QodeX Quantum, NVIDIA, RTX BBN, and U.S. Navy Researchers Forecast Unstructured PDEs on 52-Qubit IBM Hardware
AI-summarised brief · reviewed before publication
A collaborative team comprising QodeX Quantum, the University of South Carolina, the Naval Surface Warfare Center Panama City Division, NVIDIA, and RTX BBN has demonstrated gradient‑free forecasting of nonlinear partial differential equations on unstructured engineering meshes using up to 52 qubits of an IBM Quantum Heron r3 processor. The hybrid quantum reservoir network (QRN) ran circuits up to 24 × the median qubit coherence time and cut forecast error by 21‑41 % versus classical latent linear baselines on physical hardware. By pairing a pretrained graph‑neural‑network autoencoder with a fixed, untrained quantum reservoir split into memory and readout registers, the approach decouples complex boundary geometry from circuit topology. Data‑reuploading rotations, partial‑SWAP couplings, and mid‑circuit measurement/reset create a tunable fading memory that sustains evolution for up to 28 × median T1 and 24 × median T2. In ten‑step autoregressive rollouts on a cylinder‑flow dataset, hardware noise acted as a regularizer, achieving a test NRMSE of 0.257, far better than noiseless simulation (0.707) and a classical persistence baseline (1.126). The work also validates modules of QodeX’s AlphaPsi quantum workflow platform, with code and the full preprint released openly.
💡 Why It Matters
- · Demonstrating reliable PDE forecasting on noisy, unstructured quantum hardware shows a viable path for quantum‑accelerated fluid‑dynamics simulations beyond idealized grids, potentially reshaping high‑fidelity engineering analysis.
