Hotspot thermal management via thin-film evaporation

Abstract

The emerging three-dimensional vertical chip stacking architecture is expected to reduce form factor and improve performance by providing energy efficient chip design. However, increased power density and non-uniform heat generation in stacked dies offset its advantages and pose a significant thermal management challenge by creating hotspots where heat loads in excess of 1 kW/cm2 are generated from sub-millimeter areas. Furthermore, the localized heating in hotspots creates high junction temperature which can degrade the performance, reliability, and life time of electronic chips. Such ultra-high heat fluxes are challenging to remove using state-of-the-art single-phase cooling technology. Consequently, chip-level phase-change based hotspot thermal management is increasingly becoming pivotal for cooling next-generation of microelectronic devices and power amplifiers. This work experimentally characterizes capillary-limited thin-film evaporation from well-defined silicon micropillar wicks to demonstrate its potential as a thermal solution for ultra-high heat fluxes. We used contact photolithography and deep-reactive-ion-etching to create a 1×1 cm2 microstructured area. The microstructured area was surrounded by a water reservoir. Various sized thin-film heaters which were created using electron-beam evaporation and acetone lift-off were integrated on the backside of the test sample. Hotspots were emulated by locally heating a 640×620 µm2 area while background heating was emulated by heating the entire 1×1 cm2 microstructured area. The background and hotspot heaters were calibrated prior to experiment to measure temperature. All experiments were conducted in an environmental chamber which was maintained near saturated condition, i.e., saturation temperature and corresponding pressure. The working fluid, degassed de-ionized water, was transported from the surrounding water reservoir to the microstructured area passively via capillary-wicking. We dissipated ≈5.8 kW/cm2 from a 620×640 µm2 footprint when the hotspot temperature was ≈260 °C. Most importantly, when the surface dried out at ≈5.8 kW/cm2, the background temperature as well as the local temperatures 3 mm away from the hotspot were less than 50 °C. Increasing the heat flux beyond ≈5.8 kW/cm2 resulted in the formation of a dry island at the center of the hotspot which grew radially outwards. Dryout and thermal runaway occurred when viscous losses exceed the capillary pressure. Furthermore, the maximum dryout heat flux from a single hotspot decreased from ≈5.8 kW/cm2 to ≈2.9 kW/cm2 when the hotspot was assisted by a 20 W/cm2 background heating. Lastly, the dryout heat flux decreased from ≈5.8 kW/cm2 to ≈2.9 kW/cm2 per heater when three spatially distributed hotspots were created concurrently. Unlike the dryout heat flux, the total heating power increased by assisting hotspot with background heating as well as by creating spatially distributed concurrent hotspots over the microstructured area.