Confined Immersion Cooling in Microscale Gaps

Abstract

Thermal management is one of the major operational concerns for data centers and accounts for a significant fraction of total power consumption. Passive immersion cooling solutions have been explored owing to their potential for offering low overall thermal resistance in very dense rack configurations where there is no room for conventional heat sinks between printed circuit boards. Further, in practice, regions of high heat flux are localized to where processing units are positioned. Non-uniform heating, as well as local hot spots, could affect thermal performance as a result of the need for rewetting of the surface with liquid during boiling. This work explores immersion cooling in submillimeter confined liquid filled gaps with localized heat sources. Specifically, this work investigates the thermofluidic characteristics of highly confined boiling surfaces. A camera is used to visualize the two-phase flow regimes and instabilities that occur prior to critical heat flux (CHF) limits. Two distinct boiling regimes are observed (namely, intermittent boiling and partial dryout). Both the heated fraction of the area within the confined region and the gap spacing affect the CHF values and thermal performance prior to CHF. The optimum thermal performance, in terms of the surface superheat, is experimentally observed for a confinement corresponding to a Bond number of 0.2. However, at this optimum condition based on surface superheat, the CHF is significantly reduced to 27% of the unconfined CHF limit. Significant additional reductions in the CHF are also experimentally observed when the adiabatic confinement surface is extended beyond the heater edge. This additional fundamental understanding of the impact of spatial confinement on the thermal performance of immersion cooling has broad implications for two-phase thermal management solutions.