CA1302059C - Fiber-reinforced silicon nitride ceramics - Google Patents

Abstract

ABSTRACT OF THE DISCLOSURE
Very tough composites of silicon carbide fibers in silicon nitride matrices, especially reaction bonded silicon nitride matrices, can be made by precoating the fibers with pyrolytic carbon and controlling the nitri-dation or other process which forms the silicon nitride matrix so that a thickness of at least 5 nanometers of carbon remains in the composite after it is formed.
Failure of such composites is non-catastrophic.

Description

~3~12~S~
Docket P-1988c FIBER--REINFORCED SILICON NITRIDE CERAMICS

5 BACKGROUND OF THE INVENTIC)N
Field of the Invention This invention relates to the field of ceramic composites which comprise a continuous phase, also in-terchangeably called matrix, and a discontinuous phase, also interchangeably called reinforcement. The discon-tinuous phase is at least predominantly in the form of elongated fibers. Such materials are generally denoted in the art as fi~er reinforced composites. This in-~ention relates more particularly to composites with a matrix comprised predominantly of silicon nitride and : reinforcing fibers predominantly of silicon carbideO
Technical Background ~ ike almost all other ceramics, silicon nitride inherently has little ductility, extensibility, or oth-er capacity for stress relief, so that when subjectedfor e~en a short time to mechanical stresses in excess of its capacity, it normally breaks. Practical uses 13~'Z~S~
for ceramic objects generally e~pose them to discontin-uous and non-uniform mechanical loads, so that the me-chanical stress in small areas of the ceramic can easi-ly exceed the capacity of the ceramic even when the overall stress is well below a value which would lead to fracture in laboratory testing. ~ligh 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 pur-poses of this application it is useful to give it a more precise definition, with reference to a conven-tional measurement of the s-tress 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 even-tually 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 parti-cles, because of more nearly perfect crystallinity, are usually stronger and sometimes more shock resistant than bulk bodies, even of the same nominal ceramic com-:~L3~J21~
position, which are made by conventional practical pro-cesses such as powder sintering or reaction bonding.
Reinforcement, of course, need not be limited to parti~
cles 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 matri~.
In some but far from all cases, reinforcement, especially with elongated strong fibers, will prevent catastrophic fracture of a composite~ even under con-:
ditions expected to cause fracture of the matrix of the ; composite alone. This improvement in fracture resis-tance 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 ~iber reinforced composites were made by workers trying to improve strengkh or rigidity. For such purposes a strong bond between the matrix and the reinforcing fi-bers is needed, so that strong bonding has usually been a goal. For example, an improvement in modulus of rup-;~ ture for composites having a silicon nitride matrix formed by sintering was disclosed by Yajima et al. in ~ 25 U. S. Patent 4,158,687. Continuous silicon carbide fi-`~ bers formed by a special process described in U. S.
Patent 4,100,233 were used as the reinforcement, and "polycarbosilane" powder was added ko the silicon ni-tride powder to improve the bonding between the matrix and the fibers. By these means a composite body con-taining unidirectionally oriented fibers with a modulus of rupture ~denominated in this instance as "flexural strength") of 610 MPa was achieved. Good oxidation re-sistance, corrosion resistance, heat resistance, and strength at high temperatures were asserted as properties of the composites formed, but nothing was state~ about the nature of the rupture of the comp~s-.

'.
.

~3~ S~

ite.
In fiber reinforced composites with such tightbonding 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 un-desirable crack propagation can be avoided by surround-ing 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 re-inforcing fibers.
One of the earlier workers to recognize the pos-sible value of coating fibers with weakly bonded coat-ings appears to have been Warren, as exemplified by EPO
Application 0 121 797 published Oct. 17, 198g. On page 4 lines 23-25, this application states, "[P]oor fiber to matrix bonds produce tough composites while good fi-ber to matrix bonds result in brittle, flaw sensitivematerials." In the embodiment believed most relevant to the present application, the ~arren 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 de-sired final shape, overcoating with a layer of chemi-cally vapor desposited silicon carbide, heating to about 2700~F "to effect dlmensional stability between the silicon carbide/pyrolitic carbon and the sub-strate", and finally overcoating again with a chemical-ly vapor deposited silicon nitride.
secause the carbon fibers which formed the origi-nal substrate were arrayed in a fabric, felt, or simi-lar structure before coating, the coating was not uni-form around the fibers, as clearly illustrated inFigure 3 of the Warren application drawings: where two of the original fibers touched in the original array, ~L3~2~5~
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 difEerent rate from the carbon and/or silicon carhide matr:ix systems."
secause of its low coefficient of thermal expan-sion, silicon nitride has long been regarded as one of the most attractive ceramics for use in conditions re-quiring resistance to thermal stresses. Mevertheless,the low mechanical shock resistance of unreinforced silicon carbide at almost any practical service temper-atures and its low creep strength at high temperatures seriously limit its practical uses.
One of the early attempts to improve the proper-ties of silicon nitride by inclusion of other materials in it was disclosed by Parr et al. in U. S. Patent 3,222,438. This taught the inclusion of 5-10% of sili-con 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 coher-; ent silicon nitride ceramic bodies with creep resis-tance 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 aadition of cetyl alcohol as a binder and lubricant for the powder before pressing was recom-mended. The discIosure of this patent strongly recom-mended7 and the claims all required, that the reaction bonding temperature exceed 1420C, 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 'Ifavourably [sic] with those already published by others".
The use of relatively short silicon carbide fi-~3~J2~5~
bers for reinforcing ceramics was disclosed by Hough inU. S. Ratent 3,462,340. Orientation of the fibers hy mechanical or electrostatic forces was taught as an ad-vantage in this patent, but no quantitative information about the mechanical properties of the resulting com-posites was given. Moreover, the matrix of the compos-ites 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 sug-gested in the patent for obtaining silicon nitride as a "pyrolitic" product within this deEinition.
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. Patent 20 3,833,389. According to the teachings of this patent, ; the matrix was formed by sintering silicon nitride pow-der rather than by nitriding silicon metal powder, and the maximum length of the silicon carbide fiber inclu~
sions was 40 microns. A rare earth component was re-quired 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., I'Hot-Pressed SiC Whisker/Si3N4 Matrix Composites", 30 65 American Ceramic Society Bulletin 351 (1986), also utilized hot pressed silicon nitride powder with elon-gated "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. Paten-t 4,543,345 of Sep.
24, 1985 to Wei (alumina, mullite, or boron carbide ma-trices3 and U. S. Patent 4,463,058 of July 31, 1984 to ~3~!;2q;~t~j~
Hood et al. (predominantly metal matrices~.
~ composite with oriented continuous fiber sili-con carbide reinforcement was taught by Brennan et al.
in U. S. Patent 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 sili-con 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 sili-con nitride-silicon carbide composites was disclosed by Hatta et al. in U. S. Patent 4,335,217. ~ccordiny to this teaching, neither fibers nor powder of silicon carbide or silicon nitride is used as an initial con-stituent of the composite. Instead, a powdery polymer containing both silicon and carbon is mixed with sili-con metal powder, pressed, and then heated in a nitro-gen 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 Elatta patent there was also a casual ref-erence to "Conventional SiC-Si3N4 composite systems ...
fabricated by firing a mixture of silicon powder with ... SiC fibers in a nitrogen gas atmosphere at a tem-~3~t;~PS~

perature above 1220C. n No further details about howto 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 in general terms covers the same ground as the patents referenced abo~e, 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 ~ery successfully used by Yajima in U. S. Patent 4,158,687 were not satisfactory for their bonding be-cause 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 V. S. Patent 3,869,335 of Mar. 4, 1975 to Siefert. Such coatings are presumably effective be-cause the ductility of metals allows absorption of the energy of propaga~ing cracks by distortion of the met~
al. Composites with metal coated fibers are satisfac-tory for service at relatively low temperatures, but at elevated temperatures the metal coatinys can melt and thereby seriously weaken the composite. The matrices taught by Siefert were glasses, which have lower tem-perature service capability than ceramics. Thus for ceramics, temperature limitation i5 a serious disadvan-tage for metal coatings on the reinforcement.
U.S. Patent No. 4,642,271 issued February 10, 1987, Rice, teaches the use of boron nitride as a ~oating for ceramic fibers to produce a crack deflec-tion zone when ~.he coated ~ibers are incorporated into composites. Silicon carbide, alumina, and graphite fibers and silica, silicon carbide, cordierite, mull-ite, and zirconia matrices are specifically taughtOResults were highly variable. The toughness of compos-ites o silicon carbide fibers in silica matrices was .

~3~J2¢~
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 un-coated fibers.
A reason for the effectiveness of boron nitridewas suggested, and there was addit:ional relevant infor-mation, in European Patent Application No. ~ 172 082 by Societe Europeenne de Propulsion, published Feb. 19, 1986. This teaches that boron nitride coated onto fi-bers by gas phase reactions between boron and nitrogen containing gases, as well as carbon coatings produced by certain kinds of pyrolysis, is deposited on fihers in laminar form, with relatively weak bonds between laminae. Thus a crack which enters the coating will normally have its direction o~ propa~ation changed if necessary so that the crack will propagate a:long an in-terface between laminae of the coating. These inter-faces 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 decompo- -sition of aluminum butylate, are suggested.
The advantages of a crack deflection zone around reinforcing fibers in a different matrix is illustrated by ~ohn J. Brennan, "Interfacial Characterization of Glass and Glass-Ceramic Matrix/NICALON SiC Fiber Com-posites", a paper presented at the Conference on Tai-loring Multiphase and Composite Ceramics, held at Penn-sylvania State University, July 17-19, 1985. This teaches that certain processing conditions lead to com-posites 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 ~3~2~
Processing on Properties of Zirconia-Based Matrix SiC
Fiber Composites", 65 American Ceramic Society Bulletin 363 (1986). As expected from the titles of these re-ports, silicon nitride is not taught as a ma-trix by ei-ther of them.
J. W. Lucek et al., "Stability of Continuous Si-C(-O) Reinforcing Elements in Reaction sonded Silicon Nitride Process Environments", M al Matrix, Carbon, and Ceramic Matrix Composites, NASA Conference Publi-cation ~2~06, p. 27-38 (1985~, described silicon ni-tride 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 at al. re-ported, on the basis of information supplied by others,that some of kh~ fibers they used had been precoat~d with boron nitride~ Whether the fibers actually were coated has been subjected to some doubt since the orig-inal report. Lucek et al., because of government secu-rity restrictions imposed on them as a condition of thesupply 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 sllrinkage, and to partial volatilization, probably preceded by chemical reaction to give volatile products, upon heat-ing in the temperature range required for formation of RBSN. In contrast~ chemically vapor deposited (CVD~
silicon carbide fibers, also briefly studied by Lucek et al., are much less subject to change in mechanical properties with temperature.

~3~Z~;g 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 nitrid-5 ing than was -the tensile strength of similar uncoated~ fibers. Nevertheless, they further determined, by : flexural testing of composites made with varlous SiC
fibers, that (1) the strengths of RBSN composites rein~
forced with both types of PD SiC fibers were substan-~; 10 tially 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 fi-; bers, and (3) the tensile failure of the composites . 15 with both types of PD fibers was essentially cata-.~ strophic. Presumably some unascertained part of the process of ma]ci.ng the fibers into composites destroyed and/or changed the properties o~ whatever coating was on them, so that when bonded into the RBSN matrix, the coating no longer functioned effectively for crack de-flection.
At present, both CVD and PD silicon carbide fi-bers are very expensive, but it is believed that i.f significant volume demand developed, PD fibers could be ~; 25 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 versa-tile, especially in reinforcing complex shapes which ~` require strength in more than one direction and which have thin sections~ It is practically difficult to ar-:
range a single thickness of fibers in a plane in an . array which will give substantially isotropic reinforce-~: ment. It is therefore more common to use fibers within ; a single layer in nearly parallel array and to superim-pose layers of such fibers with different orientations in order to obtain substantially isotropic mechanical properties. Qbviously, if an object with a thickness 13':~Z6?5~
little more than that of one layer of CVD fibers is de-sired, such an arrangement is impossible with such fi-bers, but it could be accomplished with the PD f.ibers, which can be obtained with less than one tenth the di-ameter of the CVD fibers. The smaller and ~ore flexi-ble PD fibers also can more easily be accommodated in sharp curves of the desired compos:ite. On the other hand, the fundamentally greater thermal stability of the CVD type fibers should make their use safer in com-posites intended for su.stained high temperatureservice.
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 larg-er fiber reinforcements, and composites of both typesmade be made according to this invention.
One genera].ization which appears clea.~ Erom the background information recited above is that the prop-erties of composites of silicon nitride and silicon carbide, li]ce those of compos.ites generally, are very sensitive to the details of microstructure of the com-posite. (A similar conclusion was stated in the Fisch-bach reference already cited.) Microstructural details in turn are sensitive to the.chemical and physical characteristics of the starting materials and the pro-~ cesses 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 possib].e heretofore.
SUMMARY_ OF THE INVENTION
Silicon carbide fibers at least one millimeter inlength can be used more advantageously than short fi-bers to reinforce composites with a silicon nitride ma-trix, especially one formed by reaction bonding. The terms "silicon carbide fibers" or "SiC fibers" or gram-matical variations of these terms should be understood for the purposes of this application to include any 2~9 material in fibrous form with at least 55~ of its con- -stituent atoms comprised of silicon and carbon. This specifically includes the PD type of fibers already de-scribed above, which are known to be amorphous under some conditions and to include substantial amounts of oxygen and nitrogen atoms, and fihers with a core of some other material such as carbon.
Large diameter silicon fibers, at least as large as 140 microns in diameter, can be used with results at least equal to those obtained with smaller diameter fi-bers. Some of the composites have a modulus of rupture over 550 MPa, a Young's modulus at strains below 300 MPa of at least 375 GPa, and a non-catastrophic failure mode under mechanical stress. The moduli of the prod-ucts are particularly high if substantial regions ofthe composite produced contain long fibers that are substantially straight and mutually parallel, with an orientation transverse to the direction~sJ of the greatest strain~s) exerted on the composite during its service life.
At a minimum, the strength and number of the in-dividual 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 lc is the maximum load which the composite can sustain with-out matrix failure, Vf is the fraction of fiber area in a cross section of the composite transverse to applica-tion of force, and tf is the tensile strength per unit area of the fibers, then lc/vf should be less than or equal to tf. For safety it is preferable that 1c/vf should be substantially less than tf. As is apparent from the relation given above, if the total strength of the composite is increased and fibers of the same ten-; sile strength are used, the fiber fraction may need to ~3~

be increased to meet this criterion. For example, ifthe 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 Eraction is raised to 60%, fibers with ten-sile strength of 0.94 GPa would be adequa~e. The frac-tion of fibers will normally be between 20 and 80% by volume in the composites.
The resistance of the composites of this inven-tion 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 difEerent mechanical prop-erties from either the matrix or the rein~orcing fi-bers. In general crac~c deflec-tion occurs most effec-tively in materials with some potential new surfaces that can be formed with sma]ler inputs of energy than are required to form other possible new surfaces by fracturing the material of the deflection zone. Sur-faces that require such relatively ]ow energy to form can result from surface energy anisotropy within crys-tals, grain boundaries in polycrystalline materials, or other similar phenonena.
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 par-allel with the surfaces of the reinforcing fibers. The favored slippage surface may be, and often is, at or near the interface of the crac]~ 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 morpho-logical differences. Such an interfacial slippage 1~1 13 ~; 2 0 ~i; 9 surface is considered to be ~art of the crack deflec-tion zone for purposes of this application.
E~amples of appropriate deflection zone mate-rials include carbon that is ~hemically vapor deposited (also called pyrolytic carbon), boron nitride, and polytypes 2H~d~, 27R, 16 H, 21R, 12~, and 32H of the aluminum-nitrogen-silicon-oxygen system. Pyrolytic carbon is particularly preferred.
According to a broad aspect the invention relates to a composite, comprising (a) from 20-80~ by volume of reinforcing silicon carbide ceramic fibers at least one millimeter in average length, said fibers collectively having sufficient tensile strength to bear a load on said composite at the point of matrix failure without fiber tensile failure, (b) a matrix comprising predominantly reaction-bonded silicon nitride, and (c) crack deflection zones, having mechanical properties substantially different from those of both the matrix and the reinforcing fibers of the composite, occupying a predominant portion of the space around the reinforcing fibers, said composite having non-catastrophic failure under mechanical stress and a modulus of rupture of at least 550 MPa, wherein said crack deflection zones are comprised predominantly of a matérial with its most probable direction of slip under mechanical stress substantially parallel to the surfaces of said reinforcing fibers.

~- 15 -~3~Z~S~

BRIEF DESCRIPTION OF ~HE DRAWINGS
Figures 1-3 compare the stress-strain diagrams for composites of conventional reaction bonded silicon nitride and for two fiber reinforced composites accord-ing to this invention. Figure 4 illustrates the supe-rior strain-bearîng capahility of one of the composites of ~his invention. Fi~ures 5 and 6 shows the atomic ratio of silicon to carbon in the vicinity of the fiber-crack deflection zone-matrix interfaces in one typical com-posite according to this invention.
DESCRIPTIOM OF THE PREFERRED EMBODIMENTS
Reinforcing Fibers Two effectives type of silicon carbide fibers for making composites accordin~ to this invention are SCS-6 and SCS-2 grafles from AVCO Specialty Materials, Lowell, Massachusetts, with a diameter of about 140 microns.
Further details about these fibers are given in Example 1~ Fibers with similar characteristics may be avail-able from others. It is believed that these type of fibers gives better results tllan PD ones, when the ni-tridation cycle is long, because the fibers have more temperature stable strength, relatively low reactivity of the outer surface of the fiber with the matrix, and a relatively thick carbonaceous coating on the fibers.
This coating, especially on SCS-6 and similar`grades from other suppliers, is thick enough for a substantial amount of it to survive even long nitridation cycles at less than atmospheric pressure, so that it can function effectively as a crack deflection zone around the 15a ~3~'2~

fibers in final composites prepared with such long re-action bonding.
In light of past experience with ceramic compos-ites, it is surprising that fibers with such a large diameter as 140 microns do not appear to act as stren~th-limiting flaws in the composite structure, particularly when high volume loadings of such fihers are used. This is especially surprising in view of the substantial difference in thermal expansion coeffi-cients between the fibers and the matrix. Also surprising is that thorough densification of the matrix is not always necessary to achieve desirable composite properties, although as with other composites, rela-tively high densification of the matrix is preferable.
When thinner or potentially less costly fibers are desirable for any of the reasons already noted, on-ly PD silicon carbide fibers are now readily available.
These are commercially available from AVCO (noted above), Dow Corning, UBE Ltd. of Japan, ana Nippon Car-bon Ltd., Tokyo, Japan. These fibers have diameters between 10 and 20 microns.
Crack Deflection Zones Crack deflection zones are preferably provided by coating the reinforcing fibers with pyrolytic carbon, according to a process taught by J. V. Marzik, "CVD Fi-bers", Proceedings of the Metal and Ceramic Matrix Com-posite Processing Conference, Vol. II, p. 39-65 (Con-ference held at Battelle'~ Columbus Laboratories from 13-35 November 1984). An initial coating thic]cness of at least l micron is normally preferred, but the mosk important characteristic of the coating is its crack deflection ability after processing, not the initial coating thickness.
Green Body Assembly, Debinderiz~ and Sinteri~g The placement of fibers within the shape of the de~ired final composite may be accomplished by any means conventional in the art. When the final product ~3~2~S~

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 direc-tions, 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 fi-ber in lengths of as little RS one millimeter with rel-atively 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 fi~
bers or fiber tows to cover the desired width are sup-ported by any appropriate mechanical means in a mono-layer with the fibers substantially straight and copar-allel. This fiber array is supported in some appropri-ate fashion so that it can be coated with a slurry ofsilicon powder and at least one polymeric binder in a suitable solvent.
The pre~erred silicon powder was a technical grade, nominally 99~ pure, with a mean particle size of about 3 microns. (~ suitable material was obtained from Elkem Metals Co., Marietta, Ohio.). Although many polymeric materials, natural or synthetic, such as pol-yvinyl acetate, vegetable gums, etc. could be used as the polymeric binder, the preferred one was a plasti-cized poly~vinyl butyral), marketed as Butvar 891 byMonsanto Chemical Co., Springfield, Massachusetts.
; About 33 parts by weight of silicon, 12 parts by weight ; of polymer tincluding plasticizer), and 55 parts by weight of suitable solvent such as 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 ~3~2~
carbide fibers to a sufficient depth so as to cover the fibers after drying. The combination of fibers and slurry is dried at about 20C for about 2 hours in the ambient atmosphere, resulting in a flexible coheren~
tape from which the solvent has been substantially ex-pelled.
The tapes thus prepared may be laid up by conven-tional 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 orientat;on of the fibers within the cut lengths of tape, and mechani-cally pressed perpendicular to the planes of the tape segments in the stack, preferably under a pressure of at least 0.~ but not more than 0.7 Megapascals and at a temperature of about 100C. In a typical example, com-pressed blocks fiEty millimeters in both width and length and 6-8 mm in thickness were thus prepared.
The compressed blocks, or other bodies of any ; 20 shape, are then treated to remove the polymer binder constituent in the bodies. Preferably, this is accom-plished by heating the bodies in an inert gas atmo-sphere at a rate of temperature increase o~ about 1250C per hour to a final ~emperature of about 1250C, holding at that temperature for about fifteen minutes, and cooling by natural convection at a rate estimated to be between 100 and 200C per hour. During the heat-ing process, the flow rate of inert gas should be main-tained at a sufficient volume to sweep away any signif-icant gaseous decomposition products formed, and thebodies should be maintained under pressure. By this process the original content of polymer binder is al-most totally removed from the bodies, but because of sintering of the silicon powder particles, the bodies remain coherent.
Nitriding The debinderized bodies are converted to their ~3~J~

final ceramic form by heating the bodies in an atmo-sphere of nitrogen gas with chemical purity of at least 99.998~. Preferably, the nitriding is continued long enough to convert substantially all the elemental sili-con in the body to silicon nitride. Because siliconhas 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 difraction analysis, which could detect as little as 1 atomic per-cent, is preferred.
In the long established process of forming reac tion bonded silicon nitride ~RBSN) without reinforce-ment, it has become customary to use a lengthy nitrida-tion cycle to maximize the amount of the alpha crystalform 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.
The so-called "rate limited" nitridation process used to maximize the amount of alpha Si3N4 in unrein-; forced R~SN can be effectively used for making products according to this invention, if silicon carbide fibers with adequate stability are used. AVCO SCS~6 fibers,for example, will make effective composites according to this invention with a '1rate limited" nitridation process. This type of nitridation is normal~y done in a cold-wall vacuum furnace. The debinderized composite samples are initially heatea in vacu_ to a temperature of about 110~C. Nitrogen gas is then admitted to the furnace chamber until the total furnace pressure is ; slightly less than one atmosphere~ The temperature is then increased, initially at a rate of about 100C per hour. As the temperature rises, nitriding proceeds at a faster rate and begins to cause the pressure in the furnace to fall as nitrogen is coverted to non-volatile ~ 3~.1 2(11~
silicon nitrlde. The pressure drop is monitored by a sensor and a solenoid valve controlled by the sensor allows addi-tional nitrogen gas into the furnace only to the extent necessary to maintain the pressure between 0.3 and 0.7 atmospheres. When consumption of nitrogen is rapid, the temperature increase rate is greatly re-duced, so that substantially complete reaction normally requires 30-48 hours, The final temperature should be limited to a maxirnum of 1380C, below the melting point of silicon.
After completion of the nitriding treatment, the composite ceramic bodies are preferably cooled at a rate of not greater than 200 C per hour. The final result is a ceramic body resistant to thermal and me-chanical shocks and suitable for long term service attemperatures up to about 1200C, at least in non-oxi-dizing atmospheres, These composites contain 20-50 volume percent silicon carbi.de filaments, and the den-sity of the silicon nitride matrix component of the composites is believed to be about 1.8-2,0 gm/cm .
The superiority of the ceramic composites made with CVD SiC fibers according to this invention over conventional prior art composites is exemplified by the results of conventional laboratory testing in a four point or three point bend test geometry. (The four point test was performed with a 25 mm lower span and a 12 mm upper span; the three point test was performed with an 8:1 lower span to specimen depth ratio and a constant displacement rate of 8.5 x 10 4 mm/sec.) The test specimens exhibited an elastic deformation range in bending up to as much as 700 MPa, with Young's mod-uli of 380~420 GPa through this range of strains. The average modulus of rupture of several specimens was 580 MPa, substantially above those of most prior reports.
Finally, the mechanism of initial failure of elasticity was non-catastrophic. With all prior art SiC rein-forced R~SN composites known to the applicants, the ~ 31! 205~
first reduction of load-bearing capacity under stress usually results in complete rupture, with obvious ad-verse 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.
Example 1 For this example, the silicon carbide fibers used were obtained from AVCO Specialty Materials, Lowell, Massachuset-ts. The fibers consist primarily of a sheath about sixty microns thick of chemically vapor deposited silicon carbide on a filamentous carbon sub-strate fiber with diameter of about twenty microns. The SiC sheaths of these fibers have thermal, chemical, and mechanical properties that more closely approximate those of bulk stoichiometric sil:icon carbide than do PD
silicon carbide ~ibers. The outer surface of th~se fi-bers is then coated predominantly with pyrolytic car-bon; the coating probably also includes some silicon carbide. Two types of fibers, designated SCS-6 and SCS-2, were used for this example. The major differ-ence between these two types is the thickness of the carbon coating applied: about 1 micron for SCS-2 and 3 microns for SCS-6 Fibers were cut into 60 mm lengths and arranged on a polyethylene film backing strip in a single layer with each fiber straight and touching its neighboring ~iber(s) along essentially its entire length. Thus the fibers were substantially parallel and aligned in the same direction. The total width of the fiber array formed was about 60 mm. The fiber ends were secured to the polyethylene strip at their ends to maintain fibar alignment during coating and other processing.
A slurry coating material was prepared, having the following composition in parts by weight:
2 Propanol ~9 parts P~ly(vinyl butyral) 6 parts ~3~J2~S~

Butyl Benzyl Phthalate 6 parts Distilled water 6 parts Silicon metal powder 33 parts The silicon metal powder used had a mean particle size of about 3 microns and about 99% purity, with 0.7% iron as the principal impurity. The slurry was coated with a brush onto the previously prepared array of SiC fi-bers to a thickness of about 150 microns. The slurr~
had sufficient viscosity to remain on the fibers. Af-ter application of the slurry on the first side, thecoated composite was dried at about 20C for about two hours. The polyethylene backing then could be removed without disturbing the alignment of the fibers, and af-ter the backing was removed, the side of the array which had originally been against the polyethylene film was coated with the same slurry to a thickness o about ten microns, and again dried by the same conditions as after the first coating. The dried, doubly coated com-posite was flexible and is called a "tape".
Squares 50 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 room temperature 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 compact-ed further at 1250C and about 21 MPa pressure in an atmosphere of flowing argon gas for about fifteen min-utes. This second, hot pxessing served to sinter the silicon matrix and pyrolize and expel the fugitive or-ganic binder components: poly(vinyl butyral) and butyl benzyl phthalate. The result of this process was a co-herent composite with a silicon metal matrix and sil-icon carbide fibers.
The composite from the second pressing was then reacted over a period of 35 to 40 hours with high puri-~ ty nitrogen by the rate limited process already :

~3~.~Z~

described above. As a result of this treatmentr the silicon metal matrix was substantially quantitatively converted to silicon nitride, with a 66.5~ weight in-crease and a 22~ volume increase. The volume increase~
however, is accommodated w.~thin the pores of the sili-con matrix composite, so that no change in its external dimensions occurs during the transformation into sili-con 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 in-cluded silicon carbide fibers and the slices were pol-ished to a dimension of 3.18 x 3.18 x 50 mm with a grit 320 diamond grinding wheel. Apparent density and pre-dominant crystalline phases were determined on one ofthese samples and three were subjecte~l to a three point bending flexural stren~th test as al.ready described.
The average of three flexural tests and the other mea-surements compared with high quality commercial reac-tion bonded silicon nitride (RBSN) without fiber rein-forcement as shown in Table 1 below.
Table 1 MECHANICAL PROPERTI~S OF RBSN
. _ . ... ... _ With No Fibers With Fibers . . .
Apparent Density: 2400 Kg/m 2410 Kg/m3 Maximum Flexural Strength: 275 MPa over 600 MPa Failure Mode:Catastrophic Non-catastrophic ; Major Matrix ~ 30 Phases: alpha and beta Si3N4 same as at left __________ : The stress-strain diagrams of the three types axe shown in Figures 1-3. The presence of the reinforcing fiber leads to a higher strength and increased work of : fracture. The composites, unl.ike monolithic RBSN, can sustain substantial loads after initial matrix failure, ~3~Z~

as shown hy the retentlon of significant fiber stress after the first declines in such stress with increasing strain. Non-catastrophic failure of a composite pro-duced in this Example is also demonstrated in Figure 4 r which compares photographs of a composite according to this invention during and after severe flexural stress.
The behavior of composites with SCS-2 reinforce-ment, as shown in Figure 1, was intermediate between the other two types of composites in this example. At slightly above 0.6% fiber strain, the composite with SCS-2 exhibited substantial although not quite cata-strophic failure, while the composite with SCS-6 suf-fered no failure below 1.1% strain.
Measurement by Auger analysis of atomic fractions in the fracture zones of the two composites, shown graphically in Figure 5, established that in the com-posite with SCS-2, no area of essentially pure carbon remained around the fibers, while there was such a zone in the composite with SCS-6. Apparently the originally thinner layer of carbon in SCS-2 filaments was de-stroyed or infiltrated by silicon from the matrix dur-ing nitridation.
In the composites with SCS-6 reinforcing fiber, ; delamination during fracture occurred near the edge of the zone of essentially pure carbon present in these composites. This is consistent with the expected ef-fectiveness of this carbon, and/or its interfacss with the major components of the composite, as a crack de-flection zone.
Example 2 This example was the same as the part of Example 1 which used SCS-6 fibers~ except that the second pres-sing operation of that Example was replaced with uniax~
ial pressing at 100C and about 21 MPa in a steel mold.
The density of these composites was 2540 Kg/m3 and the flexural strength more than 875 MPa. The different pressin~ conditions resulted in composites with higher 2~

~ 3fJ~

density and flexural strength but lower tolerance for extension without significant damage~
Various physical property comparisons among the products from Examples 1-2 and unreinforced convention-al reaction bonded silicon nitride are shown in Table 2below. Values shown are averages of two or three spec-imens~ The "failure" specified in the heading of the second column of the table is the point of first de-crease in fiber stress with increasing strain. The composite samples never entirely broke, even up to 6 fiber strain, and retained significant fractions of their maximum strength even at such high strains, as shown in the table.
The work o~ fracture of the composites shown in Table 2 was calculated from the area under the load-displacement curve from the three point bend test al-ready described. The work of fracture is more than an order of magnitude hlgher for the reinforced samples.
Table 2 PROPERTIES OF CO_POSITES WITH CVD FIBERS
Product Percent Maximum Work of Percent of from Strain BendingFracture, Maximum Example to Stress, Joules/ Strength Number: Failure Giga- S~. Cm. ~ Retained pascals at 6~
- Strain l(SCS-6) 3.05 0.6~ 9 54 l(SCS-2) 3 8 0 9 10 31 2 1.9 0.~8 ~ 21 Unrein-forced 0.3 0 24 0.25 0 While the examples involve unidirectional fila-ment or fiber arrays, it should be emphasized that this is not a limitation of the invention.
Composites prepared according to this in~ention are useful for any of the uses now served by silicon ~3~}2~i9 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 alumi-num; degassing tubes and lining plates for primary alu-minum smelters; precision jigs and fixtures for solder-ing, brazing, and heat treatment p:rocesses in the man-ufacture of electronic and semiconductor goods, jewel-ry, or any other metal or glass object requiring heat treating; wear resistant fixtures for optical devicesr 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; artifici.al teeth and dental bridges; 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 stiff-ness 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 temperaturesr (2) the low weight is especially valuable because of the cost : of launching weight into space, and (3) a low coeffi-cient 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 replacemeni materials, because of corrosion resistance, absence of toxicity, and ability to bond well to animal tissue.

~L3~
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.

: S What is claimed is:

Claims (6)

1. A composite, comprising:
(a) from 20-80% by volume of reinforcing silicon carbide ceramic fibers at least one millimeter in average length, said fibers collectively having sufficient tensile strength to bear a load on said composite at the point of matrix failure without fiber tensile failure, (b) a matrix comprising predominantly reaction-bonded silicon nitride, and (c) crack deflection zones, having mechanical properties substantially different from those of both the matrix and the reinforcing fibers of the composite, occupying a predominant portion of the space around the reinforcing fibers, said composite having non-catastrophic failure under mechanical stress and a modulus of rupture of at least 550 MPa, wherein said crack deflection zones are comprised predominantly of a material with its most probable direction of slip under mechani-cal stress substantially parallel to the surfaces of said reinforcing fibers.

2. A composite as recited in claim 1, wherein said crack deflection zones are 5 nanometers thick and comprised pre-dominantly of a material selected from the group consisting of carbon, boron nitride, and poly-types 2H(d), 27R, 16H, 12H, and 32H of the aluminum-nitrogen-silicon-oxygen system.

3. A composite as recited in claim 1, wherein said crack deflection zones are composed predominantly of laminar-deposited pyrolytic carbon.

4. A composite as recited in claim 1, wherein said reinforcing fibers comprise no more than 55% of the volume of said composite, having a diameter of from 125-150 microns and consist predominantly of chemically vapor-deposited silicon carbide around a carbon core from 10-20 microns in diameter.

5. A composite according to claim 1 having a modulus of rupture of at least 550 megapascals and a Young's modulus at stresses below 300 megapascals of at least 375 gigapascals.

6. A composite according to claim 1, wherein said matrix is predominantly reaction bonded silicon nitride.