arX

iv:q

uant

-ph/

9909

012v

3 1

Nov

200

2Analysis of Quantum Functions∗

Tomoyuki Yamakami†

School of Information and TechnologyUniversity of Ottawa, Ottawa, Canada K1N 6N5

Abstract: Quantum functions are functions that can be computed by means of quantum mechanical device.This paper investigates quantum functions that take classical bit inputs. We explore basic properties and closerelationships among these quantum functions. We also discuss quantum functions that make adaptive andnonadaptive (or parallel) oracle queries.

key words: quantum Turing machine, nonadaptive query, oracle separation

1 Overture

A paradigm of a quantum mechanical computer was first proposed in the 1980s [2, 9, 16] to exercise morecomputational power over the silicon-based computer, whose development is speculated to face a physicalbarrier. Since quantum mechanics is thought to govern Nature, a computer built upon quantum physics isof great importance. A series of discoveries of fast quantum algorithms in the 1990s [20, 33] have raisedenthusiasm among computer scientists as well as physicists. These discoveries have supplied general and usefultools in programming quantum algorithms.

A quantum computer is mathematically modeled in several different manners, including quantum Turingmachines [5, 9], quantum circuits [10, 40], and topological computations [19]. This paper uses a multiple tapemodel of quantum Turing machine (referred to as QTM) along the line of expositions [1, 5, 28, 30, 38, 39] dueto its close connection to a classical off-line Turing machine (TM, for short). A quantum computation of aQTM is a series of superpositions of machine’s configurations in which any computation path can exchangeinformation with other computation paths. This phenomena is known as quantum interference. A QTM canexploit quantum interference to achieve a large volume of parallel computations.

A study of function classes has been an important subject in classical complexity theory. Search problemsand optimization problems are clearly functions and are of special interests in many practical areas of computerscience. Since the 1960s, researchers have investigated various function classes, including FP, #P [35], NPSV[31], OptP [25], SpanP [26], and GapP [14]. The treatment of functions is, however, slightly different fromlanguages (or simply called sets) because of the size of output bits. The main theme of this paper is a studyof resource-bounded quantum functions. A quantum function in general refers to any function that can becomputed by means of a quantummechanical device. For instance, the Integer Factorization Problem (i.e., givena positive integer, find all factors of it) is solved by Shor’s polynomial-time quantum algorithm [33]. Of allquantum functions, this paper focuses only on those whose inputs are classical strings. A general treatment ofquantum functions with quantum-state inputs is also possible.

We consider two categories of quantum functions. A quantum computable function computes an output of aQTM with high probability. Such a function has been used in the literature without proper names. Let FEQPand FBQP denote the collections of functions computed by polynomial-time well-formed QTMs, respectively,with certainty and with probability at least 3/4. These function classes are viewed as quantum generalizations ofthe class of polynomial-time classically computable functions. Similarly, a partial single-valued QMA-functionis a quantum variant of a single-valued NP-function [31]. A quantum probability function, on the contrary,computes the acceptance probability of a QTM. For notational convenience, #QP (“sharp” QP) denotes thecollection of such quantum functions particularly witnessed by polynomial-time well-formed QTMs. Anotherimportant quantum function is the one that computes the gap between the acceptance and rejection probabilitiesof a QTM. We call such functions quantum gap functions and use the notation GapQP to denote the collectionof polynomial-time quantum gap functions. It is shown that GapQP is the subtraction closure of #QP.

∗A preliminary version appeared in the Proceedings of the 19th International Conference on Foundations of Software Technologyand Theoretical Computer Science, Lecture Notes in Computer Science, Springer-Verlag, Vol.1738, pp.407–419, 1999.

†Most of this work was done while the author was at the Department of Computer Science, Princeton University, betweenSeptember 1997 and June 1999 and was partly supported by NSERC Fellowship as well as DIMACS Fellowship.

1

There have been developed several proof techniques in quantum complexity theory during the last decade.These techniques are crucial to our analysis of quantum functions. A quantum counting technique of Brassard etal. [7], for instance, shows that any #QP-function can be closely approximated by a certain FBQP-function. Anidea of Fenner et al. [15] leads to a striking feature of quantum gap functions: if f ∈ GapQP then f2 ∈ #QP.Moreover, a series of results by Adleman et al. [1] and Yamakami and Yao [39] draws the close connectionbetween GapQP-functions and GapP-functions. If all amplitudes are restricted to algebraic numbers, the sign(i.e., positive, zero, or negative) of the value of a GapQP-function coincides with that of a certain GapP-function. This relationship further brings a new characterization of PP in terms of GapQP-functions. As animmediate consequence, the quantum analogue of PP, called PQP, with algebraic amplitudes matches PP.

To enhance a computation of a QTM, we allow the machine to access an oracle. An oracle computationdates back to Deutsch and Jozsa [11]. Generally, one oracle query depends on its previous oracle answers. Thispattern of oracle accesses are categorized as adaptive queries. On the contrary, nonadaptive queries‡ (or parallelqueries) refer to the case where an oracle QTM prepares a list of query words along each computation pathbefore making the first query in the entire computation. For a nonadaptive query case, we use the notationFEQPA

‖ to denote the collection of FEQPA-functions that make nonadaptive queries to oracle A.In a classical bounded query model, a function class and a language class behave differently; for instance,

FPNP‖ is believed to differ from FPNP[O(logn)] whereas PNP

‖ = PNP[O(logn)] [36]. On the contrary, quantuminterference makes it possible to draw such functions and languages close together in a use of the quantumalgorithm of Deutsch and Jozsa [11]. Moreover, we can exhibit an oracle that separates FEQPA

‖ from FPA and

also construct another oracle B that makes FPB harder than FEQPB‖ . These relativized results immediately

imply that EQPA‖ * PA and PA * EQPA

‖ . This exemplifies a peculiar nature of nonadaptive queries in aquantum setting.

The study of quantum functions finds useful applications to decision problems. We show a relationshipbetween the EQP =?PP question and the closure property of #QP under maximum and minimum operators.In the course of our study, we introduce one new quantum complexity class WQP (“wide” QP), which expandsUP.

Our investigation opens an door to a largely uncultivated area of quantum functions. We initiate a prelimi-nary study of the structure of such functions; however, we leave more questions to the behavior of QTMs. Weexpect that a vigorous study of quantum functions will bring us the answers to those questions.

2 Basic Notions and Notation

We briefly give notions and notation necessary to read through this paper.Denote by N and Z, respectively, the set of all natural numbers (that is, non-negative integers) and the set of

all integers. Set N+ = N− {0}. For each d ∈ N+, let Zd = {0, 1, . . . , d− 1} and Z[d] = {−d, . . . ,−1, 0, 1, . . . , d}.Moreover, let Q, R, and C be the sets of all rational numbers, real numbers, and complex numbers, respectively.In this paper, the notation A is used to denote the set of complex algebraic numbers.

The notation [a, b] denotes the real interval between a and b. Similarly, we use (a, b] and (a, b) as well. For afinite set S, |S| denotes the cardinality of S. For an infinite set S, we say that a property P(x) holds for almostall x in S if {x ∈ S | P(x) does not hold} is a finite set.

A finite set Γ = {γ1, γ2, . . . , γk} of complex numbers is linearly independent if∑k

i=1 aiγi 6= 0 for anynon-zero k-tuple (a1, a2, . . . , ak) ∈ Qk and Γ is algebraically independent if q(γ1, γ2, . . . , γk) 6= 0 for any q ∈Q[x1, x2, . . . , xk] that is not identically 0. For a subset A of C, Q(A) denotes the field generated by all elementsin A over Q. In this paper, a polynomial with k variables means an element in N[x1, x2, . . . , xk] and thus, allpolynomials are assumed to be nondecreasing.

To describe functions, we often use the λ-notation. The notation λx.f(x) means the function f itself. Forexample, λx.(2x + 3) denotes the function that outputs 2x+ 3 on input x. For a (partial) function f , dom(f)and ran(f) denote the domain and the range of f , respectively. We write f(x) ↑ to mean that f(x) is undefined(i.e., x 6∈ dom(f)). Otherwise, we write f(x) ↓.

Throughout this paper, we freely identify any natural number with its binary representation. When wediscuss integers, we also identify each integer with its binary representation following a sign bit that indicates

‡Our quantum nonadaptive query model seems different from a quantum analogue of a truth-table reduction, which is widelyused as a nonadaptive query model in a classical setting.

2

the (positive or negative) sign§ of the integer. A rational number is also identified as a pair of integers, whichare further identified as binary integers with sign bits.

For simplicity, we use a binary alphabet Σ = {0, 1} throughout this paper unless otherwise stated. For anystring x, the length of x, denoted |x|, is the number of symbols in x. For any number n ∈ N, Σn (Σ≤n, Σ≥n,resp.) represents the set of all strings of length n (≤ n, ≥ n, resp.). Let Σ∗ =

⋃

n∈NΣn. A complexity class is a

collection of subsets of Σ∗. For any subsets A and B of Σ∗, A denotes Σ∗ −A (the complement of A), A⊕B is{0x | x ∈ A} ∪ {1x | x ∈ B} (the disjoint union of A and B), and A△B is (A− B) ∪ (B − A) (the symmetricdifference of A and B). For any complexity class C, co-C denotes the class of all complements of sets in C.

Let NΣ∗be the set of all functions that map Σ∗ to N. Similarly, we define {0, 1}Σ∗

, NN, etc. We oftenidentify a set S with its characteristic function χS , which is defined as χS(x) = 1 if x ∈ S and χS(x) = 0otherwise.

A function f from Σ∗ to Σ∗ (N, resp.) is polynomially bounded if there exists a polynomial p such that|f(x)| ≤ p(|x|) (f(x) ≤ p(|x|), resp.) for all x in Σ∗. For any two functions f and g from Σ∗ to [0, 1] andany function ǫ from N to [0, 1], we say that f ǫ(n)-approximates g if |f(x) − g(x)| ≤ ǫ(|x|) for almost all x in

Σ∗. Let F and G be any two classes of functions from Σ∗ to [0, 1]. For any f , we write f∈∼ pF if, for every

polynomial p, there exists a function g ∈ F that 1/p(n)-approximates f . The notation F ⋐p G means that

f∈∼ pG for all functions f in F . Similarly, the notation F ⋐

e G is defined using “2−p(n)-approximation” insteadof “1/p(n)-approximation.”

As a mathematical model of classical computation, we use a multi-tape off-line TM with two-way infiniteread/write tapes. A reversible TM is a deterministic TM for which each configuration has at most one prede-cessor [3, 5]. A TM moves its heads to the right, to the left, and also allows them to stay still.

Let P (E, resp.) be the class of all sets recognized by polynomial-time (linear exponential-time, resp.)deterministic TMs. Moreover, NP (UP, resp.) is the class of all sets recognized by polynomial-time nondeter-ministic TMs (unambiguous TMs, resp.). The class BPP (PP, resp.) denotes the class of all sets recognizedby polynomial-time probabilistic TMs with bounded-error (unbounded-error, resp.) probability. For othercomplexity classes, the reader should refer to the textbook, e.g., [12].

The following lemma is useful in order to simulate classical computation by a QTM.

Lemma 2.1 [3, 5] Any deterministic TM M that, on any input x, outputs a string M(x) in an output tapecan be simulated by a certain reversible TM N with polynomial slowdown such that (i) on any input x, outputs(x,M(x)) in an output tape, (ii) all heads of N move back to the start cells, and (iii) the running time of Non input x depends only on the lengths of both input x and output M(x).

Let C denote the set of complex numbers whose real and imaginary parts can be approximated to within2−n in time polynomial in n. A dyadic rational number is the number of the form 0.y for a certain binary stringy, and D denotes the set of all dyadic rational numbers. Note that N ⊆ Z ⊆ D ⊆ Q ⊆ A ⊆ C ⊆ C.

A function f from Σ∗ to Σ∗ is in FP if its values are computed by polynomial-time deterministic TMs. Forany nondeterministic TM M and any string x, let #M(x) (#M(x), resp.) denote the number of accepting(rejecting, resp.) computation paths of M on input x. A function f from Σ∗ to N is in #P if there exists apolynomial-time nondeterministic TM M such that f(x) = #M(x) for every x ∈ Σ∗ [35]. By expanding #Pnaturally, we define #E based on 2O(n)-time nondeterministic TMs. A function f from Σ∗ to Z is in GapP ifthere exists a polynomial-time nondeterministic TM M such that f(x) = #M(x) − #M(x) for every x ∈ Σ∗

[14]. The class of single-valued NP-functions, NPSV, is the collection of partial functions f from Σ∗ to Σ∗ suchthat there exists a polynomial-time nondeterministic TM M with an output tape satisfying the following: forevery x, (i) if x ∈ dom(f) then M on input x has at least one accepting computation path and M writes f(x)in its output tape on all accepting computation paths and (ii) if x 6∈ dom(f) then all computation paths endwith rejecting configurations [31].

A function f is in FPA‖ if there exists a polynomial-time deterministic oracle TM M such that, on every

input x, M makes a list (called a query list) of strings that are all queried after M finishes the list. Let C bea relativizable class of functions or sets. A set A is called a low set for C if CA ⊆ C and let low-C denote the

collection of all low sets for C. If C induces its nonadaptive relativization C(·)‖ , low-C‖ denotes the class of sets

A satisfying CA‖ ⊆ C.

A pairing function 〈·, ·〉 is assumed to be one-one on Σ∗ and polynomial-time computable with polynomial-time computable inverses. For simplicity, we assume the extra condition: |〈x, y〉| = r(1|x|+|y|) for all pairs (x, y),

§For example, we set 1 for a positive integer and 0 for a negative integer.

3

where r is a certain fixed FP-function.

3 Quantum Turing Machines

The notion of a QTM was originally introduced in [9] and fully developed by a series of expositions in [5, 38,30, 28]. For convenience, we use in this paper a general definition of QTMs¶ defined in [38], where the QTMhas k two-way infinite tapes of cells indexed by Z and its read/write heads that move along the tapes either tothe left or to the right, or the heads stay still. This model greatly simplifies the programming of QTMs.

3.1 Definition of Multiple Tape Quantum Turing Machines

A pure quantum state is a unit-norm vector in a Hilbert space (that is, a complex vector space with the standardinner product 〈·|·〉). A quantum bit (qubit, for short) is a pure quantum state in a 2-dimensional Hilbert space.We often use the standard basis {|0〉, |1〉} to represent a qubit. A quantum string (qustring, for short) of sizen is a unit-norm pure quantum state in a Hilbert space of dimension 2n. A qubit is also a qustring of size 1.For each n ∈ N+, let Φn denote the collection of all qustrings of size n and set Φ∞ =

⋃

n∈N+ Φn. The norm of

a qustring |φ〉, denoted ‖|φ〉‖, is defined as√

〈φ|φ〉.There are several useful unitary transformations used in this paper. The phase shift P maps |0〉 to −|0〉 and

|1〉 to |1〉. The Walsh-Hadamard transformation H changes |0〉 into 1√2(|0〉+ |1〉) and |1〉 into 1√

2(|0〉− |1〉). The

quantum Fourier transformation QFTn maps |m〉 to 1√2n

∑2n−1ℓ=0 e

2πimℓ2n |ℓ〉, where we identify an integer between

0 and 2n − 1 with a binary string of length n (in the lexicographic order). The last transformation H2 acts on{|0〉, |1〉, |2〉, |3〉} exactly as H ⊗H that acts on {|0〉, |1〉}2 by identifying 0 = 00, 1 = 01, 2 = 10, and 3 = 11.

Formally, a k-tape QTM M is defined as a sextuple (Q, {q0}, Qf , Σk, Γk, δ), where Σk = Σ1 ×Σ2 × · · · ×Σk,

Γk = Γ1 × Γ2 × · · · × Γk, each Σi is a finite input alphabet, Γi is a finite tape alphabet including Σi as wellas a distinguished blank symbol #, Q is a finite set of (internal) states including an initial state q0, Qf is a

set of final states with Qf ⊆ Q, and δ is a multi-valued quantum transition function mapping from Q × Γk to

CQ×Γk×{L,N,R}k

. Each value δ(p,σ) is described as a linear combination∑

αq,σ ,τ ,d |q〉|τ 〉|d〉, where the sum is

taken over all q ∈ Q, σ, τ ∈ Σk, and d ∈ {L,N,R}k, and each complex number αq,σ ,τ ,d is called an amplitudeof M . This δ induces the time-evolution operator (or matrix), which is a unitary operator conducting a singleapplication of δ to the space spanned by all configurations of M . For any nonempty subset K of C, we saythat M has K-amplitudes if the entries of its time-evolution matrix are all drawn from K. This K is called anamplitude set of M . For convenience, we assume that the last tape of M is always used as an output tape.

The following terminology comes from [5, 38]. A QTM M is in normal form if, for every qf ∈ Qf , there

exists a vector of directions d ∈ {L,N,R}k such that δ(qf ,σ) = |q0〉|σ〉|d〉 for all σ ∈ Γk and M is stationaryif, when it halts, all heads halts at the start cell. A QTM M is called clean if it is stationary in normal formand, when it halts, all tapes except for the output tape become empty. Note that a QTM M may reach finalconfigurations many times, where a finial configuration is a configuration with a final state. Hence, we needto call M synchronous if, for every x, all computation paths of M on input x reach finial configurations at thesame time. A QTM M is well-formed if its time-evolution operator preserves the ℓ2-norm. An accepting (arejecting, resp.) configuration is a finial configuration in which the output tape constitutes only a symbol “1”(“0”, resp.) in the start cell.

The running time of a QTM M on input x is defined to be the minimal number t (if any) such that, at timet, every computation path of M on x reaches a certain final configuration. We say that M on input x haltsin time t (within time t, resp.) if its running time is defined and is exactly t (at most t, resp.). We call M apolynomial-time QTM if there exists a polynomial p such that, on every input x, M halts in time p(|x|). Thisdefinition of polynomial-time computation seems restrictive but it is easier to avoid the timing problem thatarises when we modify the QTM. It is shown in [38] that any well-formed QTM with a single final state can besimulated by a certain stationary synchronous well-formed QTM in normal form with quadratic slowdown.

For any string x, the notation ρM (x) denotes the acceptance probability ofM on x: that is, the sum of everysquared magnitude of the amplitude of any accepting configuration in the final superposition of M . For anystrings x and y, we say that a QTM M on input x outputs y with probability α if |y〉 is observed in the output

¶It is proven in [40, 38, 28] that our model is polynomially “equivalent” to the more restrictive model used in [5], which issometimes called conservative [39, 38].

4

tape of the QTM (where the leftmost symbol of y is in the start cell) with probability α after M halts on inputx.

For convenience, the notation M |φ〉 with a quantum state |φ〉 denotes the final superposition of M thatstarts with |φ〉 as a superposition of inputs. By linearity, M |φ〉 =

∑

s∈S αsM |s〉 if |φ〉 =∑

s∈S αs|s〉, where Sis a set of binary strings and each αs is a complex number.

An oracle QTM is equipped with an extra tape, called a query tape and two distinguished states, a pre-querystate qp and a post-query state qa. Let A be any subset of Σ∗ (called an oracle). The oracle QTM invokes anoracle query by entering state qp with |y〉|b〉 written in the query tape, where b ∈ {0, 1} and y ∈ Σ∗. The leftmostsymbol of y is not necessarily in the start cell. In a single step, the tape content is changed into |y〉|b⊕ χA(y)〉and the machine enters state qa, where ⊕ denotes the XOR. The notation MA is used for M with oracle A andρAM (x) denotes the acceptance probability of MA on input x.

3.2 Fundamental Lemmas

For those who are not familiar with multi-tape QTMs with flexible head moves, we list several lemmas‖, givenin [38], without their proofs. These lemmas will serve the later sections.

The well-formedness of a QTMM = (Q, {q0}, Qf , Σk, Γk, δ) can be characterized by three local conditions of

its transition function δ. Let D = {0,±1}, E = {0,±1,±2}, and H = {0,±1, ♮}. Let (p,σ, τ ) ∈ Q× (Σk)2 and

ǫ ∈ Ek. Let Dǫ= {d ∈ Dk | ∀i ∈ {1, . . . , k}(|2di− ǫi| ≤ 1)} and E

d= {ǫ ∈ Ek | ∀i ∈ {1, . . . , k}(|2di− ǫi| ≤ 1)},

where d = (di)1≤i≤k and ǫ = (ǫi)1≤i≤k. Let hd,ǫ = (hdi,ǫi)1≤i≤k. We define δ[p,σ, τ |ǫ] as follows: δ[p,σ, τ |ǫ] =∑

q∈Q

∑

d∈Dǫδ(p,σ, q, τ ,d)|E

d|−1/2|q〉|h

d,ǫ〉.

Lemma 3.1 (Well-Formedness Lemma) Let k be any positive integer. A k-tape QTM (Q, {q0}, Qf , Σk, Γk, δ)is well-formed iff the following three conditions hold.

1. ‖δ(p,σ)‖ = 1 for all (p,σ) ∈ Q× Γk.

2. δ(p1,σ1) · δ(p2,σ2) = 0 for any distinct pairs (p1,σ1), (p2,σ2) ∈ Q× Γk.

3. δ[p1,σ1, τ 1|ǫ]·δ[p2,σ2, τ 2|ǫ′] = 0 for any distinct pair ǫ, ǫ′ ∈ Ek and for any pair (p1,σ1, τ 1), (p2,σ2, τ 2) ∈Q× (Γk)

2.

Another useful lemma known as Completion Lemma states that any partially defined QTM can be expandedto a standard QTM. This allows us to describe the transition of the machine only for instances of specificinterests. We say that amplitude setK is good if it is closed under the following operations: addition, subtraction,multiplication, division, complex conjugation, and square root.

Lemma 3.2 (Completion Lemma) Let K be a good amplitude set. For any k-tape polynomial-time K-amplitude QTM with a partially-defined quantum transition function δ that satisfies the well-formedness con-ditions in Lemma 3.1, there exists a k-tape polynomial-time well-formed K-amplitude QTM M ′ with the samestate set and alphabet whose transition function δ′ agrees with δ whenever δ is defined.

Reversal Lemma asserts the existence of a QTM that reverses another QTM. Let M1 and M2 be two well-formed QTMs with the same tape alphabet. Assume that M1 has a single final state. For any input x on whichM1 halts, let cx and φx be the initial configuration and the final superposition of M1 on x, respectively. Wesay that M2 reverses the computation of M1 if, for any input x on which M1 halts, M2 starts φx as its initialsuperposition and halts in a final superposition consisting entirely of configuration cx [5]. The machine M2 iscalled a reversing machine ofM1. The notationK

∗ denotes the set of the complex conjugates γ∗ for all numbersγ ∈ K.

Lemma 3.3 (Reversal Lemma) Assume that K∗ ⊆ K. Let M be a polynomial-time well-formed K-amplitudeQTM with a single final state. There exists another polynomial-time well-formed K-amplitude QTM MR thatreverses the computation of M with only quadratic slowdown.

Reversal Lemma yields the following useful lemma.

Lemma 3.4 (Squaring Lemma) Assume that K∗ ⊆ K. Let M be a polynomial-time well-formed QTM with asingle final state which, on input x, outputs b(x) ∈ {0, 1} with probability ρ(x). There exists a polynomial-time

‖Similar results for the conservative QTMs are given in [5].

5

well-formed stationary normal-form QTM N with K-amplitudes such that N on input x reaches the configura-tion, with probability exactly ρ(x)2, in which N is in finial state with x in tape 1, b(x) in tape 2, and emptyelsewhere.

The following lemma shows that any well-formed oracle QTM can be modified to a certain canonical form.For any set A, let A′ = {y01m−|y|−2 | m ≥ |y|+ 2, y ∈ A}.

Lemma 3.5 (Canonical Form Lemma) Assume that K∗ ⊆ K. Let M be a polynomial-time well-formed QTMwith K-amplitudes. Let A be any oracle. There exists a polynomial-time well-formed K-amplitude QTM N suchthat, for every x, (i) ρA

′

N (x) = ρAM (x), (ii) the length of any query word is exactly the same on all computationpaths of N on input x, and (iii) N makes exactly the same number of queries along every computation path oninput x.

A Brief Discussion on QTMs. Firstly, the choice of amplitude set K is crucial for most applications ofQTMs. Thus, we heed a special attention to amplitudes of a QTM. Throughout this paper, K denotes anarbitrary subset of C that includes {0,±1} for convenience unless otherwise stated. All quantum functionclasses discussed in this paper rely on the choice of amplitude set K. It is debatable whether C is the mostnatural choice of an amplitude set for QTMs. However, we find it more convenient to drop script K whenK = C.

Secondly, we need to address the difference between a QTM model and a model of a P-uniform family ofquantum circuits. In many proofs, we often give quantum-circuit descriptions, when we define QTMs, insteadof QTM descriptions. However, as being pointed out in [28], these two models might not always define exactlythe same complexity classes, such as EQPK and ZQPK . Hence, we need to check if the actual implementationof a given quantum circuit on a QTM is possible.

4 Quantum Functions with Classical Inputs

Over the past few decades, the function classes FP, #P, and GapP have played a major role in classicalcomplexity theory. Many old complexity classes have been redefined in terms of these function classes. Forexample, any NP-set S is characterized by S = {x | f(x) > 0} for a certain #P-function f . Similarly, anyPP-set S is written as S = {x | g(x) > 0} for a certain GapP-function g. Fenner et al. [14] further studied theextended notion of gap-definable complexity classes. These function classes continue to fascinate complexitytheoreticians.

Quantum functions are those whose outputs are produced by quantum computation in certain ways. Thesefunctions naturally expand the classical framework of computation with the help of quantum interference andentanglement. We pay special interests in classifying these quantum functions and clarifying their roles inquantum complexity theory. In particular, we focus on quantum functions whose inputs are classical binarystrings.

This paper recognizes five basic types∗∗ of quantum functions: exact quantum functions, bounded-error quan-tum functions, single-valued quantum nondeterministic function, quantum probability functions, and quantumgap functions. The first three types of quantum functions are functional generalizations of complexity classesEQP and BQP [5] and QMA [23, 24], and the last two types are quantum analogues of #P [35] and GapP [14].

4.1 Quantum Computable Functions

This section defines three types of (partial) quantum functions from Σ∗ to Σ∗ whose outcomes are computedby QTMs with high probability.

Earlier, Bernstein and Vazirani [5] introduced two important complexity classes, EQP (exact QP) and BQP(bounded-error QP), which are the collections of languages recognized by polynomial-time well-formed QTMswith C-amplitudes with certainty and with bounded-error probability, respectively. These language classes arenaturally expanded into function classes.

We begin with the functional version of EQP—the class of all quantum functions whose values are obtained,with certainty, by the measurement of the output tapes of polynomial-time well-formed QTMs. We call thempolynomial-time exact quantum computable in a fashion similar to polynomial-time computable functions.

∗∗The first type FEQP was independently introduced in [8].

6

Definition 4.1 Let FEQPK be the set of polynomial-time exact quantum computable functions with K-amplitudes; that is, there exists a polynomial-time well-formed QTM with K-amplitudes such that, on everyinput x, M outputs f(x) with probability 1. In this case, we say that M computes f with certainty.

We introduce another important function class FBQPK induced from polynomial-time well-formed K-amplitude QTMs having bounded-error (away from 1/2) probability.

Definition 4.2 A function f is polynomial-time bounded-error quantum computable with K-amplitudes if thereexists a polynomial-timeK-amplitude well-formed QTMM that, on every input x, outputs f(x) with probabilityat least 3/4. In this case, we say that M computes f with bounded-error probability. Let FBQPK denote theset of all polynomial-time bounded-error quantum functions with K-amplitudes.

The success probability 3/4 for M in Definition 4.2 can be amplified to 1− 2−q(n) for any fixed polynomialq. This is done by recursively running M in a new blank are of work tapes and then outputting the majorityvalues.

A simple example is the Integer Factorization Problem. Since Shor’s quantum algorithm [33] solves thisproblem in polynomial time with bounded-error probability, it is in FBQP.

As noted in Section 2, we drop subscript K when K = C and write FEQP for FEQPK and FBQP forFBQPK . As shown in, e.g., [5], any deterministic computation can be simulated by reversible computationwith only polynomial slowdown. Thus, we have FP ⊆ FEQPK for any amplitude set K (⊇ {0,±1}).

Lemma 4.3 FP ⊆ FEQPK ⊆ FBQPK .

The classes FEQPK and FBQPK are the functional expansions of EQPK and BQPK in the following sense:a set is in EQPK (BQPK , resp.) iff its characteristic function is in FEQPK (FBQPK , resp.).

The complexity class BQPK is known to be robust with the choice of amplitude set K. It is shown in[1, 32, 22] that BQP

C= BQPQ = BQP{0,±1,± 1√

2}. This is easily translated into function class: FBQP

C=

FBQPQ = FBQP{0,±1,± 1√2}. Unlike BQPK , EQPK is sensitive to its underlying amplitude set K. For instance,

Adleman et al. [1] showed that EQPC = EQPA∩R while Nishimura [27] proved that EQP{0,±1,± 35 ,± 4

5} collapsesto P. These results imply that FEQPC = FEQPA FEQP{0,±1,± 3

5 ,± 45} = FP.

We further expand FBQP. Earlier, Knill [24] and Kitaev [23] studied a quantum analogue of NP, namedQMA (quantum Merlin-Arthur) in [37]. Similar to partial single-valued NP-functions, we can define partialsingle-valued QMAK-functions.

Definition 4.4 A partial function f from Σ∗ to Σ∗ is called a single-valued QMA function with K-amplitudesif there exist a polynomial p and a polynomial-time well-formed quantum TM M with K-amplitudes such that,for every string x, (i) if x ∈ dom(f) then M on input |x〉|φ〉 outputs |1〉|f(x)〉 with probability at least 3/4 for acertain qustring |φ〉 of size p(|x|) and, for every string y ∈ Σp(|x|) −{f(x)} and every qustring |ψ〉 of size p(|x|),M on input |x〉|ψ〉 outputs |1〉|y〉 with probability at most 1/4 and (ii) if x 6∈ dom(f) then, for every stringy ∈ Σp(|x|) and every qustring |ψ〉 of size p(|x|), M on input |x〉|ψ〉 outputs |1〉|y〉 with probability at most 1/4.Let QMASVK be the set of all single-valued QMA-functions with K-amplitudes. The first qubit |1〉 in |1〉|y〉indicates an accepting configuration.

Similar to FBQP, we can amplify the success probability of the QTM M in Definition 4.4 from 3/4 to1− 2−q(n) in the following fashion: given input x ∈ Σn together with qustring |φ〉 ∈ Φp(n) that is sectioned intom blocks of p(n) bits, where m = 6q(n) + 1, we run this QTM m times sequentially using a new block eachtime and output the majority value at the end. This procedure works because any entanglement of two or moreblocks does not reduce the success probability of each run.

To treat partial functions in accordance with the previously defined total functions, we view the functionclasses, such as FP and FEQP, as classes of partial functions. This convention enables us state the relationshipbetween FBQPK and QMASVK as FBQPK ⊆ QMASVK . Assume that K ⊇ {0,±1,± 1√

2}. Note that, for

any set A ∈ QMA, the following function f is in QMASV: f(x) = 1 if x ∈ A, and f(x) ↑ otherwise. Clearly,dom(f) ∈ QMA. Moreover, Af = {〈x, y〉 | y ≤ f(x) ↓} is in QMA. These facts show that NPSV ⊆ QMASVK

iff NP ⊆ QMAK . Since NP ⊆ QMAK is well known, NPSV ⊆ QMASVK . Overall, we obtain the followinglemma.

Lemma 4.5 NPSV ∪ FBQPK ⊆ QMASVK if K ⊇ {0,±1,± 1√2}.

7

A function class F is said to be closed under composition if, for every f, g ∈ F , f ◦ g is also in F , wheref ◦ g = λx.f(g(x)). We claim that FEQP, FBQP, and QMASV are closed under composition. The proof iseasy and left to the reader.

Lemma 4.6 FEQP, FBQP, and QMASV are all closed under composition.

4.2 Quantum Probability Functions and Quantum Gap Functions

In classical complexity theory, the number of accepting computation paths of a nondeterministic TM is a keyto many complexity classes, including UP, NP, C=P, and PP. Valiant [35] introduced the class of functionsthat output such numbers, and coined the name #P (“sharp” P or “number” P) for this class. The class #Phas become an important subject in connection to counting classes (see a survey [17]).

In a quantum computation, its acceptance probability plays a crucial role similar to the number of acceptingcomputation paths in a classical computation. To study the behavior of the acceptance probability of a well-formed QTM, we consider quantum functions that output such probabilities. We briefly call these functionsquantum probability functions. Similar to the roles of #P, such quantum functions can characterize many knownquantum complexity classes, such as EQP, BQP, and NQP††. By abusing the existing Valiant’s notation #P,we coin the new name #QPK (“sharp” QP) for the class of polynomial-time quantum probability functions.This function class greatly expands our scope of quantum functions.

Recall that ρM (x) denotes the acceptance probability of M on input x.

Definition 4.7 A function f from Σ∗ to [0, 1] is called a polynomial-time quantum probability function with K-amplitudes if there exists a polynomial-time, well-formed QTM M with K-amplitudes such that f(x) = ρM (x)for all x. In this case, we simply say thatM witnesses f . The notation #QPK denotes the set of all polynomial-time quantum probability functions with K-amplitudes.

Restricted to {0, 1}-valued functions, quantum probability functions coincide with exact computable func-tions because their output tape contains only one qubit. Recall that {0, 1}Σ∗

denotes the set of all functionsfrom Σ∗ to {0, 1}. Thus, FEQPK ∩ {0, 1}Σ∗

= #QPK ∩ {0, 1}Σ∗for any amplitude set K.

The lemma below shows that #QPK expands #P when K ⊇ {0,±1,± 12}.

Lemma 4.8 Let K = {0,±1,± 12}. For every #P-function f , there exist two functions ℓ ∈ FP ∩ NΣ∗

and

g ∈ #QPK such that f(x) = ℓ(1|x|)g(x) for every x.

Proof. In this proof, we use the alphabet Σ4 = {0, 1, 2, 3}. Let f be any function in #P. Take a polynomialp and a polynomial-time deterministic TM M such that

f(x) = |{y ∈ Σp(|x|)4 | M(〈x, y〉) = 1}| for all x. By Lemma 2.1, we can assume that M is reversible. Define

the new QTM N as follows.

On input x, write |0p(n)〉 in a separate blank tape and apply H⊗p(n)2 . Observe |y〉 in this tape and copy

it into a storage tape to avoid any future interference. Simulate M on input 〈x, y〉.

Note that N has K-amplitudes since the above procedure can be implemented by a series of unitary operatorswith K-amplitudes. Clearly, ρM (x) equals f(x)/4p(n). It suffices to set ℓ(x) = 4p(|x|) and g(x) = ρM (x). ✷

A similar argument of the above proof works for many other amplitude setK, includingK = {0,± 35 ,± 4

5 ,±1}.Next, we show several closure properties of #QPK-functions. An ensemble {|φx〉}x∈I of qustrings is called

an ℓ-qubit source with K-amplitudes if there exists a K-amplitude well-formed clean QTM that, on every inputx, produces |φx〉 of size ℓ(|x|) in its output tape in polynomial time.

Lemma 4.9 Let ℓ be a polynomial. Let f, g ∈ #QPK and p, h ∈ FEQPK with |p(1n)| ∈ O(log n).

1. f ◦ h ∈ #QPK , where f ◦ h denotes the composition λx.f(h(x)).

2. If an ensemble Φ = {|φx〉}x∈{0,1}∗ is an ℓ-qubit source with K-amplitudes, then

λx.(∑

s:|s|=ℓ(|x|) |〈s|φx〉|2f(〈x, s〉)) is in #QPK . In particular, if K ⊇ {0,±1,± 12} then λx. 12 (f(x) + g(x))

is in #QPK .

3. λx.(∏

s:|s|=|p(1|x|)| f(〈x, s〉)) is in #QPK .††The class NQPK was introduced in [1] as a quantum analogue of NP but it has been proven that NQPQ = NQPC = co-C=P

[15, 18, 39].

8

4. λx.f(x)|h(x)| is in #QPK .

Proof. 1) Let Mh be a polynomial-time well-formed K-amplitude QTM that computes h with certainty andlet M be a polynomial-time well-formed K-amplitude QTM whose acceptance probability ρM equals f . Wecan assume without loss of generality from [5, 38] that Mh is stationary in normal form and the running timeof Mh depends only on the length of inputs. Define the new QTM N as follows.

On input x, simulate Mh. Observe the output tape after Mh halts. When |y〉 is observed, simulate Mon input y.

Note that the final superposition of Mh has the form |ψ〉|h(x)〉, where |h(x)〉 is the content of the output tape.Since |ψ〉 does not affect M ’s move, the acceptance probability ρN (x) of N is exactly ρM (h(x)). Thus, weobtain f ◦ h(x) = ρN (x).

2) Let g(x) =∑

s:|s|=ℓ(|x|) |〈s|φx〉|2f(〈x, s〉) for all x. Since {|φx〉}x∈Σ∗ is an ℓ-qubit source with K-amplitudes, let M0 be a polynomial-time K-amplitude well-formed clean QTM that, on input x, produces|φx〉. Let M be another polynomial-time K-amplitude well-formed QTM witnessing f . Consider the QTM Nthat executes the following algorithm.

On input x, simulate M0 to produce |φx〉 in a designated tape. Observe string s in this tape. Copy sinto a storage tape and then simulate M on input 〈x, s〉.

Obviously, N has K-amplitudes. Note that the probability of observing s is exactly |〈s|φx〉|2. Since copying sprevents further interference between two computations of M on input 〈x, s〉 and input 〈x, s′〉, if s 6= s′ thenρN (x) =

∑

s:|s|=ℓ(|x|) |〈s|φx〉|2ρM (〈x, s〉). Thus, g(x) = ρN (x).

The second claim follows from the fact that { 12

∑

s:|s|=2 |s〉}x∈Σ∗ is 2-qubit source with {0,±1,± 12}-amplitudes.

In this case, we use f ′, which is defined as f ′(〈x, 0b〉) = f(x) and f ′(〈x, 1b〉) = g(x) for each b ∈ {0, 1}.3) Let M be any polynomial-time K-amplitude well-formed QTM that witnesses f . Assume that M is

stationary in normal form. Consider the following QTM N .

On input x, compute m = |p(1|x|)| deterministically. Write s := 0m in a counter tape. Repeat thefollowing by incrementing lexicographically string s written in the counter tape. Copy x and s into anew blank area of a work tape and then simulate M on input 〈x, s〉. If all runs of M end with acceptingconfigurations, then accept; otherwise, reject.

Note that each run ofM is independent and does not interfere with other runs. Thus, the acceptance probability

of N equals∏

s:|s|=|p(1|x|)| ρM (〈x, s〉). Moreover, the number of runs of M is exactly 2m = 2|p(1|x|)|, which is

polynomially bounded. Hence, N runs in polynomial time.4) Let Mf be a well-formed K-amplitude QTM that witnesses f in time polynomial q. Assume that a

polynomial p satisfies |h(x)| ≤ p(|x|) for all x. Consider the following algorithm.

On input x, run Mf |h(x)| times and idle q(|x|) steps p(|x|)− |h(x)| times to avoid the timing problem.Accept x if all the first |h(x)| runs of Mf reach accepting configurations; otherwise, reject x.

Similar to the proof of 3), the above algorithm accepts x with probability exactly ρMf(x)|h(x)|. ✷

Although #QP enjoys useful closure properties as shown in Lemma 4.9, it is obviously not closed undersubtraction. In classical context, Fenner et al. [14] studied the subtraction closure of #P, named GapP.Similarly, the subtraction closure of #QP is naturally considered. We call such functions quantum gap functions.A quantum gap function is formally defined to output the difference between the acceptance and rejectionprobabilities of a well-formed QTM.

Definition 4.10 A function f from Σ∗ to [−1, 1] is called a polynomial-time quantum gap function with K-amplitudes if there exists a polynomial-time well-formed QTM M with K-amplitudes such that, for every x,

f(x) equals ‖|φ(1)x 〉‖2 − ‖|φ(0)x 〉‖2 when M on input x halts in a final superposition |φ(0)x 〉|0〉 + |φ(1)x 〉|1〉, wherethe last qubit represents the content of the output tape of M . In other words, f(x) = 2ρM (x)− 1. GapQPK isthe set of all polynomial-time quantum gap functions with K-amplitudes.

The closure properties of GapQP directly follow from those of #QP using the closely-knitted relationshipbetween #QP and GapQP shown in §5.2.

9

5 Relationships among Quantum Functions

Empowered by quantum mechanism, quantum computation can draw close quantum functions of differentnature. The relationship among these quantum functions are of special interests. In the past, several usefultechniques have been developed to analyze quantum computations. These techniques are used in this sectionto show close connections among the quantum functions.

5.1 A Relationship Between #QP and FBQP

When the range of a #QP-function f is restricted to the set {0, 1}, f belongs to FEQP. For a more generalrange, we show that every #QP-function can be approximated by a certain FBQP-function, where we viewFBQP-functions as functions from Σ∗ to D.

Grover [20] discovered a fast quantum algorithm for a database search problem. His algorithm is meantto find the location of a target key word in a large database provided that there is a unique location for thekey word. Brassard et al. [7] extended Grover’s database search algorithm and showed how to compute withǫ-accuracy the amplitude of a given superposition in O(

√N) time, where N is the size of search space, with

high probability.We rephrase the result of Brassard et al. in the following fashion. We include the proof for completeness.

Theorem 5.1 #QP ⋐p FBQP ∩DΣ∗

.

Proof Sketch. Let f be a #QP-function witnessed by a polynomial-time well-formed QTM M . By ReversalLemma, let MR denotes the reversing QTM of M . Let p be any polynomial. For convenience, we use integersbetween 0 and 2k − 1 instead of string of length k.

Let x be any string of length n. Define k = ⌈log p(n)⌉+5. Since M always outputs either 0 or 1, we need toconsider two superpositions. For each i ∈ {0, 1}, let |Φ(i)〉 be the superposition of final configurations of M oninput x, in which M writes i in the output tape. Let θ be the real number in [0, π2 ] such that sin θ = ‖|Φ(1)〉‖.Note that f(x) = sin2 θ.

Consider the following quantum algorithm Q that starts on input x.

On input x, copy x into a storage tape to remember x and then simulate M on |x〉. Produce |0〉 and

apply H⊗k to generate 1√2k

∑2k−1m=0 |m〉. Observe |m〉 and repeat the following procedure 1)-4) m times:

1. If output qubit is |1〉, change it to −|1〉; otherwise do nothing;2. simulate MR (except for the storage tape);3. if the input tape contains only |x〉 and empty elsewhere, then change it to −|x〉; otherwise do

nothing; and4. simulate M and change all |s〉 to −|s〉.

Apply QFTk, which maps |m〉 to 1√2k

∑2k−1ℓ=0 e

2πimℓ

2k |ℓ〉.

This algorithmQ has two eigenvectors |Ψ0〉 = 1√2

(

eiθ

cos θ |Φ(0)〉 − ieiθ

sin θ |Φ(1)〉)

and |Ψ1〉 = 1√2

(

e−iθ

cos θ |Φ(0)〉+ ie−iθ

sin θ |Φ(1)〉)

with their corresponding eigenvalues e2iθ and e−2iθ; namely, Q|Ψ0〉 = e2iθ|Ψ0〉 and Q|Ψ1〉 = e−2iθ|Ψ1〉).After Q, the sum of the squared norms of both |⌊ 2kθ

π ⌋〉|Ψ0〉 and |⌈ 2kθdπ ⌉〉|Ψ0〉 is at least 4

π2 . Similarly, the

squared norms of both |2k − ⌊ 2kθπ ⌋〉|Ψ1〉 and |2k − ⌈ 2kθ

π ⌉〉|Ψ1〉 sum up to at least 4π2 .

Now, we observe |ℓ〉 and define ℓ′ as ℓ′ = ℓ if ℓ ≤ 2k

2 and ℓ′ = 2k − ℓ otherwise. The probability of either

ℓ′ = ⌊ 2kθπ ⌋ or ℓ′ = ⌈ 2kθ

π ⌉ is at least 8π2 . Moreover, |θ − πℓ′

2k| < π

2kand thus, | sin2 θ − sin2 πℓ′

2k| ≤ π2

22k. Since

f(x) = sin2 θ, |f(x) − sin2 πℓ′

2k | < 14p(n) . To obtain the desired result, we must output the 1

4p(n) -approximation

of the value sin2 πℓ′

2k.

Unfortunately, our algorithm has a minor problem due to the fact that no QTM can perform QFTk exactly[28]. However, it is possible to replace QFTk by a polynomial-time well-formed QTM that 1

4p(n) -approximates

QFTk. ✷

Theorem 5.1 cannot be improved to #QP ⋐e FBQP unless BQP = PP.

Proposition 5.2 If BQP 6= PP then #QP 6⋐e FBQP.

10

Proof. We show the contrapositive. Assume that #QP ⋐e FBQP. Let A be any set in PP. There exists a

function f ∈ #P and a polynomial p such that, for every x, x ∈ A implies f(x) > 12 +2−p(|x|) and x 6∈ A implies

f(x) ≤ 12 . By Lemma 4.8, there exists two functions g ∈ #QP and ℓ ∈ FP ∩ NΣ∗

such that f(x) = g(x)ℓ(x)

for all x. Let q be a polynomial satisfying ℓ(x) ≤ 2q(|x|) for all x. Since #QP ⋐e FBQP, we can choose a

function h ∈ FBQP ∩ DΣ∗such that |g(x) − h(x)| ≤ 2−p(|x|)−q(|x|)−2 for all x. Let x ∈ Σn. On one hand, if

x ∈ A then g(x) > 1ℓ(x)(

12 + 2−p(|x|)) ≥ 1

2ℓ(x) + 2−p(n)−q(n)−2, which implies h(x) > 12ℓ(x) + 2−p(n)−q(n)−1. On

the other hand, if x 6∈ A then g(x) ≤ 12ℓ(x) . Thus, h(x) <

12ℓ(x) + 2−p(n)−q(n)−1. Clearly, h and ℓ determine the

membership of A. Since they are both in FBQP, A belongs to BQP. Therefore, BQP = PP. ✷

5.2 Relationships Between #QP and GapQP

We consider the relationship between GapQP and #QP. Quantum nature brings them closer than classicalcounterparts, GapP and #P. We first show that GapQP is the subtraction closure of #QP. More precisely, forany two sets F and G of functions, let F − G denote the set of all functions of the form f − g, where f ∈ Fand g ∈ G. We prove that GapQP = #QP − #QP. This shows another characterization of GapQP. As animmediate consequence, we have #QP ⊆ GapQP.

For any QTM M with K-amplitudes, the notation M denotes the QTM that simulates M and, when Mhalts, flip M ’s output bit. Note that M also has K-amplitudes.

Proposition 5.3 If K ⊇ {0,±1,± 12} then GapQPK = #QPK −#QPK .

Proof. Let f be any function in GapQPK . There exists a polynomial-time well-formed QTM M withK-amplitudes such that f(x) is the difference between ρM (x) and ρM (x). Since ρM and ρM are both #QPK-functions, f belongs to #QPK −#QPK . Thus, GapQPK ⊆ #QPK −#QPK .

Conversely, assume that f ∈ #QPK −#QPK . There exist two polynomial-time K-amplitude well-formedstationary QTMs Mg and Mh in normal form satisfying that f(x) = ρMg

(x) − ρMh(x) for all x. In what

follows, we use Mh instead of Mh. Generally speaking, the halting timings of Mg and Mh are different. Recallthat the running times of these machines are precisely certain polynomials, say pg and ph, in the lengths ofinputs. To synchronize the halting timing of these machines, we should attach a counter tape that behaves likea clock (counting the number of steps). When the machines halt, we force them to idle until the counter hitspg(|x|) + ph(|x|). Thus, we can assume that Mg and Mh halt at the same time. Now, consider the followingQTM N whose tape alphabet includes Σ4 = {0, 1, 2, 3}.

On input x, first write |0〉 in a separate blank tape and apply H2. Observe this symbol. When either |0〉or |1〉 is observed, simulate Mg on input x. Otherwise, simulate Mh on x.

Note that N has K-amplitudes. The acceptance probability of N is exactly 12ρMg

(x) + 12 (1 − ρMh

(x)). Thus,the gap 2ρN (x) − 1 is exactly ρMg

(x) − ρMh(x), which equals f(x). Therefore, f ∈ GapQPK . This concludes

that #QPK −#QPK ⊆ GapQPK . ✷

Since #QPK ⊆ #QPK −#QPK , we obtain the following corollary.

Corollary 5.4 #QPK ⊆ GapQPK if K ⊇ {0,±1,± 12}.

Unfortunately, it is unknown whether all nonnegative GapQP-functions are in #QP. Here, we give a partialsolution. Fenner et al. [15] recently proved that, for every GapP-function f , there exists a polynomial-timewell-formed QTM that accepts input x with probability 2−p(|x|)f(x)2 for a certain fixed polynomial p. Thisimplies that if f ∈ GapP then λx.2−p(|x|)f(x)2 ∈ #QP. Their result can be expanded to the following theorem.This exemplifies a characteristic feature of quantum gap functions. Write f2(x) = (f(x))2 for all x.

Theorem 5.5 (Squared Function Theorem) Assume that K∗ ⊆ K. If f ∈ GapQPK then f2 ∈ #QPK .

To prove Theorem 5.5, we need the following key lemma, called Gap Squaring Lemma. This is compared toSquaring Lemma.

Lemma 5.6 (Gap Squaring Lemma) Assume that K∗ ⊆ K. Let M be a polynomial-time well-formed QTMwith K-amplitudes. There exists another polynomial-time well-formed K-amplitude QTM N that, on inputs xgiven in tape 1 and empty elsewhere, halts in a final superposition in which the amplitude of configuration |x〉|1〉

11

is exactly 2ρM (x) − 1, where the last qubit |1〉 represents the content of the output tape.

Proof. Let M be the QTM given in the lemma. We define the desired QTM N as follows. First, given inputx, N simulatesM on the same input. Assume thatM halts in a final superposition |φ〉 = ∑

y αx,y|y〉|by〉, wherey ranges all (valid) configurations (except the content of the output tape) ofM and the last qubit |by〉 representsthe content of the output tape ofM . The acceptance probability ρM (x) equals

∑

y:by=1 |αx,y|2. Next, N applies

the phase shift P to |by〉 and then we obtain the superposition |φ′〉 = ∑

y:by=1 αx,y|y〉|1〉 −∑

y:by=0 αx,y|y〉|0〉.By Reversal Lemma, there exists a polynomial-time well-formed QTM MR that reverses the computation

of M . Note that MR also has K-amplitudes since K∗ ⊆ K. Now, N simulates MR starting with |φ′〉 as itsinitial superposition. Note that if we run MR on input |φ〉 then MR|φ〉 becomes the initial superposition ofM . Moreover, the inner product of |φ〉 and |φ′〉 is 〈φ|φ′〉 =

∑

y:by=1 |αx,y|2 − ∑

y:by=0 |αx,y|2, which equals

2ρM (x) − 1.At the end, N outputs 1 (i.e., acceptance) if it observes exactly the initial configuration of M on input

x; otherwise, N outputs 0 (i.e., rejection). The acceptance probability of N is exactly 〈MR|φ〉|MR|φ′〉〉 sinceMR|φ〉 is the initial superposition of M . Since N preserves the inner product, we have 〈MR|φ〉|MR|φ′〉〉 =〈φ|φ′〉 = 2ρM (x)− 1. ✷

Now, we are ready to prove Theorem 5.5.

Proof of Theorem 5.5. Let f be any GapQPK-function, which is witnessed by a certain polynomial-time well-formed QTMM withK-amplitudes. Thus, f(x) = 2ρM (x)−1 for all x. Let p be a polynomial satisfying thatMon input x halts in time p(|x|). It follows from Gap Squaring Lemma that there exists another polynomial-timeK-amplitude well-formed QTM N that starts on input (x, 1p(|x|)) and halts in a final superposition in which theamplitude of |x〉|1〉 is 2ρM (x)−1. To complete the proof, we need to add to N a reversible pre-computation thatproduces 1p(|x|) using an extra work tape. Note that any garbage left behind after the pre-computation doesnot interfere with any computation that follows. Thus, the acceptance probability of N equals (2ρM (x) − 1)2,which is clearly f2(x). Therefore, f2 belongs to #QPK . ✷

5.3 Relationships between GapQP and GapP

Quantum gap functions are closely related to their classical counterpart. The following theorem shows a differ-ently intertwined relationship between GapQPK and GapPK depending on the choice of amplitude set K. Ifwe restrict amplitudes on rational numbers, then GapQP and GapP can bear fundamentally the same compu-tational power. This indicates a limitation of quantum computation.

Let sign(a) be 0, 1, and −1 if a = 0, a > 0, and a < 0, respectively.

Theorem 5.7 1. For every f ∈ GapQPC, there exists a function g ∈ GapP such that, for every x, f(x) = 0iff g(x) = 0.

2. For every f ∈ GapQPCand every polynomial q, there exist two functions k ∈ GapP and ℓ ∈ FP ∩ NΣ∗

such that |f(x)− k(x)ℓ(1|x|)

| ≤ 2−q(|x|) for all x.

3. For every f ∈ GapQPA, there exists a function g ∈ GapP such that sign(f(x)) = sign(g(x)) for all x.

4. For every f ∈ GapQPQ, there exist two functions k ∈ GapP and ℓ ∈ FP ∩ NΣ∗such that f(x) = k(x)

ℓ(1|x|)

for all x.

In what follows, we give the proof of Theorem 5.7. Partly, we use the result by Yamakami and Yao [39],who introduced a canonical representation of the amplitude α of any configuration in the superposition of aquantum computation at time t. This representation makes it possible to encode such an amplitude into a finitesequence of integers and to simulate quantum computation by modifying each of such sequences in a classicalmanner.

Fix a well-formed QTM M and let D be the amplitude set of M . Let A = {α1, . . . , αm} be the maximalsubset of D that is algebraically independent. Define F = Q(A) and let G be the field generated by elements in{1}∪(D−A) over F . Let {β0, β1, . . . , βd−1} be a basis of G over F with β0 = 1. Let D′ = D∪{βiβj | i, j ∈ Zd}.Take a common denominator u such that, for every α ∈ D′, uα is of the form

∑

~k a~k(∏m

i=1 αki

i )βk0 , where~k = (k0, k1, . . . , km) ranges over Zd × Zm and a~k ∈ Z. Thus, there exist two constants e, d > 0 such that, theamplitude α of each configuration in the superposition of M at time t on input x, when multiplied by u2t−1,

12

can be of the form∑

~k a~k(∏m

i=1 αki

i )βk0 , where~k = (k0, k1, . . . , km) ranges over Zd × (Z[2et])

m and a~k ∈ Z. It isimplicitly shown in [39] that

∑

~k |a~k| ≤ 2p(|x|,t) for a certain fixed polynomial p.1) Let f be any function in GapQPC. By Squared Function Theorem, f2 belongs to #QPC. Yamakami and

Yao [39] essentially showed that, for every g ∈ #QPC, there exists a h ∈ GapP such that, for every x, g(x) = 0iff h(x) = 0.

Thus, there exists a function h ∈ GapP such that, for every x, f2(x) = 0 iff h(x) = 0. It immediately followsthat f(x) = 0 iff h(x) = 0.

2) Let f ∈ GapQPCand take a polynomial-time well-formed C-amplitude QTMM that witnesses f . Without

loss of generality, we can assume thatM is synchronous, stationary, and in normal form. Let q be any polynomial.We use the following lemma. For a QTM M , let ampM (x,C) denote the amplitude of configuration C of

M on input x in a final superposition at the time when M halts. The complex conjugate of M is the QTMM∗ defined exactly as M except that its time-evolution matrix is the complex conjugate of the time-evolutionmatrix of M .

Lemma 5.8 Assume that K∗ ⊆ K. Let M be a well-formed synchronous stationary K-amplitude QTM innormal form with running time T (x) on input x. There exists a well-formed QTM N such that, for every x, (1)N halts in time O(T (x)) with one symbol from {0, 1, ?} in its output tape; (2)

∑

C∈D1xampN(x,C) = ρM (x);

and (3)∑

C∈D0xampN(x,C) = 1 − ρM (x), where Di

x is the set of all final configurations, of N on x, whose

output tape consists only of symbol i ∈ {0, 1}.

Proof. The desired QTM N works as follows. On input x, N simulates M on input x; when it halts in afinal configuration, N starts another round of simulation of M∗ in a different set of tapes. After N reaches afinal configuration, we obtain two final configurations C1 and C2. Then, N deterministically checks if both finalconfigurations are identical. If C1 6= C2, then N outputs symbol “?” and halts. Now, assume that C1 = C2.If this unique configuration C1 is an accepting configuration, then N outputs 1; otherwise, it outputs 0. Notethat we obtain the amplitude ampM (x,C1) · ampM∗(x,C2), which equals |ampM (x,C1)|2. Let i ∈ {0, 1}. Forconvenience, let Ei

x denote the set of all final configurations, of M on x, in which M outputs i. Thus, the sum∑

C∈DixampN (x,C) equals

∑

C∈Eix|ampM (x,C)|2, which is exactly ρM (x). ✷

By Lemma 5.8, there exists a polynomial-time well-formed C-amplitude QTM M such that f(x) equals∑

C∈D1xampM (x,C) − ∑

C∈D0xampM (x,C), where Di

x is defined in Lemma 5.8. Let r be a polynomial that

bounds the running time of M and also satisfies |D0x ∪D1

x| ≤ 2r(|x|).Let x be any string of length n. Let ℓ(x) = 22r(n)+q(n)+1 and g(x,C) = ampM (x,C). Assume first that

there exists a function g ∈ GapP such that |g(x,C)− g(x,C)ℓ(1n) | ≤ 2−r(n)−q(n)−1, which implies |∑C∈Di

xg(x,C)−

∑

C∈Dix

g(x,C)ℓ(1n) | ≤ 2r(n) · 2−r(n)−q(n)−1 = 2−q(n)−1 for each i ∈ {0, 1}. The desired GapP-function k is defined as

k(x) =∑

C∈D1xg(x,C)−

∑

C∈D0xg(x,C). Thus, |f(x)− k(x)

ℓ(1n) | ≤ |∑

C∈D1xg(x,C)− g(x,C)

ℓ(1n) |+ |∑

C∈D0xg(x,C)−

g(x,C)ℓ(1n) | ≤ 2−q(n).

To complete the proof, we show the existence of such g. Note that each amplitude α of the transition functionof M is approximated by a certain polynomial-time TM with any desired precision since such amplitude α is inC. By simulating such a machine in polynomial time, we can get the most significant 2r(n) + q(n) + 1 bits of αso that, for every computation path P ofM on input x, we can compute the approximation ρP of the amplitudeρP of computation path P to within 2−r(n)−q(n)−1. Let h(x, P ) be the integer satisfying h(x, P ) = ℓ(1n)|ρP |.Set g(x,C) =

∑

P∈Px,Ch(x, P ), where Px,C is the collection of all computation paths of M on x that lead to

configuration C.3) Let f ∈ GapQPA. By 1), there exists a function g0 ∈ GapP such that, for every x, g0(x) = 0 iff f(x) = 0.

Let M be a polynomial-time well-formed A-amplitude QTM that witnesses f in time polynomial p. Recall thecanonical representation of any amplitude of a quantum computation. Let x be any input of length n. Since Mhas A-amplitudes, the amplitude α of each configuration in the final superposition of M in time p(n) on input

x ∈ Σn, when multiplied by u2p(n)−1, has the form∑

~k a~k(∑m

i=0 αki

i ), where ~k = (k0, k1, . . . , km) ranges overZd × (Z[2ep(n)])

m and a~k ∈ Z.We use the lemma below.

Lemma 5.9 (cf. [34]) Let α1, . . . , αm ∈ A. Let d be the degree of Q(α1, . . . , αm)/Q. There exists a constant

c > 0 that satisfies the following for any complex number α of the form∑

~k a~k(∏m

i=1 αki

i ), where ~k = (k1, . . . , km)

13

ranges over Z[N1] × · · · ×Z[Nm], (N1, . . . , Nm) ∈ Nm, and a~k ∈ Z. If α 6= 0 then |α| ≥ (∑

~k |a~k|)1−d∏m

i=1 c−dNi.

By Lemma 5.9, the amplitude α of each configuration in the final superposition, when M halts on inputx, has the squared magnitude at least 1

|u|2p(n)−1 (∑

~k |a~k|)1−d∏m

i=1 c−2edp(n). Note that

∑

~k |a~k| and |u|2p(n)−1

are both bounded above by an exponential in n. By choosing an appropriate polynomial s, we thus obtain|α| ≥ 2−s(n). This yields the 2−s(n) lower bound of |f(x)| when f(x) 6= 0.

By 2), there are two functions k ∈ GapP and ℓ ∈ FP ∩ NΣ∗satisfying that |f(x) − k(x)

ℓ(1|x|)| ≤ 2−s(|x|)−1 for

all x. Consider the case where f(x) > 0. Since f(x) > 2−s(n), it follows that k(x)ℓ(1n) ≥ 2−s(n)−1 > 0. In the case

where f(x) < 0, since f(x) < −2−s(n), we have k(x)ℓ(1n) ≤ −2−s(n)−1 < 0.

Now, the desired function g is simply defined as g(x) = g0(x) · k(x) for all x. It is easy to show that g is inGapP.

4) In 2), since M has Q-amplitudes, we can compute the amplitude ρP of computation path P and thus, we

have h(x, P ) = ℓ(1n)|ρP |. Therefore, f(x) = k(x)ℓ(1n) .

This completes the proof of Theorem 5.7.

6 Quantum Functions with an Access to Oracles

An oracle is an external device that provides computation with extra information by means of oracle queries.The role of oracles in quantum computation was recognized as far back as the early 1990s by Deutsch and Jozsa[11]. An oracle QTM seems to give more flexibility to oracle queries than a quantum circuit does with oraclegates. This section introduces relativized quantum functions that can access oracles in two different manners:adaptive and nonadaptive queries.

6.1 Adaptive Queries and Nonadaptive Queries

We first give a general resource-bounded query model for relativized quantum functions. For a later use, arestriction of the number of queries is imposed on every computation path of a given oracle QTM. Such arestriction arises the notion of bounded queries.

In what follows, r denotes an arbitrary function from N to N, and R a subset of NN. An oracle A is a subsetof Σ∗ and C denotes an arbitrary class of oracles.

Definition 6.1 Let A be any oracle. A function f is in FEQPA[r] if there exists a polynomial-time well-formedoracle QTM M such that, for every x, M on input x outputs f(x) with certainty using oracle A and makes at

most r(|x|) queries on each computation path. Let FEQPC[R] be the union of FEQPA[r]’s for all A ∈ C. The

class FEQPA[R] (FEQPC[R], resp.) is the union of FEQPA[r]’s (FEQPC[r]’s, resp.) for all r ∈ R. Conventionally,

when R = NN, we write FEQPA (FEQPC , resp.) instead of FEQPA[R] (FEQPC[R], resp.). Similar notions areintroduced to FBQP, #QP, and GapQP.

The function in Definition 6.1 is said to make adaptive (or sequential) queries since the choice of a queryword relies on the oracle answers to its previous queries. In contrast, we define a quantum function that makesnonadaptive (or parallel) queries where all query words are pre-determined before any oracle queries. Ournonadaptive query model‡‡ is an immediate adaptation of NPA

‖ (see, e.g., [36]). Every computation path Pgenerates in a designated tape a query list—a list of all query words (separated by a special separator) that arepossibly queried along computation path P before any query on this path.

There are three important issues concerning the definition of parallel queries in a quantum setting. Thefirst issue is the timing of the completion of all the query lists. Quantum interference makes it possible for twodifferent computation paths to exchange information. We should avoid the case where the first query is madeat a computation path P1 but a query list on another computation path P2 is not yet finished because thispath P2 may glance at the result of the query made on the path P1 using quantum interference. An importantrequirement of parallel queries is that we require a QTM to complete all query lists just before it enters thepre-query state for the first time in the entire computation tree of the machine. At this moment, we say that allthe query list are completed. Once the query list on each computation path is completed, M can freely deleteany word from the list but cannot add any word to the list.

‡‡The nonadaptive query model was independently introduced in [8].

14

The second issue concerns the actual queries compared to the words in a query list. In a classical setting,we can always assume that all the query words in a query list should be actually queried whether or not weuse all the oracle answers. We find it harder for a QTM to perform quantum interference if the QTM keepsunnecessary oracle answers in its tapes. Thus, the classical requirement would be relaxed so that all the querywords in a query list are not necessarily queried during a computation.

The last issue is the maintenance of query lists. Since maintaining a query list until the end of the compu-tation may prohibit any quantum interference on a computation that follows. Thus, the completed query listsmay be altered after any first query in order to make computation paths interfere with other computation pathsthat had produced different query lists.

Definition 6.2 The class FEQPA[r]‖ is the subset of FEQPA[r] with the extra condition that, on each compu-

tation path, just before M enters a pre-query state for the first time in the entire computation, it completes allquery lists. Any query list completed on each computation path must be maintained unaltered until the firstquery on this computation path but may be altered once the machine makes the first query on this computationpath. All the words in the query list may not be queried but any word that is queried must be in the query

list. The class FEQPC[r]‖ is the union of FEQP

A[r]‖ ’s over all A ∈ C. The notation FEQP

A[R]‖ (FEQP

C[R]‖ , resp.)

denotes the union of FEQPA[r]‖ ’s (FEQP

C[r]‖ , resp.) over all r ∈ R. Similar notions are introduced to FBQP,

#QP, and GapQP.

The following lemma is immediate.

Lemma 6.3 1. FEQP = FPEQP‖ = FPEQP = FEQPEQP

‖ = FEQPEQP.

2. FBQP = FPBQP‖ = FPBQP = FBQPBQP

‖ = FBQPBQP.

3. QMASV ⊆ FPQMA‖ .

4. #QP = #QPEQP‖ = #QPEQP.

5. GapQP = GapQPEQP‖ = GapQPEQP.

Proof. 1) It immediately follows that FPEQP‖ ⊆ FPEQP ⊆ FEQPEQP. To show that FEQPEQP ⊆

FPEQP‖ , let f ∈ FEQP and let p be a polynomial such that |f(x)| ≤ p(|x|) for all x. Define A = {〈x, 1i〉 |

the ith bit of f(x) is 1} and B = {〈x, 1j〉 | |f(x)| ≥ j}. The set B is necessary to determine the length of f(x).Obviously, A,B ∈ EQP by simulating the QTM that computes f . Thus, A⊕B ∈ EQP. Now, it is easy to showthat f ∈ FPA⊕B

‖ by making nonadaptive queries 〈x, 1〉, 〈x, 11〉, . . . , 〈x, 1p(|x|)〉 to both A and B.

It still remains to prove that FEQPEQP ⊆ FEQP. Let f ∈ FEQPA for a certain oracle A in EQP. Let M bea polynomial-time well-formed oracle QTM thatM on input x outputs f(x) with certainty. By Canonical FormLemma, we can assume that M is in a canonical form with oracle A. Let N be a polynomial-time well-formedQTM that recognizes A with certainty. Consider the quantum algorithm Q that simulates M on input x and,whenever it invokes a query y, simulate N on input y. This Q can be carried out on a well-formed oracleQTM since Q makes the same number of queries to oracle A with query words of the same length along everycomputation path on any input of fixed length.

2) Similar to 1) except for the proof of FBQPBQP = FBQP. This is shown in a way similar to BQPBQP =BQP [4] by amplifying the success probability of a QTM, which computes a given oracle set, from 3/4 to closeto 1 so that the cumulative error is still bounded below by 1/4 after the polynomially-many runs of this QTM.

3) Let f ∈ QMASV witnessed by a polynomial p and a polynomial-time well-formed QTMM as in Definition4.4. Let q be a polynomial satisfying |f(x)| ≤ q(|x|) for all x. Define A as the collection of all strings 〈x, 1i〉,where x ∈ Σ∗ and 0 ≤ i ≤ p(|x|), such that there exist a string y ∈ Σq(|x|) and a qustring |φ〉 ∈ Φp(|x|) satisfyingthat M on input |x〉|φ〉 outputs |1〉|y〉 with the condition that the ith bit of y must be 1. The set B is similarto A but it checks if M on input |x〉|φ〉 outputs |1〉|y〉 with |y| ≥ i. It is easy to see that A and B are in QMA.Similar to 1), making nonadaptive queries to A⊕B computes f(x) in polynomial time.

4) and 5) are similar to 1). ✷

An immediate corollary shows that EQP characterizes the low sets for #QP and GapQP in comparison withlow-#P = UP ∩ co-UP ⊆ SPP = low-GapP [14].

Corollary 6.4 EQP = low-#QP‖ = low-#QP = low-GapQP‖ = low-GapQP.

15

Proof. Clearly, low-GapQP ⊆ low-GapQP‖. Since GapQPEQP = GapQP by Lemma 6.3(5), it followsthat EQP ⊆ low-GapQP. To show that low-GapQP‖ ⊆ EQP, let A ∈ low-GapQP‖. It is easy to see that

χA ∈ GapQPA[1]‖ . Since GapQPA

‖ ⊆ GapQP, χA ∈ GapQP. Since #QP∩{0, 1}Σ∗= GapQP∩{0, 1}Σ∗

, χA also

belongs to #QP. This yields the desired conclusion that A ∈ EQP. Therefore, low-GapQP‖ ⊆ EQP. Similarly,we can show that EQP = low-#QP‖ = low-#QP. ✷

Classically, a function class and a language class exhibit a wide gap; for instance, PNP‖ = PNP[O(logn)] [36]

but FPNP‖ 6= FPNP[O(logn)] if NP 6= RP [21]. On the contrary, quantum interference draws such two classes

close together. The following lemma is an adaptation of the argument in [8], in which an quantum algorithmof Bernstein and Vazirani [5] is effectively used.

Lemma 6.5 Let R ⊆ {0, 1}Σ∗and assume that R is closed under constant multiplication. For any oracle A,

FBQPA‖ ⊆ FBQPA[R] iff BQPA

‖ ⊆ BQPA[R].

Proof. The implication from left to right is obvious. Now, assume that BQPA‖ ⊆ BQPA[R]. Let f be any

function in FBQPA‖ . We want to show that f belongs to FBQPA[R]. Let p be a polynomial that bounds the

length of the value of f . Without loss of generality, we can assume that f is length-regular (i.e., |x| = |y|implies |f(x)| = |f(y)|) since, otherwise, we set f(x) = f(x)10p(n)−|f(x)| for all x. For simplicity, assume that|f(x)| = p(|x|) for all x.

Define B = {〈x, z, b〉 | b ∈ {0, 1}, |z| = |f(x)|, b⊕ (f(x) · z) = 1}, where u · v is the dot product of u and v. It

follows from f ∈ FBQPA‖ that B is in BQPA

‖ . By our assumption, B is also in BQPA[r(n)] for a certain functionr ∈ R. Let M be a well-formed oracle QTM that recognizes B with oracle A with error probability ≤ 1/4. Wefirst amplify its success probability from 3/4 to

√14/4. For such a QTM, we apply Squaring Lemma (for an

oracle QTM) and obtain another QTM M1 that produces a final superposition of configurations, one of which

has only |x〉|z〉|χB(〈x, z, b〉)〉 written in the tapes with probability ≥ (√144 )2 = 7

8 .

The new QTM N works as follows. On input x of length n, write |0p(n)〉|1〉 and apply H⊗p(n) ⊗ H . Wethen have 2−p(n)/2

∑

z:|z|=p(n) |z〉 ⊗ |φ+〉, where |φ+〉 = 1√2(|0〉 − |1〉). For each |z〉|b〉, where b ∈ {0, 1}, run M1

with oracle A that changes |x〉|z〉|b〉 to α|x〉|z〉|b⊕ (f(x) · z)〉+ |ψx,z,b〉, where |α|2 = 78 . Thus, ‖|ψx,z,b〉‖2 = 1/8.

Apply I⊗n ⊗ H⊗p(n) ⊗ I. The final superposition becomes α|x〉|f(x)〉|φ+〉 + |ψ′〉. Since ‖|ψ′〉‖ = 18 , we can

observe |x〉|f(x)〉 with probability at least 78 − 1

8 = 34 . Obviously, N makes only O(r(n)) queries. Thus,

f ∈ FBQPA[O(r(n))] ⊆ FBQPA[R]. ✷

6.2 Oracle Separation

Berthiaume and Brassard [6] earlier constructed an oracle A such that PA 6= EQPA[1] using the algorithm of

Deutsch and Jozsa [11]. We expand their result by showing the existence of a set A such that FEQPA[1]‖ ∩

{0, 1}Σ∗* #EA, which immediately implies FEQP

A[1]‖ * FPA since FPA ⊆ #EA.

Proposition 6.6 There exists an oracle A such that FEQPA[1]‖ ∩ {0, 1}Σ∗

* #EA.

Proof. We say that a set A is good if, for every n ∈ N and every x ∈ Σn, either |A ∩ Σn2 | = |Σn2 \ A| or|A ∩Σn2 | · |Σn2 \A| = 0. For any set A and any string x ∈ Σn, let fA(x) = 2−2n2 · (|A ∩Σn2 | − |Σn2 \A|)2. Tocompute the function, consider the following oracle QTM N .

On input x of length n, write |0n2〉|1〉 in a query tape and apply H⊗n2 ⊗H . Copy the first n2 bits into

a query list. Invoke an oracle query. Delete the query list. Again, apply H⊗n2 ⊗ I. Observe the first n2

bits in the query tape. Output 1 if |0n2〉 is observed; output 0 otherwise.

The deletion of each query list is possible since the query list contains the exact copy of the first n2 bits in thequery tape. It follows by a simple calculation that NA on input x outputs fA(x) with certainty if A is drawn

from A. Thus, fA belongs to FEQPA[1]‖ ∩ {0, 1}Σ∗

for any good oracle A.

Next, we construct an oracle A ∈ A such that fA 6∈ #EA. For the construction, we need an effectiveenumeration of all 2O(n)-time bounded nondeterministic TMs. Let {Mi}i∈N be such an enumeration and define

16

{ci}i∈N to be an enumeration of natural numbers (with repetition) such that eachMi halts within time 2ci(n+1)

on all inputs of length n, independent of the choice of oracles. We construct the desired oracle A stage by stage.Initially, set n−1 = 0 and A−1 = Ø. At stage i of the construction of A, let ni denote the minimal integer

satisfying that ni−1 < ni and ci(ni + 1) < n2i − 1. Assuming Ai−1 ⊆ Σn2

i−1 , let B = Ai−1 ∪ Σn2i . Clearly,

fB(0ni) = 1. If #MBi (0ni) 6= 1, then define Ai to be B. Assume otherwise. There exists a unique accepting

computation path P ofMi on 0ni . Let QP denote the set of all words thatMi queries along computation path P .Since |QP | ≤ 2ci(ni+1) < 2n

2i−1, there is a subset C of Σn2

i such that Ai−1 ∪QP ⊆ C and |C ∩Σn2i | = |Σn2

i \C|.For this C, #MC

i (0ni) ≥ 1 but fC(0ni) = 0. Thus, we should set Ai = C. After all the stages, defineA =

⋃

i∈NAi. ✷

Proposition 6.6 demonstrates the power of the nonadaptive query class FEQPA[1]‖ against the adaptive query

class #EA. We next show the existence of an oracle such that FPA[n] ∩ {0, 1}Σ∗is more powerful than even

#QPA‖ .

For its proof, we use the concept of query magnitude from [4]. Let M be a well-formed oracle QTM, A anoracle, and x an input. For any string y, let qty(M,A, x) denote the query magnitude of y of MA on x at time t,

which is defined as the sum of squared magnitudes in the superposition of configurations conf of MA on inputx at time t such that conf is in a pre-query state with query word y. The following is a key lemma due to [4].

Lemma 6.7 LetM be a well-formed oracle QTM whose running time T (n) does not depend on the choice of or-acles. For any two oracles A and B and for any two qustrings |φ〉 and |ψ〉 of size n, |ρAM (|φ〉)−ρBM (|ψ〉)| ≤ ‖|φ〉−|ψ〉‖ +

2√

T (n)(∑T (n)−1

i=1

∑

y∈A△B

√

qiy(M,A, |φ〉)).

Theorem 6.8 There exists a set A such that FPA[n] ∩ {0, 1}Σ∗ 6⊆ #QPA‖ .

Proof. We begin with the definition of a test function fA. For any string y ∈ Σ≥3, let yA = χA(y0|y|−2)χA(y0

|y|−31)χA(y0|y|−411) · · ·χA(y1

|y|−2).

Note that |yA| = |y| − 1. Define fA(x) = χA(x0|x|A ) for all x ∈ Σ∗ and A ⊆ Σ∗. Since fA(x) ∈ {0, 1} for all x

and A, fA ∈ FPA[n] ∩ {0, 1}Σ∗.

It suffices to construct a set A such that fA 6∈ #QPA‖ . Let {Mi}i∈N and {pi}i∈N be two effective enumerations

of polynomial-time well-formed oracle QTMs and polynomials such that eachMi halts in time pi(n) on all inputsof length n independent of the choice of oracles. We build by stages a series of disjoint sets {Ai}i∈N and thendefine A =

⋃

i∈NAi.For convenience, set n−1 = 3 and A−1 = Ø. Consider stage i ∈ N. Let ni be the minimal integer such that

ni−1 < ni and 4pi(ni)4 < 2ni . In the case where Mi does not make nonadaptive queries to oracle A ∪ Ai−1 for

a certain set A ⊆ Σ2ni−2 ∪ Σ2ni−1, we set Ai as this A and go to the next stage. Hereafter, we assume thatMi makes nonadaptive queries to any oracle of the form A ∪ Ai−1 with A ⊆ Σ2ni−2 ∪ Σ2ni−1. Now, it suffices

to show the existence of a set A ⊆ Σ2ni−2 ∪ Σ2ni−1 such that χA(0ni0ni

A ) 6= ρA∪Ai−1

Mi(0ni). For readability, we

omit subscript i in what follows.

To draw a contradiction, we assume otherwise that, for any set A ⊆ Σ2n−2∪Σ2n−1, χA(0n0nA) = ρ

A∪Ai−1

M (0n).Let S be the set of all strings y ∈ Σn−1 such that some of the query lists of M on input 0n includes 0ny. Notethat S does not depend on the choice of oracles since M makes nonadaptive queries. For each y ∈ S, letqy be the sum of squared magnitudes of M ’s configurations conf in any superposition such that conf has aquery list containing word 0ny. Note that the size of each query list is at most p(n). It thus follows that∑

y∈S qy ≤ p(n)∑p(n)−1

j=1 ‖|φj〉‖2 ≤ p(n)2, where |φj〉 is the superposition of M ’s configurations at time j. By

our assumption, |S| = 2n−1 since, otherwise, we can choose an appropriate oracle A such that χA(0n0nA) 6=

ρA∪Ai−1

M (0n).Let y be any string in S and fix A such that y = 0nA. Moreover, let Ay be A except that χAy

(0ny) =

1 − χA(0ny). It follows by our assumption that ρ

A∪Ai−1

M (0n) = 1 − ρAy∪Ai−1

M (0n). By Lemma 6.7, since

|ρA∪Ai−1

M (0n) − ρAy∪Ai−1

M (0n)| = 1 as well as A△Ay = {0y}, we have∑p(n)−1

j=1 qj0ny(M,A ∪ Ai−1, 0n) ≥ 1

4p(n) .

Clearly, qj0ny(M,A ∪Ai−1, 0n) ≤ qy for each j since M makes nonadaptive queries. Thus,

∑p(n)−1j=1 qj0ny(M,A∪

Ai−1, 0n) ≤ qy ·p(n), which implies qy ≥ 1

4p(n)2 . This immediately draws the conclusion that |S| ≤ 4p(n)4. This

17

is a contradiction against |S| = 2n−1 since 4p(n)4 < 2n. ✷

By combining two oracle constructions in the proofs of Proposition 6.6 and Theorem 6.8, we obtain thefollowing corollary. This shows a quite different nature of adaptive and nonadaptive queries.

Corollary 6.9 For a certain oracle A, EQPA‖ * PA and PA * EQPA

‖ .

7 Applications to Decision Problems

The study of decision problems have been vigorously conducted in quantum complexity theory to bring infruitful results [4, 5, 15, 18, 37, 39]. These results address the strengths of quantum complexity classes. Thissection demonstrates two applications of quantum functions to decision problems and makes a bridge betweenquantum complexity classes and quantum function classes.

7.1 A Quantum Characterization of PP

Many quantum complexity classes lie within the probabilistic complexity class PP. This class PP is known tobe robust in a sense that it is characterized in many different fashions. For example, PP is characterized by twoGapP-functions; namely, PP equals the collection of sets A such that there exist two GapP-functions f and gsatisfying that, for every x, x ∈ A iff f(x) > g(x). We use a series of results in the previous sections to show anew characterization of PP in terms of #QPA and GapQPA.

Theorem 7.1 Let A be any subset of Σ∗. The following statements are equivalent.

1. A is in PP.

2. There exist two functions f, g ∈ #QPA such that, for every x, x ∈ A iff f(x) > g(x).

3. There exist two functions f, g ∈ GapQPA such that, for every x, x ∈ A iff f(x) > g(x).

Proof. 1 implies 3) Since A ∈ PP, there exist a polynomial-time deterministic TM M and a polynomial psuch that, for every x, x ∈ A iff |{y ∈ Σp(|x|) |M(x, y) = 1}| > 2p(|x|)−1. By examining the proof of Lemma 4.8,we notice the existence of a function f ∈ #QPA such that f(x) = f(x)2−p(|x|) for every x. Therefore, x ∈ A ifff(x) > 1

2 . Since #QPA ⊆ GapQPA, claim 3 follows.3 implies 2) Assume that there exist two functions f, g ∈ GapQPA such that, for every x, x ∈ A iff

f(x) > g(x). By Proposition 5.3, take four functions k0, k1, h0, h1 ∈ #QPA such that f = k0 − h0 andg = k1 − h1. Define f(x) = 1

2 (k0(x) + h1(x)) and g(x) =12 (k1(x) + h0(x)) for all x. By Lemma 4.9(2), f and g

are in #QP. It is also obvious that x ∈ A iff f(x) > g(x).2 implies 1) Assume that there exist two functions f and g in GapQPA such that, for every x, x ∈ A iff

f(x) > g(x). Define h(x) = f(x) − g(x) for all x. Note that h belongs to GapQPA by Proposition 5.3. ByTheorem 5.7(3), there exists a function k in GapP such that sign(h(x)) = sign(k(x)) for all x. This impliesthat x ∈ A iff k(x) > 0. Therefore, A is in PP. ✷

To see the robustness of PP, we consider the quantum analogue of PP.

Definition 7.2 Let PQPK be the collection of all sets A such that there exists a polynomial-time well-formedQTM with K-amplitudes satisfying: for every x, if x ∈ A then M accepts x with probability more than 1/2,and if x 6∈ A then M accepts x with probability at most 1/2.

Proposition 5.3 implies that any PQPK-set A has the form A = {x | f(x) > 0} for a certain GapQPK-function f if K ⊇ {0,±1,± 1

2}. Note that BQPK ⊆ PQPK . Thus, PQPC has uncountable cardinality since sodoes BQPC [1]. This concludes that PQPC 6= PP. However, Theorem 7.1 implies that, when K is limited to A,A falls in PP. Thus, we obtain the following.

Proposition 7.3 PQPA = PP and PQPC 6= PP.

With regard to PQPC, it is unknown even whether PQP

C= co-PQP

C. However, it seems difficult to show

the separation between PQPC and co-PQPC since this immediately implies EQPC 6= PP.

Lemma 7.4 PQPC6= co-PQP

Cimplies EQP

C6= PP.

18

Proof. We omit script C for readability. Assume that EQP = PP. Let A ∈ PQP. There exists a functionf ∈ GapQP satisfying that A = {x | f(x) > 0}. Note that f2 ∈ #QP. Consider the function g defined byg(x) = 1 if f2(x) = 0 and g(x) = −f(x) otherwise. Let B = {x | f2(x) 6= 0}. By the definition of NQP given

in [1], B belongs to NQP and thus, B is in PP. It is easy to show that g is in GapQPB[1] by making a single

query “x ∈?B” and then computing f(x) (if necessary). By our assumption, g ∈ GapQPPP[1] ⊆ GapQP. Notethat, for every x, x ∈ A implies g(x) < 0 and x 6∈ A implies g(x) > 0. This concludes that A is in co-PQP.Therefore, PQP ⊆ co-PQP. Symmetrically, we can show that co-PQP ⊆ PQP. ✷

7.2 Closure Properties of #QP

The closure properties of #P under various operators were studied in [29]. Let ◦ be any operator between twofunctions. A function class F is said to be closed under operator ◦ if, for every pair f, g in F , f ◦ g is also in F .The maximum operator max is defined by max{f, g} = λx.max{f(x), g(x)} and the minimum operator min isdefined by min{f, g} = λx.min{f(x), g(x)}. It is not known whether #QPK is closed under the maximum andminimum operators.

We introduce a new complexity class in connection to the closure property of #QPK under maximum andminimum operators. In what follows, we identify a string with a rational number (expressed as a pair of twointegers but not as a dyadic number).

Definition 7.5 A set A is in WQPK (wide QP) if there exist two functions f ∈ #QPK and g ∈ FEQPK withran(g) ⊆ (0, 1] ∩Q satisfying that f(x) = χA(x) · g(x) for every x.

Notice that we can replace #QPK in Definition 7.5 by GapQPK . Moreover, EQPK ⊆ WQPK ⊆ NQPK .Now, we show the following proposition.

Proposition 7.6 1. If EQPK = PQPK , then #QPK is closed under the maximum and the minimumoperators.

2. If #QPA is closed under the maximum and the minimum operators, then WQPA = PP.

Proof. 1) Assume that EQPK = PQPK . Let g and h be any functions in #QPK and set f(x) =max{g(x), h(x)} for all x. We define A = {x | g(x) > h(x)}, which clearly belongs to PQPK . By our as-

sumption, A is in EQPK . It is obvious that f belongs to #QPA[1]K , which is a subset of #QP

EQPK

K . Since

#QPEQPK

K = #QPK , f is in #QP. Therefore, #QPK is closed under maximum. Similarly, we can show thecase for the minimality.

2) Assume that #QPA is closed under max and min. By Proposition 7.3, A belongs to PQPA and thus,there exists a quantum function f ∈ #QPA such that, for every x, x ∈ A iff f(x) > 1/2. Let h = max{f, 12}.By our assumption, h is in #QPA. Note that λx.(h(x) − 1

2 ) is in GapQPA. Now, define k(x) = (h(x)− 12 )

2 forall x. By Squared Function Theorem, k is in #QPA. Since k ∈ #QPA, take an appropriate polynomial p suchthat k(x) ≥ 2−p(|x|) for all x (this fact is implicitly used in the proof of Theorem 5.7(3)). Finally, we definej = min{k, λx.2−p(|x|)}. Since λx.2−p(|x|) is in #QPA, j also belongs to #QPA by our closure assumption of#QPA. This j satisfies that, for every x, x ∈ A implies j(x) = 2−p(|x|) and x 6∈ A implies j(x) = 0. Thus, Abelongs to WQPA. ✷

As is shown below, WQP is in some sense a generalization of UP.

Lemma 7.7 UP ⊆ WQPK if K ⊇ {0,±1,± 12}.

Proof. Take any set A in UP. Note that χA ∈ #P. Lemma 4.8 guarantees the existence of two functionsg ∈ GapQPK and ℓ ∈ FP ∩ NΣ∗

satisfying χA(x) = ℓ(1|x|)g(x) for all x. Define h as follows: for everyx, h(x) = 1

ℓ(1|x|)if ℓ(1|x|) 6= 0 and h(x) = 1 otherwise. Thus, g(x) = χA(x)h(x) for all x. Note that

ran(h) ⊆ (0, 1] ∩Q. Clearly, h ∈ FP and thus, A belongs to WQPK . ✷

A relativized world exists where EQP and WQP are different classes.

Proposition 7.8 There exists a recursive oracle A such that EQPA 6= WQPA.

Proof. Note that UPA ⊆ WQPA for any oracle A because the proof of Lemma 7.7 relativizes. It suffices toshow that UPA * EQPA for a certain oracle A. This follows from the result of Fortnow and Rogers [18], who

19

proved that PA = BQPA 6= UPA ∩ co-UPA for a certain oracle A. ✷

Acknowledgments

The author is grateful to Harumichi Nishimura for his assistance in revising the conference version of thispaper and to Andy Yao and Yaoyun Shi for interesting discussion on quantum computation in his Princetondays.

References

[1] L. M. Adleman, J. DeMarrais, and M. A. Huang, Quantum computability, SIAM J. Comput., 26 (1997), 1524–1540.

[2] P. Benioff, The computer as a Physical system: A microscopic quantum mechanical Hamiltonian model of com-puters as represented by Turing machines, J. Stat. Phys., 22 (1980), 563–591.

[3] C. H. Bennett, Logical reversibility of computation, IBM J. Res. Develop. 17 (1973), 525–532.

[4] C. H. Bennett, E. Bernstein, G. Brassard, and U. Vazirani, Strengths and weaknesses of quantum computing,SIAM J. Comput., 26 (1997), 1510–1523.

[5] E. Bernstein and U. Vazirani, Quantum complexity theory, SIAM J. Comput., 26 (1997), 1411–1473.

[6] A. Berthiaume and G. Brassard, Oracle quantum computing, J. Modern Optics, 41 (1994), 2521–2535.

[7] G. Brassard, P. Høyer, and A. Tapp, Quantum counting, Proc. 25th International Colloquium on Automata,

Languages, and Programming, Lecture Notes in Computer Science, Vol.1443, pp.820–831, 1998.

[8] H. Buhrman and W. van Dam, Quantum bounded query complexity, LANL quant-ph/9903035, 1999.

[9] D. Deutsch, Quantum theory, the Church-Turing principle, and the universal quantum computer, Proc. Roy. Soc.London, Ser. A, 400 (1985), 97–117.

[10] D. Deutsch, Quantum computational networks, Proc. R. Soc. London, Ser. A, 425 (1989), 73–90.

[11] D. Deutsch and R. Jozsa, Rapid solution of problems by quantum computation, Proc. R. Soc. London, Ser. A, 439(1992), 553–558.

[12] D. Du and K. Ko, Theory of Computational Complexity, John Wiley & Sons, Inc., 2000.

[13] S. Fenner, L. Fortnow, S. Kurtz, and L. Li, An oracle builder’s toolkit, in Proc. 8th IEEE Conference on Structure

in Complexity Theory, pp.120–131, 1993.

[14] S. Fenner, L. Fortnow, and S. Kurtz, Gap-definable counting classes, J. Comput. and System Sci. 48 (1994),116–148.

[15] S. Fenner, F. Green, S. Homer, and R. Pruim, Determining acceptance probability for a quantum computation ishard for the polynomial hierarchy, Proceedings of the Royal Society of London, Ser.A, 455, 3953–3966, 1999.

[16] R. Feynman, Simulating Physics with computers, Intern. J. Theoret. Phys., 21 (1982), 467–488.

[17] L. Fortnow, Counting Complexity, in Complexity Theory Retrospective II, eds. L. Hemaspaandra and A. Selman,Springer, 1997.

[18] L. Fortnow and J. Rogers, Complexity limitations on quantum computation, in Proc. 13th IEEE Conference on

Computational Complexity, pp.202–209, 1998.

[19] M. H. Freedman, A. Kitaev, M. J. Larsen, and Z. Wang, Topological quantum computation. LANL quant-ph/0101025, 2001.

[20] L. Grover, A fast quantum mechanical algorithm for database search, in Proc. 28th ACM Symposium on Theory

of Computing, pp.212–219, 1996.

[21] B. Jenner and J. Toran, Computing functions with parallel queries to NP, Theoret. Comput. Sci., 141 (1995),175–193.

[22] A. Kitaev, Quantum computations: algorithms and error correction, Russian Math. Surveys, 52 (1997), 1191-1249.

[23] A. Kitaev, “Quantum NP”, public talk at the 2nd Worship on Algorithms in Quantum Information Processing,DePaul University, 1999.

[24] E. Knill, Quantum randomness and nondeterminism, Technical Report LAUR-96-2186, Los Alamos National Lab-oratory, 1996. LANL quant-ph/9610012.

[25] M. Krentel, The complexity of optimization problems, J. Comput. System Sci., 36 (1988), 490–509.

[26] J. Kobler, U. Schoning, and J. Toran, On counting and approximation, Acta Inf., 26 (1989), 363–379.

20

[27] H. Nishimura, On quantum computation with some restricted amplitudes, in Proc. 19th Annual Symposium on

Theoretical Aspect of Computer Science, Lecture Notes in Computer Science, Vol.2285, pp.311–322, 2002.

[28] H. Nishimura and M. Ozawa, Computational complexity of uniform quantum circuit families and quantum Turingmachines, Theoret. Comput. Sci., 276 (2002), 147–181.

[29] M. Ogiwara and L. Hemachandra, A complexity theory for feasible closure properties, J. Comput. System Sci., 46(1993), 295–325.

[30] M. Ozawa and H. Nishimura, Local transition functions of quantum Turing machines, RAIRO Theor. Inform.

Appl., 34 (2000), 379–402.

[31] A. Selman, X. Mei-Rui, and R. Book, Positive relativizations of complexity classes, SIAM J. Comput., 12 (1983),565–579.

[32] P. W. Shor, Fault-tolerant quantum computation, in Proc. 37th IEEE Symposium on Foundations of Computer

Science, pp.56–65, 1996.

[33] P. W. Shor, Polynomial-time algorithms for factorization and discrete logarithms on a quantum computer, SIAM

J. Comput., 26 (1997), 1484–1509.

[34] K. B. Stolarsky, Algebraic numbers and Diophantine Approximations, Marcel Dekker, 1974.

[35] L. G. Valiant, The complexity of computing the permanent, Theoret. Comput. Sci., 8 (1979), 410–421.

[36] K. W. Wagner, Bounded query classes, SIAM J. Comput., 19 (1990), 833–846.

[37] J. Watrous, Succinct quantum proofs for properties of finite groups, in Proc. 41st Annual Symposium on Founda-

tions of Computer Science, pp.537–546, 2000.

[38] T. Yamakami, A foundation of programming a multi-tape quantum Turing machine, in Proc. 24th International

Symposium on Mathematical Foundations of Computer Science, Lecture Notes in Computer Science, Vol.1672,pp.430–441, 1999.

[39] T. Yamakami and A. C. Yao, NQPC=co-C=P, Inform. Process. Lett., 71 (1999), 63–69.

[40] A. C. Yao, Quantum circuit complexity, in Proc. 34th IEEE Symposium on Foundation of Computer Science,

pp.352–361, 1993.

21