QUANTUM TURING MACHINES

COMPUTATIONS AND MEASUREMENTS

STEFANO GUERRINI∗ , SIMONE MARTINI§ , AND ANDREA MASINI¶

arXiv:1703.07748v1 [cs.LO] 22 Mar 2017

Abstract. We propose a new formulation of Quantum Turing Ma-

chines, as an extension of those proposed by Bernstein and Vazirani.
For this new class of Quantum Turing Machines, both finite and in-
finite computations are meaningful—an infinite computation does not
correspond trivially to a divergent function. Moreover, we propose a
natural observation protocol for the new QTMs, that does not modify
the probability of the possible outcomes of the machines. Finally, we
use QTMs to define a class of quantum computable functions—any such
a function is a mapping from a general quantum state to a distribution
of probability of natural numbers.

1. Introduction
Quantum computability still needs a general theory of functions, akin to
the one developed at the mid of the twentieth century for classical machines,
and more adherent to Feynman’s original motivations of simulating physics
by computers [11].
Indeed, so far quantum computing has been studied mainly as a new,
efficient paradigm for classical, discrete functions. In this way, many suc-
cesses have been achieved in computational complexity (e.g., the definition
of quantum complexity classes and their relations with classical ones) and
algorithm design (e.g., Shor factorisation algorithm [28, 29]).
At the same time, it has been left unanswered the question of how (quan-
tum) physics might be naturally simulated, and not just encoded, by a
computing machine. In this direction, the first step is to look at quantum
computation devices as machines computing more general functions than
the traditional discrete functions over natural numbers (which is enough,
in the classical setting, since any other discrete domain may be classically
encoded into N). It is clear, however, that classical computations should
remain a particular case of the quantum one, and therefore one has to cope

Date: September 3, 2022.

∗
Partially supported by the Project ELICA (ref. ANR-14-CE25-0005), of the ANR
program “Fondements du numérique (DS0705) 2014”.
§
Partially supported by the Italian “National Group for Algebraic and Geometric
Structures, and their Applications” (GNSAGA-INDAM).
¶
Partially written while at LIPN, Institut Galilée, Université Paris 13, Sorbonne Paris
Cité as visiting researcher.
1
2 S. GUERRINI, S. MARTINI, AND A. MASINI

with the unavoidable constraints established by classical computability—

undecidability of program termination, first, and non-recursive enumerabil-
ity of total functions, then. As a result, in a general theory of quantum
computable functions one is forced to a framework where also non-total
functions are present, as a results of non-terminating, or infinite, computa-
tions.
A natural starting point are Quantum Turing Machines. We are of course
aware that other computational models might be considered, but we are also
convinced that, in the development of a quantum computability theory, we
cannot avoid QTMs. For instance, we cannot restrict to quantum circuits—
which on the other hand are the simplest setting for quantum computa-
tional complexity—since circuits can only represent classes of terminating
computations. Moreover, in defining families of circuits one must ensure
enumerability of such families, or equivalently, that they may be computed
by another device—usually, a classical Turing Machine, thus begging the
question.
We therefore start with an analysis of QTMs, and in particular of the
QTMs of Bernstein and Vazirani [3]. However, the standard presentation
of QTMs as devices computing discrete classical functions does not fit our
goal. Since unitarity of evolution forbids that a QTM may halt in a given
state, several constraints are imposed to guarantee that in a computation
starting on a single “natural number”, after a finite number of steps a sin-
gle result may be read, thus mimicking—in a sense—terminating classical
discrete computations. It is our assumption, instead, that the natural in-
put of a quantum computation should be a generic element of the Hilbert
space `2 (B) (for a discrete set B; N for instance), thus, in general, an infi-
nite denumerable sum of simple configurations. Moreover, we want to allow
meaningful infinite computations, where the global output is obtained only
as a limit.
One of the technical contributions of this paper is a definition of QTMs
in which, when a computation of the machine enters into a “final state”,
its evolution continues remaining in that final state, without changing the
output written on the tape. This kind of evolution is obtained by enriching
the machines by means of a suitable counter which plays no role during
the standard evolution of the machine, but starts to be increased when the
computation enters into a final state. Indeed, as already remarked, unitary
evolution of a QTM forces it to keep modifying its configuration—there
cannot be a halting state. Therefore, once a final state is reached, we assume
that the only possible “final evolution” of the machine is to increment by 1
the counter, leaving all the rest of the configuration (the internal state, the
position of the head, and the content of the tape) unchanged.
Such an approach might remind the so-called “ancilla” proposed in the
literature. However, it is known that if the ancilla is introduced in a naive
way, no room is left for any interesting evolution of the QTM—see, for
instance [16], whose approach leads the authors to argue that the constraint
 QUANTUM TURING MACHINES: COMPUTATIONS AND MEASUREMENTS 3

that the output, modulo the ancilla, does not change leaves as only possible
machines with unitary evolution those which never enter into a final state.
Our definition of the evolution of QTMs shows instead that we can get a
unitary evolution in which, once achieved, an output information is definitely
preserved, allowing a computing schema in which the output is constructed
incrementally. The key tool for this result is the separation of the space of the
states of the final evolutions—those performed by branches of computation
which have already produced an output (since they reached a final state),
and which are characterised by a non zero counter—from the space of the
states of the main evolution of the machine—the ones in which the input
is transformed into the output, and for which the counter is set to zero.
We stress that the above final behaviour cannot be restricted to a unique
final state, but it must be extended to a set of generalised final states—
that we name target states in the paper—that behave as sinks from which
the machine cannot get out, and in which the computation continues in a
fixed way that accords with the unitarity restriction, without modifying the
content of the tape. In a sense, these additional target states play the same
role of error states in classical deterministic TMs, allowing to complete the
behaviour of a machine when it is not completely specified.
Once resolved the question of the final evolution of the machine, it re-
mains open the question of the evolution before the initial state. Indeed, by
inverting the unitary operator describing the evolution of a QTM, (possi-
ble) configurations of the machines must be defined before the initial state.
To solve this problem, Bernstein and Vazirani back-connect the final states
of their machines to the initial state. According to our previous discussion,
such an approach is incompatible with our goal of indefinitely preserving the
ability to read an output after it has been produced. Moreover, as we shall
discuss in more details later, such an approach can be reasonably taken into
account only in the case of the so-called well-behaved QTMs of Bernstein
and Vazirani, in which all the superposed configurations of the machine en-
ter the final state at the same time. To tackle (pre-)initial evolution, we
introduce a notion of source states, dual to the target states, and among
which there is an initial state in which we assume to set the machine to
start a computation. As for the case of target states, this general notion of
source state fixes at the same time the problem of the reverse evolution from
the initial state, and the possibility to complete a partially specified machine
in which, in order to satisfy the local conditions characterising the unitary
condition, some additional transitions must be added to those describing
the desired function. If needed, such missing transitions can be taken from
source states that, being not reachable by any other state, would not inter-
fere with the desired behaviour that we want to observe when starting from
a valid configuration.
By keeping distinct the initial and the final evolution of the machine
from the main one, we preserve one of the key properties of Bernstein and
Vazirani’s approach: QTMs can be characterised by a set of local conditions
4 S. GUERRINI, S. MARTINI, AND A. MASINI

on their transition function (see Theorem 12). Such conditions are the same
of Bernstein and Vazirani. Thus, any Bernstein and Vazirani QTM M can be
immediately extended into one of our QTMs whose main transition function
corresponds to that of M .
In conclusion, we show that it is possible to define QTMs such that:
(i) they take as input a general quantum state—that is, a denumerable
superposition of simple configurations representing natural numbers; and (ii)
they may have meaningful infinite behaviours, whose “output” is obtained
as a limit of finite portions of the computation. We stress that we are not
interested in QTMs as devices computing functions over natural numbers.
Instead, we see a QTM as defining a function from a general quantum state
to a distribution of probability of natural numbers. We believe this could
be the starting point of a general theory of quantum computable functions.
1.1. Content of the paper. Despite the availability of a large corpus of
results1, quantum computability still lacks a general treatment akin to clas-
sical computability theory. Taking as a reference model (quantum) Turing
machines, one of the main obstacles is that while it is obvious how to under-
stand a classical Turing machine (TM) as a device computing a numerical
function, the same is not so for a quantum Turing machine (QTM).
In a naı̈ve but suggestive way, a QTM may be described as a classical
TM which, at any point of its computation, evolves into several different
classical configurations, each characterised by a complex amplitude. Such
different configurations should be imagined as simultaneously present, P “in
superposition”—a simultaneity formally expressed as a weighted sum di Ci
of classical configurations Ci , with complex coefficients di . Even when start-
ing from a classical configuration, in a QTM, there is not a single result, but
a superposition from which we can read off several numerical “results” with
certain probabilities2. Moreover, QTMs never have genuinely finite compu-
tations and one should therefore find a way to define when and how a result
can be read off.
We propose a notion of “function computable by a quantum Turing ma-
chine,” as a mapping between superpositions of initial classical configura-
tions3 to probability distributions of natural numbers, which are obtained
(in general) as a limit of an infinite QTM computation.
Before reaching this point, however, we must go back to the basics, and
look to the very notion ofPQTM. Because, if it is true that configurations of a
QTM are superpositions di Ci of classical configurations, quantum physics
1See, in the large literature, [10, 18, 20, 23, 30] for fundamentals results, [3] for the
foundations of quantum complexity, or [1, 15, 6, 7, 5, 8, 17, 26, 27, 31] for more language
oriented papers.
2 We cannot observe the entirety of a superposition without destroying it. But if we
insist on observing it, then we will obtain one of the classical configurations Ci , with
probability |di |2 .
3
More precisely, the domain of our functions will be the Hilbert space `21 (N) of square
summable, denumerable sequences of complex numbers, with unitary norm.
 QUANTUM TURING MACHINES: COMPUTATIONS AND MEASUREMENTS 5

principles impose severe constraintsPon the possible evolutionsP of such ma-

chines. First, in any superposition di Ci , we must have |di |2 = 1. Sec-
ond, there cannot be any finite computation—we may of course name a state
as “final” and imagine that we read the result of the computation (whatever
this means) when the QTM enters such a state, but we must cope with
the fact that the computation will go on afterwards. Moreover, since any
computation of a QTM must be reversible [2], in the sense that the operator
describing the evolution of the QTM must be invertible, we cannot neither
force the machine to loop on its final configuration. On the other hand, be-
cause of reversibility, even the initial configuration must have a predecessor.
Summing up, an immediate consequence of all the above considerations is
that every state must have at least one incoming and one outgoing transi-
tion and that such transitions must accord to several constraints forced by
quantum physics principles. In particular, transitions must enter the initial
state—since a priori it might be reached as the evolution from a preceding
configuration — and exit the final state—allowing the machine configuration
to evolve even after it has reached the final result.
The reversibility physical constraints are technically expressed by the re-
quirement that the evolution operator of a QTM be unitary. If we now want
to use a QTM to compute some result, we are still faced with the problem
of when (and how, but for the moment let’s postpone this) reading such a
final result, given that the computation evolves also from the final state, and
that, without further constraints, it might of course evolve in many, different
possible ways. Bernstein and Vazirani in their seminal paper [3] (from now
on we shall refer to this paper as “B&V”) first define (non unitary) QTMs;
select then the “well-formed” QTMs as the unitary-operator ones; and de-
fine finally “well-behaved” QTMs as those which produce a superposition in
which all classical configurations are simultaneously (and for the first time)
in the final state. What happens after this moment, it is not the concern of
B&V.
Our goal is to relax the requirement of simultaneous “termination”, allow-
ing meaningful computations to reach superpositions in which some classical
configurations are final (and give a “result”), and some are not. Those which
are not final, should be allowed to continue the computation, possibly reach-
ing a final state later. The “final result” will then be defined as a limit of
this process. In order to assure that at every step of the computation the
superposition of the final configurations is a valid approximation of the limit
result, we must guarantee that once a final state is entered, the “result” that
we read off is not altered by the further (necessary, by unitarity) evolution.
To obtain this, we restrict the transition function out of a final state, without
violating unitarity. We obtain this goal by using an integer counter:

• Once a final state is entered, the counter starts counting the number
of steps during which the computation has been looping into the final
state; that is, when in a final state, at each step, the machine just
6 S. GUERRINI, S. MARTINI, AND A. MASINI

increases the counter by 1, leaving unchanged the state, the tape,

and the position of the head.
• Dually, in an initial state, the counter gives the number of steps
before the machine might actually start its main computation. That
is, when a machine is in initial state with a counter greater than 0,
it rollbacks to another initial configuration which differ for the value
of the counter only, since it has been decreased by 1.
• Finally during the normal or main evolution (i.e. the quantum evo-
lution as defined by B&V) the counter does not play any role and its
value is zero.
In the paper we will apply this approach to a generalisation of initial/final
states, called here source and target states.
After the definition of QTMs and of the corresponding functions, we will
discuss their expressive power, comparing them to the QTMs studied in
the literature. The QTMs of B&V form a robust class, but meaningful
computations are defined only for classical inputs (a single natural number
with amplitude 1). Moreover, their QTMs “terminate” synchronously—
either all paths in superpositions enter a final state at the same time, or all
of them diverge. As a consequence, there is no chance to study—and give
meaning—to infinite computations. More important, the class of “sensible”
QTMs (in B&V’s terminology: the well-formed, well-behaved, normal form
QTMs) is not recursively enumerable, since the constraint of “simultaneous
termination” is undecidable.
In Deutsch’s original proposal [10], any quantum TM has an explicit ter-
mination bit which the machine sets when entering a final configuration.
While it is guaranteed that final probabilities are preserved, the observa-
tion protocol requires that one checks termination at every step, since the
machine may well leave the final state (and change the tape). Deutsch’s ma-
chines could in principle be used to define meaningful infinite computations,
but we know of no such an attempt.
In our analysis: (i) there is no termination bit: a quantum configuration
is a genuine superposition of classical configurations; (ii) any computation
path (that is, any element of the superposition) evolves independently from
the others: any path may terminate at its own time, or may diverge; (iii)
infinite computations are meaningful; (iv) we may observe the result of the
computation in a way akin to Deutsch’s one, but with the guarantee that
once a final state is entered, the machine will not change the correspond-
ing “result” during a subsequent computation; (v) the class of QTMs is
recursively enumerable, thus opening the door to a quantum computability
theory which may follow some of the classical developments.

2. Quantum Turing Machines

In this section we define quantum Turing machines. We assume the reader
be familiar with classical Turing machines (in case, see [9]).
 QUANTUM TURING MACHINES: COMPUTATIONS AND MEASUREMENTS 7

As for classical Turing Machine (TM), a Quantum Turing Machine (QTM)

has a tape (a sequence of cells) containing symbols (one in each cell) from
a finite tape alphabet Σ, which includes at least the symbols 1 and 2: 1
is used to code natural numbers in unary notation, while 2 is the blank
symbol. We shall consider computations starting from tapes containing a
sequence of n + 1 symbols 1 (the encoding of the natural number n); thus,
in the following, for any n ∈ N, we shall use n to denote the string 1n+1 . By
the greek letters α, β, eventually indexed, we shall instead denote strings in
Σ∗ , and by αβ we shall denote the concatenation of α and β. Finally, we
shall use λ to denote the empty string.
2.1. Plain configurations. The basic elements to describe the configura-
tion of a QTM are the finite sequences of symbols on the tape (as usual,
we assume that only a finite portion of the tape contains non-blank sym-
bols), the current internal state of the machine, and the current position
of the head reading a symbol in a cell of the tape. More precisely, a plain
configuration of a given QTM M is a triple hα, q, βi ∈ Σ∗ × Q × Σ∗ , s.t.:
(1) q ∈ Q is the current state, where Q is the finite set of the internal
states of M ;
(2) β ∈ Σ+ is the right content of the tape, where Σ is the tape alphabet
of the machineM . The first symbol of β is the one in the current
cell of the tape, that is, the one read by the head of M . In detail,
β = uβ 0 , where the current symbol u is the content of the current
cell and β 0 is the longest string on the tape ending with a symbol
different from 2 and whose first symbol (if any) is written in the cell
immediately to the right of the current cell; by convention, when the
current cell and all the right content of the tape is empty, we shall
also write hα, q, λi instead of hα, q, 2i;
(3) α is the left content of the tape. That is, it is either the empty string
λ, or it is the longest string on the tape starting with a symbol differ-
ent from 2, and whose last symbol is written in the cell immediately
to the left of the current cell.
According to this definition, in a configuration hα, q, βi the string α does
not start with a 2, and β does not end with a 2. However, since it will be
useful to manipulate configurations which are extended with blank cells to
the right (of the right content) or to the left (of the left content), we equate
configurations up to the three equivalence relations induced by the following
equations
α 'l 2α β 'r β2
0 0
hα, q, βi ' hα , q, β i when α 'l α0 and β 'r β 0
2.2. Hilbert space of configurations. A quantum configuration of a
QTM is not a simple configuration as the ones of a TM, but a weighted
superposition of configurations: a vector of the Hilbert space `2 (C), where
C is a suitable set of simple configurations as the plain ones defined above.
8 S. GUERRINI, S. MARTINI, AND A. MASINI

We recall that, for any denumerable set B, `2 (B) is the Hilbert space of
square summable B-indexed sequences of complex numbers
( )
X
φ:B→C| |φ(C)|2 < ∞
C∈B
p
equipped with an inner product h. | .i and the euclidean norm kφk = hφ | φi,
and that `21 (B) denote the set of vectors {φ | φ ∈ `2 (B) & kφk = 1}.
By using Dirac notation, we shall write |φi to denote the vector of `2 (B)
corresponding to the function φ : B → C. Moreover, for every C ∈ B, we
shall write |Ci to denote the vector corresponding to C, that is, the function
equal to 1 on C, and equal to 0 on the other elements
2
P of B. Finally, we
remark that, any vector |φi of ` (B) can be written as i∈I di |Ci i, for some
denumerable set of indexes I s.t. {Ci | i ∈ I} ⊆ B and di ∈ C, for every i ∈ I.
For more details on the basic notions of Hilbert spaces, see Appendix A.
2.3. Transitions of a QTM. Given a quantum configuration φ of a QTM
M , the main idea introduced by Deutsch [10] is that the machine evolves into
another quantum configuration |ψi = U |φi, where U is a unitary operator
on the Hilbert space of the configurations of M . By linearity, this also means
that, if C is the set of simple configurations on which we define the Hilbert
space of quantum configuration `2 (C) of M , the operator U , and then the
behaviour of M , is completely determined by the value of U on the elements
of C. B&V [3] refine this point by assuming that, as in a classical TM, for
every C ∈ C, the transition from |Ci to a new configuration U |Ci depends
only on the current state of M and on the current symbol of C, and that
U |Ci is formed of configurations obtained by replacing the current symbol
u of C with a new one, and by moving the tape head to the left or the right.
Therefore, if we denote by D = {L, R} the set of the possible movements of
the tape (where L stands for left and R for right), and q is the current state
of M , we have X
U (|Ci) = δ(q, u)(p, v, d) |Cp,v,d i
p,v,d
where p ranges over the states of M , v ranges over its tape alphabet,
δ(q, u)(p, v, d) ∈ C, and Cp,v,d is the new configuration obtained from C
by changing the current state from q to p, by replacing the current symbol
u with v, and by moving the tape head in the direction d.
2.4. Initial and final configurations. In B&V, the transition rule de-
scribed above applies to every state of the machine. However, as already
remarked in the introduction, this implies a severe restriction on the ma-
chines that one can actually consider as valid ones, since the problem of
how to read the output forces to ask that, in a “well-behaved” QTM, all the
states of the configuration in superposition be final.
The key point is that a QTM cannot stop into some final configuration
Cf , since the unitarity of U requires that U (|Cf i) be defined. On the other
 QUANTUM TURING MACHINES: COMPUTATIONS AND MEASUREMENTS 9

hand, we cannot resort to the assumption that after reaching some final
configuration Cf , the computation loops on it, as this would imply that Cf
could not be reached from any other configuration D ∈ C: if U (|Cf i) = |Cf i,
then |Cf i = U −1 (|Cf i), and U (|Di)(Cf ) = hU(|Di) | Cf i = D | U−1 (|Cf i) =
hD | Cf i = 0, for D 6= Cf . Because of this, when a final configuration Cf
is reached, the machine must evolve into a different configuration which
preserves the output written on the tape, preserving at the same time the
unitarity of the evolution operator U .
To solve the above problem, we assume that, for every final configuration
Cf we have a denumerable set of indexed configurations hCf , ni, obtained
by adding a counter n ∈ N to Cf . The configuration hCf , 0i plays the usual
role of the plain configuration Cf and is the only one that can be reached
from a non final configuration; for every n ∈ N, the configuration hCf , n + ii
is instead the only successor of hCf , ni, that is, U (|Cf , ni) = |Cf , n + 1i,
where |Cf , ni is the base vector corresponding to hCf , ni. In other words,
the counter n is initialised to 0 and plays no role until the state is not final;
when a branch of the computation enters a final state, the counter is still at
0, but it is increased by 1 at each following step.
Final configurations are not the only configurations on which it would be
useful to loop. In fact, in some cases, in order to assure that the operator
U is unitary, we need to introduce additional target configurations that
behave has sinks, from which the machine cannot get out (see the example
in subsection 5.3, whose graph representation is given in Figure 4). We have
then to consider a whole set Qt of target states, containing the final state
qf , s.t. for every plain configuration Ct = hα, qt , βi, with qt ∈ Qt , we have a
set of indexed configurations hCt , ni, for n ∈ N, s.t. U (|Ct , ni) = |Ct , n + 1i.
Finally, let us remark that when U is unitary, its inverse U −1 is unitary
too, and can be applied to any initial configuration. Therefore, even if we
assume that a computation always starts from some initial configuration
Ci = hα, qi , βi, such a Ci must have a predecessor Ci,1 = U −1 (|Ci i), and
more generally a n-predecessor Ci,n = U −n (|Ci i). As already done for fi-
nal configurations, we associate to every Ci a set of indexed configurations
hCi , ni, with n ∈ N, s.t. U −1 (|Ci , ni) = |Ci , n + 1i, for n ∈ N. For n > 0,
we get then U (|Ci , ni) = |Ci , n − 1i; while U (|Ci , 0i) has the usual behaviour
expected from the machine on the plain initial configuration Ci . As for final
state and final configurations, it is useful to generalise the above behaviour
to a set of source configurations corresponding to a set Qs of source states
which contains the initial state qi .

2.5. Pre Quantum Turing Machines. In order to formally define QTMs,

we define first a notion of pre Quantum Turing Machine, or pQTM. As we
shall see later (Definition 7), a QTM is a pQTM whose evolution operator
is unitary. Since the behaviour on target states and on source states whose
counter is greater than 0 is fixed, in order to completely describe a pQTM, it
suffices to give a transition function δ0 : ((Q \ Qt ) × Σ) → `21 ((Q \ Qs ) × Σ × D)
10 S. GUERRINI, S. MARTINI, AND A. MASINI

which, for every configuration C with current state q and current symbol u,
gives the weight δ0 (q, u)(p, v, d) of |Cp,v,d i in the superposed configuration
reached from |Ci, where Cp,v,d is obtained by replacing v for u, by moving
the tape in the d direction, and by changing the current state to p.
Definition 1 (Pre Quantum Turing Machine). Given a finite set of states
Q and an alphabet Σ, a pre Quantum Turing Machine (pQTM) is a tuple
M = hΣ, Q, Qs , Qt , δ0 , qi , qf i
where
• Qs ⊆ Q is the set of source states of M , and qi ∈ Qs is a distinguished
source state named the initial state of M ;
• Qt ⊆ Q is the set of target states of M , and qf ∈ Qt is a distinguished
target state named the final state of M ;
• δ0 : ((Q \ Qt ) × Σ) → `2 ((Q \ Qs ) × Σ × D) is the quantum transition
function of M , where D = {L, R}.
2.6. Configurations. We have already said that, in order to properly deal
with source and target states, we have to associate a counter to them. For the
sake of uniformity, we shall add a counter to every plain configuration. Even
if, its value get stuck to 0 for every non-source or non-target configuration.
Remark 2 (The counter). The counter can be seen as an additional device
(for instance, as an additional tape or as a counting register) or directly
implemented by a suitable extension of the basic Turing machine (for in-
stance, by extending the tape alphabet). None of these implementations is
standard or has a direct influence in what will be presented in the following.
The key issue pointed out in the above discussion is that, for every target
configuration C of the QTM M , we need to have a denumerable space of
configurations isomorphic to {C} × N, in which the evolution of M from
C is confined, and s.t. UMk (|Ci) is in bijection with the k-th element hC, ki

of this space. Analogously, for source configurations, when considering the

−1
inverse time evolution operator UM . A more detailed discussion of the
implementation of the counter is given in Appendix B.
Definition 3 (configurations). Let M = (Σ, Q, Qs , Qt , δ0 , q0 , qf ) be a pQTM.
A configuration of M is a quadruple hα, q, β, ni ∈ Σ∗ × Q × Σ∗ × N, where
hα, q, βi is a plain configuration, and n is a counter associated to the config-
uration s.t. n = 0, when q 6∈ Qs ∪Qt . A configuration of M is a source/target
configuration when the corresponding state is a source/target state, and it is
a final/initial configuration when the current state is final/initial. We have
the following notations:
• CM is the set of the configurations of M .
• CsM and CtM are the sets of the source and of the target configurations
of M , respectively
f in
• Cinit
M and CM are the sets of the initial and final configurations of
M , respectively.
 QUANTUM TURING MACHINES: COMPUTATIONS AND MEASUREMENTS 11

• C0M is the of the configurations hα, q, β, 0i of M .

In the following, the index M in CM and in the other names indexed by
the machine may drop when clear from the context.
2.7. Quantum configurations. The evolution of a pQTM is described by
superpositions of configurations in CM . If B ⊆ CM is a set of configurations,
a superposition of configurations in B is a vector of the Hilbert space `2 (B)
(see, e.g., [4, 25]). Quantum configurations of a pQTM M are the elements
of `21 (CM ) (namely, the unit vectors of `2 (CM )). Since there is no bound on
the size of the tape in a configuration, the set CM is infinite and the Hilbert
space of the configurations `2 (CM ) is infinite dimensional.
Definition 4 (quantum configurations). Let M be a pQTM. The elements
of `21 (CM ) are the q-configurations (quantum configurations) of M .
We shall use Dirac notation (see Appendix A) for the elements φ, ψ of
`2 (CM ), writing them |φi , |ψi.
Definition 5 (computational basis). For any set of configurations B ⊆ CM
and any C ∈ B, let |Ci : B → C be the function

1 if C = D
|Ci (D) =
0 if C 6= D.
The set CB(B) of all such functions is a Hilbert basis for `2 (B) (see, e.g.,
[20]). In particular, following the literature on quantum computing, CB(CM )
is called the computational basis of `2 (C). Each element of the computational
basis is called base q-configuration.
With a little abuse of language, we shall write |Ci ∈ |φi when φ(C) 6= 0.
The span of CB(B), denoted by span(CB(B)), is the set of the finite linear
combinations with complex coefficients of elements of CB(B); span(B) is a
vector space, but not a Hilbert space. In order to get a Hilbert space from
span(CB(B)) we have to complete it, and `2 (B) is indeed the unique (up to
isomorphism) completion of span(CB(B)) (see [3]). As a consequence, any
linear operator U on span(CB(B)) has a unique extension on `2 (B), and,
when U is unitary, its extension is unitary too.
For some basic definitions, properties and notations on Hilbert spaces with
denumerable basis, see Appendix A. In particular, subsection A.1 presents a
synoptic table of the so-called Dirac notation that we shall use in the paper.
2.8. Time evolution operator and QTM. In order to define the evolu-
tion operator of a pQTM M , it suffices to give its behaviour on the compu-
tational basis CB(CM ). In particular, we have to distinguish three cases:
(1) C ∈ C0M \ CtM .
Let Cp,v,d ∈ C0M \ CsM be the configuration obtained by leaving to 0
the counter, by replacing the symbol u in the current cell with the
symbol v, by moving the head on the d direction, and by setting the
12 S. GUERRINI, S. MARTINI, AND A. MASINI

machine into the new state p. In detail, if C ' hαcw, q, uβ, 0i we

have
(
hαwv, p, β, 0i when d = R
Cp,v,d '
hα, p, wvβ, 0i when d = L.

and we define
X
W0,M (|Ci) = δ0 (q, u)(p, v, d) |Cp,v,d,k i
(p,v,d)∈(Q\Qs )×Σ×D

where δ0 is the quantum transition function of M .

(2) C ∈ CsM \ C0M .
Let C−1 ∈ CsM be the source configuration obtained by decreasing
by 1 the counter of C, we define

Ws,M (|Ci) = |C−1 i

(3) C ∈ CtM .
Let C+1 ∈ CtM \ C0M be the target configuration obtained by increas-
ing by 1 the counter of C, we define

Wt,M (|Ci) = |C+1 i

We have then three linear operators

W0,M : span(CB(C0M \ CtM )) → span(CB(C0M \ CsM ))

Ws,M : span(CB(CsM \ C0M )) → span(CB(CsM ))
Wt,M : span(CB(CtM )) → span(CB(CtM \ C0M ))

defined on three disjoint subspaces corresponding to a partition of the whole

space span(CB(CM )). Moreover, since even the images of these three oper-
ators are disjoint and cover the whole space span(CB(CM )), their sum

WM = W0,M + Ws,M + Wt,M

defines an automorphism on the linear space of q-configurations

WM : span(CB(C)) → span(CB(C))

By completion, WM extends in a unique way to an operator on the Hilbert

space of q-configurations.

Definition 6 (time evolution operator). The time evolution operator of M

is the unique extension

UM : `2 (CM ) → `2 (CM )

of the linear operator WM : span(CB(C)) → span(CB(C)).

 QUANTUM TURING MACHINES: COMPUTATIONS AND MEASUREMENTS 13

As for the operator WM , the time evolution operator UM can be decom-

posed into three operators
U0,M : `2 (C0M \ CtM ) → `2 (C0M \ CsM )
Us,M : `2 (CsM \ C0M ) → `2 (CsM )
Ut,M : `2 (CtM ) → `2 (CtM \ C0M )
s.t.
UM = U0,M + Us,M + Ut,M
Definition 7 (QTM). A pQTM is a Quantum Turing Machine (QTM)
when its time evolution operator UM is unitary.
2.9. Computations. A computation of a QTM is an iteration of its evo-
lution operator on some q-configuration. Since the time evolution operator
of a QTM is unitary, it preserves the norm of its argument and maps q-
configurations into q-configurations. By the way, this holds for the inverse
−1
UM of the time evolution operator too.
Fact 8. Let M be a QTM. If |φi ∈ `21 (CM ), then UM
i |φi ∈ `2 (C ), for every
1 M
i ∈ Z.
Definition 9 (initial and final configurations). A q-configuration |φi is
initial when |φi ∈ `21 (Cinit 2 f in
M ) and is final when |φi ∈ `1 (CM ). By |ni we
denote the initial configuration hλ, q0 , n, 0i ∈ Cinit
M .

Definition 10 (computations). Let M be a QTM and let UM be its time evo-

lution operator. For an initial q-configuration |φi ∈ `21 (Cinit ), the computation
of M on |φi is the denumerable sequence {|φi i}i∈N s.t.
(1) |φ0 i = |φi;
(2) |φi i = UMi |φi.

Clearly, any computation of a QTM M is univocally determined by its

initial q-configuration. The computation of M on the initial q-configuration
M.
|φi will be denoted by K|φi

Remark 11. The definition of time evolution operator ensures that the fi-
nal configurations reached along a computation are stable and do not in-
terfere with other branches of the computation in superposition, which
may enter into a final configuration later. Indeed, given a configuration
|φi = |φf i + |φnf i, where |φf i ∈ `2 (Cf in ) and |φnf i does not contain any
final configuration, let ψ = U i |φi = U i |φf i + U i |φnf i. Any final con-
figuration in U i |φnf i has a value of the counter less than i, while any
final configuration |C, ki ∈ |φi formed of a plain configuration C and a
counter k is mapped into a configuration |C, i + ki ∈ ψ. Moreover, |C, ki
and |C, i + ki have the same coefficient in |φi and |ψi, respectively, since
hψ | C, i + ki = U i |φi , U i |C, ki = hφ | C, ki.
14 S. GUERRINI, S. MARTINI, AND A. MASINI

2.10. Local conditions for unitary evolution. In analogy of the main

approaches in literature [3, 23], it is possible to state a set of local condition
for the quantum transition function δ0 of pQTMs in order to ensure that
the time evolution operator is unitary.
Theorem 12. Let M be a pQTM with quantum transition function δ0 .
The time evolution operator UM : `2 (CM ) → `2 (CM ) of M is unitary iff δ0
satisfies the local conditions:
(1) for any (q, a) ∈ (Q \ Qt ) × Σ
X
|δ0 (q, a)(p, b, d)|2 = 1
(p,b,d)∈(Q\Qs )×Σ×D

(2) for any (q, a), (q 0 , a0 ) ∈ (Q \ Qt ) × Σ with (q, a) 6= (q 0 , a0 )

X
δ0 (q 0 , a0 )(p, b, d)∗ δ0 (q, a)(p, b, d) = 0
(p,b,d)∈(Q\Qs )×Σ×D

(3) for any (q, a, b), (q 0 , a0 , b0 ) ∈ (Q \ Qt ) × Σ2

X
δ0 (q 0 , a0 )(p, b0 , L)∗ δ0 (q, a)(p, b, R) = 0
p∈(Q\Qs )

The proof of the above theorem follows the main steps of the proofs given
in the literature for B&V QTMs [3, 23]. First of all, one proves that the
time evolution operator UM is an isometry of `2 (C), and then that UM is
surjective. The first step is rather straightforward, since one can easily find
the adjoint UM∗ of U
M and prove that it is a left-inverse of UM iff the local
conditions hold (we recall that an operator U is an isometry iff U ∗ U = 1,
where U ∗ is the adjoint of U ). The second step is the hard part of the proof.
Both [3] and [23] provide involved proofs of the fact that U ∗ is surjective.
A previous version of the present paper included a direct and much simpler
proof of Theorem 12—instead of proving that UM ∗ is surjective, we show
∗
that UM is the right inverse of UM . We omit here the details of the proof
of Theorem 12, and we refer the interested reader to the above mentioned
version of the paper4 available on ArXiV [12].
2.11. A comparison with Bernstein and Vazirani’s QTMs: part 1.
We refer to B&V for the precise definitions of the QTMs used in that paper.
For the sake of readability, we informally recall the notion of what they call
well formed, stationary, normal form QTMs (B&V-QTMs in the following).
A B&V-QTM M = hΣ, Q, δ, q0 , qf i is defined as our QTM (with one source
state and one target state only) with the following differences:
(1) the set of configurations coincides with all possible classical config-
urations, namely all the set Σ∗ × Q × Σ∗ .
4The definition of QTM in [12] is slightly different then the one given here. In particular,
it differs for the evolution in source and target states. However, the definition of QTM in
[12] can be just seen as a particular implementation of the one given here (see Appendix B),
and the proofs in the appendix of [12] easily adapt.
 QUANTUM TURING MACHINES: COMPUTATIONS AND MEASUREMENTS 15

(2) no superposition is allowed in the initial q-configuration (it must be

a classical configuration hα, q, βi with amplitude 1);
(3) let |Ci be such an initial configuration and let
j
k = min{j | UM |Ci contains a final configuration}
If such a k exists, then (i) all the configurations in UM k |Ci are final;
(ii) for all i < k, UM i |Ci does not contain any final configuration.
We say in this case that the QTM halts in k steps in UM k |Ci;

(4) if a QTM halts, then the tape head is on the start cell of the initial
configuration;
(5) there is no counter and the transitions out of the final state or into
the initial state are replaced by loops from the final state into the
initial state, that is, δ(qf , a)(q0 , a, R) = 1 for every a ∈ Σ. There-
fore, because of the local unitary conditions, that must hold in the
final state also, these are the only outgoing transitions from the fi-
nal state, and the only incoming ones into the initial state, that is,
δ(qf , a)(q 0 , a0 , d) = 0 if (q 0 , a0 , d) 6= (q0 , a, R) and δ(q 0 , a0 )(q0 , a, d) = 0
if (q 0 , a0 , d) 6= (qf , a, R).
Theorem 13. For any B&V-QTM M there is a QTM M 0 s.t. for each initial
configuration |Ci, if M with input |Ci halts in k steps in a final configuration
|φi = UMk |Ci, then U k |Ci = |φi.
M0

Proof. The QTM M 0 has the same states of M , the initial state q0 is its
only source state, and the final state qf is its only target state. Therefore,
if M = hΣ, Q, δ, q0 , qf i, we take M 0 = hΣ, Q, {q0 }, {qf }, δ0 , q0 , qf i, where δ0
is the restriction of δ to ((Q \ Qt ) × Σ) → `2 ((Q \ Qs ) × Σ × D), that is, for
each q 6= qf and a ∈ Σ, we have δ0 (q, a)(p, b, d) = δ(q, a)(p, b, d), for every
(p, b, d) ∈ (Q \ Qs ) × Σ × D. Since the local unitary conditions hold for δ
and, in a B&V-QTM, δ(q, a)(q0 , b, d) = 0, when q 6= qf , the unitary local
conditions hold for δ0 too.
By construction, if UM i |Ci is not final for 0 ≤ i < k, then |φ i = U k |Ci =
k M
k
UM 0 |Ci. In particular, this holds when |φk i is the final configuration of the
B&V-QTM M . 

3. Quantum Computable Functions

In this section we address the problem of defining the concept of quantum
computable function in an “ideal” way, without taking into account any
measurement protocol. The problem of the observation protocol will be
addressed in Section 4. Here we show how each QTM naturally defines a
computable function from the sphere of radius 1 in `2 (CM ) to the set of
(partial) probability distributions on the set of natural numbers.

3.1. Probability distributions.

Definition 14 (Probability distributions).
16 S. GUERRINI, S. MARTINI, AND A. MASINI

(1) A partial probability distributionP (PPD) of natural numbers is a func-

tion
P P : N → R [0,1] such that n∈N P(n) ≤ 1.
(2) If n∈N P(n) = 1, P is a probability distribution (PD).
(3) P and P1 denote the sets of all the PPDs and PDs, respectively.
(4) If the set {n : P(n) 6= 0} is finite, P is finite.
(5) Let P0 , P00 be two PPDs, we say that P0 ≤ P00 (P0 < P00 ) iff for each
n ∈ N, P0 (n) ≤ P00 (n) (P0 (n) < P00 (n)).
(6) Let P = {Pi }i∈N be a denumerable sequence of PPDs; P is monotone
iff Pi ≤ Pj , for each i < j.
Remark
P 15. In the following, we shall also use the notation P(⊥) = 1 −
n∈N P(n). By definition, 0 ≤ P(⊥) ≤ 1, and a PPD is a PD iff P(⊥) = 0.
We also stress that ≤ is a partial order of PPDs and that any PD P is
maximal w.r.t. to ≤, since P ≤ P0 iff P = P0 , for any PPD P0 .
Definition 16 (limit of a sequence of PPDs). Let P = {Pi }i∈N be a sequence
of PPDs. If for each n ∈ N there exists ln = limi→∞ Pi (n), we say that
limi→∞ Pi = P : N → R[0,1] , with P(n) = ln .
3.2. Monotone sequences of probability distributions.
Proposition 17. Let P = {Pi }i∈N ⊆ P be a monotone sequence of PPDs.
it is the supremum P of P ;
F
(1) P
limi→∞FPi exists and
( P)(n) =
F P
(2) n∈N Pi (n) i∈N ;
F n∈N
(3) P ∈ P.
Proof. Since Pi (n) ≤ 1, every non-decreasing sequence {Pi (n)}i∈N has a
supremum sup{Pi (n)}i∈N = limi→∞ Pi (n). Thus, limi→∞ Pi is defined and
Pi ≤ limi→∞ Pi , for every i ∈ N. On the other hand, for any P0 s.t.
Pi ≤ P0 for i ∈ N, we have limi→∞ Pi (n) = sup{Pi (n)}i∈N ≤ P0 (n), for
every n ∈ N; namely, limi→∞ Pi ≤ P0 . We can then conclude (item 1) that
P = limi→∞ Pi .
F
P)(n) for
F
Let us now prove item 2. First of all, since 0 ≤ P i (n) ≤ (
i, n ∈ N, we have n∈N ( P)(n) ≥ sup
P F  P
n∈N Pi (n) i∈N and
( )  
X X 
sup Pi (n) ≥ sup Pi (n)
 
n∈N i∈N n≤k
i∈N
X X G
= sup {Pi (n)}i∈N = ( P)(n)
n≤k n≤k
for any k ∈ N. Thus,
( )  
X X G 
sup Pi (n) ≥ sup ( P)(n)
 
n∈N i∈N n≤k
k∈N
X G X G
= lim ( P)(n) = ( P)(n)
k→∞
n≤k n∈N
 QUANTUM TURING MACHINES: COMPUTATIONS AND MEASUREMENTS 17

Summing up, we can conclude, as

( )
X G X X G
( P)(n) ≥ sup Pi (n) ≥ ( P)(n)
n∈N n∈N i∈N n∈N
P
Finally, by hypothesis, n∈N Pi (n) ≤ 1, for any i ∈ N. Therefore,
( P)(n) = sup
P F P
F n∈N n∈N Pi (n) i∈N ≤ 1. Which proves (item 3) that
P ∈ P. 
3.3. PPD sequence of a computation. The computed output of a QTM
will be defined (Definition 22) as the limit of the sequence of PPDs obtained
along its computations.
Definition 18 (probability and q-configurations). Given a configuration
C = hα, q, β, ni, let val[C] be the number of 1 in αβ. For any |φi ∈ `2 (C),
let us define P|φi : N → R≥0 s.t.
X
P|φi (n) = |eC |2
C∈Cf in ,val[C]=n
P
when |φi = C∈C eC |Ci.
Proposition 19. If |φi is a q-configuration, P|φi is a PPD. Moreover, it is
a PD iff |φi is final.
Proof. Let |φi = |φ0 i + |φf i with |φ0 i ∈ `2 (CM \ CfMin ), and |φf i ∈ `2 (CfMin ).
By the definitionPof P|φi , it is readily seen that P|φi = P|φf i and that
P
n∈N P|φi (n) = n∈N P|φf i (n) = k|φf ik ≤ 1. Therefore, P|φi is a PPD.
Moreover, since k|φf ik = 1 iff k|φ0 ik = 0. We see that P|φi is a PD iff
|φ0 i = 0, that is, iff |φi = |φf i. 
Definition 20. For any q-configuration |φi, we shall say that P|φi is the
PPD associated to |φi, and we shall denote by PK M the sequence of PPDs
|φi
M = {|φ i}
{P|φi i }i∈N associated to the computation K|φi i i∈N .

The PPD sequence of any QTM computation is monotone. In the simple

proof of this key property (Theorem 21) we see at work all the constraints
on the time evolution of a QTM M .
(1) That when in a final (target) configuration, the machine can only
increment the counter; as a consequence, the val of final (target)
configurations does not change.
(2) That when entering for the first time into a final (target) state, the
value of the counter is initialised to 0.
(3) That when in a final (target) configuration |φ, ni, the counter gives
the number of steps n since M is looping into the plain configuration
φ.
We stress that the last two properties defuse quantum interference between
final configurations reached in a different number of steps.
18 S. GUERRINI, S. MARTINI, AND A. MASINI

M
Theorem 21 (monotonicity of computations). For any computation K|φi
of a QTM M , the sequence of PPDs PK M is monotone.
|φi

Proof. Let us prove that P|φi ≤ PU |φi , for every |φi ∈ `2 (CM ). Let |φi =
|φ0 i + k∈N |φf,k i, with |φ0 i ∈ `2 (CM \ CfMin ), and |φf,k i ∈ `2 (CfMin ∩ CkM ),
P

where CkM is the set of the configurations of M whose P counter is equal to

k ∈ N. For every n ∈ N, we see that P|φi (n) = k∈N P|φf,k i (n). Let
P
|ψi = |ψ0 i+ k∈N |ψf,k i = U |φi. By the definition of U , wePsee that U |φ0 i =
|ψ0 i + |ψf,0 i and U |φf,k i = |ψf,k+1 i. Therefore, P|ψi (n) = k∈N P|ψf,k i (n) =
P
P|ψf,0 i (n) + k∈N P|φf,k i (n) = P|ψf,0 i (n) + P|φi (n) ≥ P|φi (n).

3.4. Computed output. We can now come back to the definition of the
computed output of a QTM computation. The easy case is when a computa-
tion reaches a final q-configuration |ψi ∈ Cf in (meaning that all the classical
computations in superposition are “terminated”)—in this case the computed
output is the PD P|ψi . The QTM keeps computing and transforming |ψi
into other configurations, but all these configurations have the same PD.
However, we want to give meaning also to “infinite” computations, which
never reach a final q-configuration, yet producing some final configurations
in the superpositions. For this purpose, we define the computed output as
the limit of the PPDs yielded by the computation.
Definition 22 (computed output of a QTM). Let K|φi M = {|φ i}
i i∈N be the
computation of the QTM M on the initial q-configuration |φi. The computed
output of M on the initial q-configuration |φi is the PPD P = limi→∞ P|φi i ,
M , or by the notation M
which we shall also denote by lim K|φi |φi → P.

Let us remark that, by Proposition 17 and Theorem 21, the limit in the
above definition is well-defined for any computation. Therefore, a QTM has
a computed output for any initial q-configuration |φi.
Definition 23 (finitary computations). Given a QTM M , a q-configuration
|φi is finite if it is an element of span(CB(CM )). A computation K|φi M =

{|φi i}i∈N is finitary with computed output P if there exists a k s.t. |φk i is
final and P|φk i = P.
Proposition 24. Let K|φi M = {|φ i}
i i∈N be a finitary computation with com-
puted output P, that is, M|φi → P.
(1) There exists a k, such that for each j ≥ k, |φj i is final and P|φj i = P.
(2) P is a PD.
Proof. By definition, there is a k s.t. |φk i is final. Let P|φk i = P. By
Proposition 19, P is a PD. By monotonicity, P ≤ P|φj i , for every j ≥ k.
Thus, since any PD is maximal for ≤ (see Remark 15), P = P|φj i , for every
j ≥ k. 
 QUANTUM TURING MACHINES: COMPUTATIONS AND MEASUREMENTS 19

M , we can then distinguish the following cases:

Given a computation K|φi
M is finitary. In this case M
(1) K|φi |φi → P ∈ P1 ; the output of the
computation is then a PD and is determined after a finite number
of steps;
M is not finitary, but M
(2) K|φi |φi → P ∈ P1 . The output is a PD and is
determined as a limit;
M is not finitary, and M
(3) K|φi |φi → P ∈ P \ P1 (the sum of the probabil-
ities of observing natural numbers is p < 1). Not only the result is
determined as a limit, but we cannot extract a PD from the output.
The first two cases above give rise to what Definition 25 calls a q-total
function. Observe, however, that for an external observer, cases (2) and (3)
are in general indistinguishable, since at any finite stage of the computation
we may observe only a finite part of the computed output.
For some examples of QTMs and their computed output, see Section 5.

3.5. Quantum partial computable functions. We want our quantum

computable functions to be defined over a natural extension of the natu-
ral numbers. Recall that, for any n ∈ N, n denotes the string 1n+1 and
that |ni = |hλ, q0 , n, 0ii. When using a QTM for computing a function, we
stipulate that initial q-configurations are superpositions of initial classical
configurations of the shape  |ni. Such q-configurations are naturally isomor-
phic to the space `21 = φ : N → C | n∈N |φ(n)|2 = 1 of square summa-
P
ble,
Pdenumerable
P sequences with unitary norm, under the bijective mapping
ν( dk nk ) = dk nk .
Definition 25 (partial quantum computable functions).
(1) A function f : `21 → P is partial quantum computable (q-computable)
if there exists a QTM M s.t. f (x) = P iff Mν(x) → P.
(2) A q-partial computable function f is quantum total (q-total) if for
each x, f (x) ∈ P1 .
QCF is the class of partial quantum computable functions.

4. Observables
While the evolution of a closed quantum system (e.g., a QTM) is re-
versible and deterministic once its evolution operator is known, a (global)
measurement of a q-configuration is an irreversible process, which causes the
collapse of the quantum state to a new state. Technically, a measurement
corresponds to a projection on a subspace of the Hilbert space of quantum
states. For the sake of simplicity, in the case of QTMs, let us restrict to mea-
surements observing if a configuration belongs to the subspace described by
some set of base configurations B. The effect of such a measurement is
summarised by the following:
20 S. GUERRINI, S. MARTINI, AND A. MASINI

Measurement postulate
Given a set of configurations B ⊆ C, a measurement
P observ-
ing if a quantum configuration |φi = C∈C eC |Ci belongs
to the subspace generated P by CB(B) gives a positive an-
swer with a probability p = C∈B |eC |2 , equal to the square
of the norm of the projection of |φi onto `2 (B), causing at
the same time a collapse of the configuration into the nor-
1
malised projection C∈B p− 2 eC |Ci; dually, with probability
P

1−p = C6∈B |eC |2 , it gives a negative answer and a collapse

P

onto the subspace `2 (C \ B) orthonormal to `2 (B), that is,

1
into the normalised configuration C6∈B (1 − p)− 2 eC |Ci.
P

Because of the irreversible modification produced by any measurement

on the current configuration, and therefore on the rest of the computation,
we must deal with the problem of how to read the result of a computation.
In other words, we need to establish some protocol to observe when a QTM
has eventually reached a final configuration, and to read the corresponding
result.

4.1. The approach of Bernstein and Vazirani. We already discussed

how B&V’s “sensible” QTMs are machines where all the computations in
superposition are in some sense terminating, and reach the final state at the
same time (are “stationary”, in their terminology). More precisely, Defini-
tion 3.11 of B&V reads: ”A final configuration of a QTM is any configura-
tion in [final] state. If when QTM M is run with input x, at time T the
superposition contains only final configurations, and at any time less than T
the superposition contains no final configuration, then M halts with running
time T on input x.”
This is a good definition for a theory of computational complexity (where
the problems are classical, and the inputs of QTMs are always classical) but
it is of little use for developing a theory of effective quantum functions.
Indeed, inputs of a B&V-QTM must be classical—we cannot extend by
linearity a B&V-QTM on inputs in `21 , since there is no guarantee whatsoever
that on different inputs the same QTM halts with the same running time.

4.2. The approach of Deutsch. Deutsch [10] assumes that QTMs are
enriched with a termination bit T . At the beginning of a computation, T is
set to 0, then, the machine sets this termination bit to 1 when it enters into
a final configuration. If we write |T = ii for the function that returns 1 when
the termination bit is set to i, and 0 otherwise, a generic q-configuration of
a Deutsch’s QTM can be written as
X X
|φi2 = |T = 0i ⊗ eC |Ci + |T = 1i ⊗ dD |Di
C6∈Cf in D∈Cf in
 QUANTUM TURING MACHINES: COMPUTATIONS AND MEASUREMENTS 21

The observer periodically measures T in a non destructive way (that is,

without modifying the rest of the state of the machine).
P of T gives2 the value 0, |φi collapses
(1) If the result of the measurement
(with a probability equal to C6∈Cf in |eC | ) to the q-configuration
P
0
|T = 0i ⊗ C6∈Cf in eC |Ci
ψ = qP
2
C6∈Cf in |eC |

and the computation continues with |ψ 0 i.

(2) If the result of thePmeasurement of T gives the value 1, |φi collapses
(with probability D∈Cf in |dD |2 ) to
P
00 |T = 1i ⊗ D∈Cf in dD |Di
ψ = qP
2
D∈Cf in |dD |

and, immediately after the collapse,

P the observer makes a further
D∈Cf in dD |Di
measurement of the component pP 2
in order to read-
D∈Cf in |dD |
back a final configuration.
Note that Deutsch’s protocol (in an irreversible way) spoils at each step
the superposition of configurations. The main point of Deutsch’s approach is
that a measurement must be performed immediately after some computation
enters into a final state. In fact, since at the following step the evolution
might lead the machine to exit the final state modifying the content of
the tape, we would not be able to measure at all this output. In other
words, either the termination bit acts as a trigger that forces a measurement
each time it is set, or we perform a measurement after each step of the
computation.

4.3. Our approach. The measurement of the output computed by our

QTMs can be performed by following a variant of Deutsch’s approach. Be-
cause of the particular structure of the transition function of our QTMs, we
shall see that we do not need any additional termination bit, that a measure-
ment can be performed at any moment of the computation, and that indeed
we can perform several measurements at distinct points of the computation
without altering the result (in terms of the probabilistic distribution of the
observed output).
Given a q-configuration |φi = |φf i + |φnf i, where |φf i ∈ `2 (Cf in ) and
|φnf i ∈ `2 (C \ Cf in ), our output measurement tries to get an output value
from |φi by the following procedure:
(1) first of all, we observe the final states of |φi, forcing the q-configuration
to collapse either into the final q-configuration |φf i / k|φf ik, or into
the q-configuration |φnf i / k|φnf ik, which does not contain any final
configuration;
22 S. GUERRINI, S. MARTINI, AND A. MASINI

(2) then, if the q-configuration collapses into |φf i / k|φf ik, we observe
one of these configurations, say |Ci, which gives us the observed
output val[C] = n, forcing the q-configuration to collapse into the
final base q-configuration (ec /|ec |) |Ci;
(3) otherwise, we leave unchanged the q-configuration |φnf i / k|φnf ik ob-
tained after the first observation, and we say that we have observed
the special value ⊥.
Summing up, an output measurement of |φi may lead to observe an output
value n ∈ N associated to a collapse into a base final configuration |Ci ∈ |φi
s.t. val[φ] = n or to observe the special value ⊥ associated to a collapse into
a q-configuration which does not contain any final configuration.
Definition 26 (output observation). An output observation with collapsed
q-configuration |ψi and observed output x ∈ N P ∪ {⊥} is the result of an
output measurement of the q-configuration |φi = C∈C eC |Ci. Therefore, it
is a triple |φi ↓x |ψi s.t.
(1) either x = n ∈ N, and
eC
|ψi = |Ci with C ∈ Cf in and val[C] = n and ec 6= 0
|eC |
(2) or x = ⊥, and
|φnf i X
|ψi = where |φnf i = eC |Ci and |φnf i =
6 0
k|φnf ik
C6∈Cf in

The probability of an output observation is defined by

|eC |2
(
if x = n ∈ N
Pr{|φi ↓x |ψi} =
k|φnf ik2 if x = ⊥

Remark 27. Let e |Ci ↓x |φi, with C ∈ Cf in and val[C] = n. Since e |Ci is a
q-configuration, |e| = 1. By the definition of output observation, x = n and
|φi = (e/|e|)U |Ci = e |Di, with |Di = U |Ci ∈ CB(Cf in ) and val[D] = n.
Moreover, Pr{e |Ci ↓x |φi} = |e|2 = 1.
Remark 28. For every pair |φi ↓x1 |ψ1 i and |φi ↓x2 |ψ2 i of distinct output
observations, ψ1 and ψ2 are in the orthonormal subspaces generated by the
two disjoint sets B1 , B2 ⊆ C, where Bi = {C ∈ C | |Ci ∈ |ψi i}.
Definition 29 (observed run). Let M be a QTM and UM be its time evo-
lution operator. For any monotone increasing function τ : N → N (that is,
τ (i) < τ (j) for i < j):
(1) a τ -observed run of M on the initial q-configuration |φi is a sequence
{|φi i}i∈N s.t.:
(a) |φ0 i = |φi;
(b) UM |φh i ↓xi |φh+1 i, when h = τ (i) for some i ∈ N;
(c) |φh+1 i = UM |φh i otherwise.
 QUANTUM TURING MACHINES: COMPUTATIONS AND MEASUREMENTS 23

(2) A finite τ -observed run of length k is any finite prefix of length k + 1

of some τ -observed run. Notation: if R = {|φi i}i∈N , then R[k] =
{|φi i}i≤k .
Remark 30. We stress that, given a τ -observed run R = {|φi i}i∈N :
(1) either it never obtains a value n ∈ N as the result of an output
observation, and then it never reaches a final configuration;
(2) or it eventually obtains such a value collapsing the q-configuration
into a base final configuration e |Ci s.t. |e| = 1 and val[C] = n, and
from that point onward all the configurations of the run are base
final configurations e |Cj i = e U j |Ci s.t. val[Cj ] = n, and all the
following observed outputs are equal to n (see Remark 27).
Definition 31. Let R = {|φi i}i∈N be a τ -observed run.
(1) The sequence {xi }i∈N s.t. |φh i ↓xi |φh+1 i, with h = τ (i), is the output
sequence of the τ -observed run R.
(2) The observed output of R is the value x ∈ N ∪ {⊥} (notation: R ↓x )
defined by:
(a) x = n ∈ N, if xi = n for some i ∈ N;
(b) x = ⊥ otherwise.
(3) For any k, the output sequence of the finite τ -observed run R[τ (k)+1]
is the finite sequence {xi }i≤k and xk is its observed output.
Definition 32 (probability of a run). Let R = {|φi i}i∈N be a τ -observed
run.
(1) For k ∈ N, the probability of the finite τ -observed run R[k] is induc-
tively defined by
(a) Pr{R[0]} = 1; 
when k = τ (i)
Pr{R[k]} Pr{|φk i ↓xi |φk+1 i} for some i ∈ N


(b) Pr{R[k+1]} =


Pr{R[k]} otherwise
(2) Pr{R} = limk→∞ Pr{R[k]}.
We remark that Pr{R} is well-defined, since 1 ≥ Pr{R[i]} ≥ Pr{R[j]} > 0,
for every i ≤ j. Therefore,
1 ≥ Pr{R} = lim Pr{R[k]} = inf{Pr{R[k]}}k∈N ≥ 0.
k→∞

Remark 33. Let R = {φi }i∈N be a τ -observed run s.t. R ↓n , for some n ∈ N.
As observed in Remark 30, for some k, we have R[τ (k)] ↓n and R[τ (j)] ↓⊥ ,
for j < k; moreover, for i > τ (k), |φi i = e |Ci i with |e| = 1, Ci ∈ Cf in ,
and val[Ci ] = n. As a consequence, Pr{R[τ (k) + 1]} = Pr{R[i]}, for any
i > τ (k) (since, by Remark 27, Pr{ φτ (i) ↓n φτ (i)+1 } = 1) and Pr{R} =
limi→∞ R[i] = Pr{R[τ (k) + 1]}.
24 S. GUERRINI, S. MARTINI, AND A. MASINI

Definition 34 (observed computation). The τ -observed computation of a

QTM M on the initial q-configuration |φi, is the set KM
|φi,τ of the τ -observed
runs of M on |φi with the measure Pr : P(K|φi,τ ) → C defined by
M

X
Pr B = Pr{R}
R∈B

for every B ⊆ KM
|φi,τ .

By K[k]M|φi,τ we shall denote the set of the finite τ -observed runs of length
k of M on |φi, with the measure Pr on its subsets (see Definition 34).
It is immediate to observe that the set KM |φi,τ naturally defines an infinite
tree labelled with q-configurations, where each infinite path starting from
the root |φi corresponds to a τ -observed run in KM |φi,τ .

Lemma 35. Given R1 , R2 ∈ KM |φi,τ , with R1 = {φ1,i }i∈N 6= {φ2,i }i∈N = R2 ,

there is k ≥ 0 s.t.
(1) φ1,i = φ2,i for i ≤ τ (k), that is, R1 [τ (k)] = R2 [τ (k)];
(2) for i > τ (k), the q-configurations φ1,i 6= φ2,i are in two orthonormal
subspaces generated by two distinct subsets of C.
Proof. Let R1 [h] = R2 [h] be the longest common prefix of R1 and R2 . Since
they both starts with |φi, such a prefix is not empty; moreover, by the
definition of τ -observed run, it is readily seen that h = τ (k), for some k.
Let us now prove item 2. If we take Ba,j = {C ∈ C | |Ci ∈ |ψa,h+1+j i}, for
a = 1, 2 and j ∈ N, we need to prove that B1,j ∩ B2,j = ∅, for every j. By
construction, φ1,h+1 6= φ2,h+1 and, for a = 1, 2, φh ↓xa φa,h+1 , for some xa ,
with φh = φa,h . Moreover, B1,0 ∩ B2,0 = ∅ (see Remark 28), at least one of
the two observed values x1 , x2 is not ⊥, and one of the two q-configurations
|φ1,h+1 i , |φ2,h+1 i is a final base q-configuration. W.l.o.g., let us assume that
x1 ∈ N and let |φ1,h+1 i = e |C, ni, for some final plain configuration C and
some n ∈ N. Then, |φ1,h+1+j i = e U j |C, ni = e |C, n + ji (see Remark 30)
and B1,j = {|C, n + ji}, for j ∈ N. Thus, to prove the assertion, it suffices
to show that |C, n + ji 6∈ B2,j , for j ∈ N. Let us distinguish two cases:
(1) x2 ∈ N and |ψ2,h+1 i = e0 |D, mi, where D is a final plain config-
uration and m ∈ N. As already seen for |ψ1,h+1+j i, for j ∈ N, we
have |ψ2,h+1+j i = e0 U j |D, mi = e0 |D, m + ji and B2,j = {|D, m + ji}.
Therefore, |C, n + ji 6∈ B2,j , since |C, n + ji =
6 |D, m + ji, for |C, ni =
6
|D, mi by construction.
(2) |ψ2,h+1 i ∈ `2 (C \ Cf in ). Let |D, mi ∈ B2,j ∩ Cf in , where D is a plain
configuration and m ∈ N. By induction on j, it is readily seen that
m < j ≤ n + j. Thus, |D, mi = 6 |C, n + ji.

Lemma 36. Let K|φi M = {φ }
i i∈N be the computation of the QTM M on the
initial q-configuration |φi and KM|φi,τ be the τ -observed computation on the
 QUANTUM TURING MACHINES: COMPUTATIONS AND MEASUREMENTS 25

same initial configuration. For every k ∈ N, we have that

X p
|φk i = Pr{R} |ψR i
R∈K[k]M
|φi,τ

where |ψR i is the last q-configuration of the finite run R of length k.

Proof. By definition, φ = φ0 and R = {φ} with Pr{R} = 1, and ψR = φ is

the only run of length 0 in KM |φi,τ . Therefore, the assertion trivially holds
for k = 0.
Let us then prove the assertion by induction on k. By definition and the
induction hypothesis

X p
|φk+1 i = UM |φk i = Pr{R} UM |ψR i
R∈K[k]M
|φi,τ

We have two possibilities:

(1) k 6= τ (i), for any i. In this case, there is a bijection between the runs
of length k and those of length k +1, since each run R0 ∈ K[k +1]M |φi,τ
is obtained from a run R ∈ K[k]|φi,τ with last q-configuration |ψR i,
M

by appending to R the q-configuration |ψR0 i = UM |ψR i. Moreover,

since by definition, Pr{R0 } = Pr{R}, we can conclude that

X p X p
|φk+1 i = Pr{R} UM |ψR i = Pr{R0 } |ψR0 i
R∈K[k]M
|φi,τ
R0 ∈K[k+1]M
|φi,τ

(2) k = τ (i), for some i. In this case, every R ∈ K[k]M |φi,τ with last q-
0
configuration |ψR i generates a run R of length k +1 for every output
observation |ψR i ↓x |ψR0 i, where R0 is obtained by appending |ψR0 i
to R. Therefore, let R = {|ψi i}i≤k ∈ K[k]M|φi,τ and

BR = {{ψi }i≤k+1 | |ψk i ↓x |ψk+1 i}

By applying Definition 26, we easily check that

X p
UM |ψR i = Pr{|ψR i ↓x |ψR0 i} |ψR0 i
R0 ∈BR
26 S. GUERRINI, S. MARTINI, AND A. MASINI

Thus, by substitution, and Pr{R} Pr{|ψR i ↓x |ψR0 i} = Pr{R0 }

X p
|φk+1 i = Pr{R} UM |ψR i
R∈K[k]M
|φi,τ
X X p p
= Pr{R} Pr{|ψR i ↓x |ψR0 i} |ψR0 i
R∈K[k]M R0 ∈BR
|φi,τ
X p
= Pr{R0 } |ψR0 i
R0 ∈ BR
S
R∈K[k]M
|φi,τ
X p
= Pr{R0 } |ψR0 i
R0 ∈K[k+1]M
|φi,τ

BR = K[k + 1]M
S
since R∈K[k]M |φi,τ .
|φi,τ


We are finally in the position to prove that our observation protocol is
compatible with the probability distributions that we defined as computed
output of a QTM computation.
Theorem 37. Let K|φi M = {φ }
i i∈N be the computation of the QTM M on
the initial q-configuration |φi and KM |φi,τ be the τ -observed computation on
the same initial configuration. For every n ∈ N:
(1) P|φk i (n) = Pr{R ∈ K[k]M|φi,τ | R ↓n }, for k = τ (i) + 1 and i ∈ N;
(2) PK M (n) = Pr{R ∈ K|φi,τ | R ↓n }.
M
|φi

Proof. Let uspstart with the first item. By Lemma 36, we know that |φk i =
P
R∈K[k]M Pr{R} |ψR i, where |ψR i is the last q-configuration of R. Since
|φi,τ

k = τ (i) + 1, for some i, we also know that either ψR ∈ C \ Cf in or |ψR i =

uR |CR i with CR ∈ Cf in and |uR | = 1. Therefore,
2
X p
P|φk i (n) = Pr{R} uR |CR i
R∈B[k,n]

where
B[k, n] = {R ∈ K[k]M
|φi,τ | R ↓n }

= {R ∈ K[k]M
|φi,τ | |ψR i = uR |CR i with val[CR ] = n}

By Lemma 35, we know that for every R1 , R2 ∈ B[k, n], we have |CR1 i 6=
|CR2 i. Therefore
X X
P|φk i (n) = Pr{R}| uR |2 = Pr{R} = PrB[k, n]
R∈B[k,n] R∈B[k,n]

since |uR | = 1. Which concludes the proof of the first item of the assertion.
 QUANTUM TURING MACHINES: COMPUTATIONS AND MEASUREMENTS 27

In order to prove the second item, let B[ω, n] = {R ∈ KM

|φi,τ | R ↓n }. We
have that
X X
Pr{R ∈ KM |φi,τ | R ↓ n } = Pr{R} = lim Pr{R}
i→∞
R∈B[ω,n] R∈B[ω,n]
R[τ (i)+1]↓n
X
(∗)
= lim Pr{R[τ (i) + 1]}
i→∞
R∈B[ω,n]
R[τ (i)+1]↓n
X
(∗∗)
= lim Pr{R0 } = lim Pr B[τ (i) + 1, n]
i→∞ i→∞
R0 ∈B[τ (i)+1,n]

since: (∗) Pr{R} = Pr{R[τ (i)+1]}, when R[τ (i)+1] ↓n (see Remark 33); (∗∗)
there is a bijection between B[τ (i) + 1, n] and {R ∈ B[ω, n] | R[τ (i) + 1] ↓n }
(see Remark 30) mapping every R0 ∈ B[τ (i) + 1, n] with last q-configuration
u |Ci into R = {ψj }j∈N ∈ {R ∈ B[ω, n] | R[τ (i) + 1] ↓n } s.t R0 = R[τ (i) + 1]
j−τ (i)−1
and |ψj i = u UM |Ci for j > τ (i).
Therefore, by the (already proved) first item of the assertion
Pr{R ∈ KM lim Pr B[τ (i)+1, n] = lim P|φτ (i)+1 i (n) = PK M (n)
|φi,τ | R ↓n } = i→∞ i→∞ |φi

5. A comparison with Bernstein and Vazirani’s QTMs: part 2

In view of Theorem 13, we may say that our QTMs generalise B&V-
QTMs, since the computation of any B&V-QTM can be simulated by a
corresponding QTM with the same transition function (up to transitions
entering/leaving the initial/final state). The general framework, however, is
substantially modified, and the “same” machine behaves in different ways
in the two approaches. In this section, we give two simple examples of this.

-1 +1

a ↦ b, d, x
qs p q qt

Figure 1. Transitions of a QTM

5.1. QTM transition graphs. Let us represent a QTM M by means of

a transition graph (a directed graph) whose nodes are the states of M ,
and whose arrows give its transition function (see Figure 1). Namely, if
δ0 (p, a)(q, b, d) = x 6= 0, the graph of M contains an arrow from the node
of p to the node of q, labelled by the tuple (a 7→ b, d, x). Every non-target
28 S. GUERRINI, S. MARTINI, AND A. MASINI

node has at least an outgoing edge labelled by such a tuple, for any symbol
a of the tape alphabet.
In addition to the arrows of δ0 , to represent the looping transitions of
target nodes, the graph contains a self-loop labelled by +1 on any target
node qt : to denote the fact that the only transition of any target config-
uration is the one that increases the counter by 1, without modifying the
rest of the configuration. Dually, every source node qs has a self-loop la-
belled by −1. The self-loop of qs is not its only outgoing arrow, since the
corresponding counter decreasing transition applies to source configurations
with a counter greater than 0 only; indeed, when the counter of a source
configuration reaches the 0, the δ0 transition function applies. On the other
hand, the self-loop is the only incoming arrow of any source state qs , since
no transition from another state can enter into it. The situation is dual for
target states, for which the self-loop is the only outgoing arrow.
See Figure 4 for some examples of source and target nodes of a transition
graph: the state s and q0 (which is also initial) are source states; the states
p and qf (which is also final) are target states.
5.2. A classical reversible TM with quantum behaviour. In Figure 2,
we give the transition graph of a B&V-QTM corresponding to a reversible
TM: any reversible TM M can be transformed into a B&V-QTM by assuming
that the weight of any transition of M is 1, and by adding a back-transition
from the final state to the initial state.

☐↦1,R,1

q0 1↦☐,R,1 1↦☐,R,1 qf

☐↦1,R,1

a↦a,R,1

Figure 2. A reversible TM as a QTM là Bernstein and Vazirani

In the example in Figure 2, q0 is the initial state and qf is the final one,
and they are connected by an arrow (a 7→ a, R, 1) from qf to q0 , where
a ∈ {2, 1}. When started on an initial tape containing n + 1, the machine
M erases two symbols 1 from the tape, leaving then n symbols 1 on it (we
recall that n is a sequence of n + 1 symbols 1); while for n = 0, the machine
M loops indefinitely on the middle state q, after erasing the unique symbol
1 on the tape. Summing up, M computes the predecessor n − 1 (with
 QUANTUM TURING MACHINES: COMPUTATIONS AND MEASUREMENTS 29

probability 1), for n > 0, while it diverges (with probability 1) for n = 0.

But, if we feed M with a non classical input as |ψi = √12 |0i + √12 |2i, then
M fails to give an answer according to B&V’s framework, since it reaches a
q-configuration in which final and non-final base configurations superpose.
To transform M into a QTM according to our formalism (see Theo-
rem 13), it suffices to replace the arrow from qf to q0 by two self-loops:
one (labelled −1) on the initial state q0 , and one (labelled +1) on the final
state qf . We obtain then the QTM in Figure 3. By the definition of com-
puted output, we can see that M √1 |0i+ √1 |2i → {1 7→ 1/2; n 7→ 0, if n 6= 1};
2 2
namely, with probability 1/2 the QTM halts with computed output 1; while
with probability 1/2 it diverges.

☐↦1,R,1

q0 qf
1↦☐,R,1 1↦☐,R,1

q
-1 +1

☐↦1,R,1

Figure 3. A reversible TM as a QTM

5.3. A PD obtained as a limit. The example in Figure 4 shows a QTM

which produces a PD only as an infinite limit. The tape alphabet of the
machine M is Σ = {$, 1, 2}, the set of its source states is Qs = {q0 , s}, the
set of its target states is Qt = {qf , p}, the state q0 is initial, the state qf is
final. In the figure, the symbol a stands for any symbol of the alphabet but
$, that is, a ∈ {1, 2}.
The machine M applies properly on initial configurations hλ, q0 , $n, 0i
(i.e., we assume that the sequence coding the number n is preceded by a $,
and that this is the only $ on the tape), whose corresponding base vector
will be denoted by |$ni. On such inputs, after moving from the initial state
q0 to q1 , M either ends up in the final state qf with probability 1/2, or it
loops on q1 with probability 1/2. We remark the source state s and the
target state p: such states do not play any role when the machine computes
on $n, since none of them can be reached. Nevertheless, they must be added
to deal with error situations, and to satisfy the local unitary conditions, in
order to get a proper unitary time evolution for M .
A simple calculation shows that M|$ni → P, with P = {n + 1 7→ 1; m 7→
0, if m 6= n + 1}; namely, on the input $n, M computes with probability
30 S. GUERRINI, S. MARTINI, AND A. MASINI

-1 +1
a↦a,R,1/√2

$↦$,R,1 a↦a,R,1/√2
q0 q1 qf

a↦a,R,1/√2
$↦$,R,1
a↦a,R,1
$↦$,R,1
a↦a,R, -1/√2

p s

-1 +1

Figure 4. A QTM computing the successor function (the

identity w.r.t. the tape) as a limit

1 the successor n + 1. (The machine is indeed the identity w.r.t. the tape,
since no transition of the machine changes any symbol on the tape; however,
since n is a sequence of n + 1 symbols 1, the computation leaves all these
symbols 1 on the tape, leading then to a computed output of n + 1.) We
stress that the PD P is obtained as a limit, since for every j ∈ N, PU j |$ni is
a PPD s.t. PU j |$ni (n + 1) < 1 and PU j |$ni (m) = 0, for m 6= n + 1. Of course,
this does not mean that we have to wait an infinite time to readback the
result! A correct way to interpret this fact is that for each n ∈ N, and each
 ∈ (0, 1/2], there exists k ∈ N s.t, for every j > k, 1− < PU j |$ni (n+1) ≤ 1.

6. Related works
In addition to the already mentioned proposals by B&V and Deutsch that
we have already extensively analysed, we discuss here some other papers
related to our work.
6.1. On quantum extensions of Turing Machines, and of related
complete formalisms.
• Strictly following the B&V approach, Nishimura and Ozawa [20, 21]
study the relationship between QTMs and quantum circuits (ex-
tending previous results by Yao [30]). They show that there exists
a perfect computational correspondence between QTMs and uni-
form, finitely generated families of quantum circuits. Such a corre-
spondence preserves the quantum complexity classes EQP, BQP and
ZQP.
 QUANTUM TURING MACHINES: COMPUTATIONS AND MEASUREMENTS 31

• Perdrix [24] proposes a new way to deal with quantum extensions

of Turing Machines. The basic idea is reminiscent of the quantum-
data/classical-control paradigm coined by Selinger [26, 27]. In fact,
in Perdrix QTM’s, the only quantum component is the tape whereas
the control is completely classical.
• Dal Lago, Masini, and Zorzi [5, 6] extend the quantum-data/classical-
control paradigm to a type free quantum λ-calculus that is proven to
be in perfect correspondence with the QTMs of B&V. Following the
ideas of the so called Implicit Computational Complexity, the au-
thors propose an alternative way to deal with the quantum classes
EQP, BQP, and ZQP.

6.2. On the readout problem. The following papers address the problem
of how to readout the result of a quantum computation. Since this is a key
question in the definition of any quantum computing formalism, they deserve
some deeper attention.
We recall, however, that our main interest is in QTMs as devices com-
puting distributions of probability, and not functions over natural numbers.
• Myers [19] tries to show that it is not possible to define a truly quan-
tum general computer. The article highlights how the B&V approach
fails on truly quantum data. In fact, in such a case it is impossible
to guarantee the synchronous termination of all the computations
in superposition. Consequently, the use of a termination bit spoils
the quantum superposition of the computation. This defect was well
known, and it is for this reason that B&V did not define a general
notion of quantum computability, but rather a notion sufficient to
solve—in a quantum way— only classical decision problems. Myers’s
criticism does not apply to our approach. Our QTMs are fully quan-
tum, and they have an observational protocol of the result that does
not depend on the synchronous termination of the computations in
superposition.
• In an unpublished note, Kieu and Danos [14] claim that: “For halt-
ing, it is desirable of the dynamics to be able to store the output,
which is finite in terms of qubit resources, invariantly (that is, un-
changed under the unitary evolution) after some finite time when the
desirable output has been computed.” Unfortunately, it is not possi-
ble to enter into a truly invariant final quantum configuration—only
a machine starting in a final configuration and computing the iden-
tity can accomplish this constraint. We overcome the problem by
introducing a feasible (i.e., correct from a quantum point of view)
notion of invariant, w.r.t. the readout, of final configurations. In this
way, even if the final configuration changes, the output we read from
that configuration does not change.
• In another unpublished note, Linden and Popescu [16] address the
problem of how to readout the result of a general quantum computer.
32 S. GUERRINI, S. MARTINI, AND A. MASINI

The authors write: “We explicitly demonstrate the difficulties that

arise in a quantum computer when different branches of the compu-
tation halt at different, unknown, times”, implicitly referring to the
problems in extending the approach of B&V to general quantum in-
puts (see again [19], discussed above). In the first part of the work,
the authors show that the problem cannot be solved by means of the
so-called “ancilla”. The ancilla is an additional information added
to the main information encoded by a configuration of the quantum
machine. The idea is that, once a final state is reached, the machine
keeps modifying the ancilla only. The authors show that the ancilla
approach destroys the quantum capabilities of quantum machines,
since only classical computations can survive to this treatment of
ancilla. Even if reminiscent of the ancilla, our approach is techni-
cally different—the problems addressed by Linden and Popescu do
not apply—since we carefully tailor the space of the possible con-
figurations of the machines, allowing the ancilla to play a role only
during its final and initial evolution (see also the discussion on the
ancilla in the introduction, section 1, p. 2).
In the second part of the work, the authors launch a strong attack
against the use of termination bit, the solution originally proposed
by Deutsch and successively refined by Ozawa [22]. The authors
try to argue that the approach proposed by Deutsch/Ozawa cannot
work. In fact, they show that even if it is true that once the ter-
mination bit is set to 1 it remains firmly with such a value forever,
any terminal configuration cannot be frozen, and keeps evolving ac-
cording to the Hamiltonian of the system. Once again, our proposal
does not have the defect depicted in the paper, because, far away
to force a final configuration to remain stable, only the readout of a
final configuration is stable in our approach.
• Hines [13] shows how to ensure simultaneous coherent halting, pro-
vided that termination is guaranteed for a restricted class of quantum
algorithms. This kind of approach based on coherent halting is inten-
tionally not followed in our paper. Indeed, as previously remarked,
we are interested in treating systems which includes, as a particu-
lar case, all classical computable functions—we cannot restrict to
terminating computations.
• Miyadera and Ohya [18] discuss the notion of probabilistic halting.
In particular, they write “... the notion of halting is still probabilis-
tic. That is, a QTM with an input sometimes halts and sometimes
does not halt. If one can not get rid of the possibility of such a prob-
abilistic halting, one can not tell anything certain for one experiment
since one can not say whether an event of halting or non-halting oc-
curred with probability one or just by accident, say with probability
1040 .” Therefore, they wonder about the existence of any algorithm
 QUANTUM TURING MACHINES: COMPUTATIONS AND MEASUREMENTS 33

to decide whether or not a QTM probabilistic halts. With no sur-

prise, they conclude that such an algorithm cannot exist. In fact,
since the non-probabilistic halting of a QTM corresponds to the
simultaneous halting of all the superposing branchings of its compu-
tation, an algorithm deciding the probabilistic halting would decide
the simultaneous termination of two classical reversible machines (by
combining them into a unique QTM), which is clearly an undecidable
problem. In a sense, the question of probabilistic or non-probabilistic
halting is irrelevant for our approach. In our QTMs, the result of
a computation is defined as a limit, and any computation converges
to some result. Accordingly, the only possible readouts that we can
get are approximations of such a result. At the same time, since we
show that a repeated-measures protocol can retrieve the distribution
associated to the output of a computation, we can accept to say that
our approach is probabilistic.

7. Conclusions and further work

We find surprising that in the thirty years since [10] a theory of quantum
computable functions did not develop, and that the main interest remained
in QTMs as computing devices for classical problems/functions. This in
sharp contrast with the original (Feynman’s and Deutsch’s) aim to have a
better computing simulation of the physical world.
As always in these foundational studies, we had to go back to the ba-
sics, and look for a notion of QTM general enough to encompass previous
approaches (for instance, simulation of B&V-QTMs, Theorem 13), and still
sufficiently constrained to allow for a neat mathematical framework (for in-
stance, monotonicity of quantum computations, Theorem 21, a consequence
of the particular way final states are treated in order to defuse quantum
interference once such states are entered). While several details of the pro-
posed approach may well change during further study, we are particularly
happy to have a recursive enumerable class of QTMs. This may allow a
fresh look to the problem of a quantum universal machine, and, therefore,
to obtain some of the “standard” theorems of classical computability theory
(s-m-n, normal form, recursion, etc.). These themes, as well as those related
to the various degrees of partiality of quantum computable functions will be
the subject of forthcoming papers.

References
[1] T. Altenkirch, J. Grattage, J. K. Vizzotto, and A. Sabry. An algebra of pure quantum
programming. In 3rd International Workshop on Quantum Programming Languages,
volume 170 of Electronic Notes in Theoretical Computer Science, pages 23 – 47, 2007.
[2] C. H. Bennett. Logical reversibility of computation. IBM J. Res. Develop., 17:525–
532, 1973.
[3] E. Bernstein and U. Vazirani. Quantum complexity theory. SIAM J. Comput.,
26(5):1411–1473, 1997.
34 S. GUERRINI, S. MARTINI, AND A. MASINI

[4] J. B. Conway. A course in functional analysis, volume 96 of Graduate Texts in

Mathematics. Springer-Verlag, New York, second edition, 1990.
[5] U. Dal Lago, A. Masini, and M. Zorzi. On a measurement-free quantum lambda
calculus with classical control. Mathematical Structures in Computer Science
(doi:10.1017/S096012950800741X), 19(2):297–335, April 2009.
[6] U. Dal Lago, A. Masini, and M. Zorzi. Quantum implicit computational complexity.
Theoret. Comput. Sci., 411(2):377–409, 2010.
[7] U. Dal Lago, A. Masini, and M. Zorzi. Confluence results for a quantum lambda
calculus with measurements. Electronic Notes in Theoretical Computer Science,
270(2):251–261, 2 2011.
[8] V. Danos, E. Kashefi, and P. Panangaden. The measurement calculus. J. ACM,
54(2):Art. 8, 45 pp. (electronic), 2007.
[9] M. Davis. Computability and unsolvability. McGraw-Hill Series in Information Pro-
cessing and Computers. McGraw-Hill Book Co., Inc., New York, 1958.
[10] D. Deutsch. Quantum theory, the Church-Turing principle and the universal quantum
computer. Proceedings of the Royal Society of London Ser. A, A400:97–117, 1985.
[11] R. P. Feynman. Simulating physics with computers. Internat. J. Theoret. Phys., 21(6-
7):467–488, 1981/82. Physics of computation, Part II (Dedham, Mass., 1981).
[12] S. Guerrini, S. Martini, and A. Masini. Towards A theory of quantum computability.
CoRR, abs/1504.02817, 2015.
[13] P. Hines. Quantum circuit oracles for abstract machine computations. Theor.
Comput. Sci., 411(11-13):1501–1520, 2010.
[14] T. D. Kieu and M. Danos. The Halting problem for universal quantum computers.
1998.
[15] U. D. Lago and M. Zorzi. Wave-style token machines and quantum lambda calculi.
CoRR, abs/1502.04774, 2015.
[16] N. Linden and S. Popescu. The Halting Problem for Quantum Computers. Technical
Report quant-ph/9806054, Jun 1998.
[17] A. Masini, L. Viganò, and M. Zorzi. Modal deduction systems for quantum state
transformations. J. Mult.-Valued Logic Soft Comput., 17(5-6):475–519, 2011.
[18] T. Miyadera and M. Ohya. On halting process of quantum turing machine. Open
Systems & Information Dynamics, 12(3):261–264, 2005.
[19] J. M. Myers. Can a universal quantum computer be fully quantum? Phys. Rev. Lett.,
78(9):1823–1824, 1997.
[20] H. Nishimura and M. Ozawa. Computational complexity of uniform quantum circuit
families and quantum turing machines. Theor. Comput. Sci., 276(1-2):147–181, 2002.
[21] H. Nishimura and M. Ozawa. Perfect computational equivalence between quantum
Turing machines and finitely generated uniform quantum circuit families. Quantum
Inf. Process., 8(1):13–24, 2009.
[22] M. Ozawa. Unconventional Models of Computation: Third International Conference,
UMC 2002 Kobe, Japan, October 15–19, 2002 Proceedings, chapter Halting of Quan-
tum Turing Machines, pages 58–65. Springer Berlin Heidelberg, Berlin, Heidelberg,
2002.
[23] M. Ozawa and H. Nishimura. Local transition functions of quantum Turing machines.
Theor. Inform. Appl., 34(5):379–402, 2000.
[24] S. Perdrix and P. Jorrand. Classically controlled quantum computation. Math.
Structures Comput. Sci., 16(4):601–620, 2006.
[25] S. Roman. Advanced linear algebra, volume 135 of Graduate Texts in Mathematics.
Springer, New York, third edition, 2008.
[26] P. Selinger. Towards a quantum programming language. Mathematical Structures in
Computer Science, 14(4):527–586, 2004.
[27] P. Selinger and B. Valiron. A lambda calculus for quantum computation with classical
control. Math. Structures Comput. Sci., 16(3):527–552, 2006.
 QUANTUM TURING MACHINES: COMPUTATIONS AND MEASUREMENTS 35

[28] P. W. Shor. Algorithms for quantum computation: discrete logarithms and factoring.
In 35th Annual Symposium on Foundations of Computer Science (Santa Fe, NM,
1994), pages 124–134. IEEE Comput. Soc. Press, Los Alamitos, CA, 1994.
[29] P. W. Shor. Polynomial-time algorithms for prime factorization and discrete loga-
rithms on a quantum computer. SIAM Rev., 41(2):303–332 (electronic), 1999.
[30] A. Yao. Quantum circuit complexity. In Proceedings of the 34th Annual Symposium
on Foundations of Computer Science, pages 352–360, Los Alamitos, California, 1993.
IEEE Press.
[31] M. Zorzi. On quantum lambda calculi: a foundational perspective. Mathematical
Structures in Computer Science, pages 1–89, 2 2015.
36 S. GUERRINI, S. MARTINI, AND A. MASINI

Appendix A. Hilbert spaces with denumerable basis

Definition 38 (Hilbert space of configurations). Given a denumerable set
B, with `2 (B) we shall denote the infinite dimensional Hilbert space defined
as follow.
The set of vectors in `2 (B) is the set
( )
X
φ | φ : B → C, 2
|φ(C)| < ∞
C∈B
and equipped with:
(1) An inner sum + : `2 (B) × `2 (B) → `2 (B)
defined by (φ + ψ)(C) = φ(C) + ψ(C);
(2) A multiplication by a scalar · : C × `2 (B) → `2 (B)
defined by (a · φ)(C) = a · (φ(C));
(3) An inner product5 h·, 2
P·i : ` (B) × `2 (B) → C
∗
defined by hφ, ψi = C∈B φ(C) ψ(C);
p
(4) The Euclidian norm is defined as kφk = hφ, φi.
The Hilbert space `2 = `2 (N) is the standard Hilbert space of denu-
merable dimension—all the Hilbert spaces with denumerable dimension are
isomorphic to it. `21 is the set of the vectors of `2 with unitary norm.
Definition 39 (computational basis). The set of functions
CB(B) = {|Ci : C ∈ B, |Ci : B → C}
such that for each C

1 if C = D
|Ci (D) =
0 if C 6= D
is called computational basis of `2 (B).
We can prove that [25]:
Theorem 40. The set CB(B) is an Hilbert basis of `2 (B).
Let us note that the inner product space span(CB(B)) defined by:
( n )
X
span(CB(B)) = ci Si | ci ∈ C, Si ∈ CB(B), n ∈ N .
i=1

is a proper inner product subspace of `2 (B), but it is not an Hilbert Space

(this means that CB(B) is not an Hamel basis of `2 (B)).
The completion of span(CB(B)) is a space isomorphic to `2 (B).
By means of a standard result in functional analysis we have:
Theorem 41.
5The condition P 2 P ∗
C∈B |φ(C)| < ∞ implies that C∈B φ(C) ψ(C) converges for every
pair of vectors.
 QUANTUM TURING MACHINES: COMPUTATIONS AND MEASUREMENTS 37

(1) span(CB(B)) is a dense subspace of `2 (B);

(2) `2 (B) is the (unique! up to isomorphism) completion of span(CB(B)).
Definition 42. Let V be a complex inner product space, a linear application
U : V → V is called an isometry if hU x, U yi = hx, yi, for each x, y ∈ V;
moreover if U is also surjective, then it is called unitary.
Since an isometry is injective, a unitary operator is invertible, and more-
over, its inverse is also unitary.
Definition 43. Let V be a complex inner product vectorial space, a linear
application L : V → V is called bounded if ∃c > 0 ∀x |Lx| ≤ c||x||.
Theorem 44. Let V be a complex inner product vectorial space, for each
bounded application U : V → V there is one and only one bounded application
U ∗ : V → V s.t. hx, U yi = hU ∗ x, yi. We say that U ∗ is the adjoint of U .
It is easy to show that if U is a bounded application, then U is unitary
iff U is invertible and U ∗ = U −1 .
Theorem 45. Each unitary operator U in span(CB(B)) has an unique ex-
tension in `2 (B) [3].
A.1. Dirac notation. We conclude this brief digest on Hilbert spaces, by
a synopsis of the so-called Dirac notation, extensively used in the paper.

mathematical notion Dirac notation

inner product hφ, ψi hφ | ψi
vector φ |φi
dual of vector φ hφ|
i.e., the linear application dφ
defined as dφ (ψ) = hφ, ψi note that hφ | ψi = hφ| (|ψi)

Let L be a linear application, with hφ| L |ψi we denote hφ | Lψi.

Appendix B. Implementation of the counter

A TM purist might argue that the counter adds to the machine a device
with a denumerable set of symbols, or with a denumerable set of states,
which is not in the spirit of the finite representability of TMs. However, it is
an easy exercice to implement the counter directly into the QTM, and in the
following we shall briefly describe two ways to do it. Nevertheless, we stress
that none of these implementations can be seen as natural or standard, and
that, on the other hand, one could completely ignore the problem, assuming
to add a clock to implement the counter. For these reasons, in the paper,
we have preferred to give the more abstract solution, instead of any more
concrete implementation.
38 S. GUERRINI, S. MARTINI, AND A. MASINI

B.1. Extra symbols. A first possibility, that we followed in a previous

version of the paper, is to duplicate the symbol alphabet Σ by adding a set
of extra tape symbols Σ = {a | a ∈ Σ}: a new symbol a, for any a ∈ Σ
(including the blank 2). In this way, when in a final state, the machine
replaces any symbol a with the corresponding extra symbol a, and moves to
the right. Dually, when in a source state, if the current cell contains an extra
symbol a, the machine replaces the current symbol with the corresponding
symbol a and moves to the right; otherwise, when the current symbol is
a ∈ Σ, it behaves as specified by the main transition function δ0 . The legal
configurations are then restricted to three possible cases:
(1) hα, q, βi
(2) hαγ, qt , βi
(3) hα, qs , γβi
∗
where: αβγ ∈ Σ∗ , γ ∈ Σ is the sequence of extra symbols obtained by
replacing any symbol a of γ with the corresponding extra symbol a; q is any
state; qt is a target state; qs is a source state. It is readily seen that, to
obtain an isomorphism between QTMs with extra symbols and QTMs with
counters, it suffices to take the following bijection of configurations (where
we use the same symbols as above):
(1) hα, q, βi 7→ hα, q, β, 0i
(2) hαγ, qt , βi 7→ hα, qt , γβ, |γ|i
(3) hα, qs , γβi 7→ hαγ, qs , β, |γ|i
where |γ| denotes the length of γ.

B.2. Additional counter tape. Another possibility is to add a second

tape to the machine. The alphabet of this counter tape contains only one
symbol ∗, in addition to the blank 2 corresponding to the empty cell. A
counter containing a value of n corresponds then to a tape with n symbols
∗, see Figure 5.

α β α β

qs qt

* * * * * *
⎧
⎨
⎩
⎧
⎨
⎩

n n

Figure 5. Implementation of the counter by means of a sec-

ond tape with only one non blank symbol ∗.

Adding/subtracting 1 to the counter corresponds to write/delete a ∗ sym-

bol. To implement these operations by a single step, it suffices that:
 QUANTUM TURING MACHINES: COMPUTATIONS AND MEASUREMENTS 39

(1) When the machine is in a source state qs , the counter head is on

the rightmost ∗ of the counter tape, if n > 0, or on an empty cell,
if n = 0. If the current counter symbol is a ∗, any transition in the
state qs replaces such a ∗ with a 2, and moves the counter head to
the left, until the current counter symbol becomes a 2, in which case
the machine starts its main evolution.
(2) When the machine is in a target state qt , the counter head is on the
first empty cell of the counter tape to the right of the sequence of
∗. Any transition in the state qt replaces then the 2 in the current
counter cell with a ∗, and moves the counter head to the right.
(3) When the state is neither a source nor a target state, the counter
tape is empty, and any transition leaves the counter tape unchanged.
Stefano Guerrini, LIPN, UMR 7030 CNRS, Institut Galilée, Université Paris13,
Sorbonne Paris Cité
E-mail address: [email protected]

Simone Martini, Dipartimento di Informatica – Scienza e Ingegneria, Uni-

versità di Bologna, and Inria Sophia-Antipolis
E-mail address: [email protected]

Andrea Masini, Dipartimento di Informatica, Università di Verona

E-mail address: [email protected]