The microelectronics industry has been steadily improving processing performance for many years, and each advance increases the transistor count and density. This trend is expected to continue as described by Moore’s Law with a consequence of increased heat dissipation per unit volume. Some modern microelectronics packages have localized volumetric heat generation exceeding 2 kW/cm3, which demands a thermal management system that provides sufficient global cooling and control of hot spots on the die. As modern electronics packages progress towards stacked integrated circuits and 3D transistor technologies, passive and active cooling techniques can take advantage of the device construction to use flow boiling in microchannels or pool boiling on structured surfaces to meet the increased cooling needs.
Nucleate boiling can provide high heat fluxes while maintaining the surface temperature. However heat transfer in boiling is generally limited by a critical heat flux (CHF) that is defined by a maximum heat removal before transitioning from nucleate to film boiling. In previous works, there have been several successful methods for delaying CHF, such as structured, acoustic interfacial excitation, and flow boiling. Many of these techniques can be further optimized by numerical simulation. The proposed research will numerically investigate arrays of small heaters with partitions and open microchannels to create localized boiling and provide flow control to prevent excessive coalescence of bubbles immediately over the heated surface and delay the transition to film boiling.
Many multiphase flow simulations have excessive run times, and to reduce this computational expense, parallel direct numerical simulations are being developed for multi-GPU architecture to handle larger domains to study dynamics of many bubbles in addition to single bubble dynamics. The numerical method uses lattice Boltzmann for the fluid flow solver, a particle level-set for interface tracking, and finite-differences for heat transfer. These numerical methods have been chosen since they are amenable to parallelization, and the particle level-set method provides better resolution of interfacial features than standard level-set methods. The simulations will be validated by comparing to experimental work already underway. The simulations will be used to identify improved methods of delaying CHF and management of hot spots and for optimization of heater surface designs.