./. Appl. Ceram. Technol, 7/3/276-290(2010) DO:10IJI174-7402.200902422x International Journal o pplied Ceramic TECHNOLOGY ceramic Product D Effects of Fiber Architecture on Matrix Cracking for Melt-Infiltrated SiC/SiC Composites Ohio Aerospace Institute, 22800 Cedar Point Road, Cleveland, Ohio 44142 James A DiCarlo and James D. Kiser NASA Glenn Research Center, 21000 Brookpark Road, Cleveland, Ohio 44135 Hee Mann Yun Matech GSM, 31304 Via Colinas, Suite 102, Westlake Village, California 91362 The matrix cracking behavior of slurry cast melt-infiltrated SiC matrix composites consisting of Sylramic-iBN fibers with wide variety of fiber architectures were compared. The fiber architectures included 2D woven, braided, 3D orthogonal, and angle interlock architectures. Acoustic emission was used to monitor in-plane matrix cracking during unload-reload tensile tests. Two key parameters were found to control matrix-cracking behavior: the fiber volume fraction in the loading direction and the area of the weakest portion of the structure, that is, the largest tow in the architecture perpendicular to the loading direction. Empirical models that support these results are presented and discussed. Introduction originally started as a NASA Glenn IRD Project and was continued Silicon carbide fiber-reinforced silicon carbide ce- under NASA's ARMD Supersonics program. ramic matrix composites (SiC/SiC CMC)are actively be ing pursued for high-temperature structural applications No daim to U.S. Government works. gine combustor liners, turbine components
Effects of Fiber Architecture on Matrix Cracking for Melt-Infiltrated SiC/SiC Composites Gregory N. Morscher* Ohio Aerospace Institute, 22800 Cedar Point Road, Cleveland, Ohio 44142 James A. DiCarlo and James D. Kiser NASA Glenn Research Center, 21000 Brookpark Road, Cleveland, Ohio 44135 Hee Mann Yun Matech GSM, 31304 Via Colinas, Suite 102, Westlake Village, California 91362 The matrix cracking behavior of slurry cast melt-infiltrated SiC matrix composites consisting of Sylramic-iBN fibers with a wide variety of fiber architectures were compared. The fiber architectures included 2D woven, braided, 3D orthogonal, and angle interlock architectures. Acoustic emission was used to monitor in-plane matrix cracking during unload–reload tensile tests. Two key parameters were found to control matrix-cracking behavior: the fiber volume fraction in the loading direction and the area of the weakest portion of the structure, that is, the largest tow in the architecture perpendicular to the loading direction. Empirical models that support these results are presented and discussed. Introduction Silicon carbide fiber-reinforced silicon carbide ceramic matrix composites (SiC/SiC CMC) are actively being pursued for high-temperature structural applications such as engine combustor liners, turbine components, Int. J. Appl. Ceram. Technol., 7 [3] 276–290 (2010) DOI:10.1111/j.1744-7402.2009.02422.x Ceramic Product Development and Commercialization Funding for this work originally started as a NASA Glenn IRD project and was continued under NASA’s ARMD Supersonics program. *gregory.n.morscher@nasa.gov Journal compilation r 2009 The American Ceramic Society No claim to U.S. Government works
Efects of Fiber Architecture on SiC/SiC Composites 277 and exhaust nozzles. 2 These applications will require by the high-performance Sylramic-iBN SiC fiber cur- fiber architectures that can not only provide the com- rently represents the state-of-the-art in high-temperature ponent shape forming capability properties, but also the SiC/SiC composites because of its high-thermal and optimum in tensile strength, creep-rupture properties, structural performance at use temperatures beyond those and thermal conductivity in multiple directions. of current metallic alloys. The key in-plane mechanical Although conventional 2D woven lay-up architectures properties of interest are the elastic modulus, the onse offer a good degree of shape capability, they currently stress at which through-thickness matrix cracks forn ult in poorer CMC performance due to such issues as and the ultimate strength of as-fabricated panels at room reduced in-plane strength related to stress risers at ends temperature. Particular focus for this study was the onset low through-the-thickness tens matrIX cracking stress because it is above this stress that strength, shear strength, and thermal conductivity re- life-limiting environmental degradation of SiC compos- lated to the CMC need for weak fiber-matrix interfaces ites occurs at elevated(>600oC)temperature for matrix crack defection For this reason. more com Other studies on matrix cracking in 3D plex 3D weaves and braids with through-the-thickness SiC/SiC composites have focused on the chemically va- fiber reinforcement are currently being pursued. por infiltrated(CVI)matrix system" or the polymer-in At the current stage of development for SiC/Sic filtrated pyrolysis(PIP)-derived matrix system, both of CMC, it is thus important to develop an understanding those studies with low-modulus SiC-based fiber rein- of the fber-architecture effects on key properties in or- forcement. This study is distinguished in that high-mod der to better design the architectures both for shape and ulus polycrystalline SiC fibers were the primary fiber ent performance. Ty ners and reinforcement, a wider of fiber architectures were modelers will initially need fber-architecture compared including unbalanced fiber proportions in dif nability and associated composite processing models ferent directions, the composites were all of the MI va- in order to decide on what architectures can provide all riety, and the emphasis is on use and design based on the special shape features of the component. To matrix cracking stress. Unbalanced architectures were mize machining issues, such as added cost and fiber chosen in order to maximize the fiber content in one di- damage, it is generally desirable that SiC/SiC compo- rection, which is expected to be potentially desirable for nent be fabricated with near-net shape. With possible applications which require higher load-carrying ability in fiber architecture types in mind, designers would then one direction. In a future paper, it will be demonstrated need constituent properties and composite propert how improvements in the fiber architecture increased models in order to down-select which architecture stress capability at high temperatures in oxidizing envi- type would allow the component to best meet its struc ronments(G.N. morsche tural performance requirements. For high-temperature applications and thin-walled near net shaped compo- nents,a key CMC material design goal is to typically Experimental Procedure seek tensile strength and thermal conductivity through he component walls that are as high as possible in order The four SiC/SiC composite panels fabricated to withstand high thermal gradients. 3D woven and for this study were all processed with the slurry cast braided architectures offer enhanced capability in this Power Systems Composites, New regard by enabling directional tailoring of properties, ark, DE)0. CVI BN fiber coating, followed by CVI both in-plane and out-of-plane, using fibers of hi Sic infiltration (a few micrometers in thickness), SiC strength and high conductivity. However, adding fibers particulate slurry infiltration, and molten Si infiltra out-of-plane can sometimes degrade in-plane properties tion. These panels were tailored to be unbalanced where high structural performance is most needed with a higher fber volume fraction in one in-plane With these modeling needs in view, this study has direction than in the other orthogonal in-plane direc- sought to vary the fiber architecture and volume fraction tion. As-fabricated panel dimensions of the four pan- in five SiC/SiC slurry-cast melt-infiltrated(MI) panels in els were approximately 150 mm x 80 mm x 2mm order to measure the effects of multidirectional architec- Detailed descriptions of the as-produced architectures tures and fiber content on key in-plane mechanical prop- for these panels are shown in the top portion of erties. This ceramic composite system when reinforced Table I. In all cases, the in-plane X and/or Y fibers
and exhaust nozzles.1,2 These applications will require fiber architectures that can not only provide the component shape forming capability properties, but also the optimum in tensile strength, creep–rupture properties, and thermal conductivity in multiple directions. Although conventional 2D woven lay-up architectures offer a good degree of shape capability, they currently result in poorer CMC performance due to such issues as reduced in-plane strength related to stress risers at ends of plies,3 and to low through-the-thickness tensile strength, shear strength, and thermal conductivity related to the CMC need for weak fiber–matrix interfaces for matrix crack deflection. For this reason, more complex 3D weaves and braids with through-the-thickness fiber reinforcement are currently being pursued.4–6 At the current stage of development for SiC/SiC CMC, it is thus important to develop an understanding of the fiber-architecture effects on key properties in order to better design the architectures both for shape and final component performance. Typically, designers and process modelers will initially need fiber-architecture formability and associated composite processing models in order to decide on what architectures can provide all the special shape features of the component. To minimize machining issues, such as added cost and fiber damage, it is generally desirable that SiC/SiC component be fabricated with near-net shape. With possible fiber architecture types in mind, designers would then need constituent properties and composite property models in order to down-select which architecture type would allow the component to best meet its structural performance requirements. For high-temperature applications and thin-walled near net shaped components, a key CMC material design goal is to typically seek tensile strength and thermal conductivity through the component walls that are as high as possible in order to withstand high thermal gradients. 3D woven and braided architectures offer enhanced capability in this regard by enabling directional tailoring of properties, both in-plane and out-of-plane, using fibers of high strength and high conductivity. However, adding fibers out-of-plane can sometimes degrade in-plane properties where high structural performance is most needed. With these modeling needs in view, this study has sought to vary the fiber architecture and volume fraction in five SiC/SiC slurry-cast melt-infiltrated (MI) panels in order to measure the effects of multidirectional architectures and fiber content on key in-plane mechanical properties. This ceramic composite system when reinforced by the high-performance Sylramic-iBN SiC fiber currently represents the state-of-the-art in high-temperature SiC/SiC composites because of its high-thermal and structural performance at use temperatures beyond those of current metallic alloys. The key in-plane mechanical properties of interest are the elastic modulus, the onset stress at which through-thickness matrix cracks form, and the ultimate strength of as-fabricated panels at room temperature. Particular focus for this study was the onset matrix cracking stress because it is above this stress that life-limiting environmental degradation of SiC composites occurs at elevated (46001C) temperatures.7 Other studies on matrix cracking in 3D architecture SiC/SiC composites have focused on the chemically vapor infiltrated (CVI) matrix system8 or the polymer-in- filtrated pyrolysis (PIP)-derived matrix system,9 both of those studies with low-modulus SiC-based fiber reinforcement. This study is distinguished in that high-modulus polycrystalline SiC fibers were the primary fiber reinforcement, a wider variety of fiber architectures were compared including unbalanced fiber proportions in different directions, the composites were all of the MI variety, and the emphasis is on use and design based on matrix cracking stress. Unbalanced architectures were chosen in order to maximize the fiber content in one direction, which is expected to be potentially desirable for applications which require higher load-carrying ability in one direction. In a future paper, it will be demonstrated how improvements in the fiber architecture increased stress capability at high temperatures in oxidizing environments (G. N. Morscher, unpublished data). Experimental Procedure The four SiC/SiC composite panels fabricated for this study were all processed with the slurry cast MI technique (GE Power Systems Composites, Newark, DE)10,11: CVI BN fiber coating, followed by CVI SiC infiltration (a few micrometers in thickness), SiC particulate slurry infiltration, and molten Si infiltration. These panels were tailored to be unbalanced with a higher fiber volume fraction in one in-plane direction than in the other orthogonal in-plane direction. As-fabricated panel dimensions of the four panels were approximately 150 mm 80 mm 2 mm. Detailed descriptions of the as-produced architectures for these panels are shown in the top portion of Table I. In all cases, the in-plane X and/or Y fibers www.ceramics.org/ACT Effects of Fiber Architecture on SiC/SiC Composites 277
enational Journal of Applied Ceramic Technology Vol.7,No.3,2010 C -F-③ 8目 云8面8面面 。 E 曾 §g 手 988
Table I. Description of Different Architectures for MI SiC/SiC Panels Architecture description No. plies, test direction Tow epcm, test direction Panel thickness (mm) Fiber type (no. of combined tows), volume fraction and predominate angle to test direction X (warp) Y (fill) Z (stuffer) SiC/SiC panel type (reference) AI-UNI 1D angle interlock through-the-thickness, Y-aligned, unbalanced 6.5 11 2.0 ZMI and rayon o0.05, 901 Sylramic-iBN (1) fo 5 0.23, 01 3DO-Un-R 3D orthogonal, unbalanced Y 5 test direction 7 9.8 1.53 Sylramic-iBN (1) B0.10, 901 Sylramic-iBN (1) fo 5 0.28, 01 Rayon o0.05, 901 3DO-Un-Z 3D orthogonal, unbalanced Y 5 test direction 7 9.8 1.58 Sylramic-iBN (1) B0.10, 901 Sylramic-iBN (1) fo 5 0.27, 01 ZMI o0.05, 901 LTLAI 3D layer-to-layer angle interlock 3 5.5 0.96 Sylramic-iBN (1) B0.10, 901 Sylramic-iBN (1) fo 5 0.10, 01 MI SiC/SiC panels from previous studies 2D 5HS10,14 2D five-harness satin balanced 0/90 lay-up 4–8 4.9–8.7 1.5–2.2 Sylramic-iBN (1) fo 5 0.12–0.2, 01 Sylramic-iBN (1) 0.12–0.2, 901 2D 5HS DT10 2D five-harness satin (double tow), balanced 0/90 lay-up 8 3.9 2.1 Sylramic-iBN (2) fo 5 0.19, 01 Sylramic-iBN (2) 0.19, 901 Braid15 2D triaxial braid (double tow); cut/tooled into panel, X 5 axial, Y 5 hoop 5 test direction 4 4.7 1.8 Sylramic-iBN (2) B0.1, 901 Sylramic-iBN (2) fo 5 0.26, 7231 3DO-Bal-R-Y6 3D orthogonal, nearly balanced; Y 5 test direction 7 7.9 1.95 Sylramic (2) B0.18, 901 Sylramic (1) fo 5 0.20, 01 rayon o0.05, 901 3DO-Bal-Z-Y6 3D orthogonal, nearly balanced; Y 5 test direction 7 7.1 2.05 Sylramic (2) B0.18, 901 Sylramic (1) fo 5 0.17, 01 ZMI o0.05, 901 3DO-Bal-Z-X6 3D orthogonal, nearly balanced; X 5 test direction 8 3.9 2 Sylramic (2) fo 5 0.18, 01 Sylramic (1) 0.17, 901 ZMI o0.05, B901 ‘‘Double Tow’’ refers to two tows woven or braided together which means Nfibers/tow 5 1600; emboldened fo values refer to test direction. MI, melt-infiltrated. 278 International Journal of Applied Ceramic Technology—Morscher, et al. Vol. 7, No. 3, 2010
wwceramics. org/ACT Efects of Fiber Architecture on SiC/SiC Composites were the high-performance Sylramic-iBN SiC fibers, (0.5 um thick) and MI SiC matrix. For the prey which display high strength, high creep-rupture ous 3D orthogonal panels where rayon fibers were resistance,and high thermal conductivity. These fi- used, the preferred NASA conversion process for bers were originally produced as Sylramic SiC fibers in Sylramic-iBN preforms was not used because of initial 800 filament tow form by Dow Corning, Midland, MI. concern related to degradation of the architectures The tows were then heat treated at NASA either as in- Thus the in-plane fibers for these panels were only dividual tows or preferably within a textile-formed pre- Sylramic fibers which, other than a significant differ- form in a nitrogen-containing environment to convert ence in creep and rupture behavior, show essentially the Sylramic fibers into Sylramic-ibN fibers the same elastic and tensile strength behavior as the uch as the Ube Sylramic-iBN fibers when incorporated into MI SiC/ (Yamaguchi, Japan)"Tyranno ZMI"SiC and/or rayon SiC composites fiber tows were incorporated into the initial The densities of all of the par Imens architectures because of their low-modulus and better evaluated in this study were >2.85 g/cm. This implies ormability, for example, as the warp-weaver fibers. For that most of the porosity is resident within the tow those architectures that used the polymer-based rayon minicomposites(5% total porosity) typical of good fiber, significant degradation was observed in these MI infiltration as observed in Fig. I fibers due to its fugitive nature at the high fiber It should also be noted in Table i that in contrast to coating and matrix processing temperatures with only the other panels, the in-plane Sylramic-iBN fibers of the a remnant char remaining that was subsequently coated Braid panel were not oriented in the testing direction. with BN and SiC during matrix processing. The in situ For this panel, a 2D tri-axial braided architecture was architectures for the four SiC/SiC panels fabricated in first formed on a 50 mm diameter mandrel with the this study can be seen in Fig. 1 which shows tensile hoop fibers +67 from the axial direction. The ar- specimen cross-sections transverse to the test direction chitecture was then cut axially and laid up in tooling to Table I lists the key properties related to each of these form the final panel with the hoop fibers oriented +23 architectures from the Y loading direction As described in the top half of Table I, one panel Tensile specimens, 12.6 mm wide and 150 mm for this study, AI-UNI, was fabricated in order to ap- long(see Table I for average thickness values),were roach a unidirectional ID composite. The Al-UNI, an machined into a dogbone shape where the gage section angle interlock architecture was woven with a high frac- width and length were 10 and 40 mm, respectively tion of Sylramic-iBN fibers(0. 23)aligned in the fill The ends of the tensile bars were encased in a wire or test direction and both rayon and ZMI warp weaver mesh to alleviate grip stresses and bending moments hrough-the-thickness fibers(a small fraction of each). at and near the pneumatic pressure grips. Room tem- Two other panels, 3DO-Un-R and 3DO-Un-Z, were perature tensile tests were performed according to fabricated with 3D-orthogonal fber architectures where ASTM C1275 using a universal testing machine Sylramic fibers were woven in different proportions in (Model 8562, Instron, Canton, MA). Specimens were he warp and fill directions, and either rayon fibers or loaded at a constant rate( kN/min). Two clip-on strain ZMI fibers, respectively, were used as the through-thick- gages(2.5% max strain) were attached (one on each ness or Z direction fber tows. For these 3D orthogonal face)and the average of the two strain gages was used for panels, the final volume fractions of the Sylramic-ibn the tests. Unload-reload interruptions were also per fibers were - 0. 1 in the X direction, and -0 27 in the Y formed, usually at least two per test, in order to deter direction (test direction). The fifth panel, LTLAL, was a mine the residual compressive stress in the composite Sylramic-iBN layer-to-layer angle interlock consisting of three layers and a low fiber volume fraction in the yor Modal acoustic emission was also monitored dur- est direction ing the room temperature tensile tests. A Digital Wave Also shown in the bottom half of Table I are Fracture Wave Detector with two wide-band pass(50- MI SiC/SiC panels from earlier studies that will be 2000 kHz) frequency sensors also from Digital Wave used for property comparison and model development Corporation (Model B1025, Englewood, CO)was in this study. All of these panels were processed in the used. The two AE sensors were placed on the face of same manner regarding the BN fiber coating the specimen, one on each side of the gage section,ap-
were the high-performance Sylramic-iBN SiC fibers, which display high strength, high creep–rupture resistance, and high thermal conductivity.12 These fi- bers were originally produced as Sylramict SiC fibers in 800 filament tow form by Dow Corning, Midland, MI. The tows were then heat treated at NASA either as individual tows or preferably within a textile-formed preform in a nitrogen-containing environment to convert the Sylramic fibers into Sylramic-iBN fibers.13 In some cases, other fiber types such as the Ube (Yamaguchi, Japan) ‘‘Tyranno ZMI’’ SiC and/or rayon fiber tows were incorporated into the initial architectures because of their low-modulus and better formability, for example, as the warp-weaver fibers. For those architectures that used the polymer-based rayon fiber, significant degradation was observed in these fibers due to its fugitive nature at the high fibercoating and matrix processing temperatures with only a remnant char remaining that was subsequently coated with BN and SiC during matrix processing. The in situ architectures for the four SiC/SiC panels fabricated in this study can be seen in Fig. 1 which shows tensile specimen cross-sections transverse to the test direction. Table I lists the key properties related to each of these architectures. As described in the top half of Table I, one panel for this study, AI-UNI, was fabricated in order to approach a unidirectional 1D composite. The AI-UNI, an angle interlock architecture was woven with a high fraction of Sylramic-iBN fibers (B0.23) aligned in the fill or test direction and both rayon and ZMI warp weaver through-the-thickness fibers (a small fraction of each). Two other panels, 3DO-Un-R and 3DO-Un-Z, were fabricated with 3D-orthogonal fiber architectures where Sylramic fibers were woven in different proportions in the warp and fill directions, and either rayon fibers or ZMI fibers, respectively, were used as the through-thickness or Z direction fiber tows. For these 3D orthogonal panels, the final volume fractions of the Sylramic-iBN fibers were B0.1 in the X direction, and B0.27 in the Y direction (test direction). The fifth panel, LTLAI, was a Sylramic-iBN layer-to-layer angle interlock consisting of three layers and a low fiber volume fraction in the Y or test direction. Also shown in the bottom half of Table I are MI SiC/SiC panels from earlier studies that will be used for property comparison and model development in this study. All of these panels were processed in the same manner regarding the BN fiber coating (B0.5 mm thick) and MI SiC matrix. For the previous 3D orthogonal panels where rayon fibers were used, the preferred NASA conversion process for Sylramic-iBN preforms was not used because of initial concern related to degradation of the architectures. Thus the in-plane fibers for these panels were only Sylramic fibers which, other than a significant difference in creep and rupture behavior, show essentially the same elastic and tensile strength behavior as the Sylramic-iBN fibers when incorporated into MI SiC/ SiC composites. The densities of all of the panels and specimens evaluated in this study were 42.85 g/cm3 . This implies that most of the porosity is resident within the tow minicomposites (B5% total porosity) typical of good MI infiltration as observed in Fig. 1. It should also be noted in Table I that in contrast to the other panels, the in-plane Sylramic-iBN fibers of the Braid panel were not oriented in the testing direction. For this panel, a 2D tri-axial braided architecture was first formed on a 50 mm diameter mandrel with the hoop fibers 7671 from the axial direction.15 The architecture was then cut axially and laid up in tooling to form the final panel with the hoop fibers oriented 7231 from the Y loading direction. Tensile specimens, 12.6 mm wide and 150 mm long (see Table I for average thickness values), were machined into a dogbone shape where the gage section width and length were B10 and B40 mm, respectively. The ends of the tensile bars were encased in a wire mesh to alleviate grip stresses and bending moments at and near the pneumatic pressure grips. Room temperature tensile tests were performed according to ASTM C127516 using a universal testing machine (Model 8562, Instron, Canton, MA). Specimens were loaded at a constant rate (4 kN/min). Two clip-on strain gages (2.5% max strain) were attached (one on each face) and the average of the two strain gages was used for the tests. Unload–reload interruptions were also performed, usually at least two per test, in order to determine the residual compressive stress in the composite matrix. Modal acoustic emission was also monitored during the room temperature tensile tests. A Digital Wave Fracture Wave Detector with two wide-band pass (50– 2000 kHz) frequency sensors also from Digital Wave Corporation (Model B1025, Englewood, CO) was used. The two AE sensors were placed on the face of the specimen, one on each side of the gage section, apwww.ceramics.org/ACT Effects of Fiber Architecture on SiC/SiC Composites 279
280 International yournal of Applied Ceramic Techmolog Vol.7,No.3,2010 Y⊕ (d) Fig. 1. Cross-sections of MI SiC/SiC specimens:( a)A/,(b)3DO-Un-R(c) 3DO-Un-Z and(d)LTLAl. The widths of the specimens are 10mm(= warp direction and y= fill direction). A high magnification image of3Do-Un-Z(e)shows some of the details of the MI microstructure proximately 50-60 mm from one another. The two aE curred in the gage section were sorted out using a sensors were synchronized, that is, both sensors would threshold voltage crossing technique"and used for record the waveform from the same event at the same analysis according to the ocala each event based time if either sensor was triggered. Events which oc- on the speed of sound of the extensional wave, which
proximately 50–60 mm from one another. The two AE sensors were synchronized, that is, both sensors would record the waveform from the same event at the same time if either sensor was triggered. Events which occurred in the gage section were sorted out using a threshold voltage crossing technique17 and used for analysis according to the location of each event based on the speed of sound of the extensional wave, which Fig. 1. Cross-sections of MI SiC/SiC specimens: (a) AI-UNI, (b) 3DO-Un-R, (c) 3DO-Un-Z and (d) LTLAI. The widths of the specimens are 10 mm (X 5 warp direction and Y 5 fill direction). A high magnification image of 3DO-Un-Z (e) shows some of the details of the MI microstructure. 280 International Journal of Applied Ceramic Technology—Morscher, et al. Vol. 7, No. 3, 2010
wwceramics. org/ACT Efects of Fiber Architecture on SiC/SiC Composites was determined posttest from events which occurred fraction in the loading direction increases the point of outside of the sensors, 0, 7 nonlinearity in the stress-strain curve, that is, both the Because the panels in this study varied considerably composite proportional limit stress(PLS)and strain in- in fiber architecture and thickness, and because the ef- crease with fo. Note that in Fig. 2a the 2D 5HS N24A fective fiber volume fraction in the loading or test di- woven composite is considered to be the standard bal- rectI parameter in data analyses, a anced material developed at NASA Glenn Research geometric approach is used in determining fo by the Ce enter unde he Ultra Efficient Energy Technology program and is considered the state-of-the-art material fo=(Nply)(Nows/ply )(Nr/tow )(TR)/(nw) (1) now being investigated for various engine applica- of the Here Ply is the number of plies or layers through sured from the stress-strain curves are listed in table il or layer; Frow is the number of fibers per tow(800 for including elastic modulus, ultimate tensile strength Syl-iBN), R is the average fiber radius (5 um for Syl (UTS), and residual stress on the matrix due to ther- width. Nrowwply can be determined from the architecture of the Si as determined from the hysteresis technique of fabrication parameter tow end Steen. 8 There was some variation in elastic modulus number of fiber tows in a ply per cm oriented in a give The higher fiber volume fraction composites, especially orthogonal direction) by multiplying epcm/10(the fac those with fewer 90 minicomposites had the highest tor 10 to convert to mm)times the width. This results elastic moduli t This would be expected because the in a simplification of Eq (1)to lowest moduli element of a textile composite is the stiff- ness contribution of 90 minicomposites. The UTS f o=(Nply )(epcm)(N/ow )(IR )/10t(2) and strain properties of the composites also varied con- siderably as expected because volume fraction of the fi- It is important to note that in Eq(2), the volume bers in the loading direction varied consideral fraction fo is based on the thickness measurement used In table ii. the calculated stress on the fibers at to determine stress for a specimen and not the average failure is also listed; that is, the UTS divided by fiber thickness of a panel. The Nply, epcm, and range of fraction in the loading direction, fo. As a reference, the average stress on the fibers at failure Table I. For the Braid panel, where the load-bearing 2400+200 MPa for the standard balanced 2D five-har- Sylramic-iBN fibers were at angle +23 to the loading ness satin N24A composites over a wide range of fiber direction,the f. value included both sets of bias tows volume fractions. 0 For the composites fabricated in this (i. e, Eq(2)multiplied by 2) study, several specimens did not fail in the gage section Some of the specimens were polished along the as indicated by the ">"sign in Table Il; consequently, length in order re matrix crack density. The the absolute ultimate tensile stress for these composites plished specimens were then etched with a Cfa plasma is not known. Even so, all of the composites exhibited at 500 Watts for 30 min in order to enhance the matrix fiber stresses of at least 2000 MPa approaching that of cracks in the CVI SiC region of the microstructure. the standard N24A composites. However, similar fber Average matrix crack densities were measured over dis stress levels were not observed in some of the earlier ances over a length of at least 5 mm studies. The fibers in the braided composite>were aligned at a 23 angle to the loading direction where Results fibers would be subject to bending and shear. The bal- anced 3D orthogonal composites also had low fiber Room Temperature Stress-Strain Behavior stresses at failure. the reason for which could not b Some representative tensile stress strain curves of individual specimens from some of the panels are shown in Fig. 2a. The general trend in Fig. 2a, as observed in mposite refers to the composite entity consisting of the BN-coated Sic the earlier study, is clear: increasing the fiber volum fiber tow that is overcoated with the CVI SiC matrix. For orthogonal architectures. these sites run in the 0°and90° direction
was determined posttest from events which occurred outside of the sensors.10,17 Because the panels in this study varied considerably in fiber architecture and thickness, and because the effective fiber volume fraction in the loading or test direction, fo, is a key parameter in data analyses, a geometric approach is used in determining fo by the following relation: fo ¼ ðNplyÞðNtows=plyÞðNf=towÞðpR2 f Þ=ðtwÞ ð1Þ Here Nply is the number of plies or layers through the thickness; Ntows/ply is the number of tows per ply or layer; Nf/tow is the number of fibers per tow (800 for Syl-iBN), Rf is the average fiber radius (5 mm for SyliBN), t is the specimen thickness, and w is the specimen width. Ntows/ply can be determined from the architecture fabrication parameter ‘‘tow ends per cm’’ (epcm, the number of fiber tows in a ply per cm oriented in a given orthogonal direction) by multiplying epcm/10 (the factor 10 to convert to mm) times the width. This results in a simplification of Eq. (1) to fo ¼ ðNplyÞðepcmÞðNf=towÞðpR2 f Þ=10t ð2Þ It is important to note that in Eq. (2), the volume fraction fo is based on the thickness measurement used to determine stress for a specimen and not the average thickness of a panel. The Nply, epcm, and range of t values for the orthogonal panels can be found in Table I. For the Braid panel, where the load-bearing Sylramic-iBN fibers were at angle 7231 to the loading direction, the fo value included both sets of bias tows (i.e., Eq. (2) multiplied by 2). Some of the specimens were polished along the length in order to measure matrix crack density. The polished specimens were then etched with a CF4 plasma at 500 Watts for 30 min in order to enhance the matrix cracks in the CVI SiC region of the microstructure. Average matrix crack densities were measured over distances over a length of at least 5 mm. Results Room Temperature Stress–Strain Behavior Some representative tensile stress strain curves of individual specimens from some of the panels are shown in Fig. 2a. The general trend in Fig. 2a, as observed in the earlier study,10 is clear: increasing the fiber volume fraction in the loading direction increases the point of nonlinearity in the stress–strain curve, that is, both the composite proportional limit stress (PLS) and strain increase with fo. Note that in Fig. 2a the 2D 5HS N24A woven composite is considered to be the standard balanced material developed at NASA Glenn Research Center under the Ultra Efficient Energy Technology program and is considered the state-of-the-art material now being investigated for various engine applications.11 Some of the composite mechanical properties measured from the stress–strain curves are listed in Table II, including elastic modulus, ultimate tensile strength (UTS), and residual stress on the matrix due to thermal expansion of the different phases and solidification of the Si as determined from the hysteresis technique of Steen.18 There was some variation in elastic modulus. The higher fiber volume fraction composites, especially those with fewer 901 minicomposites had the highest elastic moduli.w This would be expected because the lowest moduli element of a textile composite is the stiffness contribution of 901 minicomposites.19 The UTS and strain properties of the composites also varied considerably as expected because volume fraction of the fi- bers in the loading direction varied considerably. In Table II, the calculated stress on the fibers at failure is also listed; that is, the UTS divided by fiber fraction in the loading direction, fo. As a reference, the average stress on the fibers at failure is typically 24007200 MPa for the standard balanced 2D five-harness satin N24A composites over a wide range of fiber volume fractions.10 For the composites fabricated in this study, several specimens did not fail in the gage section as indicated by the ‘‘4’’ sign in Table II; consequently, the absolute ultimate tensile stress for these composites is not known. Even so, all of the composites exhibited fiber stresses of at least 2000 MPa approaching that of the standard N24A composites. However, similar fiber stress levels were not observed in some of the earlier studies.10 The fibers in the braided composite15 were aligned at a 231 angle to the loading direction where fibers would be subject to bending and shear. The balanced 3D orthogonal composites also had low fiber stresses at failure, the reason for which could not be explained.6 w The term minicomposite refers to the composite entity consisting of the BN-coated SiC fiber tow that is overcoated with the CVI SiC matrix. For orthogonal architectures, these minicomposites run in the 0o and 90o direction. www.ceramics.org/ACT Effects of Fiber Architecture on SiC/SiC Composites 281
ional Journal of Applied Ceramic Technology Vol.7,No.3,2010 700 3DO Un-Z fo 0.27 600 3DO UnR fo =0.28 fo=0.23 Braid: fo= 0.26 4 2D 5HS N24A 400 3DO Balz Warp: o=0. 17 300 200 2D 5HS 4epcm 10 LTL Al fo =0.1 0 0.1 0.2 0.3 04 05 0.6 b) 0.8 fo Al UNI =0.1 4 2D 5HS N24A o=0.19 E0.6 05 DO Un-R fo=0.28 0.1 fo=0.27 0 100 300 Stress, MPa Fig. 2. Representative (a) tensile stress-strain curves and (b)acoustic emission activity versus stress of specimens tested in this study an compared from earlier studies(see Table D) Acoustic Emission and Matrix Cracking Bebavior trix cracks in sic fiber reinforced mi SiC matrix com posites for different fiber architectures. 10, 17and different les abish ed Au neity imatixseda king lpersptries oh ibed con the ihig esm prsir f mhe oude evers are was implemented. It is important to note that only aE which occur in the gage section not associated with ul events that originate in the gage section are used in the timate failure. It was shown in the earlier studies that the analysis. It has been established in the earlier studies that cumulative AE energy parameter relates directly with the the highest energy events, here referred to as"loud observed fiber-bridged matrix crack density. This is be events, "correspond to large fiber-bridged transverse ma-
Acoustic Emission and Matrix Cracking Behavior In order to quantify matrix cracking properties, the well-established AE technique used in earlier studies10,17 was implemented. It is important to note that only AE events that originate in the gage section are used in the analysis. It has been established in the earlier studies that the highest energy events, here referred to as ‘‘loud events,’’ correspond to large fiber-bridged transverse matrix cracks in SiC fiber reinforced MI SiC matrix composites for different fiber architectures6,10,17 and different fiber-containing composites.20 The loud events are de- fined as the highest order of magnitude energy events which occur in the gage section not associated with ultimate failure. It was shown in the earlier studies that the cumulative AE energy parameter relates directly with the observed fiber-bridged matrix crack density. This is because the cumulative energy parameter is dominated by Fig. 2. Representative (a) tensile stress–strain curves and (b) acoustic emission activity versus stress of specimens tested in this study and compared from earlier studies (see Table I). 282 International Journal of Applied Ceramic Technology—Morscher, et al. Vol. 7, No. 3, 2010
wwceramics. org/ACT Efects of Fiber Architecture on SiC/SiC Composites the loud events even though typically only 20% of all AE events are loud. These high-energy events are two to four 98求8 a orders of magnitude higher in energy than the majority of AE events. They do not correspond to tunnel micro- cracks, fiber sliding, or even fber breakage because in order to create these high energy acoustic pulses, a large amount of surface area must be created which can only be attributed to the formation of large if not through-thick ness fber-bridged transverse matrix cracks The AE activity was plotted as normalized cumu s点3R三 lative AE energy(NCAE= cumulative AE energy of all the events up to a given stress divided by the total AE a energy of all the events of the test) versus stress in Fig 2b. This essentially represents a relative distribution of 组|三三三2三空 stress. If the final mat山m由 这2斗 value can be multiplied by the NCaE to get an,c mated stress-dependent matrix crack density which can then be used to model stress-strain behavior. Impor tant aE properties measured for some of the specimens are shown in Table il 当当学器盐显斗 nergy events dominate the AE energy curve,)am a cated in Fig 2b by the extrapolation of the steep slope of the NCae curve with stress to the abscissa, is generally referred to as the ae onset stress for matrIX because it closely represents the occurrence of fiber- bridged matrix cracks which either are considered to be through-thickness matrix cracks ( TTMC) or at least 名到保溪两叫mAm矩顺m the Note that there if any loud AE events prior to the ae onset stress a graphical offset technique is often applied to II lists both the ae onset stress and the 0.005% offset stress, that is, where the offset line E x Strain +0.005% intersects the stress-strain curve. There is good corre- MI composite system as was observed in reference 21 会5§%出 Note that for lower modulus fiber-containing MI tems, the aE onset stress is typically lower than PLS a3 There were a few exceptions to the equivalence of PLS and ae onset stress for the architectures co The 3DO-bal-z composites had very large ZMI tows as
the loud events even though typically only 20% of all AE events are loud. These high-energy events are two to four orders of magnitude higher in energy than the majority of AE events. They do not correspond to tunnel microcracks, fiber sliding, or even fiber breakage because in order to create these high energy acoustic pulses, a large amount of surface area must be created which can only be attributed to the formation of large if not through-thickness fiber-bridged transverse matrix cracks. The AE activity was plotted as normalized cumulative AE energy (NCAE 5 cumulative AE energy of all the events up to a given stress divided by the total AE energy of all the events of the test) versus stress in Fig. 2b. This essentially represents a relative distribution of fiber-bridged matrix cracks for a given specimen with stress. If the final matrix crack density is known, that value can be multiplied by the NCAE to get an estimated stress-dependent matrix crack density which can then be used to model stress–strain behavior.11 Important AE properties measured for some of the specimens are shown in Table II. The onset of significant AE (when individual high energy events dominate the AE energy curve), as indicated in Fig. 2b by the extrapolation of the steep slope of the NCAE curve with stress to the abscissa, is generally referred to as the AE onset stress for matrix cracking because it closely represents the occurrence of fiberbridged matrix cracks which either are considered to be through-thickness matrix cracks (TTMC) or at least large fiber-bridged cracks which intersect with the surface face or edge of the composite. Another AE related stress is noted in Table II: the stress at which the first AE event occurred in the gage section. The first AE event represents the stress at which 901 tunnel cracks begin to form in these composites.19 Note that there usually are only a few if any loud AE events prior to the AE onset stress. A graphical offset technique16 is often applied to the stress–strain curve to determine the onset of through-thickness matrix cracking stress or PLS. Table II lists both the AE onset stress and the 0.005% offset stress, that is, where the offset line E Strain10.005% intersects the stress–strain curve. There is good correspondence between the two techniques for this Syl-iBN/ MI composite system as was observed in reference 21. Note that for lower modulus fiber-containing MI systems, the AE onset stress is typically lower than PLS.20 There were a few exceptions to the equivalence of PLS and AE onset stress for the architectures compared here. The 3DO-bal-Z composites had very large ZMI tows as Table II. Room Temperature Mechanical Properties SiC/SiC panel type (reference) Fiber fraction (fo) in load direction E (GPa) UTS (MPa) Stress on fibers at failure (MPa) First AE event stress (MPa) AE onset stress, MPa {# loud events prior} 0.005% offset stress (MPa) Residual Stress, MPa ( ) signifies compression AI-UNI 0.23 305 74 4472 42052 223 720 301 79 {10} 322 77 78 77 3DO-Un-R 0.28 275 79 4575 42053 208 725 270 714 {6} 261 716 55 715 3DO-Un-Z 0.27 262 79 596 2207 163 733 245 713 {1} 230 712 50 LTLAI 0.10 125 204 2040 48 80 {1} 89 20 2D 5HS N24A14# 0.19 250 463 2362 95 182 175 50 2D 5HS DT10 0.19 197 480 2526 136 78 164 76 {2} 142 50 Braid15 0.26 250 710352 718 B1352 109 726 213 72 245 714 60 3DO-Bal-R-Y6 0.20 238 336 1680 160 180 {1} 150 47 3DO-Bal-Z-Y6 0.17 248 317 1864 35 119 {0} 78 40 3DO-Bal-Z-X6 0.18 205 322 1788 47 126 {0} 82 32 For data with scatter, the scatter is based on tests of two different specimens. All other data is based on a single specimen. #The NASA N24A composite system is a 7.9 epcm, 8 ply, 2D five-harness architecture. It is shown here for comparison and since it is considered to be a ‘‘standard’’ material by NASA.11 www.ceramics.org/ACT Effects of Fiber Architecture on SiC/SiC Composites 283
International Journal of Applied Ceramic Technology-Morscher, et al. Vol.7,No.3,2010 3501+AUN ■3 DO Unbalanced 300 3DO Bal Fill 200 x LTLAI 0.06 150 ■+xo△口 Fig 3. Matrix cracking stress()and strain(b)versus fiber volume fraction in the loading direction for the different composites. the Z-direction reinforcement which was prone to low observe the matrix cracks. Because the matrix is in com- stress tunnel cracks confined to the ZMI minicomposite pression, matrix cracks are impossible to locate without which accounted for the nonlinearity. The degree of the plasma etching. Unfortunately plasma etching re- nonlinearity is dependent on several factors: the number moves most of the Si from the SiC-particulate, Si teroy of cracks and whether they are micro or macro in na- of the matrix so that cracks can only be observed in ture, volume fraction of fibers, fiber and matrix elastic CVI SiC of the minicomposites. Figure 4 shows some moduli, interfacial sliding stress, straightness of fibers, representative longitudinal sections of a 3D orthogonal tc. The graphical offset technique does not account for specimen and the AI-UNI specimen. Typical matrix the physical dependence of nonlinearity which will re- cracks are shown in Fig, 4c. Note that the cracks were ate to fiber-bridged matrix cracking differently for the very faint even after etching; however, the cracks were factors listed above and so it is not surprising that the observed to traverse through-the-thickness of the speci- two techniques vary for some cases. aE does measure men. The crack densities estimated for the 3DO-Un-R the occurrence of fiber-bridged matrix cracks and is and AI-UNI specimens were at least eight and nine considered to be the more desired method for deter- cracks/mm, respectively. These crack densities are similar ning the onset stress for through-thickness or fiber- to those measured in other Syl-iBN MI composites It should be noted that the plasma etch of the 3DO-Un The effect of fiber volume fraction on matrix crack- Z specimen was not as successful. Although some cracks ress and strain is shown in Fig. 3a and b, respec- could be observed, the surface was stained to a degree The general trend is that increasing volume that reliable crack density measurements could be made. fraction of fibers in the loading direction results in higher matrix cracking stresses and higher matrix crack ing strains, a roughly linear dependence for both.0. aNalysis However, there is considerable scatter in the data ±50 MPa in stress and~±0.03% in strain. There appears to be a general relationship between matrix cracking and fiber volume fraction of fibers in Optical Microscopy the load direction (or fibers that would be bridging a crack in the case of the braid panel). However, there still Longitudinal sections of the specimens were cut, is considerable scatter. A second effect related to matrix lished. and etched as in earlier studies 6. I0 in order to racking postulated in Morscher et al. was that large
the Z-direction reinforcement which was prone to low stress tunnel cracks confined to the ZMI minicomposite which accounted for the nonlinearity.6 The degree of nonlinearity is dependent on several factors: the number of cracks and whether they are micro or macro in nature, volume fraction of fibers, fiber and matrix elastic moduli, interfacial sliding stress, straightness of fibers, etc. The graphical offset technique does not account for the physical dependence of nonlinearity which will relate to fiber-bridged matrix cracking differently for the factors listed above and so it is not surprising that the two techniques vary for some cases. AE does measure the occurrence of fiber-bridged matrix cracks and is considered to be the more desired method for determining the onset stress for through-thickness or fiberbridged matrix cracking. The effect of fiber volume fraction on matrix cracking stress and strain is shown in Fig. 3a and b, respectively. The general trend is that increasing volume fraction of fibers in the loading direction results in higher matrix cracking stresses and higher matrix cracking strains, a roughly linear dependence for both.10,13 However, there is considerable scatter in the data, B750 MPa in stress and B70.03% in strain. Optical Microscopy Longitudinal sections of the specimens were cut, polished, and etched as in earlier studies6,10 in order to observe the matrix cracks. Because the matrix is in compression, matrix cracks are impossible to locate without the plasma etching. Unfortunately plasma etching removes most of the Si from the SiC-particulate, Si region of the matrix so that cracks can only be observed in the CVI SiC of the minicomposites. Figure 4 shows some representative longitudinal sections of a 3D orthogonal specimen and the AI-UNI specimen. Typical matrix cracks are shown in Fig. 4c. Note that the cracks were very faint even after etching; however, the cracks were observed to traverse through-the-thickness of the specimen. The crack densities estimated for the 3DO-Un-R and AI-UNI specimens were at least eight and nine cracks/mm, respectively. These crack densities are similar to those measured in other Syl-iBN MI composites.6,10 It should be noted that the plasma etch of the 3DO-UnZ specimen was not as successful. Although some cracks could be observed, the surface was stained to a degree that reliable crack density measurements could be made. Analysis There appears to be a general relationship between matrix cracking and fiber volume fraction of fibers in the load direction (or fibers that would be bridging a crack in the case of the Braid panel). However, there still is considerable scatter. A second effect related to matrix cracking postulated in Morscher et al. 6 was that larger 0 50 100 150 200 250 300 350 0 AE Onset Stress, MPa AI UNI 3DO Unbalanced braid 2D 5HS 2D 5HS - double tow 2D 5HS N24A 3DO Bal Y 3DO Bal Fill LTLAI 0 0.02 0.04 0.06 0.08 0.1 0.12 0 AE Onset Strain, % 3DO Unbalanced AI UNI 2D 5HS 2D 5HS - double tow 2D 5HS N24A braid 3DO Bal Y 3DO Bal Fill LTL AI fo 0.1 0.2 0.3 0.1 0.2 0.3 fo Fig. 3. Matrix cracking stress (a) and strain (b) versus fiber volume fraction in the loading direction for the different composites. 284 International Journal of Applied Ceramic Technology—Morscher, et al. Vol. 7, No. 3, 2010
wwceramics. org/ACT Efects of Fiber Architecture on SiC/SiC Composites (a) 90 Syl-iBN Stress Direction Z Rayon tow SyriAN 1 mm R L ZM Stress Direction CVI SIC Matrix Cracks microstructures of longitudinal polished sections of failed(a)3D orthogonal co 3DO-Un-R) and (b)through-the-thickness angle interlock composite with ZMI and Rayon warp weaver(Al-UND. Two faint matrix cracks propagating through the CVI portion of a minicomposite for an Al-UNI sPecimen are shown in(c). For(c), the image brightness had to be decreased and contrast increased significantly in order to enhance the cracks. The specimens were plasma etched which is the reason for the rather dull appearance of the MI region of the matrix. unbridged regions of transverse fiber tow minicompos- the 90 tow minicomposites at stresses just below ites perpendicular to the loading direction would result TTMC formation. This was vindicated to some of in lower matrix cracking stress if the fiber volume frac tent for the 3D orthogonal"Balanced"composites tion was the same. The rationale for this was that the Morscher et al6 where it was demonstrated that the 3D observation that the sources of TTMC matrix cracks in orthogonal composites woven with larger Z fiber tows nse MI composites are the tunnel cracks formed in had lower matrix cracking stresses than 3D orthogonal
unbridged regions of transverse fiber tow minicomposites perpendicular to the loading direction would result in lower matrix cracking stress if the fiber volume fraction was the same. The rationale for this was that the observation that the sources of TTMC matrix cracks in dense MI composites are the tunnel cracks formed in the 901 tow minicomposites at stresses just below TTMC formation.21 This was vindicated to some extent for the 3D orthogonal ‘‘Balanced’’ composites of Morscher et al. 6 where it was demonstrated that the 3D orthogonal composites woven with larger Z fiber tows had lower matrix cracking stresses than 3D orthogonal Fig. 4. Representative microstructures of longitudinal polished sections of failed (a) 3D orthogonal composite with Rayon as the Z fiber (3DO-Un-R) and (b) through-the-thickness angle interlock composite with ZMI and Rayon warp weavers (AI-UNI). Two faint matrix cracks propagating through the CVI portion of a minicomposite for an AI-UNI specimen are shown in (c). For (c), the image brightness had to be decreased and contrast increased significantly in order to enhance the cracks. The specimens were plasma etched which is the reason for the rather dull appearance of the MI region of the matrix. www.ceramics.org/ACT Effects of Fiber Architecture on SiC/SiC Composites 285
