Saved in:
| Hovedforfatter: | |
|---|---|
| Format: | Recurso digital |
| Sprog: | |
| Udgivet: |
Zenodo
2026
|
| Fag: | |
| Online adgang: | https://doi.org/10.5281/zenodo.19799694 |
| Tags: |
Tilføj Tag
Ingen Tags, Vær først til at tagge denne postø!
|
Indholdsfortegnelse:
- <p>The managed-fusion tiling (MFT) framework decomposes a quantum many-body problem on n qubits into smaller tiles plus interface corrections, with cost dominated by the largest tile width w(n) ≤ n. The framework’s practical utility depends on choosing a partition that yields small fusion widths — a choice that, in general settings, is not known in advance. This work introduces remixing: an iterative protocol that uses the fusion-channel correction ∆ = Eexact − (EA + EB ) as a diagnostic signal of correlation across a candidate boundary, and updates the partition by gradient descent on |∆| until a low-correction configuration is found. The protocol is validated on three problem classes. On one-dimensional Ising chains with engineered bond strengths, |∆| tracks bond strength at the cut with Pearson correlation r = +0.985, and hidden strong-bond positions are recovered with 10/10 success across random instances. On a 3×4 Heisenberg lattice with two strong-coupling clusters, the cluster-respecting partition ranks first of 462 candidate partitions, and Pearson correlation between strong-bond cuts and |∆| is +0.91 (p < 10^−175). On molecular Hamiltonians of H2 and LiH in the STO-3G basis, the correction-minimal partitions correctly identify spin-pair preservation as the dominant correlation in closed-shell configurations. The intrinsic tile-only formulation — restricting the Hamiltonian to within-tile terms with no reference state — is shown to be essential; reference-projection alternatives produce methodology-dependent rankings that do not reflect the underlying correlation structure.</p>