Quantum Reservoir Computing with Spin Chain Dynamics

Part of ARPA Quantum Logical Systems — QONDRA  ·  qondra@arpacorp.net

---

What this is

spintronic-qrc uses disordered XXZ spin chains as fixed quantum reservoirs for temporal machine learning. The reservoir dynamics belong to the same family of spin Hamiltonians that describe antiferromagnetic spintronic materials — including Mn₃Sn, which spinq-vqe simulates with VQE on a Kagome lattice. Only a classical Ridge output layer is trained.

Two parallel research threads:

• Spin-chain QRC: encode inputs → evolve reservoir → Pauli readout → Ridge regression on NARMA-10 and Mackey-Glass benchmarks.
• Spintronic gate set: PennyLane gates mapped from Larmor precession, exchange coupling, spin-orbit torque, and damping.

Structure

spintronic-qrc/
├── src/spintronic_qrc/
│   ├── reservoir.py   # XXZ chain Hamiltonian + device helpers
│   ├── pipeline.py    # End-to-end encode → evolve → readout → Ridge QRC loop
│   ├── encoder.py     # Input injection (local / global)
│   ├── readout.py     # Pauli expectation observables
│   ├── trainer.py     # Ridge output layer
│   ├── gates.py       # Spintronic-inspired gate set
│   ├── tasks.py       # NARMA-10, Mackey-Glass generators
│   └── utils.py       # Plotting palette
├── notebooks/         # Research notebooks
├── figures/           # Generated plots
├── data/              # Benchmark CSVs
├── benchmarks/        # Scripts vs classical Echo State Networks
├── docs/              # Guides and API reference → docs/README.md
├── OVERVIEW.md        # Research narrative and literature context
└── REFERENCES.md      # Bibliography

Install

Use the workspace venv at the Spintronics program root:

# From Spintronics/ (parent of this repo)
.venv\Scripts\activate          # Windows
source .venv/bin/activate       # Linux / macOS

pip install -e "./spintronic-qrc[dev]"

Optional extras:

Extra	Packages	When
[tuning]	Optuna	Hyperparameter sweeps (N, τ, W)
[open]	QuTiP	Open-system / Lindblad QRC
[crossval]	Qiskit + Aer	Cross-validate dynamics vs PennyLane
[viz]	Plotly	Interactive notebook plots
[notebooks]	dev + tuning + viz	Typical notebook workflow
[all]	everything above	Full research stack

pip install -e "./spintronic-qrc[notebooks]"
pip install -e "./spintronic-qrc[open,crossval]"   # open-system / validation

Requires Python ≥ 3.11. Core: PennyLane 0.39+, JAX, NumPy, SciPy, scikit-learn, Matplotlib.

Notebooks

#	Notebook	Notes
01	01_xxz_dynamics.ipynb	Spin chain time evolution
02	02_qrc_narma10.ipynb	Full QRC pipeline on NARMA-10
03	03_qrc_mackey_glass.ipynb	Chaotic attractor prediction
04	04_memory_capacity.ipynb	Quantum memory capacity
05	05_spintronic_gate_set.ipynb	Custom gates + depth benchmarks
06	06_open_system_qrc.ipynb	QuTiP + dissipation

Tests

pip install -e "./spintronic-qrc[dev]"
pytest spintronic-qrc/tests/ -v

See docs/testing.md for the full guide.

Docs

→ OVERVIEW.md — research narrative, key results, and literature context.
→ docs/README.md — physics background, API reference, notebook guide.
→ REFERENCES.md — full bibliography.

References

See REFERENCES.md for the full bibliography.
Key: Fujii & Nakajima (2017), Nakajima (2021), Dambre et al. (2012), Mujal et al. (2021).

---

License: MIT  ·  Contact: qondra@arpacorp.net