Research Projects

Rapid growth of cloud computing and AI has led to an exponential increase in data generation and processing demands, placing significant thermal loads on data centers. These high-performance systems operate continuously, producing immense heat that must be efficiently managed to ensure reliability, performance, and energy efficiency. Without effective cooling, hardware failure risks operational costs; making advanced cooling solutions essential for sustaining the scalability and sustainability of modern digital infrastructure. Modern liquid-cooled data centers use direct-to-chip cooling technologies to manage high heat densities more efficiently than traditional air-cooling. While these systems offer superior thermal performance and reduced energy consumption, they also introduce challenges, particularly the risk of liquid leakage. Leaks can damage sensitive electronic components, cause system downtime, and increase maintenance complexity. Ensuring leak-proof designs, robust monitoring, and safe coolant selection remains critical to the widespread adoption and reliability of liquid cooling solutions. The risk for liquid spillage/leakage can be eliminated if the entire rack/node cooling system is maintained at negative pressure (below atmospheric pressure) instead of the conventional positive pressure (above atmospheric pressure) system. The system, described in this proposal, implements this ideology to ensure leakproof operation of the cooling unit. A 3-tank system is proposed where the differential vacuum pressure between reservoir and main/auxiliary tank drives the liquid coolant flow through the server rack. The system operation requires a continuous vacuum pressure in the reservoir tank whereas the main and auxiliary tanks will have fluctuating pressures based on the directionality of coolant flow. (More details on system operation is provided in the section on “Research Methodology”.) As the fluid flow is under vacuum, any leakage in the pipeline will lead to ingress of air into the coolant distribution system instead of liquid coolant outgress. This completely eradicates the issue of liquid coolant spillage making this technology safe and reliable for large scale high-end semiconductor components cooling. In this project a lab-scale prototype of a 3-tank system will be designed, fabricated, assembled and tested with a proposed cooling capacity of 4 kW. The present system only requires a vacuum pump and a liquid pump for primary coolant flow through heat exchanger. The coolant flow through the server rack (secondary loop) occurs without any pump. This also makes it challenging to control the coolant flow rate through individual nodes of the server. Hence, an optimised algorithm has to be developed to control the pressure differential across the node for maintaining the desired coolant flow rate. As part of the present work, a detailed parametric study of the cooling unit will be undertaken and subsequently a control algorithm will be developed to ensure desirable coolant flow through individual nodes. The system simulation model developed as part of this study will be validated with the results from the parametric studies. This model will serve as a future reference for scale up of the system. For practical implementation of vacuum pressure cooling systems, the cold plate design of individual nodes needs to be revisited. Traditional cold plates are designed with large pressure drops across them, which is not feasible with negative pressure cooling units. Therefore, the performance of existing cold-plate designs will be first evaluated at vacuum pressures using CFD studies. Based on the CFD simulation results, suitable design modifications will be made to reduce pressure drop while maintaining high heat transfer coefficients. Finally as part of this project, a lab scale cold-plate will be fabricated and integrated with the negative pressure cooling unit to test its performance.