M.B. Ruggles-Wrenm, N.R. Szymczak/Composites: Part A 39 (2008)1829-1837 835 cross-ply composites, longitudinal compressive damage and frac from constant stress-rate test data suggest that delayed failure of ture typically involve axial splitting of the matrix, buckling of the the N720/A CMC at 1200C in steam is governed by the power fibers, and kink banding or shear banding [48-51]. Lankford [48 law type subcritical crack growth. These results point to the envi- reported that for a 0/90 crossply CMC, compressive failure initiated ronmentally assisted subcritical crack growth as the mechanism ith nucleation of axial cracks between adjacent fibers in the 90 behind the degraded compressive creep performance in steam. plies. These cracks grow subcritically to gradually form shear zones, which induce 0 ply flexure and cause buckling and kinking 3.4. Composite microstructure of the 0 fibers, leading to local fiber fracture and subsequent com- posite failure. For porous-matrix composites, the matrix is excep- Fracture surfaces of the N720/ Specimens tested in compression tionally weak and the fibers bear most of the load Once the 0 with the stress-rate of 25 MPa s at 1200C in air and in steam are bundles buckle, profuse matrix microcracking takes place, result- shown in Fig 8a and b, respectively. Brushy fracture surfaces indic ing in the loss of fiber stabilization and consequently the loss of ative of fibrous fracture are produced in both environments. How the composite 's load-bearing capacity. Composite failure is then ever, the specimen tested in air exhibits a considerably longer ached damage zone (46 mm) than the specimen tested in steam Considering that the shear d bending fracture of fiber bun- 22 mm). As seen in Fig. 9a, the fracture surface topography of dles governs the fracture of the composite in compression, and the N720/A specimen tested in air becomes less serrated at the recalling that Choi and Bansal [29] successfully used the power- stress rate of 0.0025 MPa/s: a shorter damage zone of w28 mm is w slow crack growth model to account for the rate dependency also observed. The contrast between the fracture surfaces produced of shear strength, we conjecture that this model may be used to de at the different stress rates in steam( Figs. 8b and 9b)is more strik scribe the dependence of the compressive strength on stress-rate. ing. The fracture surface produced at 0.0025 MPa s in steam has vir- Following the procedure in [36, 37] the experimental data obtained tually no regions of fibrous fracture, coordinated fiber failure is in monotonic compression tests in steam were plotted as log(com- prevalent. A much shorter damage zone of 12 mm is produced pressive strength) vs log(applied stress-rate). Despite the limited Fracture surface produced in compression test to failure con- number of compression tests conducted at different stress rates in ducted on N720/A specimen that had achieved compressive creep steam, the fit of the data to Eq (4)was good with the coefficient of run-out at-60 MPa in air(see Fig 10a)is similar to that obtained correlation in regression of 0.989. This implies that the delayed fail- in compression test conducted on as-processed material in air at ure of the N720/A composite in compression at 1200Cin steam can 0.0025 MPa/s Uncorrelated fiber fracture and a somewhat smaller be described by the empirical power-law in the formof Eq (1). Hence damage zone (32 mm)are observed. Conversely, the fracture sur- the scg parameters n= 11 and d=88.3 were determined by a linear face of the specimen tested in creep at -60 MPa in steam(see regression analysis from the slope and intercept, respectively. Using Fig 10b)does not have a"brushy " appearance and shows a shorter these parameters and Eg (6). the compressive creep lifetimes were damage zone of approximately 16 mm. Fracture surfaces domi- calculated for a range of applied compressive stress levels. nated by coordinated fiber failure and shorter damage zones pro- The predicted creep lifetimes in steam are shown in Fig. 5 as a duced in tests of longer duration conducted in steam (L.e. solid line. A good agreement between the predictions and the compression test at 0.0025 MPa/s and creep tests at-40 and perimental results indicates that in both compression creep tests -60 MPa) suggest that matrix has densified. Yet the denser matrix and in constant stress-rate compression tests at 1200C in steam, did not serve to improve the mechanical performance in compres the slow crack growth is indeed the governing failure mechanism sion as may be expected. On the contrary the most severe degra- for N720/A CMC. The dependence of compressive strength on load dation of the compressive properties and performance was ing rate together with the reasonable prediction of creep lifetimes observed in tests of longer duration conducted in steam. 181 18 mm Fig 9. Fracture surfaces of N720/A specimens tested in compression with the stress-rate of 0.0025 MPa/s at 1200C:(a) in air and (b)in steam.cross-ply composites, longitudinal compressive damage and frac￾ture typically involve axial splitting of the matrix, buckling of the fibers, and kink banding or shear banding [48–51]. Lankford [48] reported that for a 0/90 crossply CMC, compressive failure initiated with nucleation of axial cracks between adjacent fibers in the 90 plies. These cracks grow subcritically to gradually form shear zones, which induce 0 ply flexure and cause buckling and kinking of the 0 fibers, leading to local fiber fracture and subsequent com￾posite failure. For porous-matrix composites, the matrix is excep￾tionally weak and the fibers bear most of the load. Once the 0 bundles buckle, profuse matrix microcracking takes place, result￾ing in the loss of fiber stabilization and consequently the loss of the composite’s load-bearing capacity. Composite failure is then reached. Considering that the shearing and bending fracture of fiber bun￾dles governs the fracture of the composite in compression, and recalling that Choi and Bansal [29] successfully used the power￾law slow crack growth model to account for the rate dependency of shear strength, we conjecture that this model may be used to de￾scribe the dependence of the compressive strength on stress-rate. Following the procedure in [36,37] the experimental data obtained in monotonic compression tests in steam were plotted as log (com￾pressive strength) vs log (applied stress-rate). Despite the limited number of compression tests conducted at different stress rates in steam, the fit of the data to Eq. (4) was good with the coefficient of correlation in regression of 0.989. This implies that the delayed fail￾ure of the N720/A composite in compression at 1200 C in steam can be described by the empirical power-law in the form of Eq.(1). Hence the SCG parameters n = 11 and D = 88.3 were determined by a linear regression analysis from the slope and intercept, respectively. Using these parameters and Eq. (6), the compressive creep lifetimes were calculated for a range of applied compressive stress levels. The predicted creep lifetimes in steam are shown in Fig. 5 as a solid line. A good agreement between the predictions and the experimental results indicates that in both compression creep tests and in constant stress-rate compression tests at 1200 C in steam, the slow crack growth is indeed the governing failure mechanism for N720/A CMC. The dependence of compressive strength on load￾ing rate together with the reasonable prediction of creep lifetimes from constant stress-rate test data suggest that delayed failure of the N720/A CMC at 1200 C in steam is governed by the power￾law type subcritical crack growth. These results point to the envi￾ronmentally assisted subcritical crack growth as the mechanism behind the degraded compressive creep performance in steam. 3.4. Composite microstructure Fracture surfaces of the N720/A specimens tested in compression with the stress-rate of 25 MPa/s at 1200 C in air and in steam are shown in Fig. 8a and b, respectively. Brushy fracture surfaces indic￾ative of fibrous fracture are produced in both environments. How￾ever, the specimen tested in air exhibits a considerably longer damage zone (46 mm) than the specimen tested in steam (22 mm). As seen in Fig. 9a, the fracture surface topography of the N720/A specimen tested in air becomes less serrated at the stress rate of 0.0025 MPa/s; a shorter damage zone of 28 mm is also observed. The contrast between the fracture surfaces produced at the different stress rates in steam (Figs. 8b and 9b) is more strik￾ing. The fracture surface produced at 0.0025 MPa/s in steam has vir￾tually no regions of fibrous fracture, coordinated fiber failure is prevalent. A much shorter damage zone of 12 mm is produced. Fracture surface produced in compression test to failure con￾ducted on N720/A specimen that had achieved compressive creep run-out at 60 MPa in air (see Fig. 10a) is similar to that obtained in compression test conducted on as-processed material in air at 0.0025 MPa/s. Uncorrelated fiber fracture and a somewhat smaller damage zone (32 mm) are observed. Conversely, the fracture sur￾face of the specimen tested in creep at 60 MPa in steam (see Fig. 10b) does not have a ‘‘brushy” appearance and shows a shorter damage zone of approximately 16 mm. Fracture surfaces domi￾nated by coordinated fiber failure and shorter damage zones pro￾duced in tests of longer duration conducted in steam (i. e. compression test at 0.0025 MPa/s and creep tests at 40 and 60 MPa) suggest that matrix has densified. Yet the denser matrix did not serve to improve the mechanical performance in compres￾sion as may be expected. On the contrary, the most severe degra￾dation of the compressive properties and performance was observed in tests of longer duration conducted in steam. Fig. 9. Fracture surfaces of N720/A specimens tested in compression with the stress-rate of 0.0025 MPa/s at 1200 C: (a) in air and (b) in steam. M.B. Ruggles-Wrenn, N.R. Szymczak / Composites: Part A 39 (2008) 1829–1837 1835