Experimental and numerical investigation of standalone PCM-based cooling for data centers

The rapid, AI-driven expansion of data center (DC) infrastructure has made cooling—consuming 30–40 % of facility electricity—a critical energy bottleneck. Phase change energy storage (PCES) offers a mechanistically distinct alternative: encapsulated phase-change materials (PCMs) absorb latent heat through solid–liquid transitions, enabling passive diurnal load shifting that harnesses ambient thermal resources for next-generation DC cooling. Yet prior work remains confined to auxiliary buffering or bench-scale demonstrations, with no generalizable framework for standalone PCM cooling at full DC scale under transient climatic forcing. Here, we present the first integrated experimental and three-dimensional transient computational framework evaluating a fully passive, macro-scale PCES architecture sized for an operational DC under real meteorological conditions. Our systematic parametric analysis reveals the following findings. (1) PCM thickness exhibits a non-monotonic optimum governed by competition between thermal penetration depth and latent heat capacity, establishing that thickness must match the diffusion length scale, not simply be minimized. (2) Air-PCM residence time, not convective intensity, is the rate-limiting factor for cooling effectiveness. (3) We develop multivariate empirical correlations for outlet temperature and liquid fraction as the first closed-form predictive tools for full-scale PCES–DC design. (4) Under diurnal meteorological forcing, the optimized system achieves ∼5 °C peak cooling, maintains supply air below the recommended threshold with only brief exceedance, regenerates fully overnight without mechanical assistance, and reduces annualized operating costs by 77 %. These findings decisively reposition PCES from auxiliary buffer to viable primary cooling strategy, establishing a generalizable framework for passive, zero-carbon DC thermal management.