journal J.Am. Ceram.Soc,812]329-3602000) Fugitive Interfacial Carbon Coatings for Oxide/Oxide Composites Kristin A. Keller, * T Tai-lI Mah, * T Triplicane A. Parthasarathy, f and Charles M. Cooke? Air Force Research Laboratory, Materials and Manufacturing Directorate, AFRL/MLLN, Wright-Patterson AFB The effectiveness of fugitive interfacial carbon coatings in prevents fiber/matrix reaction during processing an NextelM 720-based composites was investigated. Dense defines th of the interfacial gap. In the case of an idealized >90%) matrix( calcium aluminosilicate,0°and±45°)com p, the fiber would be unconstrained after carbon ites and porous matrix (mullite/alumina, eight-harness removal, with no load transfer from the matrix to the fiber(Fig atin fabric) composites were fabricated and tensile tested in I(a)). The fibers would easily pull out from the fractured matrix two control conditions(uncoated or carbon-coated) and with and the corresponding pull-out lengths would be extremely long the carbon removed ( fugitive interface). Results indicated that In reality, however, load transfer occurs by the combined effects arbon removal in dense matrix composites did not signifi- of mechanical interlocking, intermittent fiber/matrix bonding, and antly change unidirectional composite strength, even after roughness of both the fiber and matrix. Figure 1(b) illustrates this long-term exposure at 1000C. For porous matrix composites, scenario, where a crack is shown approaching a rough fiber/matrix composite strength was independent of the fiber/matrix inter- interface. As the crack approaches the interface, matrix strain face, even after exposure at 1150%C for 500 h in air. results in sliding at the fiber/matrix interface, leading to point contacts between the fiber and matrix where mechanical interlock- ing and possible bonding occurs. Upon further stressing, the fiber L. Introduction Qr is well-known that the properties of ceramic composites can be experienced ahead of the crack tip results in the loading of the fiber otinv The primary approach toward this control has been could be activated resulting in the failure of the fiber. Pull-out lengths in this case would be much shorter and would depend on provide interfaces that ultimately increase the strain-to-failure of urface roughnesses. The viability of this concept has been the composites through crack deflection and fiber pullout, A demonstrated previously variety of coatings have been developed for use in ceramic-matri One of the most important factors associated with these coatings composites(CMCs)- with vary ing degrees of success, however, is the thickness of the carbon, which defines the gap. If the BN and carbon-4 remain the most widely used interface too wide. there would be little interaction between fiber and materials matrix. Presently, there is littile quantitative information as to an Carbon forms a weak interface between the fiber and the matrix optimal thickness. It is expected that the optimal thickness is in a composite, as first demonstrated in the NicalonM(Nippon tem-specific because of several factors, such as coefficient of Carbon Co., Tokyo, Japan) fiber/glass matrix composites, where thermal expansion mismatch and roughness effects of the fiber and matrIX the carbon layer is formed in situ. 4 The problem inherent with Composites with a fugitive gap at the interface also can be these Nicalon-based composites is the oxidation of the in situ thought of as one extreme in a range of composites wsa rated carbon layer at elevated temperatures, which results in a rapid illing of the interface with a SiO, reaction product from the ifferent distributions in the matrix. If the pores are concer fiber. This strongly bonded interface, in turn, leads to cata- at the interface, the interface is a gap. If the pores are distributed but still very close to the interface, the interface has a porous strophic failure of the specimens. However, work completed by interlayer. 7, io If the pores are distributed uniformly over the entire Plucknett et al.- indicated that Nicalon/calcium aluminosilicate (CAS)composite strength could be retained if the in situ carbon matrix, the composite is a porous matrix composite. Note that the was removed at a temperature below SiO, formation(450C) overall porosity in these composites varies significantly because of ave not This suggests that, for composites containing oxidatively stable the difference in the volume of th porous material. There fibers and matrices, there is a possibility of using a gap at the been any studies aimed at understanding the relative merits of interface to protect the fiber and achieve good strength. This these variations in composite design. In this work, we attempt to principle can be used to build composites with fugitive coat- compare the strength of a porous matrix composite with that of one The fugitive coating concept relies on the retention of a carbon Another objective of the present work is to examine the benefits and effectiveness of using fugitive coatings in both dense matrix coating during composite processing and its subsequent removal and porous matrix through oxidation. This removal leaves an unbonded fiber/matrix osites. both uncoated fibers and carbon- coated fibers embedded in dense and porous matrices are used as interface and, for most choices of coating thickness and coefficient control composites. The effect of fugitive gap thickness on the p. The carbon primarily modulus and strength of both unidirectional and +45 composites is studied. Finally, the relative merits of dense matrix composites th fugitive coatings and porous matrix composites are discussed F. Zok--contributing edito in light of the presented results I. Experimental Procedures () Dense Matrix Com rted by the Air Fe Laboratory under No.F3361596 Fiber tows(Nextel M 720, 3M, St. Paul, MN) were nominally coated with either a 0.04 or a 0.02 um thick carbon coa UES, Inc, Dayton, (Synterials, Inc Herndon, VA). These coatings were deposited
Fugitive Interfacial Carbon Coatings for Oxide/Oxide Composites Kristin A. Keller,* ,† Tai-Il Mah,* ,† Triplicane A. Parthasarathy,* ,† and Charles M. Cooke† Air Force Research Laboratory, Materials and Manufacturing Directorate, AFRL/MLLN, Wright–Patterson AFB, Ohio 45433 The effectiveness of fugitive interfacial carbon coatings in Nextel™ 720-based composites was investigated. Dense (>90%) matrix (calcium aluminosilicate, 0° and 645°) composites and porous matrix (mullite/alumina, eight-harness satin fabric) composites were fabricated and tensile tested in two control conditions (uncoated or carbon-coated) and with the carbon removed (fugitive interface). Results indicated that carbon removal in dense matrix composites did not significantly change unidirectional composite strength, even after long-term exposure at 1000°C. For porous matrix composites, composite strength was independent of the fiber/matrix interface, even after exposure at 1150°C for 500 h in air. I. Introduction I T IS well-known that the properties of ceramic composites can be optimized if control over the fiber/matrix interface can be gained.1–4 The primary approach toward this control has been through the application of fiber coatings, which provide weakly bonded interfaces that ultimately increase the strain-to-failure of the composites through crack deflection and fiber pullout. A variety of coatings have been developed for use in ceramic-matrix composites (CMCs)5–17 with varying degrees of success; however, BN18–21 and carbon22–24 remain the most widely used interface materials. Carbon forms a weak interface between the fiber and the matrix in a composite, as first demonstrated in the Nicalon™ (Nippon Carbon Co., Tokyo, Japan) fiber/glass matrix composites, where the carbon layer is formed in situ. 23,25 The problem inherent with these Nicalon-based composites is the oxidation of the in situ carbon layer at elevated temperatures, which results in a rapid filling of the interface with a SiO2 reaction product from the fiber.26 This strongly bonded interface, in turn, leads to catastrophic failure of the specimens. However, work completed by Plucknett et al.27 indicated that Nicalon/calcium aluminosilicate (CAS) composite strength could be retained if the in situ carbon was removed at a temperature below SiO2 formation (;450°C). This suggests that, for composites containing oxidatively stable fibers and matrices, there is a possibility of using a gap at the interface to protect the fiber and achieve good strength. This principle can be used to build composites with fugitive coatings.13,28 The fugitive coating concept relies on the retention of a carbon coating during composite processing and its subsequent removal through oxidation. This removal leaves an unbonded fiber/matrix interface and, for most choices of coating thickness and coefficient of thermal expansion mismatch, a gap. The carbon primarily prevents possible fiber/matrix reaction during processing and defines the width of the interfacial gap. In the case of an idealized smooth, straight gap, the fiber would be unconstrained after carbon removal, with no load transfer from the matrix to the fiber (Fig. 1(a)). The fibers would easily pull out from the fractured matrix and the corresponding pull-out lengths would be extremely long. In reality, however, load transfer occurs by the combined effects of mechanical interlocking, intermittent fiber/matrix bonding, and roughness of both the fiber and matrix. Figure 1(b) illustrates this scenario, where a crack is shown approaching a rough fiber/matrix interface. As the crack approaches the interface, matrix strain results in sliding at the fiber/matrix interface, leading to point contacts between the fiber and matrix where mechanical interlocking and possible bonding occurs. Upon further stressing, the fiber is prevented from slipping due to the roughness. The stress experienced ahead of the crack tip results in the loading of the fiber in an isolated region between the contact points. Flaws in the fiber could be activated, resulting in the failure of the fiber. Pull-out lengths in this case would be much shorter and would depend on surface roughnesses. The viability of this concept has been demonstrated previously.28 One of the most important factors associated with these coatings is the thickness of the carbon, which defines the gap. If the gap is too wide, there would be little interaction between fiber and matrix. Presently, there is little quantitative information as to an optimal thickness. It is expected that the optimal thickness is system-specific because of several factors, such as coefficient of thermal expansion mismatch and roughness effects of the fiber and matrix. Composites with a fugitive gap at the interface also can be thought of as one extreme in a range of composites with pores of different distributions in the matrix. If the pores are concentrated at the interface, the interface is a gap. If the pores are distributed but still very close to the interface, the interface has a porous interlayer.7,10 If the pores are distributed uniformly over the entire matrix, the composite is a porous matrix composite.29 Note that the overall porosity in these composites varies significantly because of the difference in the volume of the porous material. There have not been any studies aimed at understanding the relative merits of these variations in composite design. In this work, we attempt to compare the strength of a porous matrix composite with that of one with a fugitive gap. Another objective of the present work is to examine the benefits and effectiveness of using fugitive coatings in both dense matrix and porous matrix composites. Both uncoated fibers and carboncoated fibers embedded in dense and porous matrices are used as control composites. The effect of fugitive gap thickness on the modulus and strength of both unidirectional and 645° composites is studied. Finally, the relative merits of dense matrix composites with fugitive coatings and porous matrix composites are discussed in light of the presented results. II. Experimental Procedures (1) Dense Matrix Composites Fiber tows (Nextel™ 720, 3M, St. Paul, MN) were nominally coated with either a 0.04 or a 0.02 mm thick carbon coating (Synterials, Inc., Herndon, VA). These coatings were deposited F. Zok—contributing editor Manuscript No. 189762. Received November 5, 1998; approved July 15, 1999. Supported by the Air Force Research Laboratory under Contract No. F33615-96- C-5258. *Member, American Ceramic Society. † UES, Inc., Dayton, OH 45432. J. Am. Ceram. Soc., 83 [2] 329–36 (2000) 329
330 crosshead displacement were recorded by a software program MATRIX. FIBER MATRIX FIBER (TESTWORKSTM, MTS), while load and strain measurements were plotted using an X-y recorder. Tests were completed on both the 0°and±45° specimens. Elastic moduli and ultimate failure strengths of the composites were extracted from the measurements The fracture surfaces were also examined using Sem GAP→ (2) Porous Matrix Composites Nextel 720 fabric(eight-harness satin fabric) was carbon-coated by Oak Ridge National Laboratory using a CVD process. The -.I um thick coating appeared smooth and continuous. There was some isolated bridging of the coating, particularly at crossover ints in the weave. Control samples were also produced using coated. desized Nextel 720 fabric Five circular layers of fabric were stacked, bonded at the edges TRIX FIBER MATRIXFIBER Ising cellulose(MethocelTM, Dow Chemical Co., Midland, Mi), and subsequently hot molded to form a preform. This preform was then placed in a pressure infiltration vessel. An aqueous-based, nitric acid-stabilized slurry containing Al,O3(AKP-53, Sumitomo Chemical Co, Ltd, Tokyo, Japan) and mullite(MUL-SM, G=0 Baikowski International, Charlotte, NC)was prepared. The slurry composition was chosen to yield an 85/15 by weight mixture of AL,O, and SiO,, which matched the fiber composition. Both commercial powders were dispersed and sedimented prior to use, producing an average particle size of 0. 15 um for the Al,O3 and 0.38 um for the mullite(Model LS230 Particle Size Analyzer, Coulter Co., Miami, FL). The slurry was then poured into the Bridging Fiber filtration vessel, where matrix infiltration was assisted by applying air pressure(-0.55 MPa)to one side and vacuum(-003 MPa)to the other. As reported for other pressure-infiltrated ceramics, 30,3 Fig. 1. Concept of load transfer is illustrated for a composite using a infiltration followed parabolic rate kinetics, indicating that the cake thickness grew in proportion to the square root of time is straight, the fiber is protected from a matrix crack, but there is no load After the majority of the liquid in the slurry had been filtered through the unit (6 h, the composite was removed and allowed where shear tractions are forced to develop slightly away from matrix crack to dry in air overnight. The samples were then heated at 300oCfor faces, resulting in load transfer onto fiber 30 min in air to remo ss volatiles from the system. Samples containing the "C" fabric were subsequently sintered in vacuum (13.3 Pa), while the uncoated fabric samples sintered in air using a static vapor deposition(CVD) process In both cases, the samples were held at 1200C for 2 h fiber bridging naximum coating thickness. Al The sintered composites were sliced into straight-sided speci- coatings resulted in extensive mens for tensile testing using a slow-speed saw and a diamond f the fibers tows. The coated fibers were afering blade. Samples were -7.6 cm in length, with a width of sing scanning electron microscopy(SEM) 1. 27 cm. The thickness varied somewhat, with an average devia- The carbon-coated fiber tows were filament wound using a tion of -0.15 cm. Tapered epoxy tabs were placed on the calcium aluminosilicate(CAS, Corning, Inc, Corning, NY), water, specimens, producing a 2. 54 cm length gauge section. a portion of and binder(RobondtM, Rohm and Haas Co., Philadelphia, PA) the samples were tabbed after exposure at 1150C in air for either slurry. A small amount of surfactant(TritonTM X-100, LabChem 100or500h. retting of the carbon-coated tows. After drying, the tapes were mentioned test framp o Inc, Pittsburgh, PA)was added to the slurry to improve the Tensile testing of ites was completed on the afore- 4.45 kN load cell. The load- however, no strain gauges were hot-pressed at 1050.C for 15 min with an applied pressure of 13.8 used. The resultant comp MPa Composites containing uncoated Nextel 720 fiber were also fiber tow strengths, because composite properties were assumed to produced for comparison purposes, using the same procedures be fiber-dominated. The fracture surfaces of the samples were also The hot-pressed composites were diamond machined either into examined using optical microscopy and SEM dog-bone specimens(Bomas, Inc )or into straight-sided spec mens. The specimen shape(dog-bone versus straight sided) did not appear to affect test results. After machining, samples containing (3)Nextel 720 Tow Testing each carbon coating thickness and fiber orientation were expos Fiber tow strengths were measured for the Nextel 720 roving to air at 1000C for 500 h. Some samples with a 0.04 um carbon and for tows extracted from the fabric, both uncoated and"C"tows coating were also tested after oxidative exposure at 650C for 24 h were tested in each case. These tows were taped to 10.16 cm x in flowing O2. From this point, sample designations are"C for the 5.08 cm cardboard frames with 2.54 cm squares cut in the centers. carbon coating, e.g., Nextel 720/0.04 um"C /CAS Quick-setting epoxy was used to the sample ends, with Strain gauges (Model EA-06-125BZ-350, Micro- irtually no epoxy running into the gauge section. The epoxy cured Measurements, Inc, Raleigh, NC) were mounted on both sides of for >12 h The samples were loaded into the tensile test frame, and each specimen to measure true strain and to monitor possible the cardboard sides were cut prior to test initiation. A crosshead bending from misalignment. Some samples were tabbed with speed of 0.0127 cm/min and a 445N(100 lb)load cell were used epoxy, while others were tested using cardboard to line the for the tests. The maximum load was recorded and used in hydraulic wedge grip faces. Negligible slipping was observed in calculating tow strength, assuming an average cross-sectional area ceeded at a crosshead speed of 0. 127 for the fibers and a constant number of fibers. The average area for (Model SintechTM 20/G, MTS, Research the fibers has been shown to be a reasonable estimation, particu- a 4.45 kN (1000 lb)load cell. Load and larly as the number of tests increase. The results of the tow tests
using a static chemical vapor deposition (CVD) process, in which fiber bridging limited maximum coating thickness. Attempts at producing 0.1 mm thick coatings resulted in extensive cementing of the fibers within the tows. The coated fibers were examined using scanning electron microscopy (SEM). The carbon-coated fiber tows were filament wound using a calcium aluminosilicate (CAS, Corning, Inc., Corning, NY), water, and binder (Robond™, Rohm and Haas Co., Philadelphia, PA) slurry. A small amount of surfactant (Triton™ X-100, LabChem, Inc., Pittsburgh, PA) was added to the slurry to improve the wetting of the carbon-coated tows. After drying, the tapes were stacked in either a 0° or a 645° orientation and then vacuum hot-pressed at 1050°C for 15 min with an applied pressure of 13.8 MPa. Composites containing uncoated Nextel 720 fiber were also produced for comparison purposes, using the same procedures. The hot-pressed composites were diamond machined either into dog-bone specimens (Bomas, Inc.) or into straight-sided specimens. The specimen shape (dog-bone versus straight sided) did not appear to affect test results. After machining, samples containing each carbon coating thickness and fiber orientation were exposed to air at 1000°C for 500 h. Some samples with a 0.04 mm carbon coating were also tested after oxidative exposure at 650°C for 24 h in flowing O2. From this point, sample designations are “C” for the carbon coating, e.g., Nextel 720/0.04 mm “C”/CAS. Strain gauges (Model EA-06-125BZ-350, MicroMeasurements, Inc., Raleigh, NC) were mounted on both sides of each specimen to measure true strain and to monitor possible bending from misalignment. Some samples were tabbed with epoxy, while others were tested using cardboard to line the hydraulic wedge grip faces. Negligible slipping was observed in either case as testing proceeded at a crosshead speed of 0.127 cm/min on a test frame (Model Sintech™ 20/G, MTS, Research Triangle Park, NC) using a 4.45 kN (1000 lb) load cell. Load and crosshead displacement were recorded by a software program (TESTWORKS™, MTS), while load and strain measurements were plotted using an X–Y recorder. Tests were completed on both the 0° and 645° specimens. Elastic moduli and ultimate failure strengths of the composites were extracted from the measurements. The fracture surfaces were also examined using SEM. (2) Porous Matrix Composites Nextel 720 fabric (eight-harness satin fabric) was carbon-coated by Oak Ridge National Laboratory using a CVD process. The ;0.1 mm thick coating appeared smooth and continuous. There was some isolated bridging of the coating, particularly at crossover points in the weave. Control samples were also produced using uncoated, desized Nextel 720 fabric. Five circular layers of fabric were stacked, bonded at the edges using cellulose (Methocel™, Dow Chemical Co., Midland, MI), and subsequently hot molded to form a preform. This preform was then placed in a pressure infiltration vessel. An aqueous-based, nitric acid-stabilized slurry containing Al2O3 (AKP-53, Sumitomo Chemical Co., Ltd., Tokyo, Japan) and mullite (MUL-SM, Baikowski International, Charlotte, NC) was prepared. The slurry composition was chosen to yield an 85/15 by weight mixture of Al2O3 and SiO2, which matched the fiber composition. Both commercial powders were dispersed and sedimented prior to use, producing an average particle size of 0.15 mm for the Al2O3 and 0.38 mm for the mullite (Model LS230 Particle Size Analyzer, Coulter Co., Miami, FL). The slurry was then poured into the filtration vessel, where matrix infiltration was assisted by applying air pressure (;0.55 MPa) to one side and vacuum (;0.03 MPa) to the other. As reported for other pressure-infiltrated ceramics,30,31 infiltration followed parabolic rate kinetics, indicating that the cake thickness grew in proportion to the square root of time. After the majority of the liquid in the slurry had been filtered through the unit (;6 h), the composite was removed and allowed to dry in air overnight. The samples were then heated at 300°C for 30 min in air to remove excess volatiles from the system. Samples containing the “C” fabric were subsequently sintered in vacuum (13.3 Pa), while the uncoated fabric samples were sintered in air. In both cases, the samples were held at 1200°C for 2 h. The sintered composites were sliced into straight-sided specimens for tensile testing using a slow-speed saw and a diamond wafering blade. Samples were ;7.6 cm in length, with a width of 1.27 cm. The thickness varied somewhat, with an average deviation of ;0.15 cm. Tapered epoxy tabs were placed on the specimens, producing a 2.54 cm length gauge section. A portion of the samples were tabbed after exposure at 1150°C in air for either 100 or 500 h. Tensile testing of the composites was completed on the aforementioned test frame using a 4.45 kN load cell. The load– displacement curve was recorded; however, no strain gauges were used. The resultant composite strengths were correlated with the fiber tow strengths, because composite properties were assumed to be fiber-dominated. The fracture surfaces of the samples were also examined using optical microscopy and SEM. (3) Nextel 720 Tow Testing Fiber tow strengths were measured for the Nextel 720 roving and for tows extracted from the fabric; both uncoated and “C” tows were tested in each case. These tows were taped to 10.16 cm 3 5.08 cm cardboard frames with 2.54 cm squares cut in the centers. Quick-setting epoxy was used to secure the sample ends, with virtually no epoxy running into the gauge section. The epoxy cured for $12 h. The samples were loaded into the tensile test frame, and the cardboard sides were cut prior to test initiation. A crosshead speed of 0.0127 cm/min and a 445 N (100 lb) load cell were used for the tests. The maximum load was recorded and used in calculating tow strength, assuming an average cross-sectional area for the fibers and a constant number of fibers. The average area for the fibers has been shown to be a reasonable estimation, particularly as the number of tests increase.32 The results of the tow tests Fig. 1. Concept of load transfer is illustrated for a composite using a fugitive coating for fiber protection. (a) An idealized case, where the gap is straight, the fiber is protected from a matrix crack, but there is no load transfer on to the fiber. (b) In reality, interfacial roughness presents a case where shear tractions are forced to develop slightly away from matrix crack faces, resulting in load transfer onto fiber. 330 Journal of the American Ceramic Society—Keller et al. Vol. 83, No. 2
February 2000 ugitive Interfacial Carbon Coatings for Oxide/Oxide Composites were used in evaluating the strengths of both dense and porous 200+ matrIx composites. 醞As- processed 1000°C/500h/ 150 IlL. Results (I Dense Matrix Composites (A) Microstructure: The CVD"C 100 smooth and continuous There was fiber contact during the coating ocess, which was unavoidable because of the static position of the fibers during deposition. This contact led to significant areas of exposed fiber surface because of spalling of the bridged coating 50 Such exposure could be detrimental if the fiber and matrix wer reactive and/or if exposure was excessive, leading to substantial Iber/matrix sintering. The "C thicknesses were not directly measured. There were some variations in the coating thickness Calculated 0.02 micron 0.04 micron within the tow: however this was not considered in the current work. Because of possible coating variations, any demonstrated property dependence on carbon thickness should be considered qualitative CAs glass-ceramic (anorthite) melted at >1400C, but the 250 hot-pressing temperature was kept below this to minimize strength degradation of the fibers, which was significant at >1300C. Some As-Processed pertinent physical properties of the Nextel 720 fiber and CAS △200 N Oxidized 650 C/24 h/Oxygen matrix are given in Table L 端1000°c/500h/air Microstructural evaluation of the hot-pressed samples revealed that the fibers were relatively well dispersed, with 30-35 vol% 150 fibers. The bulk porosity, arising primarily from incomplete binder burnout, was determined to be -5%-10% using the archimedes technique. The composites were black in the as-processed condi- tion, and a small amount of glassy phase appeared optically to be left in the matrix. The appearance of the Nextel 720/0.04 um C"/CAS sample changed to white after oxidative heat treatment at 650C, indicating that the carbon had burned away. The oxidation onditions were chosen based on the model of Cawley et al. After both the 650 c and the long-term heat treatment, the Uncoated 0.02 micron 0.04 micron microstructures of the composites remained essentially unchanged, aside from the removal of the carbon. No large pores were seen in (b) the composites, indicating that CO, evolution and removal did not Fig. 2.(a)Measured elastic modulus ause significant damage, as might be expected from the pressure CAS composites(>90% dense)ar No significant difference can be seen m the0° Nextel720~ ith the calculated value various conditions.(b) The gap formed by the oxidation of"C was too thin to be Ultimate strengths of composites Uncoated specimens exhibit extremely imaged in the SEM; further imaging was impaired by the rounding poor strengths of the fiber edges caused by the different polishing rates of th constituents. a dark ring was seen around the fiber but it was not ear whether it was a polishing artifact or an actual gap different conditions studied. The first column shows the modulu Transmission electron microscopy redicted through rule-of-mixtures calculations (see Section usable for the oxidized samples because of difficulties associated iV(1A). Comparing the control composites, viz., 0.02 and 0.04 with sample preparation. Indications of gap stability in this work were therefore evaluated through mechanical testing of the com- m"C "specimens, with the composites of interest, viz., the same composites tested after a 1000C, 500 h heat treatment in air, posites and the resultant fiber pullout observed on the fracture shows that there is not a significant difference in the modulus surfaces values. This indicates that the presence of"C its thickness, or its (B) Mechanical Properties: The test results of the 0%(uni- removal has little effect on the measured modulus directional) composites are shown in Fig. 2. In Fig. 2(a), the elastic In Fig. 2(b), the ultimate failure strengths of the composites moduli calculated from the stress-strain plots are shown for the under different conditions are compared. The uncoated composites clearly have very poor strength; in fact, many specimens failed before testing. The "C" composites exhibit much better strengths Table I. Select Physical Properties of CAs and Nextel 720 the 0.04 um C composites are seen to have slightly better composite strengths than the 0.02 um" composites, particularly operty CAS glass-ceramic Nextel 720 if the tow strengths are taken into consideration. The strengths of Cao-Al,Ox2Si02 85 wt% Al the composites of interest, where the "C was burned away, are also shown in Fig. 2(b). These composites were tested after the Density C -removal heat treatment(650C, 24 h, O2) and after a long Elastic modulus(GP term(1000 C, 500 h, air) heat treatment. It can be seen that oefficient of composite strength decreases slightly with carbon removal; how- ever, the strengths after"C" removal of the 0.04 um composites Flexure strengt 124 Tensile strength(GPa) are comparable to those of the 0.02 um"C" composites Fabric strength(MPa) In Fig. 3, the results of tests on the t45composites are show (fill), 3 in. gauge(Ib/in) Fig 3(a), the
were used in evaluating the strengths of both dense and porous matrix composites. III. Results (1) Dense Matrix Composites (A) Microstructure: The CVD “C” appeared generally smooth and continuous. There was fiber contact during the coating process, which was unavoidable because of the static position of the fibers during deposition. This contact led to significant areas of exposed fiber surface because of spalling of the bridged coating. Such exposure could be detrimental if the fiber and matrix were reactive and/or if exposure was excessive, leading to substantial fiber/matrix sintering. The “C” thicknesses were not directly measured. There were some variations in the coating thickness within the tow; however, this was not considered in the current work. Because of possible coating variations, any demonstrated property dependence on carbon thickness should be considered qualitative. CAS glass-ceramic (anorthite) melted at .1400°C, but the hot-pressing temperature was kept below this to minimize strength degradation of the fibers, which was significant at .1300°C. Some pertinent physical properties of the Nextel 720 fiber and CAS matrix are given in Table I. Microstructural evaluation of the hot-pressed samples revealed that the fibers were relatively well dispersed, with 30–35 vol% fibers. The bulk porosity, arising primarily from incomplete binder burnout, was determined to be ;5%–10% using the Archimedes technique. The composites were black in the as-processed condition, and a small amount of glassy phase appeared optically to be left in the matrix. The appearance of the Nextel 720/0.04 mm “C”/CAS sample changed to white after oxidative heat treatment at 650°C, indicating that the carbon had burned away. The oxidation conditions were chosen based on the model of Cawley et al.33 After both the 650°C and the long-term heat treatment, the microstructures of the composites remained essentially unchanged, aside from the removal of the carbon. No large pores were seen in the composites, indicating that COx evolution and removal did not cause significant damage, as might be expected from the pressure of an entrapped gas. The gap formed by the oxidation of “C” was too thin to be imaged in the SEM; further imaging was impaired by the rounding of the fiber edges caused by the different polishing rates of the constituents. A dark ring was seen around the fiber, but it was not clear whether it was a polishing artifact or an actual gap. Transmission electron microscopy (TEM) was also not readily usable for the oxidized samples because of difficulties associated with sample preparation. Indications of gap stability in this work were therefore evaluated through mechanical testing of the composites and the resultant fiber pullout observed on the fracture surfaces. (B) Mechanical Properties: The test results of the 0° (unidirectional) composites are shown in Fig. 2. In Fig. 2(a), the elastic moduli calculated from the stress–strain plots are shown for the different conditions studied. The first column shows the modulus predicted through rule-of-mixtures calculations (see Section IV(1A)). Comparing the control composites, viz., 0.02 and 0.04 mm “C” specimens, with the composites of interest, viz., the same composites tested after a 1000°C, 500 h heat treatment in air, shows that there is not a significant difference in the modulus values. This indicates that the presence of “C”, its thickness, or its removal has little effect on the measured modulus. In Fig. 2(b), the ultimate failure strengths of the composites under different conditions are compared. The uncoated composites clearly have very poor strength; in fact, many specimens failed before testing. The “C” composites exhibit much better strengths; the 0.04 mm “C” composites are seen to have slightly better composite strengths than the 0.02 mm “C” composites, particularly if the tow strengths are taken into consideration. The strengths of the composites of interest, where the “C” was burned away, are also shown in Fig. 2(b). These composites were tested after the “C”-removal heat treatment (650°C, 24 h, O2) and after a longterm (1000°C, 500 h, air) heat treatment. It can be seen that composite strength decreases slightly with carbon removal; however, the strengths after “C” removal of the 0.04 mm composites are comparable to those of the 0.02 mm “C” composites. In Fig. 3, the results of tests on the 645° composites are shown. In Fig. 3(a), the measured moduli are compared for the control and the fugitive “C” composites. A decrease in modulus occurs with Table I. Select Physical Properties of CAS and Nextel 720 Property CAS glass-ceramic Nextel 720 Composition CaO–Al2O3–2SiO2 85 wt% Al2O3, 15 wt% SiO2 Density 2.8 3.4 Elastic modulus (GPa) 98 262 Coefficient of thermal expansion (31026 /°C) 5 6 Flexure strength (MPa) 124 Tensile strength (GPa) 2.1 Fabric strength (MPa) (fill), 3 in. gauge (lb/in.) 230 Fig. 2. (a) Measured elastic modulus values for the 0° Nextel 720/“C”/ CAS composites (.90% dense) are compared with the calculated value. No significant difference can be seen between various conditions. (b) Ultimate strengths of composites. Uncoated specimens exhibit extremely poor strengths. February 2000 Fugitive Interfacial Carbon Coatings for Oxide/Oxide Composites 331
VoL. 83. No. 2 200 a 1000°C/500hair 150 Uncoated 0.02 micron 0.04 micron 100um 150 N1000°C/500hair 100 0.04 micron us values for off-axis Nextel 720/C/CAS asing carbon thickness. indicate that"C and fugitive samples are superior to creasing "C thickness, and the modulus drops even further as the"C is removed This indicates that the off-axis load transfer is not as efficient as in the 0.orientation. In Fig 3(b), the ultimate strengths in the off-axis are compared. The fugitive"C"compos- tes appear to be nearly as good as the"C composites and 100m ignificantly better than the uncoated composites Figures 4 and 5 show fracture surfaces of some typical control Fig. 4. SEM of Nextel 720/CAS unidirectional (0%) composite fracture and fugitive composites(uncoated, 0.02 um"C, and 0.04 um C") tested in the o°and±45° orientations. Generally,the uncoated composites exhibit brittle fracture surfaces with virtually matrix, and 30 vol% porosity(ASTM C20-922). There is some no fiber pullout in both orientations(Figs. 4(a) and 5(a)). For the variability(<5 vol%)in fiber volume percentage from sample to "C and fugitive composites, the fracture surfaces show significant sample, which contributes to scatter in the tensile test data. fiber pullout. The extent of fiber pullout follows the same trend as the tensile strength values. The incorporation of the 0.02 and 0.04 (B) Mechanical Properties: The results of the tensile tests, own in Fig. 7, revealed that, for the porous matrix composites um"C results in a major increase in fiber pull-out lengths(Figs the as-processed strengths of the samples containing uncoated 4(b) and 5(b)). After the 1000@C heat treatment that removes the abric were superior to samples containing"C" fabric. Subsequent arbon, the pull-out lengths diminish noticeably(Figs. 4(c)and tow testing of the Nextel 720"C" fabric revealed that there was a trength loss of >3(Table D), which could be attributed to fiber uncoated sample degradation during the coating process. To truly evaluate the (2) Porous Matrix Composites omposite strengths, therefore, it was necessary to normalize these (A) Microstructure: Microstructural analysis of the porous trengths with respect to the measured tow properties prior to matrix composites reveals areas of fine, dispersed porosity in the natrix. Aside from these pores, however, there are also large shrinkage cracks in the matrix(Fig. 6), as seen in other works Standard Test Methods for Apparent Porosity completed on pressure-infiltrated composites. The density of the ater, ASTM Designation C20-92. American So 金 West Conshohocken, P
increasing “C” thickness, and the modulus drops even further as the “C” is removed. This indicates that the off-axis load transfer is not as efficient as in the 0° orientation. In Fig. 3(b), the ultimate strengths in the off-axis are compared. The fugitive “C” composites appear to be nearly as good as the “C” composites and significantly better than the uncoated composites. Figures 4 and 5 show fracture surfaces of some typical control and fugitive composites (uncoated, 0.02 mm “C”, and 0.04 mm “C”) tested in the 0° and 645° orientations. Generally, the uncoated composites exhibit brittle fracture surfaces with virtually no fiber pullout in both orientations (Figs. 4(a) and 5(a)). For the “C” and fugitive composites, the fracture surfaces show significant fiber pullout. The extent of fiber pullout follows the same trend as the tensile strength values. The incorporation of the 0.02 and 0.04 mm “C” results in a major increase in fiber pull-out lengths (Figs. 4(b) and 5(b)). After the 1000°C heat treatment that removes the carbon, the pull-out lengths diminish noticeably (Figs. 4(c) and 5(c)) but are still significant when compared with those of the uncoated sample. (2) Porous Matrix Composites (A) Microstructure: Microstructural analysis of the porous matrix composites reveals areas of fine, dispersed porosity in the matrix. Aside from these pores, however, there are also large shrinkage cracks in the matrix (Fig. 6), as seen in other works completed on pressure-infiltrated composites.29 The density of the composites is typically 2.2 g/cm3 , with ;30 vol% fibers, 40 vol% matrix, and 30 vol% porosity (ASTM C20-92‡ ). There is some variability (,5 vol%) in fiber volume percentage from sample to sample, which contributes to scatter in the tensile test data. (B) Mechanical Properties: The results of the tensile tests, shown in Fig. 7, revealed that, for the porous matrix composites, the as-processed strengths of the samples containing uncoated fabric were superior to samples containing “C” fabric. Subsequent tow testing of the Nextel 720 “C” fabric revealed that there was a strength loss of .1⁄3 (Table II), which could be attributed to fiber degradation during the coating process. To truly evaluate the composite strengths, therefore, it was necessary to normalize these strengths with respect to the measured tow properties prior to ‡ “Standard Test Methods for Apparent Porosity, Water Absorption, Apparent Specific Gravity, and Bulk Density of Burned Refractory Brick and Shapes by Boiling Water,” ASTM Designation C20-92. American Society for Testing and Materials, West Conshohocken, PA. Fig. 3. (a) Modulus values for off-axis (645°) Nextel 720/“C”/CAS samples. There is a decrease in modulus with increasing carbon thickness. (b) Off-axis strengths indicate that “C” and fugitive samples are superior to uncoated composites. Fig. 4. SEM of Nextel 720/CAS unidirectional (0°) composite fracture surfaces: (a) uncoated, (b) 0.02 mm “C”, (c) fugitive (0.02 mm “C” initial). 332 Journal of the American Ceramic Society—Keller et al. Vol. 83, No. 2
333 0.5 mml Fig. 6. Optical micrograph showing a typical microstructure of Nextel 720 porous matrix composites, There is distributed porosity in the matrix, along with large shrinkage cracks d150 100 Fig. 7. Tensile test results for Nextel 720/mullite/alumina porous matrix mposites. Samples were tested in the as-processed condition and after heat treatments at 1150 C for 100 and 500 h. Overall strengths of Ten posites with uncoated fabric were superior to those with"C"fabric ile testing of fiber tows revealed an y loss in strength in the "C fabric. (See Fig. 8 for normalized composite strength data. roving were coated by indep sources, it appeared the strength loss was associated ertain deposition conditions. The degradation mechanism is the scope of this paper and is not discussed further 1 mm IV. Discussion Fig. 5. SEM of Nextel 720/CAS off-axis(+45)composite fracture surfaces: (a)uncoated, (b)0.04 um"C",(c) fugitive(0.04 um "C" initial) In the case of dense CAs matrix composites, the measured modulus, ultimate fracture strength, ugitive"C" composites, when compared with those of the control significant difference can be seen between the uncoated fibers, o composites(uncoated and""composites), showed that the ugitive coating concept is a mechanically via able conce pt fibers, and fugitive composites Nextel 720-reinforced(>90% dense)CAs matrix composites In all cases, the composites exhibited nonl inearity prior to failure These composites were also found to be quite stable at 1000.C for The long-term heat treatment of the porous matrix composites did not <500 h. For the porous matrix(85/15 Al2O SiO2, -30% overal appear to affect the composite strength. In fact, the average strength porosity, -43% matrix porosity) composites, there was no sign was slightly higher than in the as-processed case. The porous matrix icant difference between the uncoated, "C, and fugitive compos- composites also exhibited bundle pullout in all cases. A typical ites. The following is a more detailed discussion of the results to fracture surface is shown in Fig 9. The presence or absence of"C" did ain further insights from the data. not affect the appearance of the composite fracture surface, and little difference was seen after the long-term heat treatment (I Dense Matrix Composites (3) Nextel 720 Tow Testing (A Modulus: The modulus of a unidirectional composite is Tow test results are incorporated in Table Il. These results well-known to obey the simple rule of mixtures, viz indicate that the Nextel 720 as-received roving and fabric dis- pl strengths to the roving coated with 0.02 um carbon (~750 lowever, both the 0.04 um"C roving and"C Using the values listed in Table l, for a fiber volume percentage of fabric strength losses of >A. Because the fabric and 32.5+2.5, we obtain a composite modulus of 151.3 4 GPa. The
comparison. The normalized data are shown in Fig. 8. No significant difference can be seen between the uncoated fibers, “C” fibers, and fugitive composites. In all cases, the composites exhibited nonlinearity prior to failure. The long-term heat treatment of the porous matrix composites did not appear to affect the composite strength. In fact, the average strength was slightly higher than in the as-processed case. The porous matrix composites also exhibited bundle pullout in all cases. A typical fracture surface is shown in Fig. 9. The presence or absence of “C” did not affect the appearance of the composite fracture surface, and little difference was seen after the long-term heat treatment. (3) Nextel 720 Tow Testing Tow test results are incorporated in Table II. These results indicate that the Nextel 720 as-received roving and fabric displayed similar strengths to the roving coated with 0.02 mm carbon (;750 MPa). However, both the 0.04 mm “C” roving and “C” fabric exhibited strength losses of .1⁄3. Because the fabric and roving were coated by independent sources, it appeared the strength loss was associated with certain deposition conditions. The degradation mechanism is beyond the scope of this paper and is not discussed further. IV. Discussion In the case of dense CAS matrix composites, the measured modulus, ultimate fracture strength, and fractography of the fugitive “C” composites, when compared with those of the control composites (uncoated and “C” composites), showed that the fugitive coating concept is a mechanically viable concept for the Nextel 720-reinforced (.90% dense) CAS matrix composites. These composites were also found to be quite stable at 1000°C for #500 h. For the porous matrix (85/15 Al2O3/SiO2, ;30% overall porosity, ;43% matrix porosity) composites, there was no significant difference between the uncoated, “C,” and fugitive composites. The following is a more detailed discussion of the results to gain further insights from the data. (1) Dense Matrix Composites (A) Modulus: The modulus of a unidirectional composite is well-known to obey the simple rule of mixtures,34 viz., Ec 5 EfVf 1 EmVm (1) Using the values listed in Table I, for a fiber volume percentage of 32.5 6 2.5, we obtain a composite modulus of 151.3 6 4 GPa. The Fig. 5. SEM of Nextel 720/CAS off-axis (645°) composite fracture surfaces: (a) uncoated, (b) 0.04 mm “C”, (c) fugitive (0.04 mm “C” initial). Fig. 6. Optical micrograph showing a typical microstructure of Nextel 720 porous matrix composites. There is distributed porosity in the matrix, along with large shrinkage cracks. Fig. 7. Tensile test results for Nextel 720/mullite/alumina porous matrix composites. Samples were tested in the as-processed condition and after heat treatments at 1150°C for 100 and 500 h. Overall strengths of composites with uncoated fabric were superior to those with “C” fabric. Tensile testing of fiber tows revealed an ; 1⁄3 loss in strength in the “C” fabric. (See Fig. 8 for normalized composite strength data.) February 2000 Fugitive Interfacial Carbon Coatings for Oxide/Oxide Composites 333