Abstract The increasing power densities of processors and server racks necessitate advanced thermal management solutions for data centers. Pumped two-phase (2P) direct-to-chip (DTC) cooling leverages boiling of a dielectric refrigerant and offers high efficiency, reliability, and scalability. However, a comprehensive framework to analyze thermal contributions in 2P DTC systems is lacking due to fundamental differences between the heat transfer characteristics of 2P and single-phase (1P) DTC cooling. In this work, a systematic thermal analysis is conducted for 2P DTC systems coupled with chilled facility water (FW). The end-to-end temperature difference between processor case and FW inlet is adopted as the system performance metric. Refrigerant temperatures at the cold plate outlet and condenser inlet are introduced as intermediate nodes to break down the end-to-end temperature difference into three contributions from cold plate, vapor line pressure drop, and condenser. Using this framework, system-level comparisons between representative 1P and 2P DTC systems are performed. A correlation-based microchannel model is used to analyze cold plate performance with PG25 in 1P mode and R515B in 2P mode. The two systems exhibit comparable performance at a processor heat flux of 100 W/cm2. Under the representative conditions considered, the 2P system enables a 5 °C higher FW inlet temperature at 200 W/cm2, suggesting potential for reduced facility-side cooling energy consumption. This work establishes a quantitative framework for analyzing 2P DTC systems and supports their large-scale adoption in data centers.
Capillary-fed thin-film evaporation is a promising route for passive thermal management, enabling high-performance cooling devices like heat pipes and vapor chambers. The wick design is crucial, and high performance is defined by sustaining large heat fluxes at low overall thermal resistance. Previous modeling efforts on evaporator wicks have predominantly focused on silicon pillar arrays, which provide limited guidance for microporous wicks. Microporous wicks have been widely used in passive thermal management owing to their capability of dissipating high heat flux by increasing the effective evaporation area and sustaining the liquid supply. In this work, we take copper inverse opal (CIO) as a representative microporous wick and develop a predictive numerical pore-scale evaporation model together with an optimization framework. The model captures pore-scale complexity by computing local contact angles from surface normals to set a dry-out criterion and by evaluating thermal resistance with a spherical cap-based interfacial area model. We then couple the model with a multi-objective genetic algorithm to optimize axially graded CIO wicks at a fixed overall thickness, where the pore geometry is varied section by section along the flow direction, to increase the critical heat flux (CHF) while lowering the overall thermal resistance. For designs with fixed thickness, pore-size graded CIO wicks outperform uniform-pore designs, achieving up to a 33.8% increase in CHF and a 62% reduction in thermal resistance. This framework can be extended to different microporous wicks and offers guidelines for capillary-fed evaporative cooling systems.
Ensuring the vibration reliability of electronic packages is crucial for their long-term operation, especially in high-reliability environments. The research investigates the effects of boundary conditions, component placement, and the number of components on the fatigue lifetime of printed circuit board assemblies (PCBA). While traditional approaches relied on numerical models for estimating fatigue life, the current research utilized advanced machine learning models, extreme gradient boosting (XGBoost), and stochastic gradient boosting (SGB), to enhance accuracy and efficiency. The finite element analysis (FEA) and experimental results were rigorously validated by matching natural frequencies and strain values, confirming the robustness of the models in simulating real-world conditions. The machine learning models demonstrated exceptional performance, with XGBoost and SGB achieving high R2 values of 0.957 and 0.978 and low mean absolute percentage error (MAPE) scores of 0.1587 and 0.1322, respectively. The machine learning predicted and numerically estimated fatigue lifetime was further validated through experimentally resonance-based fatigue analysis. The research reveals that strategic component placement and careful consideration of boundary conditions can significantly improve the fatigue lifetime of PCBA. By integrating Shapley additive explanations (SHAP), the research enhances the interpretability of machine learning models, providing insights that contribute to optimizing the design and performance of electronic packages under vibration loading.
The growing demand for higher computational efficiency in data centers and AI hardware is accelerating the adoption of three dimensional (3D) heterogeneous integration and immersion-cooling technologies. Oil immersion offers significantly enhanced heat dissipation compared to conventional air cooling, improving thermal stability and enabling higher power densities. However, the thermo-mechanical environment created by oil immersion also introduces new reliability challenges. In 3D packages, mismatches in material properties like Coefficient of Thermal Expansion (CTE) and Elastic Modulus (E) can induce interfacial stresses that lead to through-silicon via (TSV) cracking, printed circuit board (PCB) warpage, and solder-joint fatigue failures. This study investigates the influence of oil-immersion cooling on the thermomechanical reliability of solder joints and TSV structures in advanced 3D packages. Plastic strain accumulation in solder joints under extreme thermal cycling is analyzed to evaluate the likelihood of crack initiation and propagation. Additionally, J-integral computations are performed to quantify crack-tip driving forces in TSVs across different PCB configurations, providing a comprehensive assessment of package lifetime reliability. To further enhance TSV resilience, this work examines alternative via-fill materials-including tungsten, silver, nickel, carbon nanotubes, aluminum, molybdenum, gold, graphene, and annealed copper which are possible alternatives-and compares their crack-propagation tendencies using J-integral-based metrics. Overall, this research not only identifies reliability risks associated with oil-immersion environments but also proposes material-level mitigation strategies.
