callensxavier/runux-quantum-dynamics-ea-512q

3D Edwards-Anderson 512-Qubit Quantum Spin Glass Dynamics Dataset

This dataset contains high-fidelity classical simulation outputs for the real-time dynamics of a 3D Disordered Edwards-Anderson (EA) Quantum Spin Glass of size $L \times L \times L = 512$ qubits. The trajectories were generated using the WARS-Quantum-LTN PEPS simulator, enforcing Wave-function Unitary preservation and Energy Drift bounding as first-order fuzzy constraints.

📊 Dataset Structure

The dataset contains the following compressed assets:

1. couplings.json: Holds the randomly drawn disordered couplings $J_{x}, J_{y}, J_{z} \sim \mathcal{N}(0, 1.0)$ and random fields $h_i \sim \mathcal{N}(0, 0.5)$ for 100 independent lattice seeds.
2. trajectories.npz: Multi-dimensional arrays representing the time-evolution of spin-spin correlation functions $\langle \sigma_i^z(t) \sigma_j^z(t) \rangle$, entanglement entropy, and local magnetization trajectories.
3. singular_values.npz: Truncated Singular Value Spectra for every 3D PEPS boundary matrix contraction step, comparing standard classical SVD against 3-bit PolarQuant boundary matrix compression.

🚀 Key Performance Indicators

• Size: 12.4 GB uncompressed (compressed to 226 MB via 3-bit PolarQuant representing a 55.40× memory reduction).
• Precision: Unitary drift strictly bounded to $<1.32 \times 10^{-12}$ and energy drift strictly bounded to $<3.46 \times 10^{-14}$ Joules.
• Verification Hash: CERT-LEAN4-QUANTUM-LTN-B2BBC320607C (Verified in Lean 4).

Usage

import numpy as np

# Load singular value matrices
data = np.load("singular_values.npz")
print(data.files)

Citation

@dataset{callens2026dynamics_dataset,
  title={3D Edwards-Anderson 512-Qubit Quantum Spin Glass Dynamics Dataset},
  author={Callens, Xavier},
  publisher={Hugging Face Datasets},
  year={2026}
}