J.Am.cerm.Soc.83|6]1441-49(2000 urna Intermediate-Temperature Stress Rupture of a Woven Hi-Nicalon BN-Interphase, siC-Matrix Composite in Air Gregory N Morscher*f Ohio Aerospace Institute, Cleveland, Ohio 44 14 Janet Hurst* and David Brewer NASA Glenn Research Center. Cleveland Ohio 44135 Woven Hi-Nicalon"-reinforeed melt-infiltrated Sic-matrix composites when the matrix is cracked in an oxidizing environ- composites were tested under tensile stress-rupture conditions ment is proportional to time with an exponent of --y4(ro in air at intermediate temperatures. A comprehensive exami- which would be expected if the critical flaw was controlled nation of the damage state and the fiber properties at failure by passive oxidation layer growth on a SiC fiber. I4 With these was performed. Modal acoustic emission analysis was used to time exponents, C-interphase composites have been observed to monitor damage during the experiment. Extensive microscopy have relatively short lives at intermediate temperatures once the of the composite fracture surfaces and the individual fiber matrix is cracked and the interphases separating the load-bearing fracture surfaces was used to determine the mechanisms leading to ultimate failure. The rupture properties of these fibers from the matrix have been removed omposites were significantly worse than expected compared BN-interphase systems (single-tow CVI Sic minicomposites le to the oxidation of the BN interphase. Oxidation oc tested in tension"and CVI SiC-matrix composites tested in through the matrix cracks that intersected the surfa flexure)are superior to C- interphase systems due to the greater e or edge of a tensile bar. These oxidation reactions resulted in strong tability of bn in oxidizing environments. The time exponents onding of the fibers to one another at regions of near determined for single-tow minicomposites tested at 700. and fiber-to-fiber contact. It was found that two regimes for 950C in air range from -0.025 to -0.038, respectively. This rupture exist for this material: a high-stress regime where rder-of-magnitude reduction in time exponent would result in rupture occurs at a fast rate and a low-stress regime where hundreds or even thousands of hours improvement in lifetimes for rupture occurs at a slower rate. For the high-stress regime, the stresses above the matrix cracking stress of SiC/Sic composites matrix damage state consisted of through-thickness cracks. compared with composites with a carbon interphase The average fracture strength of fibers that were pulled out For these systems, a relatively short(micrometers)distance of (the final fibers to break before ultimate failure) was con nterphase is recessed along the length of the fibers away from the trolled by the slow-crack-growth rupture criterion in the matrix crack, compared with C(millimeters ). BN reacts with O iterature for individual Hi-Nicalon fibers For the low-stress and H,O in the environment because of the formation of gaseous egime, the matrix damage state consisted of microcracks (H-B-O containing gases) and condensed-phase oxidation prod- which grew during the rupture test. The average fracture ucts(B2O, liquid or borosilicate liquids or glasses). 5, 6 Yet, this strength of fibers that were pulled out in this regime was the recession and reaction product formation still cause fiber degra- same as the average fracture strength of individual fibers dation" and/or fiber-matrix bonding resulting in local stress pulled out in as-produced composites tested at room temper concentratorson the fibers in the region of the matrix crack ature In this study, the intermediate-temperature rupture properties of BN-interphase SiC/Sic composites will be extended to a melt- infiltrated(Mi) matrix composite. MI composites are much denser than CVI-matrix composites and as a result have higher matrix cracking stresses than CVI-SiC-matrix composites. The tensile susceptibility of Sic-fiber-reinforced Sic-matrix compos- es to strength degradation at intermediate stress rupture properties of a woven Hi-Nicalon fiber, BN- (~500°tolo0°C) and constant applied stress nterphase, MI SiC-matrix composite was determined at interme- environments is well known-o The root cause of th diate temperatures in air. Some rupture data for this composite system have been taken by Brewer et al., who have tested this ment"is due to the reaction of the oxidizing environment with the material under low-cycle-fatigue(LCF)conditions(2 h hold stress ber-matrix interphase material For carbon interphase systems 1-4, 6-8, 10 the environmental followed by an unload-reload cycle)at 815.C in air. reactions with the interphase torms a gap between the fiber and the time will be the matrix damage state for the given applied stress and matrix oxidize. 1-13 The stress-rupture strength, or, of these composite system s occurs over a range of stress. The onset of microcrack formation in the large matrix regions and 90 bundles (tows) occurs at a low stress (-75 MPa). As stress is in- creased,20,2 microcracks grow and/or new cracks are formed that T.A. Parthasarathy-contributing editor ntersect load-bearing bundles, which eventually results in the formation of through-thickness cracks with increasing stress. The damage accumulation for this material system was fairly wel antified with modal acoustic emission(AE) monitoring ipt No 189507 Received March 11, 1999, approved November 30, 1999. with unload-reload tensile hysteresis experiments at room temper ber. American Ceramic or Research associate at NASA Glenn Research Center. Cleveland. Ohio. ature.In this study, AE monitoring similar to that performed for
Intermediate-Temperature Stress Rupture of a Woven Hi-Nicalon, BN-Interphase, SiC-Matrix Composite in Air Gregory N. Morscher* ,† Ohio Aerospace Institute, Cleveland, Ohio 44142 Janet Hurst* and David Brewer NASA Glenn Research Center, Cleveland, Ohio 44135 Woven Hi-NicalonTM-reinforced melt-infiltrated SiC-matrix composites were tested under tensile stress-rupture conditions in air at intermediate temperatures. A comprehensive examination of the damage state and the fiber properties at failure was performed. Modal acoustic emission analysis was used to monitor damage during the experiment. Extensive microscopy of the composite fracture surfaces and the individual fiber fracture surfaces was used to determine the mechanisms leading to ultimate failure. The rupture properties of these composites were significantly worse than expected compared with the fiber properties under similar conditions. This was due to the oxidation of the BN interphase. Oxidation occurred through the matrix cracks that intersected the surface or edge of a tensile bar. These oxidation reactions resulted in strong bonding of the fibers to one another at regions of near fiber-to-fiber contact. It was found that two regimes for rupture exist for this material: a high-stress regime where rupture occurs at a fast rate and a low-stress regime where rupture occurs at a slower rate. For the high-stress regime, the matrix damage state consisted of through-thickness cracks. The average fracture strength of fibers that were pulled out (the final fibers to break before ultimate failure) was controlled by the slow-crack-growth rupture criterion in the literature for individual Hi-Nicalon fibers. For the low-stress regime, the matrix damage state consisted of microcracks which grew during the rupture test. The average fracture strength of fibers that were pulled out in this regime was the same as the average fracture strength of individual fibers pulled out in as-produced composites tested at room temperature. I. Introduction THE susceptibility of SiC-fiber-reinforced SiC-matrix composites to strength degradation at intermediate temperatures (;500° to 1000°C) and constant applied stress in oxidizing environments is well known.1–10 The root cause of this “embrittlement” is due to the reaction of the oxidizing environment with the fiber–matrix interphase material. For carbon interphase systems,1–4,6–8,10 the environmental reactions with the interphase forms a gap between the fiber and the matrix which can be filled with SiO2 reaction product as the fiber and matrix oxidize.11–13 The stress-rupture strength, sr , of these composites when the matrix is cracked6 in an oxidizing environment is proportional to time with an exponent of ;21⁄4 (sr } t 21/4) which would be expected if the critical flaw was controlled by passive oxidation layer growth on a SiC fiber.8,14 With these time exponents, C-interphase composites have been observed to have relatively short lives at intermediate temperatures once the matrix is cracked and the interphases separating the load-bearing fibers from the matrix have been removed. It has been demonstrated that the stress-rupture properties of BN-interphase systems (single-tow CVI SiC minicomposites tested in tension4,5 and CVI SiC-matrix composites tested in flexure9 ) are superior to C-interphase systems due to the greater stability of BN in oxidizing environments. The time exponents determined for single-tow minicomposites tested at 700° and 950°C in air range from 20.025 to 20.038, respectively.4,5 This order-of-magnitude reduction in time exponent would result in hundreds or even thousands of hours improvement in lifetimes for stresses above the matrix cracking stress of SiC/SiC composites compared with composites with a carbon interphase. For these systems, a relatively short (micrometers) distance of interphase is recessed along the length of the fibers away from the matrix crack, compared with C (millimeters). BN reacts with O2 and H2O in the environment because of the formation of gaseous (H–B–O containing gases) and condensed-phase oxidation products (B2O3 liquid or borosilicate liquids or glasses).15,16 Yet, this recession and reaction product formation still cause fiber degradation4 and/or fiber–matrix bonding resulting in local stress concentrators17 on the fibers in the region of the matrix crack. In this study, the intermediate-temperature rupture properties of BN-interphase SiC/SiC composites will be extended to a meltinfiltrated (MI) matrix composite. MI composites are much denser than CVI-matrix composites and as a result have higher matrix cracking stresses than CVI-SiC-matrix composites.18 The tensile stress rupture properties of a woven Hi-Nicalon fiber, BNinterphase, MI SiC-matrix composite was determined at intermediate temperatures in air. Some rupture data for this composite system have been taken by Brewer et al., 19 who have tested this material under low-cycle-fatigue (LCF) conditions (2 h hold stress followed by an unload–reload cycle) at 815°C in air. An important consideration for understanding composite lifetime will be the matrix damage state for the given applied stress. Damage accumulation for this Hi-Nicalon/BN/MI SiC woven composite system18 occurs over a range of stress. The onset of microcrack formation in the large matrix regions and 90° bundles (tows) occurs at a low stress (;75 MPa). As stress is increased,20,21 microcracks grow and/or new cracks are formed that intersect load-bearing bundles, which eventually results in the formation of through-thickness cracks with increasing stress. The damage accumulation for this material system was fairly well quantified with modal acoustic emission (AE) monitoring coupled with unload–reload tensile hysteresis experiments at room temperature.18 In this study, AE monitoring similar to that performed for T. A. Parthasarathy—contributing editor Manuscript No. 189507. Received March 11, 1999; approved November 30, 1999. *Member, American Ceramic Society. † Senior Research Associate at NASA Glenn Research Center, Cleveland, Ohio. J. Am. Ceram. Soc., 83 [6] 1441–49 (2000) 1441 journal
Journal of the American Ceramic Society-Morscher et al. ne room-temperature experiments will be performed during the field emission scanning electron microscopy(FESEM)(Hitachi stress-rupture experiments to locate and quantify the extent of $4500, Tokyo, Japan). For the FESEM, a voltage of 5 kv was used, which required no conductive coating. General observation rere made on the fracture surfaces. Individual fiber fracture surfaces were also obtained from various regions of the fracture Il. Experimental Procedure surface to perform fiber fractography analysis. Electron dispersive Several properties of the composite constituents are described in spectroscopy(EDS)(Kevex 4460 pulse processor with Quantum Table I. The material consisted of eight plies of woven 17 ultrathin window and IXRF Systems, Inc, software)was per- ends-per-inch Hi-Nicalon, a 0.5 um BN interphase, and an MI formed on matrix, interphase, and fiber surfaces to determine the SiC matrix. The matrix was processed in several steps. First, a thin extent of oxidation Microscopy was performed on samples tested (2 um) layer of Sic was applied by chemical vapor infiltration in this study as well as on two fracture surfaces of specimens that to the BN-coated woven preform. A SiC particle containing slurry were obtained from Brewer et al i9 was infiltrated into the porous network. This was followed by infiltration of molten silicon that nearly filled the porous network Therefore, the matrix was predominantly Sic with some Si lL. Results and discussion Most tensile stress-rupture tests were performed using a screw- driven universal testing machine (Instron 4502, Instron, Canton, Mechanical Data MA). One test was performed on a hydraulic universal The stress-rupture results versus time are shown in Fig. 1. The machine(Instron 8500; for this test, an induction-heated Sic experiments were performed with Tmax equal to 960 and 815.C element furnace was used). The test specimen dimensions were 2.1 of the hot zone. The furnace profile is shown in Fig. 2. The temperature poxy tabs(0.08 in. thick)were glued to the ends of the bars. The where the first upture specimens failed was 880.C. The est specimens were gripped with hydraulic grips in the tabbed mens tested with Tmax =815C failed in the hottest region of the furnace except for the tests that were stopped before failure Corp, Englewood, CO) were attached to the ends of the tensile (138 h) The 815C fast fracture strength is also plotted in Fig. 1.9 This was done to ensure that the temperature that the sensors Combining the two sets of data, there appear to be two rupture xperienced was maintained at -25C. A resistance-heated fur- egimes: a more severe rate of rupture occurs at higher stresses, nace(MoSi, elements)was used to heat the center section of the ereas a less severe rate of rupture occurs at lower stresses. There pecimens. The furnace dimension in the tensile direction was 75 is a slight difference in the transition stress between these two nm; however, the zone of maximum uniform temperature, Tn egimes for the two sets of data. The data from Brewer et al. had (see Fig. 2). better rupture properties and a higher transition stress(-165 MPa) 4 AE monitoring was performed with a Digital Wave Fracture compared with the material tested in this study, which had a ave Detector(Digital Wave Corp. The AE setup and wave form transition stress of -150 MPa. The high-stress regime will analysis is given in greater detail in Ref. 22. It is important to note referred to as regime I and the low-stress regime will be referred that the location of the events could be determined from the speed to as regime Il. For the sake of comparison, the time exponent mentally, and the difference in times of arrival of the wave forms regime l Is--00s e I is --0.13 and the time exponent for corresponding to the same event on the two aE sensors stress-strain curves for several stress-rupture experiments In all of the tests, a very small constant load of 100N(3.8 MPa and for a room-temperature fast fracture test of a specimen from o s applied under load control as the furnace was heated up to the the same panel are shown in Fig. 3. The initial loading of the ed temperature to account for the thermal expansion of the stress-rupture experiment is identical to that of the room- material. SiC contact extensometers were then applied to the edge temperature stress-strain curve up to the rupture stress after which of the tensile bar. The sample was loaded to the predetermined additional strain is accumulated with time. Note that the transition tensile load at 0.25 mm/min. The sample was then held at the stress between the two rupture regimes is at the beginning of the predetermined load until failure or 138 h. AE was monitored from “knee” in the stress-strain cu the initial heatup until the failure of the sample, if it failed in the The"knee"is associated with increased damage in the matrix in allotted time To determine the damage accumulation, several of the ruptured stress-strain behavior obtained for the ref. 19 material is also omposites were cut and polished along the length of the tensile plotted in Fig. 3. The room-temperature stress-strain curves are bar Composite lengths of 40 mm on one side or the other of the ery similar except that the knee" in the curve occurs at a lower fracture surface were mounted in epoxy resin and polished to a 1 stress for the material tested in this study compared with the um finish. These polished portions included the length of com- material tested by brewer et al. In fact, for both materials, the lower temperatures. The polished samples were plasma etched stress-strain curve, i. e, through-thickness cracking. Therefore, the with hF to enhance the matrix cracks in the CVI SiC. The crack higher stress rupture rate appears to occur for a matrix state of pacing was determined over 10 mm lengths of the tensile bar. through-thickness cracking, whereas the low-stress rupture rate The fracture surfaces were examined by conventional scanning appears to occur for a matrix damage state of non-throug electron microscopy(SEM)(Jeol JSM 840A, Tokyo, Japan)and/or thickness microcracks Table L. Composite Constituent Properties Constituent Material (GPa) Details Fiber Hi-Nicalon' 0.34 280 5 Harness Satin itride Matrix VI SIC -2 mm thickness MI SIC + Si Process temp~1400°C TNippon Carbon, Tokyo, Japan. Chemical vapor deposition process
the room-temperature experiments will be performed during the stress-rupture experiments to locate and quantify the extent of damage. II. Experimental Procedure Several properties of the composite constituents are described in Table I. The material consisted of eight plies of woven 17 ends-per-inch Hi-Nicalon, a ;0.5 mm BN interphase, and an MI SiC matrix. The matrix was processed in several steps. First, a thin (;2 mm) layer of SiC was applied by chemical vapor infiltration to the BN-coated woven preform. A SiC particle containing slurry was infiltrated into the porous network. This was followed by infiltration of molten silicon that nearly filled the porous network. Therefore, the matrix was predominantly SiC with some Si. Most tensile stress-rupture tests were performed using a screwdriven universal testing machine (Instron 4502, Instron, Canton, MA). One test was performed on a hydraulic universal testing machine (Instron 8500; for this test, an induction-heated SiC element furnace was used). The test specimen dimensions were 2.1 mm thickness, 12.6 mm width, and 150 mm length. Graphiteepoxy tabs (0.08 in. thick) were glued to the ends of the bars. The test specimens were gripped with hydraulic grips in the tabbed region. Wide band (50 kHz to 2 MHz) AE sensors (Digital Wave Corp., Englewood, CO) were attached to the ends of the tensile specimens (edge-on) within the grips with quick setting epoxy. This was done to ensure that the temperature that the sensors experienced was maintained at ;25°C. A resistance-heated furnace (MoSi2 elements) was used to heat the center section of the specimens. The furnace dimension in the tensile direction was 75 mm; however, the zone of maximum uniform temperature, Tmax, was only ;15 mm (see Fig. 2). AE monitoring was performed with a Digital Wave Fracture Wave Detector (Digital Wave Corp.) The AE setup and wave form analysis is given in greater detail in Ref. 22. It is important to note that the location of the events could be determined from the speed of sound of the damaged material, which was determined experimentally, and the difference in times of arrival of the wave forms corresponding to the same event on the two AE sensors. In all of the tests, a very small constant load of 100 N (3.8 MPa) was applied under load control as the furnace was heated up to the desired temperature to account for the thermal expansion of the material. SiC contact extensometers were then applied to the edge of the tensile bar. The sample was loaded to the predetermined tensile load at 0.25 mm/min. The sample was then held at the predetermined load until failure or 138 h. AE was monitored from the initial heatup until the failure of the sample, if it failed in the allotted time. To determine the damage accumulation, several of the ruptured composites were cut and polished along the length of the tensile bar. Composite lengths of ;40 mm on one side or the other of the fracture surface were mounted in epoxy resin and polished to a 1 mm finish. These polished portions included the length of composite exposed to the hot zone region as well as regions exposed to lower temperatures. The polished samples were plasma etched with HF to enhance the matrix cracks in the CVI SiC. The crack spacing was determined over ;10 mm lengths of the tensile bar. The fracture surfaces were examined by conventional scanning electron microscopy (SEM) (Jeol JSM 840A, Tokyo, Japan) and/or field emission scanning electron microscopy (FESEM) (Hitachi S4500, Tokyo, Japan). For the FESEM, a voltage of 5 kV was used, which required no conductive coating. General observations were made on the fracture surfaces. Individual fiber fracture surfaces were also obtained from various regions of the fracture surface to perform fiber fractography analysis. Electron dispersive spectroscopy (EDS) (Kevex 4460 pulse processor with Quantum ultrathin window and IXRF Systems, Inc., software) was performed on matrix, interphase, and fiber surfaces to determine the extent of oxidation. Microscopy was performed on samples tested in this study as well as on two fracture surfaces of specimens that were obtained from Brewer et al.19 III. Results and Discussion (1) Mechanical Data The stress-rupture results versus time are shown in Fig. 1. The experiments were performed with Tmax equal to 960° and 815°C. The specimens tested at Tmax 5 960°C failed outside of the hot zone. The furnace profile is shown in Fig. 2. The temperature where the first two rupture specimens failed was ;880°C. The specimens tested with Tmax 5 815°C failed in the hottest region of the furnace except for the tests that were stopped before failure (138 h). The 815°C fast fracture strength is also plotted in Fig. 1.19 Combining the two sets of data, there appear to be two rupture regimes: a more severe rate of rupture occurs at higher stresses, whereas a less severe rate of rupture occurs at lower stresses. There is a slight difference in the transition stress between these two regimes for the two sets of data. The data from Brewer et al. had better rupture properties and a higher transition stress (;165 MPa) compared with the material tested in this study, which had a transition stress of ;150 MPa. The high-stress regime will be referred to as regime I and the low-stress regime will be referred to as regime II. For the sake of comparison, the time exponent (Introduction) for regime I is ;20.13 and the time exponent for regime II is ;20.03. The stress–strain curves for several stress-rupture experiments and for a room-temperature fast fracture test of a specimen from the same panel are shown in Fig. 3. The initial loading of the stress-rupture experiment is identical to that of the roomtemperature stress–strain curve up to the rupture stress after which additional strain is accumulated with time. Note that the transition stress between the two rupture regimes is at the beginning of the “knee” in the stress–strain curve. The “knee” is associated with increased damage in the matrix in the form of through-thickness cracking. The room-temperature stress–strain behavior obtained for the Ref. 19 material is also plotted in Fig. 3. The room-temperature stress–strain curves are very similar except that the “knee” in the curve occurs at a lower stress for the material tested in this study compared with the material tested by Brewer et al.19 In fact, for both materials, the transition stress corresponds to the beginning of the “knee” in the stress–strain curve, i.e., through-thickness cracking. Therefore, the higher stress rupture rate appears to occur for a matrix state of through-thickness cracking, whereas the low-stress rupture rate appears to occur for a matrix damage state of non-throughthickness microcracks. Table I. Composite Constituent Properties Constituent Material Volume fraction Elastic modulus (GPa) Details Fiber Hi-Nicalon† 0.34 280 5 Harness Satin Interphase Boron nitride 0.10 ? ;0.5 mm thick Matrix CVI‡ SiC 0.18 425 ;2 mm thickness MI SiC 1 Si 0.34 345 Process temp ; 1400°C Porosity 0.04 † Nippon Carbon, Tokyo, Japan. ‡ Chemical vapor deposition process. 1442 Journal of the American Ceramic Society—Morscher et al. Vol. 83, No. 6
Stress Rupture of a Woven Hi-Nicalon, BN-Inmterphase, SiC-Matrix Composite examples are shown in Fig. 4 of the AE source locations plotted Fast fracture Transition versus time for an experiment performed at 960C(163 MPa)and Brewer et al one performed at 815C(148 MPa). The location was determined by the difference in times of arrival of the initial extensional wave rtion of the ae wave form based on the accumulate Transition This study (reduction in elastic modulus).22,23 The data were sorted so that only events which occurred in the +30 mm region of the sample Rupture 815C LCF were analyzed. The data were also sorted according to the energy of each ae event. The events with the two highest decades of ae 815c energy are plotted separately and the events with the three lowest decades of AE energy are not shown in Figs. 4(a)and(b) Several details are worth noting from the AE studies. First, the 100L ailure location was accurately predicted-, from the location 0.001 1001000 analysis(final AE event) for both specimens as shown in Fig. 4. In fact, the locations of all five rupture specimens determined by the Fig. 1. Stress-rupture data from this study and LCF data from Ref. 19 AE method were within tI mm of the actual measured location. plotted verses time. Also plotted is the 815C fast fracture strength of the This confirms the accuracy of the ae source locations as pertain- material tested in Ref. 19 ng to real physical phenomena. It is also clear that there were a number of high-energy events for both specimens occurring at approximately the same location as the final failure event before the failure time. These high-energy AE events must have been caused by the formation of large cracks or the growth of existing racks at that location 800 Second, Fig. 5 shows the cumulative number of events and cumulative Ae energy from each event versus time for three 227 MPa 163 MPa different rupture experiments. For the lowest-stress(regime n) specimen, most of the AE events and the ae energy occurred after reaching the rupture load. For the highest-stress (regime I)speci- Location men, most of the events and most of the AE energy occurred ~870C during loading. For the middle-stress specimen(regime D), most of the events occurred after loading whereas most of the energy 400 occurred during loading. It has been shown in other studies that the -20-10 formation of large matrix cracks correlates well with the largest Furnace Distance, mm energy events, and not necessarily the number of Therefore, regime I material was fairly well cracked befo Fig. 2. Furnace profile and location of failure for first two stress-rupture load and some additional cracking occurred afterward, whereas experiments. most of the cracking occurring in the regime ll material occurred after the set load was reached Third, the location of damage is apparently temperature depen- dent based on the ae activity as a function of location, i.e RT Fast Fracture temperature profile. In Fig. 4(a), AE activity is significantly less in ef.19 the Tma=960C region after 1000 s compared with the regions RT Fast just outside the Tma =960.C region. Conversely, in Fig 4(b) Fracture more AE activity occurs in the Tmax =815C region. The crack spacing was determined for several ruptured composites. Table Il lists the crack spacing data for different temperature regimes along the length of the rupture sample based on the known temperature d not f Regime l profile of the furnace from three different rupture specimens. It is evident that more cracking occurred in the 700% to 900oC exposed region of the composites than at regions exposed to higher or lower temperatures. The matrix crack spacing was also shown in Refs. 22 and 24 to correlate well with the cumulated AE energy along the length of the specimen. It was also observed for the higher-temperature regions of the Fig 3. Stress-strain behavior of material tested at room temperature and 140 and 148 MPa regime Il specimens that cracks only penetrated for several rupture tests. The failure location of the RT fast fracture three or four plies into the material. Sometimes, cracks emanating specimen occurred in the grips; therefore, the ultimate strength is greater from both sides of the specimen would link up via a longitudinal crack. The specimen tested at 148 MPa that lasted for 100 h from the Ref. 19 study also only showed cracks, for the most part, which Because of this correlation, an additional experiment was started at the surface and penetrated only into the outer two plies Fibers were always observed to bridge the oxidized matrix cracks performed where the composite was precracked at 230 MPa at in all of the specimens observed with only a few exceptions. Since room temperature and tested at 815C at 1 10 MPa( Fig. 1). It was thought that this rupture condition could result in a failure time most AE activity occurs during loading or in the first few hours of corresponding to the rupture rate for regime I(30 h). However, the test (Fig. 5), matrix crack growth must occur because of the specimen lasted much longer than expected and after 138 h the relaxation of the fibers due to oxidation of the interphase or fiber xperiment was stopped. Thus the damage state does not fully creep"during the first few hours of the test. For the specimens explain the rate of rupture strength degradation. tested at these temperatures, very little fiber creep if any would be expected; therefore, oxidation of the interphase(discussed be- low) is the like mechanism causing fiber relaxation in the crack :
Because of this correlation, an additional experiment was performed where the composite was precracked at 230 MPa at room temperature and tested at 815°C at 110 MPa (Fig. 1). It was thought that this rupture condition could result in a failure time corresponding to the rupture rate for regime I (;30 h). However, the specimen lasted much longer than expected and after 138 h the experiment was stopped. Thus the damage state does not fully explain the rate of rupture strength degradation. (2) Acoustic Emission Data and Matrix Cracking Modal AE was used to monitor the damage occurring along the length of the rupture specimens during the experiment. Two examples are shown in Fig. 4 of the AE source locations plotted versus time for an experiment performed at 960°C (163 MPa) and one performed at 815°C (148 MPa). The location was determined by the difference in times of arrival of the initial extensional wave portion of the AE wave form based on the accumulated damage (reduction in elastic modulus).22,23 The data were sorted so that only events which occurred in the 630 mm region of the sample were analyzed. The data were also sorted according to the energy of each AE event. The events with the two highest decades of AE energy are plotted separately and the events with the three lowest decades of AE energy are not shown in Figs. 4(a) and (b). Several details are worth noting from the AE studies. First, the failure location was accurately predicted22,23 from the location analysis (final AE event) for both specimens as shown in Fig. 4. In fact, the locations of all five rupture specimens determined by the AE method were within 61 mm of the actual measured location. This confirms the accuracy of the AE source locations as pertaining to real physical phenomena. It is also clear that there were a number of high-energy events for both specimens occurring at approximately the same location as the final failure event before the failure time. These high-energy AE events must have been caused by the formation of large cracks or the growth of existing cracks at that location. Second, Fig. 5 shows the cumulative number of events and cumulative AE energy from each event versus time for three different rupture experiments. For the lowest-stress (regime II) specimen, most of the AE events and the AE energy occurred after reaching the rupture load. For the highest-stress (regime I) specimen, most of the events and most of the AE energy occurred during loading. For the middle-stress specimen (regime I), most of the events occurred after loading whereas most of the energy occurred during loading. It has been shown in other studies that the formation of large matrix cracks correlates well with the largestenergy events18,23 and not necessarily the number of events. Therefore, regime I material was fairly well cracked before the set load and some additional cracking occurred afterward, whereas most of the cracking occurring in the regime II material occurred after the set load was reached. Third, the location of damage is apparently temperature dependent based on the AE activity as a function of location, i.e., temperature profile. In Fig. 4(a), AE activity is significantly less in the Tmax 5 960°C region after ;1000 s compared with the regions just outside the Tmax 5 960°C region. Conversely, in Fig. 4(b), more AE activity occurs in the Tmax 5 815°C region. The crack spacing was determined for several ruptured composites. Table II lists the crack spacing data for different temperature regimes along the length of the rupture sample based on the known temperature profile of the furnace from three different rupture specimens. It is evident that more cracking occurred in the 700° to 900°C exposed region of the composites than at regions exposed to higher or lower temperatures. The matrix crack spacing was also shown in Refs. 22 and 24 to correlate well with the cumulated AE energy along the length of the specimen. It was also observed for the higher-temperature regions of the 140 and 148 MPa regime II specimens that cracks only penetrated three or four plies into the material. Sometimes, cracks emanating from both sides of the specimen would link up via a longitudinal crack. The specimen tested at 148 MPa that lasted for 100 h from the Ref. 19 study also only showed cracks, for the most part, which started at the surface and penetrated only into the outer two plies. Fibers were always observed to bridge the oxidized matrix cracks in all of the specimens observed with only a few exceptions. Since most AE activity occurs during loading or in the first few hours of the test (Fig. 5), matrix crack growth must occur because of relaxation of the fibers due to oxidation of the interphase or fiber creep25 during the first few hours of the test. For the specimens tested at these temperatures, very little fiber creep if any would be expected;26 therefore, oxidation of the interphase (discussed below) is the likely mechanism causing fiber relaxation in the crack wake and matrix microcrack growth. It should be noted that there was a significant amount of AE events which occurred near the 630 mm region of the composite Fig. 1. Stress-rupture data from this study and LCF data from Ref. 19 plotted verses time. Also plotted is the 815°C fast fracture strength of the material tested in Ref. 19. Fig. 2. Furnace profile and location of failure for first two stress-rupture experiments. Fig. 3. Stress–strain behavior of material tested at room temperature and for several rupture tests. The failure location of the RT fast fracture specimen occurred in the grips; therefore, the ultimate strength is greater than the failure stress shown in this figure. June 2000 Stress Rupture of a Woven Hi-Nicalon, BN-Interphase, SiC-Matrix Composite 1443
Journal of the American Ceramic Society-Morscher et al. Vol. 83. No 6 ture.oC 0 200400600 80010000 600 -10 200040006000800010000 0100002000030000400005000060000 TIME at abest hot of e temperature. Te A versus time for (a)an applied stress of 163 MPa at 960C hot zone temperature and (b) an applied stress of 148 MP AE event location data are separated open squares(second highest order of magnitude energy events)and filled circles ents). The lowest energy aE data are not shown. Also plotted is location versus furnace temperature profile (line) 815C;148MPa 00 (Regime Il) 400 s fied oxidation product. For the fracture surface shown in Fig. 6(a) approximately 10 tows were embrittled, corresponding to -1 960c;163MPad815c;27MP of the load-bearing tows. For another sample tested at 165 MPa time-2.8 hrs. I time=0.2 hrs. +200 Regime I) and a rupture time of 2.8 h with the same type of picture frame appearance, 18 tows were embrittled corresponding to -34% of 200 the load-bearing tows There was little evidence of oxidation at near fiber contact 100002000030000400005000060000 regions for the interior fibers that pull out under regime I Time. sec conditions. The transition from embrittled fibers to pulled-out Fig. 5. The cumulative number of AE events and cumulative AE as a function of time for three rupture specimens. The square the thickness crack for the short-time rupture specimens. The oxida- number of events curves indicates the time at which the load reached the tion front is presumably controlled by the kinetics of O2 and H=O set load for the rupture experiment. The two shorter time experiment data ingress and consumption when reacting with the BN and SiC. The fracture surface for the low stress (Fig. 7)is muc the fracture takes place on several different crack surface"planes showed that this was a relatively sometimes separated by up to 3 mm in the loading direction. small amount of excess AE energy and corresponded to no Second, instead of forming a picture frame, it appears that these smaller local cracks grew into the interior of the sample so that in this region probably occurs due to microcracks resulting from some interior bundles were embrittled whle some surtace bundles the bending moments associated with the nearness of the test to the ingle tow separating the embrittled fibers from the apparently pristine fibers. Figure 7(c)also shows a typical oxidized crack surface where borosilicate liquid/glass had formed to fill the (3) Microscopy of Composite Fracture Surfaces interphase regions surrounding the fibers and the crack surface Two representative fracture surfaces are shown in Figs. 6 and 7 from regimes I and Il, respectively. The fracture surface of the regime I tested specimen(Fig. 6) is characterized by a"picture (4 EDS Analysis frame"type of appearance. The outer layer of tows on the face EDS was performed on individual fiber fracture surfaces from and edges of the cross section have little or no pullout whereas the two composite fracture surfaces, one from regime I and one from interior fibers have all pulled out. The outer embrittled tows still regime Il, and the oxygen- to-carbon(O: C)peak-height ratio was have bn surrounding most of the fiber except where fibers are in measured. For the regime I s fracture surface(Fig. 6)the near contact to one another (Fig. 6(b). The oxidation reactions time to failure was only 1.3 h. The O: c ratio for pulled-out fibers occur first at the fiber/bn interphase probably enhanced by the was 0.25+ 0.01 from three different bundles. The O: C ratio for presence of a thin carbon layer on the fiber surface resulting from mbrittled fiber fracture surfaces ranged from 0.24 to 0.41 from the high-temperature MI process temperature. 27,28 Since fibers are seven different bundles. The O: C ratio for the matrix separated by only a fraction of a micrometer, the larger BN surface and around the embrittled bundles ranged from 0.5 to 0.8. The area exposed to the environment in these near contact regions pulled-out fibers had lower O: C ratios indicating that they failed esults in a greater amount of Bn volatilization and oxide last; however, there was only a small difference in O: C ratios formation. For every fiber-to-fiber contact area of an embrittled between the pulled-out and embrittled fibers tow, an oxide bonds the two fiber together at the closest points and For the regime II specimen fracture surface, the time to failure a vacant area exists between the oxide and remaining bn inter 1 14 h. The pulled-out fibers again had the lowest O: C ratios, hase (Fig. 6(b). This is very similar to the observation of 0.75 0.25. For the embrittled fiber bundles only one corner of Ogbuji the composite cross section, two load-bearing bundles, showed Nearly every fiber fractu is observed to emanate from high O: C ratios(-8)indicative of a relatively thick SiO, scale and ig. 6(b). A sin early fiber failure. The rest of the embrittled bundles showed only
test specimen. AE energy analysis showed that this was a relatively small amount of excess AE energy and corresponded to no measurable increase in crack density.24 The increased AE activity in this region probably occurs due to microcracks resulting from the bending moments associated with the nearness of the test specimen to the grips. (3) Microscopy of Composite Fracture Surfaces Two representative fracture surfaces are shown in Figs. 6 and 7 from regimes I and II, respectively. The fracture surface of the regime I tested specimen (Fig. 6) is characterized by a “picture frame” type of appearance.14 The outer layer of tows on the face and edges of the cross section have little or no pullout whereas the interior fibers have all pulled out. The outer embrittled tows still have BN surrounding most of the fiber except where fibers are in near contact to one another (Fig. 6(b)). The oxidation reactions occur first at the fiber/BN interphase probably enhanced by the presence of a thin carbon layer on the fiber surface resulting from the high-temperature MI process temperature.27,28 Since fibers are separated by only a fraction of a micrometer, the larger BN surface area exposed to the environment in these near contact regions results in a greater amount of BN volatilization and oxide formation.16 For every fiber-to-fiber contact area of an embrittled tow, an oxide bonds the two fiber together at the closest points and a vacant area exists between the oxide and remaining BN interphase (Fig. 6(b)). This is very similar to the observation of Ogbuji.28 Nearly every fiber fracture mirror is observed to emanate from these near fiber-contact regions (Fig. 6(b)). A similar observation was made for Hi-Nicalon/BN/CVI SiC minicomposites tested at slightly higher temperatures (900° to 1050°C)4,5 where fracture mirrors emanated from fiber–matrix bonds formed by the solidified oxidation product. For the fracture surface shown in Fig. 6(a), approximately 10 tows were embrittled, corresponding to ;19% of the load-bearing tows. For another sample tested at 165 MPa and a rupture time of 2.8 h with the same type of picture frame appearance, 18 tows were embrittled corresponding to ;34% of the load-bearing tows. There was little evidence of oxidation at near fiber contact regions for the interior fibers that pull out under regime I conditions. The transition from embrittled fibers to pulled-out fibers represents the oxidation reaction front in the throughthickness crack for the short-time rupture specimens. The oxidation front is presumably controlled by the kinetics of O2 and H2O ingress and consumption when reacting with the BN and SiC.14 The fracture surface for the low stress regime tested specimens (Fig. 7) is much different from the regime I tested specimen. First, the fracture takes place on several different crack surface “planes” sometimes separated by up to 3 mm in the loading direction. Second, instead of forming a picture frame, it appears that these smaller local cracks grew into the interior of the sample so that some interior bundles were embrittled while some surface bundles had fiber pullout. Figure 7(b) shows a fairly sharp transition in a single tow separating the embrittled fibers from the apparently pristine fibers. Figure 7(c) also shows a typical oxidized crack surface where borosilicate liquid/glass had formed to fill the interphase regions surrounding the fibers and the crack surfaces. (4) EDS Analysis EDS was performed on individual fiber fracture surfaces from two composite fracture surfaces, one from regime I and one from regime II, and the oxygen-to-carbon (O:C) peak-height ratio was measured. For the regime I specimen fracture surface (Fig. 6) the time to failure was only 1.3 h. The O:C ratio for pulled-out fibers was 0.25 6 0.01 from three different bundles. The O:C ratio for embrittled fiber fracture surfaces ranged from 0.24 to 0.41 from seven different bundles. The O:C ratio for the matrix regions in and around the embrittled bundles ranged from 0.5 to 0.8. The pulled-out fibers had lower O:C ratios indicating that they failed last; however, there was only a small difference in O:C ratios between the pulled-out and embrittled fibers. For the regime II specimen fracture surface, the time to failure was 114 h. The pulled-out fibers again had the lowest O:C ratios, 0.75 6 0.25. For the embrittled fiber bundles only one corner of the composite cross section, two load-bearing bundles, showed high O:C ratios (;8) indicative of a relatively thick SiO2 scale and early fiber failure. The rest of the embrittled bundles showed only Fig. 4. Load and AE event location versus time for (a) an applied stress of 163 MPa at 960°C hot zone temperature and (b) an applied stress of 148 MPa at 815°C hot zone temperature. The AE event location data are separated open squares (second highest order of magnitude energy events) and filled circles (highest order of magnitude energy events). The lowest energy AE data are not shown. Also plotted is location versus furnace temperature profile (line). Fig. 5. The cumulative number of AE events and cumulative AE energy as a function of time for three rupture specimens. The square on the “number of events” curves indicates the time at which the load reached the set load for the rupture experiment. The two shorter time experiment data sets are offset on the time scale for clarity. 1444 Journal of the American Ceramic Society—Morscher et al. Vol. 83, No. 6
1445 Table Il. Temperature Dependence of Matrix Crack Spacing emperature range along length of rupture specimen >900°C 00-700° <500° Hot zon 960°C Hot zone=815°C 0.4 0.6 Hot zone=960°1.10.6 Not determine 163MPa:2.8h 815 2.0 3.4 No cracks observed 140MPa,49h fracture en fracture surface, embrittled fibers and pulled- out fibers. Also the fracture mirrors were determined for one of the room-temperature tested tensile specimens. The strength of the fibers, Us is related to the mirror radiu of=Ar-1/2 where A is the fracture mirror constant. The data to be compared on a relative stress basis, i.e., the tensile rupture sti divided by the room-temperature ultimate tensile stress. For this reason, only the mirror radius corresponding to the average fiber strength,re, needs to be determined. The normalized rupture stress from fracture mirrors can then be determined from the relationship Em(T, ryon(RT c(RT C(, Voids where (l, n)refers to the rupture condition and(rt) refers to ro temperature. At least 40 fibers were used for each distribution. The normalized strengths from fracture mirrors are plotted in Formation Fig. 8 for two regime I and two regime ll specimens. Also plotted in Fig. 8 are stress conditions pertaining to the applied load condition for each rupture specimen in the form of the computed fiber stress normalized by the room-temperature ultimate strength. These include the applied stress, the stress on the pulled-out fibers if the embrittled fibers failed first (load-shed stress), and the rupture failure stress condition for individual as-produced Hi- Nicalon fibers at the same time-temperature condition. To determine the load-shed to the pulled-out fibers, the number of tows that were embrittled on the composite fracture surface was K 12. 0um estimated as explained above and the load those tows would hay carried was added to the applied load of the pulled-out tows. Fig. 6. (a) Fracture surface of regime I specimen tested at 185 MPa. The For both sets of data(Figs. 8(a)and(b), the average fiber upture time was 1.3 h.(b)Higher-magnification region showing em strengths from fracture mirrors of embrittled fibers were 75% of brittled fibers the room-temperature ultimate strength except for the longest time to failure specimen(63%). However, the average fiber strengths from fracture mirrors of the pulled-out fibers were quite different moderately higher O: C ratios than the pulled-out for the two different rupture regimes The o: c ratio of the matrix surface around all It should be noted that the pulled-out fibers fail independently bundles was on the order of 4 to 10 with the e and over fairly long gauge lengths since they do pull out. On the regions where parts of a bundle appeared to be other hand, the embrittled fibers only fail in or very near the matrix crack plane and the fracture origin is always correlated with a Therefore. it can be concluded that embrittled fiber failure neighboring, strongly bonded, fiber. Therefore, the pulled-out fiber occurred before pulled-out fiber failure. However, the time period trengths represent the actual fiber failure strength of the remain- etween most embrittled fiber failure and pulled-out fiber failure fibers during the catastrophic failure event. However, the was fairly short. The embrittled fibers must have been bridging the measured fiber strengths from fracture mirrors of the embrittled fibers are not straightforward since fiber failure is correlated with When the embrittled fibers did fail, the load they were carryi neighboring fibers was shed onto the remaining fibers. To determine the criterion For regime I composites, the pulled-out fiber fiber failure in the different regions of the composite fracture much lower than for regime Il composites. The mirror surface, fiber fractography was used strengths for regime I pulled-out fibers correspond the load-shed stresses and the as-produced fiber rupture stresses for the given rupture experiments(Fig 8(a). However, the regime (5) Fiber fractography II pulled-out fiber strengths correspond very well with the as- The fracture mirrors of individual fibers were used to determine produced pristine fiber stress as well as the load-shed stress(Fig the relative(average) strengths of fibers on the ruptured 8(b)
moderately higher O:C ratios than the pulled-out fibers, 1.6 6 0.8. The O:C ratio of the matrix surface around all of the embrittled bundles was on the order of 4 to 10 with the exception of a few regions where parts of a bundle appeared to be “sealed” by the oxidation product. Therefore, it can be concluded that embrittled fiber failure occurred before pulled-out fiber failure. However, the time period between most embrittled fiber failure and pulled-out fiber failure was fairly short. The embrittled fibers must have been bridging the matrix cracks for considerable periods of time before failing. When the embrittled fibers did fail, the load they were carrying was shed onto the remaining fibers. To determine the criterion for fiber failure in the different regions of the composite fracture surface, fiber fractography was used. (5) Fiber Fractography The fracture mirrors of individual fibers were used to determine the relative (average) strengths of fibers on the ruptured composite fracture surfaces. Two groups of fracture surfaces were observed for each specimen fracture surface, embrittled fibers and pulledout fibers. Also, the fracture mirrors were determined for one of the room-temperature tested tensile specimens. The strength of the fibers, sf , is related to the mirror radius, rc, by the relationship29–31 sf 5 Arc 21/ 2 (1) where A is the fracture mirror constant. The data were to be compared on a relative stress basis, i.e., the tensile rupture stress divided by the room-temperature ultimate tensile stress. For this reason, only the mirror radius corresponding to the average fiber strength, rc, needs to be determined. The normalized rupture stress from fracture mirrors can then be determined from the relationship sfm(T, t)/sfm~RT! 5 $rc(RT)/rc(T, t!}1/ 2 (2) where (T,t) refers to the rupture condition and (RT) refers to room temperature. At least 40 fibers were used for each distribution. The normalized strengths from fracture mirrors are plotted in Fig. 8 for two regime I and two regime II specimens. Also plotted in Fig. 8 are stress conditions pertaining to the applied load condition for each rupture specimen in the form of the computed fiber stress normalized by the room-temperature ultimate strength. These include the applied stress, the stress on the pulled-out fibers if the embrittled fibers failed first (load-shed stress), and the rupture failure stress condition for individual as-produced HiNicalon fibers at the same time–temperature condition.26 To determine the load-shed to the pulled-out fibers, the number of tows that were embrittled on the composite fracture surface was estimated as explained above and the load those tows would have carried was added to the applied load of the pulled-out tows. For both sets of data (Figs. 8(a) and (b)), the average fiber strengths from fracture mirrors of embrittled fibers were ;75% of the room-temperature ultimate strength except for the longest time to failure specimen (63%). However, the average fiber strengths from fracture mirrors of the pulled-out fibers were quite different for the two different rupture regimes. It should be noted that the pulled-out fibers fail independently and over fairly long gauge lengths since they do pull out. On the other hand, the embrittled fibers only fail in or very near the matrix crack plane and the fracture origin is always correlated with a neighboring, strongly bonded, fiber. Therefore, the pulled-out fiber strengths represent the actual fiber failure strength of the remaining fibers during the catastrophic failure event. However, the measured fiber strengths from fracture mirrors of the embrittled fibers are not straightforward since fiber failure is correlated with neighboring fibers. For regime I composites, the pulled-out fiber strengths were much lower than for regime II composites. The fracture mirror strengths for regime I pulled-out fibers correspond very well with the load-shed stresses and the as-produced fiber rupture stresses for the given rupture experiments (Fig. 8(a)). However, the regime II pulled-out fiber strengths correspond very well with the asproduced pristine fiber stress as well as the load-shed stress (Fig. 8(b)). Table II. Temperature Dependence of Matrix Crack Spacing Rupture specimen Temperature range along length of rupture specimen .900°C 700–900°C 500–700°C ,500°C Hot zone 5 960°C 0.3 0.2 0.5 0.6 227 MPa; 0.2 h Hot zone 5 815°C 0.4 0.6 0.7 185 MPa; 1.3 h Hot zone 5 960°C 1.1 0.6 1.2 Not determined 163 MPa; 2.8 h Hot zone 5 815°C 1.0 1.7 40 148 MPa; 14.8 h Hot zone 5 815°C 2.0 3.4 No cracks observed 140 MPa; 49 h Fig. 6. (a) Fracture surface of regime I specimen tested at 185 MPa. The rupture time was 1.3 h. (b) Higher-magnification region showing embrittled fibers. June 2000 Stress Rupture of a Woven Hi-Nicalon, BN-Interphase, SiC-Matrix Composite 1445