AI and the World's Hard Problems · Quantum physics & quantum computing · arXiv · 2026-10-26
Distributed Quantum-Centric Supercomputing with Multi-Layered Generative Learning for Intermolecular Interactions
Quantum-centric supercomputing (QCSC) methods are limited for large active spaces by qubit counts, circuit depth, sampling cost and large-scale diagonalization. The authors propose a distributed framework that splits molecular problems into parallel subproblems and reconstructs the global solution with a two-step aggregation, plus a multi-layered generative model. For water and formamide dimers it cut the diagonalization dimension by up to 95% and improved energy accuracy by an order of magnitude over conventional SQD, and for a 120-qubit water hexamer it split the problem into six 20-qubit subsystems with results in good agreement with classical reference methods.