PhD - Leuven | Just now
High-performance computing (HPC) chips now operate at extreme power densities, with local heat fluxes reaching the kW·cm⁻² regime. To keep pace, the ecosystem is shifting from traditional heat sinks and heat pipes to package-integrated liquid cooling. Yet today’s in-package coolers remain constrained by the thermal resistance between silicon and coolant and by the pumping power and volume penalties of complex microchannel features. Passive flow disruptors (roughness, ribs, grooves, manifolds) can raise heat-transfer coefficients, but they also increase flow resistance (increased pumping power), complicate mass fabrication, and cannot adapt to transient hot spots.
This PhD explores a fundamentally different, active approach: ultrasound-mediated enhancement of convective heat transfer via acoustic streaming. By launching acoustic waves into the coolant, we can generate steady flows (micro-vortices/jets) that convect heat away from hot surfaces on demand. Recent advances in fast and thermoacoustic streaming suggest that significant flow speeds are feasible, opening the door to compact, and controllable heat extraction.
The central hypothesis is that co-design of the acoustic field, fluidic geometry, and package thermal path can lower junction-to-sink resistance and elevate local heat-transfer coefficients without increasing total system power or footprint. To test this, the student will (i) develop analytical models and perform Multiphysics simulations (CFD, acoustics, heat transfer) to predict flow patterns and temperature fields in realistic microchannel and manifold-on-chip layouts; and (ii) build experimental test vehicles to validate predictions via flow visualization and heat transfer measurements. Success will be measured by demonstrable reductions in junction-to-sink thermal resistance and/or increases in local heat-transfer coefficients at equal or lower system power and volume relative to conventional solutions.
Required background: Physics, Engineering Science
Type of work: 60% modeling/simulation, 30% experimental, 10% literature
Supervisor: Houman Zahedmanesh
Daily advisor: Bart Weekers, Grim Keulemans, Ben Jones
The reference code for this position is 2026-043. Mention this reference code on your application form.