i) i=1 n2i i) j=1
=
 0, otherwise,

 1, if (q1 , n11 , n12 , . . . , n1r ) = (q2 , n21 , n22 , . . . , n2r ),
=
 0, otherwise.

As well, we have
X ′ ′ ′
δ2 (p, n1 , n2 , . . . , nr , σ, q1 , n11 , n12 , . . . , n1r )
′ ′ Pr−1 ′
p,n1 ,n2 ,...,n′r , i=1
ni ≤1
′ ′ ′
×δ∗ (p, n1 , n2 , . . . , nr , σ, q2 , n21 , n22 , . . . , n2r )

 1, if (q , n , n , . . . , n ) = (q , n , n , . . . , n ),
1 11 12 1r 2 21 22 2r
=
 0, otherwise.


Therefore, VσM2 is a unitary operator on l2 (CM2 ). Furthermore, we denote CM2 by

CM2 = {|qi|n1 i|n2 i . . . |nr i : q ∈ Q, ni ∈ N, i = 1, 2, . . . , r}.

19

Then we can extend VσM2 to be a unitary operator on l2 (CM2 ) by defining as follows: for any

|qi|n1 i|n2 i . . . |nr i ∈ CM2 \CM2 ,
′ ′ ′ ′
δ2 (q, n1 , n2 , . . . , nr , σ, q , n1 , n2 , . . . , nr )

 1, if (q, n1 , n2 , . . . , nr ) = q ′ , n′ , n′ , . . . , n′ ),
1 2 r
=
 0, otherwise,

and therefore,

VσM2 |qi|n1 i|n2 i . . . |nr i = |qi|n1 i|n2 i . . . |nr i



for any |qi|n1 i|n2 i . . . |nr i ∈ CM2 \CM2 .

Finally, we show that for any x ∈ Σ∗1 ,

M1 M2
Paccept (x) = Paccept (x). (25)

For any σ ∈ Σ1 ∪ {#, $}, and k ≥ 0, 0 ≤ i ≤ r − 1,

VσM1 |qi|kr + ii
X ′ ′ ′ ′
= δ1 (q, kr + i, σ, p, k r + i )|pi|k r + i i
p,|k ′ r+i′ −kr−i|≤r
X ′ ′ ′
= δ2 (q, 0, . . . , 0, 1, 0, . . . , 0, k, σ, 0, . . . , 0, 1, 0, . . . , 0, k )|pi|k r + i i, (26)
′ ′
| {z } | {z }
p,(k ,i )∈S(k,i) i i′
′ ′ ′ ′ ′ ′ ′
where S(k, i) = {(k , i ) : k = k, 0 ≤ i ≤ r − 1, or 0 ≤ k = k − 1, r − 1 ≥ i ≥ i, or k =

k + 1, r − 1 ≤ i ≤ i}. For any |qi|n1 i . . . |nr i ∈ CM2 ,

VσM2 |qi|n1 i . . . |nr i


X ′ ′ ′ ′ ′
= δ2 (q, n1 , n2 , . . . , nr , σ, n1 , . . . , nr )|q i|n1 i . . . |nr i
Pr−1 ′
q′ , i=1
ni ≤1
X r
X r
X
′ ′ ′ ′ ′
= δ1 (q, ni i, σ, q , nj j)|q i|n1 i . . . |nr i. (27)
Pr−1 ′ i=1 j=1
q′ , i=1
ni ≤1

M
Therefore, Eq. (25) exactly follows from the definitions of VσM and Paccept , and Eqs. (26,27)
above. 2

Appendix IV. The proof of Theorem 4

Let |Σ1 | = r−2. Then we define a quantum 2(t+1)-counter machine M2 = (Q2 , Σ2 , δ2 , q0 , q2a , q2r )
that is allowed to count at each move with numbers 0, ±1, . . . , ±r simulating M1 , where

20
′ ′
Q2 = Q1 ∪{q0 , q0 } with q0 , q0 6∈ Q1 , q2a = q1a , q2r = q1r , Σ2 = Σ1 , and δ2 is defined as follows:
Suppose that Σ1 = {σ1 , σ2 , . . . , σr−2 } and bijective mapping e : Σ1 ∪ {B1 } → {1, 2, . . . , r − 1}.

(i) For any σ ∈ Σ1 , r − 1 ≥ ni ≥ 1, 1 ≤ i ≤ 2t + 1, we define

1)
δ2 (q0 , 0, . . . , 0, #, q0 , 0, . . . , 0, 1) = 1; (28)

δ2 (q0 , 0, . . . , 1, #, q0 , 0, . . . , 0, 0) = 1; (29)

2)

δ2 (q0 , n1 , n2 , . . . , ni−1 , 0, . . . , 0, i, σ, q0 , n1 , n2 , . . . , ni−1 , e(σ), 0, . . . , 0, i + 1) = 1; (30)

for instance, δ2 (q0 , 0, . . . , 0, 1, σ, q0 , k(σ), 0, . . . , 0, 2) = 1; and when i = 2t + 2, we define

δ2 (q0 , n1 , n2 , . . . , n2t+1 , 2t + 2, σ, q0 , 0, . . . , 0, 2t + 2) = 1, (31)

and for any 2 ≤ j ≤ 2t + 2

δ2 (q0 , 0, . . . , 0, j, σ, q0 , 0, . . . , 0, j − 1) = 1; (32)

3) for any 1 ≤ l ≤ n, and 1 ≤ i ≤ l + 1, then

δ2 (q0 , n1 , n2 , . . . , nl , 0, . . . , 0, i, B2 , q0 , n1 , n2 , . . . , nl , 0, . . . , 0, i − 1) = 1; (33)

δ2 (q0 , n1 , n2 , . . . , nl , 0, . . . , 0, 0, B2 , q0 , 0, . . . , 0, n1 , n2 , . . . , nl , 0, . . . , 0) = 1; (34)
| {z }
t+1

for 0 ≤ s ≤ l,

δ2 (q0 , 0, . . . , 0, n1 , n2 , . . . , nl , 0, . . . , 0, s, B2 , q0 , 0, . . . , 0, n1 , n2 , . . . , nl , 0, . . . , 0, s + 1) = 1,
| {z } | {z }
t+1 t+1
(35)
and

δ2 (q0 , 0, . . . , 0, n1 , n2 , . . . , nl , 0, . . . , 0, l + 1, B2 , q0 , n1 , n2 , . . . , nl , 0, . . . , 0, l + 1) = 1. (36)
| {z }
t+1

4) for any 1 ≤ l ≤ n, and 0 ≤ i ≤ t, then

δ2 (q0 , 0, . . . , 0, n1 , n2 , . . . , nl , 0, . . . , i, B3 , q0 , 0, . . . , 0, n1 , n2 , . . . , nl , 0, . . . , i + 1) = 1; (37)
| {z } | {z }
t+1 t+1

21
δ2 (q0 , 0, . . . , 0, n1 , n2 , . . . , nl , 0, . . . , t + 1, B3 ,
| {z }
t+1

q0 , e(B1 ), . . . , e(B1 ), n1 , n2 , . . . , nl , e(B1 ), . . . , e(B1 ), t + 1) = 1 (38)
| {z }
t+1

which implies that all e(B1 )’s correspond to the blank from cell −t to cell t in the simulated
quantum Turing machine M1 ;


δ2 (q0 , e(B1 ), . . . , e(B1 ), n1 , n2 , . . . , nl , e(B1 ), . . . , e(B1 ), t + 1, B3 , (39)
| {z }
t+1
p0 , e(B1 ), . . . , e(B1 ), n1 , n2 , . . . , nl , e(B1 ), . . . , e(B1 ), t + 1) = 1; (40)
| {z }
t+1

for t + 1 ≥ j ≥ 1,

δ2 (p0 , e(B1 ), . . . , e(B1 ), n1 , n2 , . . . , nl , e(B1 ), . . . , e(B1 ), j, B3 ,


| {z }
t+1
p0 , e(B1 ), . . . , e(B1 ), n1 , n2 , . . . , nl , e(B1 ), . . . , e(B1 ), j − 1) = 1 (41)
| {z }
t+1

and

δ2 (p0 , e(B1 ), . . . , e(B1 ), n1 , n2 , . . . , nl , e(B1 ), . . . , e(B1 ), 0, B3 ,


| {z }
t+1
q0 , 0, . . . , 0, n1 , n2 , . . . , nl , 0, . . . , 0) = 1 (42)
| {z }
t+1

where p0 is the initial state in M1 .

(ii) If δ1 (p0 , σ1 , σ, p1 , d) = c, then when d = R,

δ2 (p0 , e(B1 ), . . . , e(B1 ), e(σ1 ), . . . , e(σk ), e(B1 ), . . . , e(B1 ), t + 1, B4 ,


| {z }
t+1
p1 , e(B1 ), . . . , e(B1 ), e(σ), e(σ2 ), . . . , e(σk ), e(B1 ), . . . , e(B1 ), t + 2)
| {z }
t+1
= c; (43)

when d = L,

δ2 (p0 , e(B1 ), . . . , e(B1 ), e(σ1 ), . . . , e(σk ), e(B1 ), . . . , e(B1 ), t + 1, B4 ,


| {z }
t+1
p1 , e(B1 ), . . . , e(B1 ), e(σ), e(σ2 ), . . . , e(σk ), e(B1 ), . . . , e(B1 ), t)
| {z }
t+1
= c. (44)

22
(iii) If δ1 (p, σ, τ, q, d) = c, then for any 1 ≤ nj ≤ r − 1, j = 1, 2, . . . , i − 1, i + 1, . . . , 2t + 1,
when d = R,

δ2 (p, n1 , n2 , . . . , ni−1 , e(σ), ni+1 , . . . , n2t+1 , i, B4 ,

q, n1 , n2 , . . . , ni−1 , e(τ ), ni+1 , . . . , n2t+1 , i + 1) = c; (45)

when d = L,

δ2 (p, n1 , n2 , . . . , ni−1 , e(σ), ni+1 , . . . , n2t+1 , i, B4 ,

q, n1 , n2 , . . . , ni−1 , e(τ ), ni+1 , . . . , n2t+1 , i − 1) = c. (46)

(iv) For any p ∈ Q2 , and any j ∈ N,

δ2 (p, n1 , . . . , n2t , j, $, p, n1 , . . . , n2t , j) = 1, (47)

which means that after reading endmarker $ a computation ends.


(b) (b)
Next we will show that δ2 satisfies the well-formedness conditions. Denote CM1 , CM2 ,
(b,+)
and CM2 by
(b)
CM1 = {|qi|τ i|ii : q ∈ Q1 , i ∈ [−t, t]Z , τ ∈ (Σ1 ∪ {B1 })[−t,t]Z },
(b)
CM2 = {|qi|n1 , . . . , n2t+1 i|n2t+2 i : q ∈ Q2 , 0 ≤ ni ≤ r − 1, i = 1, 2, . . . , 2t + 1, 1 ≤ n2t+2 ≤
2t + 1},
(b,+)
CM2 = {|qi|n1 , . . . , n2t+1 i|n2t+2 i : q ∈ Q2 , 1 ≤ ni ≤ r −1, i = 1, 2, . . . , 2t+1, 1 ≤ n2t+2 ≤
2t + 1}.

Since we assume that QTM M1 will end within t steps for any input string x with |x| ≤ n,
(b) (b)
the evolution operator UM1 restricted on l2 (CM1 ) is as: for any |qi|τ i|ii ∈ CM1 ,
X
UM1 |qi|τ i|ii = δ1 (q, τ (i), σ, p, d)|pi|τ (i, σ)i|i + di (48)
p,σ,d∈{−1,1}

where
 d = −1 and d = 1 are identified with d = L and d = R, respectively, τ (i, σ)(j) =
 σ, if j = i, (b)
and UM1 is exactly unitary on subspace l2 (CM1 ) of l2 (CM1 ), where CM1
 τ (j), otherwise,
(b)
is the set of all configurations of QTM M1 . Note that if |qi|τ i|ii ∈ CM1 , then we view
τ (j) = B1 for |j| > t. We now show that δ2 satisfies the W-F conditions Eqs. (9,10) by
means of further extensions. We consider them by dividing the following cases.

(1) For the endmarker #, δ2 satisfies the W-F conditions. For any |qi|n1 , . . . , n2t+1 i ∈
(b)
CM2 , if (q, n1 , . . . , n2t+2 ) 6= (q0 , 0, . . . , 0) or (q0 , 0, . . . , 0, 1), then we define δ2 satisfies the W-F

23
conditions.
δ2 (q, n1 , . . . , n2t+2 , #, q, n1 , . . . , n2t+2 ) = 1; (49)

otherwise,
′ ′ ′
δ2 (q, n1 , . . . , n2t+2 , #, q , n1 , . . . , n2t+2 ) = 1. (50)

Then by combining the definitions Eqs.(28,29) of 1) above with (1) it is easy to check that
for input #, δ2 satisfies the W-F conditions Eqs. (9,10).

(2) For any input symbol σ ∈ Σ1 , δ2 satisfies the W-F conditions. For any |qi|n1 , . . . , n2t+2 i ∈
(b) ′ ′
CM2 , if (q, n1 , n2 , . . . , n2t+2 ) 6= (q0 , n1 , . . . , ni−1 , 0, . . . , 0, i) or (q0 , 0, . . . , 0, j) for any 1 ≤ i ≤

2t + 1 and 1 ≤ j ≤ 2t + 2 with nj > 0, j = 1, 2, . . . , i − 1, then we define

δ2 (q, n1 , n2 , . . . , n2t+2 , σ, q, n1 , n2 , . . . , n2t+2 ) = 1; (51)

and for the other cases we define

′ ′ ′ ′
δ2 (q, n1 , n2 , . . . , n2t+2 , σ, q , n1 , n2 , . . . , n2t+2 ) = 0. (52)

As well, it is ready to check that for any input σ ∈ Σ1 , δ2 satisfies the W-F conditions Eqs.
(b) (b)
(9,10) in CM2 , and therefore, VσM2 defined by Eq. (7) is a unitary operator on l2 (CM2 ).
(b)
(3) For tape symbol B2 , δ2 satisfies the W-F conditions. For any |qi|n1 , . . . , n2t+2 i ∈ CM2 ,
if
′ ′ ′
(q, n1 , n2 , . . . , n2t+2 ) 6= (q0 , n1 , n2 , . . . , nl , 0, . . . , 0, i) or (q0 , 0, . . . , 0, n1 , . . . , nl , 0, . . . , s) for
| {z }

any 0 ≤ l ≤ n and 0 ≤ s ≤ l + 1 with 0 ≤ i ≤ l + 1 and nj , nj > 0, j = 1, 2, . . . , l, then we
define
δ2 (q, n1 , n2 , . . . , n2t+2 , B2 , q, n1 , n2 , . . . , n2t+2 ) = 1, (53)

and for the other cases we define

′ ′ ′ ′
δ2 (q, n1 , n2 , . . . , n2t+2 , B2 , q , n1 , n2 , . . . , n2t+2 ) = 0. (54)

Then by combining 3) and the above definitions of δ2 we can easily know that for input B2 ,
(b)
δ2 satisfies the W-F conditions on CM2 , and therefore, VBM2 2 defined by Eq. (7) is a unitary
(b)
operator on l2 (CM2 ).
(b)
(4) For tape symbol B3 , δ2 satisfies the W-F conditions. For any |qi|n1 , . . . , n2t+2 i ∈ CM2 ,
if

24
′ ′ ′
(q, n1 , n2 , . . . , n2t+2 ) 6= (q0 , 0, . . . , 0, n1 , n2 , . . . , nl , 0, . . . , 0, i), or
| {z }
t+1
(p0 , e(B1 ), . . . , e(B1 ), n1 , n2 , . . . , nl , e(B1 ), . . . , e(B1 ), j) with 0 ≤ j ≤ t + 1,
| {z }
t+1
′ ′ ′
or (q0 , e(B1 ), . . . , e(B1 ), n1 , . . . , nl , e(B1 ), . . . , e(B1 ), t + 1)
| {z }
t+1

for any 0 < l ≤ n, any 0 < nj < r − 1, j = 1, 2, . . . , l, and t + 1 ≥ i ≥ 0, then we define

δ2 (q, n1 , n2 , . . . , n2t+2 , B3 , q, n1 , n2 , . . . , n2t+2 ) = 1, (55)

and for the other cases we define δ2 = 0 for input B3 .

Similar to the discussion of (3) above, we know that VBM3 2 defined by Eq. (7) is unitary
(b)
on l2 (CM2 ).

(5) Finally, we consider the case of input B4 . If mapping


(b) (b,+)
g : CM1 → CM2
is defined by

g|qi|τ i|ii = |qi|e(τ (−t)), e(τ (−t + 1)), . . . , e(τ (t))i|i + t + 1i,
(b) (b,+)
then g is a bijective mapping from CM1 to CM2 . By means of the definition δ2 for input B4 ,
(b)
we know that the unitarity of UM1 on l2 (CM1 ) implies that VB4 is also a unitary operator on
(b,+)
l2 (CM1 ). Indeed, if
X
UM1 |qi|τ i|ii = δ1 (q, τ (i), σ, p, d)|pi|τiσ i|i + di, (56)
p∈Q1 ,σ∈Σ1 ,d∈{−1,1}

then

VBM4 2 g(|qi|τ i|ii) = VBM4 2 |qi|e(τ (−t)), e(τ (−t + 1)), . . . , e(τ (i)), . . . , e(τ (t))i|i + t + 1i
X
= δ2 (q, e(τ (−t)), e(τ (−t + 1)), . . . , e(τ (i)), . . . , e(τ (t)),
p∈Q1 ,σ∈Σ1 ,d∈{−1,1}
B4 , p, e(τ (−t)), e(τ (−t + 1)), . . . , e(σ), . . . , e(τ (t))

|pi|e(τ (−t)), e(τ (−t + 1)), . . . , e(σ), . . . , e(τ (t))i


X
= δ1 (q, τ (i), σ, p, d)|pig(|pi|τiσ i|i + di). (57)
p∈Q1 ,σ∈Σ1 ,d∈{−1,1}

(b,+)
Therefore, VBM4 2 is unitary on l2 (CM1 ). Of course, VBM4 2 can be extended to be a unitary
(b)
operator on l2 (CM1 ).

The rest is to see that PaM1 (x) = PaM2 (x) for any input string x ∈ Σ∗1 . It exactly follows
from the above definitions regarding δ2 , and therefore, this completes the proof. 2

25

You might also like