A Novel Concept for Air Removal in Two-Phase Immersion Cooling Systems

Abstract A 10 kW scale model of a decoupled immersion cooling rig is constructed in order to serve as a testbed for immersion cooling, using 3M FC3284 dielectric cooling fluid. A species separator is constructed and demonstrates an ability to remove air from the flowfield before the condensable gases enter the condenser vessel, verified with Schlieren photography. The condenser underperformed significantly compared to initial sizing calculations using the NTU method, and film thickness of FC3284 liquid on the surface of the condenser was determined to be the cause due to low thermal conductivity of the liquid. The average film thickness on the surface of the condenser is calculated. In addition to the performance detriment of the film, air is also shown to reduce the condenser’s performance. The height of a transient stratification line is measured and compared against condenser power. Condenser efficacy losses are large and variable based on the concentration of air in the condenser vessel. A low vs high-mounted boiler is investigated. The mounting of the boiler has an effect on how much vapor is lost during a maintenance event. Finally, a comparison of the test rig’s overall cooling efficiency is made with various air-cooled datacenters by tracking energy consumption to cool a given IT load. This also translates to a reduction in carbon emissions.