Metal Matrix Composites for Thermal Management of Power Electronics

Kevin Moores

Figure: Model of an AlSiC baseplate which would carry an array of power devices with AlN substrates. The baseplate is formed via a molding process in which tapered pin fins are extruded from the bottom face of the baseplate. Forced liquid cooling is applied to the fins to cool the high power electronics.

Background

Power electronics describes the sub-set of microelectronic devices used to control and condition voltage and current. They can be found in everything from car radios to DC power supplies, to systems in the space shuttle, and everywhere in between. In contrast to the high integration of logic and memory devices, traditional power electronics are still characterized by their relatively low circuit density, their use of thick film metallization traces for handling high current levels (up to 100A), and the requisite use of higher thermal conductivity packaging materials to dissipate the considerable amount of heat generated in these packages. However, recent initiatives such as the PEBB project are aimed at promoting the evolution of power electronics into a form that is more like that of the microprocessor. That is, to create families of standardized power modules which are highly integrated, smaller and cheaper than their predecessors with ever increasing functional densities (up to 750 A). However to achieve this end, considerable improvements in the thermal performance of power electronics packaging and related cooling schemes will be needed.

The primary reasons for cooling electronics are to maintain electrical performance (which may be a function of temperature) within a given envelope of operation, and to minimize thermally induced mechanical stresses due to CTE mismatches between the packaging material and the silicon chip. Choosing which materials to use in a given package usually involves a trade-off between these two concerns. Those materials which provide the best thermal conductivity (i.e. metals) help to minimize the temperature rise from ambient to the chip, but their CTE's are invariably much higher than that of silicon. Conversely, most materials with comparable CTE's are generally poor thermal conductors.

Heat flux from today's power devices is generally less than 75 W/cm² at the chip. Through the use of conductive substrates and heat spreaders, dissipation at the heat sink can be reduced to less than 1 W/cm². At this level of heat flux, natural or forced air convection is often sufficient to maintain the device temperature within acceptable limits. However, the next generation of power packages described in the literature are predicted to exhibit heat fluxes on the order of 100 to 300 W/cm² at the die, and perhaps higher. To dissipate such high loads without increasing the size of the package dramatically, will require significantly improved thermal performance and cooling schemes.

One approach to be considered is the use of MMC (metal matrix composite) materials such as AlSiC. The marriage of Al and SiC provides the relatively high thermal conductivity of a metal, with the relatively low CTE of an ceramic. Recent studies have looked at using MMC's to replace the copper clad substrates/heat spreaders of today's technology. However, the need to couple this arrangement with a copper or aluminum, liquid flow through heat sink (which would be required for the predicted heat flux levels) still presents the problems of CTE mismatch at the interface to the heat sink. Because it is possible to form MMC's into somewhat complex geometries, the heat sink portion of the package could be integrated directly into the AlSiC substrate. This would essentially reduce the module to a (3) material structure with silicon chip, the AlSiC substrate, and some attach material between the two. Since the Silicon and AlSiC have similar CTE's, the thermal stresses induced at the interface should be considerably reduced in comparison to today's package configurations or to those studied thus far in the literature.

Objectives

The project has been divided into a series of tasks. They include:

Task 1: Baseline Calculations of Thermal Performance

Using both analytical and numerical calculations, a baseline performance level will be determined for the existing package topology (ie with separate heat sink). This will be used for comparison with enhanced designs determined later in the study. The calculations will include a thermal resistance network analysis to determine the approximate temperature rise along the packaging stack. FEA and CFD analysis will be used to estimate the performance at the heat sink end.

Task 2: Combined Conduction and Fluid Flow Modeling for Cooling Performance Evaluation of Flow Through Heat Sink

A full 3-D computational fluid dynamics analysis will be performed to determine the hydraulic and thermal performance of the baseplate with integral tapered pin fins. A unit cell approach will be used, taking advantage of symmetry of the array, to minimize computational requirements. A parametric study will be performed to determine the effect of various design parameters on the overall performance of the baseplate. These will include:

• Cooling Fluid
• Pin Height
• Flow Rate
• Pin Diameter
• Heat Flux
• Pin Spacing

Task 3: Design of Baseplate to Heat Sink Attachment

This will involve the mechanical design of the flow through cold plate assembly. The following key activities will be carried out:

• The effect of the casing (heat sink) material will be examined. Since the cold plate in many applications may be mounted near a hot local environment (e.g., engine housing), a metal casing (heat sink) will result in extraneous thermal loads. A polymer housing may be preferable for such applications. From Task 2 above there will be some guidance on the thermal performance implications of the two housing materials. For both polymer and aluminum housings, the choice of manufacturing methods will be evaluated. A web-based search will be carried out to find possible vendors for the housings.

• Methods to attach the baseplate and casing (heat sink) will be studied. The most commonly used joining technique may be an elastomeric O-ring, along with bolts at the four corners. The choice of materials for the O-ring, as well as the O-ring groove design will be systematically evaluated. For polymeric casing (heat sink), the use of threaded metal inserts will be evaluated and design guidelines provided.

• Connectors for the fluid inlet and outlet will be evaluated. In this task, the experience existing in the literature for the flow through avionics modules will be utilized. Connectors that allow for easy connection and disconnection of the fluid flow lines are generally used in such applications.

• Stress estimations for the baseplate following assembly will be made. It is expected that some guidance on the prevailing concerns (if any) will be available from the actual hardware components to be used in the prototype testing experiments described under Task 4 below, with which some overlap in timing is expected. Possible sources of stress to be looked at include the internal pressure due to the fluid, thermally induced stresses due to the electrical power dissipation, and mechanical stresses due to bolting.

Task 4: Experimental Evaluation of a Cold Plate Assembly

Based on Tasks 1 to 3, a prototype cold plate assembly will be fabricated and thermally evaluated in an existing flow loop facility at CALCE. This facility is equipped with a pump to circulate distilled water through a cold plate assembly. The temperature of the inlet water is controlled through a regulated bath. A flow meter allows the determination of the flow rate. Pressure drop measurements and temperature measurements can be made using various sensors, connected to a data-acquisition system.

The baseplate and casing (heat sink) will be provided by LEC. The test module will be assembled and instrumented at CALCE. Heating will be provided by etched film thermo-foil heaters. These will be attached to the upper side of the baseplate, using a thermally conductive adhesive.

The validation will be carried out for selected flow rates and power dissipation levels. For each test, thermocouple measurements at selected locations will be made and compared with the model predictions. Once validated, the model will be usable for the design of cold plates for a variety of applications.