J.Am. Ceram.Soe,88146-153(2005) Dol:10.l11551-2916.2004.00029x Matrix Cracking in 3D Orthogonal Melt-Infiltrated SiC/SiC Composites with Various Z-Fiber Types Ohio Aerospace Institute, Cleveland, OH 44135 Hee mann Y Cleveland State University, Cleveland, OH 44115 James a. dicarlo NASA Glenn Research Center. Cleveland. OH 4413 The occurrence of matrix cracks in melt-infiltrated SiC/Sic cracking behavior in the 0o direction has been well characterized omposites with a three-dimensional (3D)orthogonal architec for different fiber types, constituent volume content, and tow ture was determined at room temperature for specimens tested in ends per cm. 7.8 tension parallel to the y-direction(perpendicular to Z-bundle Three-dimensional (3D) orthogonal(Fig. I(a)) architecture weave direction). The fiber types were Sylramic and Sylramic- SiC/SiC composites are of interest because they offer potential iBN in the X-and Y-directions and lower modulus ZMI, T300 benefits of better reproducibility, improved interlaminar me- and rayon in the Z-direction. Acoustic emission(AE) was used chanical properties, and, for the case of MI composites, po- to monitor the matrix-cracking activity. For y-direction com- tentially higher through-thickness thermal conductivity and posites, the ae data were used to determine the location (+0. 25 better matrix infiltration. It is important to understand how mm) where matrix cracks occurred in the 3D orthogonal archi- he 3D orthogonal architecture affects matrix cracking in these tecture. This enabled the determination of the stress-dependent composites. A thorough study on the accumulation of matrix matrix crack distributions for small but repeatable matrix-rich cracks and the effect of matrix crack accumulation on the unidirectional” and the matrix-poor“ ross-ply”regi ithin stress-strain behavior was performed for a 3D orthogonal com- the architecture Matrix cracking initiated at very low stresses posite with polymer impregnation and pyrolysis Sic matrix (40 MPa) in the"unidirectional""regions for the largest Z- The amount and nature of stress-dependent matrix cracking was direction fiber tow composites. Decreasing the size of the Z-fiber determined, was effectively modeled, and then was used to mod- bundle increased the stress for matrix cracking in the "unidi- rectional"regions. Matrix cracking was analyzed on the basis in 3D architecture MI SiC/SiC composites with various fiber that the source for through-thickness matrix cracks (TTMc) types in the Z-direction is not well understood. originated in the 90 or Z-fiber tows. It was found that matrix Two aspects of 3D orthogonal composites are the focus of cracking in the dimensional cross tional” regions, 一 gions was very similar to two- this study: (I)how the Z-direction tow fiber type and size affect nposites. However, in the"unidirec initiation and progression of stress-dependent matrix cracking in cracking followed a Griffith-type MI SiC/SIC composites and (2)the effect local structure has on relationship, where the stress-distribution for TTMC was matrix cracking in the different regions of the orthogonal ar- versely proportional to the square root of the height of the chitecture. When tested in the direction perpendicular to the Z- Z-fiber tows fiber tows(Y-direction, see Fig. I(b)), there are small but re- eatable matrix-rich unidirectional (UND) composite region with a Z-direction tow perpendicular to the loading direction and a 0/90 cross-ply(XPLY) composite region. With an ad- . Introduction vanced acoustic emission(AE)system, 13 the determination of tion and propagation of multiple matrix cracks in when, where, and how much matrix cracking occurred in these dense ceramic matrix composites when subjected tensile stress is necessary for high strength and osites. However the occurrence of matrix cracks at low stresses, especially for two-dimensional (2D)architectures IL. Experimental Procedure where 90 tows act as matrix flaws, may limit the structura Unload-reload tensile tests were performed on melt-infiltrated capability of some nonoxide composite systems when subjected to oxidizing environments for long times at stresses sufficient Sic matrix composite panels that were fabricated using a 3D orthogonal architecture with two different SiC fiber types in the to cause matrix cracking and hemical vapor-infiltrated(Cvi) SiC fiber-reinforced composites X-and y-direction and three different fiber types in the Z-di- fabricated from the random lay-up of 0/90 fabric, the matrix rection. In general, all the composite architectures were rein- forced in the X-y fiber directions with 10-um diameter Sylramic Sic fibers produced by Dow Corning(Midland, MI and in the Brian Cox- ting editor Z-direction by ZMI (Ube Industries, Tokyo, Japan), T300(Am- co Performance Products, Atlanta, GA), and rayon(ICF In- dustries, New York, NY). After formation into flat X=230 mm Manuscript No. 10551 Received September 23, 2003: approv by y=150 mm preform panels with Z a 2 mm thickness, all the architectures(except for the 3D with rayon Z-fiber)were Author to whom correspondence should be addressed. morschen(a gre. Senior Research Scientists residing at NASA Glenn Research Center. Cleveland, OH
Matrix Cracking in 3D Orthogonal Melt-Infiltrated SiC/SiC Composites with Various Z-Fiber Types Gregory N. Morscherw,z Ohio Aerospace Institute, Cleveland, OH 44135 Hee Mann Yunz Cleveland State University, Cleveland, OH 44115 James A. DiCarlo NASA Glenn Research Center, Cleveland, OH 44135 The occurrence of matrix cracks in melt-infiltrated SiC/SiC composites with a three-dimensional (3D) orthogonal architecture was determined at room temperature for specimens tested in tension parallel to the Y-direction (perpendicular to Z-bundle weave direction). The fiber types were Sylramic and SylramiciBN in the X- and Y-directions and lower modulus ZMI, T300, and rayon in the Z-direction. Acoustic emission (AE) was used to monitor the matrix-cracking activity. For Y-direction composites, the AE data were used to determine the location (70.25 mm) where matrix cracks occurred in the 3D orthogonal architecture. This enabled the determination of the stress-dependent matrix crack distributions for small but repeatable matrix-rich ‘‘unidirectional’’ and the matrix-poor ‘‘cross-ply’’ regions within the architecture. Matrix cracking initiated at very low stresses (B40 MPa) in the ‘‘unidirectional’’ regions for the largest Zdirection fiber tow composites. Decreasing the size of the Z-fiber bundle increased the stress for matrix cracking in the ‘‘unidirectional’’ regions. Matrix cracking was analyzed on the basis that the source for through-thickness matrix cracks (TTMC) originated in the 901 or Z-fiber tows. It was found that matrix cracking in the ‘‘cross-ply’’ regions was very similar to twodimensional cross-woven composites. However, in the ‘‘unidirectional’’ regions, matrix cracking followed a Griffith-type relationship, where the stress-distribution for TTMC was inversely proportional to the square root of the height of the Z-fiber tows. I. Introduction THE formation and propagation of multiple matrix cracks in relatively dense ceramic matrix composites when subjected to increasing tensile stress is necessary for high strength and tough composites.1 However, the occurrence of matrix cracks at low stresses, especially for two-dimensional (2D) architectures where 901 tows act as matrix flaws,2,3 may limit the structural capability of some nonoxide composite systems when subjected to oxidizing environments for long times at stresses sufficient to cause matrix cracking.4–6 For 2D melt-infiltrated (MI) and chemical vapor-infiltrated (CVI) SiC fiber-reinforced composites fabricated from the random lay-up of 0/90 fabric, the matrix cracking behavior in the 01 direction has been well characterized for different fiber types, constituent volume content, and tow ends per cm.7,8 Three-dimensional (3D) orthogonal (Fig. 1(a)) architecture SiC/SiC composites are of interest because they offer potential benefits of better reproducibility, improved interlaminar mechanical properties,9–11 and, for the case of MI composites, potentially higher through-thickness thermal conductivity and better matrix infiltration.11 It is important to understand how the 3D orthogonal architecture affects matrix cracking in these composites. A thorough study on the accumulation of matrix cracks and the effect of matrix crack accumulation on the stress-strain behavior was performed for a 3D orthogonal composite with polymer impregnation and pyrolysis SiC matrix.9 The amount and nature of stress-dependent matrix cracking was determined, was effectively modeled, and then was used to model stress-strain behavior. However, matrix-crack accumulation in 3D architecture MI SiC/SiC composites with various fiber types in the Z-direction is not well understood. Two aspects of 3D orthogonal composites are the focus of this study: (1) how the Z-direction tow fiber type and size affect initiation and progression of stress-dependent matrix cracking in MI SiC/SiC composites and (2) the effect local structure has on matrix cracking in the different regions of the orthogonal architecture. When tested in the direction perpendicular to the Z- fiber tows (Y-direction, see Fig. 1(b)), there are small but repeatable matrix-rich unidirectional (UNI) composite region with a Z-direction tow perpendicular to the loading direction and a 0/90 cross-ply (XPLY) composite region. With an advanced acoustic emission (AE) system12,13, the determination of when, where, and how much matrix cracking occurred in these different regions was accomplished. II. Experimental Procedure Unload–reload tensile tests were performed on melt-infiltrated SiC matrix composite panels that were fabricated using a 3D orthogonal architecture with two different SiC fiber types in the X- and Y-direction and three different fiber types in the Z-direction. In general, all the composite architecturesy were reinforced in the X–Y fiber directions with 10-mm diameter Sylramic SiC fibers produced by Dow Corning (Midland, MI) and in the Z-direction by ZMI (Ube Industries, Tokyo, Japan), T300 (Amoco Performance Products, Atlanta, GA), and rayon (ICF Industries, New York, NY). After formation into flat X 5 230 mm by Y 5 150 mm preform panels with Z 2 mm thickness, all the architectures (except for the 3D with rayon Z-fiber) were Journal J. Am. Ceram. Soc., 88 [1] 146–153 (2005) DOI: 10.1111/j.1551-2916.2004.00029.x 146 Brian Cox—contributing editor Research supported by NASA’s Ultra Efficient Engine Technology program. w Author to whom correspondence should be addressed. e-mail: gmorscher@grc. nasa.gov z Senior Research Scientists residing at NASA Glenn Research Center, Cleveland, OH. Manuscript No. 10551. Received September 23, 2003; approved June 17, 2004. y 3D orthogonal preforms woven by Albany International Techniweave, Inc. (Rochester, NY)
January 2005 Matrix Cracking in 3D Orthogonal Mell-Infiltrated Composites X Table I. Properties for Three Types of 3D Architectures Y- and y-tow ends direction per cm Thickness ZMIT 2050.1740.1520.032 Syl-iBN 7.9 1950.2030.1600.001 (Atlanta GA) 1000 fibers per tow, average fiber diameter= 7 um-Property Data Sheet supplied by manufacturer. ICF Industries(New York, NY), 400 fibers fiber diameter= 12 um-Data supplied by Albany International Techniweave, Inc.NASA-modified Sylramie fiber-by heat treatment of the Sylramic fiber to produce a thin, -100 nm, in layer on the surface of the fiber. Dow Corning(Midland, MI) 800 fibers per tow, average diameter 10 m-Property Data Sheet supplied by manufacturer. For all three architectures, 3.95 epcm of a double tow was woven in the x-direction. As will be discussed, a key aspect of the Z-fiber types is their tow size in the final as-processed composite, which was largest for the ZMI fibers and smallest for the rayon fibers due to significant decomposition during composite fabrication. For convenience he three different types of 3D orthogonal composites are referred <PXPLY- X to in this paper by their particular Z-fiber type. T zes all the key properties of the 3D architectures Tensile tests were performed on specimens 12-mm wide by 150-mm long with a contoured gage section(dog-bone)of width 10 mm using a universal testing machine(Instron Model 8562, Instron Ltd, Canton, MA) with an electromechanical actuator. The 3D composites were tested with the y-direction oriented in the loading direction. Glass-fiber-reinforced epoxy tabs were ounted on both sides of the specimen in the grip regions and the specimens were gripped with rigidly mounted hydraulical actuated wedge grips. A clip on strain gage, with a range of 2.5% strain over 25 4-mm gage length was used to measure the deformation of the gage section Modal ae was monitored during the tensile tests with two ig. 1. Schematic representation of (a)3D orthogonal composite and ide-band, 50 kHz to 2.0 MHz, high-fidelity sensors placed just b)regions aligned in the Y-direction. Only nine layers are represented in outside the tapered region of the dog-bone specil b). the composites actually contained 15 layers(see Fig. I(a)) uum grease was used as a couplant and mechanical clips were used to mount the sensors to the specimen. The aE waveforms were recorded and digitized using a 4-channel, fracture wa detector(FWD) produced by Digital Wave Corporation(Engle- treated at NASa to convert the Sylramic fibers to the higher wood, CO). The load and strain were also recorded with the performance Sylramic-iBN SiC fiber. 4 The Sylramic-iBN or FWD. Location of the ae events was determined from the dif- Sylramic fibers in the preforms were then coated by CVI with ference in times of arrival of the first peak, the measured stress- BN interphase coatings and Sic matrices. Porosity between the dependent speed of sound, and the distance between the two CVI-coated"SiC/SiC mini-composite"tows was then filled by sensors as in Morscher. 12,13 However, the threshold technique slurry infiltration of Sic particles, followed by molten infiltra typically used for determination of the time of arrival was not tion of Si, commonly known as melt-infiltration, or MI.The used. Instead, the time of arrival of the first peak of the exten- composite panels were very dense, usually with less than 5% sional wave for each event waveform on the two sensors was porosity, most of which was in the form of inter-tow porosity determined by manual inspection in order to get a location ac- The 3D orthogonal panels(see Fig. I)had seven layers in the curacy of less than +0.25 mm. In this way, the AE activit Y-direction and eight layers in the Y-direction and were conse- within the UNI and XPLY regions of the 3D composites or quently not balanced in terms of fiber content. In addition, two ented in the y-direction( Fig. 1(b)could be distinguished. Sections from the the gage section of the tested tensile spe woven together in the X-direction at 3.95 tow-ends-per-cm nens at least 10 mm in length were polished and then plasma (epcm) for a fiber volume fraction of 15-18%. For the Y-direc-(CF4)etched at 500 W for 30 min. Etching was required to ob- tion, single tows were woven at either 7.I or 7.9 epcm for fiber serve transverse matrix cracks: however. the etchant reacts with ractions of 17-23%(see Table D). For the out-of-plane rein- the free Si in the matrix, removing much of it, making it im- forcement, very low fiber volume fractions(<3%)were used possible to observe the extension of matrix cracks through the MI part of the matrix. Matrix cracks can only be observed in the fiber types: II um ZMI SiC fibers from(800 fibers/tow ) 7 dense CVI SiC layer between the BN and the MI matrix T300 carbon fibers (1000 fibers/tow), and 12 um polymer- derived rayon fibers from ICF Industries(400 fiberstow ). The X- Y fibers of the rayon composite were not converted to Sylramic iBN fibers because the rayon fiber is subject to decomposition at The accuracy was dependent on the resolut the separation dis- the elevated temperatures required for Sylramic-iBN treatment ured during the test from AE that occurred outside the i.e., in the grips Interphase and fabrication was performed by General Electric Power Systems tapered section of the dog-bone. As the speed of sound decreased, the accuracy of event location increased
treated at NASA to convert the Sylramic fibers to the higherperformance Sylramic-iBN SiC fiber.14 The Sylramic-iBN or Sylramic fibers in the preforms were then coated by CVI with BN interphase coatings and SiC matrices. Porosity between the CVI-coated ‘‘SiC/SiC mini-composite’’ tows was then filled by slurry infiltration of SiC particles, followed by molten infiltration of Si, commonly known as melt-infiltration, or MI.z15 The composite panels were very dense, usually with less than 5% porosity, most of which was in the form of inter-tow porosity. The 3D orthogonal panels (see Fig. 1) had seven layers in the X-direction and eight layers in the Y-direction and were consequently not balanced in terms of fiber content. In addition, two Sylramic SiC fiber tows (800 fibers per tow) were combined and woven together in the X-direction at 3.95 tow-ends-per-cm (epcm) for a fiber volume fraction of 15–18%. For the Y-direction, single tows were woven at either 7.1 or 7.9 epcm for fiber fractions of 17–23% (see Table I). For the out-of-plane reinforcement, very low fiber volume fractions (o3%) were used based on the single-tow weaving of three different Z-direction fiber types: 11 mm ZMI SiC fibers from (800 fibers/tow), 7 mm T300 carbon fibers (1000 fibers/tow), and 12 mm polymerderived rayon fibers from ICF Industries (400 fibers/tow). The X– Y fibers of the rayon composite were not converted to SylramiciBN fibers because the rayon fiber is subject to decomposition at the elevated temperatures required for Sylramic-iBN treatment. As will be discussed, a key aspect of the Z-fiber types is their tow size in the final as-processed composite, which was largest for the ZMI fibers and smallest for the rayon fibers due to significant decomposition during composite fabrication. For convenience, the three different types of 3D orthogonal composites are referred to in this paper by their particular Z-fiber type. Table I summarizes all the key properties of the 3D architectures. Tensile tests were performed on specimens 12-mm wide by 150-mm long with a contoured gage section (dog-bone) of width 10 mm using a universal testing machine (Instron Model 8562, Instron, Ltd., Canton, MA) with an electromechanical actuator. The 3D composites were tested with the Y-direction oriented in the loading direction. Glass-fiber-reinforced epoxy tabs were mounted on both sides of the specimen in the grip regions and the specimens were gripped with rigidly mounted hydraulically actuated wedge grips. A clip on strain gage, with a range of 2.5% strain over 25.4-mm gage length was used to measure the deformation of the gage section. Modal AE was monitored during the tensile tests with two wide-band, 50 kHz to 2.0 MHz, high-fidelity sensors placed just outside the tapered region of the dog-bone specimen.12,13 Vacuum grease was used as a couplant and mechanical clips were used to mount the sensors to the specimen. The AE waveforms were recorded and digitized using a 4-channel, fracture wave detector (FWD) produced by Digital Wave Corporation (Englewood, CO). The load and strain were also recorded with the FWD. Location of the AE events was determined from the difference in times of arrival of the first peak, the measured stressdependent speed of sound, and the distance between the two sensors as in Morscher.12,13 However, the threshold technique typically used for determination of the time of arrival was not used. Instead, the time of arrival of the first peak of the extensional wave for each event waveform on the two sensors was determined by manual inspection in order to get a location accuracy of less than 70.25 mm.J In this way, the AE activity within the UNI and XPLY regions of the 3D composites oriented in the Y-direction (Fig. 1(b)) could be distinguished. Sections from the the gage section of the tested tensile specimens at least 10 mm in length were polished and then plasma (CF4) etched at 500 W for 30 min. Etching was required to observe transverse matrix cracks; however, the etchant reacts with the free Si in the matrix, removing much of it, making it impossible to observe the extension of matrix cracks through the MI part of the matrix. Matrix cracks can only be observed in the dense CVI SiC layer between the BN and the MI matrix. Fig. 1. Schematic representation of (a) 3D orthogonal composite and (b) regions aligned in the Y-direction. Only nine layers are represented in (b), the composites actually contained 15 layers (see Fig. 1(a)). Table I. Properties for Three Types of 3D Architectures Z-Fiber type X- and Y-direction fiber type Y-tow ends per cm (epcm)ww Thickness (mm) fy fx fz ZMIw Syl-iBNz 7.1 2.05 0.174 0.152 0.032 T300z Syl-iBN 7.9 1.75 0.226 0.178 0.014 Rayony SylramicJ 7.9 1.95 0.203 0.160 0.001 w Ube Industries (Tokyo, Japan), 800 fibers per tow, average fiber diameter 5 11 mm—Property Data Sheet supplied by manufacturer. z Amoco Performance Products, (Atlanta, GA), 1000 fibers per tow, average fiber diameter 5 7 mm—Property Data Sheet supplied by manufacturer. y ICF Industries (New York, NY), 400 fibers per tow, average fiber diameter 5 12 mm—Data supplied by Albany International Techniweave, Inc. z NASA-modified Sylramic fiber—by heat treatment of the Sylramic fiber to produce a thin, B100 nm, in situ BN layer on the surface of the fiber. J Dow Corning (Midland, MI); 800 fibers per tow, average diameter B10 mm—Property Data Sheet supplied by manufacturer. wwFor all three architectures, 3.95 epcm of a double tow was woven in the X-direction. z Interphase and matrix fabrication was performed by General Electric Power Systems Composites (Newark, DE). J The accuracy was dependent on the resolution in the time domain, the separation distance of the sensors, and the speed of sound across the material. For this study, those parameters were 0.1 microseconds, 50 mm, and B9000 m/s for a pristine composite, respectively. As matrix cracking occurs, the speed of sound decreases. The speed of sound was measured during the test from AE that occurred outside the sensors,12 i.e., in the grips or tapered section of the dog-bone. As the speed of sound decreased, the accuracy of event location increased. January 2005 Matrix Cracking in 3D Orthogonal Melt-Infiltrated Composites 147
148 Journal of the American Ceramic Society--Morscher et al. Vol. 88. No I. Results (1) Stress-Strain behavio The stress-strain behavior for the different 3D orthogonal ten- 5 sile specimens is shown in Fig. 2. The failure stress for the 3D 9 rthogonal composites was similar; however, not all the com- 75 posites failed in the gage section There was also some difference 6 in the debonding and sliding character in the bn interphase re- gion. The ZMI composite exhibited debonding and sliding in between the Sylramic-iBN fiber and the Bn interphase(inside debonding) as is typical for most MI composites. However the other two Z-direction fiber-type composites exhibited a mixture of inside debonding and outside debonding, that is, debonding nd sliding between the fibers and Bn interphase and between 4.5 the bn interphases and the CvI Sic matrix, respectively. Out- side debonding and mixed inside and outside debonding com- 75 posites have been shown to have lower interfacial shear trengths for Mi composites than purely inside debonding com- 105 posites. Also shown in Fig. 2 is a typical Sylramic-iBN rein- forced 2D woven composite(7.9 epcm, five harness satin,).The 2D composite failure stress was significantly greater than that of (a) Stress. MPa the 3D orthogonal composites. It is currently unclear why the 3 panels were weaker than the 2D panel. However, the main interest in this study pertains to the area where matrix cracking 感 occurs, i.e., just before, at, and after the observation of nonlin- earity in the stress-strain curve (open symbols) (2) Acoustic Emission and Matrix Cracking 盟 The location of AE activity versus applied stress is plotted Fig 3(a) for the ZMI composite y-direction. The data is sepa rated into the highest decade of energy (high energy), the second symbols) highest decade of energy(mid energy), and the lowest three dec- ades of energy (low energy). The crack density for these systems with this modal AE approach has been shown to be nearly di- rectly proportional to AE energy. 12 3 Also shown in Fig. 3(a)is a schematic of the 3D orthogonal architecture commensurate with the location of the tested specimen. It is clear that initial AE Fig 3. (a) Location of acoustic emission(AE) events along the occurs in the UNI regions of the 3D orthogonal com- of a ZMI specimen oriented in the Y-direction.(b) Cumulative t stresses below 50 MPa. This is significantly lower than ergy for each 1. 5-mm region corresponding to the unidirectional D MI composites where first AE activity occurs at open symbols) and the cross-ply (XPLY)(closed symbols)regions of (a) 100+20 MPa. Initial AE activity was composed of lower energy events, probably tunnel cracks. I that form within the Z- bundle At <115 MPa, high-energy events occurred in the UNI region, signifying large matrix crack formation and growth. Sig- 175-200 MPa in the UNI region that would correspond to nificant AE activity did not occur in the XPLY region until near matrix crack saturation in the UNI region. However, for stresses greater than 140 MPa. the XPLY regions, significant cumulative AE activity begins to The AE activity in the UNI and XPLY regions is occur at a slightly higher stress(130 MPa) to the UNI regions, cumulative AE energy for each 1.5 mm length but the distribution of cumulative aE energy is over a very wide Fig. 3(b). Significant cumulative AE activity in the U. stress range and does not appear to cease even up to failure, i. e increases very rapidly starting at 115 MPa with increasing the XPly region does not appear to saturate with matrix cracks stress over a narrow stress range. AE activity diminishes above wyp to the fracture stress of the specimen. It is also apparent that was significantly higher in mag- nitude than the XPLY region. Evidently, matrix cracks in the denser, higher modulus matrix region are of a higher energy han those formed in the predominantly 90 bundles of the 2D 7.epcm XPLY region Gage Failure\ Figure 4 shows the ae data for the UNI and XPLY regions of all the composites. All the ae data for the UNI and XPLY 3DT300 regions of a specimen were combined and normalized by the total cumulative AE energy at failure of each specimen for each Gage Failure composite. The normalized cumulative AE energy was then 3D-ZMI multiplied by the final crack density (Table ID) to determine an Gage Failure estimated stress-dependent matrix crack density' for the diffe ent regions of each specimen. The matrix-cracking stress ranges e higher for T300 and rayon composites in both the UNI and KPLY regions compared with the ZMI composite with rayon opposites having the highest matrix-cracking stress range for 0.4 oth regions. Significant matrix cracking always occurred for the xPly regions at much higher stresses than the UNi regions tressstrain data for 3D orthe osites tested in the y for the same composite. Figure 5 shows regions of the(a) ZMI and a representative 2D wover The unload-reload and(b) rayon specimens and typical matrix cracks that formed loops were removed from the stress-strain curves for clarity n the UNi region
III. Results (1) Stress–Strain Behavior The stress–strain behavior for the different 3D orthogonal tensile specimens is shown in Fig. 2. The failure stress for the 3D orthogonal composites was similar; however, not all the composites failed in the gage section. There was also some difference in the debonding and sliding character in the BN interphase region. The ZMI composite exhibited debonding and sliding in between the Sylramic-iBN fiber and the BN interphase (inside debonding) as is typical for most MI composites. However, the other two Z-direction fiber-type composites exhibited a mixture of inside debonding and outside debonding, that is, debonding and sliding between the fibers and BN interphase and between the BN interphases and the CVI SiC matrix, respectively. Outside debonding and mixed inside and outside debonding composites have been shown to have lower interfacial shear strengths for MI composites than purely inside debonding composites.16 Also shown in Fig. 2 is a typical Sylramic-iBN reinforced 2D woven composite (7.9 epcm, five harness satin7 ). The 2D composite failure stress was significantly greater than that of the 3D orthogonal composites. It is currently unclear why the 3D panels were weaker than the 2D panel. However, the main interest in this study pertains to the area where matrix cracking occurs, i.e., just before, at, and after the observation of nonlinearity in the stress–strain curve. (2) Acoustic Emission and Matrix Cracking The location of AE activity versus applied stress is plotted in Fig. 3(a) for the ZMI composite Y-direction. The data is separated into the highest decade of energy (high energy), the second highest decade of energy (mid energy), and the lowest three decades of energy (low energy). The crack density for these systems with this modal AE approach has been shown to be nearly directly proportional to AE energy.12,13 Also shown in Fig. 3(a) is a schematic of the 3D orthogonal architecture commensurate with the location of the tested specimen. It is clear that initial AE activity occurs in the UNI regions of the 3D orthogonal composite at stresses below 50 MPa. This is significantly lower than typical 2D MI composites where first AE activity occurs at B100720 MPa.7,17 Initial AE activity was composed of lower energy events, probably tunnel cracks8,18 that form within the Zbundle. At B115 MPa, high-energy events occurred in the UNI region, signifying large matrix crack formation and growth. Significant AE activity did not occur in the XPLY region until stresses greater than 140 MPa. The AE activity in the UNI and XPLY regions is plotted as cumulative AE energy for each 1.5 mm length section in Fig. 3(b). Significant cumulative AE activity in the UNI region increases very rapidly starting at B115 MPa with increasing stress over a narrow stress range. AE activity diminishes above B175–200 MPa in the UNI region that would correspond to near matrix crack saturation in the UNI region. However, for the XPLY regions, significant cumulative AE activity begins to occur at a slightly higher stress (130 MPa) to the UNI regions, but the distribution of cumulative AE energy is over a very wide stress range and does not appear to cease even up to failure, i.e., the XPLY region does not appear to saturate with matrix cracks up to the fracture stress of the specimen. It is also apparent that AE energy in the UNI region was significantly higher in magnitude than the XPLY region. Evidently, matrix cracks in the denser, higher modulus matrix region are of a higher energy than those formed in the predominantly 901 bundles of the XPLY region. Figure 4 shows the AE data for the UNI and XPLY regions of all the composites. All the AE data for the UNI and XPLY regions of a specimen were combined and normalized by the total cumulative AE energy at failure of each specimen for each composite. The normalized cumulative AE energy was then multiplied by the final crack density (Table II) to determine an estimated stress-dependent matrix crack density7 for the different regions of each specimen. The matrix-cracking stress ranges are higher for T300 and rayon composites in both the UNI and XPLY regions compared with the ZMI composite with rayon composites having the highest matrix-cracking stress range for both regions. Significant matrix cracking always occurred for the XPLY regions at much higher stresses than the UNI regions for the same composite. Figure 5 shows regions of the (a) ZMI and (b) rayon specimens and typical matrix cracks that formed in the UNI region. 0 100 200 300 400 500 0 0.1 0.2 0.3 0.4 0.5 0.6 Strain, % Stress, MPa 3D-ZMI E= 248 GPa Gage Failure 3D-T300 E= 237 GPa Gage Failure 3D-Rayon E=238 GPa Radius Failure 2D 7.9epcm E = 228 GPa Gage Failure Fig. 2. Stress-strain data for 3D orthogonal composites tested in the Ydirection and a representative 2D woven composite. The unload–reload hysteresis loops were removed from the stress–strain curves for clarity. -12 -10.5 -9 -7.5 -6 -4.5 -3 -1.5 0 1.5 3 6 7.5 9 10.5 12 0 100 200 300 Location, mm High Energy Mid Energy Low Energy . Y Z . High Energy Mid Energy Low Energy 4.5 Stress, MPa High Energy Mid Energy Low Energy 0 100 200 300 400 500 600 0 50 100 150 200 250 300 350 Stress, MPa Cum AE Energy UNI regions (open symbols) XPLY regions (closed symbols) (a) (b) Fig. 3. (a) Location of acoustic emission (AE) events along the length of a ZMI specimen oriented in the Y-direction. (b) Cumulative AE energy for each 1.5-mm region corresponding to the unidirectional (UNI) (open symbols) and the cross-ply (XPLY) (closed symbols) regions of (a). 148 Journal of the American Ceramic Society—Morscher et al. Vol. 88, No. 1
January 2005 Matrix Cracking in 3D Orthogonal Melt-Infiltrated Composites 2D 7.epcm 10 ZMI UNI Y-Direction ZMI XPLY EE8普6 T300 UNI 4 XPLY Rayon XPLY 0 Remains of rayon Z-fiber Fig 4. Estimated matrix crack density, based on acoustic emission ac- tivity. for the two regions of 3D Y-direction-oriented composites and a Most of the matrix cracking for the UNI region of a gi specimen occurred over a narrow stress range with the mid tress of the distribution increasing with decreasing Z-direction size. The height of the Z-direction tow is 0.15 mm for ZM 0. 1l mm for T300, and 0.03 mm for rayon when measured from polished longitudinal sections as in Fig. 5 where the Z- Y-direction direction tow is closest to the face of the composite. When measured in the interior of the composite, the tow height in- creases to 0.40, 0.28, and 0.15 mm for ZMI. T300, and rayon, respectively. Figure 5(b)shows the small rayon tow height. Note that the rayon fiber was expected to have decomposed during (b) the cvi bn step; however, some porous carbon char remained Fig. 5. Polished longitudinal section of (a)ZMI specimen and(b)rayon in the form of what appears to be an approximately 60"fiber ecimen tested in the y-direction orientation. arrows indicate matrix tow. Fibers in the rayon composite were very straight in both the cracks UN and XPLY regions compared with the other Z-toy posites, e.g- ZMI(Fig. 5(a)). This may also contribute to the balanced 2D architecture. the volume fraction of oo mini-com- high matrix crack stresses for this composite. It was quite re- markable that the very first cracks(as detected by AE) in the posites, minit is half the total fraction of fiber, BN, and CVI Si rayon composite occurred at 170 MPa, significantly higher(at ithin the composite. The elastic modulus of the mini-compos ites. Ein can be estimated via the rule-of-mixtures from the least 40 MPa) than the other two composites or any other 2D elastic moduli of each constituent of the mini-compo MI composite tested to date. (Er≈380GPa,EBN≈60GPa, and ecviSic≈425GPa)and the volume fraction of each constituent in the loading directio (3) Stress Dependence for Matrix Cracking Appendix a describes how this was accomplished for the 3D In a prior study concerning through-thickness matrix cracking composites TTMC) in similar 2D melt-infiltrated SiC/SiC composites, it Matrix crack density versus mini-matrix stress is plotted in was shown that an important parameter controlling cracking is Fig. 6 for the 3D composites tested in the y-direction as well as the effective stress acting on the 90 tows that exist in the matrix for the single-tow 2D composite(from Morscher) Matrix crack rich regions outside of the 0 mini-composites(hereafter termed densities of the different 3D XPLY composites and the 2D wo- the""mini-matrix"region). This mini-matrix stress, ominik-matrix. ven composite have a very similar dependence on mini-matrix can be determined from simple rule-of-mixtures theory stress. However, the stress range where matrix cracking occurs in the UNI regions appears to be dependent on the tow size of (oc+Oth)/Ec-fmini Emini the Z-fiber type, i.e., the smaller the tow size(rayon fibers), the Omini-matrix (1) higher the matrix-cracking stress range. where oe is the applied composite tensile stress, oth is the resid (4) XPLY Matrix Crackin ual compressive stress in the matrix determined from the hys- For the XPLY regions of the 3D composites a teresis loops of the tensile test, ",and Ec is the composite elastic amount of low energy ae was observed at stresses below the modulus measured from the tensile stress-strain curve. For a onset of high-energy AE activity. This low-stress low AE energy activity can be attributed to the formation of tunnel cracks Table ll. Matrix Crack densities thin the 90 X-fiber tows. Tunnel cracking began at a mini matrix stress of 20 MPa in the XPly regions for both the ZMI and T300 composites. However, for the rayon composite Composite Z-fiber type density (mm-) density(mm-) tunnel cracking in the XPLY region began at -60 MPa. The 90 x-direction SiC/SiC mini-composites in the rayon compos- ZMI (Y direction) 10.2 8.8 ites were noticeably longer and thinner than the x-direction T300(r direction) Rayon(Y direction 4.8 tows in the T300 and ZMI composites. The measured maximum height of a 90 tow(hx in Table Ill and Fig. I; see Appendix A)
Most of the matrix cracking for the UNI region of a given specimen occurred over a narrow stress range with the midstress of the distribution increasing with decreasing Z-direction tow size. The height of the Z-direction tow is B0.15 mm for ZMI, 0.11 mm for T300, and 0.03 mm for rayon when measured from polished longitudinal sections as in Fig. 5 where the Zdirection tow is closest to the face of the composite. When measured in the interior of the composite, the tow height increases to 0.40, 0.28, and 0.15 mm for ZMI, T300, and rayon, respectively. Figure 5(b) shows the small rayon tow height. Note that the rayon fiber was expected to have decomposed during the CVI BN step; however, some porous carbon char remained in the form of what appears to be an approximately 60 ‘‘fiber’’ tow. Fibers in the rayon composite were very straight in both the UNI and XPLY regions compared with the other Z-tow composites, e.g., ZMI (Fig. 5(a)). This may also contribute to the high matrix crack stresses for this composite. It was quite remarkable that the very first cracks (as detected by AE) in the rayon composite occurred at B170 MPa, significantly higher (at least 40 MPa) than the other two composites or any other 2D MI composite tested to date. (3) Stress Dependence for Matrix Cracking In a prior study concerning through-thickness matrix cracking (TTMC) in similar 2D melt-infiltrated SiC/SiC composites,7,8 it was shown that an important parameter controlling cracking is the effective stress acting on the 901 tows that exist in the matrixrich regions outside of the 01 mini-composites (hereafter termed the ‘‘mini-matrix’’ region). This mini-matrix stress, smini-matrix, can be determined from simple rule-of-mixtures theory: smini-matrix ¼ ð Þ sc þ sth Ec Ec fminiEmini 1 fmini ð1Þ where sc is the applied composite tensile stress, sth is the residual compressive stress in the matrix determined from the hysteresis loops of the tensile test,7,19 and Ec is the composite elastic modulus measured from the tensile stress–strain curve. For a balanced 2D architecture, the volume fraction of 01 mini-composites, fmini, is half the total fraction of fiber, BN, and CVI SiC within the composite. The elastic modulus of the mini-composites, Emini, can be estimated via the rule-of-mixtures from the elastic moduli of each constituent of the mini-composite (Ef 380 GPa, EBN 60 GPa, and ECVI-SiC 425 GPa) and the volume fraction of each constituent in the loading direction. Appendix A describes how this was accomplished for the 3D composites. Matrix crack density versus mini-matrix stress is plotted in Fig. 6 for the 3D composites tested in the Y-direction as well as for the single-tow 2D composite (from Morscher7 ). Matrix crack densities of the different 3D XPLY composites and the 2D woven composite have a very similar dependence on mini-matrix stress. However, the stress range where matrix cracking occurs in the UNI regions appears to be dependent on the tow size of the Z-fiber type, i.e., the smaller the tow size (rayon fibers), the higher the matrix-cracking stress range. (4) XPLY Matrix Cracking For the XPLY regions of the 3D composites a significant amount of low energy AE was observed at stresses below the onset of high-energy AE activity. This low-stress low AE energy activity can be attributed to the formation of tunnel cracks within the 901 X-fiber tows. Tunnel cracking began at a minimatrix stress of B20 MPa in the XPLY regions for both the ZMI and T300 composites. However, for the rayon composite tunnel cracking in the XPLY region began at B60 MPa. The 901 X-direction SiC/SiC mini-composites in the rayon composites were noticeably longer and thinner than the X-direction tows in the T300 and ZMI composites. The measured maximum height of a 901 tow (hx in Table III and Fig. 1; see Appendix A) 0 2 4 6 8 10 12 0 100 200 300 400 500 Composite Stress, MPa Crack density, mm-1 ZMI UNI T300 UNI Rayon UNI ZMI XPLY 2D 7.9epcm T300 XPLY Rayon XPLY I Fig. 4. Estimated matrix crack density, based on acoustic emission activity, for the two regions of 3D Y-direction-oriented composites and a representative 2D woven composite. Table II. Matrix Crack Densities Composite Z-fiber type Average UNI crack density (mm1 ) Average XPLY crack density (mm1 ) ZMI (Y direction) 10.2 8.8 T300 (Y direction) 7.4 4.3 Rayon (Y direction) 4.8 4.8 Y-Direction Matrix cracks i 1 mm 1mm Y-direction Remains of rayon Z-fiber (a) (b) Fig. 5. Polished longitudinal section of (a) ZMI specimen and (b) rayon specimen tested in the Y-direction orientation. Arrows indicate matrix cracks. January 2005 Matrix Cracking in 3D Orthogonal Melt-Infiltrated Composites 149
Journal of the American Ceramic Society-Morscher et al. Vol. 88. No 2D 7.epc Onset Composite St 150- Slope=2.17 MPa m1/2 ZMI UNI 8 T300 UNI 6 ZMI XPLY Onset Minimatrix Stress 4 Slope= 1.07 MPa Rayon XPLY (a) 〔 Tow height)12,mm1l2 T300 XPLY 10 ZMI UNI Minimatrix Stress MPa Fig. 6. Estimated crack density plotted versus mini-matrix stress in the T300 and ZMI composites ranged from 0.08 to 0.19 mm (0. 14 mm on average), whereas the height for the maximum UNI height of a 90 tow in the rayon composites ranged from 0.08 to tows may be the cause for iage). The presence of thicker 90o 0. 16 mm(-0.12 mm on aver he lower matrix tunnel-cracking 占3E UNI resses in the XPLY region of the T300 and ZMI composites Note that for random lay-up architectures of 2D woven com- posites, tunnel cracking can also occur at relatively low mini 0 natrix stress(<20 MPa, see Fig. 6). For these 2D architecture contrast to the 3D architectures. there are often two contact Stress x Tow height 2, MPa-m142 a ng 90 tows with a combined height of up to c0.3 mm. These Fig. 7.(a)Onset stress for TTMC versus the square root of the inverse regions would of course not be through the width of the spec- tow height of the Z-direction tow for the UNi regions of the 3D or- imen; however, they are prime sights for low-stress tunnel or micro-crack formation thogonal composites. (b) Estimated crack density for UN regions versu There exists a 20 MPa separation in mini-matrix stress be- "stress intensity. tween the T300 and ZMI composites and the similar rayon and 2D composites(Fig. 6). There was a considerable amount of which is effectively the flaw size in the matrix. Assuming this estimation in the determination of mini-matrix stress for the to be the case, Fig. 7(a) plots the onset of TTMC versus the different regions, and error is expected even in the use of inverse square root of the height of the Z-fiber tow measured o processing parameters since some variation in local composite mm from the face of the composite for both the applied com- constituent composition will occur over the entire composite posite stress and the mini-matrix stress. A linear relationship panel. Nevertheless, it is probable that real TTMC differences exists for both stresses, confirming this implication. This can be do exist for the different XPLY regions of the 3D composites taken one step further. The estimated TTMC density can be due to the thinner 90 tow height of the rayon composite re- plotted versus a" stress-intensity"(applied stress versus the ulting in smaller local areas of unbridged matrix cracks square root of the Z-fiber tow height) as shown in Fig. 7(b) At least for crack densities below 5. the distribution of matrix 5) UNI Matrix Cracking cracks for all three 3D orthogonal composites converges for the For the UNI regions, the lack of convergence of the TTMC stress-intensity" parameter very well. As described above, significant tunnel cracking for UNI re- mation in the T300 and rayon 3D composites at stresses lower gions of the zMl composite occurred prior to Tno( ig than those required for the ZMI composites. Crack formation in the UNI regions is likely related to the size of the z-direction for the uni reg cracking prior to TTMC occurred and rayon composites trix between the 0 mini-composites. This implies the possibility of a Griffith-type relationship between the onset for TTMC acking and the Z-direction mini-composite size or height IV. Discussion In this study, two factors of 3D orthogonal composites were found that clearly dictate the nature of matrix cracking:(1) the Table Ill. 3D Architecture Average Dimensions(see Fig. 1(b) Fig. 1(b) size or height of the Z-fiber tow and (2)the local architecture of the composite. How these factors affect matrix cracking can be Z-direction fiber type iscerned by evaluating their impact on the onset stress for tun nel cracking. the onset stress for TTMC. and the stress distri- ZMI 1.50(0.12)0.14(0.02)1.5(0.23 bution for TTMC. As discussed in the following. these 1.54(0.25)0.1400.03)1.44(0.40) parameters are interrelated and key to the development of phys Rayon 1.62(0.24) 0.12(0.02) 1.34(0.39) ics-based damage and life models for textile-woven ceramic Standard deviation of at least 20 measurements composites in general and melt-infiltrated SiC/SiC composites
in the T300 and ZMI composites ranged from 0.08 to 0.19 mm (B0.14 mm on average), whereas the height for the maximum height of a 901 tow in the rayon composites ranged from 0.08 to 0.16 mm (B0.12 mm on average). The presence of thicker 901 tows may be the cause for the lower matrix tunnel-cracking stresses in the XPLY region of the T300 and ZMI composites. Note that for random lay-up architectures of 2D woven composites, tunnel cracking can also occur at relatively low minimatrix stress (o20 MPa, see Fig. 6). For these 2D architectures, in contrast to the 3D architectures, there are often two contacting 901 tows with a combined height of up to B0.3 mm. These regions would of course not be through the width of the specimen; however, they are prime sights for low-stress tunnel or micro-crack formation. There exists a 20 MPa separation in mini-matrix stress between the T300 and ZMI composites and the similar rayon and 2D composites (Fig. 6). There was a considerable amount of estimation in the determination of mini-matrix stress for the different regions, and error is expected even in the use of processing parameters since some variation in local composite constituent composition will occur over the entire composite panel. Nevertheless, it is probable that real TTMC differences do exist for the different XPLY regions of the 3D composites due to the thinner 901 tow height of the rayon composite resulting in smaller local areas of unbridged matrix cracks. (5) UNI Matrix Cracking For the UNI regions, the lack of convergence of the TTMC stress distributions appears to be due to the lack of crack formation in the T300 and rayon 3D composites at stresses lower than those required for the ZMI composites. Crack formation in the UNI regions is likely related to the size of the Z-direction mini-composites, which act as flaws in the relatively dense matrix between the 01 mini-composites. This implies the possibility of a Griffith-type relationship between the onset for TTMC cracking and the Z-direction mini-composite size or height, which is effectively the flaw size in the matrix. Assuming this to be the case, Fig. 7(a) plots the onset of TTMC versus the inverse square root of the height of the Z-fiber tow measured 0.5 mm from the face of the composite for both the applied composite stress and the mini-matrix stress. A linear relationship exists for both stresses, confirming this implication. This can be taken one step further. The estimated TTMC density can be plotted versus a ‘‘stress-intensity’’ (applied stress versus the square root of the Z-fiber tow height) as shown in Fig. 7(b). At least for crack densities below B5, the distribution of matrix cracks for all three 3D orthogonal composites converges for the ‘‘stress-intensity’’ parameter very well. As described above, significant tunnel cracking for UNI regions of the ZMI composite occurred prior to TTMC (Fig. 6), presumably due to the large tow height of the ZMI Z-tow. However, little if any tunnel cracking prior to TTMC occurred for the UNI regions of T300 and rayon composites. IV. Discussion In this study, two factors of 3D orthogonal composites were found that clearly dictate the nature of matrix cracking: (1) the size or height of the Z-fiber tow and (2) the local architecture of the composite. How these factors affect matrix cracking can be discerned by evaluating their impact on the onset stress for tunnel cracking, the onset stress for TTMC, and the stress distribution for TTMC. As discussed in the following, these parameters are interrelated and key to the development of physics-based damage and life models for textile-woven ceramic composites in general and melt-infiltrated SiC/SiC composites 0 2 4 6 8 10 12 0 100 200 300 Minimatrix Stress, MPa Crack density, mm-1 ZMI UNI ZMI XPLY Rayon XPLY T300 UNI Rayon UNI 2D 7.9epcm model T300 XPLY Fig. 6. Estimated crack density plotted versus mini-matrix stress. Table III. 3D Architecture Average Dimensions (see Fig. 1(b)) Z-direction fiber type wx (mm) hx (mm) uy (mm) ZMI 1.50 (0.12)w 0.14 (0.02) 1.45 (0.23) T300 1.54 (0.25) 0.14 (0.03) 1.44 (0.40) Rayon 1.62 (0.24) 0.12 (0.02) 1.34 (0.39) w Standard deviation of at least 20 measurements. 0 50 100 150 200 0 0.5 1 1.5 2 2.5 3 (Tow height)-1/2, mm-1/2 Stress, MPa Onset Composite Stress Slope = 2.17 MPa m1/2 Onset Minimatrix Stress Slope = 1.07 MPa m1/2 0 2 4 6 8 10 12 01 3 5 2 4 67 Stress x Tow height1/2, MPa-m1/2 Estimated Crack Density, mm-1 ZMI UNI T300 UNI Rayon UNI (a) (b) Fig. 7. (a) Onset stress for TTMC versus the square root of the inverse tow height of the Z-direction tow for the UNI regions of the 3D orthogonal composites. (b) Estimated crack density for UNI regions versus ‘‘stress intensity.’’ 150 Journal of the American Ceramic Society—Morscher et al. Vol. 88, No. 1