MIM of thermal management materials in microelectronics(20.4-20.6)

Table of Contents

• 20.4 Tungsten-copper

• 20.4.1 Powders

• 20.4.2 Feedstock preparation

• 20.4.3 Injection molding

• 20.4.4 Debinding and sintering

• 20.4.5 Infiltration

• 20.4.6 Thermal properties

• 20.4.7 Example applications

• 20.5 Molybdenum-copper

• 20.5.1 Powders

• 20.5.2 Feedstock preparation

• 20.5.3 Injection molding

• 20.5.4 Debinding and sintering

• 20.5.5 Infiltration

• 20.5.6 Thermal properties

• 20.5.7 Example applications

• 20.6 Conclusions

20.4 Tungsten-copper

MIM of W combined with 10-20 wt% Cu was first patented in 1991, but much learning has taken place since then. As with MIM Cu, the primary challenge is attaining high sintered densities and high conductivities. Densification of W-Cu is greatly hindered by the extremely low solubility of W in Cu. Densification can be enhanced by the addition of transition metal elements, such as Ni, Co, and Fe, which have substantial solubility for W, but are detrimental to the thermal conductivity. Sintering of high-purity W-Cu to near full density requires a submicron W particle size. A few W-Cu powders suitable for MIM are commercially available. Elemental W and Cu powders can be bought separately, but the method of producing the composite powder is critical. Alternatively, W powder can be injection molded and then infiltrated with Cu.

Successful processing of W-Cu parts via MIM requires a tailored particle size distribution for both moldability and sinterability. The key requirements for processing highthermal conductivity W-Cu components via MIM are described in the following section.

20.4.1 Powders

Almost all commercially available W powders are produced from oxide reduction.Non-oxide impurities are generally less than 0.05 wt%, as required for high-thermal conductivities. Sizes are typically 3-5μm, but particle sizes below 1μm, needed for liquid-phase sintering of W-Cu, are commercially available. Composite W-Cu powders can be produced by co-reducing tungsten oxides with copper oxides. Such powders are also commercially available and provide a homogeneous distribution of the Cu that is ready for compounding without further processing. Mixing of reduced W and Cu powders is possible but requires special care.

Table 20.4 W-Cu powder characteristics

Oxide-reduced W powders must be deagglomerated, typically by milling, to produce suitable MIM feedstock. Otherwise the solids loading will be too low at a moldable viscosity. The deagglomerated W powder can be mixed with Cu powder using a double-cone, v-cone, or Turbula  mixer, but this adds a process step and requires a Cu particle size near that of the W powder. If the Cu particle size is much larger than the particle size of the W powder, pools of liquid form that are less effective for densification. High-purity Cu powders with particles less than 10μm (-10μm) are expensive. Instead, coarser Cu powders, such as the ones given in Table 20.2, can be combined with the W powder during milling.

Rod milling, ball milling, planetary milling, and attritor milling have all been used to prepare W-Cu powders. These methods are listed in order of increasing energy input. Coarser Cu powders can be used as the milling intensity increases. Higher energy milling processes, especially attritor milling, can embed W particles into larger Cu particles to produce a homogeneous composite powder, but the particle size and distribution depend on several milling parameters, including speed, time, and powder loading.

A significant obstacle to milling elemental powders is the increased likelihood of contamination from the milling media or liner. Steel media and liners are generally unsuitable, since the resulting Fe and Si contamination is highly detrimental to the thermal conductivity. Cemented carbide media with cemented carbide or polyethylene liners have been used more successfully but can be cost prohibitive for production quantities.

Some or all the Cu powder that is milled with the reduced W powder can be replaced with copper oxide, usually cuprous oxide (Cu2O). Cuprous oxide powders are lower cost than reduced Cu powders and can provide good homogeneity when milled with W powders. The copper oxide is reduced during sintering. A combination of Cu and Cu2O can form a eutectic liquid at 1065℃ that may enhance sintering in wet hydrogen. Reduction experiments have shown that the Cu2O is mostly reduced in dry hydrogen below this temperature during heating.

The characteristics of example components for preparing W-Cu powders are given in Table 20.4 along with those of a co-reduced W-15wt% Cu powder. Representative scanning electron micrographs of these powders are given in Fig. 20.13. The W powder can be combined with a 10μm Cu powder by mixing or milling. It can also be milled with a Cu2O powder or the Cu powders given in Table 20.2. Rod milling is not sufficiently energetic to break down Cu2O particles but can be used to obtain a homogeneous distribution with a 25μm Cu powder as shown in the micrographs in Fig. 20.14. Tungsten particles encapsulate Cu particles in co-reduced powders which can be directly used to produce feedstock in the as-supplied condition.

Fig. 20.13 Scanning electron micrographs of (A) submicron W powder. (B) -10μm Cu powder. (C) Cu2O powder. (D) Co-reduced W-15Cu powder

Fig. 20.14 Scanning electron micrographs of (A) rod-milled W-17.12Cu2O powder. (B) Rodmilled W-15Cu powder

Table 20.5 Examples of MIM W-Cu feedstock preparation 

20.4.2 Feedstock preparation

Demonstrations of MIM W-Cu have been almost exclusively with wax-polymer binders. A notable exception is the use of a nonaqueous binder, such as cyclohexane, and a dispersant, such as an acrylic acid-based polyelectrolyte, which can be frozen after molding and debound by sublimation. The small particle sizes required for densification have poor packing characteristics and consequently the solids loadings of these powders for injection molding are relatively low, generally ranging from 52 vol% to 58 vol %, giving tooling scale-up factors of 1.22-1.18 for a sintered density of 95% of theoretical. Because of the dispersion difficulties, high shear rate continuous mixers are preferred for compounding. A summary of various binder compositions and compounding details is given in Table 20.5.

20.4.3 Injection molding

While any of the feedstocks shown in Table 20.5 can be used to injection mold test bars, several factors must be considered for molding complex thermal management components with thin walls and large numbers of pins or feedthroughs. Wide particle size distributions produced by lightly attritor milling a submicron W powder with a coarse Cu powder are recommended for improved molding behavior. For heat sinks with pins, a higher-strength binder is generally required to prevent them from breaking upon ejection. This necessitates a high fraction of backbone polymer (polypropylene or polyethylene) or one with a high molecular weight.

For electronic packages, finding a suitable gate location for smooth flow through thin walls is often a challenge due to flatness requirements on key surfaces and the need for core pulls. The high conductivity of the feedstock increases the risk of premature solidification of the melt, resulting in non-fill. Mold-filling simulations can be used to optimize gate location and type, for example concentric or film gate, for specific part geometries.

20.4.4 Debinding and sintering

Debinding of W-Cu is typical of other MIM materials. A combination of solvent debinding to dissolve the major binder component and thermal debinding to pyrolize the remaining backbone polymer through the open pore space is the preferred technique for the wax-polymer systems described in Table 20.5. However, thermal debinding alone or a combination of wicking and thermal debinding can also be used. Hydrogen is required to reduce both tungsten oxides and copper oxides. As noted in the previous section, reduction of copper oxides occurs in the range from 550℃ to 680℃. Sintering in dry hydrogen reduces tungsten oxides at 800℃. A typical debinding cycle includes a slow (about 2℃/min) ramp rate to 500℃ to remove the binder and then further heating to 800-950℃ for oxide reduction and to produce sufficient strength for handling.

The effects of sintering at various temperatures for 2 h in dry hydrogen on the density and oxygen content of W-15Cu are shown in Fig. 20.15. The density does not significantly increase until the temperature exceeds 1100℃, which is above the melting temperature of Cu. Even at 1100℃, the oxygen content is still about 0.25 wt%, but it decreases linearly with sintering temperature. Near full density is achieved at a sintering temperature of 1300℃ with an oxygen content of 150 ppm. An example micrograph showing the composite microstructure with 1μm grains is given in Fig. 20.16.

Fig. 20.15 Effect of sintering temperature on the density and oxygen content of W-15Cu

Fig. 20.16 Micrograph of W-15Cu sintered at 1300℃ for 60min

Liquid phases generally enhance sintering, but the extremely low solubility of W in Cu severely limits densification. Achievement of full density requires a submicron W particle size to promote solid-state densification at temperatures below 1400℃ to avoid significant Cu evaporation. The effects of W particle size and sintering temperature on the density of W-10Cu are shown in Fig. 20.17. As the Cu content increases, the dependence of the sintered density on particle size decreases, as shown in Fig. 20.18.

Fig. 20.17 Effect of W BET (Brunauer-Emmett-Teller) particle size and sintering temperature on the sintered density of W-10Cu after compacting at 70MPa and sintering for 1 h in hydrogen

Fig. 20.18 Effect of W BET particle size on the density of W, W-10Cu, and W-20Cu after compacting at 70MPa and sintering for 1 h in hydrogen at 1400℃

Solid-state densification of the W skeleton can continue even as full density is approached. Since few pores remain for the liquid Cu to fill, further skeletal sintering forces it to the surface, resulting in Cu "bleed-out" An example of this phenomenon is shown in Fig. 20.19. Although skeletal sintering increases with higher temperatures, less Cu bleed-out is seen as the temperature reaches 1400℃ since liquid Cu at the surface rapidly evaporates. Over-sintering increases the W to Cu ratio and results in a Cu depletion layer at the surface. Transition metal impurities, especially Fe, can accelerate sintering of the W skeleton at lower temperatures, leading to greater amounts of bleedout. Both excess Cu and depleted Cu at the surface of MIM parts are often associated with part warpage and can cause plating problems.

Fig. 20.19 Example of Cu bleed-out at edges of a W-Cu specimen

20.4.5 Infiltration

Instead of liquid-phase sintering a mixture of W and Cu powders, W-Cu composites can also be produced by placing pressed Cu powder or wrought Cu pieces in contact with a porous W preform and heating the combination above the melting temperature of Cu. In a dry hydrogen atmosphere, the liquid Cu will infiltrate the W preform. Infiltration has been used to produce electrical contacts for many decades, but these applications require a low degree of shape complexity and the preform can be produced by die pressing W powder at high compaction pressures. The preform generally gives very little dimensional change during infiltration and is infiltrated with a volume fraction of Cu approximately equal to its porosity. Since high green densities are not possible with injection molding due to the limitation of the critical solids loading, the preform must be presintered to the density corresponding to the W volume fraction in the final composite. This requires a fine starting powder or high sintering temperature since sintering activators, such as Co, Ni, or Fe are highly detrimental to the thermal conductivity. A process map for the effect of W particle size and sintering temperature on density is given in Fig. 23.12 in Chapter 23.

Infiltration can be combined with presintering in a single thermal cycle. To achieve W contents of 80 wt% or more, a submicron W particle size is required to densify the preform at temperatures below 1500℃ after the Cu melts and infiltrates it. In this case of infiltration sintering, the injection-molded W preform displays 12%-14% shrinkage depending on the sintering temperature and Cu content. Densities above 99% of theoretical can be achieved.

20.4.6 Thermal properties

The thermal properties of MIM W-Cu depend mostly on the composition, but porosity and microstructure are also factors. Several models have been developed to estimate the thermal conductivity of composites. The rule of mixtures and the inverse rule of mixtures are the simplest and provide upper and lower boundaries on the predicted thermal conductivity. A more recent model considers the effect of grain shape on the thermal conductivity of liquid-phase-sintered composites. The predictions of these models for the thermal conductivity of pure, pore-free W-Cu versus Cu content are plotted in Fig. 20.20 in comparison to experimental results. Despite best efforts to maintain high purities and high densities, most reported values are still below the predictions of the inverse rule of mixtures.

Fig. 20.20 Model predictions for the effect of Cu content on the thermal conductivity of pure, pore-free W-Cu

Impurities are highly detrimental to the thermal conductivity. Based on the Wiedemann-Franz relationship and Nordheim's Rule, the predicted effect of Fe impurities on the thermal conductivity of W-10Cu and W-15Cu is plotted in Fig. 20.21. As the transition metal content increases, the effect of Cu volume fraction decreases, with little predicted difference in the thermal conductivity of W-10Cu and W-15Cu at impurity levels of about 0.25wt% or higher. These impurity levels are typically associated with intentionally added sintering aids. Decreases in thermal conductivity of 40W/(mK) are predicted with impurity levels of 0.1wt%, which may result from unintentional contamination.

Fig. 20.21 Model predictions for the effect of transition metal impurities on the thermal conductivity of pore-free W-Cu

Porosity of 1%-3% is typical for MIM W-Cu parts. Model predictions for the effects of porosity on the thermal conductivity of W-15Cu are shown in Fig. 20.22. Experimental data are also given for comparison showing that porosity has a relatively minor role on the thermal conductivity of W-Cu and does not explain the measurement scatter. Additional model predictions show a greater effect of an interfacial resistance with a W-W thermal boundary conductance of 108 W/(m2K) for a 1μm grain size decreasing the thermal conductivity by about 25%, in line with experimental results. Thus, larger grain sizes and lower contiguities are favorable for increasing thermal conductivity but are difficult to obtain in practice with liquid-phase sintering and are more readily produced by infiltration of a coarse W powder presintered at a high temperature.

Fig. 20.22 Model predictions for the effect of porosity on the thermal conductivity of pure W-15Cu

The thermal expansion coefficient is mostly dependent on the ratio of the components of the composite and the microstress between them. Porosity has virtually no effect on the thermal expansion of sintered components, but the effects of microstructure are less well defined. The models consider the stress-coupling of the phases and provide estimates for the thermal expansion coefficient of W-Cu for various Cu contents, as shown in Fig. 20.23. The Kerner model is closest to the average thermal expansion coefficient of W-15Cu.

Fig. 20.23 Predicted effects of Cu content on the thermal expansion coefficient of W-Cu from three different models

20.4.7 Example applications

Most W-Cu heat sinks and heat spreaders are produced from infiltrated sheets, but MIM enables fabrication of complex geometries, such as chip submounts and stem heat sinks for optoelectronic devices, base metals and fin-shaped heat sinks for integrated circuits, bases for multi-chipped boards, and microelectronic packages. Example heat spreaders are shown in Fig. 20.24. The heat spreader in Fig. 20.24A is similar to a design fabricated from infiltrated sheets but demonstrates that threaded fasteners can be incorporated in a net shaped MIM component. The part in Fig. 20.24B is an example of a complex boiler designed to transfer heat from a high-powered microprocessor to a liquid cooling column.

Fig. 20.24 Example W-Cu (A) Heat spreaders. (B) Boilers

Parts that lack a flat surface require support during sintering. Since they undergo significant sintering shrinkage, fixtures must be designed to support both the green sample and the much smaller sintered part. For example, the components shown in Fig. 20.24 were supported by alumina substrates in which a rectangular section was machined to the green dimensions of the rectangular section on the bottom of the parts. The depth of this rectangular slot was identical to the thickness of the rectangular area of the sintered part. In this way, the outer edges of the sintered part are supported at the end of the sintering cycle. Despite a high liquid volume fraction, the rigid W skeleton that develops early during sintering of W-Cu resists slumping. In comparison, tungsten heavy alloys, in which the liquid phase has substantial solubility for W, show significant slumping and loss of dimension precision at similar liquid volume fractions, as further discussed in Chapter 23.

Pilot production of the parts shown in Fig. 20.24 was able to hold tolerances on critical dimensions to a standard deviation
0.1%, while all dimensions could be kept within a standard deviation of
0.3%. W-Cu prototype packages similar to that shown in Fig. 20.3A have been produced with tolerances of ±0.1% with a dimensional yield of 57%. A W-Cu housing for a highfrequency circuit achieved tolerances of 0.06%-0.22%, 0.25%-0.29%, and 0.23%-0.35% for the length, width, and height, respectively. Warpage becomes a greater concern as the sintered density increases, so densities are usually limited to about 97% of theoretical. This density is sufficient for hermeticity with little detriment to the material properties.

20.5 Molybdenum-copper

MIM processing of Mo-Cu is similar to that of W-Cu but has been less researched. As with W, the low solubility of Mo in Cu hinders densification, and transition metal elements that enhance densification are detrimental to the thermal conductivity. In comparison to the W particle sizes used for W-Cu, slightly larger Mo particle sizes can be used for both liquid-phase sintering and for infiltration, but the particle sizes of Mo powders with commercial availability are limited. Because of the higher ductility of Mo and its subsequent inability to be attrited to smaller particle sizes, fewer options are available for producing composite Mo-Cu powders. Successful processing of Mo-Cu parts via MIM requires specific attention to the particle size distribution, particle morphology, and thermal cycle as described in the following section.

20.5.1 Powders

Like W powders, Mo powders are generally produced from oxide reduction; however, the range of sizes available is much narrower. Particle sizes are typically 2-4μm, but production of finer powders is difficult. Deagglomeration by milling is generally required, but high-energy processes, such as attritor milling, are less successful since Mo is more ductile than W and tends to form platelets, which pack poorly. Mixing with a Cu powder with a particle size similar to that of the Mo powder is generally required for high sintered densities. Coreduction of molybdenum oxides and copper oxides is possible, but such co-reduced powders are not commercially available. In all cases, high-purity powders are needed to achieve high-thermal conductivities.

The characteristics of two MIM grade Mo powders are given in Table 20.6. Scanning electron micrographs of the powders are presented in Fig. 20.25. The low pycnometer density of the 2.5 μm Mo powder is due to oxide impurities, which can be reduced during sintering.

Fig. 20.25 Scanning electron micrographs of (A) 2.5μm Mo powder. (B) 4.1μm Mo powder

20.5.2 Feedstock preparation

Wax-polymer binder systems have proven suitable for injection molding of both elemental Mo and Mo-15 Cu compositions. A plot showing the torque required to mix Mo powders at different solids loadings with a waxpolymer binder is shown in Fig. 20.26. From this plot, the critical solids loading was estimated at 64 vol% for the 4.1μm Mo powder and 60 vol% for the 2.5μm Mo powder. The coarser Mo powder can be used for molding Mo skeletons at a solids loading of 62 vol% for subsequent infiltration. The finer Mo powder is required for liquid-phase sintering of Mo mixed with Cu, and the mix can be molded at a solids loading of 58 vol% (tool scale-up factor of 1.18 for a sintered density of 95% of theoretical). Milling techniques tend to deform the Mo particles and significantly decrease solids loadings.

Fig. 20.26 Mixing torque versus solids loading for two MIM grade Mo powders showing critical solids loadings at 60 vol% for the 2.5μm Mo powder and 64 vol% for the 4.1μm Mo powder

20.5.3 Injection molding

Coarser Mo powders provide better molding behavior than mixes of finer Mo powders with Cu. Heat spreaders and electronic packages have been injection molded from elemental Mo feedstock, although increased packing pressure was needed to prevent cooling cracks along the edges of the thicker section opposite the threaded studs of the component shown in Fig. 20.24A and at the base of the studs. The same parts could not be molded from Mo-Cu feedstock without cooling cracks. Cracking was also a problem with molding Mo-Cu rectangular bars but could be eliminated with sufficient packing pressure. The packing pressure did not transfer well to the Mo-Cu parts and this was attributed to the high conductivity of the feedstock combined with a less than optimal particle size distribution. Similar problems with molding AlN heat spreaders with the same binder system were greatly alleviated by widening the particle size distribution . Thus, adjustments to the particle size distribution of Mo-Cu mixes may also improve their moldability.

20.5.4 Debinding and sintering

Debinding of Mo-Cu is similar to that of W-Cu. Hydrogen is required during thermal debinding to reduce both Mo oxides and Cu oxides. Copper hinders the reduction of molybdenum oxides, increasing its initial reduction temperature into the range of 750-800℃.

The solubility of Mo in Cu is much higher than the solubility of W in Cu, but it is still low enough to make liquid-phase sintering to full density difficult for the Mo contents of interest for thermal management applications. As for W-Cu, the early formation of a solid refractory metal skeleton slows the densification rate to that of solid-state diffusion . In this case, the Cu contributes little to densification and simply fills pore space. The liquid Cu hinders densification since it enhances growth of the Mo grains more than for W-Cu. As the grains become larger, the driving force for densification decreases. Thus, Mo-Cu must be sintered at temperatures comparable to those used for the solid-state sintering of Mo.

For liquid-phase sintering of Mo-18Cu, sintering temperatures well above the melting temperature of Cu are required to achieve high sintered densities. These high temperatures can result in significant weight loss due to Cu evaporation. The effects of sintering temperature on the final densities and weight losses of Mo-18Cu are shown in Fig. 20.27 for the Mo powders described in Table 20.6 after liquid-phase sintering for 4 h. High sintered densities require the 2.5μm Mo powder and a sintering temperature of 1400℃. Higher temperatures result in significant Cu evaporation and a decrease in relative sintered densities. The effectiveness of the 4 h sintering time is displayed in Fig. 20.28, which compares the sintering of Mo-18 Cu with that of elemental Mo. Even with the 2.5μm Mo powder, a density of only 85.2% of theoretical is achieved for Mo-18Cu after 1 h at a sintering temperature of 1400℃, but the density increases to 95.7% of theoretical after 4 h at 1400℃. The slow sintering kinetics are unusual for liquid-phase sintering and are more typical of solid-state sintering of the elemental Mo powders, which continue to densify as the sintering time is increased from 1 to 4 h.

Fig. 20.27 Effect of particle size and sintering temperature on the density of Mo-18Cu. The hold time at the sintering temperature was 4 h. Weight loss is mostly due to Cu evaporation

Fig. 20.28 Effect of particle size and sintering time at 1400℃ on the density of elemental Mo and Mo-18Cu

Fig. 20.29 Microstructure of injection molded and liquid-phase-sintered Mo-15Cu. The sample was sintered at 1450℃ for 3 h in H2.

The high tolerances demanded by the microelectronics industry require precision sintering. The heating rate and temperature gradients in the furnace can result in distortion. For example, a heating rate of 10℃/min gives warping of Mo heat spreaders. This distortion can be eliminated by decreasing the heating rate to 2℃/min. But often the dimensions of the sintered parts are related to previous processing stages, requiring tight quality control of every step of the injection molding process, starting from the reception of the powder.

20.5.5 Infiltration

Infiltration is an attractive processing route for Mo-Cu, since the sintering behavior of the Mo skeleton is improved when the Cu is not present. Also, coarser powders can be used, since sintering temperatures above 1400℃ can be utilized without the concerns of Cu evaporation. Further, the Mo skeletal density can be controlled by changing the sintering cycle to allow infiltration with different amounts of Cu. A process map for the effect of Mo particle size and sintering temperature on density is given in Fig. 23.14 in Chapter 23. For the sizes of Mo powders typically available, sintering temperatures of 1400-1800℃ are needed to achieve porosity levels equivalent to the target Cu volume fractions for thermal management applications.

A 4.1μm Mo powder has been injection molded and sintered at 1450℃ for 8 h to a density of about 83% of theoretical for infiltration with 15 wt% wrought oxygen-free high-conductivity (OFHC) Cu. Its microstructure is shown in Fig. 20.30. The density was greater than 97% of theoretical and the grain size was 4.3μm. Despite the longer sintering time at 1450℃, the grain size was smaller than that of the liquid-phase sintered sample in Fig. 20.29. Thus, the presence of Cu during liquid-phase sintering contributes significantly to grain growth, although it has little effect on densification.

Fig. 20.30 Microstructure of injection molded and infiltrated Mo-15Cu

20.5.6 Thermal properties

The same factors of Cu content, impurities, porosity, and microstructural features that control the thermal conductivity of W-Cu also apply to Mo-Cu. The sintering temperature and cooling rate can also affect the thermal conductivity of Mo-Cu due to the higher solubility of Mo in Cu, which increases from much less than 1 wt% at 1150℃ to 1.5wt% at 1400℃. At the higher end of this range, the amount of Mo dissolved in the Cu is sufficient to lower the thermal conductivity if the cooling rate is rapid enough to prevent it from precipitating. For example, liquid-phase sintering or infiltrating at a temperature of 1400℃ and furnace cooling, results in thermal conductivities of 110-130W/(m K) despite the use of high-purity Cu and Mo powders and careful avoidance of impurities during subsequent processing steps. However, thermal conductivities of 160W/(m K) can be achieved by infiltrating at 1150℃ and furnace cooling. This thermal conductivity is very close to the prediction of 169W/(m K) for this system.

Cooling at a controlled rate of 1℃/min from a sintering temperature of 1400℃ to 1050℃ (below the melting temperature of Cu) increases the thermal conductivity from 110W/(m K) to over 140W/(m K) by enabling more of the Mo to precipitate. Further improvements in the thermal conductivity should be possible with a slower cooling rate or with an isothermal hold at 1150℃ during the cooling cycle. Such an effect is not observed for the W-Cu system, since even at high temperatures, the solubility of W in Cu does not exceed 10-3 wt%. The dissolution of 1 wt% Mo in Cu could easily decrease the thermal conductivity by 30W/(m K), even if it is much less detrimental than Fe. For comparison, based on its effects on Cu and W-Cu, 0.1 wt% Fe is predicted to decrease the thermal conductivity of Mo-15Cu by over 30W/(m K).

20.5.7 Example applications

As with W-Cu, most Mo-Cu is sold as infiltrated sheets, but MIM enables net-shape processing of more complex parts. Infiltrated Mo-15Cu heat sinks processed via powder injection molding are shown in Fig. 20.31. The density of these parts was 95% of theoretical, with practically all closed pores. No problems were encountered with infiltrating the threaded studs of the heat spreader. Thus, complex Mo-Cu heat sinks can be fabricated via injection molding, although an infiltration process has proven more successful than liquid-phase sintering. Less shrinkage for improved dimensional control and a lower cost Cu source are also benefits of infiltration.

Fig. 20.31 Picture of an infiltrated Mo-15Cu heat spreader and a small transistor package

20.6 Conclusions

High-thermal conductivity materials, Cu, W-Cu, and Mo-Cu can be processed by MIM. Copper can be solid-state sintered to near full density, but care is needed to avoid hydrogen swelling. Near full density can be achieved for W-Cu and Mo-Cu by either liquid-phase sintering or infiltration. The commercial availability of composite W-Cu powders with excellent sintering behavior makes liquid-phase sintering preferable for W-Cu. Because of the poor molding characteristics of fine Mo powders and their poor densification during liquid-phase sintering, infiltration techniques are recommended for Mo-Cu. With proper powder selection and good control of sintering cycles and impurities, thermal properties close to model predictions can be achieved. MIM allows fabrication of heat sink geometries that are difficult to produce with other metal-working technologies. Novel structures, such as a heat pipe with a highconductivity casing surrounding a porous wick, can be directly fabricated into complex shapes.