US 20030024611 A1
Disclosed are methods and materials for preparing metal matrix composite (MMC) components that have low weight, good thermal conductivity and a controllable in-plane coefficient of thermal expansion. One embodiment of the invention features a metal matrix composite that includes a metal alloy and random in-plane discontinuous fibers. In some embodiments, the metal alloy includes aluminum, copper or magnesium. In certain embodiments, the metal matrix composite includes additives that enable solution hardening. In other embodiments, the metal matrix composite includes additives that enable precipitation hardening. Another embodiment of the invention features a method of manufacturing a metal matrix composite. The method includes contacting random in-plane discontinuous fibers with a binder, and pressurizing the random in-plane discontinuous fibers and the binder to form a bound preform. The preform is pressurized to a pressure greater than the molten metal capillary breakthrough pressure of the bound preform. Subsequently, the bound preform is placed in a mold, infiltrated with a molten infiltrant, and the molten infiltrant is cooled to form the metal matrix composite.
1. A metal matrix composite comprising:
a metal alloy; and
random in-plane discontinuous fibers, wherein the random in-plane discontinuous fibers comprise carbon.
2. The metal matrix composite of
3. The metal matrix composite of
4. The metal matrix composite of
5. The metal matrix composite of
6. The metal matrix composite of
7. The metal matrix composite of
8. The metal matrix composite of
9. The metal matrix composite of
10. The metal matrix composite of
11. The metal matrix composite of
12. The metal matrix composite of
13. The metal matrix composite of
14. The metal matrix composite of
15. The metal matrix composite of
16. The metal matrix composite of
17. The metal matrix composite of
18. The metal matrix composite of
19. The metal matrix composite of
20. An article of manufacture comprising the metal matrix composite of
21. A metal matrix composite comprising:
a metal alloy; and
random in-plane discontinuous fibers, wherein the random in-plane discontinuous fibers comprise carbon and are uniformly distributed within the metal matrix composite, and
wherein the metal matrix composite has a volume fraction of the random in-plane discontinuous fibers in a range of approximately 0.15 to approximately 0.6.
22. A metal matrix composite comprising:
a metal alloy consisting essentially of aluminum, silicon and magnesium, wherein
the silicon is approximately 5 wt % to approximately 20 wt % of the metal alloy, and
the magnesium is approximately 0.1 wt % to approximately 2 wt % of the metal alloy; and
random in-plane discontinuous graphite fibers uniformly distributed within the metal matrix composite.
23. A metal matrix composite comprising:
a metal alloy consisting essentially of copper, chromium and zirconium, wherein
the chromium is approximately 0.3 wt % to approximately 2 wt % of the metal alloy, and
the zirconium is approximately 0.1 wt % to approximately 1 wt % of the metal alloy; and
random in-plane discontinuous graphite fibers uniformly distributed in the metal matrix composite.
24. A method of manufacturing a metal matrix composite, the method comprising the steps of:
contacting random in-plane discontinuous fibers with a binder;
pressurizing the random in-plane discontinuous fibers and the binder to form a bound preform, wherein the random in-plane discontinuous fibers and the binder are pressurized to a pressure greater than the capillary breakthrough pressure of the bound preform;
placing the bound preform in a mold;
infiltrating the bound preform with a molten infiltrant under a pressure at least equal to the capillary breakthrough pressure; and
cooling the molten infiltrant to form the metal matrix composite.
25. The method of
placing a second bound preform adjacent to the bound preform in the mold prior to the step of infiltrating;
contacting a surface of the bound preform with a surface of the second bound preform; and
removing the binder prior to the step of infiltrating to merge the surface of the bound preform with the surface of the second bound preform.
26. The method of
heating the bound preform in the mold;
evacuating the bound preform in the mold to create a reduced pressure within the bound preform; and
transporting a charge of the molten infiltrant into the mold while maintaining the reduced pressure within the preform.
27. The method of
forming a preform of random in-plane discontinuous fibers,
wherein the step of forming the preform comprises agitating discontinuous fibers to promote a random in-plane orientation.
28. The method of
liquefying the binder; and
solidifying the binder to form the bound preform.
29. The method of
 This invention was made with government support under Grant No. N00167-99-C-0072. The government has certain rights in the invention.
 The present invention relates to methods of forming metal matrix composites for thermal or structural applications, and the resulting compositions. More specifically, the invention relates to methods of infiltration casting to form metal matrix composites with controlled thermal expansion and mechanical properties, and the resulting compositions.
 Technological developments from cellular phones to imaging satellites push current semiconductor device capabilities to their performance limits. In particular, modern devices often must dissipate a great amount of heat. Integrated circuit devices typically require integration with heat sinks due to the potentially deleterious effects of heat generated by the device. A semiconductor die, typically a portion of a silicon wafer, can be directly attached to a heat sink. More commonly, the die is encased in a ceramic package that protects the die and provides electrical connections.
 Common ceramic package materials include aluminum oxide, aluminum nitride and beryllium oxide. The coefficient of thermal expansion of the semiconductor die and the ceramic package are purposely matched to avoid thermal cycle induced mechanical stress failures. Thermal cycling arises during power up and power down cycles in combination with resistive heating due to current flow in the device.
 Heat sinks are commonly fabricated from metals, for example copper, molybdenum, tungsten and aluminum. A metal heat sink is often plated with nickel prior to attachment to a ceramic package at an elevated temperature, for example, via brazing. Alternatively, silver-filled adhesives, or other conductive metal powder-filled adhesives, can be used for bonding.
 Choosing a metal or other material for a heat sink often involves a trade-off between desirable and undesirable properties. For example, aluminum and copper have high thermal conductivity, but coefficients of thermal expansion several times greater than that of a ceramic package or semiconductor die. During power cycling of the integrated circuit, resistive heating causes the temperature of the integrated circuit, and the attached heat sink, to fluctuate. Consequently, such metals apply mechanical stress to the heat sink bonding material during power cycling. The differential expansion of the heat sink relative to the ceramic package or semiconductor die can cause failure of the bond material or cracking of the package or die.
 In contrast, some metals, such as tungsten and molybdenum, have relatively small coefficients of thermal expansion. Although such metals can permit a reliable bond, they have lower thermal conductivity than aluminum or copper substrates and they are difficult to electroplate. Further, tungsten and molybdenum are undesirable for applications that require minimal weight.
 Composites of copper and tungsten, or of copper and molybdenum, can partially mitigate these deficiencies. These composites can be made by powder metallurgical methods, such as infiltrating copper into a sintered body of tungsten or molybdenum, or sintering a mixed powder of the two metals. It is difficult, however, to obtain an elongated plate by rolling a sintered ingot of tungsten or molybdenum. Alternatively, layers of metal can be joined by cladding or lamination of sheets. Cladded and laminated products require precise machining, which is difficult and increases costs.
 As an alternative to an all metal heat sink, some heat sinks combine a sintered ceramic with a metal matrix. The fabrication process involves the formation of a ceramic preform, for example, by sintering silicon carbide powder. The ceramic preform microstructure typically has a predetermined void volume fraction that is subsequently filled with molten metal, typically aluminum. An aluminum ceramic heat sink can employ copper-based inserts to improve its thermal conductivity. Such heat sinks, however, can be difficult to machine and are usually limited in their ability to match coefficients of thermal expansion with integrated circuits.
 As another alternative, a metal matrix composite can include an inorganic fiber material. Infiltration of fibers has its own difficulties, for example, problems with fiber wetting and non-uniform fiber distribution. In addition, molten metal infiltration of fibers under pressure can displace the fibers due to the fiber breakthrough pressure threshold. Further, it is often difficult to control fiber volume fraction, and thus to obtain a desired property of the composite. These factors have limited use of metal matrix fiber composites as heat sinks.
 As the semiconductor industry continues to implement ever increasing semiconductor die sizes and transistor densities to permit enhanced integrated circuit complexity, the heat generated by state-of-the-art integrated circuits also increases. Thus, the challenge of coping with resistive heating is expected to become an ever more central concern in integrated circuit design.
 Beyond the electronics industry, precision motion and control components and other mechanical hardware must be light weight, stiff, and damp unwanted vibrations. In many instances, conventional materials, such as aluminum and copper, are unable to meet the performance demands of many emerging technologies.
 It has been discovered that a metal matrix composite (“MMC”) that includes random in-plane discontinuous carbon fibers and a method of forming an MMC from a pressure-formed preform can solve many problems of prior art heat sinks. The invention can overcome numerous problems, such as: fiber collapse during molten metal infiltration; coefficient of thermal expansion (“CTE”) mismatch; heat sink weight; limits on range of CTE values; difficulty in obtaining high fiber density; limits in control of fiber orientation; machinability of a heat sink; and/or heat sink cost. The invention addresses these problems through the use of one or all of the following: preforms prepared with pressures greater than the breakthrough pressure used during metal infiltration; in-plane oriented fibers; short fibers; and carbon fibers.
 Use of random in-plane discontinuous fibers permits a high fiber volume fraction in the MMC (“in-plane” as used herein is understood as the X-Y plane, for example, the plane parallel to the bonded surface of a heat sink). Further, by using in-plane oriented fibers, substantially all of the fibers can contribute to the control of the CTE in the X-Y plane. Though Z-direction CTE is not controlled by in-plane fibers, such control is generally unnecessary for heat sink applications because the integrated circuit or other object is attached to an X-Y oriented surface of the heat sink.
 Use of these in-plane oriented fibers permits selection of a CTE over a wide range of values. A desired volume fraction of in-plane oriented fibers is selected to obtain a desired CTE. By orienting substantially all fibers in the X-Y plane, a very high fiber volume fraction can be obtained. This permits selection of volume fraction over a wide range and a corresponding ability to select a wide range of CTE values.
 Carbon fibers, in particular graphite fibers, have excellent mechanical and thermal properties for use in heat sinks of the invention. In combination with an aluminum or other light metal, such heat sinks typically are easily machined, have excellent heat conductivity, and are lightweight. An aluminum and graphite fiber MMC of the invention thus realizes the advantages of aluminum—lightweight, easy machinability and good heat conduction—in combination with the advantages of graphite fibers—high Young's Modulus, small to negative CTE, high tensile strength, high thermal conductivity and strong damping properties.
 The invention also solves the problem of non-uniform fiber distribution within the MMC. A preform that includes fibers and a binder can be prepared via application of a pressure in a preform mold that is greater than the molten alloy breakthrough pressure for the preform. The binder maintains the compressed configuration of the fibers in the preform while the preform is removed from the preform mold and placed in a metal infiltration mold. The metal infiltration mold can maintain the compressed fiber configuration upon removal of the binder, if the infiltration mold is sized and shaped to conform to the preform. Because the fiber configuration remains in its compressed state, it is substantially undisturbed during infiltration of molten metal at the molten metal breakthrough pressure.
 In a broad aspect, the invention features a metal matrix composite that includes a metal alloy and random in-plane discontinuous fibers. The random in-plane discontinuous fibers may be carbon, and preferably are graphite. The fibers typically are milled, and preferably are ball milled. In preferred embodiments, the metal alloy includes aluminum, copper or magnesium.
 In one embodiment, the metal matrix composite has a volume fraction of random in-plane discontinuous fibers in a range of approximately 0.15 to approximately 0.6. In another embodiment, a minority of the random in-plane discontinuous fibers are oriented out of plane by an angle greater than 10°. In a preferred embodiment, the random in-plane discontinuous fibers are uniformly distributed within the metal matrix composite.
 In certain embodiments, the metal matrix composite may include a component that enables solution hardening. In other embodiments, the metal matrix composite may include a component that enables precipitation hardening. In one preferred embodiment, the metal matrix composite includes aluminum, silicon and magnesium. In another preferred embodiment, the metal matrix composite includes copper, chromium and zirconium.
 In another aspect, the invention provides a method of manufacturing a metal matrix composite. The method includes contacting random in-plane discontinuous fibers with a binder, and pressurizing the random in-plane discontinuous fibers and the binder to form a bound preform. In the later step, the random in-plane discontinuous fibers and the binder are pressurized to a pressure greater than the capillary breakthrough pressure of the bound preform. Subsequently, the bound preform typically is placed in a mold, heated under a vacuum to remove the binder, then heated to above the metal liquidus and infiltrated with a molten infiltrant. The molten infiltrant is then cooled to form the metal matrix composite.
 In another embodiment, the method includes placement of a second bound preform adjacent to the bound preform in the mold prior to infiltration with the molten infiltrant. A surface of the bound preform contacts a surface of the second bound preform, and removal of the binder prior to infiltration causes the contacted surfaces of the two preforms to merge, creating one continuous, metal matrix composite.
 The binder may be removed (called debindering) prior to infiltration with a molten infiltrant, e.g., via evaporation. Alternatively, the binder may partially or completely remain in the preform during infiltration. For example, a volatile component of a binder may be removed prior to infiltration, leaving a residue in the preform.
 Reference to the figures herein is intended to provide a better understanding of the methods and apparatus of the invention but are not intended to limit the scope of the invention to the specifically depicted embodiments. The drawings are not necessarily to scale, emphasis instead being placed upon illustrating the principles of the invention. Like reference characters in the respective figures typically indicate corresponding parts.
 It should be understood that the order of the steps of the methods of the invention is immaterial so long as the invention remains operable, i.e., e.g., a preform is made prior to infiltration of the preform. Moreover, two or more steps may be conducted simultaneously.
 The foregoing, and other features and advantages of the invention, as well as the invention itself, will be more fully understood from the description, drawings, and claims which follow.
FIG. 1 is a graph that shows the variations of CTE with in-plane volume fraction for a theoretical model and for experimental embodiments of aluminum and copper matrix composites of the invention.
FIGS. 2a and 2 b are scanning electron micrographs of an experimental embodiment of a metal matrix composite having an aluminum matrix and graphite fibers. FIG. 2a shows a cross-section through the X-Y plane. FIG. 2b shows a cross-section through the X-Z plane.
FIGS. 3a-3 e illustrate formation of a preform according to an embodiment of the invention.
FIG. 3a is a cross-sectional illustration of dispensing fibers and binder into a preform mold base portion.
FIG. 3b shows the fibers and binder residing in the preform mold base portion.
FIG. 3c shows the fibers and binder under compression in the preform mold.
FIG. 3d shows binding of the fibers by the binder.
FIG. 3e shows a completed, bound preform after removal from the preform mold.
FIG. 4 illustrates an embodiment of a stacked preform having three individual preforms layered with two graphite foils between the preforms.
FIGS. 5a and 5 b illustrate an embodiment of forming a larger preform from a combination of smaller preforms. FIG. 3a shows a stack of preforms without any intermediate layers. FIG. 3b shows the larger preform after merging of the interfaces of the smaller preforms.
FIG. 6 illustrates an embodiment of the stacked preform of FIG. 4 in a metal infiltration mold.
FIG. 7 illustrates an embodiment of a horizontally oriented preform in a metal infiltration mold.
 A metal matrix composite (“MMC”) that includes random in-plane discontinuous carbon fibers and a method of forming an MMC from a pressure-formed preform can solve many problems of prior art heat sinks. The composite and method can alleviate such problems as: fiber collapse during molten metal infiltration; coefficient of thermal expansion (“CTE”) mismatch; heat sink weight; limits on range of CTE values; difficulty in obtaining high fiber density; limits in control of fiber orientation; machinability of a heat sink; and/or heat sink cost. The following describes various embodiments of the invention that may include, for example, preforms prepared with pressures greater than the breakthrough pressure used during metal infiltration, random in-plane oriented fibers, short or discontinuous fibers, and carbon fibers.
 As used herein, “molten metal infiltration” is understood to mean any casting process with or without an externally applied pressure to facilitate infiltration of a mold vessel cavity that contains a preform. Examples of pressure infiltration casting include, but are not limited to, pressure infiltration casting such as the Advanced Pressure Infiltration Casting (APIC™) process as described in U.S. Pat. Nos. 5,322,109; 5,553,658; and 5,983,973; high throughput pressure infiltration casting as described in U.S. Pat. No. 6,148,899; squeeze casting; and die-casting.
 As used herein, “metal” is understood to mean a metal or metal alloy. Examples of common metals or metal alloys are, among others, aluminum, aluminum alloys, bronze, beryllium, beryllium alloys, chromium, chromium alloys, cobalt, cobalt alloys, copper, copper alloys, gold, iron, iron alloys, steel, magnesium, magnesium alloys, nickel, nickel alloys, lead, lead alloys, copper, tin, tin alloys such as tin-bismuth and tin-lead, zinc, zinc alloys, superalloys such as International Nickel 100 (IN-100) or International Nickel 718 (IN-718), and combinations thereof.
 As used herein, “molten infiltrant”, “liquid infiltrant,” “molten metal,” or “liquid metal” is understood to mean a respective material which is at least at or above approximately its liquidus temperature.
 As used herein, “fugitive” is understood to mean substantially removable, i.e., removable to a great extent.
 As used herein, “preform” is understood to mean a fibrous, non-metallic material such as, e.g., an oxide, a boride, a nitride, a carbide or a form of carbon which is to be infiltrated with an infiltrant. Infiltration of a preform by a molten metal followed by solidification produces a metal matrix composite (MMC).
 As used herein, “bound preform” is understood to mean a preform in which the fibers are held in a more or less fixed physical relationship due to the action of a binder material.
 As used herein, “preform mold vessel” and “preform mold” are understood to mean any container capable of holding or applying pressure to preform materials during formation of a preform.
 As used herein, “metal infiltration mold vessel” and “metal infiltration mold” are understood to mean any container capable of holding a preform, and confining the preform and molten metal during metal infiltration of the preform.
 As used herein, “in-plane” is understood to mean the X-Y plane or the plane normal to the Z direction in an X-Y-Z coordinate system. It is also understood to mean the plane that is parallel to the bonded surface of a heat sink. This is commonly referred to as the “base” plane in the electronics industry.
 Overview of Materials in an MMC Component
 Some embodiments of MMC components of the invention are well suited as heat sinks for use with a variety of integrated circuit semiconductor and ceramic packaging materials. These components have relatively low density, high thermal conductivity and a coefficient of thermal expansion (“CTE”) that can be controlled over a wide range to match a companion integrated circuit material. Properties of common semiconductor and packaging materials are illustrated in Table I. Table I also shows the preferred heat sink CTE ranges for a good match with each of the listed materials.
 In one embodiment, an MMC includes random in-plane discontinuous fibers. Use of discontinuous fibers, particularly fibers less than approximately 1 mm in length, permits good control of the volume fraction of the fibers in the finished MMC. Further, for in-plane oriented fibers, substantially all of the fibers contribute to control of CTE in the X-Y plane.
 An MMC of the invention may include fibers of a narrow or wide range of fiber lengths. In some embodiments, an MMC includes fibers as short as approximately 30 μm. In some embodiments, an MMC includes chopped fibers as long as approximately 12 mm.
 Some applications do not require control of Z-direction CTE, for example, heat sinks. For these applications, MMC components can be fabricated with a desired X-Y plane CTE over a very wide range by selection of a corresponding fiber volume fraction. Hence, the invention provides heat sinks that can be CTE-matched to a wide variety of integrated circuit materials. Fibers of various materials may be used. For example, the fibers may include silicon carbide. In preferred embodiments, MMC components include carbon fibers, preferably graphite fibers, in a light metal matrix. Carbon fibers may, e.g., be prepared from a pitch or pan precursor. Preferred embodiments employ pitch precursor fibers due to a superior elastic modulus and thermal conductivity relative to pan precursor fibers.
 Graphite fibers have excellent mechanical and thermal properties for use in heat sinks of the invention. In combination with a light metal, e.g., aluminum, magnesium, or copper, or their alloys, such heat sinks are easily machined, have excellent heat conductivity and are very lightweight. An aluminum and graphite fiber MMC of the invention thus realizes the advantages of aluminum—lightweight, easy machinability and good heat conduction—in combination with the advantages of graphite fibers—high Young's Modulus, small to negative CTE, high tensile strength, high thermal conductivity and strong damping properties.
 In addition to high thermal conductivity, graphite fibers have the unusual feature of a high modulus of elasticity combined with a negative coefficient of thermal expansion. Thus, these high modulus, negative CTE fibers can be embedded within a matrix metal alloy to restrain the matrix from expanding to its full extent during heating. The fibers further prevent excessive contraction during cooling from a processing temperature, and contribute to in-plane thermal conductivity. By combining graphite fibers with a metal such as Al and by controlling the volume fraction of fibers and the orientation, one can design a material with a specified CTE.
 In some embodiments, a heat sink of the invention can be soldered or brazed to an integrated circuit package to improve heat transfer. The heat sink has a CTE that is preferably chosen to be slightly greater in value than the package so that the package is under compression at room or operating temperature. Package cracking or failure of the heat sink/package bond is less likely under this condition.
 The CTE of the heat sink is also preferably chosen in view of the temperature range to be used during attachment of the heat sink to a ceramic package. For example, a eutectic copper-silicon braze alloy requires a temperature of 780° C. during attachment of a nickel plated copper alloy matrix heat sink to a metallized ceramic package. A gold-germanium braze alloy requires a temperature of 380° C. during attachment of an aluminum matrix heat sink to a ceramic package.
 The discontinuous fibers can be inorganic. Preferably, the discontinuous fibers are carbon-based. The fibers are more preferably graphite. The fibers can be chopped. In some embodiments the fibers are less than 25 mm in length. In some embodiments, the fibers are preferably less than 1 mm in length, more preferably less than 0.75 mm. In some embodiments, the fibers are milled. In preferred embodiments, the fibers are ball milled. In a preferred embodiment, the fibers graphite and have an average length of approximately 0.2 mm (200 μm) and a diameter of approximately 10 μm.
 Fiber Orientation and Fiber Volume Fraction
 Theoretical considerations can assist in the selection of a fiber volume fraction that will provide a desired in-plane CTE. For example, a uniaxial laminate-based theory (see, e.g., R. S. Schapery, “Thermal Expansion Coefficients of Composite Materials Based on Energy Principles,” J. Composite Materials, Vol. 2, pages 380-404, (1968)) approximates the CTE of a composite material as:
 α11 is the CTE of the composite in the orthogonal direction (parallel to the 0° fibers in an uniaxial laminate, units of ppm/° K),
 αf is the axial CTE of the fiber (units of ppm/° K),
 αm is the CTE of the matrix (e.g., 24 ppm/° K for Al and 16 ppm/° K for Cu)
v f is the volume fraction of fibers oriented parallel to the 0° axis,
 (1-vf) is the matrix volume fraction,
 Ef is the Young's modulus of elasticity for graphite fiber (units of GPa), and
 Em is the Young's modulus of elasticity for the Matrix (e.g., 69 GPa for Al and 110 GPa for Cu).
 One can model orthogonal in-plane CTE properties, i.e., as obtained with two sets of orthogonally oriented in-plane fibers, by adding a second laminate of fibers oriented at 90° in the above model. Such a model should be considered a lower bound approximation for orthogonally oriented fiber reinforced metals.
 One can approach the random in-plane condition by extension of the theoretical model to a multi-laminate composite. This will give an approximate CTE for randomly oriented in-plane fibers where the X and Y components of in-plane fibers are equal.
 For most electronic thermal management applications when a matching CTE is important, it is desirable to maintain orthogonal CTE values, i.e., the CTE in the 0° or X direction should equal the CTE in the 90° or Y direction. When the in-plane CTE is, for example, 8 ppm/° K in all in-plane directions, one can simply say that the in-plane CTE is 8 ppm/° K.
 Balanced 0°-90° composites can be achieved by use of orthogonally oriented in-plane fibers. Fibers in an MMC can be orthogonally oriented by weaving or by stacking alternating plies of uniaxially wrapped fibers to form a laminate of orthogonally oriented layers.
 In preferred embodiments of an MMC, substantially uniform in-plane orthogonal properties are obtained by use of in-plane randomly oriented discontinuous fibers. Preferably, very few of the fibers are oriented out of plane, i.e., in the Z axis direction. By controlling the volume fraction of a preform, whether it includes woven, wrapped or random-in-plane discontinuous fibers, one can obtain a graphite reinforced MMC having a selected value of CTE.
 Theoretical curves are plotted in FIG. 1 for an MMC that includes an aluminum matrix or a copper matrix, with fiber properties corresponding to those of P-120 graphite fibers, a fiber in the ThermalGraph® family of products available from BP Amoco (Alpharetta, Ga.). These fibers have an average length and width of approximately 200 μm and 10 μm, respectively, Ef=827 GPa, αf=1.45 ppm/° K, and density=2.17 g/cc. Other useful fibers in the ThermalGraph® family include those sold under the developmental names DKA X and DKD X. Similar fibers are available from other suppliers, such as Conoco Carbon Fibers (Houston, Tex.).
 Theoretical curves can be used to assist control of the CTE of a reinforced metal alloy by selecting an appropriate fiber volume fraction. For example, referring to FIG. 1, a CTE in a range of 4 to 12 ppm/° K would require a corresponding fiber volume fraction between approximately 0.40 and 0.18. Similarly, a P-120 reinforced copper alloy MMC would require selection of fiber volume fraction between 0.4 and 0.14 to obtain CTE values in the same range.
 More generally, MMC components can be prepared with a wide range of CTE values. For example, a CTE can be zero or negative in value, or can be 12 ppm/° K or greater in value. The ability to densely pack fibers permits fiber volume fractions to be chosen, preferably, from within a range of approximately 0.15 to 0.60. For some applications, fiber volume fraction may lie outside this range.
 For aluminum-based MMCs, a wide range of useful properties may be obtained with a fiber volume fraction in a range at least as broad as approximately 0.15 to approximately 0.55. For example, a fiber volume fraction of 0.30 provides a CTE of approximately 8.0 ppm/° K, while a fiber volume fraction of 0.40 provides a CTE of 4.0 ppm/° K.
 Deviation of a portion of the fibers in an MMC from in-plane orientation reduces the observed CTE from that predicted by theory. Deviations from in-plane randomness in experimental samples cause the in-plane CTE of a reinforced metal alloy to be greater than that predicted theoretically. Accordingly, empirical calibration curves can be constructed that are based on experimental data. Consequently, the CTE versus volume fraction curves are more accurate for manufacturing purposes than curves obtainable from theoretical relationships such as given in Equation 1.
 The graph in FIG. 1 shows two CTE volume fraction curves for aluminum matrix and copper matrix MMC samples prepared with P-120 graphite fiber preforms. The aluminum MMC curve was obtained from CTE measurements obtained during cooling of samples from 400° C. to 50° C. (800° C. to 50° C. for the copper matrix samples). As shown, P-120 reinforced aluminum samples were prepared having fiber volume fractions in a range of 0.2 to 0.4. The CTE of the resulting samples varied between 13 and 4 ppm/° K.
 The degree of deviation from theory is at least in part due to a population of fibers with some out of plane orientation. To the extent that a fiber is oriented out of plane, the elastic restraint of the fiber on the in-plane CTE of the matrix is reduced.
 The microstructure of a sample aluminum alloy MMC with 0.30 volume fraction of P-120 fibers is shown in the scanning electron micrographs of FIGS. 2a and 2 b. The sample was prepared by blending 743 grams of DKDX fibers and 119 grams of Carbowax® polyethelyne glycol 8000 from Union Carbide Chemical and Plastics Co. (Danbury, Conn.) as a binder.
 The blend was loaded into a 190.5 mm×190.5 mm press die mold and pressed to a thickness of 31.7 mm. Subsequent to heating to a temperature in a range of 80°-100° C. to liquefy the binder, cooling to 10° C. solidified the binder to produce a stable, bound preform. The bound preform was infiltrated with aluminum alloy containing 12.5% silicon and 0.4% magnesium. To obtain these micrographs, polished sections were taken from the sample MMC along the X-Y plane and along the X-Z plane.
 The X-Y section of FIG. 2a shows a substantial number of fibers laying parallel to the X-Y plane. In the X-Z section of FIG. 2b, only a few fibers lie with any component of their orientation parallel to the Z direction. Quantitative analysis of several micrographs showed that over 84% of the fibers laid within 10° of the X-Y plane, and that over 96% of the fibers laid within 300 of the X-Y plane.
 In one embodiment of an MMC of the invention, less than half of the fibers are oriented out of the X-Y plane by more than 10°. In a more preferred embodiment, less than 25% of the fibers are oriented out of plane by more than 10°. In a more preferred embodiment, less than 20% of the fibers are oriented out of plane by more than 10°. In a further preferred embodiment, less than 15% of the fibers are oriented out of plane by more than 10°.
 Preparation of a Preform
 Properties of an MMC are affected by both the volume fraction and the orientation of the fibers. For liquid metal infiltrated composites, preferred embodiments use a fibrous preform that does not “swim” or become disturbed during the inrush of molten metal. These embodiments include a preform that is stable and that does not lose its shape or fiber distribution during the infiltration process.
 In some embodiments, a stable preform is obtained by densely packing a preform mold. After removing the packed and bound preform from the preform mold, the preform is typically placed into a metal infiltration mold, in preferred embodiments, a steel can. Preferably, the preform completely fills the metal infiltration mold or the mold cavity of the metal infiltration mold.
 When the preform is heated, the binder may begin to release the fibers. The fibers can then relax and press against the metal infiltration mold. Some binders evaporate during heating. After the binder is removed, e.g., with the assistance of an applied vacuum, and the preform has reached a molten metal infiltration temperature, molten metal infiltration can take place.
 In one embodiment, woven fibers in the form of a fabric cloth are cut and loaded directly into a metal infiltration mold for subsequent pressure infiltration. Since a fabric has a discrete thickness, controlling the thickness of an MMC component formed from fabric is difficult. For volume fractions above the natural woven volume fraction of a fabric, the fabric is compressed and clamped into a mold, increasing tooling costs. Moreover, it is typically difficult to pack woven fabrics to a fiber volume fraction greater than approximately 0.45. Conversely, loading molds with fabric to a fiber volume fraction less than 0.40 can lead to non-uniform distribution of the fiber plies. Even when preform cloth plies have been well compressed into a mold, non-uniform ply loading can result in warping after removal from the mold.
 An alternative embodiment uses a continuous fiber preform. Such a preform may be fabricated by drum winding continuous fibers, and fixing the wound fibers onto a transfer sheet by applying a fugitive binder. Such plies can be stacked with orthogonal orientations, or more mixed orientations, for example, including plies oriented at 45° to other plies. These embodiments can include fiber volume fractions of approximately 0.55 to 0.6 or more.
 Another alternative preform material includes a paper-like product produced from chopped discontinuous fibers. The material includes a fugitive binder to provide stability and facilitate handling. In a preferred embodiment, the fibers in the “paper” are randomly orientated in the X-Y plane. This material can be compressed to a desired fiber volume fraction and further stabilized with additional binder.
 In another embodiment, a preform of the invention provides substantially uniform fiber distribution after molten metal infiltration. The preform is typically prepared by compressing fibers and a binder in a preform mold. In a preferred embodiment, the fibers are discontinuous. The fibers and binder usually are mixed, then compressed at a pressure that is greater than the molten alloy breakthrough pressure for the finished preform. Prior to metal infiltration, the binder maintains the compressed configuration of the fibers in the preform so the preform can be removed from the preform mold and placed in a metal infiltration mold. In certain applications, a bound preform may be stored for some time prior to metal infiltration. Preferably, the bound preforms are stored below room temperature.
 The binder is often removed from the preform while the preform resides in the metal infiltration mold. Under the constraints of the metal infiltration mold, the preform can maintain its compressed fiber configuration. Molten metal is then infiltrated into the preform. Because the fiber configuration remains in its compressed state, it is substantially undisturbed during infiltration of molten metal at the molten metal breakthrough pressure, and proper in-plane orientation of the fibers is maintained.
 In some embodiments, the in-plane distribution of fibers in the preform mold is enhanced prior to compression and fixation with a binder. For example, the preform mold can be agitated, such as by vibration, prior to compression. Vibration may also break up clumps, or “hair balls”, of fiber. Further, the compression of the fibers and the binder also serves to enhance the in-plane orientation of the fibers, as well as increase the volume fraction of fibers in the preform.
 Some embodiments utilize thin sheets of paper-like or felt-like material formed from random in-plane oriented fibers. In a preferred embodiment, a sheet is produced with a volume fraction in a range of approximately 0.05 to approximately 0.20. The sheets may be weighed to select a proper amount for a desired preform. The sheets may then be placed in a preform mold and compressed to obtain a final desired fiber volume fraction in the preform.
 Various binder materials can be employed to form and maintain a preform. A binder material is generally required to maintain fibers in a desired orientation and state of compression. For example, water may be a binder which is set via freezing. Solid forms of polyethylene glycol (“PEG”) may be a binder as well as acrylic. Solid binder materials usually are heated to approximately or above their melting point, then cooled to solidify. One suitable PEG material is Carbowax® polyethylene glycol 8000 from Union Carbide Chemical and Plastics Co. (Danbury, Conn.), which is liquefied at approximately 80°-100° C.
 Some binders are fully removed prior to molten metal infiltration. Other binder materials partially or fully remain during and after molten metal infiltration. For example, a phenolic-based binder may have a volatile component removed prior to molten metal infiltration. For example, the volatile component leaves in vapor form, leaving behind a carbon-based residue. In some embodiments, non-binder organic materials may also escape from the preform during evaporation of some or all of a binder material.
FIGS. 3a through 3 e illustrate in cross-section one embodiment of the formation of a preform 10. Referring to FIG. 3a, a material dispenser 22 randomly dispenses discontinuous fibers 11 and a binder 12 into a preform mold base portion 20. The binder 12 can be dispensed as discontinuous particles as shown, or can be mixed with the fibers by other means. For example, the fibers can be coated with binder subsequently or prior to distribution of the fibers into the preform mold. One or more dispensers may be employed. In other embodiments, fibers and binder are dispensed from different dispensers.
 Referring to FIG. 3b, to promote a random, in-plane distribution of the fibers, the preform mold base portion 20 often is agitated. For example, the base portion 20 can be vibrated after filling with fibers 11 and binder 12. Preferably, agitation is applied continuously during the dispensing of the fibers and binder into the preform mold.
 Referring to FIG. 3c, a preform mold cap portion 21 is placed in contact with the fibers 11 and binder 12. Pressure is applied via the preform mold cap portion 21 to the mixture of fibers 11 and binder 12 to compress the mixture at a pressure greater than the breakthrough pressure required to infiltrate the formed preform with the appropriate molten metal. The binder 12 serves to fix the configuration of fibers obtained in this compressed state by adhering neighboring fibers 11 to one another to produce the preform 10 as shown in FIG. 3d. FIG. 3e shows the preform 10 after it is removed from the preform mold base portion 20.
 In a preferred embodiment, in a tumble mill to uniformly mix the fibers and binder. The mixture then may be redispersed in a rotary brush mill to untangle the fiber clusters. A measured portion of the mixture is selected by weighing for production of an MMC of a desired size and fiber volume fraction.
 The weighed portion of fiber and PEG binder is placed in a preform mold base portion. After leveling the mixture by vibration, the preform mold is closed and the preform is compressed to a predetermined volume to obtain the desired volume fraction of fibers in the complete preform.
 When forming a preform from a mixture of PEG powder and fibers having an average length of approximately 200 μm, pressure of approximately 450 psi is required to obtain a fiber volume fraction of 30%, while a pressure of approximately 1,100 psi is required to obtain a volume fraction of 40%. These pressures exceed typical liquid metal capillary breakthrough pressures of the resulting preform.
 Heating to approximately 85° C. melts the PEG. Subsequent cooling allows the PEG to solidify and bind the fibers to one another. The preform can then be stored under refrigeration for later use.
 Metal Matrix Materials
 Various metals and metal alloys can be used in the invention depending on the particular application. Aluminum, copper and magnesium are preferred. With applications involving aluminum or an aluminum alloy, it is desirable to add silicon to the metal to reduce the reactivity of the metal with graphite, which undesirably forms aluminum carbide. For example, an MMC formed from 6061 aluminum alloy with 0.45 volume fraction of graphite fibers and had approximately 4.0% carbide formation after pressure infiltration casting. Fabrication of an MMC from an aluminum alloy having approximately 7.0% by weight silicon, reduced carbide formation to approximately 0.5%. Using 12.5% silicon in the aluminum alloy, further reduced the carbide formation to approximately 0.3%.
 Thus, in embodiments which include aluminum, silicon is preferably added. In preferred embodiments, the alloy includes at least approximately 7% by weight silicon, and more preferably approximately 12.5% by weight silicon. The eutectic composition for an aluminum-silicon alloy has 12.5% silicon. In addition to reducing the activity of carbon in aluminum, addition of silicon reduces the melting point of the alloy. This, in turn, further decreases the kinetics of carbide formation, provided that metal infiltration temperatures also are reduced.
 The matrix alloy should also be able to withstand micro-scale deformation that can occur during thermal cycling. In some applications, such as integrated circuit heat sinks, an MMC experiences large temperature cycles during use. Micro-deformation during thermal cycling can cause thermal ratcheting, i.e., a change in dimension of the MMC after each thermal cycle. An accumulation of dimensional changes can lead to damage, for example, of an electronic assembly attached to an MMC heat sink. It is thus desirable to limit deformation over the use temperature ranges of an MMC, such as from approximately −30° C. to approximately 150° C.
 To reduce thermal cycling deformation, the metal matrix alloy can be hardened. In certain embodiments, magnesium is added to aluminum. In combination with silicon, the magnesium provides an age hardenable aluminum alloy due to formation of precipitates during cooling from, for example, a brazing temperature. The precipitation hardened matrix suffers from less plastic deformation during thermal cycling.
 One embodiment of an aluminum-based MMC of the invention includes 2.0% by weight or more magnesium in the alloy. A preferred embodiment includes magnesium in a range of approximately 0.1% to approximately 1.0%. More preferred embodiments include magnesium in a range of approximately 0.2% to 0.5%, or approximately 0.3% to 0.4%. It should be understood that the aluminum alloy can contain minor impurities, for example, iron, manganese and titanium.
 Other embodiments of the invention use a copper alloy and graphite fibers. Because the bond strength between copper and graphite is extremely low, chromium is often included in the alloy. The chromium reacts with carbon in the fibers to form chromium carbide at the fiber-metal interface, which aids the bonding between the alloy and the fibers.
 One embodiment includes approximately 5.0% by weight or more chromium in the alloy. A preferred embodiment includes chromium in a range of approximately 0.3% to 5.0%. A more preferred embodiment includes chromium in a range of approximately 0.3% to 2.0%. Other more preferred embodiments include chromium in a range of approximately 0.5% to 1.5%, or approximately 0.7% to 1.0%.
 Another embodiment includes a copper alloy having improved yield strength through addition of zirconium. Zirconium promotes solid solution hardening and reduces thermal ratcheting. A preferred embodiment includes zirconium in a range of approximately 0.1% to 2.0% by weight. A more preferred embodiment includes zirconium in a range of approximately 0.1% to 1.0%. Other more preferred embodiments include zirconium in a range of approximately 0.1% to 0.5%, or approximately 0.12% to 0.3%. Alloy additions, such as those described above, have a minimal effect on the thermal conductivity of an alloy.
 Using methods known in the prior art, it is typically difficult to control MMC volume fraction by stacking woven cloth preforms into a mold. Since graphite fibers are not easily wetted by some alloys, such as aluminum, magnesium and copper, pressure infiltration is used to overcome the wettability difficulty. However, a pressure infiltrated metal exerts a compressive pressure on a preform prior to capillary breakthrough and subsequent infiltration. In some cases, the infiltrating metal compresses the preform and causes veining and gross displacement of the preform to the mold wall upon breakthrough into less dense regions. Obtaining a particular desired volume fraction of fibers is also difficult when using stacked wrapped lamina to make a preform.
 As described above, methods of the invention overcome these difficulties by providing a preform which has been compressed at a pressure above that experienced during infiltration casting. For example, by pressing preforms to a pressure greater than the capillary breakthrough pressure, activating a binder to constrain the preform, loading the fiber and binder into a fixed volume mold, removing the binder by evacuation and heating just prior to pressure infiltration casting, a metal matrix composite can be manufactured free of breakthrough defects and at a controlled volume fraction reinforcement.
 Preforms of the invention can be infiltrated individually or collectively. FIG. 4 illustrates an embodiment of multiple preforms prepared for infiltration. The preforms 10 are layered with separator sheets 15, for example, graphite foil sheets, to permit production of more than one MMC component during one infiltration cycle. The separator sheets may also be, e.g., slices of graphite, graphite coated steel sheets, or colloidal graphite coated sheets. The separator sheets ease separation of the MMC components after cooling of the metal.
 In another embodiment, as illustrated in FIG. 5a, one or more preforms 10 are placed adjacent to each other without separator sheets. Hence, surfaces of the preforms 10 are in direct contact. After packing the layered preforms 10 into a metal infiltration mold vessel, the binder is removed, for example, by heating. As illustrated in FIG. 5b, upon release of the binder, the contacted surfaces of the preforms 10 can merge with one another and permit formation of an effectively larger preform and ultimately larger MMC component.
 A preform or stack of preforms can be infiltrated with a molten metal by any method known to one skilled in the art. In one embodiment, as illustrated in FIG. 6, a stack of preforms 10 layered with separator sheets 15 is placed in a metal infiltration vessel 30. A filter 33 is placed on top of the stack to prevent premature infiltration of the preform, especially if the preform is evacuated prior to introduction of the metal. Note, however, that alternative arrangements of preforms in mold vessels are possible which may not required a filter, for example, use of gated top plates or caps.
 In another embodiment, illustrated in FIG. 7, a preform 10 is horizontally positioned in a metal infiltration vessel 30. A cap 33 with gates 39, for admission of molten metal, is placed on the preform 10. The cap may be held in place by means known in the art which includes welding. In embodiments where a high volume fraction of fibers is desired, the preform(s) typically need to be isolated in a confined space so that upon removal of the binder, the fibers maintain their position, orientation and compactness.
 One can employ a mold release agent in the vessel. For aluminum alloy and magnesium alloy, the mold release agent preferably is one or more layers of colloidal carbon, e.g., colloidal graphite or boron nitride, which is dispersed in a suitable volatile vehicle. However, other ceramic slurry coatings may be used. For copper alloy, a slurry of zirconium oxide in a slightly acidic vehicle sold under the trade name Zircwash™ may be used. Other parting compounds may be used as mold release agents or washes such as boron nitride or graphite foil.
 In one embodiment, a preform is tightly loaded into a molten metal infiltration vessel. The preform is heated to remove binder via evaporation. Upon removal of the binder, compressive stresses stored in the preform cause the preform to relax against the walls of the vessel. Since the preform is constrained by the walls of the vessel, the vessel walls now maintain a compressive stress on the preform that is greater than the breakthrough pressure of the molten metal.
 In a preferred embodiment, the process described in U.S. Pat. No. 6,148,899 is used to infiltrate molten metal into a preform. Briefly, liquid metal is transferred by vacuum siphon into the metal infiltration mold vessel which is under reduced pressure. The mold vessel is placed in an autoclave and pressurized to approximately 60 atm using nitrogen gas, forcing the molten metal into the preform.
 The mold vessel then is contacted with tin-bismuth at the eutectic composition. Heat from the vessel causes the tin-bismuth to melt. This heat transfer process increases the solidification rate of the molten metal in the preform and assists directional solidification to help eliminate shrinkage porosity.
 After cooling, the MMC component is removed from the vessel. The MMC component can than receive other processing, for example, machining into a final desired shape for use as a heat sink, or plating in preparation for some types of brazing.
 The invention may be embodied in other specific forms without departing from the spirit or essential characteristics thereof. The foregoing embodiments are therefore to be considered in all respects illustrative rather than limiting on the invention described herein. Scope of the invention is thus indicated by the appended claims rather than by the foregoing description, and all changes which come within the meaning and range of equivalency of the claims are intended to be embraced therein.
 Each of the patent documents and scientific publications disclosed hereinabove is incorporated by reference herein.