Search

Heat dissipation from three-dimensional (3D) chip stacks can cause large thermal gradients due to the accumulation of dissipated heat and thermal interfaces from each integrated die. To reduce the overall thermal resistance and thereby the thermal gradients, this publication will provide an overview of several studies on the formation of sequential thermal underfills that result in percolation and quasi-areal thermal contacts between the filler particles in the composite material. The quasi-areal contacts are formed from nanoparticles self-assembled by capillary bridging, so-called necks. Thermal conductivities of up to 2.5 W/m K and 2.8 W/m K were demonstrated experimentally for the percolating and the neck-based underfills, respectively. This is a substantial improvement with respect to a state-of-the-art capillary thermal underfill (0.7 W/m K). Critical parameters in the formation of sequential thermal underfills will be discussed, such as the material choice and refinement, as well as the characteristics and limitations of the individual process steps. Guidelines are provided on dry versus wet filling of filler particles, the optimal bimodal nanosuspension formulation and matrix material feed, and the over-pressure cure to mitigate voids in the underfill during backfilling. Finally, the sequential filling process is successfully applied on microprocessor demonstrator modules, without any detectable sign of degradation after 1500 thermal cycles, as well as to a two-die chip stack. The morphology and performance of the novel underfills are further discussed, ranging from particle arrangements in the filler particle bed, to cracks formed in the necks. The thermal and mechanical performance is benchmarked with respect to the capillary thermal and mechanical underfills. Finally, the thermal improvements within a chip stack are discussed. An 8 - or 16-die chip stack can dissipate 46% and 65% more power with the optimized neck-based thermal underfill than with a state-of-the-artcapillary thermal underfill.

An overview of the thermal management landscape with focus on heat dissipation from three-dimensional (3D) chip stacks is provided in this study. Evolutionary and revolutionary topologies, such as single-side, dual-side, and finally, volumetric heat removal, are benchmarked with respect to a high-performance three-tier chip stack with an aggregate power dissipation of 672 W. The thermal budget of 50 K can be maintained by three topologies, namely: (1) dual-side cooling, implemented by a thermally active interposer, (2) interlayer cooling with four-port fluid delivery and drainage at 100 kPa pressure drop, and (3) a hybrid approach combining interlayer with embedded back-side cooling. Of all the heat-removal concepts, interlayer cooling is the only approach that scales with the number of dies in the chip stack and hence enables extreme 3D integration. However, the required size of the microchannels competes with the requirement of low through-silicon-via (TSV) heights and pitches. A scaling study was performed to derive the TSV pitch that is compatible with cooling channels to dissipate 150 W/cm2 per tier. An active integrated circuit (IC) area of 4 cm2 was considered, which had to be implemented on the varying tier count in the stack. A cuboid form factor of 2 mm × 4 mm × 2.55 mm results from a die count of 50. The resulting microchannels of 2 mm length allow small hydraulic diameters and thus a very high TSV density of 1837 1/mm2. The accumulated heat flux and the volumetric power dissipation are as high as 7.5 kW/cm2 and 29 kW/cm3, respectively.

We demonstrate the lid-integral silicon cold-plate topology as a way to bring liquid cooling closer to the heat source integrated circuit (IC). It allows us to eliminate one thermal interface material (TIM2), to establish and improve TIM1 during packaging, to use wafer-level processes, and to ease integration in first-level packaging. We describe the integration and analyze the reliability aspects of this package using modeling and test vehicles. To compare the impact of geometry, materials, and mechanical coupling on warpage, strains, and stresses, we simulate finite element models of five different topologies on an organic land-grid array (LGA) carrier. We measure the thermal performance in terms of thermal resistance from cold-plate base to inlet liquid and obtain 15 mm2 K/W at 30 kPa pressure drop across the package. We build two different topologies using silicon cold-plates and injection-molded lids. Gasket-attached cold-plates pass an 800 kPa pressure test, and direct-attached cold-plates fracture in the cold-plate. The results advise to use a compliant layer between cold-plate and manifold lid and promise a uniformly thick TIM1 layer in the Si–Si matched topology. The work shows the feasibility of composite lids with integrated silicon cold-plates in high heat flux applications.

This work presents enhanced composite joints that support both electrical or thermal transport in electronic packages. The joints are sequentially formed by applying a nanoparticle suspension, evaporating a solvent, self-assembling of nanoparticles by capillary bridging, and the formation of so called “necks” between micron-sized features. This sequence is used to either form low temperature electrical joints under copper pillars or enhanced percolating thermal underfills with areal contacts between filler particles of the composite. The report discusses processing aspects of the capillary bridges evolution and of uniform neck formation, it discusses strategies to achieve mechanically stable necks, and it compares the performance of the achieved joints against state-of-the-art solutions. The capillary bridge evolution during liquid evaporation was investigated in copper pillar arrays and random particle beds. The vapor-liquid interface first penetrates locations of low pillar or particle density resulting in a dendritic fluid network. Once the network breaks up individual necks form. For aqueous nano-suspensions highly uniform necks with high yield were obtained by evaporation at 60°C. Isothermal conditions were preferred to yield equal neck counts at the cavity’s top and bottom surfaces. Mechanically stable silver necks required an annealing at only 150°C, dielectric necks an annealing at 140°C with a bi-modal approach. Therein polystyrene nanoparticles occupy interstitial positions in densly packed alumina necks, then melt and adhere to the alumina. The electrical necks showed a shear strength of 16 MPa, equivalent to silver joints used in power electronic packages. The thermal necks yielded a thermal conductivity of up to 3.8 W/mK, 5-fold higher than commercially available capillary thermal underfills.

Nanofluids have been proposed to improve the performance of microchannel heat sinks. In this paper, we present a systematic characterization of aqueous silica nanoparticle suspensions with concentrations up to 31 vol %. We determined the particle morphology by transmission electron microscope imaging and its dispersion status by dynamic light scattering measurements. The thermophysical properties of the fluids, namely, their specific heat, density, thermal conductivity, and dynamic viscosity were experimentally measured. We fabricated microchannel heat sinks with three different channel widths and characterized their thermal performance as a function of volumetric flow rate for silica nanofluids at concentrations by volume of 0%, 5%, 16%, and 31%. The Nusselt number was extracted from the experimental results and compared with the theoretical predictions considering the change of fluids bulk properties. We demonstrated a deviation of less than 10% between the experiments and the predictions. Hence, standard correlations can be used to estimate the convective heat transfer of nanofluids. In addition, we applied a one-dimensional model of the heat sink, validated by the experiments. We predicted the potential of nanofluids to increase the performance of microchannel heat sinks. To this end, we varied the individual thermophysical properties of the coolant and studied their impact on the heat sink performance. We demonstrated that the relative thermal conductivity enhancement must be larger than the relative viscosity increase in order to gain a sizeable performance benefit. Furthermore, we showed that it would be preferable to increase the volumetric heat capacity of the fluid instead of increasing its thermal conductivity.

We report an experimental investigation of a novel, high performance ultrathin manifold microchannel heat sink. The heat sink consists of impinging liquid slot-jets on a structured surface fed with liquid coolant by an overlying two-dimensional manifold. We developed a fabrication and packaging procedure to manufacture prototypes by means of standard microprocessing. A closed fluid loop for precise hydrodynamic and thermal characterization of six different test vehicles was built. We studied the influence of the number of manifold systems, the width of the heat transfer microchannels, the volumetric flow rate, and the pumping power on the hydrodynamic and thermal performance of the heat sink. A design with 12.5 manifold systems and 25 μm wide microchannels as the heat transfer structure provided the optimum choice of design parameters. For a volumetric flow rate of 1.3 l/min we demonstrated a total thermal resistance between the maximum heater temperature and fluid inlet temperature of 0.09 cm2 K/W with a pressure drop of 0.22 bar on a 2×2 cm2 chip. This allows for cooling power densities of more than 700 W/cm2 for a maximum temperature difference between the chip and the fluid inlet of 65 K. The total height of the heat sink did not exceed 2 mm, and includes a 500 μm thick thermal test chip structured by 300 μm deep microchannels for heat transfer. Furthermore, we discuss the influence of elevated fluid inlet temperatures, allowing possible reuse of the thermal energy, and demonstrate an enhancement of the heat sink cooling efficiency of more than 40% for a temperature rise of 50 K.

A self-contained, small-volume liquid cooling system for thin form-factor electronic equipment (e.g., blade server modules) is demonstrated experimentally in this paper. A reciprocating water flow loop absorbs heat using mesh-type microchannel cold plates and spreads it periodically to a larger area. From there, the thermal energy is interchanged via large area, low pressure drop cold plates with a secondary heat transfer loop (air or liquid). Four phase-shifted piston pumps create either a linearly or radially oscillating fluid flow in the frequency range of 0.5–3 Hz. The tidal displacement of the pumps covers 42–120% of the fluid volume, and, therefore, an average flow rate range of 100–800 ml/min is tested. Three different absorber mesh designs are tested. Thermal and fluidic characteristics are presented in a time-resolved and a time-averaged manner. For a fluid pump power of 1 W, a waste heat flux of 180 W/cm2(ΔT=67 K) could be dissipated from a 3.5 cm2 chip. A linear oscillation flow pattern is advantageous over a radial one because of the more efficient heat removal from the chip and lower hydraulic losses. The optimum microchannel mesh density is determined as a combination of low pump losses and high heat transfer rates.