J.Am. Ceran.Soc,90o3185-3193(2007 DOL: 10.1111 C 2007 The American No claim to original US government works urna In-Plane Cracking behavior and Ultimate Strength for 2D Woven and Braided Melt-Infiltrated SiC/SiC Composites Tensile Loaded in Off-Axis Fiber directions gregory n. morscher Ohio Aerospace Institute, Cleveland, Ohio Hee mann yun Matech GSM. Irvine. Califor NASA Glenn Research Center. Cleveland, ohio The tensile mechanical properties of ceramic matrix composites referred to as proportional limit stress), and fiber-pullout mech- CMC in directions off the primary axes of the reinforcing fi- anisms that lead to graceful failure and high ultimate tensile bers are important for the architectural design of CMc com- strength (UTS). This is the case because the high-modulus fi nents that are subjected to multiaxial stress states. In this bers are oriented to carry much of the tensile load applied to the study, two-dimensional (2D)woven melt-infiltrated (MD) sic composite before and after the dfls point, which is caused by Sic composite panels with balanced fiber content in the 0 ai matrix cracks 90 directions were tensile loaded in-plane in the 0 direction and (TTMc)in the CMC. However, when loaded in a direction at a at 45 to this direction. In addition, a 2D triaxially braided MI significant angle to the primary fiber axes, large reductions in Sic/SiC composite panel with a higher fiber content in the key CMC design properties such as low DFLS and UTS can +67 bias directions compared with the axial direction was occur. Whereas on-axis properties are strongly dependent on tensile loaded perpendicular to the axial direction tows (i.e 23 fiber properties, off-axis properties are strongly dependent on from the bias fibers). Stress-strain behavior, acoustic emission, matrix properties, particularly on their stiffness and load-carry nd optical microscopy were used to quantify stress-dependent ing ability, which are typically related to their porosity content matrix cracking and ultimate strength in the panels. It was ob- For example, when CMC with highly porous oxide matrices served that both off-axis-loaded panels displayed higher com- were tested off-axis, highly nonlinear stress-strain behavior and posite onset stresses for through-thickness matrix cracking elatively weak strengths were observed because the porous ma- he 2D-woven 0/90 panels loaded in the primary 0 direction. trix could not carry significant load But when the matrix stiff- d load-c be attributed to higher effective fiber fractions in the loading porous matrix, higher dFl stresses and ultimate strengths were direction, which in turn reduces internal stresses on weak regions obtained, but still not as high as the on-axis value ite tows oriented normal For CMC structural applications, high stresses for TTMC are the loading direction and or critical flaws in the matrix for enerally desired in all directions. This not only allows mainte- posite stress. Both off-axis-oriented panels also nance of composite modulus and thermal conductivity to high showed relatively good ultimate tensile strength when compared stresses but also results in greater composite life for CMC with with other off-axis-oriented composites in the literature, both on nonoxide constituents that can be degraded by environmental an absolute strength basis as well as when normalized by the permeability into TTMC. To achieve these high cracking stress- average fiber strength within the composites. Initial implications es, high-stiffness low-porosity matrices are generally preferred are discussed for constituent and architecture design to improve hich can also help to improve the CMc creep-rupture resis- the directional cracking of SiC/SiC CMC components with MI tance and thermal conductivity. One such CMC system is the matrices melt-infiltrated (MD) SiC matrix system reinforced by the syl- ramic-iBN SiC fiber that is near-stoichiometric in composition and contains a thin in sittt-grown BN layer on its surface. This L. Introducti Sic/SiC composite system is typically processed by taking a woven or braided fiber-preform, coating the fibers with another HE tensile mechanical properties of continuous fiber-rein- thin bn layer by chemical vapor infiltration(CVD, and then forced ceramic matrix composites( CMC) vary according to forming an initial SiC matrix by CV. The remaining matrix the orientation of fibers with respect to the loading direction porosity is then filled with a slurry of SiC particles. After drying CMC mechanical properties are typically measured for tensile the slurry-infiltrated preform, final matrix formation is by mel stresses in a direction parallel to one of the primary fiber axes, infiltration (MD) of liquid Si, which fills nearly all of the large hich generally results in desirable linear stress-st ores in the structure, leaving 5% closed porosity a high stress for deviation from linearity or(DFLS)(also The primary objective of this study was to measure and an- alyze the off-axis in-plane tensile stress-strain behavior of five F. Zok--contributing editor thin-walled panels with the MI Sylramic-iBNSIC system con taining two different basic fiber architectures. A and B. Four panels with architecture A, consisting of a two-dimensional (2D)-woven fabric lay-up with balanced fiber content in the 0 anuscript No. 22879 Received March 5. 2007; approved June 2.2007. uthor to whom correspondence should be addressed. e-mail: gmorscher(a and 90 directions, were tensile loaded in the 0 direction(panels grc.nasa.go Al-3)and at 45. to this direction(panel A4). Another panel 3185
In-Plane Cracking Behavior and Ultimate Strength for 2D Woven and Braided Melt-Infiltrated SiC/SiC Composites Tensile Loaded in Off-Axis Fiber Directions Gregory N. Morscherw Ohio Aerospace Institute, Cleveland, Ohio Hee Mann Yun Matech GSM, Irvine, California James A. DiCarlo NASA Glenn Research Center, Cleveland, Ohio The tensile mechanical properties of ceramic matrix composites (CMC) in directions off the primary axes of the reinforcing fi- bers are important for the architectural design of CMC components that are subjected to multiaxial stress states. In this study, two-dimensional (2D)-woven melt-infiltrated (MI) SiC/ SiC composite panels with balanced fiber content in the 01 and 901 directions were tensile loaded in-plane in the 01 direction and at 451 to this direction. In addition, a 2D triaxially braided MI SiC/SiC composite panel with a higher fiber content in the 7671 bias directions compared with the axial direction was tensile loaded perpendicular to the axial direction tows (i.e., 231 from the bias fibers). Stress–strain behavior, acoustic emission, and optical microscopy were used to quantify stress-dependent matrix cracking and ultimate strength in the panels. It was observed that both off-axis-loaded panels displayed higher composite onset stresses for through-thickness matrix cracking than the 2D-woven 0/90 panels loaded in the primary 01 direction. These improvements for off-axis cracking strength can in part be attributed to higher effective fiber fractions in the loading direction, which in turn reduces internal stresses on weak regions in the architecture, e.g., minicomposite tows oriented normal to the loading direction and/or critical flaws in the matrix for a given composite stress. Both off-axis-oriented panels also showed relatively good ultimate tensile strength when compared with other off-axis-oriented composites in the literature, both on an absolute strength basis as well as when normalized by the average fiber strength within the composites. Initial implications are discussed for constituent and architecture design to improve the directional cracking of SiC/SiC CMC components with MI matrices. I. Introduction THE tensile mechanical properties of continuous fiber-reinforced ceramic matrix composites (CMC) vary according to the orientation of fibers with respect to the loading direction. CMC mechanical properties are typically measured for tensile stresses in a direction parallel to one of the primary fiber axes, which generally results in desirable linear stress–strain behavior, a high stress for deviation from linearity or (DFLS) (also referred to as proportional limit stress), and fiber-pullout mechanisms that lead to graceful failure and high ultimate tensile strength (UTS).1 This is the case because the high-modulus fi- bers are oriented to carry much of the tensile load applied to the composite before and after the DFLS point, which is caused by the development of transverse through-thickness matrix cracks (TTMC) in the CMC. However, when loaded in a direction at a significant angle to the primary fiber axes, large reductions in key CMC design properties such as low DFLS and UTS can occur.2 Whereas on-axis properties are strongly dependent on fiber properties, off-axis properties are strongly dependent on matrix properties, particularly on their stiffness and load-carrying ability, which are typically related to their porosity content. For example, when CMC with highly porous oxide matrices were tested off-axis, highly nonlinear stress–strain behavior and relatively weak strengths were observed because the porous matrix could not carry significant load.3 But when the matrix stiffness and load-carrying ability were increased via sintering the porous matrix, higher DFL stresses and ultimate strengths were obtained, but still not as high as the on-axis value. For CMC structural applications, high stresses for TTMC are generally desired in all directions. This not only allows maintenance of composite modulus and thermal conductivity to high stresses but also results in greater composite life for CMC with nonoxide constituents that can be degraded by environmental permeability into TTMC. To achieve these high cracking stresses, high-stiffness low-porosity matrices are generally preferred, which can also help to improve the CMC creep-rupture resistance and thermal conductivity. One such CMC system is the melt-infiltrated (MI) SiC matrix system reinforced by the Sylramic-iBN SiC fiber that is near-stoichiometric in composition and contains a thin in situ-grown BN layer on its surface.4 This SiC/SiC composite system is typically processed by taking a woven or braided fiber-preform, coating the fibers with another thin BN layer by chemical vapor infiltration (CVI), and then forming an initial SiC matrix by CVI. The remaining matrix porosity is then filled with a slurry of SiC particles. After drying the slurry-infiltrated preform, final matrix formation is by melt infiltration (MI) of liquid Si, which fills nearly all of the large pores in the structure, leaving B5% closed porosity. The primary objective of this study was to measure and analyze the off-axis in-plane tensile stress–strain behavior of five thin-walled panels with the MI Sylramic–iBN/SiC system containing two different basic fiber architectures, A and B. Four panels with architecture A, consisting of a two-dimensional (2D)-woven fabric lay-up with balanced fiber content in the 01 and 901 directions, were tensile loaded in the 01 direction (panels A1–3) and at 451 to this direction (panel A4). Another panel F. Zok—contributing editor w Author to whom correspondence should be addressed. e-mail: gmorscher@ grc.nasa.gov Manuscript No. 22879. Received March 5, 2007; approved June 2, 2007. Journal J. Am. Ceram. Soc., 90 [10] 3185–3193 (2007) DOI: 10.1111/j.1551-2916.2007.01887.x r 2007 The American Ceramic Society No claim to original US government works 3185
3186 Journal of the American Ceramic Society-Morscher et al. VoL. 90. No. 10 Table I. Individual Specimen Properties From Different Composite Panels Thickness, Total fiber Test angle to Panel 2D architecture fraction ( primary fibers (GPa)(MPa) in matrix(MPa) Al 8.7 epcm 0/90, 8 ply, 5HS, single-tow 0.39 261410 7.9 epcm 090.8 ply, 5HS, single-tow A3 3.95 epcm(2)epi 0/90, 8 ply, 5HS, double-tow 2.1 000 50 A4 8.7 epcm 45/45. 8 ply, 5HS, single-tow Bl 0/+67 Braid, 4 layer, tri-axial braid, double-tow Fraction of fibers in the axial direction=0.06. Fraction of fibers in each of the +67"bias directions =0.13 epcm, ends per centimeter: UTS, ultimate tensile strength; with architecture B(panel Bl), consisting of a 2D triaxially was symmetrically located between the two primary fiber direc- braided architecture [0/+67] with balanced fiber content in the tions that contained equal volume fractions +67 bias directions and reduced fiber content in the axial di- Tensile unload-reload hysteresis tests were performed on the rection, was tensile loaded perpendicular to the axial directio dogbone specimens using a universal testing machine(Model (i.e, 23 from the bias fibers). The primary room-temperature 8562: Instron Ltd, Canton, MA) with AE monitoring as de- properties of interest were the elastic modulus, the DFLs as scribed in Morscher.6, 7 A clip-on strain gage(25.4 mm gage, measured by two off-set methods, the UTS, and the matrix- 0.25% strain) was used to measure strain in the gage section. acking behavior as monitored by acoustic emission(AE). Ini Panels Al, A2, A3, and bl were tested in load control at tial implications are discussed for architecture design to model 4 kN/min (- 200 MPa/min depending on composite thickness) and improve the directional cracking strengths of Sic/SiC CMc This loading rate is relatively slow compared with typical mono- components with MI matrices tonic fast-fracture tests and has been determined to be a good ate for ae acquisition A Fracture Wave Detector by Digital Wave Corporation II. Experimental Procedure Englewood, CO) was used to monitor AE waveforms. Three For this study, four panels with architecture A were fabricated wide-band (b1025, Digital Wave Corporation) AE sensors were by cutting 150 mm x 225 mm plies from a 2D-woven 0/90 mounted on the specimens. Two sensors were placed above and Sylramic SiC fabric with a five-harness satin weave and bal below the gage section approximately 50 mm apart from one anced number of tow ends per centimeter(epcm, i.e., the num- another. The third sensor was placed between the other two ber of fiber tows per centimeter in the woven cloth when looking sensors in the middle of the gage section. aE waveforms on all at the weave edge-on)in the 0 and 90 directions. Each single three channels were captured simultaneously when any of the tow consisted of x800 fibers with a 10 um average diameter. three sensors was triggered, i.e., the channels were synchronized For panels Al and A4, the fabric had 8.7 single-tow epcm; but This allowed for easy separation of events. Only events that the plies were cut along the 0 and 90 directions for panel Al triggered the middle sensor were used in the analysis, i.e,onl and at 450 to the orthogonal directions for panel A4. For panels each of the a panels and two specimens from panel B were A2 and a3, the plies were cut along the oand 90 directions; b the fabric had 7.9 single-tow epcm for panel A2, and 3.95 dou tested in this way for this study (see Table D) ble-tow epcm for panel A3. For each A-type panel, eight plies were then stacked and converted to Sylramic-iBN fiber at NASA Glenn. The 2D Sylramic-iBN stacks were then sent to Orthogonal fibers GE Composites, Newark, DE, for MI SiC/SiC processing For panel Bl, the architecture was first formed by creating a four-layer 0/+67 tri-axial braid on a 50-mm-diameter tubular andre. Approximately 23% of the fibers were in the axial di- rection and 77% were in the bias direction at an angle of 67 to the axial fibers. Two as-produced Sylramic fiber tows were Tensile Axis ar architecture was then removed from the mandrel and laid flat to form a 75 mm x -150 mm rectangular preform which was converted to the Sylramic-iBN fibers at NASA and then into a Sic/sic panel with typical Mi processing at GE Tensile 150-mm-long dogbone specimens with a contoured (a gage section(12.7 mm width in grip region and 10 mm width in age region) were machined from each panel. Architecture thickness, and total fiber volume fraction for all tested speci- IS mens from the five panels are listed in Table I. For panels a A2, and A3 specimens, the testing direction was along the pri- nary or 0 direction; but for panel A4, testing was at 45 to the Tensile 0o and 90 directions of the original fabric as shown in Fig. 1(a) Axis For panel B specimens, Fig. I(b) shows that the testing direction was perpendicular to the axial or 0%fiber direction, or along the hoop"direction of the original tubular architecture. Thus, for both off-axis panels of this study, the tensile loading direction Axial Fibers(below surface) FThe Syiramic fibers of this study were originally produced ochester. NH. Both he Sylramic and Syiramic-iBN SiC fibers are currently produced at ATK COI Ceramics, Fig 1. Photographs of composite surface sh ber orientations San Diego. CA and tensile axis for(a)[+45] panel A4 and (b)[0/+67] braid panel Bl
with architecture B (panel B1), consisting of a 2D triaxially braided architecture [0/767] with balanced fiber content in the 7671 bias directions and reduced fiber content in the axial direction, was tensile loaded perpendicular to the axial direction (i.e., 231 from the bias fibers). The primary room-temperature properties of interest were the elastic modulus, the DFLS as measured by two off-set methods, the UTS, and the matrixcracking behavior as monitored by acoustic emission (AE). Initial implications are discussed for architecture design to model and improve the directional cracking strengths of SiC/SiC CMC components with MI matrices. II. Experimental Procedure For this study, four panels with architecture A were fabricated by cutting 150 mm 225 mm plies from a 2D-woven 0/90 Sylramic SiC fabricz with a five-harness satin weave and balanced number of tow ends per centimeter (epcm, i.e., the number of fiber tows per centimeter in the woven cloth when looking at the weave edge-on) in the 01 and 901 directions. Each single tow consisted of B800 fibers with a 10 mm average diameter. For panels A1 and A4, the fabric had 8.7 single-tow epcm; but the plies were cut along the 01 and 901 directions for panel A1 and at 451 to the orthogonal directions for panel A4. For panels A2 and A3, the plies were cut along the 01and 901 directions; but the fabric had 7.9 single-tow epcm for panel A2, and 3.95 double-tow epcm for panel A3. For each A-type panel, eight plies were then stacked and converted to Sylramic-iBNz fiber at NASA Glenn.5 The 2D Sylramic-iBN stacks were then sent to GE Composites, Newark, DE, for MI SiC/SiC processing.4 For panel B1, the architecture was first formed by creating a four-layer 0/767 tri-axial braid on a 50-mm-diameter tubular mandrel. Approximately 23% of the fibers were in the axial direction and 77% were in the bias direction at an angle of B671 to the axial fibers. Two as-produced Sylramic fiber tows were combined in the axial and bias directions. A 75 mm-length of the tubular architecture was then removed from the mandrel and laid flat to form a 75 mm B150 mm rectangular preform, which was converted to the Sylramic-iBN fibers at NASA and then into a SiC/SiC panel with typical MI processing at GE composites. Tensile 150-mm-long dogbone specimens with a contoured gage section (12.7 mm width in grip region and 10 mm width in gage region) were machined from each panel. Architecture, thickness, and total fiber volume fraction for all tested specimens from the five panels are listed in Table I. For panels A1, A2, and A3 specimens, the testing direction was along the primary or 01 direction; but for panel A4, testing was at 451 to the 01 and 901 directions of the original fabric as shown in Fig. 1(a). For panel B specimens, Fig. 1(b) shows that the testing direction was perpendicular to the axial or 01 fiber direction, or along the ‘‘hoop’’ direction of the original tubular architecture. Thus, for both off-axis panels of this study, the tensile loading direction was symmetrically located between the two primary fiber directions that contained equal volume fractions. Tensile unload–reload hysteresis tests were performed on the dogbone specimens using a universal testing machine (Model 8562; Instron Ltd., Canton, MA) with AE monitoring as described in Morscher.6,7 A clip-on strain gage (25.4 mm gage, 0.25% strain) was used to measure strain in the gage section. Panels A1, A2, A3, and B1 were tested in load control at 4 kN/min (B200 MPa/min depending on composite thickness). This loading rate is relatively slow compared with typical monotonic fast-fracture tests and has been determined to be a good rate for AE acquisition. A Fracture Wave Detector by Digital Wave Corporation (Englewood, CO) was used to monitor AE waveforms. Three wide-band (B1025, Digital Wave Corporation) AE sensors were mounted on the specimens. Two sensors were placed above and below the gage section approximately 50 mm apart from one another. The third sensor was placed between the other two sensors in the middle of the gage section. AE waveforms on all three channels were captured simultaneously when any of the three sensors was triggered, i.e., the channels were synchronized. This allowed for easy separation of events. Only events that triggered the middle sensor were used in the analysis, i.e., only events that occurred in the gage section. One specimen from each of the A panels and two specimens from panel B were tested in this way for this study (see Table I). Table I. Individual Specimen Properties From Different Composite Panels Panel 2D architecture Thickness, mm Total fiber fraction (f) Test angle to primary fibers E (GPa) UTS (MPa) Residual stress in matrix (MPa) A1 8.7 epcm 0/90,8 ply, 5HS, single-tow 2.3 0.39 01 261 410 60 A2 7.9 epcm 0/90, 8 ply, 5HS, single-tow 2.0 0.39 01 250 463 50 A3 3.95 epcm (2) epi 0/90, 8 ply, 5HS, double-tow 2.1 0.39 01 202 444 50 A4 8.7 epcm 45/45,8 ply, 5HS, single-tow 2.4 0.36 451 233 242 40 B1 0/167 Braid, 4 layer, tri-axial braid, double-tow 1.8 0.32w 231 240 338 60 1.8 0.32 231 260 366 60 w Fraction of fibers in the axial direction 5 0.06. Fraction of fibers in each of the 1671 bias directions 5 0.13. epcm, ends per centimeter; UTS, ultimate tensile strength; 2D, two-dimensional. Fig. 1. Photographs of composite surface showing fiber orientations and tensile axis for (a) [745] panel A4 and (b) [0/767] braid panel B1. z The Sylramic fibers of this study were originally produced by Dow Corning, Midland, MI, and were woven into fabric at Albany International Techniweave, Rochester, NH. Both the Sylramic and Sylramic-iBN SiC fibers are currently produced at ATK COI Ceramics, San Diego, CA. 3186 Journal of the American Ceramic Society—Morscher et al. Vol. 90, No. 10
October 2007 Tensile Mechanical Properties of Ceramic Matrix Composites 500 7.9 epcm [Orgo From Table l, it can be seen that in-plane E values are similar 8.7 epcm [o/90 for all panels with the exception of the double-tow 5HS-woven composite(panel A3). This suggests that for the Sic/SiC system 350[0/467] double studied, the in-plane elastic modulus is not strongly dependent 3 95 epem [090 on 2D fiber architectures or on tensile-loading direction. For in- 300 f=0.39 plane UTS, the 2D-woven 0/90 panels aligned in the primary fiber axes are of course the strongest because they were test 8.7 epem [+45] parallel to the fiber direction. Nevertheless, the braided panel with over three-quarters of the fibers oriented 23 from the loading-axis displayed a high in-plane UTS; however, the [+45 panel displayed a relatively low UTS. But perhaps the most 50 striking mechanical results were the high in-plane DFLs values a3 for both the braided and the [+] panel 040.506 These high DFLS results correlate(Fig. 4 and Table In) with higher stress ranges over which high-energy ae events were re- corded in the off-axis panels. In Fig. 4. the aE activity versus Fig. 2. Stress-strain curves of off-axis and orthogonally aligned com- applied CMC stress is plotted as normalized cumulative AE en- psites with the hysteresis loop removed for clarity. An example of the ergy, which is the cumulative AE energy of each AE event up ysteresis loops for 0/+67 braid (I)is shown in the inset. a given event divided by the total AE energy of all the eve o For the MI siC/SiC system, it has been shown that this type of plot represents a good relative distribution of transverse or direction in order to ns were cut and polished along the TTMC. In addition, multiplying the final matrix crack density measured from polished sections after failure, by the normalized The polished specimens were plasma etched(CFa at 500 cumulative AE energy is a very good estimate of the actual for 30 min) in order to enhance the matrix cracks in the CvI sic stress-dependent matrix crack density versus applied stress. The of the matrix. matrix cracks were counted over lengths of neasured matrix crack densities at failure are shown in Table ll approximately 10 mm in order to obtain a matrix crack density. and the estimated matrix crack density with stress in Fig. 5. The difference in the shape of the matrix crack distribution between e double-tow and single-tow 2D-woven composites is hypoth esized to be due to the greater concentration and longer lengt II Results of back-to-back 90 minicomposites (see Section IV). Two The room-temperature tensile stress-strain curves for the off- axis-oriented SiC/SiC panels A4 and Bl and for the orthogonal ifferent off-axis-loaded specimens. The [+45] specimen has a oriented 2D-woven SiC/SiC panels Al-3 are shown in Fig. 2 ery steep curve, whereas the braided specimens were similar in Hysteresis loops used for residual stress determination have shape to the single-tow 2D composites. Neither of the off-axis been eliminated from these curves except for a braided panel composites appears to have reached matrix crack saturation be BI specimen, which is given as an example in Fig. 2. The values ause a decrease in the rate of ae activity was never achieved at of initial elastic modulus e. UTs. and residual stress on matrix her stresses. It is also interesting to note that at a composit stress of 200 MPa, the off-axis-tested braided specimens had are listed in Table I and are consistent with other composite little or no cracks: however, the double-tow 2D-woven compo panels fabricated with the same fiber types and fiber architec tures , As shown in Fig. 3, the offset strain construct method ites had -5 TTMC/mm. was also used to determine DFLs. It consists of drawing a line For this study, several aE criteria were used to evaluate stress levels near initiation of matrix cracking as these values ith the same slope as the initial elastic modulus, but offset by some amount of positive strain, where the dFls would be de- t upper design limits beyond which CMC mod- termined by the intersection of that curve with the stress-strain and axial thermal conductivity irreversibly decrease and the curve. Typical offset-strain values used are 0.005% and bility for adverse environmental effects for Sic/SiC com- 0.002%. 0 As indicated in Table II. both values were dete ites exists. These stress levels. which are indicated in Fig. 6 n this study and Table Il, are associated with(I)the first AE event, (2) the first loud ae event. which is defined as an ae event with an energy of at least one-tenth the highest energy event not corresponding to final failure of the composite, and ( 3)the effective AE onset stress, which is determined by extrapolation of the steep slope of the normalized cumulative AE energy with increasing stress back down to the zero energy axis(see Fig. 6 For the braided specimens, there was an initial increase in slope (AE activity) followed by a further increase in slope, which con- tinued at the higher rate until failure. For AE onset stress, the Stress-strain curve initial moderate AE slope(Fig. 6) was used because it was con- 0. 005% offset firmed that high-energy AE events were occurring in this regime It is evident in Table II that except for the first AE event, the 0. 002% offset other two stress measures for matrix cracking are significantl higher for the off-axis specimens than for the on-axis specimens. Underlying mechanisms and technical significance for these stress levels will be discussed in the following sections Strain. The effect of architecture and orientation on matrix cracking(or Fig3. Deviation from linearity-stress construction for a braided spec- DFLS) and UTs are key properties that need to be understood imen. Note, only the low strain portion of the stress-strain curve is for applications using MI SiC/SiC composites under multiaxial plotted ress states. In the following sections, the results shown in Figs
The tested specimens were cut and polished along the loading direction in order to measure transverse matrix cracks optically. The polished specimens were plasma etched (CF4 at 500 Watts for 30 min) in order to enhance the matrix cracks in the CVI SiC of the matrix. Matrix cracks were counted over lengths of approximately 10 mm in order to obtain a matrix crack density.7 III. Results The room-temperature tensile stress–strain curves for the offaxis-oriented SiC/SiC panels A4 and B1 and for the orthogonaloriented 2D-woven SiC/SiC panels A1–3 are shown in Fig. 2. Hysteresis loops used for residual stress determination have been eliminated from these curves except for a braided panel B1 specimen, which is given as an example in Fig. 2. The values of initial elastic modulus E, UTS, and residual stress on matrix are listed in Table I and are consistent with other composite panels fabricated with the same fiber types and fiber architectures.4,7 As shown in Fig. 3, the offset strain construct method8 was also used to determine DFLS. It consists of drawing a line with the same slope as the initial elastic modulus, but offset by some amount of positive strain, where the DFLS would be determined by the intersection of that curve with the stress–strain curve. Typical offset-strain values used are 0.005%9 and 0.002%.10 As indicated in Table II, both values were determined in this study. From Table I, it can be seen that in-plane E values are similar for all panels with the exception of the double-tow 5HS-woven composite (panel A3). This suggests that for the SiC/SiC system studied, the in-plane elastic modulus is not strongly dependent on 2D fiber architectures or on tensile-loading direction. For inplane UTS, the 2D-woven 0/90 panels aligned in the primary fiber axes are of course the strongest because they were tested parallel to the fiber direction. Nevertheless, the braided panel with over three-quarters of the fibers oriented 231 from the loading-axis displayed a high in-plane UTS; however, the [745] panel displayed a relatively low UTS. But perhaps the most striking mechanical results were the high in-plane DFLS values for both the braided and the [745] panels. These high DFLS results correlate (Fig. 4 and Table II) with higher stress ranges over which high-energy AE events were recorded in the off-axis panels. In Fig. 4, the AE activity versus applied CMC stress is plotted as normalized cumulative AE energy, which is the cumulative AE energy of each AE event up to a given event divided by the total AE energy of all the events. For the MI SiC/SiC system, it has been shown7 that this type of plot represents a good relative distribution of transverse or TTMC. In addition, multiplying the final matrix crack density, measured from polished sections after failure, by the normalized cumulative AE energy is a very good estimate of the actual stress-dependent matrix crack density versus applied stress.7 The measured matrix crack densities at failure are shown in Table II and the estimated matrix crack density with stress in Fig. 5. The difference in the shape of the matrix crack distribution between the double-tow and single-tow 2D-woven composites is hypothesized to be due to the greater concentration and longer lengths of back-to-back 901 minicomposites (see Section IV). Two different matrix crack distributions are evident for the two different off-axis-loaded specimens. The [745] specimen has a very steep curve, whereas the braided specimens were similar in shape to the single-tow 2D composites. Neither of the off-axis composites appears to have reached matrix crack saturation because a decrease in the rate of AE activity was never achieved at higher stresses. It is also interesting to note that at a composite stress of 200 MPa, the off-axis-tested braided specimens had little or no cracks; however, the double-tow 2D-woven composites had B5 TTMC/mm. For this study, several AE criteria were used to evaluate key stress levels near initiation of matrix cracking as these values typically represent upper design limits beyond which CMC moduli and axial thermal conductivity irreversibly decrease and the possibility for adverse environmental effects for SiC/SiC composites exists.11 These stress levels, which are indicated in Fig. 6 and Table II, are associated with (1) the first AE event, (2) the first loud AE event, which is defined as an AE event with an energy of at least one-tenth the highest energy event not corresponding to final failure of the composite, and (3) the effective AE onset stress,7 which is determined by extrapolation of the steep slope of the normalized cumulative AE energy with increasing stress back down to the zero energy axis (see Fig. 6). For the braided specimens, there was an initial increase in slope (AE activity) followed by a further increase in slope, which continued at the higher rate until failure. For AE onset stress, the initial moderate AE slope (Fig. 6) was used because it was con- firmed that high-energy AE events were occurring in this regime. It is evident in Table II that except for the first AE event, the other two stress measures for matrix cracking are significantly higher for the off-axis specimens than for the on-axis specimens. Underlying mechanisms and technical significance for these stress levels will be discussed in the following sections. IV. Analysis and Discussion The effect of architecture and orientation on matrix cracking (or DFLS) and UTS are key properties that need to be understood for applications using MI SiC/SiC composites under multiaxial stress states. In the following sections, the results shown in Figs. Fig. 2. Stress–strain curves of off-axis and orthogonally aligned composites with the hysteresis loop removed for clarity. An example of the hysteresis loops for 0/767 braid (1) is shown in the inset. 0 50 100 150 200 250 0 0.02 0.04 0.06 0.08 0.1 Strain, % Stress, MPa Stress-strain curve 0.002% offset 0.005% offset Fig. 3. Deviation from linearity-stress construction for a braided specimen. Note, only the low strain portion of the stress–strain curve is plotted. October 2007 Tensile Mechanical Properties of Ceramic Matrix Composites 3187
3188 Journal of the American Ceramic Society-Morscher et al. VoL. 90. No. 10 Table Il. Matrix Cracking Properties of Composite Specimens DFLS,0002% DFLS.0.005% First AE First loud ae AE onset before AE onset stress- Final matrix crack Panel offset(MPa) offset(MPa stress(MPa stress(MPa) stress(MPa otal #f loud events density(#/mm) Orthogonal oriented composites 147 174 190 4-141 l30 173 182 1-141 135 157 2-134 9.0 Off-axis oriented composites 10 4.0 4.9 195 231 193 210 3-134 7. 9 epcm [±45 7.9 epcm [/9o 8.7 epcm 0/901 8.7 epcm double-tot [0/90] 3.95 epcm Double Tow [0/901 p 2 [O/+67] braid 8.7 epcm, 00 Fig 4. Acoustic emission activity versus composite stress Fig. 5. Estimated matrix crack density with stress based on acoustic 2 and 4 and in Table I and Il will be mechanistically analyzed and compared with other data in the literature to better ur dividual 90" minicomposites and the size of two 90% minicom- tand their scientific and technical significan osites that happen to be adjacent to one another. Whenever this back-to-back tow circumstance exists. the effective width of an unbridged tunnel crack would be approximately twice the (1) Matrix Cracking in 0/90 2D-Woren Composites Tested in back-to-back individual 90 minicomposites is much more com- osite. This characteristic of the o direction non in the double-tow woven 3.95 epcm composite panel com- A variety of microscopic studies of [o/90]2D-woven CMC spec- pared with single-tow 7.9 epcm woven panels. Not only are there mens have shown that at lower stresses. initial transverse matrix nore regions of multiple 90@ minicomposites but also the length cracks are usually either"tunnel"microcracks, which occur in of these regions(distance the tow is woven over four tows before the 90 minicomposites oriented perpendicular to the 00loadir axis, and or nonsteac ady-state microcracks that are partially it is woven under the fifth tow in the five-harness satin archi- bridged due to sufficient fiber traction in the matrix crack tecture)would be approximately twice the length of single-tow to stop matrix crack propagation through-thickness. At woven composites for the five- harness satin weave because the higher stresses, these microcracks propagate through-thickness epcm of the double-tow CMC was one-half that of the single or link up with other microcracks to form TTMC over a range tow CMC. As a result, as shown in Fig 4, the double-tow woven AE methodologies have been successfully used not only uantify but also to understand and model the occurrence and tress-strain dependence of microcrack and TTMc behavior. Initial low-energy events generally correspond to tunnel micro crack formation in 90 minicomposites perpendicular to the loading direction. High-energy events, those in the upper loga- rithmic decade of energy, correspond to either large microcracks 1st Loud AE event raid &[0/90] tribution for TTMC is controlled by(I) the size distribution of 1st AE even 90 minicomposites perpendicular to the load-bearing 0 mini 5 0.31 Braid &(0r-9o) composites and (2)the in situ stress in the region of the com- 0 posite outside the load-bearing 0 minicomposite, i.e., the portion of the composite composed of 90 minicomposites and the MI matrix. For panels fabricated from the random lay-up of standard single-tow-woven fabric plies, the size distribution of 90 minicomposites can crudely vary between the size of Stress. MPa ' For the CMC of this study, a minicom of a single multi fiber tow, the CvI Fig. 6. Acoustic emission(AE) events and onset stress determination. The arrows below the x-axis indicate ae onset stress
2 and 4 and in Table I and II will be mechanistically analyzed and compared with other data in the literature to better understand their scientific and technical significance. (1) Matrix Cracking in 0/90 2D-Woven Composites Tested in the 01 Direction A variety of microscopic studies of [0/90] 2D-woven CMC specimens have shown that at lower stresses, initial transverse matrix cracks are usually either ‘‘tunnel’’ microcracks,12 which occur in the 901 minicompositesy oriented perpendicular to the 01 loading axis, and/or nonsteady-state microcracks that are partially bridged due to sufficient fiber traction in the matrix crack wake to stop matrix crack propagation through-thickness. At higher stresses, these microcracks propagate through-thickness or link up with other microcracks to form TTMC over a range of stress levels. AE methodologies have been successfully used not only to quantify but also to understand and model the occurrence and stress–strain dependence of microcrack and TTMC behavior. Initial low-energy events generally correspond to tunnel microcrack formation in 901 minicomposites perpendicular to the loading direction. High-energy events, those in the upper logarithmic decade of energy, correspond to either large microcracks and/or TTMC.13 For 2D-woven 0/90 MI SiC/SiC panels tested in the 01 direction, it has been demonstrated that the stress distribution for TTMC is controlled by (1) the size distribution of 901 minicomposites perpendicular to the load-bearing 01 minicomposites and (2) the in situ stress in the region of the composite outside the load-bearing 01 minicomposite, i.e., the portion of the composite composed of 901 minicomposites and the MI matrix.7,14 For panels fabricated from the random lay-up of standard single-tow-woven fabric plies, the size distribution of 901 minicomposites can crudely vary between the size of individual 901 minicomposites and the size of two 901 minicomposites that happen to be adjacent to one another. Whenever this back-to-back tow circumstance exists, the effective width of an unbridged tunnel crack would be approximately twice the crack width of a single minicomposite. This characteristic of back-to-back individual 901 minicomposites is much more common in the double-tow woven 3.95 epcm composite panel compared with single-tow 7.9 epcm woven panels. Not only are there more regions of multiple 901minicomposites but also the length of these regions (distance the tow is woven over four tows before it is woven under the fifth tow in the five-harness satin architecture) would be approximately twice the length of single-tow woven composites for the five-harness satin weave because the epcm of the double-tow CMC was one-half that of the singletow CMC. As a result, as shown in Fig. 4, the double-tow woven Table II. Matrix Cracking Properties of Composite Specimens Panel DFLS, 0.002% offset (MPa) DFLS, 0.005% offset (MPa) First AE stress (MPa) First loud AE stress (MPa) AE onset stress (MPa) No. of loud events before AE onset stress— total # loud events Final matrix crack density (#/mm) Orthogonal oriented composites A1 147 174 132 170 190 4–141 — A2 130 173 100 159 182 1–141 10.3 A3 135 176 128 138 157 2–134 9.0 Off-axis oriented composites A4 210 225 56 197 220 1–65 4.0 B1 232 259 83 187 215 2–95 4.9 195 231 135 193 210 3–134 — AE, acoustic emmision. 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 100 200 300 400 500 Stress, MPa Norm Cum AE Energy 7.9 epcm [0/90] [0/+67] braid 8.7 epcm, [+45] 3.95 epcm double-tow [0/90] 8.7 epcm [0/90] Fig. 4. Acoustic emission activity versus composite stress. 0 2 4 6 8 10 12 0 100 200 300 400 500 Composite Stress, MPa Estimated Crack Density, #/mm 7.9 epcm; [0/90] 3.95 epcm Double Tow [0/90] 8.7 epcm [0/90] [0/+67] braid 8.7 epcm, [+45] Fig. 5. Estimated matrix crack density with stress based on acoustic emission. 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 100 200 300 400 500 Stress, MPa Norm Cum AE Energy 8.7 epcm [0/90] 1st AE event Braid & [0/90] 1st Loud AE event Braid & [0/90] [0/+67] braid Fig. 6. Acoustic emission (AE) events and onset stress determination. The arrows below the x-axis indicate AE onset stress. y For the CMC of this study, a minicomposite consists of a single multi fiber tow, the CVI BN interphase coating, and the initial CVI SiC matrix coating associated with tow. 3188 Journal of the American Ceramic Society—Morscher et al. Vol. 90, No. 10
October 2007 Tensile Mechanical Properties of Ceramic Matrix Composites 3189 orthogonal this study D[o+67] double tow braid (hoop direction)-this study Fiber Fraction in Loading Direction emission(AE) matrix cracking(onset)stress for two- osites with fibers tows oriented perpendicular to the loading direction. panel displays the lowest onset stress and the smallest stress dis- tribution for TTMC in the on-axis panels In an effort to model architectural effects on the onset of TTMC in 2D-woven 0/90 SiC/SiC composites tested in the 0o direction, a previous study has examined the important micro- tructural factors affecting the internal stresses on the 90 mini composites since these appear to be the most critical flaws within the matrix that eventually cause TTMC. Two important facto 1 mm were identified: (1)a built-in residual stress on the matrix and (2) the applied composite stress as modified by the load shared by the 0%minicomposites. As such, one can estimate the stress on the matrix region containing the 90 minicomposites(=mini- matrix) by the following simple rule of mixtures relationship (oc +Oth)Ec-fmini em ominimatrix 1) Fig 8. Micrograph of a two-dimensional Ec posite after room temperature tensile failure urce.Most notably, there was also a very na Here,σ is the composite stress, Oth is the residual tribution for TTMC(Fig. 4), which would corres ess in the matrix Ec is the measured composite ela population(high Weibull modulus)of flaws and tic modulus from the oe curve (Table I): mini is the volume that propagate through-thickness at and slightly above the ma ically -2 times the fiber content in the 0 direction for the com-(panel A4) displayed a higher onset stress for TTMC than its can be reduced by increasing a compressive residual stress on the cracking stress and the narrow stress-distribution for matrix matrix and/or by increasing mini. As shown in Table l, one ben acking for this system is primarily due to a lack of minicom- efit of the mi sic/SiC system is a compressive residual stress on posites perpendicular to the direction of stress so that higher the as-fabricated matrix, but this stress did not vary much with tresses are required for tunnel crack propagation at an ang the 2D architectures of this study. The exact source of this stress 45 to the loading axis. There is also the possibility that the is not completely understood, but probably can urce of matrix cracking are surface flaws or pores in the ma silicon content in the matrix and to the composite fabrication trix and not the +45-oriented minicomposites Matrix cracks conditions. On the other hand, fiber content in the or primary were observed to be perpendicular to the loading direction as MI SiC/SiC system, which in turn should increase the composite stress for the onset of ttmc. that this latter mechanism can indeed be utilized is seen in Fig. 7, which plots as open circles the Ae onset cracking strength as a function of primary for the (3) Matrix Cracking in 01+67 2D-Braided Composite Tested 0o-loaded 0/90 MI SiC/SiC composites of this study and for a in the Hoop Direction For the triaxially braided specimen, approximately 77% of the fibers or a total fiber fraction of 0. 26 were oriented 23 from the loading axis. This fraction is effectively higher than that of the (2) Matrix Cracking in 0/90 2D-Woren Composites Tested 2D-woven composites loaded in the primary fiber direction, a rhe45° Direction thus appears to be one cause for the higher matrix cracking The stress-distribution for matrix cracking in the [+45]off-axis stress for the braided composite. As described above for the 0/90 panel A4 differs considerably from those of the [o/90] compos- composites, matrix cracks emanate from the region outside the ites tested in a 0 or fiber direction first ae events show that load-bearing minicomposites, e.g., from the minicomposites ori some small microcracks were formed at low stresses in the [t45 ented 90 to the loading axis. The same type of analysis can be composite (Table ID), the source of which has not been deter- applied here to the braided composites to determine whether the mined. There was considerable porosity in the matrix region stress ranges in the TTMC flaw-source regions(axial tow mini- tween the outer two plies ( Fig 8), which could have been one composites, see Fig. 1)of the matrix are similar to standard
panel displays the lowest onset stress and the smallest stress distribution for TTMC in the on-axis panels. In an effort to model architectural effects on the onset of TTMC in 2D-woven 0/90 SiC/SiC composites tested in the 01 direction, a previous study7 has examined the important microstructural factors affecting the internal stresses on the 901 minicomposites since these appear to be the most critical flaws within the matrix that eventually cause TTMC. Two important factors were identified: (1) a built-in residual stress on the matrix and (2) the applied composite stress as modified by the load shared by the 01 minicomposites. As such, one can estimate the stress on the matrix region containing the 90 minicomposites ( 5 minimatrix) by the following simple rule of mixtures relationship7 : sminimatrix ¼ ðsc þ sthÞ Ec Ec fminiEmini 1 fmini (1) Here, sc is the applied composite stress, sth is the residual stress in the matrix (Table I); Ec is the measured composite elastic modulus from the se curve (Table I); fmini is the volume fraction of the 01 minicomposites in the loading direction (typically B2 times the fiber content in the 01 direction for the composites of this study); and Emini is the effective modulus of these 01 minicomposites. Thus, the stress on the 90 minicomposites can be reduced by increasing a compressive residual stress on the matrix and/or by increasing fmini. As shown in Table I, one benefit of the MI SiC/SiC system is a compressive residual stress on the as-fabricated matrix, but this stress did not vary much with the 2D architectures of this study. The exact source of this stress is not completely understood, but probably can be related to the silicon content in the matrix and to the composite fabrication conditions. On the other hand, fiber content in the 01 or primary fiber direction, fprimary, can be increased to a large degree for the MI SiC/SiC system, which in turn should increase the composite stress for the onset of TTMC. That this latter mechanism can indeed be utilized is seen in Fig. 7, which plots as open circles the AE onset cracking strength as a function of fprimary for the 01-loaded 0/90 MI SiC/SiC composites of this study and for a previous study.7 (2) Matrix Cracking in 0/90 2D-Woven Composites Tested in the 451 Direction The stress-distribution for matrix cracking in the [745] off-axis panel A4 differs considerably from those of the [0/90] composites tested in a 01 or fiber direction. First AE events show that some small microcracks were formed at low stresses in the [745] composite (Table II), the source of which has not been determined. There was considerable porosity in the matrix region between the outer two plies (Fig. 8), which could have been one source. Most notably, there was also a very narrow stress-distribution for TTMC (Fig. 4), which would correspond to a large population (high Weibull modulus) of flaws and/or microcracks that propagate through-thickness at and slightly above the matrix cracking stress. The important fact is that this composite (panel A4) displayed a higher onset stress for TTMC than its theoretically equivalent 0/90 composite (panel A1) loaded along its primary fiber axis. It is suggested that the higher matrix cracking stress and the narrow stress-distribution for matrix cracking for this system is primarily due to a lack of minicomposites perpendicular to the direction of stress so that higher stresses are required for tunnel crack propagation at an angle 451 to the loading axis. There is also the possibility that the source of matrix cracking are surface flaws or pores in the matrix and not the 7451-oriented minicomposites. Matrix cracks were observed to be perpendicular to the loading direction as has been reported before.2 (3) Matrix Cracking in 0/767 2D-Braided Composite Tested in the Hoop Direction For the triaxially braided specimen, approximately 77% of the fibers, or a total fiber fraction of 0.26, were oriented 231 from the loading axis. This fraction is effectively higher than that of the 2D-woven composites loaded in the primary fiber direction, and thus appears to be one cause for the higher matrix cracking stress for the braided composite. As described above for the 0/90 composites, matrix cracks emanate from the region outside the load-bearing minicomposites, e.g., from the minicomposites oriented 901 to the loading axis. The same type of analysis can be applied here to the braided composites to determine whether the stress ranges in the TTMC flaw-source regions (axial tow minicomposites, see Fig. 1) of the matrix are similar to standard 100 120 140 160 180 200 220 240 0.1 0.15 0.2 0.25 0.3 Fiber Fraction in Loading Direction AE Matrix Cracking Stress, MPa 2D orthogonal [7] 2D orthogonal - this study 2D double tow orthogonal [7] 2D double tow orthogonal this study [0/+67] double tow braid (hoop direction) - this study Fig. 7. Effect of fiber fraction in the loading direction on acoustic emission (AE) matrix cracking (onset) stress for two-dimensional composites with fibers tows oriented perpendicular to the loading direction. Fig. 8. Micrograph of a two-dimensional woven [745]-oriented composite after room temperature tensile failure. October 2007 Tensile Mechanical Properties of Ceramic Matrix Composites 3189