Quantized composition penalties alter state selection in phase-field Ising models

Phase-field models seek low-energy spatial structures at fixed composition, but finite-precision Ising formulations can change which structures are selected. We investigate how rounding alters energy ordering among composition-preserving states and why protecting the composition penalty may not improve physical accuracy. Using a one-dimensional model with four allowed local field values and an exact binary encoding before rounding, we combine analytic energy decomposition with exhaustive enumeration of all configurations and tied minima. Rounding the composition penalty creates artificial preferences among feasible states; jointly rounding physical and penalty coefficients can reverse physical-energy ordering even when residual penalty energies are identical. In the calibrated comparison, exact penalty preservation reduces reversals but consumes coefficient range, making physical energy differences harder to resolve and increasing ties with inferior states. Averaged over 96 held-out instances and six prespecified precisions, the separately calibrated penalty-preserving policy lowers the fraction of encoded minima satisfying composition and the physical-energy tolerance by 2.14 percentage points relative to conventional rounding. A separate continuum comparison identifies a physical minimum excluded by the allowed field values. These results connect coefficient errors to selected structures and explain how increased inferior ties can outweigh reduced ordering reversals.

Authors

Publication Details

Journal
Research Square
Published
2026-11-19
DOI
https://doi.org/10.21203/rs.3.rs-11603792/v1
Primary Topic
Solidification and crystal growth phenomena
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

Quantized composition penalties alter state selection in phase-field Ising models

Zicong Zhu, Anping Liu, Zhiyi Liu
Research Square
Solidification and crystal growth phenomena
preprint

Quantized composition penalties alter state selection in phase-field Ising models

Zicong Zhu, Anping Liu, Zhiyi Liu
preprint en

Abstract

Phase-field models seek low-energy spatial structures at fixed composition, but finite-precision Ising formulations can change which structures are selected. We investigate how rounding alters energy ordering among composition-preserving states and why protecting the composition penalty may not improve physical accuracy. Using a one-dimensional model with four allowed local field values and an exact binary encoding before rounding, we combine analytic energy decomposition with exhaustive enumeration of all configurations and tied minima. Rounding the composition penalty creates artificial preferences among feasible states; jointly rounding physical and penalty coefficients can reverse physical-energy ordering even when residual penalty energies are identical. In the calibrated comparison, exact penalty preservation reduces reversals but consumes coefficient range, making physical energy differences harder to resolve and increasing ties with inferior states. Averaged over 96 held-out instances and six prespecified precisions, the separately calibrated penalty-preserving policy lowers the fraction of encoded minima satisfying composition and the physical-energy tolerance by 2.14 percentage points relative to conventional rounding. A separate continuum comparison identifies a physical minimum excluded by the allowed field values. These results connect coefficient errors to selected structures and explain how increased inferior ties can outweigh reduced ordering reversals.

Research Square
Solidification and crystal growth phenomena
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon