---
title: 'Cyclic Quantum Walks: Dynamics & Topology'
url: https://www.emergentmind.com/topics/cyclic-quantum-walks-cqws
type: topic
---

# Cyclic Quantum Walks: Dynamics & Topology

Searching arXiv for recent and foundational CQW papers to ground the article.
Cyclic quantum walks (CQWs) are discrete-time quantum walks defined on finite cycle graphs with periodic boundary conditions, typically implemented as coined walks on a ring of \(N\) sites. In their standard form, the Hilbert space factorizes into a position space on the cycle and a finite-dimensional coin space, and the one-step dynamics consists of a coin operation followed by a conditional shift around the ring. Within this seemingly minimal geometry, CQWs support a broad range of phenomena: exact recurrences and periodic limiting distributions, maximally entangled single-particle states, disorder-induced resilience and revival, flat bands and topological phase transitions, robust edge states on finite rings, efficient quantum-circuit realizations, and programmable photonic implementations on cyclic, cylindrical, and toroidal lattices [2410.12710], [2301.04501], [2507.17250], [2506.19024].

## 1. Formal definition and principal variants

A standard coined CQW on an \(N\)-cycle acts on a composite Hilbert space \( \mathcal{H} = \mathcal{H}_c \otimes \mathcal{H}_p \), where the coin space is usually two-dimensional and the position space is spanned by the cycle vertices \(\{|x\rangle : x=0,1,\dots,N-1\}\). A single time step is typically written as
\[
U = S(C \otimes I_p),
\]
where \(C\) is a coin unitary and \(S\) is a conditional shift implementing motion clockwise or anticlockwise modulo \(N\) [2410.12710], [2005.02447], [2301.04501].

For two-state walks, a common shift operator is
\[
S = \sum_{x=0}^{N-1}\left(|0\rangle\langle 0|\otimes |x+1 \bmod N\rangle\langle x| + |1\rangle\langle 1|\otimes |x-1 \bmod N\rangle\langle x|\right),
\]
or an equivalent convention with the two coin states associated to opposite directions [2005.02447], [2008.00316], [2412.00536]. The position probabilities at time \(t\) are obtained from \(|\psi_t\rangle = U^t |\psi_0\rangle\) by summing over coin states [2005.02447].

Several CQW generalizations recur in the literature. One is the discrete-time coined walk on an \(N\)-cycle with a one-parameter coin
\[
C(\rho)=
\begin{pmatrix}
\sqrt{\rho} & \sqrt{1-\rho}\\
\sqrt{1-\rho} & -\sqrt{\rho}
\end{pmatrix},
\]
which includes the Hadamard coin at \(\rho=1/2\) [2410.12710]. Another uses a general \(SU(2)\) or \(U(2)\) coin with amplitude and phase parameters, allowing fine control of spectral degeneracies and quasienergy bands [2412.00536], [2008.00316], [2301.04501].

A distinct branch replaces the conventional single-shift step by more structured protocols. The step-dependent cyclic quantum walk of the topological literature uses a coin \( \hat C_2(\theta,T)=e^{-i(T\theta/2)\sigma_y} \) and a conditional shift \(\hat S\), producing momentum-space blocks
\[
U(k') = e^{-i(2\pi/N)k' \sigma_z} e^{-i(T\theta/2)\sigma_y},
\]
with \(T\) acting as a step-dependence parameter [2507.17250]. The single-coin split-step cyclic quantum walk (SCSS-CQW) employs
\[
U_{\mathrm{SCSS}} = S_+ C_\gamma S_- C_\gamma,
\]
so that one and the same coin is applied twice within each step [2603.07701].

Other variants enlarge the internal space. “Lively quantum walks on cycles” introduce a qutrit coin and a long-range branch of jump length \(a\), with shift
\[
S(n,a)=\sum_x \left(|0\rangle\langle 0|\otimes |x-1\rangle\langle x| + |1\rangle\langle 1|\otimes |x+1\rangle\langle x| + |2\rangle\langle 2|\otimes |x+a\rangle\langle x|\right),
\]
thereby embedding long-range motion directly into the walk definition [1512.02802]. The Möbius quantum walk adds a separate rotation space and a coin-conditioned rotation per step, introducing a Möbius factor \(\alpha = N\theta/(2\pi)\) that modifies spectral degeneracies and limiting distributions [1706.04817]. In Szegedy’s formalism, CQWs arise when the columns of a transition matrix are related by cyclic permutations, yielding efficient quantum circuits for directed and weighted cyclic Markov chains [1609.00173].

This diversity suggests that “CQW” is best understood as a family of finite-ring quantum-walk constructions rather than a single model. What unifies them is the cyclic geometry and the resulting discrete quasi-momentum structure.

## 2. Fourier representation, spectra, and recurrence structure

The discrete Fourier basis on the cycle,
\[
|k\rangle = \frac{1}{\sqrt N}\sum_x e^{ikx}|x\rangle,
\]
or its equivalent discrete form \(k=2\pi \ell/N\), block-diagonalizes the walk into \(2\times 2\) coin-sector problems for two-state CQWs [2410.12710], [2005.02447], [2301.04501]. In the clean unit-jump case one obtains
\[
U(k)=D(k)C,\qquad D(k)=\mathrm{diag}(e^{-ik},e^{ik}),
\]
which yields bounded group velocities and a linear physical time cone [2410.12710].

For the coin family \(C(\rho)\), one finds
\[
\sin \omega(k) = -\sqrt{\rho}\sin k,
\]
up to sign convention, and \( |v_g| \le 1 \), consistent with ballistic spreading on the clean cycle [2410.12710]. In the step-dependent topological CQW, the quasienergies are
\[
E(k)=\pm \arccos[\cos k \cdot \cos(T\theta/2)],
\]
with band closings determined by \( \cos k \cos(T\theta/2)=\pm 1 \) [2507.17250]. In the SCSS-CQW, the split-step dispersion becomes
\[
E_\pm(k)=\pm \cos^{-1}\!\left[\cos k (\cos\theta)^2-(\sin\theta)^2\right],\qquad \theta=D\gamma/2,
\]
giving a different route to flat bands and topological transitions [2603.07701].

Finite cyclic geometry strongly constrains recurrence. In the clean Hadamard CQW, maximally entangled single-particle states recur with period \(4\) on \(N=4\) and \(N=8\) cycles [2410.12710], [2301.04501]. In “order from chaos” constructions, periodicity is defined by the exact condition \(U^N=I\), equivalently by requiring all eigenvalues to be roots of unity [2008.00316]. For the \(n\)-cycle with coin
\[
C_2(\rho,\alpha,\beta)=
\begin{pmatrix}
\sqrt{\rho} & \sqrt{1-\rho}e^{i\alpha}\\
\sqrt{1-\rho}e^{i\beta} & -\sqrt{\rho}e^{i(\alpha+\beta)}
\end{pmatrix},
\]
the quasi-momentum blocks are \(U_k=\mathrm{diag}(e^{i\phi_k},e^{-i\phi_k})C_2\), and periodicity reduces to rational commensurability of the associated eigenphases [2008.00316].

The long-range qutrit CQW has its own spectral periodicity mechanism. In that model, degeneracies occur precisely when \(e^{ika}=1\), and the time-averaged limiting distribution has period \(\gcd(a,n)\) whenever \(\gcd(a,n)>1\) [1512.02802]. In the Möbius quantum walk, the parameter \(\alpha\) controls degeneracy conditions of the form
\[
k = \frac{nN}{2} - k' \pm \alpha,
\]
and the limiting distribution is uniform for all \(N\) whenever \(\alpha \neq m/2\) [1706.04817].

These results establish a general pattern: CQW recurrence is not a generic feature of finite unitarity alone, but a consequence of arithmetic commensurability across the discrete \(k\)-sectors. This suggests why parity, cycle size, and coin phases repeatedly appear as decisive parameters.

## 3. Entanglement generation and single-particle structure

A major recent direction treats CQWs as generators of single-particle entanglement (SPE) between the coin and position degrees of freedom. For a pure state \(\rho(t)=|\psi(t)\rangle\langle\psi(t)|\), the coin-reduced state is \(\rho_c(t)=\mathrm{Tr}_p[\rho(t)]\), and SPE is quantified by the von Neumann entropy
\[
E(t)= -\mathrm{Tr}[\rho_c(t)\log_2 \rho_c(t)].
\]
For a qubit coin, \(E(t)=1\) corresponds to a maximally entangled single-particle state (MESPS) [2410.12710], [2301.04501].

A central result is that, for the clean Hadamard CQW on any \(N\)-cycle, the state at \(t=1\) is a single-particle Bell state for arbitrary \(\theta\) when \(\phi=\pi/2\), so that \(E_{\mathrm{av}}(1)=1\) [2410.12710]. More generally, with a balanced coin \( \rho=1/2 \), if the initial phase satisfies
\[
\gamma+\phi \in \{\pi/2,\,3\pi/2\},
\]
then the one-step CQW generates a MESPS at \(t=1\) for any cycle size \(k\) and any \(\theta\) [2301.04501].

On \(C_4\) and \(C_8\), recurrent MESPS can be produced using a single fixed coin applied at every step. For the Hadamard coin on \(C_4\) with \(\phi=\pi/2\), MESPS occur at \(t=1,5,9,\dots\), i.e. \(t\equiv 1 \pmod 4\), while on \(C_8\) the same choice yields MESPS at \(t\equiv 1 \pmod{12}\) [2301.04501]. The Fourier coin and other balanced coins can produce analogous recurrent entanglement patterns, including periods \(3\), \(4\), and \(12\), depending on the coin phases [2301.04501].

The explicit \(C_4\) Hadamard evolution illustrates the mechanism. Starting from a localized state with \(\phi=\pi/2\), the one-step state is
\[
|\psi(1)\rangle = \sqrt{2}\alpha |3,0\rangle + \sqrt{2}\beta |1,1\rangle,
\]
with orthogonal positions carrying the two coin components. The position trace therefore removes cross terms, giving \(\rho_c=I_2/2\) and hence \(E(1)=1\) [2301.04501]. At later times, entanglement oscillates: \(E(2)\approx 0.557\), \(E(4)=0\), and \(E(5)=1\) [2301.04501].

The same work also identifies effective-single and two-coin schedules on \(k\in\{3,4,5,8\}\) that produce recurrent MESPS when a strictly single-coin ordered walk is unavailable. Examples include \(HII\ldots\), \(IHI\ldots\), \(HHX\ldots\), and \(HXHX\ldots\), with MESPS periods \(4\), \(6\), \(9\), \(12\), and \(15\) depending on \(k\) and schedule [2301.04501].

A plausible implication is that CQWs provide a controlled route to hybrid entanglement engineering in bounded Hilbert spaces, with arithmetic recurrence replacing the asymptotic or many-body mechanisms common in larger systems.

## 4. Disorder, noise, resilience, and dynamical transitions

Disorder in CQWs has been studied in several distinct senses: phase disorder, coin disorder, position disorder, and static site noise. These perturbations are not equivalent.

In the disorder study on odd and even cyclic graphs, phase disorder is introduced through a phase-decorated shift, coin disorder through random site- or time-dependent coin parameters, and position disorder through a random jump length \(J(t)\) drawn from a Poisson distribution with mean \(\lambda\) [2410.12710]. The principal findings are sharply differentiated.

Phase disorder, whether static or dynamic, leaves the MESPS at \(t=1\) exactly intact for any initial state. The evolved state picks up local phases on distinct positions, but tracing out position yields
\[
\rho_c(1)=\frac12 I_2,
\]
independent of \(\theta\), \(N\), and the disorder variable [2410.12710]. Coin disorder shows a closely related but more restricted immunity: for the phase-symmetric initial coin states
\[
|q_\pm\rangle = \frac{|0\rangle \pm i|1\rangle}{\sqrt2},
\]
MESPS at \(t=1\) remain exact for any static or dynamic coin disorder strength [2410.12710].

Beyond \(t=1\), small phase or coin disorder causes only insignificant reduction of SPE at early times; numerically, \(\delta \approx 0.2\) or \(\omega \approx 0.2\) still show resilience on \(N=4\) [2410.12710]. Moderate or strong disorder can even enhance entanglement at specific times, and all three disorder types can revive SPE from zero when the clean walk is separable, for example at \(t=4,8,12\) on \(N=4\) [2410.12710].

Position disorder is qualitatively different. Randomizing the jump length modifies the shift operator itself, breaks odd-even parity on even cycles, distorts the physical time cone, destroys clean recurrences, and makes SPE more vulnerable [2410.12710]. Yet its long-time behavior is unexpectedly robust in another sense: \(E_{\mathrm{av}}(t)\) saturates to a fixed value at large \(t\), irrespective of \(\lambda\), although the limiting value depends on \(N\) [2410.12710]. This combination of short-time vulnerability and long-time stabilization is one of the most distinctive results in the current CQW disorder literature.

A related but separate noise analysis studies static site-phase noise in a homogeneous cyclic graph using a three-parameter \(SU(2)\) coin [2412.00536]. There the noisy step is
\[
U_{\mathrm{noise}}(\phi)= [D_{\mathrm{noise}}(\phi)\otimes I_{\mathrm{coin}}]U,
\]
with \(D_{\mathrm{noise}}=\sum_s e^{i\varphi_s}|s\rangle\langle s|\) and \(\varphi_s\) uniformly distributed in \([-\phi,\phi]\). The mean squared displacement is fit as \(\langle x^2(m)\rangle \approx \alpha m^\beta\), revealing a progression from nearly ballistic behavior at \(\phi=0\) with \(\beta\approx 1.996\), to super-diffusive behavior at \(\phi=\pi/10\) with \(\beta\approx 1.794\), to diffusive behavior near \(\phi=\pi/3\) with \(\beta\approx 1.032\), and to sub-diffusive behavior at \(\phi=\pi\) with \(\beta\approx 0.419\) for \(N=128\) and Hadamard coin parameters \(\{\gamma,\theta,\phi\}=\{\pi/4,0,0\}\) [2412.00536]. On finite cycles the mean squared displacement saturates when the number of steps exceeds the graph size by about an order of magnitude [2412.00536].

That work also reports a strong empirical correspondence between the average eigenstate participation ratio and mean-squared-displacement behavior, with low participation ratio correlating with localization and high participation ratio with delocalization [2412.00536]. This suggests that CQW transport under quenched disorder can often be diagnosed spectrally rather than through long-time simulation.

## 5. Topological CQWs, flat bands, and edge states

Recent work places CQWs within Floquet topological band theory. In step-dependent CQWs on cyclic graphs, the effective Hamiltonian is defined by \(U=e^{-iH}\) with
\[
H=E(k)\,\hat n(k)\cdot \vec\sigma,
\]
and quasienergies
\[
E(k)=\pm \arccos[\cos k \cos(T\theta/2)].
\]
This framework yields gapped and gapless phases, Dirac-cone-like band closings, topological flat bands, and edge states, all without split-step or split-coin protocols [2507.17250].

The flat-band condition is
\[
\cos(T\theta/2)=0 \iff \theta=(2n+1)\pi/T,
\]
which gives \(E(k)=\pm \pi/2\), independent of \(k\), for any \(N\) [2507.17250]. Rotationally symmetric flat bands, however, occur only when \(N=4n\), because the discrete momenta must include values with \(\cos k=0\) exactly [2507.17250]. Even and odd cycles therefore have qualitatively different spectral possibilities: even cycles admit \(k=\pi\) and richer gap-closing structure, while odd cycles do not [2507.17250].

The corresponding topological invariant is a discretized Zak phase or winding number. For the \(N\)-cycle, the winding number is
\[
\omega_{\theta,T,N} = \sum_{k'=0}^{N-1}\frac{\sin(T\theta/2)}{N[1-\cos^2(2\pi k'/N)\cos^2(T\theta/2)]},
\]
and the Hadamard case \(\theta=\pi/2,\,T=1\) gives \(Z=\pi\) and \(\omega=1\) [2507.17250]. Interfaces between regions with different winding numbers support localized edge states. For example, on \(N=8\) with \(T=2\), choosing \(\theta_A=7\pi/5\) at site \(0\) and \(\theta_B=\pi/3\) elsewhere creates a persistent edge state localized at the interface site [2507.17250]. These edge states remain localized under static disorder \(0<\Delta_s\lesssim 0.2\), dynamic disorder \(0<\Delta_d\lesssim 0.05\), and phase-preserving perturbations that do not change the winding sector [2507.17250].

The SCSS-CQW extends this program in a different topological direction. Its quasienergies satisfy
\[
E_\pm(k)=\pm \cos^{-1}[\cos k(\cos\theta)^2-(\sin\theta)^2],
\]
with flat \(\pi\)-bands at
\[
\gamma=(2n+1)\pi/D,
\]
for which \(U(k)=-I\) and the bands are pinned at quasienergy \(\pi\) [2603.07701]. More strikingly, the protocol yields fractional winding numbers \(\pm 1/2\) and Zak phases \(\pm \pi/2\), rather than integer invariants. For \(D=1\) and \(\gamma=\pi/2\), the winding is numerically \(0.505\) for \(N=3\), \(0.5009\) for \(N=4\), and exactly \(1/2\) for \(N=1000\); for \(D=5\) and \(\gamma=\pi/2\), one obtains the opposite fractional sector \(-1/2\) [2603.07701]. Domain walls between these fractional sectors support edge-localized bound states on finite rings, robust against dynamic and static coin disorder as well as phase-preserving perturbations [2603.07701].

A common misconception is that finite cyclic geometry is too small or too symmetric to support nontrivial topology. The topological CQW results indicate the opposite: the ring geometry discretizes momentum but does not trivialize the Floquet band structure. Instead, it renders parity, cycle divisibility, and finite-size sampling central to the topological classification.

## 6. Implementations, circuit constructions, and related extensions

CQWs have been analyzed both as abstract dynamical models and as concrete hardware primitives.

A universal-gate implementation of the discrete circular walk on \(N=2^m\) sites uses one coin qubit and an \(m\)-qubit position register. One step applies a coin gate followed by a coin-controlled modular decrement or increment on the position register, implemented through cascades of multi-controlled NOT gates [2005.02447]. In the \(m=3\) realization on IBM Q London, corresponding to \(N=8\), the full one-step walker circuit used \(87\) CNOTs, and the average fidelity over all \(8\) initial position states was \(17.42\%\) [2005.02447]. The paper attributes the strong degradation to realistic two-qubit gate errors; a naive estimate using \(p_{\mathrm{CNOT}}\approx 1.38\times 10^{-2}\) gives \((1-p_{\mathrm{CNOT}})^{87}\approx 29.85\%\), above the measured fidelity, indicating additional contributions from readout, decoherence, and mapping overhead [2005.02447]. This result is less a demonstration of scalable CQWs than a quantitative diagnosis of circuit-depth limitations in NISQ superconducting hardware.

Szegedy-type CQWs offer a different circuit model. When the columns of a transition matrix are related by cyclic permutations, the corresponding Szegedy walk can be compiled efficiently using a reflection diagonalization strategy. For circulant transition matrices, the necessary controlled shifts and state-preparation blocks can be implemented with \(O(\mathrm{polylog}\,N)\) resources, independently of sparsity [1609.00173]. This extends CQW circuit design to directed and weighted cycles, not only conventional coined walks.

Programmable photonic CQWs provide a more scalable experimental route. A reciprocal-space platform based on spatial light modulators implements arbitrary translationally invariant unitaries by discrete sampling of the Brillouin zone, thereby enforcing cyclic, cylindrical, or toroidal boundary conditions directly in momentum space [2506.19024]. In one dimension, the lattice size is set by the number of sampled \(k\)-points; in two dimensions, sampling in both directions yields toroidal topologies, while periodicity in one direction and openness in the other yields cylindrical topologies [2506.19024].

That platform realizes up to \(90\) steps in a single synthesized unitary for a \(1\)D cycle with \(N=15\), with partial refocusing at \(t=23\) and similarity \(S=0.92\pm 0.05\) between experiment and theory [2506.19024]. For \(N=3\), exact recurrences were engineered with periods \(2\) and \(3\), giving \(S\approx 0.9998\), while a near-recurrence at \(t\approx 253\) with the default coin yielded \(S\approx 0.97\) [2506.19024]. The same platform demonstrates wavepacket breathing, topology-dependent trajectories on a torus, and dimensional reduction of a cylindrical \(2\)D walk to an effective \(1\)D walk with a high-dimensional coin [2506.19024].

Photonic CQW relevance also appears in the entanglement and disorder studies, which explicitly identify polarization as a coin degree of freedom and time bins, OAM, path, or spatial modes as position encodings [2410.12710]. In those settings, phase disorder can be introduced by phase plates or fast modulators, coin disorder by changing waveplate angles, and position disorder by reconfigurable routing or delay lines [2410.12710].

Beyond implementation, several nonstandard CQW directions broaden the conceptual scope. The time-dependent-coin construction that matches arbitrary classical random walks on cycles shows that a unitary coined CQW can reproduce any target vertex-distribution sequence at measurement times, provided the local coins are allowed to vary in space and time [2103.06463]. The percolated-Grover analysis on arbitrary graphs shows that cyclic local shift operators can produce nonstationary asymptotics with periods dividing \(6\) when an edge-3-coloring exists, and that cyclic shifts avoid the trapped states that plague reflecting-shift walks [1812.02519]. Anyonic CQWs on cycles incorporate braiding with stationary anyons and show mixing behavior on finite rings that resembles standard Hadamard walks more than classical random walks, despite strong decoherence effects on infinite chains [1210.3446].

These extensions indicate that CQWs function as a meeting point for finite-graph quantum transport, Floquet topology, hybrid entanglement, programmable optics, and circuit-level quantum simulation. Their unifying theme is not merely “walking on a ring,” but exploiting the ring’s discrete momentum structure as a controlled arena for interference, periodicity, topology, and robustness.

Source: https://www.emergentmind.com/topics/cyclic-quantum-walks-cqws