Percolation-based characterization of vapor network evolution during pool boiling of a highly wetting dielectric fluid

Highly wetting dielectric fluids are increasingly used in two-phase immersion cooling because of their favorable heat transfer characteristics. However, vapor structure evolution and connectivity across boiling regimes remain insufficiently understood. In this study, pool boiling experiments were conducted using NOVEC 649 on a vertical heated surface, and percolation-based analysis was employed to characterize vapor networking spatially and temporally. Spatial development was evaluated using the normalized largest connected vapor cluster area (NLCA) and correlation length, while temporal persistence was characterized using the largest cluster fraction (LCF). With increasing heat flux, bubble coalescence promoted larger and more extended vapor structures. The NLCA approached 0.5, while the correlation length reached 0.75–0.82 mm at high heat fluxes. The LCF increased from approximately 0.4 in the bubbly regime to 0.95–0.98 in the slug-like vapor regime, indicating a persistent dominant vapor network. The transition in vapor connectivity was quantified using logistic sigmoid fitting and bootstrap analysis, identifying a transition heat flux of 60.96 kW/m 2 . Combining the LCF and mean correlation length, a quantitative boiling regime map was constructed to distinguish the bubbly, transition, and slug-like vapor regimes based on temporal and spatial vapor connectivity. These results demonstrate that boiling regime transitions can be quantitatively characterized through vapor connectivity rather than qualitative bubble morphology. The proposed framework may be extended to other working fluids and boiling conditions.