Scholl, S., Gorle, C., Houshmand, F., Asheghi, M., Goodson, K.E., Verstraete, T., “Numerical optimization of advanced monolithic microcoolers for high heat flux microelectronics,” ASME International Technical Conference and Exhibition on Packaging and Integration of Electronic and Photonic Microsystems (InterPACK) 2015, July 6 – July 9, San Francisco, CA
This study considers the optimization of a complex micro-scale cooling geometry that represents a unit-cell of a full heat sink microstructure. The configuration consists of a channel with a rectangular cross section and a hydraulic diameter of 100 um, where the fluid flows between two cooling fins connected by rectangular crossbars (50 x 50 um). A previous investigation showed that adding these crossbars at certain locations in the flow can increase the heat transfer in the microchannel, and in the present work we perform an optimization to determine the optimal location and number of crossbars. The optimization problem is defined using 12 discrete design parameters, which represent 12 crossbars at different locations in the channel that can either be turned off and become part of the fluid domain, or turned on and become part of the solid domain.The optimization was done using conjugate heat transfer computational fluid dynamics (CFD) simulations using Fluent 15.0. All possible 4096 configurations were simulated for one set of boundary conditions. The domain was discretized using about 1 million nodes combined for the fluid and solid domains and the computational time was around 1 CPU hour per case. The results show that further improvements in heat transfer are feasible at an optimized pressure drop. The results cover a range of pressure drops from 25 kPa to almost 90 kPa and the heat transfer coefficient varies from 60 to 120 kW/m2K. The configurations on the Pareto front show the trend that crossbars closer to the maximal fluid-solid interface result in a more optimal performance than crossbars positioned farther away. In addition to performing simulations for all possible configurations, the potential of using a genetic algorithm to identify the configurations that define the Pareto front was explored, demonstrating that a 80% reduction in computational time can be achieved. The results of this study demonstrate the significant increase in performance that can be obtained through the use of computational tools and optimization algorithms for the design of single phase cooling devices.