Joumal J Am Ceram Soc, 80[7]1812-20(1997) High-Temperature Transverse fracture Toughness of Nicalon-Fiber-Reinforced CAS-Il Glass-Ceramic Matrix Composite Ramazan Kahama Department of Chemical Engineering, King Fahd University of Petroleum and minerals Dhahran 31261, Saudi arabia John F. Mandell and Max C Deibert Department of Chemical Engineering, Montana State University, Bozeman, Montana Cracking parallel to the fibers in off-axis plies is usually the reinforced glass-ceramics, producing the carbon-rich inter- ng process has been associated with the(transverse)frac- bond strong enough for load transfer, yet weafer-matrix) ture toughness, defined by the critical strain energy release debond readily and allow fiber bridging during crack propaga ate, Gye. The measurement of Ge provides basic informa- tion(perpendicular to the fibers )resulting in high longitudinal tion about the transverse crack resistance. In this study, the fracture toughness in the composite. tility of the double torsion Dt) test technique to deter However, reductions of strength and strain to failure (in nine Gi in a glass-ceramic matrix composite(Nicalon/ the fiber direction) have been observed when testing at tem- CAS-Im at temperatures up to 1000.C has been demor peratures as low as 400oC. 10 In contrast to room-tempera temperature (as does the bulk matrix); however, no evi. pullout was observed at high temperatures. It has been pro- dence of an interphase oxidizing effect on crack growth posed that the transition from tough(notch- insensitive)behav (parallel to the fibers)could be found. The inevitable mis. ior at room temperature to brittle(notch-sensitive) behavior at significant matrix crack interactions with the fibers re- tion effects at interfaces exposed to the environment. 61 5 bridging the crack in the DT specimens, in contrast to the and/or increased fiber-matrix bond strength caused by ported for other geometries such as double cantilever beam The oxidation reaction is not initiated until the matrix cracks and flexure specimens. pon stressing(in the fiber direction), allowing penetration of high-temperature air. 12.13 The air then infiltrates the composite and attacks the low-strength carbon-rich fiber-matrix inter brittlement of the lot ERAMIc and glass-ceramic composites offer a great range of gitudinal fracture process..o t was reported that upon the oxi- utility on a high-temperature basis and based on other en- formed between the fiber and the matrix 14-l vironmental considerations such as oxidation resistance,ero- sion, and chemical attack from acids, etc. 2 The proper Inspite of the well-demonstrated embrittlement effect of ties of these composites depend on the combination of the reinforced glass-ceramic composites, there is not much study properties of starting materials and the fabrication proce- on the effects of temperature on fracture parallel to the fibers dure. By choosing high-strength, relatively high-modulus, continuous(or relatively long) fibers and incorporating them (transverse fracture). It was the objective of this study to in into matrices, high-strength and tough composites can be cre- vestigate the resistance of Nicalon-fiber-reinforced calcium ated if the fiber-matrix interphase is weak enough to divert aluminosilicate glass-ceramic matrix composite(Nicalon/CAS and/or bridge cracks. The ultimate strength of such a com- posite is then controlled by the in situ fiber strength, and a study of the transverse fracture toughness, the effects of high temperature air environment on fracture toughness, and also composite tensile strength well beyond that associated with the stress corrosion crack growth(if present) matrix cracking can be achieved. 3 The Nicalon-fiber reinforced glass-ceramic composites have met these require. ics is important since there are few engineering applications interphase during processing, and with no deterioration in fiber where transverse stresses are not encountered, 17, 18 The resis- tance to crack growth parallel to the fibers is especially a de- Nicalon has a unique nonstoichiometric chemistryis that ing parallel to the fibers is usually the initial form of damage in glass-ceramic matrix composites. Chemical reactions can occur multidirectional composites. 9 This cracking process has beer associated with: the(transverse) fracture toughness, defined by of energy per unit plate thickness per unit crack extension. Unstable(fast) crack growth should occur when the strain en ergy release rate, Gr, reaches the critical value, GIc, for the G. Evans--contributing editor material. Measurements of the strain energy release rate for fractures verify that the critical strain energy release rate is indeed a material parameter when fracture occurs exclusively by opening, Mode L, and when plane-strain conditions prevail 995: approved July 10, 1996. at the crack tip. 22,23 It is denoted by Gre, referring to failure in Mode I and under plain-strain conditions. Because of their
;;j ,;; 2 r.,. . .; ,.i . .. .._,...,. > .,.. J. Am. Cerarn. Soc., 80 I71 1812-20 (1597) High-Temperature Transverse Fracture Toughness of Nicalon-Fiber-Reinforced CAS-I I Glass-Ceramic Matrix Composite Ramazan Kahramant Department of Chemical Engineering, King Fahd University of Petroleum and Minerals, Dhahran 31261, Saudi Arabia John F. Mandell and Max C. Deibert Department of Chemical Engineering, Montana State University, Bozeman, Montana 59717 Cracking parallel to the fibers in off-axis plies is usually the initial form of damage in composite laminates. This cracking process has been associated with the (transverse) fracture toughness, defined by the critical strain energy release rate, GI,. The measurement of GI, provides basic inf'ormation about the transverse crack resistance. In this study, the utility of the double torsion (DT) test technique to determine GI, in a glass-ceramic matrix composite (Nicalod CAS-11) at temperatures up to 1OOO"C has been demonstrated. GI, did decrease moderately with increasing temperature (as does the bulk matrix); however, no evidence of an interphase oxidizing effect on crack growth (parallel to the fibers) could be found. The inevitable misalignment of fibers in the material was not very efficient at bridging the crack in the DT specimens, in contrast to the significant matrix crack interactions with the fibers reported for other geometries such as double cantilever beam and flexure specimens. I. Introduction ERAMIC and glass-ceramic composites offer a great range of C utility on a high-temperature basis and based on other environmental considerations such as oxidation resistance, erosion, and chemical attack from acids, etc.lV2 The properties of these composites depend on the combination of the properties of starting materials and the fabrication procedure.' By choosing high-strength, relatively high-modulus, continuous (or relatively long) fibers and incorporating them into matrices, high-strength and tough composites can be created if the fiber-matrix interphase is weak enough to divert and/or bridge cracks. The ultimate strength of such a composite is then controlled by the in situ fiber strength, and a composite tensile strength well beyond that associated with matrix cracking can be a~hieved.~ The Nicalon-fiberreinforced glass-ceramic composites have met these requirements through the development of a carbon-rich fiber-matrix interphase during processing, and with no deterioration in fiber strength. Nicalon has a unique nonstoichiometric chemisw5 that makes it particularly suited to the development of high-strength glass-ceramic matrix composites. Chemical reactions can occur between the fiber and matrix during processing of NicalonA. G. Evans-contributing editor Manuscript No. 192457. Received July 17. 1995; approved July 10. 1996. 'Author to whom correspondence should be addressed. reinforced glass-ceramics, producing the carbon-rich interphase regi~n.~."~ The carbon-rich layer forms a (fiber-matrix) bond strong enough for load transfer, yet weak enough to debond readily and allow fiber bridging during crack propagation (perpendicular to the fibers) resulting in high longitudinal fracture toughness in the c0mposite.6*~ However, reductions of strength and strain to failure (in the fiber direction) have been observed when testing at temperatures as low as 400°C.L0 In contrast to room-temperature fractures, a nearIy planar fracture surface with little fiber pullout was observed at high temperatures. It has been proposed that the transition from tough (notch-insensitive) behavior at room temperature to brittle (notch-sensitive) behavior at elevated temperatures is likely due to fiber strength degradation and/or increased fiber-matrix bond strength caused by oxidation effects at interfaces exposed to the The oxidation reaction is not initiated until the matrix cracks upon stressing (in the fiber direction), allowing penetration of high-temperature air.12J3 The air then infiltrates the composite and attacks the low-strength carbon-rich fiber-matrix interphase in such a way as to cause an embrittlement of the longitudinal fracture process.'S6 It was reported that upon the oxidation of the interphase carbon layer a stronger silica bond is formed between the fiber and the matix.'"I6 Inspite of the well-demonstrated embrittlement effect of temperature on the longitudinal behavior of the Nicalonreinforced glass-ceramic composites, there is not much study on the effects of temperature on fracture parallel to the fibers (transverse fracture). It was the objective of this study to investigate the resistance of Nicalon-fiber-reinforced calcium alurninosilicate glass-ceramic matrix composite (NicalodCASII) to crack growth parallel to the fibers. This included the study of the transverse fracture toughness, the effects of hightemperature air environment on fracture toughness, and also the stress corrosion crack growth (if present). Studying the transverse properties of fiber-reinforced ceramics is important since there are few engineering applications where transverse stresses are not en~ountered.'~.'~ The resistance to crack growth parallel to the fibers is especially a determining factor for the useful design stress range since cracking parallel to the fibers is usually the initial form of damage in multidirectional composites.'9 This cracking process has been associated with the (transverse) fracture toughness, defined by the critical strain energy release rate, GIc.20*21 G,, has the units of energy per unit plate thickness per unit crack extension. Unstable (fast) crack growth should occur when the strain energy release rate, GI, reaches the critical value, G,,. for the material. Measurements of the strain energy release rate for fractures verify that the critical strain energy release rate is indeed a material parameter when fracture occurs exclusively by opening, Mode I, and when plane-strain conditions prevail at the crack tip.22,23 It is denoted by Gn, refemng to failure in Mode I and under plain-strain conditions. Because of their ' 1812
July 1997 High-Temperature Transverse fracture Toughness l813 limited plasticity, materials like ceramics and glasses The theoretical analysis of the dt technique assu atisfy the plar ditions as long as the thickness of the crack front is straight and orthogonal to the plane of the tative of the material 23 ough to be microstructurally represen the specimen is specimen. However, in most materials the crack front profile is significantly curved, with the furthest advance along the bot tom face of the specimen. 4 Because of this, Evans2suggests I. Test Technique rather than analytically. It is important that a constant-K(stress intensity factor The consensus among most researchers, 32 has been specimen be used so that continuous crack length measure crack velocity and crack length of interest is that of the ments are not required for data reduction, since high dge of the crack, i.e the furthest advanced bottom temperature tests will take place within a furnace. the doubl location transverse fracture toughness in this study, satisfies the con- specimen dimensions are approximately as follows: wit um Tait Fry and Garrett 3 have suggested that the opti stant-K requirement. The utility of the dt technique has been demonstrated in both ceramic and in composites. 20, 2It thicknesses complicate the analysis because of the significant has been shown that it gives results similar to those of other test contact stresses generated by the interaction of the torsion arms The original DT technique was suggested by Outwater and Another area of contention for the dt test configuration is Gerry and developed by Kies and Clark. 27 As illustrated in Fig. the length of crack over which the stress intensity may be I, the test specimen is a rectangular plate with a narrow notch considered to be independent of crack length(constant-K re or crack started in one end. The specimen is loaded in four- point bending across the notch, i. e, in"double torsion. " Thus lengths greater than 0. 55w, and ligaments(uncracked portion) the crack propagates axially along the specimen The crack is greater than 0. 65W. Tait et al found a consensus over severa researchers which indicated that at least the central one-third t one-half of the specimen (for L/w =3) should offer shear deformation for the crack to develop modes Il and III constant-K conditions Shetty and Virkar4 stated that the valid failure vanishes because of the symmetric double torsion"of region decreases as the en length-to-width ratio de- 山 pecimen with respect to the crac2 creases First, the pecimen complia ance, c (inverse of the against displacement, d, curve), is linearly related to the spec Il. Experimental Procedure men crack length, a. Second, the critical load, Pc, at which crack propagation is initiated, is independent of crack length, a The materials used in this study were unidirectional ([0]16) (constant-K specimen). Therefore, the fracture toughness, GIe Nicalon fiber/CAS-Il matrix composites. They were supplied ed from the compliance by Corning, Inc in the form of square plates, approximately I UsIn cm by 15 cm and about 3 mm thick. dt tests were conducted on an instron model 8562 servo- P? dc electric universal testing machine. Tests were performed with a Gle=2 da (1) 500 lb Le Bow load cell mounted below the standard 20 000 lb Instron load cell. This lower capacity load cell was used be where t is the specimen thickness. Both Pc and dida can be cause of its better resolution in the load range of interest. An accurately determined from the test results, as will be discussed Inconel rod was attached to the load cell for load application in the Experimental Section A flat compression table was mounted to the bottom grip to The compliance relationship with respect to the crack length, hold the DT fixture base. a photograph of the loading section dClda, can also be expressed analytically by 28 of the dt test system is shown in Fig. 2 The test fixture used for the dt test was similar to the one designed by Professor Leon Chuck from the University of Day ton. A close-up of the fixture, with a composite specime mounted, appears in Fig. 3. The fore-and-aft, and side-to-side modifying the lengt由 the two pivot rollers:0m如b w(r=1-0.6302(2t/W)+1.20(2 /W)e m/ww (3) the top portion plays no role during loading. The fixture, the base plate, and the loading rod were all made from Inconel, a nickel-based superalloy, for use at ter mperatures up to 1000C. The testing machine was equipped with a slot type, 2 Zone Model"D", fumace, mounted on rails, supplied by Instron Corp This furmace which was originally designed for tensile sts was enlarged axially by plates to create a box-shaped furnace positioned over the slot furnace. Internal dimensions of the box were 15 cm x 15 cm For those tests performed at high temperatures using the en larged furnace, the temperature was read by inserting a ther mocouple wire into the furnace next to the dt specimen DT test specimen blanks of 2. 18 cm x 6.70 cm were cut from the composite plates using a diamond saw and a diamond coated grinding wheel. Initial notches were cut into one side of the specimens using the 1 mm thick diamond saw. To deter mine the compliance vs crack length calibration, dc/da, six to mens were prepared and a different notch length was machined into each specimen. The introduced notch length Fig. 1. Double torsion test configuration ranged from 1. 80 to 5.08 cm
July 1997 High-Temperature Transverse Fracture Toughness 1813 limited plasticity, brittle materials like ceramics and glasses satisfy the plane-strain conditions as long as the thickness of the specimen is large enough to be microstructurally representative of the material.23 11. Test Technique It is important that a constant-K (stress intensity factor) specimen be used so that continuous crack length measurements are not required for data reduction, since hightemperature tests will take place within a furnace. The double torsion (DT) test method, which was utilized to measure the transverse fracture toughness in this study, satisfies the constant-K requirement. The utility of the DT technique has been demonstrated in both ceramics2626 and in composites.20.21 It has been shown that it gives results similar to those of other test technique^.^^.^^ The original DT technique was suggested by Outwater and Gerry and developed by Kies and Clark.27 As illustrated in Fig. 1, the test specimen is a rectangular plate with a narrow notch or crack started in one end. The specimen is loaded in fourpoint bending across the notch, i.e., in “double torsion.” Thus the crack propagates axially along the specimen. The crack is more open at the bottom face and less open at the top face of the specimen, defining Mode I fra~t~re.~~.~~.~~*~~ The necessary shear deformation for the crack to develop Modes 11 and 111 failure vanishes because of the symmetric “double torsion” of the specimen with respect to the crack plane.2’.28 The DT test has two major features.20.21,2627,29.30 First, the specimen compliance, C (inverse of the initial slope of load, P, against displacement, d, curve), is linearly related to the specimen crack length, a. Second, the critical load, P,, at which crack propagation is initiated, is independent of crack length, a (constant-K specimen). Therefore, the fracture toughness, GI,, can be computed from the compliance calibration method by using.20.21 25.283 I dC G --- IC- 2t da where t is the specimen thickness. Both P, and dClda can be accurately determined from the test results, as will be discussed in the Experimental Section. The compliance relationship with respect to the crack length, dClda, can also be expressed analytically by28 dC 3x2 da - G Wr3 +( t) -- where G is the elastic shear modulus of the material, x, W, and t are specimen dimensions shown in Fig. 1, and +(t) is the thickness correction factor, which may be expressed by28 $(t) = 1 - 0.6302(2t/W) + 1.20(2t1W)e-“’(2‘w (3) v2 t Fig. 1. Double torsion test configuration. The theoretical analysis of the DT technique assumes that the crack front is straight and orthogonal to the plane of the specimen. However, in most materials the crack front profile is significantly curved, with the furthest advance along the bottom face of the specimen.24 Because of this, Evansz3 suggests that the compliance calibration be obtained experimentally rather than analytically. The consensus among most re~earchers~~.~~ has been that the crack velocity and crack length of interest is that of the leading edge of the crack, i.e., the furthest advanced bottom surface location. Tait, Fry, and Garrett33 have suggested that the optimum specimen dimensions are approximately as follows: width = W; length, L = 3W; and thickness = W16 to W/15. Greater thicknesses complicate the analysis because of the significant contact stresses generated by the interaction of the torsion arms (the two sides of the specimen separated by the crack). Another area of contention for the DT test configuration is the length of crack over which the stress intensity may be considered to be independent of crack length (constant-K region). Trantina30 showed that this assumption is valid for crack lengths greater than 0.55 W, and ligaments (uncracked portion) greater than 0.65W. Tait et found a consensus over several researchers which indicated that at least the central one-third to one-half of the specimen length (for UW = 3) should offer constant-K conditions. Shetty and Virkar34 stated that the valid region decreases as the specimen length-to-width ratio decreases. In. Experimental Procedure The materials used in this study were unidirectional ([O],,) Nicalon fiberlCAS-11 matrix composites. They were supplied by Coming, Inc. in the form of square plates, approximately 15 cm by 15 cm and about 3 mm thick. DT tests were conducted on an Instron Model 8562 servoelectric universal testing machine. Tests were performed with a 500 lb LeBow load cell mounted below the standard 20 OOO lb Instron load cell. This lower capacity load cell was used because of its better resolution in the load range of interest. An Inconel rod was attached to the load cell for load application. A flat compression table was mounted to the bottom grip to hold the DT fixture base. A photograph of the loading section of the DT test system is shown in Fig. 2. The test fixture used for the DT test was similar to the one designed by Professor Leon Chuck from the University of Dayton.35 A close-up of the fixture, with a composite specimen mounted, appears in Fig. 3. The fore-and-aft, and side-to-side alignment of the base and the upper fixture is adjusted by modifying the length of the two pivot rollers. Once the specimen and fixture are aligned during setup, the support fixture for the top portion plays no role during loading. The fixture, the base plate, and the loading rod were all made from Inconel, a nickel-based superalloy, for use at temperatures up to 1000°C. The testing machine was equipped with a slot type, 2 Zone Model “D”, furnace, mounted on rails, supplied by Instron Corp. This furnace which was originally designed for tensile tests was enlarged axially by attaching alumina insulation plates to create a box-shaped furnace positioned over the slot furnace. Internal dimensions of the box were 15 cm x 15 cm. For those tests performed at high temperatures using the enlarged furnace, the temperature was read by inserting a thermocouple wire into the furnace next to the DT specimen. DT test specimen blanks of 2.18 cm x 6.70 cm were cut from the composite plates using a diamond saw and a diamondcoated grinding wheel. Initial notches were cut into one side of the specimens using the 1 mm thick diamond saw. To determine the compliance vs crack length calibration, dClda, six to eight specimens were prepared and a different notch length was machined into each specimen. The introduced notch length ranged from 1.80 to 5.08 cm
1814 Journal of the American Ceramic Society-Kahraman et al Vol. 80. No. 7 a SLOPE 1 0.04 Displacement(E)mm Fig 4. Specimen compliance(C) calibration load vs displacement plot. 6.54cm 0.10cm Fig. 2. Loading section of the double torsion test system VIEW EDGE VIEW Flg. 5. Double torsion specimen for fracture testing, including notch Ig. 3. Photograph of the double torsion test fixture root detail The compliance calibration tests were conducted at a loading Before the dt tests to failure were performed the notch ti rate of 89 N/min, while plotting load vs piston displacement. were machined such that the thickness of the uncracked part of Specimens were loaded to low loads to avoid fracture during the specimen at the notch tip tapered from very thin to the full compliance calibration. Only the compliance of the specimen hickness as shown in Fig. 5. The top face of the specimen was before any appreciable crack growth occurred was of interest in the face with the longest notch, corresponding to the full thick the calibration. The sharpness of the crack tip was therefore not ness. This facilitated the initiation of a sharp precrack in the critical in determining dc/da specimen, as suggested by Tait et al. 33 Since the stress intensity A representative load(P)vs displacement(d) curve of a is inversely proportional to r,2325.3133 a reduction of the ompliance calibration test is shown in Fig. 4. Compliance, C, thickness by taper from full thickness to zero at the crack tip was determined by calculatin ing the slope of the linear portion of results in very high stress intensities at the P-d curve during loading. The nonlinearities in the lot facilitates the formation of a sharp crack at low loads. The load portion of the P-d curves may be attributed to the gradual crack can thus initiate at loads well below those required to accumulation of forces in the fixture. 5 The change in compli cause fast fracture of the full thickness. As noted by evans, ance with crack length, dClda, was calculated by performing a a sharp precrack is necessary for valid fracture toughness tests linear regression of compliance(C) vs crack length(a) in ceramics. The critical load values obtained from blunt cracks
1814 0.10 cm Journal of the American Ceramic Society--Kahraman et al. Vol. 80, No. 7 -6.69 c- -6.54 cm-----) Fig. 2. Loading section of the double torsion test system. Fig. 3. Photograph of the double torsion test fixture. The compliance calibration tests were conducted at a loading rate of 89 N/min, while plotting load vs piston displacement. Specimens were loaded to low loads to avoid fracture during compliance calibration. Only the compliance of the specimen before any appreciable crack growth occurred was of interest in the calibration. The sharpness of the crack tip was therefore not critical in determining dC/da. A representative load (P) vs displacement (d) curve of a compliance calibration test is shown in Fig. 4. Compliance, C, was determined by calculating the slope of the linear portion of the P-d curve during loading. The nonlinearities in the lowload portion of the P-d curves may be attributed to the gradual accumulation of forces in the fixture.35 The change in compliance with crack length, dC/du, was calculated by performing a linear regression of compliance (C) vs crack length (a). Displacement (6), rnm Fig. 4. Specimen compliance (0 calibration load vs displacement plot. Fig. 5. Double torsion specimen for fracture testing, including notch root detail. Before the DT tests to failure were performed, the notch tips were machined such that the thickness of the uncracked part of the specimen at the notch tip tapered from very thin to the full thickness as shown in Fig. 5. The top face of the specimen was the face with the longest notch, corresponding to the full thickness. This facilitated the initiation of a sharp precrack in the specimen, as suggested by Tait et Since the stress intensity is inversely proportional to t1n,23.25*31*33 a reduction of the thickness by taper from full thickness to zero at the crack tip results in very high stress intensities at first loading which facilitates the formation of a sharp crack at low loads. The crack can thus initiate at loads well below those required to cause fast fracture of the full thickness. As noted by Evans,23 a sharp precrack is necessary for valid fracture toughness tests in ceramics. The critical load values obtained from blunt cracks
July 1997 High-Temperature Transverse Fracture Toughness a 100% increase in g tests to failure were performed with load-point displace- ment rates ranging from 0. 1 mm/h to 0.05 mm/min. a repre sentative P-d curve for the dt tests to failure is shown in Fig 6. Failure did not occur by a sudden total catastrophic fracture of the specimen, but rather by a gradual propagation of the E crack with a gradually increasing nonlinearity(decreasing ope). Fiber bridging might occur during the test due to the inevitable misalignment of fibers in the material, and there might be fibers which cross the plane of the crack.36-38As the crack opens, these fibers will be extracted from the matrix requiring additional applied load. Once crack propagation has become well established, fiber bridging will dominate the frac as it extends, in contrast to a fairly constant load for stable 6 ture energy. This results in an increasing load to grow the crack crack propagation in homogeneous materials. IHowever, the region of nonlinearity in Fig. 6 is small, suggesting that the b. i fibers are not very efficient at bridging the crack in the dt specimens(as will be discussed in the Results section), in contrast to the significant matrix crack interactions with the fibers reported for other geometries such as double cantilever beam(dCB)6.37 and flexure specimens. 38 Still the gradual failure made it difficult to precisely determine the critical load, In order to better define Pe, the offset procedure,35 was employed. P. was determined from the P-d curve with the Flg. 7. Compliance vs crack length for glass double torsion speci secant of 5% lower slope than the original elastic slope(slope mens. of the linear portion of the curve)as shown in Fig The transverse fracture toughness, GIc, was then calculated from Eq. (1) glass slides with known fracture toughness as a calibration of Fiber-matrix bond strength measurements were done utiliz. the Dt test apparatus and procedure. Experimental details are ing the microdebonding technique, where individual fibers are in Ref. 42 The compliance vs crack length plot and the slope(dC/da) obtained by lir is shown in Fig. 7. The mental dC/da, 0.000140 N-1, is in good agreement with the calculated dcyda(using 2), which is0000136N-1.4 I. Results Note that the compliance-axis-intercept is not zero for the experimental compliance calibration. This nonzero intercept is Before performing any fracture test on Nicalon/CAS-II com- attributed to the compliance of the uncracked part of the speci posite specimens, the reliability of the DT test results was men.28 The arms( two sides across the crack) of the dt s monstrated by the preliminary tests on soda-lime-silicate men are attached not to a ri ate, but to a plate(uncracked portion of the specimen) with some compliance which is inde pendent of crack length. The compliance of an uncracked specimen(a 0)is also included in Fig. 7. It falls close to the The fracture toughness, GIe, values were calculated from Eq appears from the data that the crack experiences constant-K conditions for about 0.1 <alL<0.7. This is the expected valid range from the literature.33 The Gc data in this constant-K region scatter in the range of 5.5-7.5 J/m2 with an average value of about 6.4 J/m2 Ke values in the literature are found to be in the range of 43.44 6.35-8.3 J/m2,42 Thus, the experimental data correlate well with those found in literature (1)Composite Compliance Calibration DT compliance calibration tests were performed on compos 54 Offset Compiance ite specimens with different premachined crack lengths. The resulting sample compliance vs crack length plot of those test -squares linear regress men compliance vs crack length data was done to determine the compliance calibration, dC/da, and the zero crack length inter cept. Note again that the compliance-axis -intercept is not zero for the experimental compliance calibration, as discussed 0,000020 060.080.10 above Displacement(5), mm The experimental dC/da, the slope of C-a plot in Fig. 9, is 5.35 x 10-6 N-. The analytical dc/da was calculated to cal load P than 2%
July 1997 High-Temperature Transverse Fracture Toughness 1815 from DT specimens may be high by as much as 40% causing about a 100% increase in The tests to failure were performed with load-point displacement rates ranging from 0.1 mmfl.1 to 0.05 dmin. A representative P-d curve for the DT tests to failure is shown in Fig. 6. Failure did not occur by a sudden total catastrophic fracture of the specimen, but rather by a gradual propagation of the crack with a gradually increasing nonlinearity (decreasing slope). Fiber bridging might occur during the test due to the inevitable misalignment of fibers in the material, and there might be fibers which cross the plane of the crack.3c38 As the crack opens, these fibers will be extracted from the matrix, requiring additional applied load. Once crack propagation has become well established, fiber bridging will dominate the fracture energy. This results in an increasing load to grow the crack as it extends, in contrast to a fairly constant load for stable crack propagation in homogeneous materials.21 However, the region of nonlinearity in Fig. 6 is small, suggesting that the fibers are not very efficient at bridging the crack in the DT specimens (as will be discussed in the Results section), in contrast to the significant matrix crack interactions with the fibers reported for other geometries such as double cantilever beam (DCB)36,37 and flexure specimens.38 Still the gradual failure made it difficult to precisely determine the critical load, P,, at which crack propagation initiated. was employed. P, was determined from the P-d curve with the secant of 5% lower slope than the original elastic slope (slope of the linear portion of the curve) as shown in Fig. 6. The transverse fracture toughness, GI,, was then calculated from Eq. (1). Fiber-matrix bond strength measurements were done utilizing the microdebonding technique, where individual fibers are compressively loaded on a polished surface to produce debonding.4"+2 In order to better define P,, the offset IV. Results Before performing any fracture test on NicalodCAS-II composite specimens, the reliability of the DT test results was demonstrated by the preliminary tests on soda-limesilicate 250 200 Z 150 a U 0 h " 2 100 50 I I 0.00 002 0.04 0.06 0.08 0.10 0.12 0,14 Displacement (6), mm Fig. 6. Specimen load vs displacement plot from the double torsion tests to failure, including 5% offset compliance to determine the critical load, P,. 1 I I I I 0.00 0.0 1 0.02 0.03 0.04 0.05 0.06 Crack Lenglh (a), m Fig. 7. Compliance vs crack length for glass double torsion specimens. glass slides with known fracture toughness as a calibration of the DT test apparatus and procedure. Experimental details are in Ref. 42. The compliance vs crack length plot and the slope (dC/du) obtained by linear regression is shown in Fig. 7. The experimental dC/da, 0.000140 N-I, is in good agreement with the calculated dCYda (using Eq. (2)), which is 0.000136 N-1.42 Note that the compliance-axis-intercept is not zero for the experimental compliance calibration. This nonzero intercept is attributed to the compliance of the uncracked part of the specimen.28 The arms (two sides across the crack) of the DT specimen are attached not to a rigid plate, but to a plate (uncracked portion of the specimen) with some compliance which is independent of crack length. The compliance of an uncracked specimen (a = 0) is also included in Fig. 7. It falls close to the trend-line intercept. The fracture toughness, GI,, values were calculated from Eq. (1) and plotted against crack length as shown in Fig. 8. It appears from the data that the crack experiences constant-K conditions for about 0.1 < a/L < 0.7. This is the expected valid range from the literature.33 The GI, data in this constant4 region scatter in the range of 5.5-7.5 J/m2 with an average value of about 6.4 J/mz. K,, values in the literature are found to be in the range of 0.7-0.8 MPa - m1'2,43,44 corresponding to a GI, range of about 6.35-8.3 J/m2.42 Thus, the experimental data correlate quite well with those found in literature. (1) Composite Compliance Calibration DT compliance calibration tests were performed on composite specimens with different premachined crack lengths. The resulting sample compliance vs crack length plot of those tests is shown in Fig. 9. A least-squares linear regression of specimen compliance vs crack length data was done to determine the Compliance calibration, dC/da, and the zero crack length intercept. Note again that the compliance-axis-intercept is not zero for the experimental compliance calibration, as discussed above. The experimental dC/da, the slope of C-a plot in Fig. 9, is 5.35 x lod N-I. The analytical dC/da was calculated to be 5.45 x lo4 N-' using Eq. (2).42 The difference between the experimental and the analytical values of dC/da is less than 2%
I816 Joumal of the American Ceramic Society--Kahraman et al. Vol. 80. No. 7 Fig. 8. Critical strain energy release rate(GIe)measured on srack Flg. 9. Compliance vs crack length plot for Nicalon/CAS-lI double as a function of torsion specimens. men ness values for two Nicalon/CAS-lI composite plates are close (2) Room-Temperature Test Results to that of the CAS matrix, reported as 51.5 J/m2 at room tem Composite DT specimens were cut and notched as shown in perature. 7Schutz35 determined a lower Gis for Nicalon/CAS- Fig. 5. As stated earlier, the tapered notch facilitates the ini- Il, about 34 J/m2 with experimental compliance calibration tiation of a sharp precrack in the specimen at a load lower than (about 41 J/m2 if analytical compliance calibration was used) Pe, which is necessary for fracture toughness tests to be valid. (The difference between the analytical and experimental dc/d This was demonstrated by loading a specimen up to about 85% was about 20%, compared to 2% in this study of the critical load. The notch tip in the specimen was exam- Although matrix crack-fiber interactions, as the crack front ined under the optical microscope before and after the test. meanders and crosses over inclined fibers, are reported to While there was no change on the upper face of the specimen, dominate the measured fracture loads in conventional double up pictures of the notch tip on the bottom side of the specimen of this study the fibers do not appear to be very effective ts cantilever beam(DCB)specimens according to the res before the loading(without a sharp crack) and after the loading bridging the crack in the DT specimens. The bridging fibers in (with the introduced sharp crack)are shown in Figs. 10 and 11 o.The sharp precrack extended about 7 mm on the bottom face pressure is achieved at a crack opening of about 60 um. That crack(notch) length on the top face. This observation was in (see next section). Thus, there might not be enough crack open good agreement with the curved front profile discussed earlier ing in the DT specimen to show a significant toughness in- The fracture toughness test data obtained in this study therefore crease due to the fibers as compared to the DCB specimen results from naturally developed sharp cracks prior to failure The relationship between the transverse fracture toughne ()High-Temperature Transverse Fracture Results Gre, and the initial crack length, a, for Nicalon/CAS-lI DT The same analytical dC/da which was determined at room specimens from two plates is illustrated in Fig. 12. The crack mperature was used for high-temperature transverse fracture length used in the plot is the approximate length of the crack on toughness calculations, assuming that the composite retains its the bottom side of the specimen after the initiation of the sharp room-temperature compliance at temperatures up to 1000C crack at the notch tip. It was demonstrated above that after the This was demonstrated by determining the compliance of a initiation of a sharp crack at the notch tip at a load lower than Nicalon/CAS-IIDT specimen both at room temperature and at Pe the crack length on the bottom face becomes about 4 mm 1000oC. The change in compliance between room temperature longer than the notch length on the top face nd 1000C was small. less than 5%. This is consistent with the There seems to be no discernible trend between the lack of any significant decrease in modulus observed over this sured transverse fracture toughness and the initial crack length range. 5 The specimens tested at different temperatures had the in the range of crack lengths used. This confirms the con same(top face)crack length of about 3. 28 cm stant-K characteristics of this specimen type, as already dem- A transverse fracture toughness(Gre)vs time to failure(test- onstrated on glass specimens for about 0.1 alL <0.7. The ing time)plot of Nicalon/CAS-Il specimens for room tempera- data point for alL =0.73 may not be valid because of the long ture, 800 C and 1000@C is shown in Fig 13. There seems to b initial crack length, which might be out of the constant-k no significant effect of test rate on transverse fracture tough Considering only the specimens with 0.1 a/L <0.7, the 38-61 J/m with an average value of 48 J/m2. On the or ness at room temperature. The Gre data scatter in the ra average transverse fracture toughness value is 49 and 50 J/m hand, the transverse fracture toughness of Nicalon/CASIl for plates 1 and 3, respectively. These measured fracture tough- composite was determined to be lower than 38 J/m- for the
1816 Journal of the American Ceramic Society-Kahraman et al. Vol. 80. No. 7 f f \ I I I I 0.2 0.4 0.6 0.8 a/l Fig. 8. Critical strain energy release rate (G,c) measured on sodalime-silicate glass double torsion specimens as a function of crack length-to-specimen length ratio (dL). (2) Room-Temperature Test Results Composite DT specimens were cut and notched as shown in Fig. 5. As stated earlier, the tapered notch facilitates the initiation of a sharp precrack in the specimen at a load lower than P,, which is necessary for fracture toughness tests to be valid. This was demonstrated by loading a specimen up to about 85% of the critical load. The notch tip in the specimen was examined under the optical microscope before and after the test. While there was no change on the upper face of the specimen, a sharp crack initiated at the notch tip on the lower face. Closeup pictures of the notch tip on the bottom side of the specimen before the loading (without a sharp crack) and after the loading (with the introduced sharp crack) are shown in Figs. 10 and 11, respectively. The sharp precrack extended about 7 mm on the bottom face making the crack length on this face 4 mm longer than the crack (notch) length on the top face. This observation was in good agreement with the curved front profile discussed earlier. The fracture toughness test data obtained in this study therefore results from naturally developed sharp cracks prior to failure. The relationship between the transverse fracture toughness, G,, and the initial crack length, a, for NicalodCAS-11 DT specimens from two plates is illustrated in Fig. 12. The crack length used in the plot is the approximate length of the crack on the bottom side of the specimen after the initiation of the sharp crack at the notch tip. It was demonstrated above that after the initiation of a sharp crack at the notch tip at a load lower than P, the crack length on the bottom face becomes about 4 mm longer than the notch length on the top face. There seems to be no discernible trend between the measured transverse fracture toughness and the initial crack length in the range of crack lengths used. This confirms the constant-K characteristics of this specimen type, as already demonstrated on glass specimens for about 0.1 < a/L < 0.7. The data point for a/L = 0.73 may not be valid because of the long initial crack length, which might be out of the constant-K range. Considering only the specimens with 0.1 < a/L < 0.7, the average transverse fracture toughness value is 49 and 50 J/m2 for plates 1 and 3, respectively. These measured fracture tough- 0 0.00 L 0.01 0.02 0.03 0 Crack Length (a). m Fig. 9. Compliance vs crack length plot for NicalodCAS-I1 double torsion specimens. ness values for two NicalodCAS-I1 composite plates are close to that of the CAS matrix, reported as 51.5 J/m2 at room temperature." Sch~tz~~ determined a lower G, for NicalodCASII, about 34 J/m2 with experimental compliance calibration (about 41 J/m2 if analytical compliance calibration was used). (The difference between the analytical and experimental dC/da was about 20%. compared to 2% in this study.) Although matrix crack-fiber interactions, as the crack front meanders and crosses over inclined fibers, are reported to dominate the measured fracture loads in conventional double cantilever beam (DCB) specimen^,^^^* according to the results of this study the fibers do not appear to be very effective at bridging the crack in the DT specimens. The bridging fibers in the crack wake exert a crack closing pressure for a given crack opening and according to Kaute et a1.36 the maximum closing pressure is achieved at a crack opening of about 60 pm. That large of a crack opening is never achieved in DT specimens (see next section). Thus, there might not be enough crack opening in the DT specimen to show a significant toughness increase due to the fibers as compared to the DCB specimen. (3) High-Temperature Transverse Fracture Results The same analytical dC/da which was determined at room temperature was used for high-temperature transverse fracture toughness calculations, assuming that the composite retains its room-temperature compliance at temperatures up to 1O0O"C. This was demonstrated by determining the compliance of a NicalodCAS-11 DT specimen both at room temperature and at 1OOO"C. The change in compliance between room temperature and 1OOO"C was small, less than 5%. This is consistent with the lack of any significant decrease in modulus observed over this range.35 The specimens tested at different temperatures had the same (top face) crack length of about 3.28 cm. A transverse fracture toughness (GI,) vs time to failure (testing time) plot of NicalodCAS-I1 specimens for room temperature, 800°C and 1000°C is shown in Fig. 13. There seems to be no significant effect of test rate on transverse fracture toughness at room temperature. The GI, data scatter in the range of 38-61 J/m2 with an average value of 48 J/m2. On the other hand, the transverse fracture toughness of NicalodCAS-I1 composite was determined to be lower than 38 J/mz for the