/Reliable in-package liquid jet impingement cooling

Reliable in-package liquid jet impingement cooling

Leuven | More than two weeks ago

Develop the cooling solutions for the high-power AI systems of the future.

With increasing power density in electronic systems due to the demand for compute capabilities, the development of efficient cooling systems is crucial to limit the operating temperature levels. Significant work has been undertaken to address this issue using liquid cooling in various configurations, such as microchannel, jet impingement, stacked/double layer configurations etc. In recent years, we have developed a package level integrated jet impingement based liquid cooling solution that delivers the liquid coolant cooling directly on the chip backside and avoids the use of the thermal interface material [1], [2]. Very good thermal performance is attributed to the underlying physics involved, such as the presence of a very thin thermal boundary layer due to jet impact in the stagnant zone, and thanks to the introduction of structured surfaces to locally enhance the heat transfer.  Despite the promising cooling performance experimentally demonstrated on advanced thermal test vehicles several remaining integration and reliability challenges need to be addressed to enable a successful adoption of this cooling technology.

 

The objectives of this PhD work are to further improve the cooling performance by pushing the limit of heat transfer; to gain fundamental understanding in the long-term reliability of the liquid coolant, Si surface and cooler material; and to integrate the cooling solution into the chip package by providing the required protection to the Si surface and minimizing its thermal penalty. The proposed PhD thus seeks to develop fundamental understanding while advancing the technology of single-phase jet impingement cooling in the following areas:

 

A.    Impact of structured surfaces on heat transfer performance. The presence of structures on the heated area (die) is bound to alter the hydrodynamics in a confined jet and thus the thermal behavior. What kind of structures would be optimum under given conditions needs to be investigated. Fundamental understanding of the flow and thermal dynamics will provide insights into the impact of various parameters, such as fluid properties on the performance thereby providing a roadmap for developing of non-trivial structures such as those showing behavior independent of liquid.

 

B.    Coolant selection: evaluate the impact of the thermo-physical properties of liquid coolants or liquid-liquid solutions [3] on the cooling performance and the interaction between the coolant and Si surface and cooler material. Liquid-liquid solutions show a potential increase in heat transfer over water-based cooling due to mass transfer during phase separation. The proposed work seeks to investigate these alternative fluids in jet impingement configuration. The scientific interests lie in understanding the interaction of hydro-, thermo-, and phase transition dynamics. Further, how does the presence of interface provide means of enhancing mixing through capillary/interfacial phenomenon would be analyzed along with impact of structured surfaces in these systems.  

 

C.   Reliability of the cooling system:  The cooling solution using liquid are a closed system and are designed to run for a longer period. The interaction of liquid with solid substrate (die) can lead to alteration of the solid surface – corrosion, salt deposition etc. Even though it may be small in quantity, over a period this can escalate and result in significant performance loss. The third part of the project thus seeks to answer these questions with and without structured surface and various liquid combinations and investigate the impact of coatings to prevent this degradation.

 

  

The proposed methodology comprises of:

 

  1. Modeling and simulation where conjugate heat transfer studies would be undertaken. In addition, appropriate modeling strategies would be developed to account for reliability and incorporated with governing fluid.
  2. Use of Machine learning methodology (data driven techniques, Physics informed neural networks) to understand the phenomenon over a wider range of operating conditions.
  3. Design and fabrication of cooler demonstrators to characterize the thermal performance of the cooler, enhancement structures, coolant impact and long term effects using imec’s high resolution thermal test chip.

 

Required background: Master degree in engineering, physics or mathematics. Affinity with numerical methods.

 

Type of work: 45% modeling/simulation, 45% experimental, 10% literature

Supervisor: Tine Baelmans

Co-supervisor: Herman Oprins

Daily advisor: Deewakar Sharma

The reference code for this position is 2025-005. Mention this reference code on your application form.

Who we are
Accept marketing-cookies to view this content.
imec's cleanroom
Accept marketing-cookies to view this content.

Send this job to your email