The arXiv preprint (submitted June 2026) introduces an extended multiphysics model that couples volumetric contraction during phase change with capillary and Marangoni flows, validated on 1D Stefan problems, 2D free-surface solidification, and axisymmetric laser-melt benchmarks. Unlike earlier enthalpy-porosity schemes that neglect density jumps, the new mass-correction algorithm slashes conservation error by orders of magnitude, exposing how shrinkage-induced surface depression seeds roughness and lack-of-fusion pores. This numerical advance directly addresses a gap in industrial selective laser melting (SLM) where parts for aerospace turbines routinely exhibit 0.5–2 % linear contraction that prior CFD codes ignored, leading to underestimated residual stresses. Real-world patterns confirm the omission: studies such as Khairallah et al. (Acta Materialia, 2016) documented pore formation but attributed it solely to recoil pressure, while Tang et al. (Additive Manufacturing, 2020) linked surface topography to Marangoni flow without quantifying shrinkage-driven meniscus evolution. The preprint’s smoothed mushy-zone treatment mitigates discretization artifacts yet remains limited to single-track simulations; no multi-layer, powder-bed statistics are provided, leaving scalability to production parts untested. Consequently, manufacturers risk continued over-reliance on empirical parameter tuning rather than predictive compensation for shrinkage-induced warpage in high-value components.