The study of a novel topology-optimized heat sink for single-phase immersion cooling in data centers

• Topology optimization redesigned a heat sink for single-phase immersion cooling. • Optimized heat sink cut average CPU temperature 19.02% and nonuniformity 44.90%. • Pumping power dropped 8.30%; energy efficiency rose 115.23% over the baseline. • Assessed impacts of TDP, coolant properties, and server height on performance. As the performance demands of data centers continue to grow, the challenges associated with cooling and the energy required for thermal management have become increasingly significant. To address these issues, this study applies topology optimization to redesign the heat sink component of a single-phase immersion cooling (SPIC) system, aiming to enhance heat dissipation efficiency and thereby improve overall system performance. The topology-optimized heat sink is then compared with a conventional rectangular-fin heat sink under identical immersion cooling conditions. The results show that the optimized design reduces the average temperature of the central processing unit (CPU) by 19.02% and lowers the temperature standard deviation by 44.90%. Moreover, it requires less pumping power, leading to a 115.23% increase in the energy efficiency coefficient. These improvements are of great significance for future efforts to enhance cooling efficiency and reduce energy consumption in data centers. In addition, this study also discusses the impact of coolant inlet temperature, inlet velocity, CPU thermal design power (TDP), channel height, and different cooling media on the performance of the SPIC system under two different heat sink designs.