Geometry Representation, Not Solver Choice, Dominates Cross-Code Permeability Predic-tions for Biomimetic and TPMS Heat Pipe Wicks

Abstract Pore-scale permeability predictions for candidate heat pipe wicks are routinely reported from a single solver on a single geometry representation, leaving solver error, discretization error, and geometric error unseparated. This paper reports a permeability campaign in which four solvers spanning three modeling families are validated against a closed-form capillary-bundle benchmark and applied to a common cylindrical heat pipe core. The four are two pore-scale lattice Boltz-mann codes (FluidX3D, Palabos/MPLBM-UT), a geometry-resolved finite-volume solver (Open-FOAM), and a pore-network solver (OpenPNM). On the 5×5 capillary-bundle benchmark the three geometry-resolved solvers agree with the analytical permeability to within 3.44, 2.22, and 0.19 %; the two lattice Boltzmann codes agree with each other to 0.02 % at matched resolution. Nine wick structures were simulated: three triply periodic minimal surface (TPMS) families at two fill ratios, and three computed-tomography-derived biomimetic structures. Permeabilities span 4.18 × 10−8 to 1.67 × 10−7 m2. At both fill ratios permeability ranks Schwarz Primitive > gyroid > Schwarz Diamond while the most permeable structure is the least porous, so the two predictors disagree on ranking before any thermal data are introduced. Reported TPMS permeabilities come from an im-plicit level-set representation rather than a voxelized triangulated surface; implicit cases change by under 1 % between refinement levels, voxelized cases at the 25p fill ratio by 4.9 to 9.4 %, and the tightest lattice of the set by 27.9 %. Geometry representation, not solver choice, is the dominant source of scatter in this class of prediction, with consequences for capillary-limit design screening.