Quantum Stochastic Walks Overview

• Quantum Stochastic Walks are defined by a GKSL master equation that interpolates between coherent (unitary) and incoherent (classical) dynamics.
• They exhibit diverse transport regimes—from ballistic to diffusive behavior—by tuning the interplay between quantum interference and dissipation.
• Implemented via vectorized superoperator propagation, QSWs are key to modeling quantum transport, ranking algorithms, and quantum neural network protocols.

A quantum stochastic walk (QSW) is a quantum dynamical process on a graph defined by a Gorini–Kossakowski–Sudarshan–Lindblad (GKSL) master equation that interpolates continuously between unitary quantum walks and classical random walks, incorporating both coherent and incoherent (dissipative) dynamics. QSWs exhibit a rich phenomenology not found in purely classical or quantum walks, including regimes of ballistic, diffusive, and super-diffusive transport, and capture key features of open quantum systems. They are widely used for modeling quantum transport, ranking algorithms on graphs, and as substrates for quantum machine-learning protocols.

1. Mathematical Framework and Definition

A QSW on a finite or countably infinite graph $G=(V,E)$ is formulated as an open-system evolution of a density matrix $\rho(t)$ subject to a GKSL master equation: $$\frac{d\rho}{dt} = -i(1-\omega)[H,\rho] + \omega \sum_k \Big( L_k\,\rho\,L_k^\dagger - \frac{1}{2}\{L_k^\dagger L_k,\,\rho\} \Big)$$ where:

• $H$ is the Hermitian Hamiltonian encoding coherent hopping (typically the adjacency matrix or Laplacian of the underlying graph),
• $\{L_k\}$ are Lindblad (jump) operators modeling stochastic, environment-induced processes (e.g., incoherent jumps, dephasing, loss),
• $\omega \in [0,1]$ controls the interpolation between unitary ($\omega=0$) and classical ($\omega=1$) regimes.

The choice of $\{L_k\}$ determines the specific dissipative processes; commonly, $L_{nm} = |n\rangle\langle m|$ models jumps from node $m$ to node $n$. The vectorized form and associated superoperator formalism allow efficient simulation as: $|\rho(t)\rangle = e^{t F} |\rho(0)\rangle$ with $F$ constructed from $H$ and $\{L_k\}$ (Domino et al., 2016, Glos et al., 2018, Govia et al., 2016, Matwiejew et al., 2020).

2. Transport Regimes and Asymptotic Behavior

QSWs capture a continuum of transport regimes on graphs, governed by the asymptotic scaling of the second moment (mean squared distance) $\mu_2(t)$:

• Classical random walk ($\omega=1$): $\mu_2(t) \propto t$ (diffusive).
• Unitary quantum walk ($\omega=0$): $\mu_2(t) \propto t^2$ (ballistic).
• Quantum stochastic ($0\leq \omega <1$): For global dissipation, the ballistic scaling ($\mu_2(t)\sim 2(1-\omega)^2 t^2$) persists for any $\omega<1$: $$\alpha(\omega) = \begin{cases} 2 & 0 \leq \omega <1 \ 1 & \omega=1 \end{cases}$$ This non-analytic jump in scaling exponent $\alpha$ makes it a robust coherence measure: ballistic spreading ($\alpha=2$) signals quantum interference; diffusive scaling ($\alpha=1$) reflects classicality (Domino et al., 2016, Domino et al., 2017).

Local interaction models (one jump operator per edge) typically destroy long-range coherence in the $\omega\to1$ regime, returning to classical diffusion, while global dissipators can preserve super-diffusive or even ballistic scaling despite strong dissipation, provided the correction for “spontaneous moralization” is applied (Domino et al., 2017).

3. Dissipator Schemes and Directed-Graph Effects

QSWs are highly flexible in the choice of dissipative structure:

• Local environment: Lindblad operators $L_{v\to w}=|w\rangle\langle v|$ per edge, producing classical-like relaxation on strongly connected digraphs.
• Global environment: Single Lindblad $L=A$ (adjacency), enabling persistent ballistic transport, but leading to the “spontaneous moralization” effect where extra transitions are induced between parents of common children.
• Non-moralizing construction: Enlarged Hilbert space with corrected dissipator to exactly preserve original directed-graph topology while retaining ballistic or super-diffusive scaling (Domino et al., 2017, Glos et al., 2017, Glos et al., 2018).

The limiting behavior depends on both graph connectivity and dissipator structure:

• Relaxing (unique stationary state) when local dissipators are used on strongly/weakly connected digraphs,
• Convergent but not relaxing (multiple stationary states) in the global dissipator on undirected graphs,
• Non-convergent (cycle or oscillatory modes) in uncorrected global dissipators on certain directed graphs,
• Digraph observance (stationary support within graph sinks) restored by non-moralizing schemes (Glos et al., 2017).

4. Information Measures and Entropy Growth

QSWs display distinctive behavior in entropy production:

• Von Neumann entropy: For a QSW on networks, $S_{\text{vn}}(t;\alpha)$ exhibits long-time logarithmic growth:

$$S_{\text{vn}}(t;\alpha) \simeq d_I(\alpha)\ln t + \text{const}$$

where $d_I(\alpha)$ interpolates between the classical information dimension $d_s/2$ and an enhanced quantum value close to the spectral dimension $d_s$ for the network. Thus, QSWs generalize the notion of information dimension and allow entropy-based characterization of quantum transport universality classes (Schijven et al., 2014).

Entropy scaling is validated by exact solution on chains and fractals, and compared to hierarchy-equations-of-motion simulations.

5. Discrete-Time Quantum Stochastic Walks

Discrete-time QSWs (DTQSW) interpolate between unitary quantum walks and classical random walks via Kraus-operator-based completely positive maps: $$\mathcal{T}[\rho] = (1-p)\, U \rho U^\dagger + \frac{p}{2}\Big((I_c\otimes T)\rho (I_c\otimes T)^\dagger + (I_c\otimes T^\dagger)\rho (I_c\otimes T^\dagger)^\dagger\Big)$$ where $p$ tunes between quantum and classical evolution, and recurrence/return probabilities can show nontrivial, even nonmonotonic, dependence on $p$. For some DTQSWs with certain coin parameters, small amounts of classical randomness can lower the quantum recurrence probability below even the fully classical limit, a phenomenon tied to the subtle interplay between coherence and decoherence (Stefanak et al., 15 Jan 2025, Schuhmacher et al., 2020).

6. Simulation, Physical Realizability, and Applications

QSWs are efficiently simulated via vectorized superoperator propagation on sparse matrices, as implemented in QSWalk.jl, QSW_MPI, and QSWalk (Mathematica). For large graphs, Krylov-subspace matrix exponentiation and parallelization are essential, with tens of thousands of vertices tractable (Glos et al., 2018, Matwiejew et al., 2020, Falloon et al., 2016).

Physical realization of general QSW dynamics is fundamentally constrained. Not all Lindblad models can be microscopically engineered with standard weak-coupling, secular-approximation bath models; only certain classes of dissipators (typically involving local dephasing or decay) can be realized without fine-tuned reservoir spectral densities or complete diagonalization of the system Hamiltonian. Some QSWs, despite being well defined mathematically, are not physically implementable with straightforward open-system quantum devices. Experimental digital quantum simulation via trajectory-based unraveling (e.g., on superconducting or ion platforms) or driven stochastic protocols (e.g., photonic chips generating Haar-uniform randomness) can bypass these restrictions for a large class of graphs (Govia et al., 2016, Taketani et al., 2016, Tang et al., 2021).

Applications span environment-assisted quantum transport, graph ranking (quantum PageRank resolves classical degeneracies and can accelerate convergence), quantum neural networks with improved robustness and generalization, quantum simulation of stochastic resetting, and photonics-based disorder-randomness generators (Benjamin et al., 2022, Wang et al., 2021, Wald et al., 2020, Tang et al., 2021).

7. Physical Phenomena in Disordered and Biased Environments

QSWs on rings or in random environments illustrate quantum-enhanced localization, diffusive, and oscillatory regimes. Coherent hopping can enhance effective disorder, causing non-monotonic delocalization transitions and nontrivial spectral gap closing. The critical thresholds for underdamped relaxation (onset of oscillatory decay) are shifted due to quantum effects, as in the quantized Sinai–Derrida model (Avnit et al., 2023).

The interplay of quantum coherence, dissipation, bias, and disorder in these models reveals new transport phenomena absent in either purely quantum or classical models, including parametrically tunable crossover between localization and delocalization, and scaling of relaxation times with system parameters. This makes QSWs a central framework for studying open-quantum-network physics, algorithmic search, and quantum statistical mechanics.