Integrated CFD modeling and experimental validation of single-phase immersion cooling for data center thermal management

Driven by the escalating power densities of modern servers, single-phase immersion cooling has emerged as a critical thermal management paradigm. However, the absence of a systematic methodology for selecting optimal coolants and components remains a significant challenge. This study addresses this gap by proposing a holistic framework that integrates theoretical analysis, computational fluid dynamics (CFD) simulations, and experimental validation. First, the flow and heat transfer characteristics of oil-based and fluorocarbon-based coolants were scrutinized through boundary layer theory and the field synergy principle. Concurrently, a novel Heat Dissipation Coefficient (HDC) was formulated to quantitatively evaluate coolant performance. Second, the thermo-hydraulic behavior of a standard 1U server was rigorously evaluated via CFD simulations and validated on a dedicated experimental platform. Building upon this, heat sink geometries were optimized using a hybrid CFD-Genetic Algorithm (GA) approach. Third, a novel tank architecture was engineered to mitigate flow maldistribution. The Realizable k-ε model and the SST k-ω model were respectively employed for server flow simulations and ribbed structure optimization. The results demonstrated that the HDC index provided a high-fidelity assessment of coolant performance, aligning closely with CFD data while significantly reducing computational overhead. Comparative analysis revealed that fluorocarbon coolants exhibited superior thermal performance over silicone oil. The optimized heat sink achieved a Performance Evaluation Criterion (PEC) of 2.05, while the redesigned tank structure reduced flow maldistribution by 85%. Finally, a 1000-h compatibility test confirmed the long-term operational reliability of the SS110 coolant.