E≈S Journal of the European Ceramic Society 20(2000)2627-2636 Room and elevated temperature tensile properties of single tow Hi-Nicalon, carbon interphase, CVI SiC matrix minicomposites Martinez-Fernandez aGn. Morscher epartamento de Fisica de la Materia Condensada, Universidad de Sevilla, Spain bNASA Glenn Research Center, Ohio Aerospace Institute, MS 106-5, Cleveland, OH 44135, USA Received 20 January 2000: received in revised form 13 April 2000; accepted 29 April 2000 Abstract Single tow Hi-NicalonM,C interphase, CVI SiC matrix minicomposites were tested in tension at room temperature, 700, 950, and 1200C in air. Monotonic loading with modal acoustic emission monitoring was performed at room temperature in order to determine the dependence of matrix cracking on applied load. Modal acoustic emission was shown to correlate directly with the number of matrix cracks formed. Elevated temperature constant load stress-rupture and low-cycle fatigue experiments were per formed on precracked specimens. The elevated temperature rupture behavior was dependent on the precrack stress, the lower pre crack stress resulting in longer rupture life for a given stress. It was found that the rupture lives of C-interphase Hi-NicalonTM minicomposites were superior to C-interphase Ceramic Grade Nicalon M minicomposites and inferior to those of BN-interphase Hi-NicalonTM minicomposites. C 2000 Elsevier Science Ltd. All rights reserved Keywords: C-interphase; Composites: Mechanical properties; SiC-SiC; SiC fibres 1. Introduction However, this strength degradation at temperature, in air. and at stress is far more severe than could be behavi poor intermediate temperature tensile-rupture accounted for from these two mechanisms. For exam Carbon (NIC), Tokyo, Japan] reinforced CVI Sic NIC/SiC composite, the fully loaded gage length of the matrix composites with carbon interphases has been fibers could be approximated by demonstrated over temperatures ranging from 425 to 1000oC. 1-7 The time to failure for all these studies cor 1=aR/2/r (2) responds to a stress exponent, n, of approximately 1/4, 4 where time to failure, L, is directly proportional to com- where R is the fiber radius, f is the volume fraction of posite stress, o, to the power n: load-bearing fibers, and t is the interfacial shear strength. For the case where o equals 240 MPa, f t aoh () equals 0.16, t equals 10 MPa, and assuming a Weibull modulus m for individual fiber failure equal to 5, the This corresponds to a rupture strength degradation of maximum decrease in strength due to an increase in over 70% for rupture times less than 10 h. Two gage length(12 mm for the hot zone in the Ref 2 study) mechanisms have been put forward for the reduction in would be 45%. If more than one crack were in the hot strength of the NIC/SiC system with C interphases: the zone of the furnace the degradation in rupture strength increase in effective gage length from carbon volatiliza- due to this mechanism would be less. In addition, if the tion'and the flaw size increase due to oxide scale oxide scale was related to the flaw size, one would growthon the surface of the fibers. The latter mechan- expect an increase in flaw size(oxide scale thickness)of ism would predict a stress exponent of 1 /4 assuming approximately one order of magnitude if the rupture parabolic oxide growth at intermediate temperatures. strength was reduced from 2000 to 500 MPa(a 75% decrease)over 100 h at 700 C, assuming a Kic of 2.9 For the Ref. 2 study, SiO2 scales were not detectable on the fiber surfaces at these low temperatures, even though 0955-2219/00/S. see front matter C 2000 Elsevier Science Ltd. All rights reserved PII:S0955-2219(00)00138-2
Room and elevated temperature tensile properties of single tow Hi-Nicalon, carbon interphase, CVI SiC matrix minicomposites J. MartõÂnez-FernaÂndez a , G.N. Morscher b,* a Departamento de FõÂsica de la Materia Condensada, Universidad de Sevilla, Spain bNASA Glenn Research Center, Ohio Aerospace Institute, MS 106-5, Cleveland, OH 44135, USA Received 20 January 2000; received in revised form 13 April 2000; accepted 29 April 2000 Abstract Single tow Hi-NicalonTM, C interphase, CVI SiC matrix minicomposites were tested in tension at room temperature, 700, 950, and 1200C in air. Monotonic loading with modal acoustic emission monitoring was performed at room temperature in order to determine the dependence of matrix cracking on applied load. Modal acoustic emission was shown to correlate directly with the number of matrix cracks formed. Elevated temperature constant load stress-rupture and low-cycle fatigue experiments were performed on precracked specimens. The elevated temperature rupture behavior was dependent on the precrack stress, the lower precrack stress resulting in longer rupture life for a given stress. It was found that the rupture lives of C-interphase Hi-NicalonTM minicomposites were superior to C-interphase Ceramic Grade NicalonTM minicomposites and inferior to those of BN-interphase Hi-NicalonTM minicomposites. # 2000 Elsevier Science Ltd. All rights reserved. Keywords: C-interphase; Composites; Mechanical properties; SiC±SiC; SiC ®bres 1. Introduction The poor intermediate temperature tensile-rupture behavior in air of ceramic grade NicalonTM [Nippon Carbon (NIC), Tokyo, Japan] reinforced CVI SiC matrix composites with carbon interphases has been demonstrated over temperatures ranging from 425 to 1000C.1ÿ7 The time to failure for all these studies corresponds to a stress exponent, n, of approximately 1/4,4 where time to failure, t, is directly proportional to composite stress, , to the power n: t ÿn 1 This corresponds to a rupture strength degradation of over 70% for rupture times less than 10 h. Two mechanisms have been put forward for the reduction in strength of the NIC/SiC system with C interphases: the increase in eective gage length from carbon volatilization3 and the ¯aw size increase due to oxide scale growth4 on the surface of the ®bers. The latter mechanism would predict a stress exponent of 1/4 assuming parabolic oxide growth at intermediate temperatures. However, this strength degradation at temperature, in air, and at stress is far more severe than could be accounted for from these two mechanisms. For example, if only one crack existed in the gage section of a NIC/SiC composite, the fully loaded gage length of the ®bers could be approximated by: 1 R=2f 2 where R is the ®ber radius, f is the volume fraction of load-bearing ®bers, and is the interfacial shear strength.8 For the case where equals 240 MPa, f equals 0.16, equals 10 MPa, and assuming a Weibull modulus m for individual ®ber failure equal to 5, the maximum decrease in strength due to an increase in gage length (12 mm for the hot zone in the Ref. 2 study) would be 45%. If more than one crack were in the hot zone of the furnace the degradation in rupture strength due to this mechanism would be less. In addition, if the oxide scale was related to the ¯aw size, one would expect an increase in ¯aw size (oxide scale thickness) of approximately one order of magnitude if the rupture strength was reduced from 2000 to 500 MPa (a 75% decrease) over 100 h at 700C, assuming a KIC of 2.9 For the Ref. 2 study, SiO2 scales were not detectable on the ®ber surfaces at these low temperatures, even though 0955-2219/00/$ - see front matter # 2000 Elsevier Science Ltd. All rights reserved. PII: S0955-2219(00)00138-2 Journal of the European Ceramic Society 20 (2000) 2627±2636 * Corresponding author. E-mail address: gmorscher@grc.nasa.gov (G.N. Morscher)
2628 Martine. dez, G N. Morscher /Journal of the European Ceraic Society 20 (2000)2627-2636 the fiber fracture mirrors were observed to increase how much worse, if at all, is C interphase HN/ SiC in commensurate with the degree of minicomposite rup rupture than BN interphase HN/SiC? Therefore, the Ire strength loss. Therefore, it is evident that the impetus for to determine the rupture mechanisms causing NIC/SiC rupture with carbon behavior of minicomposites with C interphases interphases include a flaw growth or flaw creation to provide ison for HN/SiC minicomposites mechanism in addition to the two already mentioned with BN interphases as well as NIC/Sic minicomposites It is likely that the surface of the NIC fiber may have with C interphases b n modified during CVI SiC infiltration. Naslain0 describes four NIC/C/CVI SiC composite scenarios where the surface of the Si-C-O containing NIC fibers 2. Experimental procedure are altered after CVI SiC composite fabrication. All four scenarios have a complex carbon-rich layer on the Tows of 500 hn fibers were used to fabricate the fiber surface in between the Cvi deposited carbon layer minicomposites studied in this work. The tows were and the Si-C-O fiber. In some cases, SiO2 is present and mounted on graphite racks, coated with carbon and in others it is not depending on the CVI approach taken then composited with Sic by chemical vapor infiltration and/or a fiber pretreatment. Nevertheless, local carbon( Hyper-Therm Inc, Huntington Beach, CA). The fiber rich areas at the fiber surface are present that woul volume fraction(0.16+0.01)and minicomposite cross- oxidize during rupture testing. In other words, sectional area was determined based on the measured or mechanism due to surface recession or"pit formation estimated weights and densities of the minicomposite as a result of the oxidation of carbon layers or local constituents. 3 The observation of the minicomposite carbon rich regions on the fiber surface could lead to polished cross-sections indicated that the carbon inter rapid strength loss. This type of mechanism better phase was uniform with an average thickness of 0. 4 um explains the observed rapid rupture strength loss with The fracture surfaces of the minicomposites had a time for NIC/C/SiC composites. It is not known if the fibrous appearance. the Sic matrix being thicker on the more thermally stable Hi-NicalonM(HN)fibers withC outside but more uniformly distributed in the interior interphases would undergo similar degradation after than in the NIC fiber, C-interphase minicomposites CVI SiC processing previously studied. 2 In the earlier study, the minicomposite tensile stress Room temperature tensile testing was performed rupture properties of NIC reinforced SiC with a BN using an universal testing machine (Model 4502, interphase had superior elevated temperature stress- Instron, Canton, MA); the test set-up is described in rupture life in air compared to C interphase mini- detail elsewhere. 2 The minicomposites were mounted composites. This has also been shown for woven NIC/ onto cardboard tabs with epoxy. Monotonic loading Sic composites tested in flexure. It was suggested in tensile tests were performed to determine the ultimate Ref. 2 that the BN interphase NIC/SiC rupture beha- failure load and the optimum precrack load. Modal vior was superior to C interphase NIC/SiC because of acoustic emission(AE)was monitored with sensors the formation of a measurable oxide layer (0.5 H attached to the epoxy just above and below the mini- 700C after 12 h)on the fibers that had a bn inter- composite gage-section. The AE analyzer(Digital Wave phase. This was due to the enhanced oxidation of Sic Corporation, Englewood, CO) recorded and digitized when in contact with BN. This oxide layer, which was the true sound wave form for each event on both chan predominantly SiOz, protected the fibers from the nels(sensors). The number of events and location of strength degrading mechanism of the NIC fibers when each event could then be determined once the speed of directly exposed to the environment as was the case for sound was estimated The details of the ae analysis are C interphase NIC/SiC. The presence of thick oxide included in the Appendix A. scales on fibers in BN interphase NIC/SiC causing only Constant-load stress-rupture tests were run in a dead- minimal strength degradation (n-50) also is further weight load stress-rupture rig. A furnace was located at evidence against the oxide scale mechanism discussed the center of the minicomposite. The total length of the above for NIC/SiC with C interphases. It was also furnace was 35 mm with a hot zone of 12 mm. These hown in Ref 2 that the rupture behavior of Hn/BN/ tests were run at 700, 950, and 1200C. The mini- iC minicomposites were superior to NIC/BN/SiC composites were fully loaded before increasing the tem minicomposites in air minature at 100C/min up to the test temperature. The Unfortunately, in the earlier study, HN/SiC mini- minicomposites were precracked at room temperature composites with C interphases were not studied. Since with loads of 119 or 126 N, which corresponds to a HN is a more thermally stable fiber than NIC, it may be composite stress of 280 or 295 MPa, respectively expected that HN SiC with C interphases is not as sus- The fatigue tests were als o run at700,950,and ceptible to severe rupture-strength degradation as NIC/ 1200C with the universal testing machine and same SiC with C interphases. Thus, the question remains, furnace set-up. The load was cycled from a minimum
the ®ber fracture mirrors were observed to increase commensurate with the degree of minicomposite rupture strength loss. Therefore, it is evident that the mechanisms causing NIC/SiC rupture with carbon interphases include a ¯aw growth or ¯aw creation mechanism in addition to the two already mentioned. It is likely that the surface of the NIC ®ber may have been modi®ed during CVI SiC in®ltration. Naslain10 describes four NIC/C/CVI SiC composite scenarios where the surface of the Si±C±O containing NIC ®bers are altered after CVI SiC composite fabrication. All four scenarios have a complex carbon-rich layer on the ®ber surface in between the CVI deposited carbon layer and the Si±C±O ®ber. In some cases, SiO2 is present and in others it is not depending on the CVI approach taken and/or a ®ber pretreatment. Nevertheless, local carbon rich areas at the ®ber surface are present that would oxidize during rupture testing. In other words, a mechanism due to surface recession or ``pit formation'' as a result of the oxidation of carbon layers or local carbon rich regions on the ®ber surface could lead to rapid strength loss. This type of mechanism better explains the observed rapid rupture strength loss with time for NIC/C/SiC composites. It is not known if the more thermally stable Hi-NicalonTM (HN) ®bers with C interphases would undergo similar degradation after CVI SiC processing. In the earlier study,2 the minicomposite tensile stress rupture properties of NIC reinforced SiC with a BN interphase had superior elevated temperature stressrupture life in air compared to C interphase minicomposites. This has also been shown for woven NIC/ SiC composites tested in ¯exure.11 It was suggested in Ref. 2 that the BN interphase NIC/SiC rupture behavior was superior to C interphase NIC/SiC because of the formation of a measurable oxide layer (0.5 m at 700C after 12 h) on the ®bers that had a BN interphase. This was due to the enhanced oxidation of SiC when in contact with BN.12 This oxide layer, which was predominantly SiO2, protected the ®bers from the strength degrading mechanism of the NIC ®bers when directly exposed to the environment as was the case for C interphase NIC/SiC. The presence of thick oxide scales on ®bers in BN interphase NIC/SiC causing only minimal strength degradation (n50) also is further evidence against the oxide scale mechanism discussed above for NIC/SiC with C interphases. It was also shown in Ref. 2 that the rupture behavior of HN/BN/ SiC minicomposites were superior to NIC/BN/SiC minicomposites in air. Unfortunately, in the earlier study, HN/SiC minicomposites with C interphases were not studied. Since HN is a more thermally stable ®ber than NIC, it may be expected that HN/SiC with C interphases is not as susceptible to severe rupture-strength degradation as NIC/ SiC with C interphases. Thus, the question remains, how much worse, if at all, is C interphase HN/SiC in rupture than BN interphase HN/SiC? Therefore, the impetus for this work was to determine the rupture behavior of HN/SiC minicomposites with C interphases to provide a comparison for HN/SiC minicomposites with BN interphases as well as NIC/SiC minicomposites with C interphases. 2. Experimental procedure Tows of 500 HN ®bers were used to fabricate the minicomposites studied in this work. The tows were mounted on graphite racks, coated with carbon and then composited with SiC by chemical vapor in®ltration (Hyper-Therm Inc., Huntington Beach, CA). The ®ber volume fraction (0.160.01) and minicomposite crosssectional area was determined based on the measured or estimated weights and densities of the minicomposite constituents.13 The observation of the minicomposite polished cross-sections indicated that the carbon interphase was uniform with an average thickness of 0.4 mm. The fracture surfaces of the minicomposites had a ®brous appearance, the SiC matrix being thicker on the outside, but more uniformly distributed in the interior than in the NIC ®ber, C-interphase minicomposites previously studied.2 Room temperature tensile testing was performed using an universal testing machine (Model 4502, Instron, Canton, MA); the test set-up is described in detail elsewhere.2 The minicomposites were mounted onto cardboard tabs with epoxy. Monotonic loading tensile tests were performed to determine the ultimate failure load and the optimum precrack load. Modal acoustic emission (AE) was monitored with sensors attached to the epoxy just above and below the minicomposite gage-section. The AE analyzer (Digital Wave Corporation, Englewood, CO) recorded and digitized the true sound wave form for each event on both channels (sensors). The number of events and location of each event could then be determined once the speed of sound was estimated. The details of the AE analysis are included in the Appendix A. Constant-load stress-rupture tests were run in a deadweight load stress-rupture rig. A furnace was located at the center of the minicomposite. The total length of the furnace was 35 mm with a hot zone of 12 mm. These tests were run at 700, 950, and 1200C. The minicomposites were fully loaded before increasing the temperature at 100C/min up to the test temperature. The minicomposites were precracked at room temperature with loads of 119 or 126 N, which corresponds to a composite stress of 280 or 295 MPa, respectively. The fatigue tests were also run at 700, 950, and 1200C with the universal testing machine and same furnace set-up. The load was cycled from a minimum 2628 J. Marti nez-FernaÂndez, G.N. Morscher / Journal of the European Ceramic Society 20 (2000) 2627±2636
J. Martinez-Fernandez, G.N. Morscher / Journal of the European Ceramic Society 20(2000)2627-2636 stress of 25-50 MPa to the peak stress. The loading ultimate failure composite stress ranged from 280 to 350 requency was 0.01 Hz (100 s per cycle) for all the MPa; however, most of the minicomposites(approxi experiments. For the fatigue tests, the load-cycle was mately 75%)failed between 310 and 325 MPa, with an begun after reaching the set temperature. The peak load average strength of 317-+26 MPa. The average ultimate as applied prior to and during heating just as for the strength of the fibers determined by the cross-sectional rupture tests. The minicomposites were also precracked area of load-bearing fibers(fully loaded) was 2000_+200 prior to the cyclic test at 280 MPa. MPa. The strength of individual HN fibers (25.4 mm Some minicomposites were polished longitudinally gage length)is -2800 MPa whereas the average strength after testing, and the crack spacing was measured by of as-produced tows of hn(25.4 mm gage length)is optical microscopy. The fracture surfaces were observed 1700 MPa. 4 Based on Eq (2), the actual fully-loaded using a scanning electron microscope(SEM) Jeol 840A gage length is on the order of 30 to 50 mm, therefore, in (eol, Tokyo, Japan) comparison to single tow strength, no strength loss due to minicomposite fabrication occurred Fig. 2. shows the cumulative number of events recor- 3. Results ded normalized by the gage length of the minicomposite en. The nature of Ae activity was very co 3.1. Room temperature mechanical testing for the specimens tested. For several samples, the num ber of cracks was measured from polished longitudinal The load versus time curve for a typical monotonic sections. The crack spacing was determined from the loading experiment is shown in Fig. 1. The curves are number of cracks counted over a given length of linear up to approximately 84.4 N(200 MPa composite minicomposite. In the Appendix, Table Al shows stress), when matrix cracking starts to occur. The the nearly one-to-one correspondence between the occurrence of AE events(also shown in Fig. 1 )indi- number of highest energy events recorded and the cates that the change in slope is associated with matrix estimated number of cracks in the gage section of cracking. The cumulative number of AE events and the the minicomposite cumulative total energy increase at the same rate. The 3. 2. Elevated temperature mechanical testing The data from the constant load stress-rupture experiments are plotted in Fig. 3a and b as the stress applied (if fibers were fully loaded) versus the time to rupture( the arrows indicate the minicomposite did not fail). These data are plotted together with previous data from C-NIC minicomposites(Fig. 3a)and BN-HN minicomposites(Fig 3b).2 The C-HN minicomposites have longer survival times than the C-NIC mini- 50100150200250300 composites at 700 C although for higher loads the two curves tend to blend together. The C-hn mini al load vs. time curve for a monotonic loading experi. composites have shorter survival times than the BN- Imber and cumulative energy (in arbitary units)of the HN minicomposites especially at 700 and 950C. All the also shown samples broke in the hot region of the furnace. For hese tests the minicomposites were precracked at 280 MPa(composite stress). Stress-rupture tests were also performed using a lar ger precrack load In Fig 4, the C-Hn data from Fig 3 (where a precrack load of 280 MPa was used) are plot- 9 N Precrack ted compared with the data from samples precracked at 295 MPa. At 700 and 950C. the time of survival decreases with increasing precrack load. At 1200C, this 04 effect was not as pronounced as at the lower tempera ture tests. The use of a larger precrack load introduced a greater amount of scatter in the data. The precrack stresses of 280 MPa (load of 119 N) and 295 MPa(load of 126 N) correspond with 88 and 93% of the average Fig. 2. Number of events per mm vs load for several specimens tested strength, respectively. The precrack load used in a pre- at room temperature vious work for C-NIC and BN-hn was 60%(110 N)
stress of 25±50 MPa to the peak stress. The loading frequency was 0.01 Hz (100 s per cycle) for all the experiments. For the fatigue tests, the load-cycle was begun after reaching the set temperature. The peak load was applied prior to and during heating just as for the rupture tests. The minicomposites were also precracked prior to the cyclic test at 280 MPa. Some minicomposites were polished longitudinally after testing, and the crack spacing was measured by optical microscopy. The fracture surfaces were observed using a scanning electron microscope (SEM) Jeol 840A (Jeol, Tokyo, Japan). 3. Results 3.1. Room temperature mechanical testing The load versus time curve for a typical monotonic loading experiment is shown in Fig. 1. The curves are linear up to approximately 84.4 N (200 MPa composite stress), when matrix cracking starts to occur. The occurrence of AE events (also shown in Fig. 1.) indicates that the change in slope is associated with matrix cracking. The cumulative number of AE events and the cumulative total energy increase at the same rate. The ultimate failure composite stress ranged from 280 to 350 MPa; however, most of the minicomposites (approximately 75%) failed between 310 and 325 MPa, with an average strength of 31726 MPa. The average ultimate strength of the ®bers determined by the cross-sectional area of load-bearing ®bers (fully loaded) was 2000200 MPa. The strength of individual HN ®bers (25.4 mm gage length) is 2800 MPa whereas the average strength of as-produced tows of HN (25.4 mm gage length) is 1700 MPa.14 Based on Eq. (2), the actual fully-loaded gage length is on the order of 30 to 50 mm, therefore, in comparison to single tow strength, no strength loss due to minicomposite fabrication occurred. Fig. 2. shows the cumulative number of events recorded normalized by the gage length of the minicomposite specimen. The nature of AE activity was very consistent for the specimens tested. For several samples, the number of cracks was measured from polished longitudinal sections. The crack spacing was determined from the number of cracks counted over a given length of minicomposite. In the Appendix, Table A1 shows the nearly one-to-one correspondence between the number of highest energy events recorded and the estimated number of cracks in the gage section of the minicomposite. 3.2. Elevated temperature mechanical testing The data from the constant load stress-rupture experiments are plotted in Fig. 3a and b as the stress applied (if ®bers were fully loaded) versus the time to rupture (the arrows indicate the minicomposite did not fail). These data are plotted together with previous data from C±NIC minicomposites (Fig. 3a) and BN±HN minicomposites (Fig. 3b).2 The C±HN minicomposites have longer survival times than the C±NIC minicomposites at 700C although for higher loads the two curves tend to blend together. The C±HN minicomposites have shorter survival times than the BN± HN minicomposites especially at 700 and 950C. All the samples broke in the hot region of the furnace. For these tests the minicomposites were precracked at 280 MPa (composite stress). Stress±rupture tests were also performed using a larger precrack load. In Fig. 4, the C±HN data from Fig. 3 (where a precrack load of 280 MPa was used) are plotted compared with the data from samples precracked at 295 MPa. At 700 and 950C, the time of survival decreases with increasing precrack load. At 1200C, this eect was not as pronounced as at the lower temperature tests. The use of a larger precrack load introduced a greater amount of scatter in the data. The precrack stresses of 280 MPa (load of 119 N) and 295 MPa (load of 126 N) correspond with 88 and 93% of the average strength, respectively. The precrack load used in a previous work for C±NIC and BN±HN was 60% (110 N) Fig. 1. Typical load vs. time curve for a monotonic loading experiment. The number and cumulative energy (in arbritary units) of the AE events are also shown. Fig. 2. Number of events per mm vs load for several specimens tested at room temperature. J. Marti nez-FernaÂndez, G.N. Morscher / Journal of the European Ceramic Society 20 (2000) 2627±2636 2629
J. Martinez.Fernandez, G N Morscher Journal of the European Ceramic Society 20(2000)2627-2636 vo=5c 000 言 9876 □ ChnC 700C CNic(1 TTTTmTTTTTTTTmmF 0.010.101.0010.0 00100000 00.00100000 Time(hrs Fig 3. Minicomposite stress-rupture data: (a)compared with previous data on Nicalon minicomposites: (b) compared with previous data on BN- HiNic minicomposites. 2. and 78%(111 N) of the average strength, respectively. MPa, the stress on the fibers where fully loaded would Since the fraction of the average strength using the be 500 MPa). For the tests run at 700 and 950C, failure lower precrack load is closer to the previous ones, the always occurred in the hot zone of the furnace For all results with the samples precracked at 280 MPa were he tests run at 1200C. the failure occurred outside the the ones used for comparison hot zone region, about 2 cm from the center of the fur The results from the fatigue experiments are plotted in nace. In these regions, the temperature is approximately Fig. 5 together with the stress-rupture results from 900oC, based on the temperature profile of the furnace samples precracked at 280 MPa(same precrack load). The data from these tests were very similar to the data The time taken for the fatigue experiments is the total from the tests run at 950 C. time under cyclic loading conditions. At 700oC, the Some of the minicomposites did not fail after very behavior is essentially the same as the constant load long rupture times(e.g. 1000, 500, and 400 MPa for 700 conditions. The minicomposite resistance to fatigue 950, and 1200C, respectively). These tests as well as rapidly decreased at 950C. At this temperature, the some tests at shorter times were stopped before rupture samples tested at the highest peak stress failed in the and the retained strengths of the minicomposites were first or second fatigue cycle (for peak stresses over 75 determined at room temperature. The general trend was 700c:119N 00950c:119N▲ 950c.126N 00c:119N 5 Stress-nuoture 1200 C r10 TTTT TTTTT 1000100.001000.00 0010.1010010.00100.001000.00 Time(hrs sS-rupture data for two differe Fig. 5. Peak stress for fatigue experiments vs the survival time and
and 78% (111 N) of the average strength, respectively. Since the fraction of the average strength using the lower precrack load is closer to the previous ones, the results with the samples precracked at 280 MPa were the ones used for comparison. The results from the fatigue experiments are plotted in Fig. 5 together with the stress-rupture results from samples precracked at 280 MPa (same precrack load). The time taken for the fatigue experiments is the total time under cyclic loading conditions. At 700C, the behavior is essentially the same as the constant load conditions. The minicomposite resistance to fatigue rapidly decreased at 950C. At this temperature, the samples tested at the highest peak stress failed in the ®rst or second fatigue cycle (for peak stresses over 75 MPa, the stress on the ®bers where fully loaded would be 500 MPa). For the tests run at 700 and 950C, failure always occurred in the hot zone of the furnace. For all the tests run at 1200C, the failure occurred outside the hot zone region, about 2 cm from the center of the furnace. In these regions, the temperature is approximately 900C, based on the temperature pro®le of the furnace. The data from these tests were very similar to the data from the tests run at 950C. Some of the minicomposites did not fail after very long rupture times (e.g. 1000, 500, and 400 MPa for 700, 950, and 1200C, respectively). These tests as well as some tests at shorter times were stopped before rupture and the retained strengths of the minicomposites were determined at room temperature. The general trend was Fig. 3. Minicomposite stress±rupture data: (a) compared with previous data on Nicalon minicomposites;1 (b) compared with previous data on BN± HiNic minicomposites.2 . Fig. 4. Minicomposite stress±rupture data for two dierent precrack loads. Fig. 5. Peak stress for fatigue experiments vs the survival time and comparison with stress±rupture data. 2630 J. Marti nez-FernaÂndez, G.N. Morscher / Journal of the European Ceramic Society 20 (2000) 2627±2636
J. Martinez.Fernandez, G N Morscher Journal of the European Ceramic Society 20(2000)2627-2636 Retained strength at room temperature data for C-HN and BN-HN minicomposites Incomposite Temperature Applied stress Time at stress Retained of as-produced (°C) strength(MPa) strength C-HN 000000 337 337 17 BN-HN 400 1630 7867507758 The average ultimate strength(of the fibers if fully loaded) for as-produced C-HN=2000 MPa and for as-produced BN-HN=2100 MPa. higher rupture temperatures resulted in lower retained though the carbon interphase was removed by oxida- strength(Table 1) tion, resulting in a gap between the fiber and matrix At 1200 C the formation of Sio, is more evident on 33. Microstructural characterization the surface of the CVI SiC matrix and in the gap that was formerly the C interphase(Fig. 7). Two different Figs. 6 and 7 show the range of observed micro- regions are noted in this sample. The higher magnifica structural features observed on fracture surfaces after tion insets in Fig. 7 show that the crack had propagated for C-Hn minicomposite specimens tested between 700 through part of the sample prior to failure, where the and 1200C. Fig. 6 shows a micrograph of a typical matrix region is clearly oxidized. Part of the failure fracture surface for a stress-rupture test run at 700C. matrix crack surface was not oxidized. Apparently, the Due to the fibrous nature of these minicomposites frac- non-through thickness matrix crack propagated ire often occurs at different planar levels, although at through an uncracked section of matrix at ultimate each"level"the matrix fracture surfaces were flat. For minicomposite failure. However, the C interphase was short-term rupture(t<0. 1 h) there was some fiber pull- removed and fibers were strongly bonded to the matrix out. For longer rupture time at 700oC, some regions of in this region, presumably from oxidation through the minicomposite fracture surface showed relatively another matrix crack and along the vacated interphase long pullout lengths(Fig. 6). However, this increase in channel. A SiO2 scale did not cover the fiber fracture length was not uniform throughout the sample, ever surfaces because the fibers did not fail until mini omposite ultimate failure The fracture surfaces of the stress-rupture test per formed at 950C(not shown) contain little if any fiber pull out even for very short-term rupture because the 00m oxide reaction product nearly fills the gap left by the vacated interphase. Bonding between the fiber and matrix was obviously strong. 4. Discussion 4.1. Stress-rupture: comparison with previous data In Fig 3a, the stress-rupture data at 700oC is com- pared with previous data from NIC minicomposites with a carbon interphase. In both cases the mini- composite rupture is due to fiber degradation, as the Fig. 6. Typical fracture surface of C-HN minicomposite for stress matrix-fiber bonding associated with SiOz formation is ruptures at 700C(1183 MPa, 5.9 h) negligible. For this reason the better properties of C-
higher rupture temperatures resulted in lower retained strength (Table 1). 3.3. Microstructural characterization Figs. 6 and 7 show the range of observed microstructural features observed on fracture surfaces after for C-HN minicomposite specimens tested between 700 and 1200C. Fig. 6 shows a micrograph of a typical fracture surface for a stress-rupture test run at 700C. Due to the ®brous nature of these minicomposites fracture often occurs at dierent planar levels, although at each ``level'' the matrix fracture surfaces were ¯at. For short-term rupture (t<0.1 h) there was some ®ber pullout. For longer rupture time at 700C, some regions of the minicomposite fracture surface showed relatively long pullout lengths (Fig. 6). However, this increase in length was not uniform throughout the sample, even though the carbon interphase was removed by oxidation, resulting in a gap between the ®ber and matrix. At 1200C the formation of SiO2 is more evident on the surface of the CVI SiC matrix and in the gap that was formerly the C interphase (Fig. 7). Two dierent regions are noted in this sample. The higher magni®cation insets in Fig. 7 show that the crack had propagated through part of the sample prior to failure, where the matrix region is clearly oxidized. Part of the failure matrix crack surface was not oxidized. Apparently, the non-through thickness matrix crack propagated through an uncracked section of matrix at ultimate minicomposite failure. However, the C interphase was removed and ®bers were strongly bonded to the matrix in this region, presumably from oxidation through another matrix crack and along the vacated interphase channel. A SiO2 scale did not cover the ®ber fracture surfaces because the ®bers did not fail until minicomposite ultimate failure. The fracture surfaces of the stress-rupture test performed at 950C (not shown) contain little if any ®ber pull out even for very short-term rupture because the oxide reaction product nearly ®lls the gap left by the vacated interphase. Bonding between the ®ber and matrix was obviously strong. 4. Discussion 4.1. Stress±rupture: comparison with previous data In Fig. 3a, the stress±rupture data at 700C is compared with previous data from NIC minicomposites with a carbon interphase. In both cases the minicomposite rupture is due to ®ber degradation, as the matrix±®ber bonding associated with SiO2 formation is negligible. For this reason the better properties of C± Table 1 Retained strength at room temperature data for C±HN and BN±HN minicomposites Minicomposite Temperature ( C) Applied stress (MPa) Time at stress (h) Retained strength (MPa) % of as-produced strengtha C±HN 700 328 50 1389 69 700 1099 358 1114 56 950 590 94 683 34 950 655 337 732 37 950 655 337 766 38 1200 439 3 512 26 1200 439 17 544 27 1200 439 162 491 25 BN±HN 700 996 400 2260 100 816 951 324 1411 67 950 1019 252 1630 77 1200 328 882 525 25 1200 385 965 792 38 a The average ultimate strength (of the ®bers if fully loaded) for as-produced C±HN=2000 MPa and for as-produced BN±HN=2100 MPa. Fig. 6. Typical fracture surface of C±HN minicomposite for stress ruptures at 700C (1183 MPa, 5.9 h). J. Marti nez-FernaÂndez, G.N. Morscher / Journal of the European Ceramic Society 20 (2000) 2627±2636 2631
2632 J. Martinez.Fernandez, G N Morscher Journal of the European Ceramic Society 20(2000)2627-2636 No Sio.o atrix surfac pR matrix face Fig. 7. Fracture surface of a C-HN minicomposite for stress rupture at 1200.C (439 MPa, 217 h). Hn versus C-NIC are clearly related with the better strength to 2100 MPa was used for comparison thermal stability of the HN fibers. The stress-rupture as-produced fiber rupture properties with the resistance of the C-Hn minicomposites, however, is composite stress-rupture data. If the fiber degradation is inferior to BN-HN(Fig 3b) merely due to the fiber degradation observed for as- In order to compare minicomposite data at different produced fibers, the minicomposite rupture data should temperatures and fiber fast fracture data, the Larson- decrease the same as the fiber data if the starting Miller(LM)approach, Io was used. In this empirical strength of the fibers were 2100 MPa approach the effect of temperature and time are com- The C-HN and BN-HN minicomposite properties bined in one parameter: are closer to the individual fiber rupture behavior at low and high temperatures. At intermediate temperatures q= T(logar +O ( there is degradation of the minicomposite behavior compared to the individual fiber because of the embrit tlement associated with fiber-matrix bonding due to where T is the temperature(K), fr the time(h), and C SiO2 formation. Fig 8 shows the improvement in rup- the LM constant, which was found to be 22 for both ture properties of BN interphase over the C interphase NIC and Hn fibers. In Fig. 8, data obtained in this It is also clear from Fig 8 that C-NIC suffered more and previous studies are plotted as the applied rupture serious fiber degradation due to oxidation than C-HN stress versus the LM parameter(). The room tempera At 700C, the loss of carbon by oxidation results in ture(rT) failure stress was 2020, 2300, and 2100 MPa long load-bearing gage lengths for the Hn fibers. The for C-HN, C-NIC, and BN-HN, respectively. Also oxidation kinetics of a continuous carbon interphase plotted in Fig. 8 is the data for stress-rupture of as- a composite has been published in the literature. 9,20 If produced HN fibers. The room temperature strength we assume that the oxidation is controlled by the diffu of the HN fibers studied in Ref. 17 was 2800 MPa. It sion of oxygen, the recession length(5)is was shown in Ref 18 that HN-BN minicomposites with different RT strengths had the same rupture behavior at low and high temperatures as the fibers [i.e. same slope 5=klr/2 n In(o)vs q. However, the absolute rupture strengths were lower for the lower RT strength minicomposites in proportion to the absolute strength of the as-produced where kp is the parabolic rate constant and t the time fibers. Therefore, normalizing the rt as-produced fiber For an opening of 0. 4 um(carbon interface thickness )
HN versus C±NIC are clearly related with the better thermal stability of the HN ®bers. The stress-rupture resistance of the C-HN minicomposites, however, is inferior to BN±HN (Fig. 3b). In order to compare minicomposite data at dierent temperatures and ®ber fast fracture data, the Larson± Miller (LM) approach15,16 was used. In this empirical approach the eect of temperature and time are combined in one parameter: q T logtR C 3 where T is the temperature (K), tR the time (h), and C the LM constant, which was found to be 22 for both NIC and HN ®bers.17 In Fig. 8, data obtained in this and previous studies2 are plotted as the applied rupture stress versus the LM parameter (q). The room temperature (RT) failure stress was 2020, 2300, and 2100 MPa for C±HN, C±NIC, and BN±HN, respectively. Also plotted in Fig. 8 is the data for stress-rupture of asproduced HN ®bers.17 The room temperature strength of the HN ®bers studied in Ref. 17 was 2800 MPa. It was shown in Ref. 18 that HN±BN minicomposites with dierent RT strengths had the same rupture behavior at low and high temperatures as the ®bers [i.e. same slope on ln() vs. q]. However, the absolute rupture strengths were lower for the lower RT strength minicomposites in proportion to the absolute strength of the as-produced ®bers. Therefore, normalizing the RT as-produced ®ber strength to 2100 MPa was used for comparison of the as-produced ®ber rupture properties with the minicomposite stress-rupture data. If the ®ber degradation is merely due to the ®ber degradation observed for asproduced ®bers, the minicomposite rupture data should decrease the same as the ®ber data if the starting strength of the ®bers were 2100 MPa. The C±HN and BN±HN minicomposite properties are closer to the individual ®ber rupture behavior at low and high temperatures. At intermediate temperatures there is degradation of the minicomposite behavior compared to the individual ®ber because of the embrittlement associated with ®ber±matrix bonding due to SiO2 formation. Fig. 8 shows the improvement in rupture properties of BN interphase over the C interphase. It is also clear from Fig. 8 that C±NIC suered more serious ®ber degradation due to oxidation than C±HN. At 700C, the loss of carbon by oxidation results in long load-bearing gage lengths for the HN ®bers. The oxidation kinetics of a continuous carbon interphase in a composite has been published in the literature.19,20 If we assume that the oxidation is controlled by the diusion of oxygen, the recession length () is: k1=2 p t 1=2 4 where kp is the parabolic rate constant and t the time. For an opening of 0.4 mm (carbon interface thickness), Fig. 7. Fracture surface of a C±HN minicomposite for stress rupture at 1200C (439 MPa, 217 h). 2632 J. Marti nez-FernaÂndez, G.N. Morscher / Journal of the European Ceramic Society 20 (2000) 2627±2636
dez, G N. Morscher /Journal of the Europ amic Society 20 (2000)2627-2636 HN As-produced HN minicomposites at 1200C still could be due to a (o RT, 2800 MPa) slower recession of the bn interface 4.2. Dependence of the stress-rupture behavior on precrack load The load-time curve of Fig. I indicates that cracks are being created in the minicomposite up to failure. It clear then that the increase of the precrack load from ° C Experiments9so" CExperiments1200°c 280 to 295 MPa implies an increase in the number of 000 35000 40000 matrix cracks. Based on the direct relationship betwee the number of cracks and AE events(Table Al)and the Fig.8. Plot of the stress on fibers if fully loaded vS LM parameter. number of events recorded for several samples as Previous data included. See text for further discussion function of load(Fig. 2), there would be one or two cracks in the hot zone after the 119n precrack load For the 126 n precrack load, between two and four kp is 10-m2s-,and approximately 6 mm of carbon cracks would be estimated in the hot zone. A larger interphase length would be lost in about I h. This would number of cracks cause faster oxidation of the entire correspond to the entire hot zone gage length for a phase in the hot zone. This results in a larger length minicomposite with one crack in the hot zone region. of the fiber holding the maximum load (at least for short The fiber length exposed to the maximum load when the tests), more environmental degradation of the fibers, interphase was completely oxidized was about 25 times and more locations for stress concentration associated larger than the fiber length if there was no oxidation of with SiOz formation and the strong bonding between the interface(assuming one crack, as in the Introduc- fiber and matrix. all these factors would cause a decrease ion). This difference in load-bearing length will result inin the stress-rupture survival time. During the precrack decrease in strength of 47 to 28% for a Weibull mod- loading step it was rather common(approximately 25% ulus from 5 to 10. For more than one crack in the hot of the cases) for composites to fail between 290 to 295 zone, a smaller reduction in strength would occur. This MPa. This indicates that, in addition to an increase in the gage-length effect best explains the decrease of strength crack density, a greater occurrence of fiber failure would of C-Hn compared with BN-Hn since the rupture be occurring in minicomposites precracked at 295 MPa strength of the C-Hn minicomposites are approxi than at 280 MPa, and could also contribute to the poorer mately 25% reduced compared to the shorter gage stress-rupture behavior compared to samples precracked length loaded BN-HN minicomposites at 280 MPa(Fig. 6). At 1200oC, failure is dominated by The presence of fibers with no pullout on the fracture the creep-rupture behavior of the fibers, explaining the surface was observed for some of the fibers on the smaller dependence of the survival time with precrack minicomposite fracture surfaces tested at 700C and load for this temperature occurred for all of the fibers on the minicomposite frac ture surfaces tested at 950C. This indicates that strong 43. Fatigue bonding of the fibers to the matrix was a factor at these temperatures. When one or more fibers failed at or The results from fatigue experiments(Fig. 5)clearly away from the matrix crack, the added load to the indicate that it is at intermediate temperatures when the neighboring fibers in the matrix crack would be relative movement of matrix and fiber, due to fatigue, enhanced due to strong bonding resulting in local stress- affect the survival time most dramatically. This is pre- concentrations. This causes most of the fibers to pre- sumably due to the local concentration of stress during Terentially fail in the plane of the matrix crack. The fatigue on the sites where the fiber and matrix are bon longer recession distance with C interphases not only ded because of the Sio2 formation. At 1200 C, the increases the fiber gage length, it also allows fibers to minicomposite failed about 2 cm from the center of the bond more quickly compared to BN interphases. The furnace. In these regions the temperature is approxi- BN interphase only recesses a few microns separating mately 900@C. This shows that the resistance to fatigue the fiber from the matrix. For C interphase mini- is worse at intermediate temperatures than at 1200 C, omposites, after total C interphase removal, the fibers even though a greater amount of oxide reaction produc are free to move towards and contact the matrix is formed at 1200 C in the matrix cracks than at inter. At 1200 C, SiO2 formation is extensive. The stress- mediate temperatures. SiO2 at 1200C does flow to some upture properties appear to be controlled by the creep extent and may relieve some of the stress-concentrations of the fibers and are similar for both minicomposite produced at the fiber-SiOr-matrix bond compared to ystems. The differences between C and BN interphase lower temperatures where no relaxation of the glass
kp is 10ÿ8 m2 sÿ1 , 20 and approximately 6 mm of carbon interphase length would be lost in about 1 h. This would correspond to the entire hot zone gage length for a minicomposite with one crack in the hot zone region. The ®ber length exposed to the maximum load when the interphase was completely oxidized was about 25 times larger than the ®ber length if there was no oxidation of the interface (assuming one crack, as in the Introduction). This dierence in load-bearing length will result in a decrease in strength of 47 to 28% for a Weibull modulus from 5 to 10. For more than one crack in the hot zone, a smaller reduction in strength would occur. This gage-length eect best explains the decrease of strength of C±HN compared with BN±HN since the rupture strength of the C±HN minicomposites are approximately 25% reduced compared to the shorter gagelength loaded BN±HN minicomposites. The presence of ®bers with no pullout on the fracture surface was observed for some of the ®bers on the minicomposite fracture surfaces tested at 700C and occurred for all of the ®bers on the minicomposite fracture surfaces tested at 950C. This indicates that strong bonding of the ®bers to the matrix was a factor at these temperatures. When one or more ®bers failed at or away from the matrix crack, the added load to the neighboring ®bers in the matrix crack would be enhanced due to strong bonding resulting in local stressconcentrations. This causes most of the ®bers to preferentially fail in the plane of the matrix crack. The longer recession distance with C interphases not only increases the ®ber gage length, it also allows ®bers to bond more quickly compared to BN interphases. The BN interphase only recesses a few microns separating the ®ber from the matrix. For C interphase minicomposites, after total C interphase removal, the ®bers are free to move towards and contact the matrix. At 1200C, SiO2 formation is extensive. The stressrupture properties appear to be controlled by the creep of the ®bers and are similar for both minicomposite systems. The dierences between C and BN interphase HN minicomposites at 1200C still could be due to a slower recession of the BN interface. 4.2. Dependence of the stress±rupture behavior on precrack load The load±time curve of Fig. 1 indicates that cracks are being created in the minicomposite up to failure. It is clear then that the increase of the precrack load from 280 to 295 MPa implies an increase in the number of matrix cracks. Based on the direct relationship between the number of cracks and AE events (Table A1) and the number of events recorded for several samples as a function of load (Fig. 2), there would be one or two cracks in the hot zone after the 119 N precrack load. For the 126 N precrack load, between two and four cracks would be estimated in the hot zone. A larger number of cracks cause faster oxidation of the entire interphase in the hot zone. This results in a larger length of the ®ber holding the maximum load (at least for short tests), more environmental degradation of the ®bers, and more locations for stress concentration associated with SiO2 formation and the strong bonding between ®ber and matrix. All these factors would cause a decrease in the stress-rupture survival time. During the precrack loading step, it was rather common (approximately 25% of the cases) for composites to fail between 290 to 295 MPa. This indicates that, in addition to an increase in the crack density, a greater occurrence of ®ber failure would be occurring in minicomposites precracked at 295 MPa than at 280 MPa, and could also contribute to the poorer stress-rupture behavior compared to samples precracked at 280 MPa (Fig. 6). At 1200C, failure is dominated by the creep-rupture behavior of the ®bers, explaining the smaller dependence of the survival time with precrack load for this temperature. 4.3. Fatigue The results from fatigue experiments (Fig. 5) clearly indicate that it is at intermediate temperatures when the relative movement of matrix and ®ber, due to fatigue, aect the survival time most dramatically. This is presumably due to the local concentration of stress during fatigue on the sites where the ®ber and matrix are bonded because of the SiO2 formation. At 1200C, the minicomposite failed about 2 cm from the center of the furnace. In these regions the temperature is approximately 900C. This shows that the resistance to fatigue is worse at intermediate temperatures than at 1200C, even though a greater amount of oxide reaction product is formed at 1200C in the matrix cracks than at intermediate temperatures. SiO2 at 1200C does ¯ow to some extent and may relieve some of the stress-concentrations produced at the ®ber±SiO2±matrix bond compared to lower temperatures where no relaxation of the glass Fig. 8. Plot of the stress on ®bers if fully loaded vs LM parameter. Previous data included.1 See text for further discussion. J. Marti nez-FernaÂndez, G.N. Morscher / Journal of the European Ceramic Society 20 (2000) 2627±2636 2633
J. Martinez.Fernandez, G N Morscher Journal of the European Ceramic Society 20(2000)2627-2636 product would be activity in order to capture and digitize the waveform. matrix are not fully bonded as evidenced by the occurrence This approach provides greater accuracy in sorting out of fiber pullout(Fig. 7)and the differences between fatigue gage events and correlating individual events with phy- and stress rupture are not as significant as at 950%C. sical sources. 2 Fig. Al shows a schematic of the tensile test. For most tests, only two wide band(50 kHz to 2 4.4. Retained strength MHz) sensors(Model B1080, Digital Wave Corpora tion)were attached with alligator clips to the epoxy. For The retained strengths at room temperature of C-hn two experiments, four wide band sensors were attached minicomposites that did not fail during the rupture test to the epoxy, two wide-band and two resonant fre- are listed for C-HN and BN-HN' minicomposites in quency(Model"Pico", Physical Acoustics Corporation) Table 1. For rupture conditions at 700 and 900oC, the (the same as used in the earlier studies). The data was etained strength of BN-HN minicomposites are sig collected on a Digital Wave(Englewood, CO) Fracture nificantly better than for C-HN minicomposites. This is Wave Detector. After the test, the same software used most likely due to the slower recession of bn which to collect the data was used to analyze the data. For enables the fibers to be separated from the SiC matrix. experiments with only two sensors, 100 to 200 events Higher temperatures are required to form enough oxide were recorded. For experiments with four sensors, up to reaction product, for BN-HN minicomposites, in order to 600 events were recorded because the resonant fre- fill the gap between the fibers and matrix. In comparison, quency sensors were much more sensitive, especially to the fibers in C-hn minicomposites are observed to be in lower frequency AE. intimate contact with the matrix shortly after the C layer Post-test analysis consisted first of determining the was removed by oxidation at 700C(e.g. Fig. 6). After speed of sound through the sample in order to locate the 1200C rupture conditions, the vacated interphases are sources of the aE events and to sort out events which completely filled by the glass reaction products for both occurred outside of the gage section. Pencil-lead breaks ystems and the retained strength data for C-Hn and had been performed prior to a tensile test on the epoxy BN-HN minicomposites are nearly identical of an undamaged specimen outside of the sensors, so that the sound waves produced by the fracture of the pencil-lead traveled from one sensor to the next, and the 5. Conclusions ae data was saved on a separate file. The maximum difference in time of arrival, Atr from one sensor to the The stress-rupture properties of HN fiber-reinforced next was determined from the first peak of the sound CVI-SiC minicomposites with carbon interphase were wave(extensional wave) received on both sensors from tudied at temperatures ranging from 700 to 1200oC The stress-rupture lives of C-HN are by far superior to C-NIC, presumably due to the lack of fiber decom- position/reaction during CVI SiC processing that occurs for NIC fibers. This study demonstrates that C-HN minicomposites have worse mechanical properties at 700 and 950C than the BN-Hn minicomposites pre- viously studied, due to the removal of carbon inter phases and the ease with which fibers bond to the natrix. Therefore, BN interphases are more envir onmentally stable in the intermediate temperature regime. At 1200C, the improvement of the BN inter- phase over the carbon interphase was not quite as sig nificant as at the lower temperatures; the stress-rupture and fatigue properties predominantly being controlled by the fiber properties. This study also shows that Ae emission can be used on this system as a reliable quan titative method to monitor damage in this system Appendix. Acoustic emission detection and analysis of minicomposite tensile tests The AE set-up was similar to previous studies. 21 22 However, wide-band sensors were used to detect AE Fig. Al. Schematic of room temperature tensile test
product would be expected. At 700C, the ®ber and matrix are not fully bonded as evidenced by the occurrence of ®ber pullout (Fig. 7) and the dierences between fatigue and stress rupture are not as signi®cant as at 950C. 4.4. Retained strength The retained strengths at room temperature of C±HN minicomposites that did not fail during the rupture test are listed for C±HN and BN±HN2 minicomposites in Table 1. For rupture conditions at 700 and 900C, the retained strength of BN±HN minicomposites are signi®cantly better than for C±HN minicomposites. This is most likely due to the slower recession of BN which enables the ®bers to be separated from the SiC matrix. Higher temperatures are required to form enough oxide reaction product, for BN±HN minicomposites, in order to ®ll the gap between the ®bers and matrix. In comparison, the ®bers in C±HN minicomposites are observed to be in intimate contact with the matrix shortly after the C layer was removed by oxidation at 700C (e.g. Fig. 6). After 1200C rupture conditions, the vacated interphases are completely ®lled by the glass reaction products for both systems and the retained strength data for C±HN and BN±HN minicomposites are nearly identical. 5. Conclusions The stress±rupture properties of HN ®ber-reinforced CVI±SiC minicomposites with carbon interphase were studied at temperatures ranging from 700 to 1200C. The stress-rupture lives of C±HN are by far superior to C±NIC, presumably due to the lack of ®ber decomposition/reaction during CVI SiC processing that occurs for NIC ®bers. This study demonstrates that C±HN minicomposites have worse mechanical properties at 700 and 950C than the BN±HN minicomposites previously studied, due to the removal of carbon interphases and the ease with which ®bers bond to the matrix. Therefore, BN interphases are more environmentally stable in the intermediate temperature regime. At 1200C, the improvement of the BN interphase over the carbon interphase was not quite as signi®cant as at the lower temperatures; the stress±rupture and fatigue properties predominantly being controlled by the ®ber properties. This study also shows that AE emission can be used on this system as a reliable quantitative method to monitor damage in this system. Appendix. Acoustic emission detection and analysis of minicomposite tensile tests The AE set-up was similar to previous studies.21,22 However, wide-band sensors were used to detect AE activity in order to capture and digitize the waveform. This approach provides greater accuracy in sorting out gage events and correlating individual events with physical sources.22 Fig. A1 shows a schematic of the tensile test. For most tests, only two wide band (50 kHz to 2 MHz) sensors (Model B1080, Digital Wave Corporation) were attached with alligator clips to the epoxy. For two experiments, four wide band sensors were attached to the epoxy, two wide-band and two resonant frequency (Model ``Pico'', Physical Acoustics Corporation) (the same as used in the earlier studies). The data was collected on a Digital Wave (Englewood, CO) Fracture Wave Detector. After the test, the same software used to collect the data was used to analyze the data. For experiments with only two sensors, 100 to 200 events were recorded. For experiments with four sensors, up to 600 events were recorded because the resonant frequency sensors were much more sensitive, especially to lower frequency AE. Post-test analysis consisted ®rst of determining the speed of sound through the sample in order to locate the sources of the AE events and to sort out events which occurred outside of the gage section. Pencil-lead breaks had been performed prior to a tensile test on the epoxy of an undamaged specimen outside of the sensors, so that the sound waves produced by the fracture of the pencil-lead traveled from one sensor to the next, and the AE data was saved on a separate ®le. The maximum dierence in time of arrival, tx, from one sensor to the next was determined from the ®rst peak of the sound wave (extensional wave) received on both sensors from Fig. A1. Schematic of room temperature tensile test. 2634 J. Marti nez-FernaÂndez, G.N. Morscher / Journal of the European Ceramic Society 20 (2000) 2627±2636
J. Martinez.Fernandez, G N Morscher Journal of the European Ceramic Society 20(2000)2627-2636 Table al AE and matrix crack data for several room temperature tested minicomposites Specimen Gage Speed of Number Failure Crack nated number length unda(m/s) of events load (N) with highest ks in ga two decades of AE energy 928 1.67 a Extensional wave the lead-break event. As a composite forms matrix The reduction in speed of sound is about 12%. Since cracks during testing the elastic modulus decreases and the speed of sound of the extensional wave is directly pro- the speed of sound decreases resulting in an increase in portional to the square root of the elastic modulus, the Alx. For the tensile experiment Ae data, the maximum elastic modulus is reduced by approximately 23%. This difference in time of arrival was determined as a func- is not as great a reduction in elastic modulus as in woven A2 shows the decrease in speed of sound of the exten The events that were determined to occur outside of sional wave with load. This corresponds to the reduc- the gage section, i.e. events with difference in time of tion in square root of the elastic modulus of the arrival, 4t, equal to 4tx, were removed from the data specimen as matrix cracking occurs set. The location of each event along the gage length of the specimen could then be determined based on At, Alr,and the sensor which received the signal first. 22 Fig A3 shows an example of event location as a func- tion of load. The acoustic energy was determined for 11400 each event. It was observed that for the specimens where only two sensors were used, all of the AE event 11200 energies were within two decades of energy from one another. However, for the specimens where four sensors were used, the spread in AE energies was over three orders of magnitude. Also, many AE events were trig 10400 gered by the more sensitive resonant frequency trans ducers. These events would never have triggered the wide-band sensors alone at the preamplification settings used. The events were therefore filtered according to ae Fig. A2. Speed of sound vs load for specimen h3hc70 energy where only the two highest decades of events were used to relate to the accumulated d Table Al shows the event statistics along with the sured crack spacings for five specimens. Note the excel lent agreement between the estimated number of cracks over the entire gage length of the samples compared to 0 he highest two orders of magnitude AE energy events The cumulative number and energies of the sorted AE events were used for Figs. 2 and 3 References L. Heredia, F. E, McNulty, J C, Zok, F. w. and Evans, A. G, Am. Ceram.Soc,1995,78.2097-2100 Fig. A3. Location of the ae events along the length of the mini 2. Morscher. G. N, Tensile stress rupture of SiCf/ Sicm mini composite(h3hc70)during a tensile monotonic loading test as a func- mposites with carbon and boron nitride interphases at elevated
the lead-break event.22 As a composite forms matrix cracks during testing the elastic modulus decreases and the speed of sound decreases resulting in an increase in tx. For the tensile experiment AE data, the maximum dierence in time of arrival was determined as a function of load and the speed of sound determined from the distance between the two sensors divided by tx. Fig. A2 shows the decrease in speed of sound of the extensional wave with load. This corresponds to the reduction in square root of the elastic modulus of the specimen as matrix cracking occurs. The reduction in speed of sound is about 12%. Since the speed of sound of the extensional wave is directly proportional to the square root of the elastic modulus, the elastic modulus is reduced by approximately 23%. This is not as great a reduction in elastic modulus as in woven macrocomposites, 60%, that reach a state of matrix crack saturation at relatively low composite stresses.22 The events that were determined to occur outside of the gage section, i.e. events with dierence in time of arrival, Dt, equal to Dtx, were removed from the data set. The location of each event along the gage length of the specimen could then be determined based on t, tx, and the sensor which received the signal ®rst.22 Fig. A3 shows an example of event location as a function of load. The acoustic energy was determined for each event. It was observed that for the specimens where only two sensors were used, all of the AE event energies were within two decades of energy from one another. However, for the specimens where four sensors were used, the spread in AE energies was over three orders of magnitude. Also, many AE events were triggered by the more sensitive resonant frequency transducers. These events would never have triggered the wide-band sensors alone at the preampli®cation settings used. The events were therefore ®ltered according to AE energy where only the two highest decades of energy events were used to relate to the accumulated damage. Table A1 shows the event statistics along with the measured crack spacings for ®ve specimens. Note the excellent agreement between the estimated number of cracks over the entire gage length of the samples compared to the highest two orders of magnitude AE energy events. The cumulative number and energies of the sorted AE events were used for Figs. 2 and 3. References 1. Heredia, F. E., McNulty, J. C., Zok, F. W. and Evans, A. G., Oxidation embrittlement probe for ceramic-matrix composites. J. Am. Ceram. Soc., 1995, 78, 2097±2100. 2. Morscher, G. N., Tensile stress rupture of SiCf/SiCm minicomposites with carbon and boron nitride interphases at elevated temperatures in air. J. Am. Ceram. Soc., 1997, 80, 2029±2042. Fig. A2. Speed of sound vs load for specimen h3hc70. Table A1 AE and matrix crack data for several room temperature tested minicomposites Specimen Gage length (mm) Speed of sounda (m/s) Number of events Failure load (N) Crack spacing (mm) No. of events with highest two decades of AE energy Estimated number of cracks in gage length h3hc39 140 13 861 212 151 ± 135 ± h3hc40 158 15 192 579 152 0.56 238 250 h3hc70 160 11 851 149 143 1.05 149 133 h3hc76 168 13 228 132 143 0.96 132 146 h3hc77 170 12 500 87 138 1.67 87 84 a Extensional wave. Fig. A3. Location of the AE events along the length of the minicomposite (h3hc70) during a tensile monotonic loading test as a function of time. J. Marti nez-FernaÂndez, G.N. Morscher / Journal of the European Ceramic Society 20 (2000) 2627±2636 2635
2636 J. Martinez.Fernandez, G N Morscher Journal of the European Ceramic Society 20(2000)2627-2636 3. Lara-Curzio E. ferber. m. k and tortorelli. lr nitride layers in composites. J. Am. Ceram Soc., 1999, 82. 1473- Oxidation and Stress-Rupture of Nicalon TM/SiC Ci Intermediate Temperatures. Key Engineering Material 13. Morscher. G. N. Martinez-Fernandez 127-131. Trans. Tech. Publications, Switzerland, 1997. pp Determination of interfacial gle fiber microcomposite test. J. Am. Ceram Soc. 9.1083-1091 4. Lara-Curzio, E. Stress-rupture of Nicalon/ SiC continuous fiber 14. Yun, H. M. and DiCarlo, J. A, Thermomechanical behavior of ramic matrix composites in air at 950C.J. Am. Ceram. Soc. advanced SiC fiber multifilament tows. Ceram. Eng. Sci. Proc. 1997,80.3268-3272. 1996,17,61-67 5. Lipetzky, P, Stoloff, N. S. and Dvorak, G. J, Atmospher 15. Larson, F.R. and Miller, J, A time-temperature relationship for effects on high-temperature lifetime of ceramic composites. rupture and creep stresses. Trans. ASME. 1952, 74. 765 Ceran. Eng. Sci. Proc., 1997, 18. 355-362. 16. Conway, J. B, Numerical methods for creep and rupture ana- 6. Verrilli, M. J, Calomino, A. M. and Brewer, D. N, Creep- lyses. Gordon and Breach, Science Publ, New York, pp. 155- pture behavior of a Nicalon/SiC composite echanical Test Methods and Behavior of Continuous-Fiber 17. Yun H. M. and DiCarlo, J. A, Time/ temperature dependent Ceramic Composites(ASTM STP 1309). ed M. G. Jenkins. s. T Gonczy, E. Lara-Curzio, N. E. Ashbaugh and L. Zawada. ASTM,1997,pp.158-175 IT,ed.NPBansal and J.P. Singh, 1996, pp 17-20 omposie tensile strength of Sic and AlO3-based fibers. In cerami Transactions. Vol. 74. Advances in Ceramic-Matrix 7. Steyer, T.E., Zok, F. W.and Walls, D. P, Stress rupture of an 18. Morscher, G. N, The effect of static and cyclic tensile stress and enhanced Nicalon SiC composite at intermediate tempera temperature on failure for precracked Hi-Nicalon/BN/CVD SiC tures.J.Am. Ceran.Soc.,1998,81,2140-2146. minicomposites in air. Ceram. Eng. Sci. Proc., 1997, 18, 737-745 8. Curtin, w.A., Theory of mechanical properties of ceramic- 19. Filipuzzi, L. Camus, G. Naslain, R and Thebault, J, Oxidation matrix composites. J. Am. Ceram Soc., 1991, 74, 2837-2845 mechanisms and kinetics of lD-SiC/C/SiC composite materials: Sawyer, L C, Jamieson, M. Brikowski, D, Haider, M. I and L, an experimental approach. J. Am. Ceram. Soc., 1994. 77(2), Chen, R. T, Strength, structure, and fracture properties of 459-46 ceramic fibers produced from polymeric precursors: I, base-line 20. J and Cawley, studies. J. Am. Ceram. Soc., 1987, 70, 798-810 hase in a nonreactive matrix. J./m. Cera. Soc.. 1995, 78. 972- 10. Naslain, R, Fiber-matrix interphases and interfaces in ceramic atrix composites processed by CVI. Composite Interfaces, 1993, 21. Morscher. g.n. and martinez. Fernandez. j. fiber effects on 1,253-286 minicomposite mechanical properties for several silicon carbide I1. Lin, H. T. and Becher, P. F, Effect of coating on lifetime of fiber-chemically vapor-infiltrated silicon carbide matrix systems nicalon fiber-silicon carbide composites in air. Mater J.Am. Ceran.Soc.,l999,82,145-155 A231,143-150 12. Jacobson, N.S., Morscher, G. N Bryant, D. R and Tressler, R. C/SiC composite. Comp. Sci. Tech, 1999, E, High-temperature oxidation of boron nitride: Il, boron 687-697
3. Lara-Curzio, E., Ferber, M. K. and Tortorelli, P. F., Interface Oxidation and Stress±Rupture of Nicalon, TM/SiC CFCCs at Intermediate Temperatures. Key Engineering Materials, Vols. 127±131. Trans. Tech. Publications, Switzerland, 1997, pp. 1069± 1082. 4. Lara-Curzio, E., Stress±rupture of Nicalon/SiC continuous ®ber ceramic matrix composites in air at 950C. J. Am. Ceram. Soc., 1997, 80, 3268±3272. 5. Lipetzky, P., Stolo, N. S. and Dvorak, G. J., Atmospheric eects on high-temperature lifetime of ceramic composites. Ceram. Eng. Sci. Proc., 1997, 18, 355±362. 6. Verrilli, M. J., Calomino, A. M. and Brewer, D. N., Creep± rupture behavior of a Nicalon/SiC composite. In Thermal and Mechanical Test Methods and Behavior of Continuous-Fiber Ceramic Composites (ASTM STP 1309), ed. M. G. Jenkins, S. T. Gonczy, E. Lara-Curzio, N. E. Ashbaugh and L. Zawada. ASTM, 1997, pp. 158±175. 7. Steyer, T. E., Zok, F. W. and Walls, D. P., Stress rupture of an enhanced NicalonTM/SiC composite at intermediate temperatures. J. Am. Ceram. Soc., 1998, 81, 2140±2146. 8. Curtin, W. A., Theory of mechanical properties of ceramicmatrix composites. J. Am. Ceram. Soc., 1991, 74, 2837±2845. 9. Sawyer, L. C., Jamieson, M., Brikowski, D., Haider, M. I. and Chen, R. T., Strength, structure, and fracture properties of ceramic ®bers produced from polymeric precursors: I, base-line studies. J. Am. Ceram. Soc., 1987, 70, 798±810. 10. Naslain, R., Fiber-matrix interphases and interfaces in ceramic matrix composites processed by CVI. Composite Interfaces, 1993, 1, 253±286. 11. Lin, H. T. and Becher, P. F., Eect of coating on lifetime of nicalon ®ber-silicon carbide composites in air, Mater. Sci. Eng., A231, 143±150. 12. Jacobson, N. S., Morscher, G. N., Bryant, D. R. and Tressler, R. E., High-temperature oxidation of boron nitride: II, boron nitride layers in composites. J. Am. Ceram. Soc., 1999, 82, 1473± 1482. 13. Morscher, G. N., Martinez-Fernandez, J. and Purdy, M. J., Determination of interfacial properties using a single ®ber microcomposite test. J. Am. Ceram. Soc., 1996, 79, 1083±1091. 14. Yun, H. M. and DiCarlo, J. A., Thermomechanical behavior of advanced SiC ®ber multi®lament tows. Ceram. Eng. Sci. Proc., 1996, 17, 61±67. 15. Larson, F. R. and Miller, J., A time-temperature relationship for rupture and creep stresses. Trans. ASME, 1952, 74, 765. 16. Conway, J. B., Numerical methods for creep and rupture analyses. Gordon and Breach, Science Publ., New York, pp. 155± 177. 17. Yun H. M. and DiCarlo, J. A., Time/temperature dependent tensile strength of SiC and Al2O3-based ®bers. In Ceramic Transactions, Vol. 74, Advances in Ceramic-Matrix Composites III, ed. N. P. Bansal and J. P. Singh, 1996, pp. 17±26. 18. Morscher, G. N., The eect of static and cyclic tensile stress and temperature on failure for precracked Hi-Nicalon/BN/CVD SiC minicomposites in air. Ceram. Eng. Sci. Proc., 1997, 18, 737±745. 19. Filipuzzi, L., Camus, G., Naslain, R. and Thebault, J., Oxidation mechanisms and kinetics of 1D-SiC/C/SiC composite materials: I, an experimental approach. J. Am. Ceram. Soc., 1994, 77(2), 459±466. 20. Eckel, A. J. and Cawley, J. D., Oxidation of a continuos carbon phase in a nonreactive matrix. J. Am. Ceram. Soc., 1995, 78, 972± 980. 21. Morscher, G. N. and Martinez-Fernandez, J., Fiber eects on minicomposite mechanical properties for several silicon carbide ®ber-chemically vapor-in®ltrated silicon carbide matrix systems. J. Am. Ceram. Soc., 1999, 82, 145±155. 22. Morscher, G. N., Modal acoustic emission of damage accumulation in a woven SiC/SiC composite. Comp. Sci. Tech., 1999, 59, 687±697. 2636 J. Marti nez-FernaÂndez, G.N. Morscher / Journal of the European Ceramic Society 20 (2000) 2627±2636