Topological design optimization of nested channels for squeeze flow of thermal interface materials

As part of investigations into improving thermal interface performance for polymeric, particle-filled thermal interface materials, this work has accomplished a number of useful steps towards better thermal interface designs for electronic systems. Amongst original contributions, a non-contact, relatively easy-to-recreate experimental methodology for measuring bond-line thickness in electronic systems is developed and described in order to allow for experimental evaluation of designs. Perhaps more importantly, the apparatus can be used for characterization of thermal interface material constitutive behavior for use in performance predictive models, and this functionality is demonstrated with a common TIM. Another novel contribution is the development of a reduced-order model based on a resistance network for simulating squeeze flow between parallel plates that are separated by distances common in thermal interfaces. Individual resistances are computed based on smaller squeeze flow sub-problems which are then linked together with appropriate conservation equations to simulate the entire set of parallel plates. This reduced-order model allows for efficient evaluation of performance of heat spreader designs, and is extendable to many relevant fluid constitutive behavior models. The model lends itself to efficient analytical sensitivity analysis, which can be used for design analysis and optimization. The assumption of Newtonian fluid in the system simulation, along with its sensitivity analysis, is used in order to generate optimal designs in a Pareto optimal framework that weighs a measure of thermal resistance performance against squeezing flow pressure, in order to explore the tradeoffs between the conflicting objectives to find optimal designs. Families of optimal designs are generated, depending on above-stated user preference as well as initial conditions, since the design space as formulated contains local minima. The design space for the Newtonian system is explored in order to determine factors that affect performance in the various objective function metrics. Finally, the optimization framework-generated designs are compared to ad-hoc designs similar to those in literature