《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_j.1151-2916.1999.tb01734.x[1]

ournal In. Ceran. Soc, 82[1]1415-55(1999) Fiber Effects on minicomposite mechanical Properties for Several Silicon Carbide Fiber-Chemically Vapor-Infiltrated Silicon Carbide Matrix Systems Gregory N Morscher, t Department of Materials Science and Engineering, Case Western Reserve University, Cleveland, Ohio Julian Martinez-Fernandez Departamento de Fisica de la Materia Condensada, University of Seville, Seville, Spain Several different types of SiC fiber tows were coated with fiber tests while mimicking, to some degree, the larger-scale BN and composited using chemically vapor- infiltrated Sic macrocomposite tensile behavior. Single-fiber tests have been to form single-tow minicomposites. The types of SiC fib used to determine the interfacial properties for a Nicalon fiber/ included NicalonM, Hi-Nicalon'M, and the new SyiramieT CVI-SiC matrix system, a CVD-fiber/glass-matrix system, 0 were determined from unload-reload tensile hysteresis- studies, the model of Marshall 2 for fiber pullout was used to loop tests. The ultimate stress and strain properties also determine the interfacial properties of the fiber/matrix system. were determined for the minicomposites. The ultimate Also, the effect of fiber roughness on fiber sliding was deter- strengths of the newer Hi-Nicalon and Sylramic fibers were mined for the CVD-SiC/glass matrix system superior to that of Nicalon minicomposites with similar Because macrocomposite are composed of several tows in iber volume fractions. The Sylramic minicomposites had the loading direction, the minicomposite represents a subele- strength, respectively, because of the high modulus of the of minicomposites is very similar to that of macrocomposite, fiber and the rough surface of this fiber type. The apparent because the interfacial and elastic properties of the constituents interfacial shear strength increased as the stress increased are the same. The matrix and fibers also have different flaw for the Sylramic minicomposites, which also was attributed distributions, as in a macrocomposite, which account for, in to the surface roughness of this fiber accordance with the interfacial properties, the stress-dependent cracking behavior and fiber failure prior to ultimate failure L. Introduction The aspect of a macrocomposite that is foreign to minicom- posites are 90 plies, which are often low-stress crack-initiation composites to be used at temperatures >1200%C, fibers with(tunnel cracking)intersect load-bearing tows. Therefore,the better creep and rupture properties than ceramic-grade(CG absolute stress-strain behavior of minicomposites is expected to differ from macrocomposite, because of the differences in able' Fiber development has been underway, and several ven- crack morphology that result from differences in matrix-flaw dors are or will be offering higher-use-temperature SiC-based distributions and fiber architecture fibers. 2-4 Currently, SylramicTM and Hi-Nicalon TM fibers are substantially more expensive than CG NicalonTM. It is es of fiber and similar interphases and matrices and then that. as these fibers find greater use ca In this study, minicomposites were fabricated with different their cost will also de- tensile tested at room temperature, to determine the effect of crease. There is a current need to assess the effect of different fiber properties on the SIiC /BN,SiCm system. Even though the fibers for a given composite system in a cost-effective manner. absolute stress-strain behaviors of minicomposites are ex- One way to accomplish this is to fabricate and test single-tow pected to differ from macrocomposite that have been made minicomposites. Minicomposite fabrication and testing are es- with the same constituents the relative difference in stress. pecially practical for ceramic composite systems that use strain behavior of the different fiber-type minicomposites is chemically vapor infiltrated(CVI) interphases and matrices expected to translate to macrocomposite stress-strain behavior such as in many SiC fiber/BN interphase/SiC matrix systems This test approach has already been used to determine roor Il. Experimental Procedure temperature interfacial and ultimate tensile properties,6 and high-temperature stress rupture, and cyclic stress properties Single-fiber-tow composites were processed in the same way in ambient air for several fiber/interphase/matrix composite as that described by Morscher. A single-fiber tow was coate with a BN interphase, and then the coated tow was infiltrated The minicomposite test incorporates the analysis of single with SiC. The three different minicomposites are listed in Table I, with pertinent physical rties of the composite components. Note that the Bn interphase for Nicalon-fiber ites was processed by a different vendor than the K.T. Faber--contributing editor Hi-Nicalon-fiber and Sylramic-fiber minicomposites. There are two reasons for this. The nicalon fibers were coated for an earlier study and at the time when the Hi-Nicalon and syl- ramic fibers were coated. the vendor who coated the nicalon lo. 190649. Received October 6, 1997, approved April 24, 1998. fibers was not in the fiber-coating business anymore. also A-Lewis Research Center. Cleveland OH Nicalon could not be coated with the BN coating applied to the 44135-3127 Hi-Nicalon or Sylramic fibers because the BN

Fiber Effects on Minicomposite Mechanical Properties for Several Silicon Carbide Fiber–Chemically Vapor-Infiltrated Silicon Carbide Matrix Systems Gregory N. Morscher*,† Department of Materials Science and Engineering, Case Western Reserve University, Cleveland, Ohio Julian Martinez-Fernandez Departamento de Fisica de la Materia Condensada, University of Seville, Seville, Spain Several different types of SiC fiber tows were coated with BN and composited using chemically vapor-infiltrated SiC to form single-tow minicomposites. The types of SiC fiber included Nicalon™, Hi-Nicalon™, and the new Sylramic™ polycrystalline SiC fiber. The interfacial shear stresses were determined from unload–reload tensile hysteresis￾loop tests. The ultimate stress and strain properties also were determined for the minicomposites. The ultimate strengths of the newer Hi-Nicalon and Sylramic fibers were superior to that of Nicalon minicomposites with similar fiber volume fractions. The Sylramic minicomposites had the lowest strain to failure and highest interfacial shear strength, respectively, because of the high modulus of the fiber and the rough surface of this fiber type. The apparent interfacial shear strength increased as the stress increased for the Sylramic minicomposites, which also was attributed to the surface roughness of this fiber. I. Introduction I T IS evident that, in order for SiC-fiber-reinforced ceramic composites to be used at temperatures >1200°C, fibers with better creep and rupture properties than ceramic-grade (CG) Nicalon™ (Nippon Carbon, Tokyo, Japan) need to be avail￾able.1 Fiber development has been underway, and several ven￾dors are or will be offering higher-use-temperature SiC-based fibers.2–4 Currently, Sylramic™ and Hi-Nicalon™ fibers are substantially more expensive than CG Nicalon™. It is expected that, as these fibers find greater use, their cost will also de￾crease. There is a current need to assess the effect of different fibers for a given composite system in a cost-effective manner. One way to accomplish this is to fabricate and test single-tow minicomposites.5 Minicomposite fabrication and testing are es￾pecially practical for ceramic composite systems that use chemically vapor infiltrated (CVI) interphases and matrices such as in many SiC fiber/BN interphase/SiC matrix systems. This test approach has already been used to determine room￾temperature interfacial and ultimate tensile properties5,6 and high-temperature stress rupture7,8 and cyclic stress8 properties in ambient air for several fiber/interphase/matrix composite systems. The minicomposite test incorporates the analysis of single￾fiber tests while mimicking, to some degree, the larger-scale macrocomposite tensile behavior. Single-fiber tests have been used to determine the interfacial properties for a Nicalon fiber/ CVI-SiC matrix system,9 a CVD-fiber/glass-matrix system,10 and a CVD-SiC fiber/CVI-SiC matrix system.11 For all these studies, the model of Marshall12 for fiber pullout was used to determine the interfacial properties of the fiber/matrix system. Also, the effect of fiber roughness on fiber sliding was deter￾mined for the CVD-SiC/glass matrix system.10 Because macrocomposites are composed of several tows in the loading direction, the minicomposite represents a subele￾ment of the macrocomposite. The tensile stress–strain behavior of minicomposites is very similar to that of macrocomposites,5 because the interfacial and elastic properties of the constituents are the same. The matrix and fibers also have different flaw distributions, as in a macrocomposite, which account for, in accordance with the interfacial properties, the stress-dependent cracking behavior and fiber failure prior to ultimate failure. The aspect of a macrocomposite that is foreign to minicom￾posites are 90° plies, which are often low-stress crack-initiation sites13 and non-through-thickness cracking that may or may not (tunnel cracking14) intersect load-bearing tows. Therefore, the absolute stress–strain behavior of minicomposites is expected to differ from macrocomposites, because of the differences in crack morphology that result from differences in matrix-flaw distributions and fiber architecture. In this study, minicomposites were fabricated with different types of fiber and similar interphases and matrices and then tensile tested at room temperature, to determine the effect of fiber properties on the SiCf /BNi /SiCm system. Even though the absolute stress–strain behaviors of minicomposites are ex￾pected to differ from macrocomposites that have been made with the same constituents, the relative difference in stress– strain behavior of the different fiber-type minicomposites is expected to translate to macrocomposite stress–strain behavior. II. Experimental Procedure Single-fiber-tow composites were processed in the same way as that described by Morscher.7 A single-fiber tow was coated with a BN interphase, and then the coated tow was infiltrated with SiC. The three different minicomposites are listed in Table I, with pertinent physical properties of the composite components. Note that the BN interphase for Nicalon-fiber minicomposites was processed by a different vendor than the Hi-Nicalon-fiber and Sylramic-fiber minicomposites. There are two reasons for this. The Nicalon fibers were coated for an earlier study, and at the time when the Hi-Nicalon and Syl￾ramic fibers were coated, the vendor who coated the Nicalon fibers was not in the fiber-coating business anymore. Also, Nicalon could not be coated with the BN coating applied to the Hi-Nicalon or Sylramic fibers because the BN processing tem￾K. T. Faber—contributing editor Manuscript No. 190649. Received October 6, 1997; approved April 24, 1998. *Member, American Ceramic Society. † Resident Research Associate at NASA–Lewis Research Center, Cleveland, OH 44135–3127. J. Am. Ceram. Soc., 82 [1] 145–55 (1999) Journal 145

146 Journal of the American Ceramic Society-Morscher and Martines-Fernandes Vol. 82. Ne Table I. Data for Single-Tow Minicomposites That Have Been Tested ber diameter ic/3MBN/SIC 0.07-0.12 Hi-NIC/PBN/SIC 0.13-0.2 0.75(0.43) Syl/PBN/SIC 0.12-0.19 045(0.3-2) pyrolytic BN(Advanced Ceramics, CI land and density of the fiber tow, BN interphase, and CVI-SiC matrix. 'Values given in e for this coating would severely degrade the fiber Each individual minicomposite that failed was cut into sev- strength. The BN that was used to coat the Nicalon tows was eral -30 mm lengths, mounted in epoxy, and polished longi- rocessed at a low temperature (-1050oC)and a thin Sic tudinally to a I um finish to determine the matrix-crack spac layer(0.5 um thick) was applied on top of the bn by the More than half of the gauge section was used to determine nterphase vendor before SiC infiltration. The Hi-Nicalon and he matrix-crack spacing. For some of the Syl-PBN minicom- Sylramic tows were coated with a bn phase that was de sites. the cracks were not always observable after polishing at 1400C. 16 The abbreviations 3MBN and PbN will be used and required etching in a boiling Muriakami's solution. Matrix denote the interphases for the Nicalon -fiber(Nic-3MBN) Hi- cracks in the Nic-3MBN and HN-PBN minicomposites were icalon-fiber(HN-PBN), and Sylramic-fiber(Syl-PBN)mini- easily observed after the longitudinal polishing com The fiber volume fraction for the minicomposites was de termined from the weight of the fibers, the weight gain of the IlL. Results BN coating, and the weight gain of the CVI infiltration. The density of Bn used in the calculations for PBN was 1.76 g/cr Tensile-Test Results based on the measured value of a similar bn made in bulk Typical examples of stress-strain hysteresis loc p tests are orm. The density(atomic order) of Bn increased as the shown in Fig. I for the Nic-3MBN and Syl-PBN minicompos processing temperature increased. The density of 3MBN is ites. The stress-strain behavior of the HN-PBN minicompos unknown. A value of 1.5 g/cm was assumed for the 3MBN ites is not shown for clarity but would be located between that nterphase, because it was processed at a lower temperature. of the two minicomposite types shown in Fig. 1. The composite The Syl-PBN and HN-PBN minicomposites that were tested stress was determined by multiplying the fiber volume fractio had an average fiber volume fraction of -0.16. The average by the stress on the fibers if fully loaded(failure load divided volume fraction of the Nic-3MBN minicomposites was sligh by total fiber area). Figure 2 shows the final hysteresis loop just lower (0.1). However, there was a range of fiber volume prior to failure for the three minicomposite types that were fractions. as noted in Table I sted in this study. The fiber volume fraction for the Nic- o The interphase thickness was variable across the tow cross 3MBN, HN-PBN, and Syl-PBN minicomposites used for com- est variability. The outer layer of fibers could have a coating up respectively parison in this study( Figs. 2-6) were 0.12, 0.16, and 0.16 to a 3 um thick, whereas the interior fibers in a tow The ultimate of the HN-PBN and Syl-PBN mini much-thinner(-0 4 um)and more-uniform coatings (Table I composites we ately the same(450 MPa), wher calcular age value for the coating thickness was used in the c-3MBN minicomposites was -310 ons, based on the weight gain of the PBN MPa. Based on the number of fibers per tow, these strengths e The minicomposites were mounted with epoxy to cardboard correspond to an average fiber strength of.8 GPa for the emission(AE)sensors were attached to the epoxy withalleaesic GPa for the Nicalon fibers, assuming that the fibers were bear ing the full load just prior to failure. The as-produced fiber tor clips. The length of the minicomposites between the card- strengths were -2.8-3.0 GPa for all three of these fiber types board edges ranged from 60 to 150 mm. Tensile testing was performed on a universal testing machine( Model 4502, stron, Canton, MA). Tensile unload-reload hysteresis loops were performed with increasing loads until the minicomposite failed. The displacements of the upper and lower cardboard tabs were monitored with a laser extensometer(Model Zygo Syl-PBN ), Zygo train was mined from the difference between the upper-tab and lower-tab 5 edge displacements divided by the length of the minicompos- 9 ites between the edges. The energy of the acoustic events was monitored with an AE analyzer(Model LOCAN 320, Physical Acoustics, Princeton, NJ). The AE analyzer also included computer, which collected the load, strain, and AE data. Be cause of the gripping method and attachment of AE transducers mBN to the epoxy, significant bending moments could result near the grips. The minicomposites often failed in this region at stresses =50 lower than those achieved for gauge failures. Only samples that failed in the gauge section were used to determine the ultimate properties. However, tensile hysteresis analysis could Strain. still be performed with samples that failed prematurely(near

perature for this coating would severely degrade the fiber strength. The BN that was used to coat the Nicalon tows was processed at a low temperature (∼1050°C)15 and a thin SiC layer (∼0.5 mm thick) was applied on top of the BN by the interphase vendor before SiC infiltration. The Hi-Nicalon and Sylramic tows were coated with a BN phase that was deposited at 1400°C.16 The abbreviations 3MBN and PBN will be used to denote the interphases for the Nicalon-fiber (Nic-3MBN), Hi￾Nicalon-fiber (HN-PBN), and Sylramic-fiber (Syl-PBN) mini￾composites, respectively. The fiber volume fraction for the minicomposites was de￾termined from the weight of the fibers, the weight gain of the BN coating, and the weight gain of the CVI infiltration. The density of BN used in the calculations for PBN was 1.76 g/cm3 , based on the measured value of a similar BN made in bulk form.17 The density (atomic order) of BN increased as the processing temperature increased. The density of 3MBN is unknown. A value of 1.5 g/cm3 was assumed for the 3MBN interphase, because it was processed at a lower temperature. The Syl-PBN and HN-PBN minicomposites that were tested had an average fiber volume fraction of ∼0.16. The average volume fraction of the Nic-3MBN minicomposites was slightly lower (∼0.1). However, there was a range of fiber volume fractions, as noted in Table I. The interphase thickness was variable across the tow cross section.7 The higher-temperature PBN coatings had the great￾est variability. The outer layer of fibers could have a coating up to a 3 mm thick, whereas the interior fibers in a tow had much-thinner (∼0.4 mm) and more-uniform coatings (Table I). An average value for the coating thickness was used in the calculations, based on the weight gain of the PBN. The minicomposites were mounted with epoxy to cardboard tabs, as described in other studies.7,11 The epoxy section of the minicomposite was gripped with pneumatic grips, and acoustic emission (AE) sensors were attached to the epoxy with alliga￾tor clips. The length of the minicomposites between the card￾board edges ranged from 60 to 150 mm. Tensile testing was performed on a universal testing machine (Model 4502, In￾stron, Canton, MA). Tensile unload–reload hysteresis loops were performed with increasing loads until the minicomposite failed. The displacements of the upper and lower cardboard tabs were monitored with a laser extensometer (Model Zygo 1100, Zygo Corp., Middlefield, CT). The strain was deter￾mined from the difference between the upper-tab and lower-tab edge displacements divided by the length of the minicompos￾ites between the edges. The energy of the acoustic events was monitored with an AE analyzer (Model LOCAN 320, Physical Acoustics, Princeton, NJ). The AE analyzer also included a computer, which collected the load, strain, and AE data. Be￾cause of the gripping method and attachment of AE transducers to the epoxy, significant bending moments could result near the grips. The minicomposites often failed in this region at stresses lower than those achieved for gauge failures. Only samples that failed in the gauge section were used to determine the ultimate properties. However, tensile hysteresis analysis could still be performed with samples that failed prematurely (near the epoxy). Each individual minicomposite that failed was cut into sev￾eral ∼30 mm lengths, mounted in epoxy, and polished longi￾tudinally to a 1 mm finish to determine the matrix-crack spac￾ing. More than half of the gauge section was used to determine the matrix-crack spacing. For some of the Syl-PBN minicom￾posites, the cracks were not always observable after polishing and required etching in a boiling Muriakami’s solution. Matrix cracks in the Nic-3MBN and HN-PBN minicomposites were easily observed after the longitudinal polishing. III. Results (1) Tensile-Test Results Typical examples of stress–strain hysteresis loop tests are shown in Fig. 1 for the Nic-3MBN and Syl-PBN minicompos￾ites. The stress–strain behavior of the HN-PBN minicompos￾ites is not shown for clarity but would be located between that of the two minicomposite types shown in Fig. 1. The composite stress was determined by multiplying the fiber volume fraction by the stress on the fibers if fully loaded (failure load divided by total fiber area). Figure 2 shows the final hysteresis loop just prior to failure for the three minicomposite types that were tested in this study. The fiber volume fraction for the Nic- 3MBN, HN-PBN, and Syl-PBN minicomposites used for com￾parison in this study (Figs. 2–6) were 0.12, 0.16, and 0.16, respectively. The ultimate stresses of the HN-PBN and Syl-PBN mini￾composites were approximately the same (∼450 MPa), whereas the ultimate stress of the Nic-3MBN minicomposites was ∼310 MPa. Based on the number of fibers per tow, these strengths correspond to an average fiber strength of ∼2.8 GPa for the Sylramic fibers, 2.75 GPa for the Hi-Nicalon fibers, and 2.4 GPa for the Nicalon fibers, assuming that the fibers were bear￾ing the full load just prior to failure. The as-produced fiber strengths were ∼2.8–3.0 GPa for all three of these fiber types. Table I. Data for Single-Tow Minicomposites That Have Been Tested Minicomposite† Fiber diameter (mm) Elastic modulus (GPa) Volume fraction of fibers, vf ‡ Average interphase thickness§ Fiber Matrix (mm) Nic/3MBN/SiC 14 200 400 0.07–0.12 0.5 (0.4–1) Hi-Nic/PBN/SiC 13 280 400 0.13–0.21 0.75 (0.4–3) Syl/PBN/SiC 9 380 400 0.12–0.19 0.45 (0.3–2) † The assembly of the minicomposite is given in the format of fiber/interphase/matrix. Abbreviations in this format denote the following materials: ‘‘Nic’’ 4 Nicalon fiber and ‘‘Hi-Nic’’ 4 Hi-Nicalon fiber (Nippon Carbon, Tokyo, Japan); ‘‘Syl’’ 4 Sylramic polycrys￾talline SiC fiber (Dow Corning, Midland, MI); ‘‘3MBN’’ 4 low-temperature (1050°C)-deposited BN (The 3M Corp., St. Paul, MN); PBN 4 1400°C-deposited pyrolytic BN (Advanced Ceramics, Cleveland, OH); and ‘‘SiC’’ 4 the CVI-SiC matrix (B. F. Goodrich, Brecksville, OH). ‡ Determined from the mass and density of the fiber tow, BN interphase, and CVI-SiC matrix. § Values given in parentheses are the range of interphase thickness. Fig. 1. Typical stress–strain curves for the minicomposites. 146 Journal of the American Ceramic Society—Morscher and Martinez-Fernandez Vol. 82, No. 1

January 1999 Fiber Effects on Minicomposite Mechanical Properties for Several SiC/cvl-SiC Matrix System 班Nic-PBN Syl-PBN 50u Se 100 Nic-3MBN 00203040.50.6070B08 Fig. 2. Hysteresis loops for three minicomposites with different types of Sic fiber just prior to failure Evidently, the Nicalon fibers were damaged during the Cvi SiC step to a greater extent than the other fibers, as has been observed by other researchers. 8 It will be shown below that the 4015 crack spacing and interfacial shear strength are similar for the Nicalon and Hi-Nicalon minicomposites, therefore, the Fig 4. SEM micrograph of the fracture surface of the Nicalon/ strength comparison between the two minicomposite systems is 3MBN/SiC minicomposite reasonable, because the actual fiber length at peak stress-the gauge length'-is approximately the same. The Hi-Nicalon and Sylramic fibers did not incur very much, if any, strength would account for this wide scatter in the measured elastic loss after minicomposite fabrication modulu The scatter in the measured elastic modulus was approxi- The hysteresis-loop modulus, ultimate strain, and hysteresis- mately the same for all three minicomposite types(at least five loop widths were very different for the three different mini- amples for each minicomposite type were tested) and ranged composites; this is most easily observed in Fig. 2. Just prior to from 330 to 380 GPa, even though the different minicompos- failure, the hysteresis-loop modulus was 175 GPa for Syl-PBN ites consisted of fibers with different moduli. The expected 74 GPa for HN-PBN, and 47 GPa for Nic-3MBN minicom- elastic moduli(rule of mixtures)and scatter due to the range of posites. The strain to failure was.27% for Syl-PBN,0.76% fiber volume fractions for each minicomposite system would for HN-PBN, and 0.9% for Nic-3MBN minicomposites. The hysteresis-loop widths were significantly larger for the HN- PBN, and Syl-PBN minicomposites, respectively. The mea- PBN and Nic-3MBN minicomposites, compared to the Syl sured values are lower and the scatter in measured values is PBN minicomposite. All these results were expected, consid- significantly larger than what would be expected. The experi ering the fiber moduli and roughnesses, as discussed below mental setup that is used in this study does introduce some The permanent deformation on unloading(Fig. 1)for the error in measuring the displacement at low strains, because Nic-3MBN minicomposite was very large(-0.2% for a peak some self-alignment occurs for the samples and grips, which stress of-320 MPa). Some of the permanent deformation could 350000 350000 军3000Nic-3MBN Syl NiC-3MBN PBN 250000 250000 Hi-NiC- 150Dn PBN 150000 世 100000 100000 Hi-Nic- PBN 1 50000 Syl-PBN 0 100150200250300350400450500001020.3040508070809 Composite Stress, MPa Composite Strain, Fig 3. Acoustic emission(AE) for different minicomposites

Evidently, the Nicalon fibers were damaged during the CVI￾SiC step to a greater extent than the other fibers, as has been observed by other researchers.18 It will be shown below that the crack spacing and interfacial shear strength are similar for the Nicalon and Hi-Nicalon minicomposites; therefore, the strength comparison between the two minicomposite systems is reasonable, because the actual fiber length at peak stress—the ‘‘gauge length’’—is approximately the same. The Hi-Nicalon and Sylramic fibers did not incur very much, if any, strength loss after minicomposite fabrication. The scatter in the measured elastic modulus was approxi￾mately the same for all three minicomposite types (at least five samples for each minicomposite type were tested) and ranged from 330 to 380 GPa, even though the different minicompos￾ites consisted of fibers with different moduli. The expected elastic moduli (rule of mixtures) and scatter due to the range of fiber volume fractions for each minicomposite system would be 381 ± 5, 379 ± 5, and 397 ± 1 GPa for Nic-3MBN, HN￾PBN, and Syl-PBN minicomposites, respectively. The mea￾sured values are lower and the scatter in measured values is significantly larger than what would be expected. The experi￾mental setup that is used in this study does introduce some error in measuring the displacement at low strains, because some self-alignment occurs for the samples and grips, which would account for this wide scatter in the measured elastic modulus. The hysteresis-loop modulus, ultimate strain, and hysteresis￾loop widths were very different for the three different mini￾composites; this is most easily observed in Fig. 2. Just prior to failure, the hysteresis-loop modulus was 175 GPa for Syl-PBN, 74 GPa for HN-PBN, and 47 GPa for Nic-3MBN minicom￾posites. The strain to failure was ∼0.27% for Syl-PBN, 0.76% for HN-PBN, and 0.9% for Nic-3MBN minicomposites. The hysteresis-loop widths were significantly larger for the HN￾PBN and Nic-3MBN minicomposites, compared to the Syl￾PBN minicomposite. All these results were expected, consid￾ering the fiber moduli and roughnesses, as discussed below. The permanent deformation on unloading (Fig. 1) for the Nic-3MBN minicomposite was very large (∼0.2% for a peak stress of ∼320 MPa). Some of the permanent deformation could Fig. 2. Hysteresis loops for three minicomposites with different types of SiC fiber just prior to failure. Fig. 3. Acoustic emission (AE) for different minicomposites. Fig. 4. SEM micrograph of the fracture surface of the Nicalon/ 3MBN/SiC minicomposite. January 1999 Fiber Effects on Minicomposite Mechanical Properties for Several SiC/CVI-SiC Matrix Systems 147

Journal of the American Ceramic Sociery-Morscher and Martines-Fernandes Vol. 82. No (b) 0.5mm Matrix Cracks Fig. 5. SEM micrographs of the fracture surfaces of the Hi-Nicalon/PBN/CVI-SiC m incOmposite( (a)matrix crack 300 um away from the fracture urface and(b) matrix crack 1200 um away from the fracture surface) 868428K Fig. 6. SEM micrographs of the fracture surfaces of the Sylramic/PBN/SiC minicomposite((a) matrix crack is <100 um from the fracture surface Figs. 6(b)and (c)show that fiber pullout occurs throughout the minicomposite)

Fig. 5. SEM micrographs of the fracture surfaces of the Hi-Nicalon/PBN/CVI-SiC minicomposite ((a) matrix crack 300 mm away from the fracture surface and (b) matrix crack 1200 mm away from the fracture surface). Fig. 6. SEM micrographs of the fracture surfaces of the Sylramic/PBN/SiC minicomposite ((a) matrix crack is <100 mm from the fracture surface; Figs. 6(b) and (c) show that fiber pullout occurs throughout the minicomposite). 148 Journal of the American Ceramic Society—Morscher and Martinez-Fernandez Vol. 82, No. 1

January 1999 Fiber Efects on Minicomposite Mechanical Properties for Several SiC/Cvl-SiC Matrix System associated with the relief of residual stres fiber did seem to show some pullout(Figs. 6(b)and(c)). The mposite with matrix cracking. The fiber had a thermal ex- Sylramic fibers that did pull out a longer distance(Fig. 6(b)) pansion coefficient of-3. 1 x 10-b/C, according to the product were always the outer-tow fibers, which had a thicker BN literature from the manufacturer for CG on. Assumin ayer. that the matrix has a thermal expansion coefficient of -4 Another observation is the presence of matrix cracks near the 10/C and the minicomposite was processed at-1025C, the fracture surface. For a HN-PBN minicomposite, the nearest anent deformation due to full decoupling of the fiber from matrix cracks to the fracture surface were -300 and 1200 um matrix would result in a permanent strain of-0.1% away from the fracture surface(Figs. 5(a)and(b). For a Syl value, of course, is an overestimate, because complete PBN minicomposite, the matrix-crack spacing is very smal rever. it is evident that more th (70%of the sample, which indicates that this system is far AE behavior(cracking trends) with stress, however, the Syl- from matrix-crack saturation the different crack. ng dis- PBN minicomposite has a rapid crack-growth regime for tributions are shown in Fig. 8 for the HN-PBN and Syl-PBN pared to that of the HN-PBN minicompos- minicomposit e. The rate of AE energy for the Nic-3MBN minicomposite decreases at -0 32% strain(-200 MPa stress). The HN-PBN IV. Analysis and Syl-PBN minicomposites do not show a decreasing rate o AE activity. All three minicomposit aturate in matrix cracks prior to the failure load a (1) Determination of the Interfacial Shear Stress fro Hysteresis Loops Scanning electron microscopy(SEM) micrographs of typical Lamon hysteresis loops were analyzed using the approach of (2) Examination of the Fracture Surface aly and v PBN, respectively ) Significant pullout lengths were observed for the Nic-3MBN and HN-3PBN minicomposites, whereas the da Syl-PBN minicomposites only had very short pullout lengths The Nic-3MBN fracture surfaces were always more frag- mented and jagged, in comparison to the other two minicom where de is the hysteresis loop width, o the peak stress of the posite types. It seems that the Sic did not infiltrate as well into hysteresis loop, o the stress where de is measured, and o min the the tow, compared to the other two minicomposite types, which minimum stress of the hysteresis loop. The inelastic strain in- resulted in the appearance of a thicker SiC"sheath"that sur dex, is determined from the relationship rounded the infiltrated tow. This phenomenon made it difficult to determine the pullout length for the Nic-3MBN minicom -a0(8) posite. The pullout-length distributions were determined for the HN-PBN and Syl-PBN minicomposite fracture surfaces. The d=b 4T-TE mean pullout lengths were 240 and 3 um for the HN-PBN and Syl-PBN minicomposites, respectively. Even though the Syl- where R is the average fiber radius, d the average matrix-crack PBN minicomposite fiber pullout is very small, almost ever spacing(number of cracks divided by minicomposite length), f Table Il. Mechanical Properties of the minicom crack Estimated interfacial Ultimate stress shear stress(MPa) Nic/3MBN Hi-Nic/PBN 15±10 -14 0.035 440±20 65-175 0.035 450±20

be associated with the relief of residual stresses in the mini￾composite with matrix cracking. The fiber had a thermal ex￾pansion coefficient of ∼3.1 × 10−6/°C, according to the product literature from the manufacturer for CG Nicalon. Assuming that the matrix has a thermal expansion coefficient of ∼4 × 10−6/°C and the minicomposite was processed at ∼1025°C, the permanent deformation due to full decoupling of the fiber from the matrix would result in a permanent strain of ∼0.1%. This value, of course, is an overestimate, because complete decou￾pling does not occur; however, it is evident that more than half of the permanent deformation measured for the Nic-3MBN minicomposite is not from the relief of residual stress in the gauge section of the minicomposite. Instead, at least half of the permanent deformation is associated with excessive damage and alignment that occurs at the tabs because of the gripping arrangement. This result is also probably the cause of the large load decreases that occur in the Nic-3MBN stress–strain curve at higher stresses (strains) in Fig. 1. Less permanent deforma￾tion was observed for the other two systems (Fig. 2). Figure 3 shows typical AE data for the three minicomposite types, as a function of minicomposite stress and strain. The AE energy can be attributed to matrix cracking.11 The first matrix￾cracking stress and strain for the different minicomposites was determined from the onset of AE activity (Fig. 3) and is tabu￾lated in Table II. The first matrix-cracking stresses and strains are approximately the same for the three minicomposite sys￾tems. At ∼0.04% strain, there is a significant deviation from linearity on the stress–strain curve for the Nic-3MBN (Fig. 1) and HN-PBN (not shown) minicomposites; however, for the Syl-PBN minicomposites, there is no significant deviation from linearity until a strain of ∼0.1% is reached. The AE data also implies that the rate of cracking with stress is much greater for Nic-3MBN minicomposites, at least at the onset of crack￾ing. The HN-PBN and Syl-PBN minicomposites have similar AE behavior (cracking trends) with stress; however, the Syl￾PBN minicomposite has a rapid crack-growth regime for strains >0.1%, compared to that of the HN-PBN minicompos￾ite. The rate of AE energy for the Nic-3MBN minicomposite decreases at ∼0.32% strain (∼200 MPa stress). The HN-PBN and Syl-PBN minicomposites do not show a decreasing rate of AE activity. All three minicomposite systems probably do not saturate in matrix cracks prior to the failure load. (2) Examination of the Fracture Surface Scanning electron microscopy (SEM) micrographs of typical minicomposite fracture surfaces are shown in Figs. 4–6 for the three minicomposite types (Nic-3MBN, HN-PBN, and Syl￾PBN, respectively). Significant pullout lengths were observed for the Nic-3MBN and HN-3PBN minicomposites, whereas the Syl-PBN minicomposites only had very short pullout lengths. The Nic-3MBN fracture surfaces were always more frag￾mented and jagged, in comparison to the other two minicom￾posite types. It seems that the SiC did not infiltrate as well into the tow, compared to the other two minicomposite types, which resulted in the appearance of a thicker SiC ‘‘sheath’’ that sur￾rounded the infiltrated tow. This phenomenon made it difficult to determine the pullout length for the Nic-3MBN minicom￾posite. The pullout-length distributions were determined for the HN-PBN and Syl-PBN minicomposite fracture surfaces. The mean pullout lengths were 240 and 3 mm for the HN-PBN and Syl-PBN minicomposites, respectively. Even though the Syl￾PBN minicomposite fiber pullout is very small, almost every fiber did seem to show some pullout (Figs. 6(b) and (c)). The Sylramic fibers that did pull out a longer distance (Fig. 6(b)) were always the outer-tow fibers, which had a thicker BN layer. Another observation is the presence of matrix cracks near the fracture surface. For a HN-PBN minicomposite, the nearest matrix cracks to the fracture surface were ∼300 and 1200 mm away from the fracture surface (Figs. 5(a) and (b)). For a Syl￾PBN minicomposite, the matrix-crack spacing is very small (70% of the sample, which indicates that this system is far from matrix-crack saturation. The different crack-spacing dis￾tributions are shown in Fig. 8 for the HN-PBN and Syl-PBN minicomposites. IV. Analysis (1) Determination of the Interfacial Shear Stress from Hysteresis Loops The hysteresis loops were analyzed using the approach of Lamon et al.9 and Vagaggini et al.19 The hysteresis loop width was related to stress for various peak-stress hysteresis loops as follows: d« sp 2 = 2+S s sp − smin sp DS1 − s sp D (1) where d« is the hysteresis loop width, sp the peak stress of the hysteresis loop, s the stress where d« is measured, and smin the minimum stress of the hysteresis loop. The inelastic strain in￾dex, +, is determined from the relationship + = b2 ~1 − a1 f! 2 S Rf d D 4f 2 tEm (2) where Rf is the average fiber radius, d the average matrix-crack spacing (number of cracks divided by minicomposite length), f Table II. Mechanical Properties of the Minicomposites Minicomposite Average crack spacing† (mm) Estimated interfacial shear stress (MPa) First cracking Ultimate stress (MPa) Stress (MPa) Strain (%) Nic/3MBN 0.45 25 ± 10 ∼140 0.035 310 ± 20 Hi-Nic/PBN 0.58 15 ± 10 ∼140 0.035 440 ± 20 Syl/PBN 0.34 65–175 ∼140 0.035 450 ± 20 † At failure. January 1999 Fiber Effects on Minicomposite Mechanical Properties for Several SiC/CVI-SiC Matrix Systems 149

150 Journal of the American Ceramic Society-Morscher and Martines-Fernandes Vol. 82. Ne The relative loop width, de/o2, was plotted versus the rela- e stress,(o/op -omin/opx(1-olap), as shown in Fig9.The slope is 29, and T can be determined if the crack spacing is known(Eq (2)). T was determined from the hysteresis loop just prior to failure, where the matrix-crack spacing was assumed to be that of the failed minicomposite. Table II lists the average crack spacing and T for the three different minicomposites The T value that was determined for the Syl-PBN minicom 1 mm po a lower in the range of 65-175 MPa. The sample that failec (a) Syl-PBN minicomposite that failed at the highest stress(Fig. 1) had a relatively high shear stress(170 MPa). This apparent stress-dependent T behavior was not observed in the other two (2) Stress-Strain Behavior of 0--mm Different Fiber Minicomposites If T is assumed to be constant throughout the minicomposite which is probably not the case for the Syl-PBN minicomposite and will be discussed below ) the minicomposite stress-strain curves can be modeled as a function of e and T if the amount of damage is known. The displacement that occurs around the cracks(AL1) can be determined in a manner similar to that by Lissart and Lamon 18 (4 where a is a material parameter(Er/E and Is is the sliding length and can be approximated from the relation (b) R Fig. 7. Examples of transverse matrix cracks in(a)Hi-Nicalon and (b) Sylramic minicomposites Ur is the stress on the fibers(o). The strain of the ked matrix regions is simply ee o/Ee. The total mInI- he fiber volume fraction, T the interfacial shear stress, and e from the elastic properties of the composite constituents: 2bed N△L1+(lg-2)e E E where Le is the gauge length and N is the number of cracks in Em[+(I-2vEd b2=(+vE[(1+p)E+(1-v)E was determined from knowing the number of cracks at failure and estimating the number of cracks at a given stress Where v is the Poisson's ratio(considered to be the same for the from the extent of AE activity at that stress(see Appendix)by er and the matrix and equal to 0.2)and the subscripts m, f, he relationship and c respectively refer to the matrix, fiber, and composite EA(o) N(o)=Nould (7) where N(o)and N(ouly are the number of cracks for the same sample estimated at the peak hysteresis-loop stress of interest Syl-PBN avg.=034 minimum=0.03 mm 12 HN-PBN avg=0.58 mm Hi-Nic-PBN: tF 20 MPa SyPBN: T=180 MP Crack Spacing, mm 0 Fig 8. Cumulative distribution of matrix-crack spacing for HN-PBN 00050101502025 and Syl-PBN; the total length of the minicomposite over which these values were determined was -70 mm for both types of minicompos- Fig 9. Interfacial shear-stress determination from hysteresis loops

the fiber volume fraction, t the interfacial shear stress, and Em the matrix modulus. The parameters a1 and b2 are determined from the elastic properties of the composite constituents:20 a1 = Ef E (3a) b2 = ~1 + n! Em@Ef + ~1 − 2n!Ec# Ef@~1 + n!Ef + ~1 − n!Ec# (3b) where n is the Poisson’s ratio (considered to be the same for the fiber and the matrix and equal to 0.2) and the subscripts m, f, and c respectively refer to the matrix, fiber, and composite. The relative loop width, d«/s2 p, was plotted versus the rela￾tive stress, (s/sp − smin/sp)(1 − s/sp), as shown in Fig. 9. The slope is 2+, and t can be determined if the crack spacing is known (Eq. (2)). t was determined from the hysteresis loop just prior to failure, where the matrix-crack spacing was assumed to be that of the failed minicomposite. Table II lists the average crack spacing and t for the three different minicomposites. The t value that was determined for the Syl-PBN minicom￾posites was in the range of 65–175 MPa. The sample that failed at a lower stress near the epoxy had a lower shear stress. The Syl-PBN minicomposite that failed at the highest stress (Fig. 1) had a relatively high shear stress (∼170 MPa). This apparent stress-dependent t behavior was not observed in the other two minicomposite systems. (2) Stress–Strain Behavior of Different Fiber Minicomposites If t is assumed to be constant throughout the minicomposite (which is probably not the case for the Syl-PBN minicomposite and will be discussed below), the minicomposite stress–strain curves can be modeled as a function of Ef and t if the amount of damage is known. The displacement that occurs around the cracks (DL1) can be determined in a manner similar to that by Lissart and Lamon:18 DL1 = S 2ls Ef DSsc f DS2 + a 1 + aD (4) where a is a material parameter (Ef/Ec) and ls is the sliding length and can be approximated from the relation ls = sf Rf 2t (5) where sf is the stress on the fibers (sc /f ). The strain of the uncracked matrix regions is simply «c 4 sc /Ec. The total mini￾composite strain can then be determined from the relation « = NDL1 + ~Lg − 2lsN!«c Lg (6) where Lg is the gauge length and N is the number of cracks in the gauge section. The only variable that needs to be deter￾mined is N as a function of stress. N was determined from knowing the number of cracks at failure and estimating the number of cracks at a given stress from the extent of AE activity at that stress (see Appendix) by the relationship N~s! = N~sult! S EAE~s! EAE~sult! D (7) where N(s) and N(sult) are the number of cracks for the same sample estimated at the peak hysteresis-loop stress of interest Fig. 7. Examples of transverse matrix cracks in (a) Hi-Nicalon and (b) Sylramic minicomposites. Fig. 8. Cumulative distribution of matrix-crack spacing for HN-PBN and Syl-PBN; the total length of the minicomposite over which these values were determined was ∼70 mm for both types of minicompos￾ites. Fig. 9. Interfacial shear-stress determination from hysteresis loops. 150 Journal of the American Ceramic Society—Morscher and Martinez-Fernandez Vol. 82, No. 1

January 1999 Fiber Effects on Minicomposite Mechanical Properties for Several SiC/cv1-SiC Matrix Systems d measured at the minicomposite failure stress, respectively. Similarly, EA(o)and EAF(out are the cumulative AE ener- gies for the same sample measured at the peak hysteresis-loop Sylramic-PBN stress of interest and measured at the minicomposite failure stress, respectively. Equation(6) was used to best fit the stress- shear stresses of 10 and 60 MPa best fit the HN-pbn and Syl-PBN data, respectively, for the estimated number of cracks 5 30 ormed during the experiment. These values were very similar to those found by using the hysteresis-loop technique for the HN-PBN (15+ 10 MPa over the entire stress range)and Syl- PBN (65 MPa at low stresses) minicomposites Hi-NiC-PBN x Also plotted in Fig, 10 are predicted stress-strain data for value greater than three times the best-fit values. The Sy Pbn predicted curves change much less, absolutely and rela- 400 tively, compared to the HN- pbn predicted curves, as t in- Hysteresis L。opP。 ak Stres$,MPa creases. This observation implies that the estimation of T from such a model is more accurate for the HN-PBN minicomposite. Fig, II of peak applied stress on the interfacial shear stress Conversely, estimating T or predicting the stress-strain curve same incomposite for the Syl-PBN minicomposite using the above-mentioned ncreasing peak-stress hysteresis loops. approach, is not very good. It will be argued later in this paper that the stress-strain behavior of the Syl-PBN minicomposite cannot be properly modeled with a constant T, because T resistance due to fiber roughness, 2 122(ii)a large interface hanges with o for the Syl-PBN minicomposite debond energy, 19,23 or(ii) overlapping sliding zones. 18,24 If a fiber-roughness effect was responsible for the increase in the 3) Dependence of T on Composite Stress apparent""T, a real increase in sliding resistance would have T was determined from the hysteresis loops just prior to occurred and T would not be constant along the sliding length. failure. For the Syl-PBN minicomposite, a relatively low-peak A large debond energy or overlapping sliding zones both as- stress hysteresis loop(o 330 MPa)resulted in a lower T sume a constant T value. The measured increase in T for these (65 MPa) compared to that for a relatively high-peak-stress two situations would be due to sliding lengths that do not hysteresis loop(T=170 MPa for p =460 MPa). There was increase with stress as expected at a given crack because the ny difference in T values determined for different debond lengths are hindered from attaining the assumed sliding ak-stress hysteresis loops of HN-PBN or Nic-3MBN mini- lengths(Eq (5), because of the high interfacial fracture energy omposites. To determine if this was a variation in stress or or interaction with another debonded zone, respectively ample-to-sample variation, T was determined for lower-peak An increase in the measured sliding resistance due to over- stress hysteresis loops of the sample that failed at 460 MPa lapping sliding zones is unlikely because most of the cracking The crack spacing was estimated from the AE energy, as de occurred near failure(the highest 5% of stress). The load on the cribed previously and in the Appendix. Figure 1l shows T alues that have been determined from several hysteresis loop were bridging cracks; therefore, the sliding lengths could only for several Syl-PBN and HN-PBN minicomposites. There is be restricted over a very small stress range. The net reduction ittle difference in T for the HN-PBN minicomposite, as a func- in loop width(increased T)would be minimal for the case of a tion of stress, whereas the Syl-PBN minicomposite shows an restricted sliding length. In addition, the le average fiber pull-out almost-linear dependence on the minicomposite peak stress lengths are almost an order of magnitude less than the smallest Several possible explanations exist for the stress-dependent matrix-crack spacings, which indicates a very short sliding dis T behavior of Syl-PBN minicomposites:(i)increased sliding tance. It is possible for short pull-out lengths to be the result of per Weib Weibull modulus for as-produced Sylramic fiber was -5.25 which is a very low value. This result is compared to a Weibull Syl-PBN Predicted Syl modulus of-8 for as-produced Hi-Nicalon fiber. 25 If the fiber/matrix interface is characterized by a large debond energy, the slope of the hysteresis loop would be linear edicted hn at the end of unloading or reloading portion of the loop. 9 If the ■t=30MPa Syl-PBN minicomposite strain scale was increased, there does not appear to be a linear region for the hysteresis loop. How- ever, because these loops are depicted at smaller strains, the Predicted HN sensitivity of the displacement measurement may be inad- equate to distinguish between a linear or parabolic-shaped hys- HN-PBN teresis loop at the end of the unloading or reloading portion the hysteresis loop. This mechanism cannot be excluded for ∧ two reasons. First, it is known that the fiber begins to decom- pose as the carbon-rich Hi-Nicalon fiber is subjected to tem- peratures >1300.C, resulting in a very thin carbon layer on the Model surface of the fiber. 26 The interphase processing temperature Matrix was 1400C for-I min, which could have resulted in a very 010203040.06070800 thin carbon layer between the fiber and the BN. Such a carbo Strain, layer would not form on Sylramic fibers, because these fibers are stoichiometric SiC and are processed at temperatures much higher than the interphase processing temperature. Therefore, IN-PBN minicomposites(the hyster. it is likely that the actual debonding interface for HN-PBN curve has been removed ) The T value was that minicomposites is in the carbon or at the carbon/SiC interface which was The actual debonding interface for Syl-PBN minicomposites

and measured at the minicomposite failure stress, respectively. Similarly, EAE(s) and EAE(sult) are the cumulative AE ener￾gies for the same sample measured at the peak hysteresis-loop stress of interest and measured at the minicomposite failure stress, respectively. Equation (6) was used to best fit the stress– strain curve for the Syl-PBN and HN-PBN minicomposites for t. Figure 10 shows the predicted and tensile data. Interfacial shear stresses of 10 and 60 MPa best fit the HN-PBN and Syl-PBN data, respectively, for the estimated number of cracks formed during the experiment. These values were very similar to those found by using the hysteresis-loop technique for the HN-PBN (15 ± 10 MPa over the entire stress range) and Syl￾PBN (65 MPa at low stresses) minicomposites. Also plotted in Fig. 10 are predicted stress–strain data for a t value greater than three times the best-fit values. The Syl￾PBN predicted curves change much less, absolutely and rela￾tively, compared to the HN-PBN predicted curves, as t in￾creases. This observation implies that the estimation of t from such a model is more accurate for the HN-PBN minicomposite. Conversely, estimating t or predicting the stress–strain curve for the Syl-PBN minicomposite using the above-mentioned approach, is not very good. It will be argued later in this paper that the stress–strain behavior of the Syl-PBN minicomposite cannot be properly modeled with a constant t, because t changes with s for the Syl-PBN minicomposite. (3) Dependence of t on Composite Stress t was determined from the hysteresis loops just prior to failure. For the Syl-PBN minicomposite, a relatively low-peak￾stress hysteresis loop (sp 4 330 MPa) resulted in a lower t (∼65 MPa) compared to that for a relatively high-peak-stress hysteresis loop (t ≈ 170 MPa for sp 4 460 MPa). There was little if any difference in t values determined for different peak-stress hysteresis loops of HN-PBN or Nic-3MBN mini￾composites. To determine if this was a variation in stress or a sample-to-sample variation, t was determined for lower-peak￾stress hysteresis loops of the sample that failed at 460 MPa. The crack spacing was estimated from the AE energy, as de￾scribed previously and in the Appendix. Figure 11 shows t values that have been determined from several hysteresis loops for several Syl-PBN and HN-PBN minicomposites. There is little difference in t for the HN-PBN minicomposite, as a func￾tion of stress, whereas the Syl-PBN minicomposite shows an almost-linear dependence on the minicomposite peak stress. Several possible explanations exist for the stress-dependent t behavior of Syl-PBN minicomposites: (i) increased sliding resistance due to fiber roughness,21,22 (ii) a large interface debond energy,19,23 or (iii) overlapping sliding zones.18,24 If a fiber-roughness effect was responsible for the increase in the ‘‘apparent’’ t, a real increase in sliding resistance would have occurred and t would not be constant along the sliding length. A large debond energy or overlapping sliding zones both as￾sume a constant t value. The measured increase in t for these two situations would be due to sliding lengths that do not increase with stress as expected at a given crack because the debond lengths are hindered from attaining the assumed sliding lengths (Eq. (5)), because of the high interfacial fracture energy or interaction with another debonded zone, respectively. An increase in the measured sliding resistance due to over￾lapping sliding zones is unlikely because most of the cracking occurred near failure (the highest 5% of stress). The load on the fibers did not increase significantly for most of the fibers that were bridging cracks; therefore, the sliding lengths could only be restricted over a very small stress range. The net reduction in loop width (increased t) would be minimal for the case of a restricted sliding length. In addition, the average fiber pull-out lengths are almost an order of magnitude less than the smallest matrix-crack spacings, which indicates a very short sliding dis￾tance. It is possible for short pull-out lengths to be the result of a low t value and high fiber Weibull modulus. However, the Weibull modulus for as-produced Sylramic fiber was ∼5,25 which is a very low value. This result is compared to a Weibull modulus of ∼8 for as-produced Hi-Nicalon fiber.25 If the fiber/matrix interface is characterized by a large debond energy, the slope of the hysteresis loop would be linear at the end of unloading or reloading portion of the loop.19 If the Syl-PBN minicomposite strain scale was increased, there does not appear to be a linear region for the hysteresis loop. How￾ever, because these loops are depicted at smaller strains, the sensitivity of the displacement measurement may be inad￾equate to distinguish between a linear or parabolic-shaped hys￾teresis loop at the end of the unloading or reloading portion of the hysteresis loop. This mechanism cannot be excluded for two reasons. First, it is known that the fiber begins to decom￾pose as the carbon-rich Hi-Nicalon fiber is subjected to tem￾peratures >1300°C, resulting in a very thin carbon layer on the surface of the fiber.26 The interphase processing temperature was 1400°C for ∼1 min,16 which could have resulted in a very thin carbon layer between the fiber and the BN. Such a carbon layer would not form on Sylramic fibers, because these fibers are stoichiometric SiC and are processed at temperatures much higher than the interphase processing temperature.3 Therefore, it is likely that the actual debonding interface for HN-PBN minicomposites is in the carbon or at the carbon/SiC interface. The actual debonding interface for Syl-PBN minicomposites Fig. 10. Measured (solid lines) and modeled (data points) stress– strain behavior of Syl-PBN and HN-PBN minicomposites (the hyster￾esis portion of the curve has been removed). The t value was that which was used in the model. Fig. 11. Effect of peak applied stress on the interfacial shear stress. Each set of data points has been taken from the same minicomposite tensile test for increasing peak-stress hysteresis loops. January 1999 Fiber Effects on Minicomposite Mechanical Properties for Several SiC/CVI-SiC Matrix Systems 151

Journal of the American Ceramic Society-Morscher and Martines-Fernandes Vol. 82. Ne Debond Crack-tip Opening BridgingFiber Matrix Matrix Fig. 12. Illustration of the effect of interfacial roughness during progressive debonding away from a matrix crack in a composite under tension Figure 2 from Parthasarathy and Kerans, 22 reprinted with permission occurs in the PBn interphase or at the PBN/SiC interface. Such Il is characterized by an increasing T value, whereas Region Ill characterized by a constant T value. Depending on the sliding infiltrated SiC composites with BN interphases where the ma- lengths that correspond to Region ll or Region Ill for a given trix processing temperature is-1410oC.sEcond, the rougher crack, the value of T can increase or decrease as the sliding Sylramic fiber(as described later in this paper) can create length and stress increases. 22 If the total sliding distance--or at more-tortuous interface crack, which would require more en- least most of the sliding distance--was Region II for the Syl ergy to propagate, in comparison to an interface crack that is PBN minicomposite, an increasing T value would result with associated with a smooth fiber surface Increasing stress, Parthasarathy and Kerans have shown that if the fiber The difference in fiber roughness between the Hi-Nicalon roughness is larger than the Poisson contraction, the sliding and Sylramic fiber types is shown in Fig. 14 for uncoated, resistance can increase as the stress or sliding length increases. as-produced fibers. Atomic force microscopy of Sylramic and This relation is shown in Fig. 12. 2 There would be a continu- ously increasing resistance to sliding for a region of a debonded nan27 and Chawla et al. 8 respectively. For the Sylramic fiber, fiber near the interface crack tip where the actual distance that the mean, root mean square(RMS), and peak amplitude (h) a point on the fiber surface slid is less than the characteristic were 69, 18, and 90 nm, respectively. For Hi-Nicalon fiber, oughness of the fiber(Region lI in Fig. 12). For the section of the mean. RMS. and h values were 43. 5.4. and 43 nm the fiber that has slid a distance greater than the characteristic roughness of the fiber surface, near the matrix crack, the sliding The larger fiber roughness of Sylramic fiber must contribute resistance is constant(Region Ill in Fig. 12). Parthasarathy and in large part to the stress-dependent T behavior that is observed Kerans have modeled such behavior, assuming a"sawtooth for Sylramic minicomposites, because of the increased resis- fiber roughness. Figure 13 shows the expected T behavior for tance to sliding with increasing sliding length. It is also pos- an individual fiber during pullout for the two regions. Region sible that the debond energy is greater for the Sylramic mini- interfacial Friction stress RegionⅢl From Unseating Siding Fnction j Roughness-induced Friction for s> d u{A(△aT} 8=d(half-period) FL Relative Fiber/Matrix Displacement (δ) Fig. 13. Schematic sketch of the effect of progressive roughness on the interfacial friction stress at a selected point on the fiber.( Figure 3 from Parthasarathy and Kerans, 2 reprinted with permission

occurs in the PBN interphase or at the PBN/SiC interface. Such a carbon layer has been observed for Hi-Nicalon, melt￾infiltrated SiC composites with BN interphases where the ma￾trix processing temperature is ∼1410°C.27 Second, the rougher Sylramic fiber (as described later in this paper) can create a more-tortuous interface crack, which would require more en￾ergy to propagate, in comparison to an interface crack that is associated with a smooth fiber surface. Parthasarathy and Kerans22 have shown that if the fiber roughness is larger than the Poisson contraction, the sliding resistance can increase as the stress or sliding length increases. This relation is shown in Fig. 12.22 There would be a continu￾ously increasing resistance to sliding for a region of a debonded fiber near the interface crack tip where the actual distance that a point on the fiber surface slid is less than the characteristic roughness of the fiber (Region II in Fig. 12). For the section of the fiber that has slid a distance greater than the characteristic roughness of the fiber surface, near the matrix crack, the sliding resistance is constant (Region III in Fig. 12). Parthasarathy and Kerans22 have modeled such behavior, assuming a ‘‘sawtooth’’ fiber roughness. Figure 13 shows the expected t behavior for an individual fiber during pullout for the two regions. Region II is characterized by an increasing t value, whereas Region III is characterized by a constant t value. Depending on the sliding lengths that correspond to Region II or Region III for a given crack, the value of t can increase or decrease as the sliding length and stress increases.22 If the total sliding distance—or at least most of the sliding distance—was Region II for the Syl￾PBN minicomposite, an increasing t value would result with increasing stress. The difference in fiber roughness between the Hi-Nicalon and Sylramic fiber types is shown in Fig. 14 for uncoated, as-produced fibers. Atomic force microscopy of Sylramic and Hi-Nicalon as-produced fiber has been performed by Bren￾nan27 and Chawla et al., 28 respectively. For the Sylramic fiber, the mean, root mean square (RMS), and peak amplitude (h) were 69, 18, and 90 nm, respectively. For Hi-Nicalon fiber, the mean, RMS, and h values were 4.3, 5.4, and 43 nm, respectively. The larger fiber roughness of Sylramic fiber must contribute in large part to the stress-dependent t behavior that is observed for Sylramic minicomposites, because of the increased resis￾tance to sliding with increasing sliding length. It is also pos￾sible that the debond energy is greater for the Sylramic mini￾Fig. 12. Illustration of the effect of interfacial roughness during progressive debonding away from a matrix crack in a composite under tension. (Figure 2 from Parthasarathy and Kerans,22 reprinted with permission.) Fig. 13. Schematic sketch of the effect of progressive roughness on the interfacial friction stress at a selected point on the fiber. (Figure 3 from Parthasarathy and Kerans,22 reprinted with permission.) 152 Journal of the American Ceramic Society—Morscher and Martinez-Fernandez Vol. 82, No. 1

January 1999 Fiber Effects on Minicomposite Mechanical Properties for Several SiC/cvl-SiC Matrix Syster 3日-M白Y This observation would mean that for every crack formed, the amount of matrix around a crack that is incapable of further cracking is much greater for the Nic-3MBN and HN-PBN minicomposites than for the Syl-PBN minicomposites(from Eq(5) By combining these two factors, the local nature of cracking in Syl-PBN minicomposites can be deduced. Initial cracking in all three minicomposites occurs at the same stress(Fig. 3), because of the large flaws in the matrix. Most likely, there is some variabl ity in the matrix volume fraction along the length, and initial fracture occurs at regions with greater matrix frac- tions(this was observed for single-fiber microcomposites) Cracking eventually stops in these weak matrix regions for the Nic-3MBN and HN-PBN minicomposites, because the low in- terfacial shear stress blocks"large"regions of matrix from greater amounts of stress. However, cracking occurs in other egions of the matrix with smaller flaws as the matrix stress is HNAP 5 gapm increased. For the Syl-PBN minicomposite, the high interfacial shear stress enabled a larger portion of the weak matrix regions to be stressed, so cracking was concentrated more locally along the length of the minicomposite Matrix stresses that were high enough to crack/saturate other portions of the matrix were The local nature of cracking for the Syl-PBN minicomposite as well as the lower strain to failure of this fiber type, explains he lower strain to failure of this minicomposite system. Be- cause only a fraction of the matrix is actually cracked over the entire gauge length, and the sliding length is relatively small at the ultimate failure stress and strain. To obtain larger ded at each crack, a relatively small length of fiber is fully trains with this composite, lower interfacial shear st lower matrix cracking stresses, and/or higher fiber sti must be achieved o The minicomposite test has been proven to be very effective evaluating the tensile stress-strain behavior of several dif- SYLAP 5.0 kV x6.00K 5.00H ferent types of SiC fiber/CVI-SiC matrix systems. The ultimate strength, ultimate strain, and interfacial properties could be directly correlated with the fiber strengths, fiber moduli, and Fig. 14. SEM micrographs of fiber surfaces for(a)Hi-Nicalon and fiber roughnesses for the three systems studied. It was evident (b) Sylramic as-produced fibers that the lower strain to failure and large interfacial shear stresses measured for the Sylramic minicomposites were re- spectively due to the high modulus and rough surface of this composites, because of the fiber surface and/or the fiber. This result may have good and bad implications for com- lack of a carbon layer at th iber interface. A constant osite use. On the negative side higher strain to failures would interfacial shear stress co account for the stress be desired for most composite applications that require com- dependent T behavior and ot adequately model Syl- ant behavior On the positive side the smaller crack spacing ramic minicomposite behavior (and presumably smaller crack openings) could potentially have benefits for high-temperature applications if the access of (4) Matrix-Crack Spacing the environment to the interphase region was hindered, because Two factors account for cracking behavior in composites smaller matrix cracks"seal"'faster(compared to larger crack First, the matrix has a matrix-flaw distribution that is quite openings)from reactions between the matrix and the environ- large and accounts for the large stress range over which crack ment. In addition, the higher-T, higher-modulus Sylramic fiber ing occurs(Fig. 3 and Morscher et al. 1). Figure 3(a)shows system offers a higher stiffness material after damage, which that it is reasonable to assume that the matrix -flaw distribution may be important for some applications. These implications in the Syl-PBN and HN-PBN minicomposites is approximately will have to be considered and studied when implementing the same, because the AE activity of both minicomposites f these fiber types in a specific composite system for a specific lows the same trend with composite stress(matrix stress). The Nic-3MBN minicomposite may have a different matrix-flaw distribution. The coated-Nicalon tow already had a Sic layer APPENDIX on top of the BN coating prior to CVI-SiC infiltration, which or matrix infiltration and a thicker SiC region Estimation of the number of cracks from the exterior of the minicomposite(Fig. 4). This thicker SiC Acoustic Emission ion most likely resulted in a greater density of larger fl: Unfortunately, matrix cracks were not visible via stereom- which resulted in more cracking at lower stresses. Second, the croscopy during the tensile test. To determine the value of T, sliding length determines the amount of matrix around a crack the number of cracks for a given hysteresis-loop peak stress that can still be cracked. 2 The interfacial shear stress of the must be known, because crack saturation for most of the hys- Syl-PBN minicomposite, although probably not constant over teresis-loop peak stresses had not been reached. If the cumu- g length, is significantly greater than the interfacial lative AE energy could be related to the cracks produced, M ss of the Nic-3MBN and HN-PBN minicomposites. (Eq(6))could be approximated for a given hy

composites, because of the rougher fiber surface and/or the lack of a carbon layer at the BN/fiber interface. A constant interfacial shear stress could not account for the stress￾dependent t behavior and would not adequately model Syl￾ramic minicomposite behavior. (4) Matrix-Crack Spacing Two factors account for cracking behavior in composites. First, the matrix has a matrix-flaw distribution that is quite large and accounts for the large stress range over which crack￾ing occurs (Fig. 3 and Morscher et al.11). Figure 3(a) shows that it is reasonable to assume that the matrix-flaw distribution in the Syl-PBN and HN-PBN minicomposites is approximately the same, because the AE activity of both minicomposites fol￾lows the same trend with composite stress (matrix stress). The Nic-3MBN minicomposite may have a different matrix-flaw distribution. The coated-Nicalon tow already had a SiC layer on top of the BN coating prior to CVI-SiC infiltration, which resulted in poor matrix infiltration and a thicker SiC region on the exterior of the minicomposite (Fig. 4). This thicker SiC region most likely resulted in a greater density of larger flaws, which resulted in more cracking at lower stresses. Second, the sliding length determines the amount of matrix around a crack that can still be cracked.24 The interfacial shear stress of the Syl-PBN minicomposite, although probably not constant over the sliding length, is significantly greater than the interfacial shear stress of the Nic-3MBN and HN-PBN minicomposites. This observation would mean that for every crack formed, the amount of matrix around a crack that is incapable of further cracking is much greater for the Nic-3MBN and HN-PBN minicomposites than for the Syl-PBN minicomposites (from Eq. (5)). By combining these two factors, the local nature of cracking in Syl-PBN minicomposites can be deduced. Initial cracking in all three minicomposites occurs at the same stress (Fig. 3), because of the large flaws in the matrix. Most likely, there is some variability in the matrix volume fraction along the length, and initial fracture occurs at regions with greater matrix frac￾tions (this was observed for single-fiber microcomposites29). Cracking eventually stops in these weak matrix regions for the Nic-3MBN and HN-PBN minicomposites, because the low in￾terfacial shear stress blocks ‘‘large’’ regions of matrix from greater amounts of stress. However, cracking occurs in other regions of the matrix with smaller flaws as the matrix stress is increased. For the Syl-PBN minicomposite, the high interfacial shear stress enabled a larger portion of the weak matrix regions to be stressed, so cracking was concentrated more locally along the length of the minicomposite. Matrix stresses that were high enough to crack/saturate other portions of the matrix were never achieved. The local nature of cracking for the Syl-PBN minicomposite, as well as the lower strain to failure of this fiber type, explains the lower strain to failure of this minicomposite system. Be￾cause only a fraction of the matrix is actually cracked over the entire gauge length, and the sliding length is relatively small at each crack, a relatively small length of fiber is fully loaded at the ultimate failure stress and strain. To obtain larger failure strains with this composite, lower interfacial shear stresses, lower matrix cracking stresses, and/or higher fiber strengths must be achieved. V. Conclusions The minicomposite test has been proven to be very effective for evaluating the tensile stress–strain behavior of several dif￾ferent types of SiC fiber/CVI-SiC matrix systems. The ultimate strength, ultimate strain, and interfacial properties could be directly correlated with the fiber strengths, fiber moduli, and fiber roughnesses for the three systems studied. It was evident that the lower strain to failure and large interfacial shear stresses measured for the Sylramic minicomposites were re￾spectively due to the high modulus and rough surface of this fiber. This result may have good and bad implications for com￾posite use. On the negative side, higher strain to failures would be desired for most composite applications that require com￾pliant behavior. On the positive side, the smaller crack spacing (and presumably smaller crack openings) could potentially have benefits for high-temperature applications if the access of the environment to the interphase region was hindered, because smaller matrix cracks ‘‘seal’’ faster (compared to larger crack openings) from reactions between the matrix and the environ￾ment. In addition, the higher-t, higher-modulus Sylramic fiber system offers a higher stiffness material after damage, which may be important for some applications. These implications will have to be considered and studied when implementing these fiber types in a specific composite system for a specific application. APPENDIX Estimation of the Number of Cracks from Acoustic Emission Unfortunately, matrix cracks were not visible via stereomi￾croscopy during the tensile test. To determine the value of t, the number of cracks for a given hysteresis-loop peak stress must be known, because crack saturation for most of the hys￾teresis-loop peak stresses had not been reached. If the cumu￾lative AE energy could be related to the cracks produced, N (Eq. (6)) could be approximated for a given hysteresis-loop Fig. 14. SEM micrographs of fiber surfaces for (a) Hi-Nicalon and (b) Sylramic as-produced fibers. January 1999 Fiber Effects on Minicomposite Mechanical Properties for Several SiC/CVI-SiC Matrix Systems 153

Journal of the American Ceramic Society-Morscher and Martines-Fernandes Vol. 82. Ne stress. For single-fiber microcomposites, a direct relationship 12E+00 existed between the AE energy and number of cracks pre luced. Figure al shows the data from a single-fiber SCS-6 CVD-SiC fiber with a diameter of 143 um, Textron Specialt 1E+006 Materials, Lowell, MA)microcomposite tensile test.29The same tensile-test setup as that described by Morscher et al 800000 was used. The load was increased to a set load and held at that Nic-3MBN load. The cracks were opened and could be observed via ste Hi-Ni reomicroscopy. The position of each crack and the cumulative PBN AE energy at the load were recorded. The load was then in- creased and held, and the crack position and AE energy were again recorded. This procedure was continued up to a load of 45N. which did not cause failure. The number of cracks and AE energy was normalized by the normalize ea cracl 20000 mulative ae energy at 45 N and plotted in Fig. Al. The not malized number of cracks and aE energy, as a function of load Syl-PBN are almost identical, indicating a linear relationship between the ae energy and the number of cracks that are formed. the 050100150200250300350 epoxy mounting and gripping technique for this microcompos of Cracks In Gauge ite test was exactly the same as used for the estimated from crack spacing cause matrix cracks were not observable on the surface Fig. A2, nship between crack formation and AE energy fo of minicomposites, the relationship between AE energy and minicomposites. Each data point corresponds to a differe the number of cracks must be determined indirectly, i. e, after the tensile test. The crack spacing at failure was determined for each minicomposite from polished longitudinal sections of a major portion of the gauge section. The number of and interface sliding events to the cumulative AE energy is cracks formed during the tensile test in the gauge section was very small. On work with macrocomposite where it was then approximated from the crack spacing of the polished por possible to convincingly separate fiber-failure events from ma- trix-crack events, the fiber-failure events had amplitudes and failure. It was assumed that the energy of an ae source event energies that were orders of magnitude lower than the higher- from a matrix crack was proportional to the area of the mini- omposite. Therefore, absolute cumulative AE energy normal ergy approach was taken in the macrocomposite work that was ed by the minicomposite area was plotted versus the number ed in this study fiber failure would only account for-2.5% of cracks in the gauge, as shown in Fig. A2. Each data point in the total cumulative energy of the tensile test after failure Fig. A2 represents a different minicomposite that has been Matrix cracking would have accounted for at least 85% of the tested to failure. There does seem to be a linear relationsh total cumulative energy of the tensile test. This result is rea- (the dashed line in Fig. A2)between the AE energy and the sonable because the surface energy that is created from a ma- umber of cracks formed for minicomposites. However, there cantly larger than that produced by an in- significant scatter in this data. the greatest source of this dividual fiber failure. For the minicomposites, AE events that scatter is probably the inconsistency in the ae signal intensi- occur near failure( which would most likely be individual fiber es, as measured by the transducers through the epoxy for each failures)for all the minicomposites that have been tested in this mple. Because the epoxy surface is not flat, the measured study are much-lower-energy events than the events that occur signal intensity will be dependent on the transducer contact earlier in the stress-strain curve. Therefore, this assumption is area and the epoxy thickness, which varied from sample to considered to be reasonable The assumption that the ae energy is directly related to the Acknowledgments: The authors wish to thank Drs. Stanley Levine and number of cracks implies that the contribution of fiber failure and Drs. Ti Force Base(Wright-Patterson AFB, OH) for their input on the patterson Air Parthasarathy and Ron Kerans of roughness on interfacial shear stress Normalized References A. DiCarlo, ""Creep Limitations of Current Polycrystalline Ceramic Fi 0.8 Fof Cracks 比mmA小 of the l ww kasai, T. Seguchi, and K. Stoichiometric B-SiC Com- Ceram. Eng. Sci. Proc., 18 3 147-57(1997) Kumagawa, H. Yamaoka, M. Shibuya, and T. Yamamura,""Thermal and Chemical Corrosion Resistance of Newly Developed normalized -O Tyranno Fiber, Ceram. Eng. Sci. Proc., 18 3) 113-18 SN. Lissart and J. Lamon, ""Analysis of Damage Failure in Model Temperature Ceramic-Matrix Composites I. Edited by A G. Evans and bs. T. Gonczy, R C. Sprandel, and K. T. Faber, ""Tensile Tests of Miniature Fiber Reinforced Ceramic Composites 0152025303540 Tensile Load, N with Carbon and Boron Nitride Interphases at Elevated Temperatures in Air J Am Ceram Soc., 80 18 N. Morscher. *The Effect of Static and Cvelic Tensile Stress and Te Fig. Al. Relationship between crack formation and AE energy for a perature on Failure for Precracked Hi-Nicalon/BN/CVD SiC Minicomposites in Air, Ceram. Eng. Sci. Proc., 183]737-45(1997

stress. For single-fiber microcomposites, a direct relationship existed between the AE energy and number of cracks pro￾duced. Figure A1 shows the data from a single-fiber SCS-6 (CVD-SiC fiber with a diameter of 143 mm, Textron Specialty Materials, Lowell, MA) microcomposite tensile test.29 The same tensile-test setup as that described by Morscher et al.11 was used. The load was increased to a set load and held at that load. The cracks were opened and could be observed via ste￾reomicroscopy. The position of each crack and the cumulative AE energy at the load were recorded. The load was then in￾creased and held, and the crack position and AE energy were again recorded. This procedure was continued up to a load of 45 N, which did not cause failure. The number of cracks and AE energy was normalized by the normalized cracks and cu￾mulative AE energy at 45 N and plotted in Fig. A1. The nor￾malized number of cracks and AE energy, as a function of load, are almost identical, indicating a linear relationship between the AE energy and the number of cracks that are formed. The epoxy mounting and gripping technique for this microcompos￾ite test was exactly the same as used for the minicomposites of this study. Because matrix cracks were not observable on the surface of minicomposites, the relationship between AE energy and the number of cracks must be determined indirectly, i.e., after the tensile test. The crack spacing at failure was determined for each minicomposite from polished longitudinal sections of a major portion of the gauge section. The number of cracks formed during the tensile test in the gauge section was then approximated from the crack spacing of the polished por￾tion of the gauge length for the entire tested gauge length at failure. It was assumed that the energy of an AE source event from a matrix crack was proportional to the area of the mini￾composite. Therefore, absolute cumulative AE energy normal￾ized by the minicomposite area was plotted versus the number of cracks in the gauge, as shown in Fig. A2. Each data point in Fig. A2 represents a different minicomposite that has been tested to failure. There does seem to be a linear relationship (the dashed line in Fig. A2) between the AE energy and the number of cracks formed for minicomposites. However, there is significant scatter in this data. The greatest source of this scatter is probably the inconsistency in the AE signal intensi￾ties, as measured by the transducers through the epoxy for each sample. Because the epoxy surface is not flat, the measured signal intensity will be dependent on the transducer contact area and the epoxy thickness, which varied from sample to sample. The assumption that the AE energy is directly related to the number of cracks implies that the contribution of fiber failure and interface sliding events to the cumulative AE energy is very small. On work with macrocomposites where it was possible to convincingly separate fiber-failure events from ma￾trix-crack events, the fiber-failure events had amplitudes and energies that were orders of magnitude lower than the higher￾energy matrix-cracking events.30 If the same cumulative en￾ergy approach was taken in the macrocomposite work that was used in this study, fiber failure would only account for ∼2.5% of the total cumulative energy of the tensile test after failure. Matrix cracking would have accounted for at least 85% of the total cumulative energy of the tensile test. This result is rea￾sonable because the surface energy that is created from a ma￾trix crack is significantly larger than that produced by an in￾dividual fiber failure. For the minicomposites, AE events that occur near failure (which would most likely be individual fiber failures) for all the minicomposites that have been tested in this study are much-lower-energy events than the events that occur earlier in the stress–strain curve. Therefore, this assumption is considered to be reasonable. Acknowledgments: The authors wish to thank Drs. Stanley Levine and James DiCarlo of NASA–Lewis Research Center for review of the manuscript and Drs. Triplicane Parthasarathy and Ron Kerans of Wright-Patterson Air Force Base (Wright-Patterson AFB, OH) for their input on the effects of fiber roughness on interfacial shear stress. References 1 J. A. DiCarlo, ‘‘Creep Limitations of Current Polycrystalline Ceramic Fi￾bers,’’ Compos. Sci. Technol., 51, 213–22 (1994). 2 M. Takeda, Y. Imai, H. Ichikawa, T. Ishikawa, N. Kasai, T. Seguchi, and K. Okamura, ‘‘Thermomechanical Analysis of the Low Oxygen Silicon Carbide Fibers Derived from Polycarbosilane,’’ Ceram. Eng. Sci. Proc., 14 [7–8] 540– 47 (1993). 3 J. Lipowitz, J. A. Rabe, A. Zangvil, and Y. Xu, ‘‘Structure and Properties of Sylramic Silicon Carbide Fiber—A Polycrystalline, Stoichiometric b-SiC Com￾position,’’ Ceram. Eng. Sci. Proc., 18 [3] 147–57 (1997). 4 K. Kumagawa, H. Yamaoka, M. Shibuya, and T. Yamamura, ‘‘Thermal Stability and Chemical Corrosion Resistance of Newly Developed Continuous Si-Zr-C-O Tyranno Fiber,’’ Ceram. Eng. Sci. Proc., 18 [3] 113–18 (1997). 5 N. Lissart and J. Lamon, ‘‘Analysis of Damage Failure in Model Unidirec￾tional CVI Composites’’; pp. 241–46 in Ceramic Transactions, Vol. 57, High Temperature Ceramic-Matrix Composites I. Edited by A. G. Evans and R. Naslain. American Ceramic Society, Westerville, OH, 1995. 6 S. T. Gonczy, R. C. Sprandel, and K. T. Faber, ‘‘Tensile Tests of Miniature Fiber Reinforced Ceramic Composites for Screening Process-Property Rela￾tions,’’ Ceram. Eng. Sci. Proc., 18 [3] 729–36 (1997). 7 G. N. Morscher, ‘‘Tensile Stress-Rupture of SiCf /SiCm Minicomposites with Carbon and Boron Nitride Interphases at Elevated Temperatures in Air,’’ J. Am. Ceram. Soc., 80 [8] 2029–42 (1997). 8 G. N. Morscher, ‘‘The Effect of Static and Cyclic Tensile Stress and Tem￾perature on Failure for Precracked Hi-Nicalon/BN/CVD SiC Minicomposites in Air,’’ Ceram. Eng. Sci. Proc., 18 [3] 737–45 (1997). Fig. A1. Relationship between crack formation and AE energy for a single-fiber SCS-6 minicomposite. Fig. A2. Relationship between crack formation and AE energy for several single-tow minicomposites. Each data point corresponds to a different minicomposite. 154 Journal of the American Ceramic Society—Morscher and Martinez-Fernandez Vol. 82, No. 1