Professional Documents
Culture Documents
Abstract. In this paper, we present a quantum Bernoulli noises approach to quantum walks
on hypercubes. We first obtain an alternative description of a general hypercube and then,
based on the alternative description, we find that the operators ∂k∗ + ∂k behave actually as the
shift operators, where ∂k and ∂k∗ are the annihilation and creation operators acting on Bernoulli
functionals, respectively. With the above operators as the shift operators on the position space,
we introduce a discrete-time quantum walk model on a general hypercube and obtain an explicit
formula for calculating its probability distribution at any time. We also establish two limit
theorems showing that the averaged probability distribution of the walk even converges to the
uniform probability distribution. Finally, we show that the walk produces the uniform measure as
its stationary measure on the hypercube provided its initial state satisfies some mild conditions.
Some other results are also proven.
Keywords. Quantum walk; Quantum Bernoulli noises; Quantum probability; Stationary mea-
sure; Uniform measure
1 Introduction
As is known, random walks have been extensively used in modeling various physical processes
and in developing random algorithms. Recent two decades have witnessed great attention paid
to quantum walks, which are quantum analogs of random walks (see, e.g. [2, 5, 9, 13, 19] and
references therein). Quantum walks are also known as quantum random walks, and have found
wide application in quantum computation. Due to the quantum interference effects, quantum
walks greatly outperform random walks at certain computational tasks, and moreover it has
turned out that quantum walks constitute universal models of quantum computation [13].
From a perspective of mathematical physics, a quantum walk can be viewed as a quantum
dynamical system (QDS) driven by a unitary operator on the tensor space of an l2 -space associated
with a graph and an auxiliary Hilbert space. For instance, the well-known Hadamard walk can be
thought of a QDS driven by a unitary operator on the tensor space l2 (Z) ⊗ C2 , where the integer
lattice Z is essentially a graph. There are two basic types of quantum walks in the literature: the
1
type of discrete-time quantum walks and the type of continuous-times quantum walks. In this
paper, we focus on the former.
Hypercubes are a special class of regular graphs, which play an important role in computer
science. Recently there has been much interest in quantum walks on hypercubes. In their survey
paper, Aharonov et al [1] presented the ground for a theory of quantum walks on graphs including
hypercubes. Moore and Russell [12] studied the instantaneous mixing time of quantum walks on
the n-dimensional hypercube, while Kempe [6] investigated the hitting time of these walks. Alagić
and Russell [3] considered decoherence in quantum walks on hypercubes. There are many other
works concerning quantum walks on hypercubes in the literature (see, e.g. [10, 11]).
However, little attention has been paid to the problem of finding a way to produce the uniform
measure on a general hypercube via a quantum walk, which is interesting as is pointed out in [8].
In this paper, as one of our main purposes, we would like to provide a solution to such problem.
Quantum Bernoulli noises (QBN) are annihilation and creation operators acting on Bernoulli
functionals, which satisfy the canonical anti-commutation relations (CAR) in equal time. It has
turned out that QBN can play an important role in describing the irreversible evolution of an
open quantum system (especially a Fermi system) [16, 17]. In 2016, Wang and Ye introduced
a quantum walk model on the 1-dimensional integer lattice Z in terms of QBN and showed its
strong decoherence property [18].
In the present paper, as our another main purpose, we would also like to present a QBN
approach to quantum walks on hypercubes. Our main work is as follows.
• We introduce a notion of coin operator system on a general Hilbert space and obtain a
necessary and sufficient condition for an operator system to be a coin operator system.
• We establish two limit theorems showing that the averaged probability distribution of the
walk even converges to the uniform probability distribution on the vertex set of the (n + 1)-
dimensional hypercube.
• Finally, we prove that the walk produces the uniform measure as its stationary measure on
the (n + 1)-dimensional hypercube provided its initial state satisfies some mild conditions.
Our work shows that QBN can provide a way to produce the uniform measure on a general
hypercube via a quantum walk. Moreover, our work also suggests that, for a quantum walk on a
hypercube, its “components in the coin space” can be the determining factors of its probability
distributions and evolution behavior.
2
The paper is organized as follows. In Section 2, we recall some necessary notions and facts
about quantum Bernoulli noises. Our main work then lies in Section 3, where we give an alter-
native description of the (n + 1)-dimensional hypercube (Subsection 3.1), characterize our coin
operator systems (Subsection 3.2), define our quantum walk model (Subsection 3.3), examine the
probability distribution of the walk (Subsection 3.4), prove that the walk can produce the uniform
measure as its stationary measure (Subsection 3.5), and offer some examples (Subsection 3.6).
Finally in Section 4, we make some conclusion remarks.
Let F be the σ-field on Ω generated by the sequence (ζn )n≥0 , and (pn )n≥0 a given sequence of
positive numbers with the property that 0 < pn < 1 for all n ≥ 0. Then there exists a unique
probability measure P on the measurable space (Ω, F ) such that
k
Y 1+ǫj 1−ǫj
P ◦ (ζn1 , ζn2 , · · · , ζnk )−1 (ǫ1 , ǫ2 , · · · , ǫk ) = pnj2 (1 − pnj ) 2 (2.2)
j=1
We denote by h·, ·ih the usual inner product of the space h, and by k · kh the corresponding norm.
It is known [14, 15] that Z has the chaotic representation property, which implies that the system
Z = {Zσ | σ ∈ Γ} forms an orthonormal basis (ONB) for h, where Z∅ = 1 and
Y
Zσ = Zj , σ ∈ Γ, σ 6= ∅, (2.5)
j∈σ
3
where Γ = {σ ⊂ N | #σ < ∞}. The ONB Z = {Zσ | σ ∈ Γ} is known as the canonical ONB for
h. Clearly h is separable and infinite-dimensional as a complex Hilbert space.
It can be shown that [15], for each k ≥ 0, there exists a bounded operator ∂k on h such that
where ∂k∗ denotes the adjoint of ∂k , σ \ k = σ \ {k}, σ ∪ k = σ ∪ {k} and 1σ (k) the indicator of σ
as a subset of N.
The operators ∂k and ∂k∗ are usually known as the annihilation and creation operators acting
on Bernoulli functionals, respectively. And the family {∂k , ∂k∗ }k≥0 is referred to as quantum
Bernoulli noises (QBN).
A typical property of QBN is that they satisfy the canonical anti-commutation relations (CAR)
in equal-time [15]. More specifically, for k, l ≥ 0, it holds true that
and
∂k ∂k = ∂k∗ ∂k∗ = 0, ∂k ∂k∗ + ∂k∗ ∂k = I, (2.8)
4
The n-dimensional hypercube is then the graph (V (n) , E (n) ) with V (n) being the vertex set and
E (n) being the edge set, where E (n) is given by
E (n) = {x, y} | x, y ∈ V (n) , |x − y| = 1 (3.2)
It is well known that the n-dimensional hypercube (V (n) , E (n) ) is a regular graph and its degree
is exactly n. Moreover, it has 2n vertices and n × 2n−1 edges (see, e.g. [19] for more information
about a hypercube).
We next give an alternative description of a hypercube. For a nonnegative integer n ≥ 0, we
write Nn = {0, 1, 2, · · · , n} and denote by Γn its the power set, namely
Γn = {σ | σ ⊂ Nn }. (3.4)
For example, Γ1 = ∅, {0}, {1}, {0, 1} . Recall that, for σ ∈ Γn and k ∈ Nn , we use σ \ k to
mean σ \ {k} for brevity. Similarly, we use σ ∪ k, σ ∩ k, etc.
Definition 3.1. For nonnegative integer n ≥ 0, we denote by (Γn , En ) the graph with Γn being
the vertex set and En being the edge set, where En given by
En = {σ, τ } | σ, τ ∈ Γn , #(σ △ τ ) = 1 , (3.5)
The following proposition shows that the graph (Γn , En ) actually belongs to the category of
hypercubes.
Proposition 3.1. Let n ≥ 0 be a nonnegative integer. Then the graph (Γn , En ) is isomorphic to
the (n + 1)-dimensional hypercube (V (n+1) , E (n+1) ).
where 1σ (x) denotes the indicator of σ as a subset of Nn . It is easy to see that J is a bijection
from Γn to V (n+1) . Let σ, τ ∈ Γn be such that σ ∼ τ . Then #(σ △ τ ) = 1, which implies that
σ ⊂ τ with #(τ \ σ) = 1 or τ ⊂ σ with #(σ \ τ ) = 1. In the case of σ ⊂ τ with #(τ \ σ) = 1, we
have
n−1
X X
|J(σ) − J(τ )| = |1σ (k) − 1τ (k)| = 1τ (k) = 1,
k=0 k∈τ \σ
which means J(σ) ∼ J(τ ). Similarly, we can also get J(σ) ∼ J(τ ) in the case of τ ⊂ σ with
#(σ \ τ ) = 1.
Consider the graph (Γn , En ), where n ≥ 0. For vertices σ and τ ∈ Γn , as usual we use σ ∼ τ
to mean that σ and τ are adjacent, namely {σ, τ } ∈ En .
5
Proposition 3.2. Let σ, τ ∈ Γn be vertices in the graph (Γn , En ). Then, σ ∼ τ if and only if
there exists a unique k ∈ Nn such that k ∈ σ with σ \ k = τ or k ∈
/ σ with σ ∪ k = τ .
Proof. Let σ ∼ τ . Then #(σ △τ ) = 1, which implies that #(σ \ τ ) = 1 with τ ⊂ σ or #(τ \ σ) = 1
with σ ⊂ τ , which implies that there exists a unique k ∈ Nn such that k ∈ σ with σ \ k = τ or
k∈
/ σ with σ ∪ k = τ .
Conversely, if there exists a unique k ∈ Nn such that k ∈ σ with σ \ k = τ or k ∈
/ σ with
σ ∪ k = τ , then σ △ τ = {k}, hence #(σ △ τ ) = 1, which means {σ, τ } ∈ En , namely σ ∼ τ .
Corollary 3.3. Let σ ∈ Γn be a vertex in the graph (Γn , En ). Then, for k ∈ Nn , one has:
σ ∼ (σ \ k) when k ∈ σ; σ ∼ (σ ∪ k) when k ∈
/ σ.
where I denotes the the identity operator on K. Based on these equalities, one can further
know that both Ck∗ Ck and Ck Ck∗ are projection operators on K for each k with 0 ≤ k ≤ n. To
characterize a coin operator system, let us first introduce a notion as follows.
A system {Pk | 0 ≤ k ≤ n} of projection operators on the space K is called a resolution of the
P
identity if nk=0 Pk = I and Pj Pk = 0, j 6= k, 0 ≤ j, k ≤ n.
In terms of a resolution of the identity as well as a unitary operator, the next theorem provides
a necessary and sufficient condition for a system of bounded operators to be a coin operator
system.
6
(2) There exist a unitary operator U and a resolution of the identity {Pk | 0 ≤ k ≤ n} on K
such that Ck = Pk U , 0 ≤ k ≤ n.
P
Proof. “(1) ⇒ (2)”. Let U = nk=0 Ck . Then, by the definition, U is a unitary operator. Now
consider the system {Pk | 0 ≤ k ≤ n} of projection operators, where Pk = Ck Ck∗ . Clearly, it is a
resolution of the identity. For 0 ≤ k ≤ n, a straightforward calculation yields
n
X n
X
Pk U = Ck Ck∗ Cj = Ck Ck∗ Ck = Ck I − Cj∗ Cj = Ck .
j=0 j=0,j6=k
Pn Pn
“(2) ⇒ (1)”. It is easy to see that k=0 Ck = U , which means that k=0 Ck is a unitary
operator. For 0 ≤ j, k ≤ n with j 6= k, using Pj Pk = 0 gives
Recall that, for τ ∈ Γn , 1τ (k) stands for the indicator of the set τ , which allows us to define
a function ετ (k) on Nn as
ετ (k) = 2 × 1τ (k) − 1, k ∈ Nn . (3.8)
Proof. Let τ ∈ Γn be given. Using properties of the coin operator system C, we have
∗ X
n X
n Xn
2 n
X
Uτ(C) Uτ(C) = ετ (k)Ck∗ ετ (k)Ck = ετ (k) Ck∗ Ck = Ck∗ Ck = I,
k=0 k=0 k=0 k=0
hn = span{Zσ | σ ∈ Γn }. (3.10)
7
Note that Γn ⊂ Γ and #(Γn ) = 2n+1 , which implies that hn is a 2n+1 -dimensional subspace of
h, hence a closed subspace. In other words, hn itself is a Hilbert space with the inner product
h·, ·ih . It can be shown that, for all k ∈ Nn , both the annihilation operator ∂k and the creation
operator ∂k∗ leave hn invariant. This means that, for all k ∈ Nn , ∂k and ∂k∗ can be viewed as the
annihilation and creation operators on hn , respectively.
Theorem 3.6. Consider the graph (Γn , En ). Let σ, τ ∈ Γn be its vertices. Then σ ∼ τ if and
only if there exists a unique k ∈ Nn such that
Proof. Let τ ∈ Γn . Then, by Proposition 3.2, there exists a unique k ∈ Nn such that k ∈ σ with
τ = σ \ k or k ∈
/ σ with τ = σ ∪ k, which implies that
On the other hand, it follows from properties of ∂k and ∂k∗ (see (2.6) for details) that
Thus (∂k∗ + ∂k )Zσ = Zτ . Now suppose that (∂k∗ + ∂k )Zσ = Zτ . Then, we have
Zτ = (1 − 1σ (k))Zσ∪k + 1σ (k)Zσ\k ,
Remark 3.1. Let σ be a vertex in the graph (Γn , En ) and k ∈ Nn . Then, using properties of ∂k
and ∂k∗ , we have (
Zσ\k , k ∈ σ;
(∂k∗ + ∂k )Zσ =
Zσ∪k , k ∈
/ σ.
On the hand, by Corollary 3.3, σ ∼ (σ \ k) when k ∈ σ; σ ∼ (σ ∪ k) when k ∈
/ σ. Therefore, the
operator (∂k∗ + ∂k ) on hn behaves actually as a shift operator.
As mentioned above, the Hilbert space K will serve as the coin space of our quantum walk
model. In what follows, we denote by h·, ·i and k · k the inner product and norm in the tensor
space hn ⊗ K. We mention that {Zσ ⊗ ej | σ ∈ Γn , 0 ≤ j ≤ dK − 1} is an orthonormal basis for
hn ⊗ K.
Theorem 3.7. Let C = {Ck | 0 ≤ k ≤ n} be a coin operator system on the coin space K and
write
n
X
WC = (∂k∗ + ∂k ) ⊗ Ck . (3.12)
k=0
8
Proof. It follows from (2.7) and (2.8) that (∂k∗ + ∂k )2 = Ihn , 0 ≤ k ≤ n, where Ihn denotes the
identity operator on hn . Thus, by Definition 3.2 and properties of coin operator systems, we have
X
n X
n
WC ∗ WC = (∂k∗ + ∂k ) ⊗ Ck∗ (∂k∗ + ∂k ) ⊗ Ck
k=0 k=0
n
X
= (∂k∗ + ∂k )2 ⊗ Ck∗ Ck
k=0
Xn
= Ihn ⊗ Ck∗ Ck
k=0
= Ihn ⊗ IK
= I,
Definition 3.4. The quantum walk on the graph (Γn , En ) with WC being the evolution operator
is the discrete-time quantum walk that admits the following features:
• The walk takes hn ⊗ K as its state space, where hn describes the position information of
the walk, while K describes its internal degrees of freedom;
• The states of the walk are represented by unit vectors in hn ⊗ K and the time evolution of
the walk is governed by the equation
Φt+1 = WC Φt , t ≥ 0, (3.13)
where Φt denotes the state of the walk at time t ≥ 0, especially Φ0 denotes the initial state
of the walk;
K −1
dX
Pt (σ | Φ0 ) = |hZσ ⊗ ej , Φt i|2 , (3.14)
j=0
Conventionally, hn and K are known as the position space and coin space of the walk, respec-
tively, while the function σ 7→ Pt (σ | Φ0 ) on Γn is called the probability distribution of the walk
at time t ≥ 0, which usually depends on the initial state Φ0 and the evolution operator WC .
Remark 3.2. In what follows, we simply use WC to indicate the quantum walk on the graph
(Γn , En ) with WC being the evolution operator. In other words, when we say the walk WC , we
just mean the quantum walk on the graph (Γn , En ) with WC being the evolution operator.
9
3.4 Probability distribution of the quantum walk
In the present subsection, we examine properties of the walk WC from a perspective of probability
distribution. We continue to use the assumptions made in the previous subsection. Unless
otherwise specified, we always use Φ0 to mean the initial state of the walk WC .
Recall that, for σ ∈ Γn , the function εσ (·) on Nn is defined by εσ (k) = 2 × 1σ (k) − 1, where
1σ (k) is the indicator of σ. With these functions, we introduce the following vectors in the
position space hn
X Y
bσ = √ 1
Z εσ (k) Zτ , σ ∈ Γn , (3.15)
2n+1 τ ∈Γn k∈τ
Q
where k∈τ εσ (k) = 1 if τ = ∅. For σ, τ ∈ Γn , one can easily obtain the next useful formula
1 Y (−1)#(σ\τ )
hZσ , Zbτ ih = √ ετ (k) = √ , (3.16)
2n+1 k∈σ
2n+1
where #(σ \ τ ) denotes the cardinality of the set σ \ τ . Additionally, we also introduce the
following operators on the position space hn
n
Y
Aσ = I + εσ (k)(∂k∗ + ∂k ) , σ ∈ Γn , (3.17)
k=0
bσ =
Proposition 3.8. Let σ ∈ Γn be given. Then Z √ 1 Aσ Z∅ , where Z∅ is the basis vector in
2n+1
the canonical ONB for hn .
Proof. Consider the operator system Aσ = I + εσ (k)(∂k∗ + ∂k ) | 0 ≤ k ≤ n associated with
σ. According to (2.7), any two operators in Aσ are commutative. Thus, by a straightforward
calculation, we can get
X Y Y
Aσ = εσ (k) (∂k∗ + ∂k ) , (3.18)
τ ∈Γn k∈τ k∈τ
Q ∗
where k∈τ (∂k + ∂k ) = I when τ = ∅. On the other hand, for each τ ∈ Γn , by using (2.6) and
the induction method we have Y
(∂k∗ + ∂k ) Z∅ = Zτ ,
k∈τ
√
which together with (3.18) implies Aσ Z∅ = bσ , equivalently Z
2n+1 Z bσ = √ 1 Aσ Z∅ .
2n+1
Proposition 3.9. Let σ ∈ Γn be given. Then, for all k ∈ Nn , the following formula holds true
10
which, together with the commutativity of the operator system Aσ , gives
n
Y
(∂k∗ + ∂k )Aσ = (∂k∗ + ∂k ) I + εσ (k)(∂k∗ + ∂k ) I + εσ (j)(∂j∗ + ∂j )
j=0,j6=k
n
Y
= εσ (k) I + εσ (k)(∂k∗ + ∂k ) I + εσ (j)(∂j∗ + ∂j )
j=0,j6=k
n
Y
= εσ (k) I + εσ (j)(∂j∗ + ∂j )
j=0
= εσ (k)Aσ .
bσ = εσ (k)Z
(∂k∗ + ∂k )Z bσ , k ∈ Nn , σ ∈ Γ n . (3.20)
Proof. Let σ, γ ∈ Γn be such that σ 6= γ. Then, there exists k ∈ Nn such that εσ (k) 6= εγ (k),
which implies that εσ (k)εγ (k) = −1 and εσ (k) + εγ (k) = 0. Thus, we have
I + εσ (k)(∂k∗ + ∂k ) I + εγ (k)(∂k∗ + ∂k ) = (1 + εσ (k)εγ (k))I + (εσ (k) + εγ (k))(∂k∗ + ∂k ) = 0,
which, together with (3.17) as well as the commutativity of both Aσ and Aγ , yields
Aσ Aγ = I + εσ (k)(∂k∗ + ∂k ) I + εγ (k)(∂k∗ + ∂k )
n
Y n
Y
× I + εσ (j)(∂j∗ + ∂j ) I + εγ (j)(∂j∗ + ∂j )
j=0,j6=k j=0,j6=k
= 0,
which, together with Proposition 3.8 and the self-adjoint property of Aσ , further gives
bσ , Z
bγ ih = 1 1
hZ hAσ Z∅ , Aγ Z∅ ih = n+1 hZ∅ , Aσ Aγ Z∅ ih = 0.
2n+1 2
It follows directly from (3.15) that
1 X
Y
2
1 X
bσ , Z
hZ bσ ih = kZ
bσ k2 =
εσ (k) Zτ
= 1 = 1.
h
2n+1 h 2n+1
τ ∈Γn k∈τ τ ∈Γn
Therefore, the system Zbσ | σ ∈ Γn is an orthonormal system in hn . This, together with the fact
that # Zbσ | σ ∈ Γn = dim hn = 2n+1 , means that Z bσ | σ ∈ Γn actually forms an orthonormal
basis for hn . Finally, for k ∈ Nn and σ ∈ Γn , using Proposition 3.8 and Proposition 3.9 leads to
As an immediate consequence of Theorem 3.10, we have the following corollary, which shows
that all the shift operators ∂k∗ + ∂k , k ∈ Nn , have a common fixed point in the unit sphere of hn .
11
bNn . Then (∂ ∗ + ∂k )ξ ⋆ = ξ ⋆ , ∀ k ∈ Nn .
Corollary 3.11. Write ξ ⋆ = Z k
Proof. Let k ∈ Nn be given. Then εNn (k) = 2×1Nn (k)−1 = 1, which together with Theorem 3.10
gives
bNn = ξ ⋆ ,
(∂k∗ + ∂k )ξ ⋆ = (∂k∗ + ∂k )ZbNn = εNn (k)ZbNn = Z
bσ ⊗ ej | σ ∈ Γn , 0 ≤ j ≤ dK − 1}
Remark 3.3. According to Theorem 3.10, the vector system {Z
forms an orthonormal basis for the tensor space hn ⊗ K. Thus, each Φ ∈ hn ⊗ K has an expansion
of the following form
X
Φ= bσ ⊗ uσ ,
Z (3.21)
σ∈Γn
PdK −1 b
where uσ = j=0 hZσ ⊗ ej , Φiej .
(C)
Recall that, for τ ∈ Γn , the ετ -weighted sum Uτ of the coin operator system C is a unitary
operator on the coin space K (see Definition 3.3 and Proposition 3.5). The next result unveils a
(C)
link between the action of the evolution operator WC and that of Uτ .
bτ ⊗ Uτ(C) u .
bτ ⊗ u) = Z
Theorem 3.12. If τ ∈ Γn and u ∈ K, then WC (Z
bτ ⊗ u) = Zbτ ⊗ Uτ(C) u .
which, together with Definition 3.3, gives WC (Z
We are now ready to establish an explicit formula for calculating the probability distribution
of the walk at any time t ≥ 0.
Theorem 3.13. For any t ≥ 0, the probability distribution of the walk WC at time t has a
representation of the following form
1
X #(σ\τ )
(C) t
2
Pt (σ | Φ0 ) = n+1
(−1) U τ u τ
, σ ∈ Γn , (3.22)
2 K
τ ∈Γn
PdK −1
where uτ = bτ ⊗ ej , Φ0 iej for τ ∈ Γn and Φ0 is the initial state of the walk.
hZ
j=0
Proof. According to Remark 3.3, the initial state Φ0 has an expansion of the following form
X
Φ0 = bτ ⊗ uτ ,
Z (3.23)
τ ∈Γn
which together with Theorem 3.12 implies that, at time t ≥ 0, the walk’s state Φt can be expressed
as
X X t
Φt = WC t Φ0 = bτ ⊗ uτ ) =
WC t (Z Zbτ ⊗ Uτ(C) uτ .
τ ∈Γn τ ∈Γn
12
Let σ ∈ Γn be given. Then, using the above expression, we have
K −1
dX
Pt (σ | Φ0 ) = |hZσ ⊗ ej , Φt i|2
j=0
K −1
dX
X
2
= bτ ⊗ U (C) t uτ
Zσ ⊗ ej , Z τ
j=0 τ ∈Γn
K −1
dX
X
2
= bτ ih ej , U (C) t uτ
hZσ , Z τ K
j=0 τ ∈Γn
K −1 D
dX
X E 2
= ej , bτ ih Uτ(C) t uτ
hZσ , Z
K
j=0 τ ∈Γn
X
2
bτ ih Uτ(C) t uτ
=
hZσ , Z
,
K
τ ∈Γn
bτ ih =
which together with hZσ , Z √ 1 (−1)#(σ\τ ) yields
2n+1
X
2 1
t
bτ ih Uτ(C) t uτ
X
2
Pt (σ | Φ0 ) =
hZσ , Z
=
(−1)#(σ\τ ) Uτ(C) uτ
.
K 2n+1 K
τ ∈Γn τ ∈Γn
Definition 3.5. For time T ≥ 1, the T -averaged probability distribution of the walk WC is
defined as
T −1
1 X
P T (σ | Φ0 ) = Pt (σ | Φ0 ), σ ∈ Γn , (3.24)
T t=0
where Pt (σ | Φ0 ) is the probability of finding the walker on vertex σ at time t and Φ0 is the initial
state.
The next result establishes a limit theorem for the walk WC , which shows that under some
mild conditions the walk has a limit averaged probability distribution as time T goes to ∞.
PdK −1
Theorem 3.14. Suppose that, for each τ ∈ Γn , uτ = bτ ⊗ ej , Φ0 iej is an eigenvector of
hZ
j=0
(C)
Uτ with bτ being the corresponding eigenvalue. Then, on all vertex σ ∈ Γn , one has
1 h X i
lim P T (σ | Φ0 ) = n+1 1 + (−1)#(σ\τ1 )+#(σ\τ2 ) huτ1 , uτ2 iK , (3.25)
T →∞ 2
(τ1 ,τ2 )
P
where (τ1 ,τ2 ) means to sum over the set (τ1 , τ2 ) ∈ Γn × Γn | τ1 6= τ2 , bτ1 = bτ2 .
13
P
Let σ ∈ Γn be given. Then, by Theorem 3.13 as well as the equality τ ∈Γn kuτ k2K = 1, we have
1
X #(σ\τ ) t
2
Pt (σ | Φ0 ) = n+1
(−1) b u
τ τ
2 K
τ ∈Γn
1 h X X X i t
= + + (−1)#(σ\τ1 )+#(σ\τ2 ) bτ1 btτ2 huτ1 , uτ2 iK
2n+1
(τ1 ,τ2 )∈△1 (τ1 ,τ2 )∈△2 (τ1 ,τ2 )∈△3
1 h X X
= kuτ k2K + (−1)#(σ\τ1 )+#(σ\τ2 ) huτ1 , uτ2 iK
2n+1
τ ∈Γn (τ1 ,τ2 )∈△2
X t
i
+ (−1)#(σ\τ1 )+#(σ\τ2 ) bτ1 btτ2 huτ1 , uτ2 iK
(τ1 ,τ2 )∈△3
1 h X i
= 1+ (−1)#(σ\τ1 )+#(σ\τ2 ) huτ1 , uτ2 iK
2n+1
(τ1 ,τ2 )∈△2
1 X t
+ (−1)#(σ\τ1 )+#(σ\τ2 ) bτ1 btτ2 huτ1 , uτ2 iK ,
2n+1
(τ1 ,τ2 )∈△3
X T −1
1 1 X t t
+ (−1)#(σ\τ1 )+#(σ\τ2 ) huτ1 , uτ2 iK bτ b ,
2n+1 T t=0 1 τ2
(τ1 ,τ2 )∈△3
Thus
1 h X i
lim P T (σ | Φ0 ) = 1+ (−1)#(σ\τ1 )+#(σ\τ2 ) huτ1 , uτ2 iK ,
T →∞ 2n+1
(τ1 ,τ2 )∈△2
(C)
which implies that uτ is an eigenvector of Uτ with bτ being the corresponding eigenvalue. This
shows that the assumptions made in Theorem 3.14 can be satisfied.
14
As an immediate consequence of Theorem 3.14, the next theorem shows that the limit averaged
probability distribution of the walk WC even coincides with the uniform probability distribution
on Γn .
PdK −1 b
Theorem 3.15. Suppose that, for each τ ∈ Γn , uτ = j=0 hZτ ⊗ ej , Φ0 iej is an eigenvector of
(C)
Uτ with bτ being the corresponding eigenvalue, and moreover bτ1 6= bτ2 when τ1 τ2 ∈ Γn with
τ1 6= τ2 . Then, it holds true that
1
lim P T (σ | Φ0 ) = , ∀ σ ∈ Γn . (3.26)
T →∞ 2n+1
Proof. It follows from the conditions that (τ1 , τ2 ) ∈ Γn × Γn | τ1 6= τ2 , bτ1 = bτ2 = ∅, which
together with Theorem 3.14 implies (3.26).
(C)
Remark 3.5. The unitary operators Uτ | τ ∈ Γn can be viewed as the evolution operator’s
“components in the coin space”. Similarly, the vectors {uτ | τ ∈ Γn can be viewed as the
initial state’s “components in the coin space”. Theorem 3.13, Theorem 3.14 and Theorem 3.15
suggest that the walk’s “components in the coin space” are actually the determining factors of
its probability distributions.
1
κn (σ) = , σ ∈ Γn . (3.27)
2n+1
Definition 3.6. A measure ν on Γn is called a stationary measure of the walk WC if there exists
some unit vector Ψ ∈ hn ⊗ K such that
K −1
dX
ν(σ) = |hZσ ⊗ ej , WC t Ψi|2 , ∀ σ ∈ Γn , ∀ t ≥ 0. (3.28)
j=0
Furthermore, if the stationary measure ν coincides with the uniform measure κn , then we say
that the walk WC with Φ0 = Ψ produces the uniform measure as its stationary measure on Γn .
bγ ⊗ u
Theorem 3.16. For all γ ∈ Γn and all u ∈ K with kuk2K = 1, the walk WC with Φ0 = Z
produces the uniform measure as its stationary measure on Γn . In particular, the uniform measure
κn on Γn is a stationary measure of the walk WC .
15
Proof. Let γ ∈ Γn and u ∈ K with kuk2K = 1 be given. Then, for all σ ∈ Γn and t ≥ 0, by
Theorem 3.12 we have
K −1
dX K −1
dX
Zσ ⊗ ej , WC t (Zbγ ⊗ u) 2 = bγ ⊗ U (C) t ui|2
|hZσ ⊗ ej , Z γ
j=0 j=0
K −1
dX
= bγ ih ej , U (C) t u 2
hZσ , Z γ K
j=0
bγ ih |2
(Uγ(C) t u
2
= |hZσ , Z K
bγ ih |2 kuk2
= |hZσ , Z K
bγ ih |2 ,
= |hZσ , Z
bγ ih = (−1)#(σ\γ)
which together with hZσ , Z √
2n+1
gives
K −1
dX
1
bγ ⊗ u) 2 = |hZσ , Z
Zσ ⊗ ej , WC t (Z bγ ih |2 = .
j=0
2n+1
Then the walk WC with Φ0 = Ψ produces the uniform measure as its stationary measure on Γn .
Proof. Let Φ0 = Ψ and denote by bτ the eigenvalue corresponding to the eigenvector uτ . Then,
for all σ ∈ Γn and t ≥ 0, by Theorem 3.13 we immediately have
K −1
1
2
dX
X
1 X
|hZσ ⊗ ej , WC t Ψi|2 = Pt (σ | Φ0 ) = n+1
(−1)#(σ\τ ) t
b τ u τ
= kuτ k2K ,
2 K 2n+1
j=0 τ ∈Γn τ ∈Γn
X K −1
X dX
kuτ k2K = |hZbτ ⊗ ej , Ψi|2 = kΨk2 = 1
τ ∈Γn τ ∈Γn j=0
yields
K −1
dX X
1 1
|hZσ ⊗ ej , WC t Ψi|2 = kuτ k2K = .
2n+1 2n+1
j=0 τ ∈Γn
This exactly means that the walk WC with Φ0 = Ψ produces the uniform measure as its stationary
measure on Γn .
16
3.6 Examples
In the final subsection, we offer some examples to show that the assumptions made in Theo-
rem 3.15 and Theorem 3.17 can be satisfied.
Consider the graph (Γ1 , E1 ), which is actually isomorphic to the 2-dimensional hypercube. In
this case, the position space of the walk is just h1 = span Zσ | σ ∈ Γ1 . Note that
Γ1 = ∅, {0}, {1}, {0, 1} . (3.29)
Example 3.1. Take C2 as the coin space, namely K = C2 . Then, one has a coin operator system
C = {C0 , C1 } on K = C2 , where
! !
0 1 0 0
C0 = , C1 = . (3.30)
0 0 1 0
In this case, the state space of the walk is h1 ⊗ K = h1 ⊗ C2 , while the evolution operator WC takes
the form
WC = (∂0∗ + ∂0 ) ⊗ C0 + (∂1∗ + ∂1 ) ⊗ C1 . (3.31)
According to Definition 3.3, the weighted sums of the coin operator system C are
! ! ! !
(C) 0 −1 (C) 0 1 (C) 0 −1 (C) 0 1
U∅ = , U{0} = , U{1} = , U{0,1} = .
−1 0 −1 0 1 0 1 0
By careful calculations, one can immediately get spectrums of these unitary operators (matrices):
− 12
X
Φ0 = M 0 bγ ⊗ vγ ,
Z
γ∈Γ1
P
where M0 = kvγ k2C2 . Then, Φ0 is a unit vector in h1 ⊗ K = h1 ⊗ C2 , hence can serve as
γ∈Γ1
an initial state. Furthermore, for each τ ∈ Γ1 = ∅, {0}, {1}, {0, 1} , we have
K −1
dX
−1
uτ = hZbτ ⊗ ej , Φ0 iej = M0 2 vτ ,
j=0
(C) (C)
which implies that Uτ uτ = bτ uτ , namely uτ is an eigenvector of Uτ with bτ being the cor-
responding eigenvalue. Clearly, bτ1 6= bτ2 for τ1 , τ1 ∈ Γ1 with τ1 6= τ2 . This shows that the
assumptions made in Theorem 3.15 can be satisfied.
17
Example 3.2. Take C4 as the coin space, namely K = C4 . Then, one also has a coin operator
system C = {C0 , C1 } on K = C4 , where
1 0 0 0 0 0 0 0
0 1 0 0 0 0 0
0
C0 =
0 0
, C1 = . (3.32)
0 0
0
0 −1 0
0 0 0 0 0 0 0 −1
Thus, the state space of the walk is h1 ⊗ K = h1 ⊗ C4 , and the evolution operator WC takes the
form
WC = (∂0∗ + ∂0 ) ⊗ C0 + (∂1∗ + ∂1 ) ⊗ C1 . (3.33)
By Definition 3.3, one can get the weighted sums of the coin operator system C as follows:
−1 0 0 0 1 0 0 0
(C) 0 −1 0 0 (C) 0 1 0 0
U∅ = 0 , U{0} =
,
0 1 0 0 0 1 0
0 0 0 1 0 0 0 1
−1 0 0 0 1 0 0 0
(C) 0 −1 00 (C) 0 1 0 0
U{1} =
0
, U{0,1} = ,
0 −1 0
0 0
−1 0
0 0 0 −1 0 0 0 −1
which are unitary operators (matrices) on K = C4 . These operators respectively have eigenvectors
v∅ , v{0} , v{1} , v{0,1} given by
1 1 T 1 −1 T
v∅ = 0, 0, √ , √ , v{0} = 0, 0, √ , √ ,
2 2 2 2
1 1 T 1 −1 T
v{1} = √ , √ , 0, 0 , v{0,1} = √ , √ , 0, 0 .
2 2 2 2
Clearly, {v∅ , v{0} , v{1} , v{0,1} } forms an orthonormal system in K = C4 . Now put
1 X b
Ψ= Zγ ⊗ vγ . (3.34)
2
γ∈Γ1
Then, Ψ is a unit vector in h1 ⊗ K = h1 ⊗ C4 . Moreover, for each τ ∈ Γ1 = ∅, {0}, {1}, {0, 1} ,
we have
K −1
dX
bτ ⊗ ej , Ψiej = 1
uτ = hZ vτ ,
j=0
2
(C)
which implies that uτ is an eigenvector of Uτ . Clearly, for τ1 , τ2 ∈ Γ1 with τ1 6= τ2 , one has
1
huτ1 , uτ2 iK = hvτ , vτ iK = 0.
4 1 2
This shows that the assumptions made in Theorem 3.17 can also be satisfied.
18
4 Conclusion remarks
As is seen, we obtain an alternative description of the (n + 1)-dimensional hypercube. And based
on the alternative description, we find that the operators {∂k∗ + ∂k | 0 ≤ k ≤ n} behave actually as
the shift operators. This allows us to introduce a quantum walk model on the (n + 1)-dimensional
hypercube with {∂k∗ + ∂k | 0 ≤ k ≤ n} as the shift operators on the position space. We explicitly
obtain a formula for calculating the probability distribution of the walk at any time and establish
two limit theorems showing that the averaged probability distribution of the walk even converges
to the uniform probability distribution on the (n + 1)-dimensional hypercube. Finally, we prove
that the walk produces the uniform measure as its stationary measure on the (n + 1)-dimensional
hypercube provided its initial state satisfies some mild conditions.
From our work, we may come to some observations as follows: (1) QBN (quantum Bernoulli
noises) can provide a useful framework for producing the uniform measure on a general hypercube
via a quantum walk. (2) For a quantum walk on a hypercube, its “components in the coin space”
can be the determining factors of its probability distributions and evolution behavior.
Acknowledgements
The author is extremely grateful to the referees for their valuable comments and suggestions on
improvement of the first version of the present paper.
Author declarations
Conflict of interest
The author has no conflicts to disclose.
Data Availability
Data sharing is not applicable to this article as no new data were created or analyzed in this
study.
References
[1] D. Aharonov, A. Ambainis, J. Kempe and U. Vazirani, Quantum walks on graphs, In ACM
[ACM01], 2001.
[2] Y. Aharonov, L. Davidovich and N. Zagury, Quantum random walks, Phys. Rev. A 48 (1993),
1687–1690.
[3] G. Alagić and A. Russell, Decoherence in quantum walks on the hypercube, Phys. Rev. A
72 (2005), 062304.
[4] M.-J. Cantero, A. Grünbaum, L. Moral and L. Velázquez, The CGMV method for quantum
walks, Quantum Inf. Process. 11 (2012), 1149-1192.
19
[5] J. Kempe, Quantum random walks-an introductory overview, Contemp. Phys. 44 (2003),
307–327.
[6] J. Kempe, Discrete quantum walks hit exponentially faster, Probab. Theory Relat. Fields
133 (2005), 215–235.
[7] T. Komatsu and N. Konno, Stationary amplitudes of quantum walks on the higher-
dimensional integer lattice, Quantum Inf. Process. 16 (2017), Paper No. 16.
[8] N. Konno, The uniform measure for discrete-time quantum walks in one dimension, Quantum
Inf. Process. 13 (2014), 1103-1125.
[9] N. Konno, Quantum walks, In: U. Franz and M. Schürmann (eds.), Quantum Potential
Theory, Lecture Notes in Mathematics, vol. 1954, pp. 309-452, Springer, Heidelberg (2008)
[10] A. Makmal, M. Zhu, D. Manzano, M. Tiersch and H.J. Briegel, Quantum walks on embedded
hypercubes, Phy. Rev. A 90 (2014), 022314.
[11] F.L. Marquezino, R. Portugal, G. Abal and R. Donangelo, Mixing times in quantum walks
on the hypercube, Phy. Rev. A 77 (2008), 042312.
[12] C. Moore and A. Russell, Quantum walks on the hypercube, In: J.D.P. Rolim, S. Vadhan
(eds), Randomization and Approximation Techniques in Computer Science, Lecture Notes
in Computer Science, vol. 2483, Springer, Berlin (2002).
[13] R. Portugal, Quantum Walks and Search Algorithms, 2nd edn. Springer Nature, 2018.
[14] N. Privault, Stochastic analysis of Bernoulli processes, Probability Surveys 5 (2008), 435–483.
[15] C.S. Wang, H.F Chai and Y.C. Lu, Discrete-time quantum Bernoulli noises, J. Math. Phys.
51 (2010), no. 5, 053528.
[16] C.S. Wang and J.S. Chen, Quantum Markov semigroups constructed from quantum Bernoulli
noises, J. Math. Phys. 57 (2016), no.2, 023502.
[17] C.S. Wang, Y.L. Tang and S.L. Ren, Weighted number operators on Bernoulli functionals
and quantum exclusion semigroups, J. Math. Phys. 60 (2019), 113506.
[18] C.S. Wang and X.J. Ye, Quantum walk in terms of quantum Bernoulli noises. Quantum Inf.
Process. 15 (2016), 1897-1908.
[19] S.E. Venegas-Andraca, Quantum walks: a comprehensive review, Quantum Inf. Process. 11
(2012), 1015-1106.
20