The uniform measure for quantum walk on hypercube:

arXiv:2211.07948v1 [quant-ph] 15 Nov 2022

a quantum Bernoulli noises approach

Ce Wang
Yau Mathematical Sciences Center, Tsinghua University
Beijing 100084, People’s Republic of China

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

Mathematics Subject Classification. 81S25; 60H40; 81Q99.

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 obtain an alternative description of the (n + 1)-dimensional hypercube, where n is a

general nonnegative integer. And based on this alternative description, we find that the
operators {∂k∗ + ∂k | 0 ≤ k ≤ n} behave actually as the shift operators, where ∂k and ∂k∗ are
the annihilation and creation operators acting on Bernoulli functionals.

• 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.

• With {∂k∗ + ∂k | 0 ≤ k ≤ n} as the shift operators on the position space, we introduce a

quantum walk model on the (n + 1)-dimensional hypercube and obtain an explicit formula
for calculating the probability distribution of the walk at any time.

• 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.

Conventions. Throughout this paper, by a Hilbert space we means a separable complex

Hilbert space whose inner product is conjugate linear in the first variable and linear in the second
variable. If A is an bounded operator on a Hilbert space, then A∗ denotes its adjoint. For Hilbert
spaces H1 and H2 , their tensor space is written as H1 ⊗ H2 . By convention, dim H means the
dimension of a Hilbert space H.

2 Quantum Bernoulli noises

Let Ω be the set of all functions ω : N 7→ {−1, 1}, and (ζn )n≥0 the sequence of canonical projections
on Ω given by
ζn (ω) = ω(n), ω ∈ Ω. (2.1)

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

for nj ∈ N, ǫj ∈ {−1, 1} (1 ≤ j ≤ k) with ni 6= nj when i 6= j and k ≥ 1. Thus one has a

probability measure space (Ω, F , P), which is referred to as the Bernoulli space and complex-
valued random variables on it are known as Bernoulli functionals.
Let Z = (Zn )n≥0 be the sequence of Bernoulli functionals generated by the sequence (ζn )n≥0 ,
namely
ζn + qn − pn
Zn = √ , n ≥ 0, (2.3)
2 pn qn
where qn = 1 − pn . Clearly Z = (Zn )n≥0 is a sequence of independent random variables on the
probability measure space (Ω, F , P). Let h be the space of square integrable Bernoulli functionals,
namely
h = L2 (Ω, F , P). (2.4)

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

∂k Zσ = 1σ (k)Zσ\k , ∂k∗ Zσ = [1 − 1σ (k)]Zσ∪k σ ∈ Γ, (2.6)

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

∂k ∂l = ∂l ∂k , ∂k∗ ∂l∗ = ∂l∗ ∂k∗ , ∂k∗ ∂l = ∂l ∂k∗ (k 6= l) (2.7)

and
∂k ∂k = ∂k∗ ∂k∗ = 0, ∂k ∂k∗ + ∂k∗ ∂k = I, (2.8)

where I is the identity operator on h.

3 QBN approach to quantum walk on hypercube

In this section, we present our main work. We first give an alternative description of a hypercube,
and then define and characterize our coin operator systems. Based on these, we introduce our
quantum walk model on a general hypercube and examine its properties both from a perspective
of pure mathematics and from a perspective of probability theory. In particular, we show that
our quantum walk can produce the uniform measure as its stationary measure.

3.1 Alternative description of hypercube

In this subsection, we present an alternative description of a hypercube. To do so, we first recall
some general notions and notation about a graph.
A (simple) graph is a pair (V, E), where V is a nonempty set and known as the vertex set
of the graph, while E is a subset of the set {e | e ⊂ V, #e = 2} and known as the edge set of
the graph, where #e means the cardinality of e as a subset of V . For vertices v1 , v2 ∈ V , if
{v1 , v2 } ⊂ E, then it is said that v1 and v2 are adjacent (there exists an edge linking v1 and v2 ),
and written as v1 ∼ v2 . A graph (V, E) is said to be finite if its vertex set V is a finite set. For
a vertex v of a finite graph (V, E), its degree is defined as deg(v) = #{v ′ ∈ V | v ′ ∼ v}. A finite
graph (V, E) is called a regular graph if deg(v) = d for all v ∈ V , where d is some constant. In
that case, the constant d is called the degree of the regular graph (V, E).
We now recall the usual definition of a hypercube. Let n ≥ 1 be a positive integer and V (n)
the n-fold Cartesian product of the set {0, 1}, namely

V (n) = x = (x1 , x2 , · · · , xn ) | xk ∈ {0, 1}, 1 ≤ k ≤ n . (3.1)

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)

with |x − y| denoting the Hamming distance between x and y, which is given by

n
X
|x − y| = |xk − yk |, x = (x1 , x2 , · · · , xn ), y = (y1 , y2 , · · · , yn ). (3.3)
k=1

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)

where σ △ τ = (σ \ τ ) ∪ (τ \ σ) and #(σ △ τ ) means the cardinality of the set σ △ τ .

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) ).

Proof. Define a mapping J : Γn → V (n+1) in the manner as follows

J(σ) = 1σ (0), 1σ (1), · · · , 1σ (n) , σ ∈ Γn ,

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.

3.2 Coin operator system

To define a quantum walk model, one needs some operators on a Hilbert space to describe the
walker’s internal degrees of freedom. Such operators are usually known as coin operators, while
the space they act on is referred to as the coin space. This subsection defines a notion of coin
operator system and examine properties of such system from a perspective of pure mathematics.
In this subsection, we assume that n ≥ 0 is a given nonnegative integer and K is a Hilbert
space with dim K ≥ n + 1 (namely the dimension of K is not less than n + 1). We denote by
h·, ·iK and k · kK the inner product and norm in K, respectively.

Definition 3.2. A system C = {Ck | 0 ≤ k ≤ n} of bounded operators on K is called a coin

Pn
operator system if the sum k=0 Ck is a unitary operator on K and

Cj∗ Ck = Cj Ck∗ = 0, j 6= k, 0 ≤ j, k ≤ n, (3.6)

where Ck∗ denotes the adjoint of Ck .

Let C = {Ck | 0 ≤ k ≤ n} be a coin operator system on K. Then, it follows immediately that

n
X n
X
Ck∗ Ck = Ck Ck∗ = I, (3.7)
k=0 k=0

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.

Theorem 3.4. Let C = {Ck | 0 ≤ k ≤ n} be a system of bounded operators on K. Then the

following statements are equivalent:

(1) The system C = {Ck | 0 ≤ k ≤ n} is a coin operator system on K.

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

Cj∗ Ck = (Pj U )∗ Pk U = U ∗ Pj Pk U = 0, Cj Ck∗ = Pj U (Pk U )∗ = Pj U U ∗ Pk = Pj Pk = 0.

Thus C = {Ck | 0 ≤ k ≤ n} is a coin operator system.

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)

Clearly, the function ετ (k) takes values in {−1, 1}.

Definition 3.3. For τ ∈ Γn and a coin operator system C = {Ck | 0 ≤ k ≤ n} on K, we define

n
X
Uτ(C) = ετ (k)Ck (3.9)
k=0

and call it the ετ -weighted sum of the coin operator system C.

(C)
Proposition 3.5. Let C = {Ck | 0 ≤ k ≤ n} be a coin operator system on K. Then Uτ is a
unitary operator on K for each τ ∈ Γn .

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

(C) (C)
∗ (C)

where I stands for the identity operator on K. Similarly, we have Uτ Uτ = I. Thus Uτ is
a unitary operator.

3.3 Definition of the quantum walk model

In this subsection, we define our quantum walk model and examine its fundamental properties.
Throughout this subsection, we assume that n ≥ 0 is a fixed nonnegative integer and K is a
finite-dimensional Hilbert space with dK ≡ dim K ≥ n + 1. Additionally, we fix an orthonormal
basis {ej | 0 ≤ j ≤ dK − 1} for K. We will take K as the coin space for our quantum walk model.
Recall that the space h of square integrable Bernoulli functionals has an orthonormal basis
of form {Zσ | σ ∈ Γ}, which is known as its canonical ONB. We denote by hn the subspace of h
spanned by {Zσ | σ ∈ Γn }, namely

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

(∂k∗ + ∂k )Zσ = Zτ . (3.11)

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

Zτ = 1σ (k)Zσ\k + (1 − 1σ (k))Zσ∪k = (1 − 1σ (k))Zσ∪k + 1σ (k)Zσ\k .

On the other hand, it follows from properties of ∂k and ∂k∗ (see (2.6) for details) that

(∂k∗ + ∂k )Zσ = (1 − 1σ (k))Zσ∪k + 1σ (k)Zσ\k .

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 ,

which implies that k ∈ σ with τ = σ\k or k ∈

/ σ with τ = σ∪k, which together with Proposition 3.2
implies σ ∼ τ .

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

Then WC is a unitary operator on hn ⊗ K.

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,

where IK means the identity operator on K. Similarly, we have WC WC ∗ = I.

In the following, unless otherwise specified, we always assume that C = {Ck | 0 ≤ k ≤ n} is a

fixed coin operator system on the space K. We call the operator WC defined by (3.12) the unitary
operator on hn ⊗ K generated by the coin operator system C. The next definition describes our
quantum walk model on the graph (Γn , En ).

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;

• The probability Pt (σ | Φ0 ) of finding the walker on vertex σ ∈ Γn at time t ≥ 0 is given by

K −1
dX
Pt (σ | Φ0 ) = |hZσ ⊗ ej , Φt i|2 , (3.14)
j=0

where {ej | 0 ≤ j ≤ dK − 1} is the orthonormal basis for the coin space K.

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

where I is the identity operator on hn .

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

(∂k∗ + ∂k )Aσ = εσ (k)Aσ . (3.19)

Proof. Let k ∈ Nn be given. Then, by CAR in equal time (2.8), we find

(∂k∗ + ∂k ) I + εσ (k)(∂k∗ + ∂k ) = εσ (k) I + εσ (k)(∂k∗ + ∂k ) ,

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 .

This completes the proof.


Theorem 3.10. The vector system Zbσ | σ ∈ Γn forms an orthonormal basis for the position
space hn . Moreover, it holds true that

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

bσ = √ 1 (∂k∗ + ∂k )Aσ Z∅ = √ 1 εσ (k)Aσ Z∅ = εσ (k)Z

(∂k∗ + ∂k )Z bσ .
2n+1 2n+1
In summary, the theorem is true.

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

which is the desired.

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

Proof. Let τ ∈ Γn and u ∈ K be given. Then (∂k∗ + ∂k )Zbτ = ετ (k)Z

bτ for all k ∈ Nn . Thus, by
the definition of the evolution operator WC , we have
n
X n
X n
X
bτ ⊗ u) =
WC (Z [(∂k∗ + ∂k )Zbτ ] ⊗ (Ck u) = bτ ) ⊗ (Ck u) = Z
(ετ (k)Z bσ ⊗ ετ (k)Ck u,
k=0 k=0 k=0

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

This completes the proof.

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 .

Proof. Divide Γn × Γn into three parts Γn × Γn = △1 ∪ △2 ∪ △3 , where △1 = {(τ, τ ) | τ ∈ Γn },

△2 = {(τ1 , τ2 ) ∈ Γn × Γn | τ1 6= τ2 , bτ1 = bτ2 }, △3 = {(τ1 , τ2 ) ∈ Γn × Γn | τ1 6= τ2 , bτ1 6= 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

which implies that

1 h X i
P T (σ | Φ0 ) = n+1
1+ (−1)#(σ\τ1 )+#(σ\τ2 ) huτ1 , uτ2 iK
2
(τ1 ,τ2 )∈△2

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

where T ≥ 1. Note that

T −1
T
1 X t t 1 1 − bτ1 bτ2
lim bτ1 bτ2 = lim = 0.
T →∞ T T →∞ T 1 − bτ1 bτ2
t=0

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

which is the same as (3.25).

(C)
Remark 3.4. For γ ∈ Γn , let vγ be an eigenvector of Uγ with bγ being the corresponding
eigenvalue (such an eigenvector does exist because the coin space K is finite-dimensional and
(C)
Uγ is a unitary operator on K). Put
− 12
X
Φ0 = M 0 bγ ⊗ vγ ,
Z
γ∈Γn
P
where M0 = γ∈Γn kvγ k2K . Then, Φ0 is a unit vector in hn ⊗ K, hence can serve as an initial
state of the walk WC . Moreover, for each τ ∈ Γn , we have
K −1
dX
−1
uτ = hZbτ ⊗ ej , Φ0 iej = M0 2 vτ ,
j=0

(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.

3.5 The uniform measure as a stationary measure

The present subsection shows that the walk WC produces the uniform measure as its stationary
measure on Γn provided its initial state satisfies some mild conditions.
In this subsection, we assume that n ≥ 0 is a fixed nonnegative integer and the coin space K
is a finite-dimensional Hilbert space with dK ≡ dim K ≥ 2n+1 . Additionally, we assume that that
C = {Ck | 0 ≤ k ≤ n} is a given coin operator system on K. As above, Φ0 always denotes the
initial state of the walk WC below.
A probability measure (measure for short) ν on Γn is a nonnegative function ν : Γn → R+
P
satisfying that σ∈Γn ν(σ) = 1. The uniform measure κn on Γn is the measure given by

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

bγ ⊗u produces the uniform measure as its stationary

This exactly means that walk WC with Φ0 = Z
measure on Γn .
(C)
Recall that, for τ ∈ Γn , Uτ is the ετ -weighted sum of the coin operator system C (see
Definition 3.3). The next theorem shows that, even for some “complicated” initial states, the
walk WC still produces the uniform measure.
PdK −1 b
Theorem 3.17. Let Ψ ∈ hn ⊗ K be a unit vector and uτ = j=0 hZτ ⊗ ej , Ψiej for τ ∈ Γn .
Suppose further that
(C)
(1) for each τ ∈ Γn , uτ is an eigenvector of the unitary operator Uτ ;

(2) huτ1 , uτ2 iK = 0 for τ1 , τ2 ∈ Γn with τ1 6= τ2 .

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

which together with

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)

Additionally, the graph (Γ1 , E1 ) is regular and its degree is exactly 2.

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):

(C)
(C)
(C)
(C)

Spec U∅ = Spec U{0,1} = {−1, 1}, Spec U{0} = Spec U{1} = {−i, i},

where Spec(A) means the spectrum of an operator A acting on K = C2 .

(C)
Let v∅ ∈ C2 be an eigenvector of U∅ with b∅ = −1 being the corresponding eigenvalue,
2 (C)
v{0} ∈ C be an eigenvector of U{0} with b{0} = −i being the corresponding eigenvalue, v{1} ∈ C2
(C)
be an eigenvector of U{1} with b{1} = i being the corresponding eigenvalue, and v{0,1} ∈ C2 be an
(C)
eigenvector of U{0,1} with b{0,1} = 1 being the corresponding eigenvalue. Put

− 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