《复合材料 Composites》课程教学资源（学习资料）第五章 陶瓷基复合材料_melt-infiltrated

Availableonlineatwww.sciencedirect.com BCIENCEODIRECT@ COMPOSITES SCIENCE AND TECHNOLOGY ELSEVIER Composites Science and Technology 64(2004)1311-1319 Stress-dependent matrix cracking in 2D woven SiC-fiber reinforced melt-infiltrated SiC matrix composites Ohio Aerospace Institute, Brookpark, OH, US.A Received 24 February 2003: received in revised form 23 October 2003: accepted 23 October 2003 Available online 23 December 2003 Abstract The matrix cracking of a variety of SiC/SiC composites has been characterized for a wide range of constituent variation. These composites were fabricated by the two-dimensional lay-up of 0/90 five-harness satin fabric consisting of Sylramic fiber tows that were then chemical vapor infiltrated(CVn) with BN, CVI with Sic, slurry infiltrated with Sic particles followed by molten infil- tration of Si. The composites varied in number of plies, the number of tows per length, thickness, and the effective-size of the tows. his resulted in composites with a fiber volume fraction in the load-bearing direction that ranged from 0. 12 to 0. 20. Matrix cracking was monitored with modal acoustic emission in order to estimate the stress-dependent distribution of matrix cracks. It was found that the general matrix crack properties of this system could be fairly well characterized by assuming that no matrix cracks orig- inated in the load-bearing fiber, interphase, chemical vapor infiltrated SiC tow-minicomposites, i.e., all matrix cracks originate in the 90 tow regions or the large unreinforced SiC-Si matrix regions. Also, it was determined that the higher fiber-count tow composites had a much narrower stress range for matrix cracking compared to the standard tow size composites. c 2003 Elsevier Ltd. All Keywords: A Ceramic matrix composites: Stress-strain behavior: B Matrix cracking: D. Acoustic 1. Introduction since matrix cracking results in the desired stress-strain non-linearity, composite toughness, and strength prop- Sic-fiber reinforced, melt-infiltrated(MI) SiC matrix erties [3]. For non-oxide composites, such as the Sic/Sic omposites are leading candidate materials for aircraft system, the presence of matrix cracks enables oxidizing and land-based turbine engine applications such as a environments to diffuse into the interior of the composite combustor liner [1, 2]. However, for such materials to be and cause strength-degradation, especially at intermedi sed, the stress-strain behavior of these materials needs ate temperatures [4, 5]. In addition, for some BN inter to be well characterized for composites with a wide range phase composites, the degree of strength-degradation at of physical characteristics, e.g., thickness, fiber archi- intermediate temperatures is related to the number of tecture, fiber volume fraction, etc. since composite matrix cracks [6]. Therefore, it is essential that a good structures are not necessarily simple shapes but consist of understanding of the stress-dependent matrix crack thickness changes, curvature, and attachment schemes properties of a viable composite system be well-charac depending on the need of the component. In order to terized. For example, this has been done to a large extent nodel the stress-response of these materials a good un- for the Nicalon,C infil- derstanding of their matrix crack properties are required trated(CVI) Sic matrix system [7-101 a considerable amount of has occurred for the Sic fiber-reinforced. bn inter- NASA Glenn Research Center, Ohio Space phase, MI matrix system [11-13]. This development has MS-106-5. Cleveland, OH 44135 USA. Tel: included studying different fiber-types, interphases, and +1216-433-5544. gory. n. morscher(@grc. nasa. gov (G.N. Mor- Nippon Carbon Co., Tokyo, Japa 0266-3538/S- see front matter 2003 Elsevier Ltd. All rights reserved. doi: 10. 1016/j. compscitech 2003 10.02

Stress-dependent matrix cracking in 2D woven SiC-fiber reinforced melt-infiltrated SiC matrix composites Gregory N. Morscher * Ohio Aerospace Institute, Brookpark, OH, USA Received 24 February 2003; received in revised form 23 October 2003; accepted 23 October 2003 Available online 23 December 2003 Abstract The matrix cracking of a variety of SiC/SiC composites has been characterized for a wide range of constituent variation. These composites were fabricated by the two-dimensional lay-up of 0/90 five-harness satin fabric consisting of Sylramic fiber tows that were then chemical vapor infiltrated (CVI) with BN, CVI with SiC, slurry infiltrated with SiC particles followed by molten infil￾tration of Si. The composites varied in number of plies, the number of tows per length, thickness, and the effective-size of the tows. This resulted in composites with a fiber volume fraction in the load-bearing direction that ranged from 0.12 to 0.20. Matrix cracking was monitored with modal acoustic emission in order to estimate the stress-dependent distribution of matrix cracks. It was found that the general matrix crack properties of this system could be fairly well characterized by assuming that no matrix cracks orig￾inated in the load-bearing fiber, interphase, chemical vapor infiltrated SiC tow-minicomposites, i.e., all matrix cracks originate in the 90
tow regions or the large unreinforced SiC–Si matrix regions. Also, it was determined that the higher fiber-count tow composites had a much narrower stress range for matrix cracking compared to the standard tow size composites. 2003 Elsevier Ltd. All rights reserved. Keywords: A. Ceramic matrix composites; Stress–strain behavior; B. Matrix cracking; D. Acoustic emission 1. Introduction SiC-fiber reinforced, melt-infiltrated (MI) SiC matrix composites are leading candidate materials for aircraft and land-based turbine engine applications such as a combustor liner [1,2]. However, for such materials to be used, the stress–strain behavior of these materials needs to be well characterized for composites with a wide range of physical characteristics, e.g., thickness, fiber archi￾tecture, fiber volume fraction, etc. since composite structures are not necessarily simple shapes but consist of thickness changes, curvature, and attachment schemes depending on the need of the component. In order to model the stress-response of these materials a good un￾derstanding of their matrix crack properties are required since matrix cracking results in the desired stress–strain non-linearity, composite toughness, and strength prop￾erties [3]. For non-oxide composites, such as the SiC/SiC system, the presence of matrix cracks enables oxidizing environments to diffuse into the interior of the composite and cause strength-degradation, especially at intermedi￾ate temperatures [4,5]. In addition, for some BN inter￾phase composites, the degree of strength-degradation at intermediate temperatures is related to the number of matrix cracks [6]. Therefore, it is essential that a good understanding of the stress-dependent matrix crack properties of a viable composite system be well-charac￾terized. For example, this has been done to a large extent for the NicalonTM, 1 C interphase, chemical vapor infil￾trated (CVI) SiC matrix system [7–10]. A considerable amount of composite development has occurred for the SiC fiber-reinforced, BN inter￾phase, MI matrix system [11–13]. This development has included studying different fiber-types, interphases, and * Present address: NASA Glenn Research Center, Ohio Space Institute, Lewis Field, MS-106-5, Cleveland, OH 44135, USA. Tel.: +1-216-433-5512; fax: +1-216-433-5544. E-mail address: gregory.n.morscher@grc.nasa.gov (G.N. Mor￾scher). 1 Nippon Carbon Co., Tokyo, Japan. 0266-3538/$ - see front matter 2003 Elsevier Ltd. All rights reserved. doi:10.1016/j.compscitech.2003.10.022 Composites Science and Technology 64 (2004) 1311–1319 www.elsevier.com/locate/compscitech COMPOSITES SCIENCE AND TECHNOLOGY

G N. Morscher Composites Science and Technology 64 (2004)1311-1319 matrix compositions in order to maximize composite properties. Most of the development composite panels E荡 ave been processed with the sylramic- or the newer Sylramic-iBN fiber-type. The latter offers the strongest most creep-resistant, and most reliable composites to date. A variety of composite specimens were tested in this study from those developmental panels with a wide 语35 variation of the 2D woven, five-harness satin(5HS)ar- 寸R≥9g85g chitecture, e.g., changes in composite thickness, number of plies, number of tows per length, and the number of fibers per woven tow. An earlier work [14] compared the effects of these changes on the stress-strain curve in general. This study will concentrate on characterizing 甚ssss the effect these changes have on matrix cracking toward the development of a general relationship that can be used for purposes of component design and perfor- nance modeling 2. Experimental 6,55864,日 Unload-reload tensile hysteresis tests were performed R商送因后后 on ten different composite specimens, each from a dif ferent SiC/SiC composite panel, that varied in fiber tow ends per unit length, number of plies, and composite thickness. Two different fiber-types were used, Sylramic 上过过 Syl) and in situ-BN Sylramic(Syl-iBN) which is a modified Sylramic fiber that has been heat-treated [13] 兰 to form an in situ BN layer (iBN)on the fiber surface prior to composite fabrication. However, the differences 338 in the fiber-types are considered to not affect matrix cracking behavior. Both have 800 fibers per standard 当限喜 tow with an average fiber diameter of 10 um. For one composite panel, 041, two standard 800 fiber-count tows were woven together at 3.9 tow ends per cm(epcm)[15] In effect this was like having a woven tow with twice as many fibers, i.e., 1600, as the standard tow. However, 041 had the same fraction of fibers in the loading 号期可 direction as a single-tow weave of 7.9 epcm(011 in Table 1). Optical micrographs of polished longitudinal sections of the oll and the 041 composites are shown in ∞∞∞g必∞|s Fig. 1. It is evident that the double-tow woven com- posite essentially doubles the width dimension of the effective woven tow. The composites were processed by 5≥ the former Honeywell Advanced Composites(Newark, DE), currently known as General Electric Power Sys- [2]. Table 1 lists the consti ations for the composites tested. Note that there was considerable variation in fiber volume fraction and cvi 百 SiC volume fraction between the panels tested. Composite processing entails first stacking of bal anced 0/90 five-harness fabric woven from SYL or SYL iBN tows, a Bn interphase layer deposition (0.5 um) via CVi, a Sic interphase over-coating via CVI, Sic 自 3旨象 2 Dow Corning Corporation, Midland, MI

matrix compositions in order to maximize composite properties. Most of the development composite panels have been processed with the Sylramic 2 or the newer Sylramic-iBN fiber-type. The latter offers the strongest, most creep-resistant, and most reliable composites to date. A variety of composite specimens were tested in this study from those developmental panels with a wide variation of the 2D woven, five-harness satin (5HS) ar￾chitecture, e.g., changes in composite thickness, number of plies, number of tows per length, and the number of fibers per woven tow. An earlier work [14] compared the effects of these changes on the stress–strain curve in general. This study will concentrate on characterizing the effect these changes have on matrix cracking towards the development of a general relationship that can be used for purposes of component design and perfor￾mance modeling. 2. Experimental Unload–reload tensile hysteresis tests were performed on ten different composite specimens, each from a dif￾ferent SiC/SiC composite panel, that varied in fiber tow ends per unit length, number of plies, and composite thickness. Two different fiber-types were used, Sylramic (Syl) and in situ-BN Sylramic (Syl-iBN) which is a modified Sylramic fiber that has been heat-treated [13] to form an in situ BN layer (iBN) on the fiber surface prior to composite fabrication. However, the differences in the fiber-types are considered to not affect matrix￾cracking behavior. Both have 800 fibers per standard tow with an average fiber diameter of 10 lm. For one composite panel, 041, two standard 800 fiber-count tows were woven together at 3.9 tow ends per cm (epcm) [15]. In effect this was like having a woven tow with twice as many fibers, i.e., 1600, as the standard tow. However, 041 had the same fraction of fibers in the loading direction as a single-tow weave of 7.9 epcm (011 in Table 1). Optical micrographs of polished longitudinal sections of the 011 and the 041 composites are shown in Fig. 1. It is evident that the double-tow woven com￾posite essentially doubles the width dimension of the effective woven tow. The composites were processed by the former Honeywell Advanced Composites (Newark, DE), currently known as General Electric Power Sys￾tems Composites [2]. Table 1 lists the constituent vari￾ations for the composites tested. Note that there was considerable variation in fiber volume fraction and CVI SiC volume fraction between the panels tested. Composite processing entails first stacking of bal￾anced 0/90 five-harness fabric woven from SYL or SYL￾iBN tows, a BN interphase layer deposition (
0.5 lm) via CVI, a SiC interphase over-coating via CVI, SiC Table 1 Properties of tested SiC/SiC specimens Specimen (type of debondinga) Sylramic fiber typeb Tow ends per cm No. of plies Specimen thickness (mm) Crack density after specimen failure, #/mm Fiber fractionc E (GPa) rth (MPa) BN fractionc CVI SiC fractionc fmini Emini Single-tow woven composites 007 (OD) AP 4.9 8 1.63 8.0 0.15 219 )35 0.04 0.11 0.30 357 016 (ID) iBN 4.9 8 2.04 11.6 0.12 279 )35 0.03 0.23 0.38 386 017 (ID) iBN 4.9 8 1.46 12.0 0.17 224 )53 0.04 0.14 0.35 366 012 (ID) iBN 7.1 6 1.99 7.2 0.14 289 )37 0.03 0.20 0.36 379 009 (MD) AP 7.9 8 2.18 9.0 0.18 246 )55 0.04 0.11 0.33 358 011 (MD) iBN 7.9 8 2.04 10.3 0.19 228 )57 0.04 0.12 0.34 361 018 (ID) iBN 7.9 4 1.37 5.0 0.14 235 )53 0.03 0.12 0.29 364 044 (OD) iBN 8.7 8 2.14 9.5 0.20 216 )35 0.03 0.13 0.35 369 068 (ID) IBN 8.7 8 2.21 10.4 0.20 277 )67 0.04 0.15 0.39 366 Double-tow woven composite 041 (OD) IBN 3.9(2)d 8 2.07 9.0 0.19 197 )50 0.03 0.11 0.33 363 a ID, inside debonding; OD, outside debonding; MD, mixed debonding. b AP, as-produced; iBN, in situ BN. Each fiber tow consisted of 800 fibers. The average fiber diameter was 10 lm. c Volume fraction of constituent in the load-bearing direction, the total fraction would be double this amount. dTwo tows were woven together into a 3.9 epcm fabric. In effect there were the same number of 800 count tows per length as the 7.9 epcm 011 specimen; however, the effective tow size for 041 was 1600 fibers. 2 Dow Corning Corporation, Midland, MI. 1312 G.N. Morscher / Composites Science and Technology 64 (2004) 1311–1319

G N. Morscher Composites Science and Technology 64(2004)1311-1319 at 500 w for 30 min. The etchant reacts with the free si in the matrix, removing much of it, making it impossible to observe cracks in 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 3. Results and analysis 3. 1. Standard single-tow woven composites Monotonic and unload-reload stress strain data with E activity plotted as energy are shown in Fig. 2 for two different specimens from the same panel. Several aspects of Fig. 2 are characteristic of the Sylramic/BN/MI SiC system For specimens from the same panel of material, the stress-strain properties are very consistent, i.e., little scatter from specimen to specimen and little difference 器m for monotonic and unload-reload experiments. Also the ae activity is very consistent and occurs over a Fig. I Polished longitudinal sections of standard tow woven (olD) range of stress(strain). Finally, the matrix is in residual composite and double-tow woven(041)composite. compression, which is indicative of the intersection of the intercepts of the average slope of the top portion of particulate infiltration via slurry-infiltration, and finally the hysteresis loop in the positive stress-strain quadrant liquid Si infiltration [1, 2] according to Steen and Valles [18] The tensile tests were performed on 150 mm long pecimens with a contoured gage section (dog-bone 2.5mm width in grip regid idth in gage section) using a universal-testing machine (Instron Model 8562, Instron, Ltd, Canton Mass. with an elec SYL-iBN tromechanical actuator. Glass fiber reinforced epoxy 8.epcm: 8 ply: f=0.2 12 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. Tensile tests were performed in load- control at 2 kN/min Modal acoustic emission(AE)was monitored during the tensile tests with two wide -band. 50 kHz to 2.0 MHz 0050.10.150202503035 high fidelity sensors placed just outside the tapered re- gion of the dog-bone specimen. Vacuum grease was used as a couplant and mechanical clips were used to mount (b) 1 the sensors to the specimen. The AE waveforms were recorded and digitized using a 4-channel, Fracture Wave Detector(FWD) produced by Digital Wave Corpora tion(Englewood, CO). The load and strain were also E0.5 recorded. After the tensile test, the ae data was filtered 3 04 using the location software from the fwd manufac- 0.3 turer in order to separate out the ae that occurred outside of the gage section. For more information on the AE procedure and analysis, see [16, 17] Since residual compressive stresses in the matrix close dense the matrix cracks, to measure crack density, sections of the tested tensile specimens in the gage section at least Fig. 2. Typical (068)monotonic and load-unload 10 mm long were polished and then plasma(CF4)etched stress-strain behavior and(b)stress-dependent AE activity

particulate infiltration via slurry-infiltration, and finally, liquid Si infiltration [1,2]. The tensile tests were performed on 150 mm long specimens with a contoured gage section (dog-bone, 12.5 mm width in grip region and 10 mm width in gage section) using a universal-testing machine (Instron Model 8562, Instron, Ltd, Canton Mass.) with an elec￾tromechanical actuator. 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. Tensile tests were performed in load￾control at 2 kN/min. Modal acoustic emission (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 re￾gion of the dog-bone specimen. 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 Corpora￾tion (Englewood, CO). The load and strain were also recorded. After the tensile test, the AE data was filtered using the location software from the FWD manufac￾turer in order to separate out the AE that occurred outside of the gage section. For more information on the AE procedure and analysis, see [16,17]. Since residual compressive stresses in the matrix close the matrix cracks, to measure crack density, sections of the tested tensile specimens in the gage section at least 10 mm long were polished and then plasma (CF4) etched at 500 W for 30 min. The etchant reacts with the free Si in the matrix, removing much of it, making it impossible to observe cracks in 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. 3. Results and analysis 3.1. Standard single-tow woven composites Monotonic and unload–reload stress strain data with AE activity plotted as energy are shown in Fig. 2 for two different specimens from the same panel. Several aspects of Fig. 2 are characteristic of the Sylramic/BN/MI SiC system. For specimens from the same panel of material, the stress–strain properties are very consistent, i.e., little scatter from specimen to specimen and little difference for monotonic and unload–reload experiments. Also, the AE activity is very consistent and occurs over a range of stress (strain). Finally, the matrix is in residual compression, which is indicative of the intersection of the intercepts of the average slope of the top portion of the hysteresis loop in the positive stress–strain quadrant, according to Steen and Valles [18]. Fig. 1. Polished longitudinal sections of standard tow woven (011) composite and double-tow woven (041) composite. Fig. 2. Typical (068) monotonic and load–unload–reload hysteresis (a) stress–strain behavior and (b) stress-dependent AE activity. G.N. Morscher / Composites Science and Technology 64 (2004) 1311–1319 1313

G N. Morscher Composites Science and Technology 64 (2004)1311-1319 500 0+1 3.9(epcm; f 3501060:8pcm:t= 250 .epcm: f= 6:4epcm;f=0.12 50 0.5 0.6 0.7 Strain. Fig. 3. Stress-strain curves for the specimens tested and analyzed in his study. Note, unload-reload hysteresis loops have been removed for clarity Fig 4. Example of matrix cracks observed along a polished and etched longitudinal section of the tensile specimen(017) after failure at room temperatur Fig. 3 shows stress-strain curves. unload-reload loops removed, for a variety of different architecture composites,i.e, different panels. As is expected, the T12 posites tend to exhibit higher 649epcm;向=0.12 action com ultimate strengths, higher stresses for the"knee"in the 04t;3.92)epcm;017:49 epcm, I stress-strain curve, and steeper secondary slopes [14] There was considerable scatter in elastic moduli. some of 88068:8.epcm: 011;7.9 epcm: fe019 which was due to the anomalies described below The matrix crack density was determined fo 007: 4.9opcm: fat 012:7.1 eocm: f=0.13 of matrix cracking in these composites for a portion a p/ specimen after tensile testing. Fig 4 shows an example oeE pecimen cut and polished from the gage section fol lowed by a plasma-etch. No matrix cracks are visible, because of the residual compression in the matrix and 00050.10.15020.25 0.35 the higher interfacial shear stresses of composites with Strain. nic fibers, without plasma etching The energy of aE has shown to be a good measure of the relative crack density when the more accurate 017:4.9epcm0.17 "modal"AE approach is used for these types of com- 8.epcm, fa posite systems[17]. In other words, the relative amount 2 041:3.92pcm of cumulative AE energy is nearly directly related to the g relative number of matrix cracks formed. Therefore the 007:49epmo final matrix crack density measured from the composite test specimens was multiplied by the normalized cumu- lative AE energy(e.g, Fig. 2(b))for each specimen in 018:79epcm台=0.14 order to estimate the stress-dependent matrix crack distribution. the estimated crack distributions are shown in Fig. 5 for a number of specimens versus strain and stress. The strain and stress distribution for matrix 150200250300350 cracking varies from specimen to specimen considerably. ter to define the earliest formation of Fig. 5. Estimated matrix cracking(normalized AE measured during large matrix cracks is theonset"strain or stress at strain and (b)stress for standard single-tow woven composites and a which the rate of AE rapidly increases. The sudden double-tow woven composite(041). crease in AE activity is due to high-energy events that are associated with large bridged matrix cracks that with another matrix crack) of the specimen [16, 17] propagate through-the-thickness(or at least a significant Fig. 2(a) and(b) show the determination of Eonset and portion of the cross-section if a matrix crack links up onset from extrapolation of the initial high-rate AE

Fig. 3 shows stress–strain curves, unload–reload loops removed, for a variety of different architecture composites, i.e., different panels. As is expected, the higher volume fraction composites tend to exhibit higher ultimate strengths, higher stresses for the ‘‘knee’’ in the stress–strain curve, and steeper secondary slopes [14]. There was considerable scatter in elastic moduli, some of which was due to the anomalies described below. The matrix crack density was determined for each specimen after tensile testing. Fig. 4 shows an example of matrix cracking in these composites for a portion of a specimen cut and polished from the gage section fol￾lowed by a plasma-etch. No matrix cracks are visible, because of the residual compression in the matrix and the higher interfacial shear stresses of composites with Sylramic fibers, without plasma etching. The energy of AE has shown to be a good measure of the relative crack density when the more accurate ‘‘modal’’ AE approach is used for these types of com￾posite systems [17]. In other words, the relative amount of cumulative AE energy is nearly directly related to the relative number of matrix cracks formed. Therefore, the final matrix crack density measured from the composite test specimens was multiplied by the normalized cumu￾lative AE energy (e.g., Fig. 2(b)) for each specimen in order to estimate the stress-dependent matrix crack distribution. The estimated crack distributions are shown in Fig. 5 for a number of specimens versus strain and stress. The strain and stress distribution for matrix cracking varies from specimen to specimen considerably. One useful parameter to define the earliest formation of large matrix cracks is the ‘‘onset’’ strain or stress at which the rate of AE rapidly increases. The sudden in￾crease in AE activity is due to high-energy events that are associated with large bridged matrix cracks that propagate through-the-thickness (or at least a significant portion of the cross-section if a matrix crack links up with another matrix crack) of the specimen [16,17]. Fig. 2(a) and (b) show the determination of eonset and ronset from extrapolation of the initial high-rate AE Fig. 3. Stress–strain curves for the specimens tested and analyzed in this study. Note, unload–reload hysteresis loops have been removed for clarity. Fig. 4. Example of matrix cracks observed along a polished and etched longitudinal section of the tensile specimen (017) after failure at room temperature. Fig. 5. Estimated matrix cracking (normalized AE measured during the stress–strain test multiplied by measured crack density) versus (a) strain and (b) stress for standard single-tow woven composites and a double-tow woven composite (041). 1314 G.N. Morscher / Composites Science and Technology 64 (2004) 1311–1319

G N. Morscher Composites Science and Technology 64(2004)1311-1319 0.12 01· onset stra typically had lower elastic modulus values, M220+ we 0.1 Onset stress 口·g044+200 20 MPa, compared to the 250+ 30 MPa measured for ID composites. Note that the two composites with the higher stress-distributions for matrix crack ing were OD(044)and MD(Ol1) specimens (2)The matrix crack density varied from 5 to 12 cracks/ mm and did not appear to show any correspondence with the fiber volume fraction or interfacial shear strength(Table 1) 0.120.140.16 (3)Many of the composites did not fully saturate with matrix cracks. Specifically, those composites with Fig. 6. Eonset and onset versus fraction of fibers in the loading direction lower fiber volume fractions where the rate of ae activity remained high until failure(e. g, 007, 012, activity to the abscissa for strain and stress, respectively 016, and 018 in Fig. 5). Matrix crack saturation oc- The initial low energy ae prior to the AE energy in curs when the rate of cumulative Ae energy dimin crease corresponds to the formation of microcracks or ishes to near zero(slope of Fig. 2(b) plateaus at tunnel cracks that form in the 90 bundles but do not higher stress). This is evident in Fig. 5 for 00 penetrate, at least not very much, into the load-bearing 011, 068, 044, and probably 017 fibers. Eonset and onset was determined for all the speci- men data in Fig. 5. In general, Eonset and onset increase 3. 2. Normalizing matrix cracking behavior for standard with fiber volume fraction in the loading direction single-tow woren composites (Fig. 6). However, composites with the same fo but higher E have lower Eonset. Two examples are shown in It is evident that there is some relationship between Fig. 6 for specimens with a fo=0. 2(068 and 044)and fiber volume fraction and stress-distribution for matrix specimens with a fo=0.14(012 and 018). For both cracking and it would be useful to characterize matrix examples, onset was very similar but Eonset was less for cracking based on the constituent properties of the the higher E material. The stress-dependent crack den- composites and the matrix in particular. Matrix cracks sity distribution in general follows the same trend as originate within the 90 tow-minicomposites or large, Conset and occurs over a higher stress range for higher unreinforced matrix regions and then propagate fiber volume fraction through the load-bearing minicomposites [7, 8, 20). In Even though the composite panels were fabricated by other words, matrix cracks do not originate for these the same vendor over a relatively short period of time, l materials in the load-bearing fiber, interphase, CVI SIC year, there were several anomalies observed for the tow-minicomposite. If the volume fraction and elastic composites tested in this study modulus of an average""minicomposite"can be deter (1) Two types of interfacial debonding and sliding be- mined from the processing data, then the average stress haviors were present in the data set for this study in the matrix region excluding the bn and CVI SiC in and noted in Table 1. Most of the specimens exhib- the load-bearing minicomposite could be backed out ited debonding between the fiber and the BN-inter- from a rule-of-mixtures approach and a relationshi phase, this was referred to as""inside debonding between matrix cracking and matrix stress can be (ID). Two specimens exhibited debonding between established the BN-interphase and the Cvi SiC portion of the The fraction of load-bearing minicomposites, mini matrix, this was referred to"outside debonding was estimated from half of the combined fraction of fi (OD). Also, two specimens exhibited a mixture of in- ber, BN, and CVI SiC determined from the processing side and outside debonding, i. e, "mixed debonding" data sheet supplied by the composite fabricator. The (MD). This was pointed out in[14] and has been de- elastic modulus of the minicomposites, Emini, was esti scribed in greater detail in a recent paper [19]. od mated via the rule-of-mixtures from the elastic moduli of composites have much lower interfacial shear stres- each constituent of the minicomposite (Er=380 GPa, ses, 10 MPa, compared to the 70+10 MPa mea- EBN=60 GPa, and ECVLSiC 425 GPa)and the frac- sured for ID composites. Also, OD composites tion of each constituent in the loading direction. Again, a rule-of-mixtures approach can be used to"back-out the stress in the"minimatrix'"surrounding the load-bearing The interfacial shear stress. t [12, 19]:(1)fiber"push-in, a direct measure of t, and (2)best fitting the stress-strain curve for t using Eqs. (3H5)(below) and assuming the matrix crack distribution from AE energy and final matrix crack The raw data from the manufacturer is base density, an indirect measure. Both methods produced consistent each processing step. The constituent fractions determined from results processing data are tabulated in Table I for each panel

activity to the abscissa for strain and stress, respectively. The initial low energy AE prior to the AE energy in￾crease corresponds to the formation of microcracks or tunnel cracks that form in the 90
bundles but do not penetrate, at least not very much, into the load-bearing fibers. eonset and ronset was determined for all the speci￾men data in Fig. 5. In general, eonset and ronset increase with fiber volume fraction in the loading direction (Fig. 6). However, composites with the same f0 but higher E have lower eonset. Two examples are shown in Fig. 6 for specimens with a f0 ¼ 0:2 (068 and 044) and specimens with a f0 ¼ 0:14 (012 and 018). For both examples, ronset was very similar but eonset was less for the higher E material. The stress-dependent crack den￾sity distribution in general follows the same trend as ronset and occurs over a higher stress range for higher fiber volume fraction composites. Even though the composite panels were fabricated by the same vendor over a relatively short period of time,
1 year, there were several anomalies observed for the composites tested in this study: (1) Two types of interfacial debonding and sliding be￾haviors were present in the data set for this study and noted in Table 1. Most of the specimens exhib￾ited debonding between the fiber and the BN-inter￾phase, this was referred to as ‘‘inside debonding’’ (ID). Two specimens exhibited debonding between the BN-interphase and the CVI SiC portion of the matrix, this was referred to ‘‘outside debonding’’ (OD). Also, two specimens exhibited a mixture of in￾side and outside debonding, i.e., ‘‘mixed debonding’’ (MD). This was pointed out in [14] and has been de￾scribed in greater detail in a recent paper [19]. OD composites have much lower interfacial shear stres￾ses,
10 MPa, compared to the 70 10 MPa mea￾sured for ID composites. 3 Also, OD composites typically had lower elastic modulus values,
220 20 MPa, compared to the 250 30 MPa measured for ID composites. Note that the two composites with the higher stress-distributions for matrix crack￾ing were OD (044) and MD (011) specimens. (2) The matrix crack density varied from 5 to 12 cracks/ mm and did not appear to show any correspondence with the fiber volume fraction or interfacial shear strength (Table 1). (3) Many of the composites did not fully saturate with matrix cracks. Specifically, those composites with lower fiber volume fractions where the rate of AE activity remained high until failure (e.g., 007, 012, 016, and 018 in Fig. 5). Matrix crack saturation oc￾curs when the rate of cumulative AE energy dimin￾ishes to near zero (slope of Fig. 2(b) plateaus at higher stress). This is evident in Fig. 5 for 009, 011, 068, 044, and probably 017. 3.2. Normalizing matrix cracking behavior for standard single-tow woven composites It is evident that there is some relationship between fiber volume fraction and stress-distribution for matrix cracking and it would be useful to characterize matrix cracking based on the constituent properties of the composites and the matrix in particular. Matrix cracks originate within the 90
tow-minicomposites or large, unreinforced matrix regions and then propagate through the load-bearing minicomposites [7,8,20]. In other words, matrix cracks do not originate for these materials in the load-bearing fiber, interphase, CVI SiC ‘‘tow-minicomposite’’. If the volume fraction and elastic modulus of an average ‘‘minicomposite’’ can be deter￾mined from the processing data, then the average stress in the matrix region excluding the BN and CVI SiC in the load-bearing minicomposite could be backed out from a rule-of-mixtures approach and a relationship between matrix cracking and matrix stress can be established. The fraction of load-bearing minicomposites, fmini, was estimated from half of the combined fraction of fi- ber, BN, and CVI SiC determined from the processing data sheet supplied by the composite fabricator. 4 The elastic modulus of the minicomposites, Emini, was esti￾mated via the rule-of-mixtures from the elastic moduli of each constituent of the minicomposite (Ef ¼ 380 GPa, EBN ¼ 60 GPa, and ECVI–SiC ¼ 425 GPa) and the frac￾tion of each constituent in the loading direction. Again, a rule-of-mixtures approach can be used to ‘‘back-out’’ the stress in the ‘‘minimatrix’’ surrounding the load-bearing Fig. 6. eonset and ronset versus fraction of fibers in the loading direction. 3 The interfacial shear stress, s, was determined by two methods in [12,19]: (1) fiber ‘‘push-in’’, a direct measure of s, and (2) best fitting the stress–strain curve for s using Eqs. (3)–(5) (below) and assuming the matrix crack distribution from AE energy and final matrix crack density, an indirect measure. Both methods produced consistent results. 4 The raw data from the manufacturer is based on weight gains after each processing step. The constituent fractions determined from the processing data are tabulated in Table 1 for each panel. G.N. Morscher / Composites Science and Technology 64 (2004) 1311–1319 1315

G N. Morscher Composites Science and Technology 64 (2004)1311-1319 minicomposites based on processing data for each 450 composite pane 68:8.7epcm;f=02 (σ+) where oc is the composite stress, oth is the residual compressive stress in the matrix [21] which was found to be higher, in general, for higher volume fraction com- 嘉∞0 posites with inside debonding [14], and Ec is the mea sured composite elastic modulus from the a/E curve 100 012:71epcm;f=0.14 Fig. 7 shows the estimated stress-dependent matrix crack distribution versus ominimatrix. There is consider- able convergence of the matrix crack data with this 00.050.10.150.20.250.3035 normalization step even though there is a wide scatter in Strain, e and the differences in debonding and sliding character, Fig 8. Stress-strain predictions based on best-fit matrix crac i.e., interfacial shear stress. However, for specimens with for two different volume fraction composites with same t The solid lines are the measure 二N the fewer number of plies, the matrix crack distribution appears to be broader in o and ominimatrix, but still within right of the squares are the model predictions assuming 33%fewer and he range of matrix cracking exhibited by the standard 33% greater number of matrix cracks, respectively eight ply panels. Also, there appears to be a small sep- aration in gminimatrix, N10 MPa, between the lowest (4.9 epcm composites and those woven with higher (7.1-8.7) term in Eq (1). Finally, the data were compared using epcm. The matrix crack distribution for the lower epi only fo and no residual stress term. None of these at composites occurring at lower ominimatris range but with tempts resulted in the convergence of the data as well as about the same relative slopes as higher volume fraction the"minimatrix""approach composites. This perhaps does indicate an effect of ar chitecture on the nature of matrix cracking in these 3.3. Matrix cracking in double-tow woven composites composites; however, for the use of these materials in actual components, composite architecture will most The estimated crack density versus composite stress likely use the higher epcm ranges and higher fiber fr and ominimatrix for the composite specimen woven with a tions where the behavior was fairly consistent double-tow(specimen 041 in Table 1)are also shown in Other attempts were made to relate the crack-distri Figs. 5 and 7, respectively. Matrix cracking occurs over butions as well. The data were compared only using the a significantly narrower stress-range for the double-tow volume fraction of fibers in the loading direction, fo, and woven composites compared to the standard single-tow Er in Eq()instead of mini and Emini, respectively. The woven composites and at lower stresses when comparing data were compared without the use of a residual stress similar volume fraction composites(Fig. 5). This also is evident in the ominimatrix range of matrix cracking(Fig 8) where the matrix cracking in the double-tow composites occurs at significantly lower stresses. It should be noted 017:49 cpam, f=0.17 hat onset for 041 was significantly less than 01l which DaB: 8, epcm: f=D 01649pm12 has the same fo for a standard single-tow woven com posite but it had a similar Eonset(Fig. 6) 044:87epcm向=0.2 009:790pm018 4. Discussio 018;79epcm0.14 To model the stress-strain response of ceramic matrix composites, knowledge of the stress-dependent number of matrix cracks is essential [20-23]. Usually this can only be determined experimentally for the composite Mini Matrix Stress. MPa system studied. For example, Lamon and coworkers [9, 10] determined the distribution of matrix cracks in Fig. 7. Stress-dependent matrix cracking versus ominimatrix for standard tow woven composites and a double-tow woven com different parts of the 2D architecture, i.e, those ema 041). A simple best-fit of matrix crack density, up to mimimatrix= 150 nating from 90 tows or those emanating in minicom- MPa, for single-tow woven composites with epcm >7.1 is plotted as posite ligaments not adjacent to 90 bundles, as a open square function of strain for a CvI SiC matrix system. For the

minicomposites based on processing data for each composite panel: rminimatrix ¼ ð Þ rc þ rth Ec Ec fminiEmini 1 fmini
; ð1Þ where rc is the composite stress, rth is the residual compressive stress in the matrix [21] which was found to be higher, in general, for higher volume fraction com￾posites with inside debonding [14], and Ec is the mea￾sured composite elastic modulus from the r=e curve. Fig. 7 shows the estimated stress-dependent matrix crack distribution versus rminimatrix. There is consider￾able convergence of the matrix crack data with this normalization step even though there is a wide scatter in E and the differences in debonding and sliding character, i.e., interfacial shear stress. However, for specimens with the fewer number of plies, the matrix crack distribution appears to be broader in r and rminimatrix, but still within the range of matrix cracking exhibited by the standard eight ply panels. Also, there appears to be a small sep￾aration in rminimatrix,
10 MPa, between the lowest (4.9) epcm composites and those woven with higher (7.1–8.7) epcm. The matrix crack distribution for the lower epi composites occurring at lower rminimatrix range but with about the same relative slopes as higher volume fraction composites. This perhaps does indicate an effect of ar￾chitecture on the nature of matrix cracking in these composites; however, for the use of these materials in actual components, composite architecture will most likely use the higher epcm ranges and higher fiber frac￾tions where the behavior was fairly consistent. Other attempts were made to relate the crack-distri￾butions as well. The data were compared only using the volume fraction of fibers in the loading direction, f0, and Ef in Eq. (1) instead of fmini and Emini, respectively. The data were compared without the use of a residual stress term in Eq. (1). Finally, the data were compared using only f0 and no residual stress term. None of these at￾tempts resulted in the convergence of the data as well as the ‘‘minimatrix’’ approach. 3.3. Matrix cracking in double-tow woven composites The estimated crack density versus composite stress and rminimatrix for the composite specimen woven with a double-tow (specimen 041 in Table 1) are also shown in Figs. 5 and 7, respectively. Matrix cracking occurs over a significantly narrower stress-range for the double-tow woven composites compared to the standard single-tow woven composites and at lower stresses when comparing similar volume fraction composites (Fig. 5). This also is evident in the rminimatrix range of matrix cracking (Fig. 8) where the matrix cracking in the double-tow composites occurs at significantly lower stresses. It should be noted that ronset for 041 was significantly less than 011 which has the same f0 for a standard single-tow woven com￾posite but it had a similar eonset (Fig. 6). 4. Discussion To model the stress–strain response of ceramic matrix composites, knowledge of the stress-dependent number of matrix cracks is essential [20–23]. Usually this can only be determined experimentally for the composite system studied. For example, Lamon and coworkers [9,10] determined the distribution of matrix cracks in different parts of the 2D architecture, i.e., those ema￾nating from 90
tows or those emanating in minicom￾posite ligaments not adjacent to 90
bundles, as a function of strain for a CVI SiC matrix system. For the Fig. 8. Stress–strain predictions based on best-fit matrix crack density for two different volume fraction composites with same s ¼ 70 MPa. The solid lines are the measured stress–strain curves. The squares are the predicted stress–strain curves and the dashed lines to the left and right of the squares are the model predictions assuming 33% fewer and 33% greater number of matrix cracks, respectively. Fig. 7. Stress-dependent matrix cracking versus rminimatrix for standard single-tow woven composites and a double-tow woven composite (041). A simple best-fit of matrix crack density, up to rminimatrix ¼ 150 MPa, for single-tow woven composites with epcm P7.1 is plotted as open squares. 1316 G.N. Morscher / Composites Science and Technology 64 (2004) 1311–1319

G.N. Morscher / Composites Science and Technology 64(2004)1311-1319 composites tested in this study, it is evident that the strain curve [14] for a range of 2D woven architectures strain where matrix cracking occurs can vary consider- number of plies, and composite thickness. For example ably (Fig. 5(a))dependent on architecture(type of the stress-dependent matrix crack distribution for the reave, number of plies, etc. ) interfacial shear stress, and higher epi composites(>7. 1 epcm) was fit with a simple starting elastic modulus. Also, there are no minicom- linear relationship for ominimatrix above 95 MPa(Fig 8) posite ligaments independent of the rest of the structure, i.e., all load-bearing minicomposites are bonded P=0.1034 strongly through a dense matrix to the 90 tows, the where Pe is the stress-dependent matrix crack density .This primary flaw source. Also, matrix do essentially propa- data was then used to model the o/& curve over the ap gate through-the-thickness or link-up with one another propriate stress range for two specimens(012 and 068) as will be described below. Curtin et al. [21] have pro- with similar interfacial shear stress, t, of 70 MPa(Fig 8) posed a matrix cracking model based on the distribution from the approach taken in [12] based on [21,22] of flaws in the matrix, the distribution of matrix cracks E=G/E+aS(d)P/Er(a+Gth); for P-1> 28,(3) action of neighboring matrix cracks that ultimately leads where the sliding length to matrix crack saturation when the fiber sliding lengths 8=ar(o+oth)/2t extend the length of the uncracked matrix segments. An attempt was made to model the matrix crack distribu r is the fiber radius. and tion according to the Weibull model put forward by (1-f)Em/fec Curtin et al. [21]; however, it was not very robust for the ariety of composites studied here The stress-dependent matrix cracking behavior at It is evident from the analysis in this study that the room temperature can be used to model the time-de- ress-dependence for matrix cracking of 2D woven pendent elevated temperature life expectancy at stress Sylramic/BN/MI SiC composites with same number of [6]. The stress-strain curve of these composites upon fibers per tow was dependent on the stress in the ma- initial loading varies only slightly 200°C[2 trix outside of the load-bearing minicomposite, i.e, the ittle difference is observed in load-unload beh minimatrix stress, the source of flaws for matrix crack 815C, including oth, until higher stresses are reached formation in the"minimatrix'"being the 90 tows. This (>275 MPa) after considerable exposure time(30 min can be well characterized by a simple rule-of-mixtures in air at temperature during the unload-reload test relationship(Eq. (1)up to about half the saturation Therefore, it can be assumed that the matrix cracking of crack density, 5 cracks/mm. For higher fiber volume these materials upon initial loading behaves essentially fraction composites, this relates to stresses in excess of as at room temperature and the stress-depen- 250 MPa, much higher than their expected use condi- dent matrix-crack distribution can be used as the initial tions for long-time combustor applications [2]. Only a crack density at elevated temperatures upon loading few properties of the composite were required to for Evidence for this has been demonstrated for a mulate this relationship: the undamaged elastic modu- number of the panels tested in this study where speci- lus, Ec, the residual stress in the matrix, Oth, and the mens have been subjected to stress-rupture testing at properties of the load-bearing minicomposite, Eminimatrix 815C. After rupture, through-thickness cracks remain ind ominimatrix. The former two properties were found open due to the formation of solid oxidation products from a simple load-unload tensile test at room tem in the matrix crack and there is no need to etch the perature and the later two properties were estimated polished sections. The measured crack densities after from the composite processing data. If one fabricated rupture are commensurate with what would have been component out of this material, all they would require expected from the ae data and known final crack would be to fabricate a few" witness "panels of varying density measured after room temperature testing to architecture to determine the needed composite prop- failure(Fig 9). For specimens from the 007 and 044 erties and verify this stress-dependent matrix cracking composite panels, the matrix crack densities after from acoustic techniques and Possible to determine Ec stress-rupture at an intermediate stress were nearly testing witness panels would give one insight as to the AE data. For specimens from the 017 composite panel, interfacial debonding and sliding properties which the matrix crack densities after stress-rupture at an could be estimated from the stress-strain curve and intermediate stress were slightly lower by approxi stress-dependent matrix crack distribution and would mately 33% of what was to be expected from the room be necessary for modeling the stress-strain behavior temperature ae data which represents the largest de- viation from estimated crack density observed. It The understanding of the stress-dependent matrix should be noted that for the Sylramic/BN/MI system, it acking behavior can now be used to model the stress- has been shown that relatively little global AE activity

composites tested in this study, it is evident that the strain where matrix cracking occurs can vary consider￾ably (Fig. 5(a)) dependent on architecture (type of weave, number of plies, etc.), interfacial shear stress, and starting elastic modulus. Also, there are no minicom￾posite ligaments independent of the rest of the structure, i.e., all load-bearing minicomposites are bonded strongly through a dense matrix to the 90
tows, the primary flaw source. Also, matrix do essentially propa￾gate through-the-thickness or link-up with one another as will be described below. Curtin et al. [21] have pro￾posed a matrix cracking model based on the distribution of flaws in the matrix, the distribution of matrix cracks that emanate from some of those flaws, and the inter￾action of neighboring matrix cracks that ultimately leads to matrix crack saturation when the fiber sliding lengths extend the length of the uncracked matrix segments. An attempt was made to model the matrix crack distribu￾tion according to the Weibull model put forward by Curtin et al. [21]; however, it was not very robust for the variety of composites studied here. It is evident from the analysis in this study that the stress-dependence for matrix cracking of 2D woven Sylramic/BN/MI SiC composites with same number of fibers per tow was dependent on the stress in the ma￾trix outside of the load-bearing minicomposite, i.e, the minimatrix stress, the source of flaws for matrix crack formation in the ‘‘minimatrix’’ being the 90
tows. This can be well characterized by a simple rule-of-mixtures relationship (Eq. (1)) up to about half the saturation crack density,
5 cracks/mm. For higher fiber volume fraction composites, this relates to stresses in excess of 250 MPa, much higher than their expected use condi￾tions for long-time combustor applications [2]. Only a few properties of the composite were required to for￾mulate this relationship: the undamaged elastic modu￾lus, Ec, the residual stress in the matrix, rth, and the properties of the load-bearing minicomposite, Eminimatrix and rminimatrix. The former two properties were found from a simple load–unload tensile test at room tem￾perature and the later two properties were estimated from the composite processing data. If one fabricated a component out of this material, all they would require would be to fabricate a few ‘‘witness’’ panels of varying architecture to determine the needed composite prop￾erties and verify this stress-dependent matrix cracking relationship. It would also be possible to determine Ec from acoustic techniques and estimate rth. However, testing witness panels would give one insight as to the interfacial debonding and sliding properties which could be estimated from the stress–strain curve and stress-dependent matrix crack distribution and would be necessary for modeling the stress–strain behavior [14]. The understanding of the stress-dependent matrix cracking behavior can now be used to model the stress– strain curve [14] for a range of 2D woven architectures, number of plies, and composite thickness. For example, the stress-dependent matrix crack distribution for the higher epi composites (>7.1 epcm) was fit with a simple linear relationship for rminimatrix above 95 MPa (Fig. 8): qc ¼ 0:1034rminimatrix 9:8074; ð2Þ where qc is the stress-dependent matrix crack density. This data was then used to model the r=e curve over the ap￾propriate stress range for two specimens (012 and 068) with similar interfacial shear stress, s, of
70 MPa (Fig. 8) from the approach taken in [12] based on [21,22]: e ¼ r=Ec þ adðrÞqc=Efðr þ rthÞ; for q1 c > 2d; ð3Þ where the sliding length d ¼ arðr þ rthÞ=2s ð4Þ r is the fiber radius, and a ¼ ð1 f ÞEm=fEc: ð5Þ The stress-dependent matrix cracking behavior at room temperature can be used to model the time-de￾pendent elevated temperature life expectancy at stress [6]. The stress–strain curve of these composites upon initial loading varies only slightly up to 1200
C [2]. Little difference is observed in load–unload behavior at 815
C, including rth, until higher stresses are reached (>275 MPa) after considerable exposure time (
30 min) in air at temperature during the unload–reload test. Therefore, it can be assumed that the matrix cracking of these materials upon initial loading behaves essentially the same as at room temperature and the stress-depen￾dent matrix-crack distribution can be used as the initial crack density at elevated temperatures upon loading. Evidence for this has been demonstrated for a number of the panels tested in this study where speci￾mens have been subjected to stress-rupture testing at 815
C. After rupture, through-thickness cracks remain open due to the formation of solid oxidation products in the matrix crack and there is no need to etch the polished sections. The measured crack densities after rupture are commensurate with what would have been expected from the AE data and known final crack density measured after room temperature testing to failure (Fig. 9). For specimens from the 007 and 044 composite panels, the matrix crack densities after stress-rupture at an intermediate stress were nearly ex￾actly what was expected from the room temperature AE data. For specimens from the 017 composite panel, the matrix crack densities after stress-rupture at an intermediate stress were slightly lower by approxi￾mately 33% of what was to be expected from the room temperature AE data which represents the largest de￾viation from estimated crack density observed. It should be noted that for the Sylramic/BN/MI system, it has been shown that relatively little global AE activity G.N. Morscher / Composites Science and Technology 64 (2004) 1311–1319 1317

G N. Morscher Composites Science and Technology 64 (2004)1311-1319 erformed on 3D architecture composites with double 017;RTp=120° tows in the load-bearing direction and single-tows per pendicular to the loading-direction, the identical narrow 007: RTD =8 stress-range versus ominimatrix was found to exist [25] tress This implies that the narrow stress range for matrix cracking was due to the larger fiber-bridged tow size in he loading-direction because only single-tows were used 0.41stress-rupture. in the 90 direction for these 3D composites. However, the absolute stress-range for matrix cracking would still be dependent on the type of woven architecture and 044RTp=9.50° concentration of composite constituents Finally, it should be noted that the specific results of this study only apply to the Sylramic/BN/MI properties Stress. MPa from this vendor. Changing the fiber, interphase, matrix, Fig. 9. Crack density measured in the load-bearing minicomposites or possibly even the vendor could alter the absolute nature after stress-rupture at 815. C in air compared with the AE activity at of matrix cracking of that composite system compared to room temperature for specimens from the same panel for two different what was found in this study. However, the methodology and approach taken here would be expected to be appli cable for a different composite system as a way to char occurs after initial loading for stress-rupture tests at acterize the stress-dependent crack distribution 815C[24, i.e., little matrix crack formation or growth occurs after attaining the constant stress condition. This confirms that matrix cracks in these composites are 5. Co through-thickness, i.e., progress through the MI matrix, even though not visible after plasma-etch A relationship for matrix cracking versus applied The differences between the double-tow woven and stress was effectively established by assuming that matrix single-tow woven matrix cracking behaviors were strik- cracks originate in the region of the material outside of ing. The double-tow woven composites have shown the load-bearing minicomposite for the 2D woven Syl higher composite strengths for the same volume fraction ramic fiber, BN interphase, melt-infiltrated SiC matrix omposites [17]. However, the relatively lower stress composite system. This was confirmed for a number of range for matrix cracking may be detrimental for use of composite specimens processed with different fiber ar these composites at higher stresses due to strength-de- chitectures and numbers of plies resulting in varied grading oxidation processes. The much narrower matrix concentrations of composite constituents; however, all crack distribution for the double-tow composites ap having the same essential flaw-producer, a standard proaches the more classical assumption of a single ma- sized 90 fiber tow. When an effective larger woven fiber trix crack stress [3]. The narrow matrix cracking stress tow(two standard tows woven together)was used,a range must be due to the larger fiber-count tow. Two different stress-dependent matrix crack distribution possibilities seem apparent: (I)the height of the 90 observed. A simple mathematical relationship for stress bundles is large creating larger effective minimatrix flaws dependent matrix cracking was used to describe the and/or(2) larger but fewer and more spaced out fiber- stress-dependent matrix crack density from which the bridged regions. The dimensions of the 90 bundle, i. e, stress-strain response for a composite component for height and width, were measured for a number of the areas of different thickness, numbers of plies, and/or 2D standard tow composites tested (007, 012, 017, 011, 044) fiber architectures could be modeled. The stress-depen and the 041 double-tow woven composite from polished dent matrix crack distribution also provides the starting wenty-five 90 tows were measured from crack density for modeling the elevated temperature each specimen. The size for all of the standard tow properties of these composites when subjected to me- woven composites fell into a range of h=0. 13+0.01 chanical loads at temperature mm and w=1. 14+0.04 mm. The size for the 041 double-tow woven composite was h=0.12 and w 2. 34. For each individual specimen the scatter was x10%. Acknowledgements Therefore, the height of the tow, presumably the im- portant dimension for flaw size was nearly identical for This work was supported by the Ultra Efficient Engine the standard tow and double-tow woven composites. Technology (UEET) program at NASA Glenn Research However, the load-bearing area of the tow was of course Center. The author also appreciated the suggestion by double for the double-tow woven with twice the number Dr William Curtin of Brown University to assume matrix of fibers per effective load-bearing tow. In recent tests cracking occurs outside the load-bearing minicomposite

occurs after initial loading for stress-rupture tests at 815
C [24], i.e., little matrix crack formation or growth occurs after attaining the constant stress condition. This confirms that matrix cracks in these composites are through-thickness, i.e., progress through the MI matrix, even though not visible after plasma-etch. The differences between the double-tow woven and single-tow woven matrix cracking behaviors were strik￾ing. The double-tow woven composites have shown higher composite strengths for the same volume fraction composites [17]. However, the relatively lower stress range for matrix cracking may be detrimental for use of these composites at higher stresses due to strength-de￾grading oxidation processes. The much narrower matrix crack distribution for the double-tow composites ap￾proaches the more classical assumption of a single ma￾trix crack stress [3]. The narrow matrix cracking stress range must be due to the larger fiber-count tow. Two possibilities seem apparent: (1) the height of the 90
bundles is large creating larger effective minimatrix flaws and/or (2) larger but fewer and more spaced out fiber￾bridged regions. The dimensions of the 90
bundle, i.e., height and width, were measured for a number of the standard tow composites tested (007, 012, 017, 011, 044) and the 041 double-tow woven composite from polished micrographs. Twenty-five 90
tows were measured from each specimen. The size for all of the standard tow woven composites fell into a range of h ¼ 0:13 0:01 mm and w ¼ 1:14 0:04 mm. The size for the 041 double-tow woven composite was h ¼ 0:12 and w ¼ 2:34. For each individual specimen the scatter was
10%. Therefore, the height of the tow, presumably the im￾portant dimension for flaw size was nearly identical for the standard tow and double-tow woven composites. However, the load-bearing area of the tow was of course double for the double-tow woven with twice the number of fibers per effective load-bearing tow. In recent tests performed on 3D architecture composites with double￾tows in the load-bearing direction and single-tows per￾pendicular to the loading-direction, the identical narrow stress-range versus rminimatrix was found to exist [25]. This implies that the narrow stress range for matrix cracking was due to the larger fiber-bridged tow size in the loading-direction because only single-tows were used in the 90
direction for these 3D composites. However, the absolute stress-range for matrix cracking would still be dependent on the type of woven architecture and concentration of composite constituents. Finally, it should be noted that the specific results of this study only apply to the Sylramic/BN/MI properties from this vendor. Changing the fiber, interphase, matrix, or possibly even the vendor could alter the absolute nature of matrix cracking of that composite system compared to what was found in this study. However, the methodology and approach taken here would be expected to be appli￾cable for a different composite system as a way to char￾acterize the stress-dependent crack distribution. 5. Conclusions A relationship for matrix cracking versus applied stress was effectively established by assuming that matrix cracks originate in the region of the material outside of the load-bearing minicomposite for the 2D woven Syl￾ramic fiber, BN interphase, melt-infiltrated SiC matrix composite system. This was confirmed for a number of composite specimens processed with different fiber ar￾chitectures and numbers of plies resulting in varied concentrations of composite constituents; however, all having the same essential flaw-producer, a standard sized 90
fiber tow. When an effective larger woven fiber tow (two standard tows woven together) was used, a different stress-dependent matrix crack distribution was observed. A simple mathematical relationship for stress￾dependent matrix cracking was used to describe the stress-dependent matrix crack density from which the stress–strain response for a composite component for areas of different thickness, numbers of plies, and/or 2D fiber architectures could be modeled. The stress-depen￾dent matrix crack distribution also provides the starting crack density for modeling the elevated temperature properties of these composites when subjected to me￾chanical loads at temperature. Acknowledgements This work was supported by the Ultra Efficient Engine Technology (UEET) program at NASA Glenn Research Center. The author also appreciated the suggestion by Dr. William Curtin of Brown University to assume matrix cracking occurs outside the load-bearing minicomposite. Fig. 9. Crack density measured in the load-bearing minicomposites after stress-rupture at 815
C in air compared with the AE activity at room temperature for specimens from the same panel for two different panels. 1318 G.N. Morscher / Composites Science and Technology 64 (2004) 1311–1319

G N. Morscher Composites Science and Technology 64(2004)1311-1319 References [12 Yun HM, DiCarlo JA. In: Krenkel w, Naslain R, Schneider H, editors. High temperature ceramic matrix composites. Weinhein l] Johnson AM, Bartlett BJ, Troha WA. In: Billig FS, editor rmany: Wiley/VCH: 2001. p.99-105 Thirteenth International Symposium for Air Breathing Engines [13] Yun HM, Gyekenyesi JZ, Chen YL, Wheeler DR, DiCarlo JA. ISABE97-7179,vol.2;1997.p.132l-8 Ceram Eng Sci Proc 2001: 22(3): 521-31 [14 Morscher GN, Eldridge JI. In: Ravi-Chandar et al. editors. 3 Aveston J, Cooper GA, Kelly A. In: The Properties of Fibre Advances in Fracture Research Proceedings of International Composites, Conference Proceedings, National Physics Labora- Conference of Fracture 10 on compact tory, Guilford, UK. Teddington, UK: IPC Science and Technol- nce. 2001 ogy Press Ltd; 1971.p 15-24 [15 Yun HM, DiCarlo JA. unpublished data. Heredia FE, McNulty JC, Zok FW, Evans AG. J Am Ceram Soc [16 Morscher GN. Comp Sci Tech 1999: 59(5): 687-97 [17 Morscher GN. In: Thompson DO, Chimenti DE, editors. 5 Morscher GN, Hurst J, Brewer D. J Am Ceram Soc 2000: 83(6): Review of progress in quantitative nondestructive evaluation, rol. 18A. New York: Kluwer Academic/Plenum Publishers; 1999. 6 Morscher GN, Cawley JD. J Eur Ceram Soc 2002; 22(14-15): 419-26 [18] Steen M, Valles JL. In: Jenkins MG, editor. ASTM STP 1309 [7 Pluvinage P, Parvizi-Majidi A, Chou Tw. J Mater Sci 1996: 31: est Conshohocken, PA: American Society for Testing and Materials: 1997. P. 49-65. [8 Guillaumat L, Lamon J. Multi-fissuration de composites SiC/ [19 Morscher GN, Yun HM, Thomas-Ogbuji L, DiCarlo JA.JAm Sic. Revue des composites et des materiaux avances 1993: 3: 159- Ceram Soc [in print] 71 20 Domergue JM, Heredia FE, Evans AG. J Am Ceram Soc 1996: [9 Guillaumat L, Lamon J Comp Sci Tech 1996: 56: 803-8 [10 Lamon J, Thommeret B, Percevault C J Eur Ceram Soc 1998: 18 21 Curtin WA, Ahn BK, Takeda N. Acta Mater 1998: 46(10):3409- 1797-808 [ll] DiCarlo JA, Yun HM, Morscher GN. Thomas-Ogbuji LU 22Pryce AW, Smith PA Br Ceram Trans 1993: 92(2): 49-54. ranmkic matrix com pos tes weinheim. germ ang wileyvche (24 Morscher GN, Hurst 3. Ceram Eng Sci proc 2001: 22(3):547-52. 2001.p.777-82. 5 Morscher GN, Yun HM, Di

References [1] Johnson AM, Bartlett BJ, Troha WA. In: Billig FS, editor. Thirteenth International Symposium for Air Breathing Engines, ISABE 97-7179, vol. 2; 1997. p. 1321–8. [2] Brewer D. Mater Sci Eng A 1999;A261:284–91. [3] Aveston J, Cooper GA, Kelly A. In: The Properties of Fibre Composites, Conference Proceedings, National Physics Labora￾tory, Guilford, UK. Teddington, UK: IPC Science and Technol￾ogy Press Ltd; 1971. p. 15–24. [4] Heredia FE, McNulty JC, Zok FW, Evans AG. J Am Ceram Soc 1996;79(5):1181–8. [5] Morscher GN, Hurst J, Brewer D. J Am Ceram Soc 2000;83(6): 1441–9. [6] Morscher GN, Cawley JD. J Eur Ceram Soc 2002;22(14–15): 2777–88. [7] Pluvinage P, Parvizi-Majidi A, Chou TW. J Mater Sci 1996;31: 232–41. [8] Guillaumat L, Lamon J. Multi-fissuration de composites SiC/ SiC. Revue des composites et des materiaux avances 1993;3:159– 71. [9] Guillaumat L, Lamon J. Comp Sci Tech 1996;56:803–8. [10] Lamon J, Thommeret B, Percevault C. J Eur Ceram Soc 1998;18: 1797–808. [11] DiCarlo JA, Yun HM, Morscher GN, Thomas-Ogbuji LU. In: Krenkel W, Naslain R, Schneider H, editors. High temperature ceramic matrix composites. Weinheim, Germany: Wiley/VCH; 2001. p. 777–82. [12] Yun HM, DiCarlo JA. In: Krenkel W, Naslain R, Schneider H, editors. High temperature ceramic matrix composites. Weinheim, Germany: Wiley/VCH; 2001. p. 99–105. [13] Yun HM, Gyekenyesi JZ, Chen YL, Wheeler DR, DiCarlo JA. Ceram Eng Sci Proc 2001;22(3):521–31. [14] Morscher GN, Eldridge JI. In: Ravi-Chandar et al. editors. Advances in Fracture Research Proceedings of International Conference of Fracture 10 on compact disk, Honolulu. Elsevier Science; 2001. [15] Yun HM, DiCarlo JA. unpublished data. [16] Morscher GN. Comp Sci Tech 1999;59(5):687–97. [17] Morscher GN. In: Thompson DO, Chimenti DE, editors. Review of progress in quantitative nondestructive evaluation, vol. 18A. New York: Kluwer Academic/Plenum Publishers; 1999. p. 419–26. [18] Steen M, Valles JL. In: Jenkins MG, editor. ASTM STP 1309. West Conshohocken, PA: American Society for Testing and Materials; 1997. p. 49–65. [19] Morscher GN, Yun HM, Thomas-Ogbuji L, DiCarlo JA. J Am Ceram Soc [in print]. [20] Domergue JM, Heredia FE, Evans AG. J Am Ceram Soc 1996; 79(1):161–70. [21] Curtin WA, Ahn BK, Takeda N. Acta Mater 1998;46(10):3409– 20. [22] Pryce AW, Smith PA. Br Ceram Trans 1993;92(2):49–54. [23] Lissart N, Lamon J. Acta Mater 1997;45(3):1025–44. [24] Morscher GN, Hurst J. Ceram Eng Sci Proc 2001;22(3):547–52. [25] Morscher GN, Yun HM, DiCarlo J. J Am Ceram Soc [submitted]. G.N. Morscher / Composites Science and Technology 64 (2004) 1311–1319 1319