Quantum Memristor Meshes

• Quantum memristor meshes are networks of quantum devices exhibiting nonlinear, history-dependent responses that enable neuromorphic quantum computing.
• Their operation relies on diverse principles such as floating-gate charge trapping, photonic interferometric feedback, and quantum-dot tunneling modeled by governing equations.
• Mesh architectures offer scalable, reconfigurable topologies that support in-memory processing, robust entanglement, and efficient quantum neural-network implementations.

Quantum memristor-based meshes are networks of interconnected quantum memristive elements—devices whose quantum-coherent transport and internal memory dynamics admit nonlinear, history-dependent responses. These meshes form the hardware substrate for neuromorphic quantum computing architectures, combining in situ learning with quantum resources such as coherence, entanglement, and nonlocal connectivity. The diversity of available realizations, encompassing topological insulator phases, photonic interferometric feedback, quantum-dot charge storage, and Josephson SQUID circuits, enables a variety of design strategies. Key metrics include memristive hysteresis, resistance levels, memory retention, quantum-coherent transport properties, and, in mesh networks, the entanglement structure and scalability.

1. Quantum Memristor Realizations and Principle of Operation

Quantum memristors generalize classical resistive memory to systems where the nonvolatile state variable influences and is influenced by quantum transport. Three archetypes have been experimentally and theoretically characterized:

• Topological Field-Effect Memristors: Inverted InAs/Ga$_{0.68}$In$_{0.32}$Sb/InAs trilayer quantum wells function in the quantum spin Hall regime, exhibiting dissipationless, phase-coherent edge channels ("helical edges") within a topological gap ($\Delta_{\mathrm{topo}} \simeq 27$ meV). A robust, tunable memristor emerges via intrinsic floating-gate charge trapping at the oxide/quantum well interface, controllably shifting threshold voltages and toggling the transport between helical edge- and incoherent bulk-dominated states (Meyer et al., 21 Nov 2025).
• Photonic Quantum Memristors: Single-photon interferometric circuits, integrating Mach–Zehnder interferometers with internal phase shifters, realize memristive behavior as a nonlinear feedback loop: the internal phase variable $\phi_t$ (affecting optical reflectivity and thus output signal) is adaptively updated in response to the history of measurement outcomes. This explicit feedback emulates the memory effect over quantum states of light and serves as a “quantum synaptic widget” (Selimović et al., 25 Apr 2025).
• Quantum-Dot Memristors: Charge storage in a secondary quantum dot (QD2) capacitively influences the conductance of a primary transport dot (QD1), resulting in history-dependent conductance and stochastic switching. The full dynamics are governed by a Lindblad master equation for the double-dot plus reservoir system, giving rise to nonvolatile, hysteretic I–V loops and quantum-jump statistics (Li et al., 2016).

2. Governing Equations for Memory Dynamics and Transport

All quantum memristor architectures are formulated in terms of (1) an internal state variable (trapped charge, phase, occupation), (2) a transport law coupling quantum evolution and system response, and (3) a memory-update protocol—either through coherent dynamics (e.g., phase-space flow) or stochastic tunneling/measurement-induced jumps.

Examples:

• Topological Field-Effect Memristor:
  • Floating-gate charge-voltage relation:

  $$Q_{\rm trap}(t) = C_{\rm eff}[V_{\rm FG}(t) - V_{\rm TG}(t)]$$

  Resistance state:

  $$V(t) = R_m[V_{\rm FG}(t)] I(t), \quad R_m = \frac{dV}{dI}\Big|_{I\to0}$$

  Switching dynamics:

  $$\frac{dQ_{\rm trap}}{dt} = f(I(t),V_{\rm FG}(t)) - \frac{Q_{\rm trap}}{\tau_{\rm leak}}$$

• Photonic Quantum Memristor:

$$\phi_{t+1} = \phi_t + f(\phi_t,p_2(t)), \quad f(\phi, p) = 2\arccos\sqrt{(1-\eta)\cos^2(\phi/2) + \eta p} - \phi$$

where $p_2(t)$ is the output photon-detection probability in a dedicated port, feeding back to update the internal state.

• Quantum-Dot System:
  • Open-quantum Lindblad evolution for double-dot occupancy ($n_1$, $n_2$), tunneling rates $\gamma_1$, $\gamma_2$, and capacitive shift $E_C$ (memory variable $x=n_2$):

  $$I(V,n_2) = \begin{cases} G_\mathrm{ON}(V) V & n_2=0 \ G_\mathrm{OFF}(V) V & n_2=1 \end{cases}$$

  with stochastic jump dynamics in $n_2$ under bias.

These formulations reveal the duality of memory: encoded either in floating charges, quantum phases, or occupation numbers, and always exerting dynamic feedback on quantum-coherent transport channels.

3. Mesh Architectures: Topology, Coupling, and Scalability

Large-scale quantum memristor meshes are naturally modeled as graphs $G = (V, E)$, with vertices representing memristive nodes and edges the quantum-coherent links (via edge channels, optical couplers, interdot capacitances, or inductive SQUID couplers). Important architectural aspects are:

• Interconnect Parasitics and Coherence: For topological (QSHI-based) systems, coherence preservation requires inter-node spacings not to exceed the phase coherence length $L_\phi$, with low-loss connections such as epitaxial metal or graphene (Meyer et al., 21 Nov 2025). In photonic and superconductor-based circuits, mode-matching and minimization of cross-talk/parasitic inductances are critical (Kumar et al., 2022, Selimović et al., 25 Apr 2025).
• Coupling Strength and Nonlinearity: Optical meshes are reconfigurable through variable beam-splitter strengths; quantum-dot meshes leverage capacitive cross-couplings and tunable tunneling rates; superconducting meshes exploit engineered inductances and flux biasing. Tuning the mesh to the edge of chaos (in the sense of reservoir computing theory) can maximize nonlinear memory effects (Selimović et al., 25 Apr 2025).
• Node Count and Resource Scaling: The scaling of modes, couplers, and phase shifters varies by realization. In photonic architectures, a fully connected $N$-node mesh uses $3N$ modes, $3N$ MZIs, and up to $O(N^2)$ inter-node couplers. Sparser random graphs can provide improved computational capacity per resource (Selimović et al., 25 Apr 2025).

4. Quantum Correlations and Entanglement in Meshes

Quantum meshes enable not only classical memory and learning, but also multipartite entanglement and nontrivial quantum correlations between nodes. This is most explicit in superconducting SQUID-based meshes:

• Hamiltonian Structure: Each node is a harmonic mode (in CA–SQUID circuits, with suppressed Josephson energy at half flux quantum), coupled to others via $g_{jk}(a_j+a_j^\dagger)(a_k+a_k^\dagger)$.
• Entanglement Measures: Reduced density matrices, partial transposition, and negativity quantify bipartite and multipartite entanglement (e.g., tripartite negativity $\mathcal{N}_3$). In triangular meshes, maxima in entanglement and form factor are correlated; in linear chains, their evolution is out of phase (Kumar et al., 2022).
• Monogamy and Scalability: Quantum monogamy constraints bound the distribution of entanglement across the mesh, affecting the balance between local memory and network-wide quantum correlations. Generalizations to larger graphs entail intricate monogamy and negativity structures, with design implications for quantum neural networks.

5. Performance Metrics and Operational Regimes

Experimentally extracted and theoretically postulated metrics include:

Device Type	ON/OFF Ratio	Coherence/Quantization	Nonlinearity Origin	Scalability
Topological QSHI	$\lesssim$22.5	$R_{\rm edge}\to h/2e^2$	Charge-trapping at floating gate	III–V wafer, $\sim$sub-$\mu$m lithography
Photonic	(benchmarked, $N=1$)	Single-photon, unitary Q	Interferometric feedback	$O(N)$ modes, $O(N)$ MZIs, $O(N^2)$ couplers
Quantum-dot	$E_C\gg kT$, tunable	Stochastic quantum jumps	Tunneling-induced switching	Dense planar/3D arrays, cross-capacitance
SQUID (supercond.)	Implied (via $\Gamma_j$)	Coupled oscillator modes	Time-dependent relaxation	Graph topology, monogamy, inductive couplers

• Memory Retention: Retention times $\tau_\mathrm{leak}$ (floating gate), tunable by spectral gap engineering (quantum dots), or flux-biasing (SQUIDs), set the volatility.
• Switching Speed and Endurance: For topological QSHI, $ms$–$s$ range for charge trapping; photonic feedback limited by measurement rate; quantum-dot transitions in the $Hz$–$GHz$ regime. Endurance is projected from analogous classical architectures.
• Error Resilience: Topological protection in QSHI ensures robustness to nonmagnetic disorder; memory states stored as trapped charges or phase variables are long-lived at low T.

6. Applications and Prospects for Quantum Neuromorphic Networks

Quantum memristor meshes are targeted as substrates for neuromorphic quantum computing, unifying in-memory computation with quantum-enhanced resources:

• Co-localization and In-Memory Processing: Coincident memory (floating charge/state variable) and quantum-coherent signal processing within each node facilitate local learning and non-von Neumann architectures (Meyer et al., 21 Nov 2025).
• Reservoir and Neural Network Computing: Meshes of quantum memristors simulate recurrent neural networks, with demonstrated performance gains in nonlinear prediction and time-series tasks for single-node devices (Selimović et al., 25 Apr 2025). Scaling up to larger graphs is predicted to increase memory capacity and nonlinear transformation power, subject to the complexities of optical loss, topological control, and quantum noise.
• Entanglement-Aware Design: Triangular and higher-connectivity mesh motifs, which maximize multipartite entanglement and uniform nonlinearity, are structurally favored for reservoir computing and quantum neural-network primitives (Kumar et al., 2022).
• Scalability and Integration: III–V quantum wells, glass photonics, silicon quantum dots, and superconducting circuits all support scalable fabrication, though each presents distinct engineering tradeoffs in coherence, addressability, and classical–quantum interfacing.

A plausible implication is that the further engineering of inter-node coupling, distributed coherence, and resource overheads will define the frontier for practical quantum neuromorphic mesh processors, leveraging the dual capabilities of memory and quantum-coherent processing in a single, tunable architecture.