Physically Inspired Particle Framework for Recalescence-Induced Self-Jumping Ice Dynamics

<title>Abstract</title> <p>This work presents a reduced-order, particle-based physical model for recalescence-induced self-jumping ice dynamics, where recalescence denotes the rapid latent-heat-driven temperature recovery that occurs immediately after nucleation in a supercooled droplet.Rather than focusing on visual modeling alone, the proposed framework captures the essential momentum-generation mechanism arising from asymmetric vaporization during the early stage of freezing.Unlike prior graphics-oriented studies that mainly addressed static ice appearance or geometric growth, our approach reproduces the transient dynamics driven by thermal asymmetry and imbalanced vapor momentum during recalescence.The proposed system automatically detects nucleation sites from the evolving surface temperature field, propagates an asymmetric recalescence front, and estimates the jump direction and magnitude by accumulating directional impulse within an arc-limited region around the front tip.This accumulated impulse is formulated as a reduced-order evaporation-recoil proxy, serving as a phenomenological approximation of the reaction forces generated by asymmetric vapor flux observed in physical experiments.After detachment, the model transitions to rigid-body motion while preserving the original shape, enabling consistent analysis of post-detachment dynamics.By generalizing front propagation using a boundary-aware geodesic parameterization, the framework naturally extends from 2D silhouettes to arbitrary 3D triangular meshes.Through experiments on 2D image masks, noncircular shapes, and multiple sampling schemes (Poisson-disk and grid sampling), as well as on complex 3D meshes such as dino, bunny, statue, dragon, and teapot, we demonstrate that the proposed model consistently produces shape-independent, reproducible, and curvature-aware self-jumping behavior.Overall, the proposed framework provides a computationally tractable reduced-order representation of recalescence-driven impulse generation and detachment dynamics, bridging experimental observations and simulation without resorting to full multiphase or thermo-fluid solvers.</p>