Compressed and navigable electronic manifolds in large Hilbert spaces

This repository provides the source data and reviewer evidence package accompanying the manuscript:

“Compressed and navigable electronic manifolds in exponentially large Hilbert spaces”

This work addresses a central assumption in quantum chemistry and quantum computing: that exponential Hilbert-space growth places strongly correlated electronic-structure problems beyond practical classical computation and into the domain of quantum computers or large high-performance-computing infrastructures.

The manuscript presents evidence that, for the structured ground states examined here, the physically relevant determinant manifold remains highly compressed and navigable. The central result is a FeMoco CAS(54,54) calculation performed on a single desktop workstation (Intel i5-13600K, NVIDIA RTX 4090, 32 GB RAM), recovering a physically consistent wavefunction using 748,000 determinants from a formal determinant space of approximately 3.79 × 10³⁰. This corresponds to an approximately 5.1 × 10²⁴-fold compression of the formal space.

The significance of this result is not that FeMoco has been solved to definitive chemical accuracy against the most advanced multi-GPU DMRG calculations. Rather, it shows that a problem long framed as requiring quantum hardware or HPC-scale resources can be brought into the personal-desktop regime when the accessible electronic manifold is targeted directly.

The DTQW–LICP framework combines structured graph propagation on the Slater determinant graph with local energetic filtering. It is validated against full configuration interaction at machine precision across small benchmark systems and extended through iron–sulfur clusters to the full FeMoco cofactor. The recovered FeMoco wavefunction satisfies electron-number conservation, singlet spin symmetry, and the expected Fe–Mo 3d / bridging sulfide 3p correlated-orbital structure.

Included materials:
- Selected numerical results and convergence trajectories
- Parameter sweeps used in the main text and Supplementary Materials
- Figure-generation scripts and selected source data
- Reviewer evidence files supporting validation and benchmarking

This package is intended to provide transparency, facilitate peer review, and enable independent verification of the reported results. Complete reproduction code, molecular integrals, and precomputed FeMoco results are available to editors and reviewers upon request under confidential peer review.