Physically Inspired Visual Simulation of Recalescence-Driven Ice Formation and Self-Jumping

This study presents a lightweight particle-based framework for visually simulating self-jumping ice driven by the recalescence process during freezing. Unlike previous graphics studies that focused on icicle growth or optical rendering, our method models the dynamic coupling between thermal asymmetry and evaporative momentum that leads to spontaneous detachment. The system automatically identifies the nucleation site from the evolving temperature field, propagates asymmetric freezing fronts, and accumulates directional impulses within arc-limited regions to determine the jump direction and magnitude. After detachment, the dynamics are reduced to a rigid-body phase while preserving the original frozen silhouette for visual consistency. The extended solver also supports arbitrary shapes defined by input mask images, demonstrating robust jumping behavior across non-circular geometries and sampling patterns. The proposed framework balances physical plausibility and artistic controllability, offering an efficient and visually coherent approach to recalescence-driven ice simulation.