PHYSICAL REVIEW X 3, 041007 (2013)

Degree Distribution in Quantum Walks on Complex Networks

Mauro Faccin,1,* Tomi Johnson,1,2 Jacob Biamonte,1 Sabre Kais,3,4 and Piotr Migdał5,1
1
Institute for Scientific Interchange, Via Alassio 11/c, 10126 Torino, Italy
2
Clarendon Laboratory, University of Oxford, Parks Road, Oxford OX1 3PU, United Kingdom
3
Department of Chemistry, Physics, and Birck Nanotechnology Center, Purdue University, West Lafayette, Indiana 47907, USA
4
Qatar Environment and Energy Research Institute (QEERI), Doha, Qatar
5
ICFO–Institut de Ciències Fotòniques, 08860 Castelldefels (Barcelona), Spain
(Received 29 May 2013; revised manuscript received 1 August 2013; published 24 October 2013)
In this theoretical study, we analyze quantum walks on complex networks, which model network-based
processes ranging from quantum computing to biology and even sociology. Specifically, we analytically
relate the average long-time probability distribution for the location of a unitary quantum walker to that of
a corresponding classical walker. The distribution of the classical walker is proportional to the distribution
of degrees, which measures the connectivity of the network nodes and underlies many methods for
analyzing classical networks, including website ranking. The quantum distribution becomes exactly equal
to the classical distribution when the walk has zero energy, and at higher energies, the difference, the so-
called quantumness, is bounded by the energy of the initial state. We give an example for which the
quantumness equals a Rényi entropy of the normalized weighted degrees, guiding us to regimes for which
the classical degree-dependent result is recovered and others for which quantum effects dominate.

DOI: 10.1103/PhysRevX.3.041007 Subject Areas: Complex Systems, Quantum Physics

I. INTRODUCTION A widely applicable analysis of unitary quantum walks

has remained illusive because of a strong dependence on
A quantum walk on a network is a fundamental natural
the initial state that is exacerbated by the lack of conver-
process [1–5] since the quantum dynamics of any discrete
gence to a steady state (which is not necessarily the case for
system can be reexpressed and interpreted as a single-
particle quantum walk, which is capable of performing open quantum walks [20,21]). To overcome these difficul-
universal quantum computation [6]. Quantum walks are ties and obtain a result that is relevant beyond specific
also of increasing relevance outside physics. As well as initial states and walk durations, we consider a quantity
being a powerful tool for studying transport in quantum that characterizes each walk over long times and relate this
systems [7–10], e.g., the transport of energy through bio- to another quantity that characterizes each initial state.
logical complexes or man-made solar cells, quantum walks Specifically, we characterize a walk by the long-time av-
have been proposed as a means of analyzing classical erage probability distribution of finding the walker at each
sociological networks [11–14]. Fully understanding these node [16,22], which captures all knowledge of the location
phenomena, and others, for networks with nontrivial top- of the walker in the absence of knowledge about when the
ologies, requires the merging of the methods of complex walk began. The initial state is naturally characterized by
networks and quantum mechanics [15]. the energy, which gives a total ordering of the initial
While analytical results have been obtained for some states. We show that for low energies the long-time average
specific topologies, such as starlike [16–18], regular, or probability distribution is equal to the normalized distri-
semiregular [19] networks, progress in analyzing quantum bution of degrees in the network. Specifically, the energy
walks on complex networks has largely been based on bounds the trace distance between the two distributions.
numerics, leaving open the possibility that many conclu- This provides a wide class of quantum walks on complex
sions are not representative of all regimes. In this article, networks with an analytically tractable low-energy regime.
we instead discover analytical properties of continuous- Our result is achieved by mapping the properties of the
time unitary quantum walks of arbitrary topology that ground state of a quantum walk to the steady state of a
follow from the topology of the underlying complex corresponding classical walk—in particular, a classical
network. walk whose steady state represents the connectivity of
nodes as determined by their degree. Such walks are
used by search engines, e.g., Google, to rank websites
[23,24]. Our result extends the importance of the concept
*[email protected] of degree from classical systems [25–30], ranging from
the sociological to the ecological [31–33], to quantum
Published by the American Physical Society under the terms of
the Creative Commons Attribution 3.0 License. Further distri- systems [7–10,34–36].
bution of this work must maintain attribution to the author(s) and As a case study, we analytically and numerically
the published article’s title, journal citation, and DOI. study the walk for a range of model complex network

2160-3308=13=3(4)=041007(8) 041007-1 Published by the American Physical Society

FACCIN et al. PHYS. REV. X 3, 041007 (2013)

structures, including the Barabàsi-Albert (BA), Erdős-

Rényi (ER), Watt-Strogatz (WS), and random geometric
(RG) networks. We repeat this analysis for several real-
world networks, specifically, a karate club (KC) social
network [37], the email (EM) network of Universitat
Rovira i Virgili (URV) [38], the C. elegans metabolic
(CE) network [39], and a coauthorship (CA) network of
scientists [40]. To compare these networks, we start from
an evenly distributed initial state. We find an additional
connection to the degrees for this case, namely, that the
quantumness of the walk is itself controlled by the het-
erogeneity of the degrees, which we quantify in terms of
a Rényi entropy.
In Sec. II, we formulate and study the problem analyti-
cally, first for a stochastic walk and then for a quantum
walk. Following this, in Sec. III, we confirm our analytical FIG. 1. Relating stochastic and quantum walks. An undirected
results for the quantumness of a quantum walk numerically weighted network (graph) G is represented by a symmetric, off-
and explore the way in which the quantum long-time diagonal and non-negative adjacency matrix A. There is a map-
average deviates from the corresponding classical distri- ping from A (by summing columns) to the diagonal matrix D,
with entries given by the weighted degree of the corresponding
bution. We conclude with a discussion in Sec. IV.
node. The node degrees are proportional to the steady-state
probability distribution of the continuous-time stochastic walk
II. WALKS FRAMEWORK (with uniform escape rate from each node) generated by HC ¼
We consider a walker moving on a connected network of LD
1 , where L ¼ D
A is the Laplacian. The steady-state
probabilities, represented by the vector j
0 i, are proportional
N nodes, with each weighted undirected edge between
to the node degrees. We generate a corresponding continuous-
nodes i and j described by the element Aij of the off- time unitary quantum walk by the Hermitian operator HQ ¼
diagonal adjacency matrix A. The matrix is symmetric D
1=2 LD
1=2 , which is similar to HC . The probability of being
(Aij ¼ Aji ) and has real, non-negative entries. We use in a node in the stochastic stationary state j
0 i and the proba-
P
Dirac notation and represent A ¼ ij Aij jiihjj in terms of bility arising from the quantum ground state are equal and
N orthonormal vectors jii. proportional to the node degree.
The network gives rise to both a quantum walk and a
corresponding classical walk. The classical stochastic equal to zero. We assume throughout this work that the
walk SðtÞ ¼ e
HC t is generated by the infinitesimal sto- walk is connected, i.e., that it is possible to transition

1
chastic (see, e.g., Refs. [41–43]) operator HP C ¼ LD , from any node to any other node through some series of
where L ¼ D
A is the Laplacian and D ¼ i di jiihij is allowed transitions. In this case, there is a unique eigen-
defined by its diagonal elements, the degrees, di ¼ vector j
0 i ¼ jPC i with 0 ¼ 0, and i > 0 for all
P
j Aij . For this classical walk, the total rate of leaving iPÞ 0 [41,44–46]. This (normalized) eigenvector jPC i ¼
each node is identical. The corresponding unitary quan- i ðPC Þi jii describes the steady-state distribution
tum walk UðtÞ ¼ e
iHQ t is generated by the Hermitian
d
operator HQ ¼ D
1=2 LD
1=2 . For this quantum walk, the ðPC Þi ¼ P i : (1)
energies hijHQ jii at each node are identical. The gener- j dj

ators HC and HQ are similar matrices, related by HQ ¼ In other words, the process is ergodic, and after long
D
1=2 HC D1=2 . This mathematical framework, repre- times, the probability of finding the walker at any node i
sented in Fig. 1, underpins our analysis. is given purely by the importance of the degree di of that
As we will describe in Sec. II A, the long-time behavior node in the network underlying the process.
of the classical walk generated by HC has been well
explained in terms of its underlying network properties, B. Quantum walks
specifically, the degrees di . Our goal in Sec. II B is to
determine the role this concept plays in the quantum When considering quantum walks on networks, it is
walk generated by HQ . natural to ask what the long-time behavior of a quantum
walker is [11,15,22,47]. The unitary evolution will not
drive the system towards a steady state. Therefore, to
A. Classical walks
obtain a static picture, we consider the long-time average
In the classical walk, the probability Pi ðtÞ of being at probability ðPQ Þi of being on node i, which reads
node i atP time t evolves as jPðtÞi ¼ SðtÞjPð0Þi, where
jPðtÞi ¼ i Pi ðtÞjii. The stationary states of the walk are 1 ZT
ðPQ Þi ¼ lim dthijUðtÞð0ÞUy ðtÞjii: (2)
described by eigenvectors j
ki i of HC with eigenvalues i T!1 T 0

041007-2
DEGREE DISTRIBUTION IN QUANTUM WALKS ON . . . PHYS. REV. X 3, 041007 (2013)

For ease of comparison with jPC i, we P will also write the This expression for the quantumness in Eq. (6) enables
distribution in Eq. (3) as a ket jPQ i ¼ i ðPQ Þi jii. Unlike us to make some physical statements about a general initial
the classical case, Eq. (2) depends on the initial state ð0Þ. state. By realizing that j0 i is the ground state of zero
Interference between subspaces of different energy van- energy 0 ¼ 0 and the gap  ¼ miniÞ0 i in the energy
ishes in the long-time average, so we obtain an expression spectrum is nonzero for a connected network [41,44–46],
for the probability ðPQ Þi in terms of the energy-eigenspace the above implies a bound E=  " for the quantumness "
projectors
j of the Hamiltonian HQ , of the walk in terms of the energy E ¼ trfHQ g of the
X initial state. The bound is obtained through the following
ðPQ Þi ¼ hij
j ð0Þ
j jii: (3) steps:
j X X
P E ¼ trfHQ g ¼ j trf
j ð0Þg   trf
j ð0Þg
Here,
j ¼ k jj ihj j
k k projects onto the subspace
jÞ0 jÞ0
spanned by the eigenvalues jkj i of HQ , corresponding to
the same eigenvalue j . In other words, the long-time ¼ ð1
trf
0 ð0ÞgÞ ¼ : (7)
average distribution is a mixture of the distributions ob- The equation above demonstrates that the classical sta-
tained by projecting the initial state onto each eigenspace. tionary probability distribution will be recovered for low
Because of the similarity transformation HQ ¼ energies. A utility of this result is that it connects the long-
D
1=2 HC D1=2 , the classical HC and quantum HQ genera- time average distribution to a simple physical property of
tors share the same eigenvalues i  0 and have the walk, the energy, which provides a total ordering of all
eigenvectors related by jki i ¼ D
1=2 j
ki i up to their possible initial states.
normalizations. In particular, the unique eigenvectors cor-
responding to 0 ¼ 0 are j
0 i ¼ Dj1iP and j0 i ¼ D1=2 j1i C. Degree distribution and quantumness
up to their normalizations, with j1i ¼ i jii. Therefore, the Quantumness is a function of both the degrees of the
probability vector describing the outcomes of a measure- network nodes and the initial state. To compare the quan-
ment of the quantum ground-state eigenvector j0 i in tumness of different complex networks, we fix the initial
the node basis is the classical steady-state distribution state ð0Þ. For our example, we choose the even superpo-
j
0 i ¼ jPC i. pffiffiffiffi
sition state ð0Þ ¼ jð0Þihð0Þj with jð0Þi ¼ j1i= N .
The state vector jPC i appears in Eq. (3) for the quantum This state has several appealing properties; for example, it
long-time average distribution jPQ i with weight is invariant under node permutations and independent of
h0 jð0Þj0 i. Accordingly, we split the sum in Eq. (3) the arrangement of the network.
into two parts: The first part we call the ‘‘classical term’’ In this case, the quantumness is given by the expression
jPC i, and the second we call the ‘‘quantum correction’’ pffiffiffi
jP~Q i, given as h di2
"¼1
; (8)
hdi
jPQ i ¼ ð1
"ÞjPC i þ "jP~Q i: (4) P pffiffiffi
P
The normalized quantum correction jP~Q i ¼ i ðP~Q Þi jii is
where
P pffiffiffiffiffi hdi ¼ i di =N is the average degree and h di ¼
i di =N is the average root degree of the nodes. As such,
given by the quantumness depends only on the degree distribution of
1X the network and increases with network heterogeneity.
ðP~Q Þi ¼ hij
j ð0Þ
j jii; (5) This statement is quantified by writing the quantumness
" jÞ0
  
1 d
and the weight " ¼ 1
exp H1=2 P i ; (9)
N j dj
" ¼ 1
h0 jð0Þj0 i (6) in terms of the Rényi entropy
that we call quantumness is a function of both the degrees, X 
1
through j0 i, and the initial state. Hq ðfpi gÞ ¼ ln pqi ; (10)
1
q i
We can think of the parameter ", which controls the
P
classical-quantum mixture, as the quantumness of jPQ i for where di = j dj ¼ ðPC Þi are the normalized degrees.
the following three reasons. First, the proportion of the To obtain an expression in terms of the (perhaps) more
elements in ðPQ Þi that corresponds to the quantum correc- familiar Shannon entropy H1 [obtained by taking the
tion is ". Second, the trace distance between the normal- q ! 1 limit of Eq. (10)], we recall that the Rényi entropy
ized distribution ðPC Þi and the unnormalized distribution is nonincreasing with q [48], which leads to the upper
ð1
"ÞðPC Þi forming the classical part of the quantum bound
result is also ". Last, using a triangle inequality, the trace   
1 di
distance between the normalized distributions ðPC Þi and "  1
exp H1 P : (11)
ðPQ Þi is upper bounded by 2". N j dj

041007-3
FACCIN et al. PHYS. REV. X 3, 041007 (2013)

The quantumness approaches this upper bound in the limit quantumness of a connected network is more restricted.
that M nodes have uniform degree di ¼ Mhdi=N and all It is maximized by a walk based on a star network—where
others have di ¼ 0. This limit is never achieved unless a single node is connected
pffiffiffiffiffiffiffiffiffiffiffiffiffi to all others. For a walk of this
M ¼ N and " ¼ 0, e.g., in a regular network. Physically, type, " ¼ 1=2
N
1=N  1=2.
" ¼ 0 for a regular network because the symmetry of the Next, in Sec. III, we confirm the above analytical find-
Hamiltonian HQ implies that its eigenvectors are evenly ings numerically and, at the same time, numerically study
distributed. The only eigenvector of this type that is posi- the form of the quantum correction jP~Q i given by Eq. (5)
tive is the initial state jð0Þi, which because of the Perron- for a range of complex network topologies.
Frobenius theorem, must also be the ground state jð0Þi ¼
j0 i. Therefore, E ¼ 0 and so, from Eq. (7), " ¼ 0. III. NUMERICAL RESULTS
In another limit, the quantumness takes its maximum A. Artificial network topologies
value " ¼ ðN
2Þ=N  1 when the degrees of two nodes
are equal and much larger than those of the others (note We consider walkers on model networks, each with a
that the symmetry of A prevents the degree of a single node fundamentally different complex network topology. To
from dominating). In the case that Aij 2 f0; 1g, i.e., the start, we consider nonweighted binary networks Aij 2
network underlying the walks is not weighted, the f0; 1g with N ¼ 500 nodes and average degree hdi  6. If

FIG. 2. Long-time average probability and degree for nodes in a complex network. Eight networks are considered: BA, ER, WS, RG,
KC, EM, CE, and CA. We plot the classical ðPC Þi (red dashed line) and quantum ðPQ Þi (black +) probabilities against the degree di for
every node i. We overlay this with a plot of the average degree distribution PðdÞ against d for each network type (grey solid line), when
known, along with the distribution for the specific realization used (grey +). Alongside the BA network, we also plot ðPQ Þi for the
optimized BA (BA-opt) network, in which the internode weights of the BA network are randomly varied in a Monte Carlo algorithm to
reach " ¼ 0:6 (orange x). We do not include a plot of the degree distribution for this network.

041007-4
DEGREE DISTRIBUTION IN QUANTUM WALKS ON . . . PHYS. REV. X 3, 041007 (2013)

a disconnected network is obtained, only the giant

component is considered. Specifically, we consider the BA
scale-free network [25], the ER [49] and the WS [50] small-
world networks, and the RG (on a square) [51], a network
without the scale-free or small-world characteristics.
The long-time average probability of being on each node
i is plotted against its degree di for a quantum (PQ ) and a
stochastic (PC ) walk in Fig. 2. The two cases are nearly
identical for these binary networks and the evenly distrib-
uted initial state, illustrating that the quantumness " is
small. We have in fact calculated the quantumness directly
for each network type, yielding " ¼ 0:130, 0.043, 0.016,
and 0.040 for the BA, ER, WS, and RG networks, respec-
tively. Within these, the BA network shows the highest
quantum correction. This result is expected since the BA
network has a more heterogeneous degree distribution. The
WS network, which is well known to have quite uniform
degrees [52], is accordingly the network with the lowest
quantum correction.
For many of the network types, the typical quantumness
can be obtained from the expected (thermodynamic limit)
degree distribution. In the BA network, the degree distri- FIG. 3. Quantumness and degree entropy. The value of "
bution approximately obeys the continuous probability against H1 [Eq. (10)] for the nine different networks considered
in Fig. 2, as well as the random regular (RR) network (a network
density PðdÞ ¼ hdi2 =2d3 [25]. Integrating this to find the
with the same degree for each node—in this case, we consider a
moments results in " ¼ 1=9, which is independent of the regular graph with each node of degree six) and star (ST)
average degree hdi and is compatible with our numerics. networks (black +). We also plot " and H1 for the network
The degree distributions of the ER and RG networks both obtained in several iteration steps, each randomly varying an
approximately follow the Poissonian distribution PðdÞ  internode weight of the BA network, for an increasing number of
hdid e
hdi =d! for large networks, which explains the simi- iteration steps (bottom to top, grey to orange x). The quantum-
larity of their quantumness " values. For hdi ¼ 6, we ness " increases and the entropy H1 decreases with step number.
recover "  0:046, which is compatible with the values The red dashed line represents the upper bound of Eq. (11).
for the particular networks we generated. From the general
form, calculating the quantumness numerically and per-
forming a best fit, we find that "  1 hdi
2 , with fitting value without being ensured by the bound only (see
parameters 1 ¼ 0:429 and 2 ¼ 1:210. Table I).
The size of the quantum effects can be enhanced by Finally, our numerical calculations reveal the behavior
introducing heterogeneous weights Aij within a network. of the quantum part P~Q of the long-time average node
We have done this for a BA network by using several occupation. We find that the quantum part enhances the
iterations of the following procedure. A pair of connected long-time average probability of being at nodes with
nodes is randomly selected, and then the associated weight small degree relative to the classical part. More precisely,
is doubled or halved at random. As anticipated, the effect is
to increase the discrepancy between the classical and
TABLE I. Quantumness, energy, and gap. We show the quan-
quantum dependence of the long-time average probability tumness " and its upper bound E= (the ratio of the energy and
on degree, illustrated in Fig. 2. As the number of iterations the gap), for each of the nine networks considered in Fig. 2.
is increased, the quantumness follows the bound given in
Eq. (11), as shown in Fig. 3. In fact, most networks are Type " E=
found to be close to saturating this bound, especially for BA 0.1299 0.5583
low quantumness. ER 0.0431 0.1734
Further, the energy E ¼ h0 jHQ j0 i of the given RG 0.0396 11.2875
pffiffiffiffiffiffiffiffiffihas a simple expression E ¼ 1
ð1=NÞ
initial state
P WS 0.0164 0.0846
ij Aij = di dj , which allows us to determine the extent BA-opt 0.6092 844.9181
to which the bound E=  " is saturated by comparing KC 0.1204 1.3471
the values of E= and ". We find that for some networks, CE 0.2247 4.7622
e.g., the BA, ER, and WS networks, the bound is quite EM 0.1987 1.5449
restrictive and reasonably saturated. However, for the CA 0.1138 39.8535
other networks, we find that quantumness takes a low

041007-5
FACCIN et al. PHYS. REV. X 3, 041007 (2013)

ðP~Q Þi =ðPC Þi exhibits roughly ðdi Þ
3 scaling, with 3  1, Ref. [30]). Therefore, we now study the behavior of the
as shown in Fig. 4. Interestingly, there is a correlation quantumness and gap for topologies present in a variety of
between the amount of enhancement, given by 3 , and real-world networks, as described in Sec. I: a karate club
the type of complex network. The network types with (KC) social network [37], the e-mail (EM) network of URV
smaller diameters (order of increasing diameter: BA, then University [38], the C. elegans metabolic (CE) network
ER and WS, then RG) have the smallest 3 , and the [39], and a coauthorship (CA) network of scientists [40].
quantum parts enhance the low-degree nodes least. The values of the quantumness and comparison against
Moreover, the enhancement 3 seems to be quite indepen- the entropic upper bound are shown in Fig. 3. Despite
dent of the internode weights. Thus, our numerics show a the variety of topologies, we again find that the quantum-
qualitatively common quantum effect for a range of com- ness is consistently small. Therefore, the classical and
plex network types. Quantitative details vary between the quantum distributions are very close, as shown in Fig. 2.
network types but appear robust within each type. Additionally, the quantum correction exhibits the same
generic behavior as observed for the artificial networks;
B. Real-world network topologies Fig. 4 shows an enhancement of the probability of being in
The models of networks examined in the previous sub- nodes of small degree. Interestingly, the quantumness of
section have very specific topologies and, therefore, degree real-world networks is appreciably smaller than enforced
distributions, and they do not capture the topological prop- by the bound of Eq. (7), with E=" taking large values, as
erties of all real-world networks (for details, see Chap. 9 of shown in Table I.

FIG. 4. Quantum effects. The ratio of the quantum ðP~Q Þi and classical ðPC Þi probabilities plotted against degree di (black +) for every
i, for the networks considered in Fig. 2. We also plot the best-fitting curve (red dashed line) to these data of the form ðP~Q Þi =ðPC Þi /
ðdi Þ3 , whose exponent 3 is given in the plot.

041007-6
DEGREE DISTRIBUTION IN QUANTUM WALKS ON . . . PHYS. REV. X 3, 041007 (2013)

Environment and Energy Research Institute (QEERI),

IV. DISCUSSION
where part of this research was completed. P. M. would
We have found an analytical expression for the average like to acknowledge the Spanish MINCIN/MINECO
long-time probability distribution for the location of a low- project TOQATA (FIS2008-00784), EU Integrated
energy quantum walker on a complex network of arbitrary Projects AQUTE and SIQS, and HISTERA project
topology. Specifically, we have shown that this is equal to DIQUIP.
the distribution arising in the steady state of a correspond-
ing classical walk, equal to the normalized degrees. As
well as providing an analytical solution for low-energy
walks, our result will allow the benchmarking of other
methods for studying quantum walks on complex net- [1] Martı́ Cuquet and John Calsamiglia, Entanglement
works, a field in which numerical analysis is typically the Percolation in Quantum Complex Networks, Phys. Rev.
Lett. 103, 240503 (2009).
only viable option.
[2] S. Perseguers, M. Lewenstein, A. Acı́n, and J. I. Cirac,
The stationary state of the classical walk generated by
Quantum Random Networks, Nat. Phys. 6, 539 (2010).
the asymmetrically normalized Laplacian HC is closely [3] S. Perseguers, D. Cavalcanti, G. J. Lapeyre, M.
connected to the ranking of nodes within a network, as Lewenstein, and A. Acı́n, Multipartite Entanglement
used by Google. Therefore, our results indicate that the Percolation, Phys. Rev. A 81, 032327 (2010).
long-time average probability distribution of a quantum [4] Richard P. Feynman, Robert B. Leighton, and Matthew
walk under HQ , with the energy of the initial state as a free Sands, The Feynman Lectures on Physics (Addison-
parameter, could provide a means of interpolating between Wesley, Reading, MA, 2005), 2nd ed.
classical and quantum [11,12] ranking of real-world net- [5] Richard P. Feynman and A. R. Hibbs, Quantum Mechanics
works. This idea also connects nicely with the work of and Path Integrals, International Series in Pure and
Ref. [14] in which the authors numerically simulate driving Applied Physics (McGraw-Hill, New York, 1965).
a quantum system to its ground state, with the quantum [6] Andrew M. Childs, Universal Computation by Quantum
Walk, Phys. Rev. Lett. 102, 180501 (2009).
system chosen such that its ground state represents the
[7] E. Farhi and S. Gutmann, Quantum Computation and
Google ranking vector. Decision Trees, Phys. Rev. A 58, 915 (1998).
For the evenly distributed initial state, the quantumness [8] F. Caruso, A. W. Chin, A. Datta, S. F. Huelga, and M. B.
(loosely speaking, the difference between the classical and Plenio, Highly Efficient Energy Excitation Transfer in
quantum distributions) only depends on the degrees. Light-Harvesting Complexes: The Fundamental Role of
Together with our result for the low-energy distribution, Noise-Assisted Transport, J. Chem. Phys. 131, 105106
this result shows that the degree distribution can be as (2009).
important and illuminating in quantum walks as in their [9] M. Mohseni, P. Rebentrost, S. Lloyd, and A. Aspuru-
classical counterparts. Our numerical examples also show Guzik, Environment-Assisted Quantum Walks in
that for remarkably diverse network types, quantum effects Photosynthetic Energy Transfer, J. Chem. Phys. 129,
are qualitatively similar; they act to reduce the degree 174106 (2008).
dependence of the average probability of a walker being [10] Yuan-Chung Cheng and Graham R. Fleming, Dynamics of
Light Harvesting in Photosynthesis, Annu. Rev. Phys.
found on a node.
Chem. 60, 241 (2009).
Our presentation focused on a walker state  that is a [11] E. Sánchez-Burillo, J. Duch, J. Gómez-Gardeñes, and D.
time average over a unitary evolution. However, to con- Zueco, Quantum Navigation and Ranking in Complex
clude, note that our analytical solution to the expected node Networks, Nat. Sci. Rep. Ochanomizu Univ. 2, 605 (2012).
occupation hijjii holds whenever a significant portion of [12] G. D. Paparo and M. A. Martin-Delgado, Google in a
 is in the ground-state subspace of HQ , i.e., when  ¼ Quantum Network, Sci. Rep. 2, 444 (2012).
P

0 
0 is small. To see this, one can always~ use
1 [13] Silvano Garnerone, Paolo Zanardi, and Daniel A. Lidar,
j
j ¼ 1 to expand hijjii ¼ ð1
"ÞðPC Þi þ "ðPQ Þi .
Adiabatic Quantum Algorithm for Search Engine Ranking,
Similarly, the bound E ¼ trfHQ g  " will always Phys. Rev. Lett. 108, 230506 (2012).
[14] Silvano Garnerone, Thermodynamic Formalism for
hold [Eq. (7)]. In particular, these results are independent
Dissipative Quantum Walks, Phys. Rev. A 86, 032342
of whether  is obtained by a unitary or a nonunitary walk. (2012).
For example, the steady state  of a walker equilibrating [15] O. Mülken and A. Blumen, Continuous-Time Quantum
with a low-temperature bath has a small "; thus, hijjii is Walks: Models for Coherent Transport on Complex
proportional to the degree. Networks, Phys. Rep. 502, 37 (2011).
[16] Oliver Muelken, Inefficient Quantum Walks on Networks:
ACKNOWLEDGMENTS The Role of the Density of States, arXiv:0710.3453.
[17] Oliver Mülken, Veronika Bierbaum, and Alexander
We thank Michele Allegra, Ville Bergholm, Stephen Blumen, Coherent Exciton Transport in Dendrimers and
Kirkland, Giovanni Petri, and Simone Severini for fruitful Continuous-Time Quantum Walks, J. Chem. Phys. 124,
discussions. J. D. B. would like to thank the Qatar 124905 (2006).

041007-7
FACCIN et al. PHYS. REV. X 3, 041007 (2013)

[18] Chengzhen Cai and Zheng Yu Chen, Rouse Dynamics of a [36] Salvador Elias Venegas-Andraca, Quantum Walks for
Dendrimer Model in the # Condition, Macromolecules Computer Scientists, Synthesis Lectures on Quantum
30, 5104 (1997). Computing 1, 1 (2008).
[19] S. Salimi, Continuous-Time Quantum Walks on Semi- [37] Wayne W. Zachary, An Information Flow Model for
Regular Spidernet Graphs via Quantum Probability Conflict and Fission in Small Groups, J. Anthropol. Res.
Theory, Quantum Inf. Process. 9, 75 (2010). 33, 452 (1977).
[20] Herbert Spohn, An Algebraic Condition for the Approach [38] Roger Guimera, Leon Danon, A Diaz-Guilera, Francesc
to Equilibrium of an Open N-Level System, Lett. Math. Giralt, and Alex Arenas, Self-Similar Community
Phys. 2, 33 (1977). Structure in a Network of Human Interactions, Phys.
[21] James D. Whitfield, César A. Rodrı́guez-Rosario, and Rev. E 68, 065103 (2003).
Alán Aspuru-Guzik, Quantum Stochastic Walks: A [39] Jordi Duch and Alex Arenas, Community Detection in
Generalization of Classical Random Walks and Quantum Complex Networks Using Extremal Optimization, Phys.
Walks, Phys. Rev. A 81, 022323 (2010). Rev. E 72, 027104 (2005).
[22] Oliver Mülken, Antonio Volta, and Alexander Blumen, [40] Mark E. J. Newman, Finding Community Structure in
Asymmetries in Symmetric Quantum Walks on Two- Networks Using the Eigenvectors of Matrices, Phys.
Dimensional Networks, Phys. Rev. A 72, 042334 (2005). Rev. E 74, 036104 (2006).
[23] Michalis Faloutsos, Petros Faloutsos, and Christos [41] John C. Baez and Jacob Biamonte, A Course on Quantum
Faloutsos, On Power-Law Relationships of the Internet Techniques for Stochastic Mechanics, arXiv:1209.3632.
Topology, SIGCOMM Comput. Commun. Rev. 29, 251 [42] T. H. Johnson, S. R. Clark, and D. Jaksch, Dynamical
(1999). Simulations of Classical Stochastic Systems Using Matrix
[24] R. Albert, H. Jeong, and A. L. Barabási, Internet: Product States, Phys. Rev. E 82, 036702 (2010).
Diameter of the World-Wide Web, Nature (London) 401, [43] J. C. Baez and B. Fong, A Noether Theorem for
130 (1999). Markov Processes, J. Math. Phys. (N.Y.) 54, 013301
[25] A.-L. Barabási and R. Albert, Emergence of Scaling in (2013).
Random Networks, Science 286, 509 (1999). [44] Joel Keizer, On the Solutions and the Steady States of a
[26] Réka Albert and Albert-László Barabási, Statistical Master Equation, J. Stat. Phys. 6, 67 (1972).
Mechanics of Complex Networks, Rev. Mod. Phys. 74, [45] Peter Lancaster and Miron Tismenetsky, Theory of
47 (2002). Matrices (Academic Press, New York, 1985), Vol. 2.
[27] M. E. J. Newman, The Structure and Function of Complex [46] James R. Norris, Markov Chains, 2008 (Cambridge
Networks, SIAM Rev. 45, 167 (2003). University Press, Cambridge, England, 1998).
[28] Mark Newman, Networks: An Introduction (Oxford [47] Dorit Aharonov, Andris Ambainis, Julia Kempe, and
University Press, New York, NY, 2010). Umesh Vazirani, Quantum Walks on Graphs, in
[29] D. J. de Solla Price, Networks of Scientific Papers, Science Proceedings of the 33rd Annual ACM Symposium on
149, 510 (1965). Theory of Computing (Organization ACM, New York,
[30] Ernesto Estrada, The Structure of Complex Networks: NY, 2001), pp. 50–59.
Theory and Applications (Oxford University Press, [48] C. Beck and F. Schögl, Thermodynamics of Chaotic
New York, NY, 2011). Systems: An Introduction (Cambridge University Press,
[31] Stanley Wasserman and Katherine Faust, Social Network Cambridge, England, 1993).
Analysis. Methods and Applications (Cambridge [49] P. Erdős and A. Rényi, On the Evolution of Random
University Press, Cambridge, England, 1994). Graphs, in Publication of the Mathematical Institute of
[32] Duncan J. Watts, Peter Sheridan Dodds, and M. E. J. the Hungarian Academy of Sciences (Hungarian Academy
Newman, Identity and Search in Social Networks, of Science, Budapest, Hungary, 1960), pp. 17–61.
Science 296, 1302 (2002). [50] D. J. Watts and S. H. Strogatz, Collective Dynamics of
[33] M. E. J. Newman, Spread of Epidemic Disease on ‘‘Small-World’’ Networks, Nature (London) 393, 440
Networks, Phys. Rev. E 66, 016128 (2002). (1998).
[34] Zoltan Zimboras, Mauro Faccin, Zoltan Kadar, James [51] Mathew Penrose, Random Geometric Graphs (Oxford
Whitfield, Ben Lanyon, and Jacob Biamonte, Quantum University Press on Demand, Oxford, England, 2003),
Transport Enhancement by Time-Reversal Symmetry Vol. 5.
Breaking, Sci. Rep. 3, 2361 (2013). [52] Alain Barrat and M. Weigt, On the Properties of
[35] J. Kempe, Quantum Random Walks: An Introductory Small-World Network Models, Eur. Phys. J. B 13, 547
Overview, Contemp. Phys. 44, 307 (2003). (2000).

041007-8