Speaker
Description
Collective neutrino flavor oscillations driven by neutrino–neutrino interactions are a quantum many-body dynamics problem amenable to digital quantum simulation, but near-term superconducting hardware is constrained by circuit depth, two-qubit errors, and topology-dependent routing overhead. Recent three-flavor circuit constructions (including qubit and qutrit encodings) demonstrate small-system feasibility while indicating that recoverable neutrino observables can be dominated by compilation choices rather than the underlying Hamiltonian model. We propose a compilation digital twin that learns hardware-aware strategy selection for three-flavor collective-neutrino time-evolution circuits under a fixed resource budget. Given physics descriptors (system size, coupling regime, evolution time, and target observables such as flavor-survival probabilities and flavor–flavor correlators) and backend descriptors (connectivity and calibrated error rates), the twin predicts an action set (including initial qubit layout, routing/optimization settings, and Trotter term ordering/grouping) to minimize a task objective combining two-qubit gate count, depth, and noisy-simulation observable error relative to ideal evolution. The digital twin is trained on offline compilation sweeps labeled by a task-driven score. We benchmark twin-chosen compilation against standard heuristics across multiple IBM backends and out-of-distribution coupling/time regimes, showing improved preservation of collective-oscillation signatures at matched depth and reduced backend-dependent variability, bridging many-body neutrino simulation proposals to executable NISQ experiments.
| I read the instructions above | Yes |
|---|