Here is the abstract you requested from the imaps_2019 technical program page. This is the original abstract submitted by the author. Any changes to the technical content of the final manuscript published by IMAPS or the presentation that is given during the event is done by the author, not IMAPS.
|A 3D Simulation Tool for Electronics Thermal Graphite Applications|
|Keywords: graphite, thermal simulation, heat spreader|
|Graphite has been used to remove heat from electronic devices for several decades. Its high thermal conductivity, low weight, and anisotropic nature make it a unique engineering material. The earliest uses began with highly oriented pyrolytic graphite (HOPG) in low volume applications. Flexible natural graphite foils, comprised of expanded and oriented flake graphite, were invented in the 1960s and used as gasket material, but the higher conductivity versions (400-to-600 W/m*K) found their way into the mass electronics cooling market with the advent of plasma displays and laptop computers in the 2000s. Meanwhile, synthetic pyrolytic graphite sheets made from polymer precursors were introduced to limited aerospace markets in the 1990s, but the biggest adoption came with the drive to make thin smart phones in the 2010s. Along the way, thermal engineers leveraged the compliant surface and lateral thermal conductivity of graphite to replace liquid and paste interfaces in certain TIM applications. Sizing an interposer, heat sink, or spreader presents an engineering challenge in any isotropic material, but the challenge increases when the material is orthotropic. Specifically, graphite foils conduct heat in a range of one hundred to more than four hundred times higher in the molecular plane than they do through the plane. This extreme ratio in a diminishingly thin layer can confound the best simulation tools and regularly challenges experienced modelers. This is particularly the case when the flexible nature of the graphite is bent in a fold, hinge, or wrapper. Even the simple task of selecting the optimal thickness from a range of standard grades required laboratory mock ups to evaluate their effectiveness. While the electronics cooling community has produced several general software tools for simulating heat transfer in electronics, most of which contain the capacity for anisotropic thermal conductivity in materials, a need was identified for a graphite-centric modeling tool that enabled the comparison of graphite-specific characteristics without taxing the modeler with system-level cooling parameters. Specialized models have been created to optimize board-level component placement, heat pipe sizing, and heat sink fin configuration, so it seemed appropriate to create one that could optimize the area and thickness of multi-layer passive heat spreaders. One of the first questions addressed by a thermal engineer for an electronics system will be “Would this application be better served by using [an isotropic] metal plate or [anisotropic] graphite heat spreader?”, followed shortly by “Since graphite comes in discrete thicknesses, which grade [thickness and thermal conductivity] will solve the problem most cost effectively?”. Compounding the calculation are the binary constraints of limiting the source (component) temperature to maximize component life while minimizing any external surface temperature to protect any human skin it might contact. The human element of using a simulation tool was also considered in the design of the tool, with an eye on minimal complexity and background information. Rather than constrain the user to drop-down lists of materials, a hovering “tool tips” bubble appears over each input value, with guidelines or sample values. While computational fluid dynamics (CFD) is frequently preferred in system level analysis where fans or liquid cooling systems are present, the effect on the fluid flow by flat plates of various thicknesses and thermal conductivity values is essentially nil, so a heat transfer coefficient (HTC) was determined to sufficiently represent the ambient environment. While conjugate heat transfer due to flowing fluid can introduce asymmetry and will influence the accuracy of the result, it is not likely to change substantially, or directionally, whether the spreader element is twenty-five micrometers thick, or two hundred and fifty. Thus, the computation domain of the tool was limited to only the heat source and spreader, and any laminates therein attached. Second-order finite elements were selected to capture the extreme gradients adjacent to the heat source, and a user-selectable swept grid layer efficiently fills the volume with hexahedral elements while enabling the user to compensate for grid-induced errors. Errors of fourteen percent have been observed in under-meshed CFD prototypes, and even single layer finite element meshes introduce small errors relative to optimized models. It has also been observed that they converge more slowly than a multi-layer discretization, a counter-intuitive finding that reverses in the case of isotropic materials. With this specialized tool, a new grid can be specified and a new model run in less than 90 seconds, enabling a designed experiment of 2x2x3 configurations (i.e., two footprints, two grades of material, and three thicknesses) to be run and summarized in less than one hour. Although the tool was intended as a calculator for assisting applications engineers to size graphite spreaders for customer, it has been used for marketing and for material development purposes to evaluate the merit of alternate thicknesses and grades, and to identify opportunities in the product portfolio. For example, smart phone manufacturers were calling for thinner and thinner grades of graphite – as thin as 10um – to fit within ever thinner devices. The practical limitations of handling such a thin material notwithstanding, they soon began to laminate the highest conductivity (1500 W/m*K) synthetic graphite sheets into two, three, and multi-layer composites with material costs that reflected the premium materials, additional processing steps, and intermediate layers. Comparing the performance of these exotic materials with single layer pure graphite materials became a topic among graphite suppliers, and the calculator enabled scores of variants to be simulated before any process changes were attempted. The result of the study was the development of thick (100um, 150um, and thicker) flexible graphite heat spreaders with moderately high (1000 W/m*K) conductivity. Limitations of the tool include transient temperature evaluation, as many electronic devices heat in bursts (IGBTs, televisions, smartphones) and extending time-to-temperature can prevent triggering of a thermal switch that throttles the application. Although the specific heat of graphite is unremarkable, its high thermal conductivity enables laminates to serve as thermal capacitors that can mitigate heat spikes; the ability to model such a feature may be added in future versions. The ability to model complex geometric shapes, such as out-of-plane surfaces, was considered, but the extra demands on the user (defining curve radii, cutout locations, etc.) was not seen to add value during the initial development.|
|Rick Beyerle, Senior Development Engineer
NeoGraf Solutions, LLC