Novel thermal management solutions for extreme heat flux cooling
- 50 - 60% of all electronic failures in the field are attributed to thermal issues. Increasingly power dense electronics devices of the future also dissipates an exponential amount of waste heat flux, which when inadequately cooled, could be catastrophic. Not only does it reduce performance, but it also decreases the reliability, robustness and life-span of the package. To solve this issue, research into novel extreme heat flux thermal management solutions is necessary. High performance cooler development and its successful and reliable integration with heat producing power chips, has been determined to be the most promising and energy efficient path towards achieving next generation power packages. This dissertation also identifies an all-silicon coolers directly die-attached to silicon chips as the promising near junction cooling scheme of the future, this integration tactic enabling the extraction of the maximum amount of performance, efficiently from the package. This dissertation first introduces the readers to the UV-laser tool, which is able to quickly prototype, high performance cooler chips, however debris produced during laser rastering, is a major reliability issue and causes subsequent bonding failure between the chips. We propose a novel method of using a temporary, sacrificial polymer protective coating, which collected the debris during processing and removed it perfectly. The bonding method which will be used to integrate these cooler chips, is studied next - varying temperature and pressure was used to identify ideal process conditions during ultra-thin (1 um) layer eutectic bond reaction between Au and Sn. Additionally, bond metal squeeze-out, which is the primary source of bond weakening and failure, is also characterized and a simple method suggested to predict and control overflow. The next section of the thesis discusses passive cooling solutions which show "passive" surface tension driven flow and spreads heat from a tiny hotspot to a much larger area using a liquid-vapor phase-change loop. The limit to passive cooler performance has been found to be surface tension driven flow rate within the small pores of the wick microstructure, and thus enhancing capillary driven transport has been of much interest to the microfluidic cooling community. This dissertation uses a UV-laser to easily create high functional, hybrid pin fin structures with uniformly distributed polyp like roughness, using a process much simpler and cost effective than traditional hybridization methods. Upon comparing these hybrid structures with their smooth counterparts, we observe 40 - 116% enhancement in transport. Two models were also set up to capture the effect of roughness of altering wicking rates of square pillar arrays, which performed surprisingly well and was also able to explain the results shown by outlier designs. The next section in this thesis discusses "active" cooling solutions which become indispensable in extreme heat flux (> 500 W/cm2) cooling scenarios. This dissertation first proposes a 2-level manifold concept that shortens flow path within a complex microfluidic device network and promotes massive input energy savings. This input energy savings translates into high efficiency over large areas, and thus makes these 2-level manifolds ideal for scale up, which is an important focus of the modern electronics community. This has been validated theoretically and numerically in great detail, showing that 2-level manifolds are at least 5 x more efficient that their 1-level counterparts. However, it was also identified that 2-level Manifolded coolers are extremely complex to fabricate using conventional cleanroom techniques. To address this, a double-sided anisotropic, deep Si etching recipe was developed, that was seen to be robust, reliable and repeatable - using this recipe we were able to successfully fabricate extreme area (600 mm2) devices with nominal channel dimensions, ~ 10 um. Realizing the difficulty and cost associated with Silicon processing, a detailed techno-economic feasibility study was performed, which revealed that even though all-Si coolers are still significantly expensive to manufacture, the performance metric it provided blows every other single phase cooler out of the park and justifies its use in high compute scenarios. With a few more years of process characterization, recipe development, manifolded coolers should be ready to be deployed commercially to large area power electronics. The penultimate chapter in the thesis serves to solve some of the limitations in silicon processing and aims to make it more versatile. It delineates an interesting pattern transfer technique that draws inspiration from grayscale lithography and multi-lithography. It cleverly combines these two processes to reliably create 3D, multi-level, hybrid, hierarchical structures in silicon with ease as compared to conventional methods which can only make single-level features. Several multi-level structures made using this method has been demonstrated - this method also vastly simplified or improved many issues faced in previous chapters. It solved issues related to bonding failures from UV-laser processing debris and promotes technology scale up of both passive coolers and active coolers by providing easy methods of multi-level structure creation. This finding is extremely fortunate in 2023, since hybrid features have been recently found to vastly improve device performance metrics and efficiencies in a variety of applications like microfluidics, biology, biomimetics, catalysis, sorption, desalination etc. Finally, the thesis summarizes its contributions and contextualizes their need and technology readiness by citing potential advanced technologies around the world, that could immensely benefit from superior cooling. It simultaneously discusses future paths towards the unified goal of improving power density in current electronics and methods that would likely be used to achieve it. High performance, scalable cooling solutions were found to be at the forefront of research and development that will enable this performance jump while simultaneously leaving the world cleaner and greener for future generations.
|Type of resource
|electronic resource; remote; computer; online resource
|1 online resource.
|Goodson, Kenneth E, 1967-
|Goodson, Kenneth E, 1967-
|Santiago, Juan G
|Degree committee member
|Degree committee member
|Santiago, Juan G
|Stanford University, School of Engineering
|Stanford University, Department of Mechanical Engineering
|Statement of responsibility
|Submitted to the Department of Mechanical Engineering.
|Thesis Ph.D. Stanford University 2023.
- © 2023 by Sougata Hazra
- This work is licensed under a Creative Commons Attribution Non Commercial 3.0 Unported license (CC BY-NC).
Also listed in
Loading usage metrics...