Fiber-Reinforced Silicon Nitride Composite Ceramics

1. Field of the Invention

This invention relates to the field of ceramic composites which comprise a continuous phase, also interchangeably called matrix, and a discontinuous phase, also interchangeably called reinforcement. The discontinuous phase is at least predominantly in the form of elongated fibers. Such materials are generally denoted in the art as fiber reinforced composites. This invention relates more particularly to composites with a matrix comprised predominantly of silicon nitride and reinforcing fibers predominantly of silicon carbide.

2. Technical Background

Like almost all other ceramics, silicon nitride inherently has little ductility, extensibility, or other capacity for stress relief, so that when subjected for even a short time to mechanical stresses in excess of its capacity, it normally breaks. Practical uses for ceramic objects generally expose them to discontinuous and non-uniform mechanical loads, so that the mechanical stress in small areas of the ceramic can easily exceed the capacity of the ceramic even when the overall stress is well below a value which would lead to fracture in laboratory testing. High stresses in small areas cause cracks to form, and because cracks themselves concentrate stress at their tips, a single initial crack can propagate entirely across a ceramic object, causing its catastrophic failure.

Although the term "catastrophic" is often used loosely to describe the failure of materials, for purposes of this application it is useful to give it a more precise definition, with reference to a conventional measurement of the stress induced in a material by mechanical strain. For most materials, including ceramics, the relation between stress and strain is linear at low strains. Increased strain leads eventually to a value, called the yield strain, at which the rate of increase of stress with increasing strain begins to fall below the value it had at very low strains. For typical unreinforced ceramics, the yield strain coincides with fracture of the ceramic, so that the stress falls essentially to zero. Failure of a body is defined as catastrophic for purposes of this application if the stress on the body at a strain 10% higher than the yield strain is less than 20% of the stress on the body at a strain 2% less than the yield strain.

A method well known in general terms in the art for improving the mechanical stability of typically brittle ceramics such as silicon nitride is reinforcing the ceramic with inclusions of other material, often another ceramic. Small ceramic fibers or other particles, because of more nearly perfect crystallinity, are usually stronger and sometimes more shock resistant than bulk bodies, even of the same nominal ceramic composition, which are made by conventional practical processes such as powder sintering or reaction bonding. Reinforcement, of course, need not be limited to particles of the same composition as the matrix, and often it is advantageous to utilize a different composition for some particular property in which it is superior to the matrix.

In some but far from all cases, reinforcement, especially with elongated strong fibers, will prevent catastrophic fracture of a composite, even under conditions expected to cause fracture of the matrix of the composite alone. This improvement in fracture resistance from fiber reinforcement is believed to result primarily from three mechanisms generally recognized in general terms in the prior art: load transfer, crack bridging, and debonding.

Until the recent past, most new types of fiber reinforced composites were made by workers trying to improve strength or rigidity. For such purposes a strong bond between the matrix and the reinforcing fibers is needed, so that strong bonding has usually been a goal. For example, an improvement in modulus of rupture for composites having a silicon nitride matrix formed by sintering was disclosed by Yajima et al. in U.S. Pat. No. 4,158,687. Continuous silicon carbide fibers formed by a special process described in U.S. Pat. No. 4,100,233 were used as the reinforcement, and "polycarbosilane" powder was added to the silicon nitride powder to improve the bonding between the matrix and the fibers. By these means a composite body containing unidirectionally oriented fibers with a modulus of rupture (denominated in this instance as "flexural strength") of 610 MPa was achieved. Good oxidation resistance, corrosion resistance, heat resistance, and strength at high temperatures were asserted as properties of the composites formed, but nothing was stated about the nature of the rupture of the composite.

In fiber reinforced composites with such tight bonding as illustrated by this Yajima patent, cracks resulting from concentrated mechanical stresses in the matrix tend to propagate into the fibers and crack them as well. Recent workers have discovered that such undesirable crack propagation can be avoided by surrounding the reinforcing fibers with a crack deflection zone. The crack deflection zone should have mechanical properties which will cause most cracks which propagate into the zone from the matrix either to be arrested or to follow a path which will keep them away from the reinforcing fibers.

One of the earlier workers to recognize the possible value of coating fibers with weakly bonded coatings appears to have been Warren, as exemplified by EPO Application 0 121 797 published Oct. 17, 1984. On page 4 lines 23-25, this application states, "[P]oor fiber to matrix bonds produce tough composites while good fiber to matrix bonds result in brittle, flaw sensitive materials." In the embodiment believed most relevant to the present application, the Warren application teaches forming an array of carbon fibers, coating them while they are in the array with a layer of pyrolitic carbon, machining the resulting porous body to the desired final shape, overcoating with a layer of chemically vapor desposited silicon carbide, heating to about 2700.degree. F. "to effect dimensional stability between the silicon carbide/pyrolitic carbon and the substrate", and finally overcoating again with a chemically vapor deposited silicon nitride.

Because the carbon fibers which formed the original substrate were arrayed in a fabric, felt, or similar structure before coating, the coating was not uniform around the fibers, as clearly illustrated in FIG. 3 of the Warren application drawings: where two of the original fibers touched in the original array, the coating apparently could not penetrate between them.

The pyrolytic carbon layer deposited according to the teachings of Warren was so weakly bonded that the "fibers were free to move at a different rate from the carbon and/or silicon carbide matrix systems."

Because of its low coefficient of thermal expansion, silicon nitride has long been regarded as one of the most attractive ceramics for use in conditions requiring resistance to thermal stresses. Nevertheless, the low mechanical shock resistance of unreinforced silicon carbide at almost any practical service temperatures and its low creep strength at high temperatures seriously limit its practical uses.

One of the early attempts to improve the properties of silicon nitride by inclusion of other materials in it was disclosed by Parr et al. in U.S. Pat. No. 3,222,438. This taught the inclusion of 5-10% of silicon carbide powder among silicon metal powder which was to be converted to a solid ceramic body by treatment with nitrogen gas at a sufficiently high temperature to promote the conversion of the silicon to its nitride. This process, termed reaction bonding, produced coherent silicon nitride ceramic bodies with creep resistance significantly improved over those made without the silicon carbide powder additions. The bodies to be fired were formed from powders by cold pressing in a die set, and the addition of cetyl alcohol as a binder and lubricant for the powder before pressing was recommended. The disclosure of this patent strongly recommended, and the claims all required, that the reaction bonding temperature exceed 1420.degree. C., the melting point of silicon, during part of the bonding cycle. The modulus of rupture for the composite bodies formed was not given, being described merely as comparing "favourably [sic] with those already published by others".

The use of relatively short silicon carbide fibers for reinforcing ceramics was disclosed by Hough in U.S. Pat. No. 3,462,340. Orientation of the fibers by mechanical or electrostatic forces was taught as an advantage in this patent, but no quantitative information about the mechanical properties of the resulting composites was given. Moreover, the matrix of the composites taught by this patent was limited to "pyrolitic" materials. The term "pyrolitic" was not particularly clearly defined in the patent specification, but it was apparently restricted to materials having all their chemical constituent elements derived from a gas phase in contact with the hot reinforcing filaments and a mold-like substrate which determined the inner shape of the body to be formed. No method was taught or suggested in the patent for obtaining silicon nitride as a "pyrolitic" product within this definition.

A use of very short fibers of silicon carbide to reinforce ceramic composites having a silicon nitride matrix was taught by Komeya et al. in U.S. Pat. No. 3,833,389. According to the teachings of this patent, the matrix was formed by sintering silicon nitride powder rather than by nitriding silicon metal powder, and the maximum length of the silicon carbide fiber inclusions was 40 microns. A rare earth component was required in the matrix in addition to silicon nitride, and the highest modulus of rupture (denominated as "breaking strength") was 375 megapascals (hereinafter MPa). A much more recent publication, P. Shalek et al., "Hot-Pressed SiC Whisker/Si.sub.3 N.sub.4 Matrix Composites", 65 American Ceramic Society Bulletin 351 (1986), also utilized hot pressed silicon nitride powder with elongated "whiskers" of silicon carbide as reinforcement, but these whiskers still are no more than 0.5 mm in length. Use of silicon carbide whiskers in still other matrices is taught in U.S. Pat. No. 4,543,345 of Sep. 24, 1985 to Wei (alumina, mullite, or boron carbide matrices) and U.S. Pat. No. 4,463,058 of July 31, 1984 to Hood et al. (predominantly metal matrices).

A composite with oriented continuous fiber silicon carbide reinforcement was taught by Brennan et al. in U.S. Pat. No. 4,324,843. The matrix specified by Brennan was a crystalline ceramic prepared by heating a glassy, non-crystalline powder of the same chemical composition as the matrix desired in the composite. This description of the matrix appears to exclude silicon nitride, which was not taught in the patent as a matrix material. In fact, the broadest claim of this patent required a matrix of metal aluminosilicates or mixtures thereof. Perhaps for this reason, the highest modulus of rupture noted in this patent for any of its product was less than 100 MPa.

Still another microstructural variation for silicon nitride-silicon carbide composites was disclosed by Hatta et al. in U.S. Pat. No. 4,335,217. According to this teaching, neither fibers nor powder of silicon carbide or silicon nitride is used as an initial constituent of the composite. Instead, a powdery polymer containing both silicon and carbon is mixed with silicon metal powder, pressed, and then heated in a nitrogen atmosphere. The polymer gradually decomposes under heat to yield silicon carbide, while the silicon powder reacts with nitrogen to yield silicon nitride. The composition of the final composite is described as "comprising crystals of beta-silicon carbide, alpha-silicon nitride, and beta-silicon nitride . . . forming interwoven textures of beta-silicon carbide among said alpha-silicon nitride and beta-silicon nitride crystals without chemical bonding to provide micro gaps . . . for absorption of thermal stresses." The highest reported modulus of rupture for these composites was 265 MPa.

In this Hatta patent there was also a casual reference to "Conventional SiC-Si.sub.3 N.sub.4 composite systems . . . fabricated by firing a mixture of silicon powder with . . . SiC fibers in a nitrogen gas atmosphere at a temperature above 1220.degree. C." No further details about how to make such allegedly conventional composites were given in the specification, however.

Much of the non-patent literature in the field of silicon nitride-silicon carbide composites, which i general terms covers the same ground as the patents referenced above, was reviewed by Fischbach et al. in their final report to the Department of Energy under Grants ET-78-G-01-3320 and DE-FG-01-78-ET-13389. These investigators found that the types of fibers reported as very successfully used by Yajima in U.S. Pat. No. 4,158,687 were not satisfactory for their bonding because of a tendency for the interior of these fibers to debond from the sheath layer of the fibers during nitridation.

Metal coatings for ceramic reinforcing fibers are taught in U.S. Pat. No. 3,869,335 of Mar. 4, 1975 to Siefert. Such coatings are presumably effective because the ductility of metals allows absorption of the energy of propagating cracks by distortion of the metal. Composites with metal coated fibers are satisfactory for service at relatively low temperatures, but at elevated temperatures the metal coatings can melt and thereby seriously weaken the composite. The matrices taught by Siefert were glasses, which have lower temperature service capability than ceramics. Thus for ceramics, temperature limitation is a serious disadvantage for metal coatings on the reinforcement.

U.S. Application Ser. No. 700,246 filed Feb. 11, 1985 by Rice, now U.S. Pat. No. 4,642,271, the use of boron nitride as a coating for ceramic fibers to produce a crack deflection zone when the coated fibers are incorporated into composites. Silicon carbide, alumina, and graphite fibers and silica, silicon carbide, cordierite, mullite, and zirconia matrices are specifically taught. Results were highly variable. The toughness of composites of silicon carbide fibers in silica matrices was dramatically increased by a coating of about 0.1 micron of boron nitride, but the same type of coated fibers in zirconia or cordierite produced little improvement in composite toughness compared with composites of uncoated fibers.

A reason for the effectiveness of boron nitride was suggested, and there was additional relevant information, in European Patent Application No. 0 172 082 by Societe Europeenne de Propulsion, published Feb. 19, 1986. This teaches that boron nitride coated onto fibers by gas phase reactions between boron and nitrogen containing gases, as well as carbon coatings produced by certain kinds of pyrolysis, is deposited on fibers in laminar form, with relatively weak bonds between laminae. Thus a crack which enters the coating will normally have its direction of propagation changed if necessary so that the crack will propagate along an interface between laminae of the coating. These interfaces are parallel to the fiber surface, so that the crack is usually prevented from entering the fiber. Carbon and silicon carbide fibers in silicon carbide matrices are specifically taught by this application, and other matrices, such as alumina formed by decomposition of aluminum butylate, are suggested.

The advantages of a crack deflection zone around reinforcing fibers in a different matrix is illustrated by John J. Brennan, "Interfacial Characterization of Glass and Glass-Ceramic Matrix/NICALON SiC Fiber Composites", a paper presented at the Conference on Tailoring Multiphase and Composite Ceramics, held at Pennsylvania State University, July 17-19, 1985. This teaches that certain processing conditions lead to composites in which a carbon rich layer forms around SiC reinforcing fibers, and the carbon rich layer acts as a crack deflection zone. Similarly, advantages for boron nitride coatings on reinforcing fibers are taught by B. Bender et al., "Effect of Fiber Coatings and Composite Processing on Properties of Zirconia-Based Matrix SiC Fiber Composites", 65 American Ceramic Society Bulletin 363 (1986). As expected from the titles of these reports, silicon nitride is not taught as a matrix by either of them.

J. W. Lucek et al., "Stability of Continuous SiC (-O) Reinforcing Elements in Reaction Bonded Silicon Nitride Process Environments", Metal Matrix, Carbon, and Ceramic Matrix Composites, NASA Conference Publication #2406, p. 27-38 (1985), described silicon nitride matrices reinforced with silicon carbide fibers about 10-25 microns in diameter. These SiC fibers were derived from organo-silicon polymer starting materials. High strength, non-brittle composites were not achieved with these silicon carbide fibers. Lucek et al. reported, on the basis of information supplied by others, that some of the fibers they used had been precoated with boron nitride. Whether the fibers actually were coated has been subjected to some doubt since the original report. Lucek et al., because of government security restrictions imposed on them as a condition of the supply of the allegedly coated fibers, did not attempt to characterize the composites they had prepared to a sufficient extent to determine whether boron nitride, or any other material, actually was present around the polymer derived (PD) silicon carbide fibers they used after their composites had been made with the allegedly coated fibers.

The PD SiC fibers are known to be subject to some recrystallization, with accompanying volume shrinkage, and to partial volatilization, probably preceded by chemical reaction to give volatile products, upon heating in the temperature range required for formation of reaction bonded silicon nitride (RBSN). In contrast, chemically vapor deposited (CVD) silicon carbide fibers, also briefly studied by Lucek et al., are much less subject to detrimental changes while nitriding.

Lucek et al. determined that the tensile strength of their allegedly boron nitride-coated PD SiC fibers was degraded substantially less by exposure of the fibers to the temperature and atmosphere of nitriding than was the tensile strength of similar uncoated fibers. Nevertheless, they further determined, by flexural testing of composites made with various SiC fibers, that (1) the strengths of RBSN composites reinforced with both types of PD SiC fibers were substantially less than those of composites reinforced with CVD SiC fibers, (2) the strength of such composites made with the allegedly coated PD fibers was even less than that of similar composites with uncoated PD fibers, and (3) the tensile failure of the composites with both types of PD fibers was essentially catastrophic. Presumably some unascertained part of the process of making the fibers into composites destroyed and/or changed the properties of whatever coating was on them, so that when bonded into the RBSN matrix, the coating no longer functioned effectively for crack deflection.

At present, both CVD and PD silicon carbide fibers are very expensive, but it is believed that if significant volume demand developed, PD fibers could be made at much lower costs than CVD ones. There are also fundamental advantages to the smaller diameter of the PD fibers: smaller fibers are more flexible and versatile, especially in reinforcing complex shapes which require strength in more than one direction and which have thin sections. It is practically difficult to arrange a single thickness of fibers in a plane in an array which will give substantially isotropic reinforcement. It is therefore more common to use fibers within a single layer in nearly parallel array and to superimpose layers of such fibers with different orientations in order to obtain substantially isotropic mechanical properties. Obviously, if an object with a thickness little more than that of one layer of CVD fibers is desired, such an arrangement is impossible with such fibers, but it could be accomplished with the PD fibers, which can be obtained with less than one tenth the diameter of the CVD fibers. The smaller and more flexible PD fibers also can more easily be accommodated in sharp curves of the desired composite. On the other hand, the fundamentally greater thermal stability of the CVD type fibers should make their use safer in composites intended for sustained high temperature service.

For these reasons, it is advantageous to provide strong, tough RBSN composites with PD or other small diameter SiC fiber reinforcements as well as with larger fiber reinforcements, and composites of both types made be made according to this invention.

One generalization which appears clear from the background information recited above is that the properties of composites of silicon nitride and silicon carbide, like those of composites generally, are very sensitive to the details of microstructure of the composite. (A similar conclusion was stated in the Fischbach reference already cited.) Microstructural details in turn are sensitive to the chemical and physical characteristics of the starting materials and the processes used to convert the starting materials into a coherent composite body. Little predictability about the mechanical toughness of new and different composite microstructures has been possible heretofore.

Silicon carbide fibers at least one millimeter in length can be used more advantageously than short fibers to reinforce composites with a silicon nitride matrix, especially one formed by reaction bonding. The terms "silicon carbide fibers" or "SiC fibers" or grammatical variations of these terms should be understood for the purposes of this application to include any material in fibrous form with at least 55% of its constituent atoms comprised of silicon and carbon. This specifically includes the PD type of fibers already described above, which are known to be amorphous under some conditions and to include substantial amounts of oxygen and nitrogen atoms, and fibers with a core of some other material such as carbon. The moduli of the composite products reinforced with silicon carbide fibers are particularly high if substantial regions of the composite produced contain long fibers that are substantially straight and mutually parallel, with an orientation transverse to the direction(s) of the greatest strain(s) exerted on the composite during its service life.

At a minimum, the strength and number of the individual fibers in the matrix should be high enough so that the fibers collectively are capable of bearing the load on the composite after matrix failure. Thus if necessary a complete load transfer from the matrix to the fibers can occur without mechanical failure due to tensile overloading. In mathematical terms, if 1.sub.c is the maximum load which the composite can sustain without matrix failure, v.sub.f is the fraction of fiber area in a cross section of the composite transverse to application of force, and t.sub.f is the tensile strength per unit area of the fibers, then 1.sub.c /v.sub.f should be less than or equal to t.sub.f For safety it is preferable that 1.sub.c /v.sub.f should be substantially less than t.sub.f. As is apparent from the relation given above, if the total strength of the composite is increased and fibers of the same tensile strength are used, the fiber fraction may need to be increased to meet this criterion. For example, if the total composite has a matrix failure in tension at 552 MPa and the fiber fraction is only 30%, a fiber tensile strength of at least 1.84 GPa is needed. If the fiber fraction is raised to 60%, fibers with tensile strength of 0.94 GPa would be adequate. The fraction of fibers will normally be between 20 and 80% by volume in the composites.

The resistance of the composites of this invention to catastrophic failure can be made particularly high by assuring the presence in the final composite of crack deflection zones surrounding substantially all the reinforcing fibers. A crack deflection zone is a region with significantly different mechanical properties from either the matrix or the reinforcing fibers. In general crack deflection occurs most effectively in materials with some potential new surfaces that can be formed with smaller inputs of energy than are required to form other possible new surfaces by fracturing the material of the deflection zone. Surfaces that require such relatively low energy to form can result from surface energy anisotropy within crystals, grain boundaries in polycrystalline materials, or other similar phenomena.

The crack deflection zone will often differ in chemical composition from both the reinforcing fibers and the matrix, but a difference in morphology could also be sufficient. A highly preferred characteristic of a crack deflection zone is the presence of at least one favored slippage surface running approximately parallel with the surfaces of the reinforcing fibers. The favored slippage surface may be, and often is, at or near the interface of the crack deflection zone with either the fibers or the matrix, as a result of such phenomena as compositional changes at the interface, thermal expansion coefficient mismatches, or morphological differences. Such an interfacial slippage surface is considered to be part of the crack deflection zone for purposes of this application.

Examples of appropriate deflection zone materials include carbon that is chemically vapor deposited (also called pyrolytic carbon), boron nitride, and polytypes 2H(d), 27R, 16 H, 21R, 12H, and 32H of the aluminum-nitrogen-silicon-oxygen system. Pyrolytic carbon is particularly preferred.

In co-pending Application Serial No. 893,747, of which this application is a continuation-in-part, composites with reinforcing fibers greater than 100 microns in diameter were described in detail. This application concentrates on composites with smaller fiber reinforcements.

FIG. 1 shows the time and temperature program for short nitriding according to this invention.

FIG. 2 depicts the stress-strain relationship for certain composites of silicon carbide fiber, uncoated or with various coatings, in an RBSN matrix.

FIG. 3 depicts the stress strain relationship for a composite made according to this invention, with non-catastrophic failure.

FIG. 4 depicts the stress-strain relationship for another composite according to this invention and compares it with one made with the same initial components but different nitriding conditions, which gives a different result.

FIGS. 5 and 6 are photomicrographs of a cross section through composites of the type with stress-strain curves depicted in FIG. 4.

Two effectives type of PD silicon carbide fibers for making composites according to this invention are available commercially under the trade name Nicalon fibers from Nippon Carbon Ltd., Tokyo, Japan. These fibers, described by the supplier as derived from polycarbosilane precursors, have diameters between 8 and 20 microns, and are available in two grades: standard and ceramic. The latter, which has a higher bulk density and less oxygen impurity, is generally preferred over the standard grade for making composites according to this invention. Another effective type of fiber is one supplied by Dow Corning (not yet commercially) under the designation MPDZ, Lot No. 33050-22-1. These fibers are reported to be derived from methyl poly(disilyl azane). MPDZ fibers made composites with greater toughness when used according to this invention than did the Nicalon fibers. Other fibers not specifically tested may well be equally effective.

Crack deflection zones may be advantageously provided by at least two different methods: (1) coating the reinforcing fibers with a suitable material to serve as a crack deflection zone and then preserving and/or improving the properties of the coating during nitridation, or (2) generating the crack deflection zone during nitridation from the materials of the fiber, the matrix, and/or the nitriding atmosphere. Either method is capable of giving good results; which one is preferable depends on the desired results and on the type of fibers used.

Coating: A suitable coating is pyrolytic carbon, applied according to a process taught by J. V. Marzik, "CVD Fibers", Proceedings of the Metal and Ceramic Matrix Composite Processing Conference, Vol. II, p. 39-65 (Conference held at Battelle's Columbus Laboratories from 13-15 November 1984). An initial coating thickness of at least 1 micron is normally preferred, but the most important characteristic of the coating is its crack deflection ability after processing, not the initial coating thickness.

In-Situ Generation: It has been found that PD SiC fibers normally decrease in diameter during nitridation. The mechanism for the size reduction is not clear, but surface mass loss, reduction in fiber void volume, and chemical reaction are all reasonable possibilities. The decrease in diameter may produce an empty or almost empty space around the fibers after some nitridation cycles, as in illustrated in some of the subsequent examples. In some other examples, the space between the dense part of the reinforcing fiber and the bulk of the matrix is filled with a variable amount of material presumably derived from reaction among the fiber, the matrix, and/or the nitriding atmosphere during nitridation. In at least one case, the zone around the fibers contained silicon nitride whisker, although not enough to fill the space. Either a thin gap or a zone filled with porous material such as randomly oriented whiskers can provide an effective crack deflection zone.

The placement of fibers within the shape of the desired final composite may be accomplished by any means conventional in the art. When the final product is expected to be subject to stresses in use primarily along a single direction, the fibers should be arranged as much as practicable transverse to that direction, so that the expected stress will have to bend the fibers in order to distort the body they are reinforcing. For applications with stresses applied in various directions, it may be advantageous to use several layers, with parallel orientation of fibers within each layer, but different directions between layers. In many cases, however, it will be adequate to utilize the fiber in lengths of as little as one millimeter with relatively random orientation.

Many final product structures can be effectively assembled from thin flat "tapes" containing oriented fibers. To make such tapes, a sufficient number of fibers or fiber tows to cover the desired width are supported by any appropriate mechanical means in a monolayer with the fibers substantially straight and coparallel. This fiber array is supported in some appropriate fashion so that it can be coated with a slurry of silicon powder and at least one polymeric binder in a suitable solvent.

The preferred silicon powder was a technical grade, nominally 99% pure, with a mean particle size of about 3 microns. (A suitable material was obtained from Elkem Metals Co., Marietta, Ohio.). Although many polymeric materials, natural or synthetic, such as polyvinyl acetate, vegetable gums, etc. could be used as the polymeric binder, the preferred one was a plasticized poly(vinyl butyral), marketed as Butvar 891 by Monsanto Chemical Co., Springfield, Massachusetts. About 26 parts by weight of silicon, 8.6 parts by weight of polymer (including plasticizer), 2 parts by weight of acetone, and 63.4 parts by weight of a suitable solvent such as isopropyl alcohol are mixed together. The mixture is coated by any appropriate means, such as hand application, spraying, painting, a curtain coater, etc. over the prepared array of silicon carbide fibers to a sufficient depth so as to cover the fibers after drying. The combination of fibers and slurry is dried at about 20.degree. C. for about 2 hours in the ambient atmosphere, resulting in a flexible coherent tape from which the solvent has been substantially expelled.

The tapes thus prepared may be laid up by conventional means to fit any desired final shape. To make samples of composites for testing, suitable lengths of the tape thus made were cut, stacked one atop another while preserving a common direction of orientation of the fibers within the cut lengths of tape, and mechanically pressed perpendicular to the planes of the tape segments in the stack, preferably under a pressure of at least 0.4 but not more than 0.7 Megapascals and at a temperature of about 100.degree. C. In a typical example, compressed blocks fifty millimeters in both width and length and 6-8 mm in thickness were thus prepared.

The compressed blocks, or other bodies of any shape, are then treated to remove the polymer binder constituent in the bodies. Preferably, this is accomplished by heating the bodies in an inert gas atmosphere at a rate of temperature increase of about 1150.degree. C. per hour to a final temperature of about 1150.degree. C., holding at that temperature for about fifteen minutes, and cooling by natural convection at a rate estimated to be between 100.degree. and 200.degree. C. per hour. During the heating process, the flow rate of inert gas should be maintained at a sufficient volume to sweep away any significant gaseous decomposition products formed, and the bodies should be maintained under pressure. By this process the original content of polymer binder is almost totally removed from the bodies, but because of sintering of the silicon powder particles, the bodies remain coherent.

The debinderized bodies are converted to their final ceramic form by heating the bodies in an atmosphere of nitrogen gas with chemical purity of at least 99.998%. Preferably, the nitriding is continued long enough to convert substantially all the elemental silicon in the body to silicon nitride. Because silicon has a much lower melting point than silicon nitride, substantial residual elemental silicon can limit high temperature serviceability of the final composites. Complete absence of silicon in an X ray diffraction analysis, which could detect as little as 0.1 atomic percent, is preferred.

In the long established process of forming reaction bonded silicon nitride (RBSN) without reinforcement, it has become customary to use a lengthy nitridation cycle to maximize the amount of the alpha crystal form of silicon nitride in the final product. Crystals of the alpha form were believed to form a stronger product than those of the beta form, which predominate in products prepared at higher nitrogen gas pressure and correspondingly shorter times. Processing cycles 100 hours long with a maximum temperature of 1410.degree. C. are common for making RBSN monolithic bodies.

Such long nitridation cycles as used for monolithic RBSN can sometimes be effective for making composites according to our invention, but they are rarely if ever preferred. Instead much shorter cycles, as little as 8 hours at a maximum temperature of 1350.degree. C., are generally preferred. If development of a whisker-containing crack deflection zone is desired, an intermediate length cycle such as 48 hours length may be preferred. Specific modes of practicing the invention are given in the examples below.

Nitridation for the preparation of composites according to this invention is normally done in a cold-wall vacuum furnace, by either a flow-controlled or a pressure-controlled method.

For a flow-controlled nitridation, which is preferred for in situ generation of crack deflection zones, the debinderized composite samples are initially heated in vacuo to a temperature of about 1100.degree. C. Nitrogen gas is then admitted to the furnace chamber until the total furnace pressure has reached a desired initial value. Gas flow is then interrupted and the temperature increased at a rate of about 100.degree. C. per hour. As the temperature rises, the pressure in the furnace initially rises but later falls as nitrogen is converted to non-volatile silicon nitride. The pressure drop is monitored by a sensor, and when the pressure has dropped to a pre-determined trigger value, a solenoid valve controlled by the sensor allows additional nitrogen gas into the furnace at a constant flow rate for the remainder of the nitridation.

A pressure controlled nitriding cycle is preferred for use with fibers coated with carbon coatings. In this method, the nitrogen gas pressure is controlled throughout nitridation at a fixed pre-determined value. Initial nitrogen gas pressures between 0.55 and 200 atmospheres, times between 6 and 48 hours, and final temperatures between 1325.degree. and 1400 .degree. C. could be used for such nitriding cycles in the practice of this invention. The only necessary conditions are that sufficient tensile strength be maintained in the fibers to meet the strength criterion already given above and that the properties of the interface between the fibers and the matrix, including any crack deflection zone present, are adequate to prevent catastrophic failure of the composites produced after nitriding. A pressure controlled nitriding cycle especially preferred for composites with coated Nicalon fibers is one with an initial nitrogen gas pressure of at least one atmosphere, a time of not more than fifteen hours between first heating above room temperature and the beginning of cooling, a maximum temperature not more than 1375.degree. C., and no more than ten hours above 1200.degree. C.

After completion of the nitriding treatment, the composite ceramic bodies are preferably cooled at a rate of not greater than 200.degree. C. per hour. The final result is a ceramic body resistant to thermal and mechanical shocks and suitable for long term service at temperatures up to about 1200.degree. C., at least in non-oxidizing atmospheres. These composites contain 20-50 volume percent silicon carbide filaments, and the density of the silicon nitride matrix component of the composites is believed to be about 1.8-2.0 gm/cm.sup.3.

The mechanism of initial failure of elasticity of composites made according to this invention is noncatastrophic. With all prior art SiC reinforced RBSN composites known to the applicants, the first reduction of load-bearing capacity under stress usually results in complete rupture, with obvious adverse consequences for the integrity of any structure composed of such materials.

The scope and variety of the invention can be further appreciated from the following examples.

Ceramic grade PD SiC fibers from Nippon Carbon were used for all these examples. Uncoated fibers were designated as Example 1; the coatings and corresponding example numbers are shown in Table 1 below.

Coating thicknesses shown in Table 1 were estimated by scanning electron microscopy (SEM).

Composites with fibers of each of the types shown in Table 1 were prepared by first winding, or laying up by hand, the fiber tow unidirectionally to form an array with essentially parallel fibers one tow thick. The fiber ends were secured with tape to maintain fiber alignment during coating and other processing.

A slurry coating material was prepared, having the following composition in parts by weight:

Squares 75 mm on each side were cut from the tape. Approximately eight of these squares were stacked with fiber directions in all squares the same, and the stack was then pressed at 100.degree. C. in a steel die at a pressure of about 21 MPa. The polymer bonded composite formed by this first pressing was then transferred to a graphite hot pressing die and compacted further at 1150.degree. C. and about 21 MPa pressure in an atmosphere of flowing argon gas for about fifteen minutes. This second, hot pressing served to sinter the silicon matrix and pyrolize and expel the fugitive organic binder components: poly(vinyl butyral) and butyl benzyl phthalate. The result of this process was a coherent composite with a silicon metal matrix and silicon carbide fibers.

The sintered silicon bodies were then nitrided under the conditions shown in FIG. 1 to form the final composites. As a result of this treatment, the silicon metal matrix was substantially quantitatively converted to silicon nitride, with a 66.5% weight increase and a 22% volume increase. The volume increase, however, is accommodated within the pores of the silicon matrix composite, so that no change in its external dimensions occurs during the transformation into silicon nitride.

The silicon nitride composite was sliced into test samples along two sets of perpendicular planes, each of which was parallel to the direction of the included silicon carbide fibers. These samples were used for strength tests, made in a three point bend configuration with a span of 62 mm, at a constant displacement rate of 0.,127 mm/min. Span to depth ratios varied from 16:1 to 46:1, depending on sample thickness. The as-nitrided surfaces of the samples were not disturbed, except to the extent necessary to cut the samples to the size needed for the strength testing. The samples were used for both quantitative strength evaluations, as shown in Table 2, and for qualitative assessments of failure behavior.

The short nitriding cycle produced composites with non-catastrophic failure with carbon coated fibers (Example 4). Using 48 flow controlled nitriding condifor composites containing the same coated fibers produced composites with catastrophic failure. All other types of coated or uncoated fiber, i.e., Examples 1-3 and 5, yielded composites with catastrophic failure, even with short nitriding cycles. Examination of the interfaces in the composites by Auger spectroscopy and the SEM indicated the presence of a carbon layer at least 60 nanometers thick around the SiC fibers in the non-brittle composite of Example 4.

The yield strains, maximum sustainable strains, and the stresses at the yield strains for the composites of Examples 1-5 (all with nitriding conditions as shown in FIG. 1, except for Example 3 which was nitrided for 48 hours at a maximum nitrogen pressure of 0.7 atmospheres and maximum temperature of 1380.degree. C., are shown in Table 2 below. The figures in the table were determined at a displacement rate of 0.12 mm/min.

Stress-strain curves for the composites produced by Examples 1-5 are shown in FIGS. 2 and 3. The non-catastrophic failure of the composite of Example 4 allowed it to reach a maximum sustainable stress at 0.396% strain, almost twice its yield strain, while for all the other composites shown in FIG. 2, the yield stress was the maximum sustainable stress.

All the composites made in Examples 1-5 exhibited a significant degree of load transfer from the matrix to the fiber during testing, but in all of them except Example 4, there was apparently an insufficiently effective crack deflection zone around the fibers to result in non-catastrophic failure. This indicates that fibers which retain strength after processing, load transfer, and effective crack deflection zones around the fibers are all important in achieving the toughest composites.

For this example, uncoated ceramic grade Nicalon fibers as in Example 1 above were used, but the nitriding cycle was increased to 48 hours in length. The composite produced in this example had improved toughness, compared to the product from Example 1. Microscopic examination of the composite indicated development of a zone, partially filled with a material distinct from either the fibers or the matrix, around the fibers in this composite, while no such zone was observed in the product from Example 1. This appears to account for the difference in toughness. The general physical properties of this composite were not as good as that from Example 4, however. It is believed that this difference results from degradation of the tensile strength of the fibers during the relatively long nitridation cycle.

This was the same as Example 6, except that standard rather than ceramic grade Nicalon fibers were used. The zone around the fibers after nitridation in this case had randomly oriented silicon nitride whiskers which were only very poorly if at all consolidated with the silicon nitride matrix, Qualitatively, the composites from this example were evaluated as slightly tougher than those of Example 6 and considerably tougher than those of Example 1, but they still sufferred catastrophic failure when stressed.

For these examples, uncoated MPDZ fiber was used instead of Nicalon fiber. A 48 hour flow controlled nitriding cycle was used for Example 8 and an 8 hour pressure controlled cycle for Example 9. Other processing was the same as for Example 4. Stress-strain curves of the resulting products are shown in FIG. 4. The composite with the longer nitriding cycle had noncatastrophic failure. The difference from Example 6 is believed to be due to the greater resistance of MDPZ fibers to tensile strength loss under nitriding conditions. With MDPZ fiber, short nitridation cycles such as used for Example 4 give composites subject to catastrophic failure. This is believed to be due to the fact that the shorter nitridation time does not result in the generation of a gap around the fibers in the composite, as the longer time does. The difference is apparent in cross-sections of the composites shown in FIGS. 5 and 6.

While the examples involve unidirectional filament or fiber arrays, it should be emphasized that this is not a limitation of the invention.

Composites prepared according to this invention are useful for any of the uses now served by silicon nitride, including but not limited to: thermocouple sheaths, riser stalks for low pressure die casting, crucibles, and furnace tapping seals and plugs for foundries for non-ferrous metals, particularly aluminum; degassing tubes and lining plates for primary aluminum smelters; precision jigs and fixtures for soldering, brazing, and heat treatment processes in the manufacture of electronic and semiconductor goods, jewelry, or any other metal or glass object requiring heat treating; wear resistant fixtures for optical devices, nose guides and electrode holders for electrodischarge machining, or guides and templates for electrochemical machining; welding nozzles and insulators; components of pumps, valves, or vessels for handling corrosive chemicals and abrasive mixtures; artificial teeth and dental bridges; low signature structural fairings and similar structures for aircraft and other vehicles which will be unusually difficult to detect by radar; and components for engines which can operate at higher temperatures than engines with all metal combustion containment chambers.

Composites with reaction bonded silicon nitride are particularly useful as structural ceramics, because of their combination of light weight with high stiffness and a low coefficient of thermal expansion. This combination of properties is particularly valuable for structures to be used in space, where (1) the absence of oxygen avoids one of the major limitations of RBSN in terrestrial environments, the susceptibility of RBSN to oxidative degradation at high temperatures, (2) the low weight is especially valuable because of the cost of launching weight into space, and (3) a low coefficient of thermal expansion is particularly valuable for structures exposed to sunlight on only one side.

RBSN composites are also useful for electrical applications because of their strength and dielectric properties, and as biological replacement materials, because of corrosion resistance, absence of toxicity, and ability to bond well to animal tissue.

The greater strength and toughness of composite bodies made according to this invention will make them useful in additional applications now avoided with silicon nitride bodies subject to catastrophic failure.