Availableonlineatwww.sciencedirect.com SCIENCE E噩≈S ournal of the European Ceramic Society 24(2004)565-578 www.elsevier.com/locate/jeurceramsoc Thermal degradation of an oxide fibre(Nextel 720)/ aluminosilicate composite M.-L. Anttia,*E. Lara-CurziobR. Warren a Division of Engineering Materials, Luled University of Technology, 97187 Luled, Sweden mEtals and Ceramics Division, Oak Ridge National Laboratory, Oak Ridge, TN 37830-6064, USA Materials Science Group, Division of Innovation, Production and Management, Malmo University, 20506 Malmo, Sweden Received 5 January 2002: received in revised form 15 March 2003: accepted 6 April 200 The effect of thermal exposure on the microstructure and tensile stress-strain behaviour has been investigated for composites of woven continuous oxide fibres(Nextel 720)in a porous aluminosilicate matrix. The tensile tests were carried out on straight-sided centre hole notched plates with 0/90 and +45 orientations. The as-received material was slightly notch sensitive in that the net section fracture stress decreased somewhat with increasing hole diameter but much less than predicted for an ideally elastic, fully notch-sensitive material. After exposure at 1100C and for long time at 1000C in air the composite was embrittled. In the o composite this resulted in a reduced fracture strength, a reduced strain to failure as well as a reduced fracture toughness and damage zone size. After exposure for 100 h at 1 100C(the most extreme exposure applied) the material also became significantly more notch sensitive and had failure characteristics similar to those of a monolithic ceramic. The 45 composite was also embrit- tled which resulted in a reduced strain to failure but an increase in fracture strength. Density measurements and observations on the microstructure and fracture surfaces indicated that the embrittlement was due mainly to localised densification of the matrix and an increase in fibre/matrix bonding C 2003 Elsevier Ltd. All rights reserved Keywords: Composites; Fibres; Mechanical properties; Nextel fibres; Thermal degradation 1. Introduction the matrix and the crack deflects into a plane parallel to the loading direction. ' In particular oxide/oxide CFCCs It is now well established that continuous fibre rein- exploiting this principle attract interest as candidate forced ceramic composites(CFCCs)can be made to materials for use in combustors, exhibiting damage tol- have non-brittle fracture behaviour and improved erance combined with inherent oxidation resistance damage tolerance by introduction of a suitably weak Their anticipated ability to operate at higher tempera fibre/matrix interface which provides a more favourable tures than the superalloys used today is expected to lead path for the extension of matrix cracks than penetrating to an increased efficiency and a reduced need for cooling the fibre. To achieve this an interphase is often coated air, as well as a decreased emission of NO gases. The separately onto the fibres during processing and this life requirements of the material are high, the aim being involves both an increased production cost as well as an several hundred thousand hours at at least 1100C added complexity. It has recently been demonstrated When designing a structural component, the effect of that similar crack-deflecting behaviour can also be holes and notches on the mechanical properties is an achieved by means of a finely distributed porosity in the important practical factor to be taken into considera matrix instead of a separate interphase between matrix tion. For this reason the mechanical strength of CFCCs and fibres -Delamination has been shown to occur in has frequently been studied in terms of the notch strength and notch sensitivity. The present work con- Corresponding author. TeL : +49-920-492093; fax: +46-920 cerns the notch strength behaviour of a commercially 491084. available composite consisting of laminated, woven E-mail address: marta-lena antti(@ mb luth. se(M.-L. Antti) mullite/alumina fibres (Nextel 720) in a porous 0955-2219/03/S. see front matter C 2003 Elsevier Ltd. All rights reserved. doi:10.1016S0955-221903)00250-4
Thermal degradation of an oxide fibre (Nextel 720)/aluminosilicate composite M.-L. Anttia,*, E. Lara-Curziob, R. Warrenc a Division of Engineering Materials, Lulea˚ University of Technology, 97187 Lulea˚, Sweden bMetals and Ceramics Division, Oak Ridge National Laboratory, Oak Ridge, TN 37830-6064, USA c Materials Science Group, Division of Innovation, Production and Management, Malmo¨ University, 20506 Malmo¨, Sweden Received 5 January 2002; received in revised form 15 March 2003; accepted 6 April 2003 Abstract The effect of thermal exposure on the microstructure and tensile stress–strain behaviour has been investigated for composites of woven continuous oxide fibres (Nextel 720) in a porous aluminosilicate matrix. The tensile tests were carried out on straight-sided, centre hole notched plates with 0/90 and 45 orientations. The as-received material was slightly notch sensitive in that the net section fracture stress decreased somewhat with increasing hole diameter but much less than predicted for an ideally elastic, fully notch-sensitive material. After exposure at 1100 C and for long time at 1000 C in air the composite was embrittled. In the 0/90 composite this resulted in a reduced fracture strength, a reduced strain to failure as well as a reduced fracture toughness and damage zone size. After exposure for 100 h at 1100 C (the most extreme exposure applied) the material also became significantly more notch sensitive and had failure characteristics similar to those of a monolithic ceramic. The 45 composite was also embrittled which resulted in a reduced strain to failure but an increase in fracture strength. Density measurements and observations on the microstructure and fracture surfaces indicated that the embrittlement was due mainly to localised densification of the matrix and an increase in fibre/matrix bonding. # 2003 Elsevier Ltd. All rights reserved. Keywords: Composites; Fibres; Mechanical properties; Nextel fibres; Thermal degradation 1. Introduction It is now well established that continuous fibre reinforced ceramic composites (CFCCs) can be made to have non-brittle fracture behaviour and improved damage tolerance by introduction of a suitably weak fibre/matrix interface which provides a more favourable path for the extension of matrix cracks than penetrating the fibre. To achieve this an interphase is often coated separately onto the fibres during processing and this involves both an increased production cost as well as an added complexity. It has recently been demonstrated that similar crack-deflecting behaviour can also be achieved by means of a finely distributed porosity in the matrix instead of a separate interphase between matrix and fibres.13 Delamination has been shown to occur in the matrix and the crack deflects into a plane parallel to the loading direction.1 In particular oxide/oxide CFCCs exploiting this principle attract interest as candidate materials for use in combustors, exhibiting damage tolerance combined with inherent oxidation resistance. Their anticipated ability to operate at higher temperatures than the superalloys used today is expected to lead to an increased efficiency and a reduced need for cooling air, as well as a decreased emission of NOx gases. The life requirements of the material are high, the aim being several hundred thousand hours at at least 1100 C. When designing a structural component, the effect of holes and notches on the mechanical properties is an important practical factor to be taken into consideration. For this reason the mechanical strength of CFCCs has frequently been studied in terms of the notch strength and notch sensitivity. The present work concerns the notch strength behaviour of a commercially available composite consisting of laminated, woven mullite/alumina fibres (Nextel 720) in a porous 0955-2219/03/$ - see front matter # 2003 Elsevier Ltd. All rights reserved. doi:10.1016/S0955-2219(03)00250-4 Journal of the European Ceramic Society 24 (2004) 565–578 www.elsevier.com/locate/jeurceramsoc * Corresponding author. Tel.: +49-920-492093; fax: +46-920- 491084. E-mail address: marta-lena.antti@mb.luth.se (M.-L. Antti).
M.L. Antti et al. Journal of the European Ceramic Society 24(2004)565-578 aluminosilicate(AS) matrix. The results have been pre- However, in a study of edge-notched specimens of a 0/ sented in general terms in an earlier study. 4 The focus of 90 Nextel 610/AS composite it was found that the fail this study is the thermal instability of the composite ure mode changed from multiple matrix fracture at microstructure when exposed to elevated temperatures room temperature to self-similar crack growth at 950C. and in particular how this influences notch strength and This change led to a significantly increased notch sensi- notch sensitivity.Both090°and±45° orientations tivit were investigated Fibre degradation as a possible source of composite In common with polymer matrix composites, degradation should also be considered. Nextel 720 fibres matrix composites and several non-oxide CF( consist of alumina grains with an approximate diameter porous matrix oxide/oxide composites exhibit me of 0. 1 um distributed among larger (0.5 um) mullite notch sensitivity. -12 For example, centre hole plate grains consisting of many smaller subg samples with 0/90 orientation at room temperature gations of the response of the fibres to thermal exposure exhibit a moderate decrease in net section strength with are somewhat conflicting. Deleglise et al. 5 observed increasing hole diameter. The strength falls towards a significant strength degradation only above 1400C for limiting value for large hole diameters that generally lies 5 h exposure times while Milz et al. 6 observed severe between 60 and 80% of the unnotched strength. The degradation after 2 h at 1300oC. Petry and Mah" elatively good notch tolerance is associated with the report a small loss in strength after exposure for 2 h at development of an extensive zone of damage developing 1100C. The causes of degradation are also not clear; from the hole with increasing load which effectively surface grooving, structural coarsening 7 and local reduces the stress intensity at the hole. In dense matrix impurity enrichment have been suggested. 6 The con composites the damage consists of multiple tensile flicting observations could as well be due to differences cracking of the matrix combined with fibre/matrix in the fibre batches used as to experimental differences. debonding In porous matrix composites, shear damage of the matrix is thought to make a significant contribu- tion. Composites with +45 orientation are notch 2. The notch sensitivity test insensitive. 2 12 Their fracture is dominated by shear failure of the matrix. During the composite fracture As already indicated the main objective of the present process the fibre tows can separate without significant study was to investigate the degradation at high tem- fibre fracture. A function of the fibres is to aid the peratures of oxide/oxide, porous matrix composites and extension of matrix damage thereby reducing the stress in particular through its effect on the notch sensitivity intensity. The nature of the centre hole notch test is measured in terms of the tensile properties of centre discussed in more detail in Section 2 hole panels. This form of notch sensitivity test, though Reports on the thermal degradation of porous matrix, of practical relevance, is somewhat complex in inter all-oxide CFCCs are scarce. Levi et al. report, for an pretation. In an isotropic, elastically deforming solid alumina fibre(Nextel 610) reinforced mullite/alumina the maximum stress intensification at the edge of a cir composite, little change in unnotched tensile behaviour cular notch in an infinitely wide plate loaded in uniaxial after up to 100 h in air at 1200 oC. However, a 2 h tension is a factor of three regardless of hole size. 8 Thus exposure at 1300C led to reduced fracture stress and the strength of an ideally, fully notch sensitive material reduced strain to failure. The authors attributed this to will be reduced by a factor of three when such a notch is fibre degradation rather than to changes in the matrix introduced. A fully notch insensitive material will not be ind suggested that the use of the more stable Nextel 720 weakened since the stress intensification will be dis fibre should lead to improved thermal stability. Jurf and sipated, for example, by plastic deformation. In reality Butner3observed a decrease to about 70% unnotched most materials, including long-fibre reinforced compo- strength of the 0/90 material studied here after 1000 h at sites, exhibit intermediate behaviour leading to a mod- 100oC. It was suggested that the primary explanation erate loss of strength that is generally dependent on hole for the loss in strength is matrix densification rather size. The dependence on hole size arises because the than degradation of fibre strength. 3 Long-term ageing circular notch affects and interacts with the material in has also been performed by Siemens-Westinghouse. 4 its immediate vicinity, thereby altering the distribution At 1100C the unnotched strength after 100 h was of the stress intensification; the volume of affected reduced by over 25% and after 2600 h by over 55%.A material scales with the hole size. 8 An added compli heat-treatment of 3000 h at 1000C led to a decrease in cation in experimental situations is that the stress strength of about 23%. The suggested explanation of intensification of a circular hole is affected by the width the degradation was embrittlement due to densification of the tested plate, decreasing from a factor of three to a of the matrix and possibly a phase change. 4 The pre- factor of two as the ratio of hole size to plate width(a/w) sent authors are not aware of any earlier reports on the approaches 1, thus necessitating a finite width correc notch strength or notch sensitivity of such composites. tion. It is also to be noted that the stress intensification
aluminosilicate (AS) matrix. The results have been presented in general terms in an earlier study.4 The focus of this study is the thermal instability of the composite microstructure when exposed to elevated temperatures and in particular how this influences notch strength and notch sensitivity. Both 0/90 and 45 orientations were investigated. In common with polymer matrix composites, carbon matrix composites and several non-oxide CFCCs,58 porous matrix oxide/oxide composites exhibit moderate notch sensitivity.912 For example, centre hole plate samples with 0/90 orientation at room temperature exhibit a moderate decrease in net section strength with increasing hole diameter. The strength falls towards a limiting value for large hole diameters that generally lies between 60 and 80% of the unnotched strength. The relatively good notch tolerance is associated with the development of an extensive zone of damage developing from the hole with increasing load which effectively reduces the stress intensity at the hole. In dense matrix composites the damage consists of multiple tensile cracking of the matrix combined with fibre/matrix debonding. In porous matrix composites, shear damage of the matrix is thought to make a significant contribution.2,12 Composites with 45 orientation are notch insensitive.2,12 Their fracture is dominated by shear failure of the matrix. During the composite fracture process the fibre tows can separate without significant fibre fracture. A function of the fibres is to aid the extension of matrix damage thereby reducing the stress intensity. The nature of the centre hole notch test is discussed in more detail in Section 2. Reports on the thermal degradation of porous matrix, all-oxide CFCCs are scarce. Levi et al.2 report, for an alumina fibre (Nextel 610) reinforced mullite/alumina composite, little change in unnotched tensile behaviour after up to 100 h in air at 1200 C. However, a 2 h exposure at 1300 C led to reduced fracture stress and reduced strain to failure. The authors attributed this to fibre degradation rather than to changes in the matrix and suggested that the use of the more stable Nextel 720 fibre should lead to improved thermal stability. Jurf and Butner13 observed a decrease to about 70% unnotched strength of the 0/90 material studied here after 1000 h at 1100 C. It was suggested that the primary explanation for the loss in strength is matrix densification rather than degradation of fibre strength.13 Long-term ageing has also been performed by Siemens-Westinghouse.14 At 1100 C the unnotched strength after 100 h was reduced by over 25% and after 2600 h by over 55%. A heat-treatment of 3000 h at 1000 C led to a decrease in strength of about 23%. The suggested explanation of the degradation was embrittlement due to densification of the matrix and possibly a phase change.14 The present authors are not aware of any earlier reports on the notch strength or notch sensitivity of such composites. However, in a study of edge-notched specimens of a 0/ 90 Nextel 610/AS composite it was found that the failure mode changed from multiple matrix fracture at room temperature to self-similar crack growth at 950 C. This change led to a significantly increased notch sensitivity.11 Fibre degradation as a possible source of composite degradation should also be considered. Nextel 720 fibres consist of alumina grains with an approximate diameter of 0.1 mm distributed among larger (0.5 mm) mullite grains consisting of many smaller subgrains.15 Investigations of the response of the fibres to thermal exposure are somewhat conflicting. Dele´glise et al.15 observed significant strength degradation only above 1400 C for 5 h exposure times while Milz et al.16 observed severe degradation after 2 h at 1300 C. Petry and Mah17 report a small loss in strength after exposure for 2 h at 1100 C. The causes of degradation are also not clear; surface grooving, structural coarsening17 and local impurity enrichment have been suggested.16 The con- flicting observations could as well be due to differences in the fibre batches used as to experimental differences. 2. The notch sensitivity test As already indicated the main objective of the present study was to investigate the degradation at high temperatures of oxide/oxide, porous matrix composites and in particular through its effect on the notch sensitivity measured in terms of the tensile properties of centre hole panels. This form of notch sensitivity test, though of practical relevance, is somewhat complex in interpretation. In an isotropic, elastically deforming solid, the maximum stress intensification at the edge of a circular notch in an infinitely wide plate loaded in uniaxial tension is a factor of three regardless of hole size.18 Thus the strength of an ideally, fully notch sensitive material will be reduced by a factor of three when such a notch is introduced. A fully notch insensitive material will not be weakened since the stress intensification will be dissipated, for example, by plastic deformation. In reality most materials, including long-fibre reinforced composites, exhibit intermediate behaviour leading to a moderate loss of strength that is generally dependent on hole size. The dependence on hole size arises because the circular notch affects and interacts with the material in its immediate vicinity, thereby altering the distribution of the stress intensification; the volume of affected material scales with the hole size.18 An added complication in experimental situations is that the stress intensification of a circular hole is affected by the width of the tested plate, decreasing from a factor of three to a factor of two as the ratio of hole size to plate width (a/w) approaches 1,19 thus necessitating a finite width correction. It is also to be noted that the stress intensification 566 M.-L. Antti et al. / Journal of the European Ceramic Society 24 (2004) 565–578
M.-L. Antti et al. /Journal of the European Ceramic Society 24(2004)565-578 is altered to some extent by anisotropy of elastic lay-ups (0/90 and +45)were produced from rectan gular plates of 300x 300 mm. Specimens with 0/90 fibre A monolithic ceramic might be expected to be close to orientation were obtained from two plates(A and B) an ideally notch sensitive material. However, here the processed in different batches; +45 specimens were measured strength will depend on the interaction of the taken from a third plate(C) stress field around the hole with the flaw population in The composites were plain weave with a unit cell of the material thus again introducing a notch size effect. approximately 1200 x 900 microns(Fig. 1). The warp Moreover, for such brittle solids the quantitative and weft fibre tows had average widths of 250 and 340 assessment of notch sensitivity is somewhat impractical microns, respectively. Furthermore, they exhibited up to since a unique value of the notch-free strength is difficult 0.5 deviation from the nominal orientation with respect to determine due to the stochastic nature of their to the side of the test bars in both the 0/90 and +450 fracture stress sam A number of semi-empirical models have been pro- posed to describe the notch sensitivity of long fibre, 3.2. Experimental procedure minate composites including the effective crack model proposed by Waddoups et al., I the point stress model2 Tensile tests were performed on centre circular not- and the average stress model. Primarily developed for ched straight-sided specimens with a length of 100 mm polymer matrix composites they have also proved for tests at ambient temperature and 200 mm for tests at to long fibre reinforced ceramics. Above all high-temperature. Their width was 12.5 mm. The cir- they can be used to predict the effect of hole size on the cular centre hole with its axis normal to the Lw surface basis of a limited number of experimental measurements (Fig. 2) was drilled with a carbide drill. The ratio in combination with fitting procedures. More recently, between hole diameter and specimen width(a/w) was more physically realistic models based on crack bridging one of three nominal values (0.1, 0.25 and 0.4). For mechanisms have been presented which also successfully practical reasons the actual values sometimes deviated predict the effects of notch geometry. However, appli- somewhat from the nominal cation of these models requires knowledge of the bridging forces, which usually implies additional experimental measurement. In the present work the results have been interpreted in terms of the Waddoups approach, not only because of its relative computational convenience but also because it was considered to be appropriate for the composite degraded by heat treatment which was assumed to have become embrittled making measure ment of an unnotched strength difficult 3. Experimental details 3.. Materia The material studied was a commercially available composite( Composite Optics, Inc, San Diego, USA) Fig. I. Fibre fabric structure of the composites. Loading direction + consisting of Nextel 720 fibres in a porous aluminosili cate matrix in the form of 3 mm thick plates. The plates LW surface consisted of 12 0/90 woven layers, with a density of 2.6 g/cm and a fibre volume of approximately 48% The fibre fabric is infiltrated with the matrix in a sol-gel process. After drying with a so-called vacuum-bag technique under low pressure and low temperature, the LT surfac composite is pressureless sintered. 3 Intermediate re- infiltration of pyrolysis steps are not necessary. The A⊙ matrix in these types of composites is characterised by a ansverse loadin porosity level around 30-40%o which renders the direction x Parallel loading matrix sufficiently weak to enable damage tolerance during loading. Test specimens of two different fibre Fig. 2. Definition of test bar surfaces
is altered to some extent by anisotropy of elastic properties.20 A monolithic ceramic might be expected to be close to an ideally notch sensitive material. However, here the measured strength will depend on the interaction of the stress field around the hole with the flaw population in the material thus again introducing a notch size effect. Moreover, for such brittle solids the quantitative assessment of notch sensitivity is somewhat impractical since a unique value of the notch-free strength is difficult to determine due to the stochastic nature of their fracture stress. A number of semi-empirical models have been proposed to describe the notch sensitivity of long fibre, laminate composites including the effective crack model proposed by Waddoups et al.,21 the point stress model22 and the average stress model.22 Primarily developed for polymer matrix composites they have also proved applicable to long fibre reinforced ceramics. Above all they can be used to predict the effect of hole size on the basis of a limited number of experimental measurements in combination with fitting procedures. More recently, more physically realistic models based on crack bridging mechanisms have been presented which also successfully predict the effects of notch geometry.23 However, application of these models requires knowledge of the bridging forces, which usually implies additional experimental measurement. In the present work the results have been interpreted in terms of the Waddoups approach, not only because of its relative computational convenience but also because it was considered to be appropriate for the composite degraded by heat treatment which was assumed to have become embrittled making measurement of an unnotched strength difficult. 3. Experimental details 3.1. Material The material studied was a commercially available composite (Composite Optics, Inc, San Diego, USA) consisting of Nextel 720 fibres in a porous aluminosilicate matrix in the form of 3 mm thick plates. The plates consisted of 12 0/90 woven layers, with a density of 2.6 g/cm3 and a fibre volume of approximately 48%. The fibre fabric is infiltrated with the matrix in a sol-gel process. After drying with a so-called vacuum-bag technique under low pressure and low temperature, the composite is pressureless sintered.13 Intermediate reinfiltration of pyrolysis steps are not necessary. The matrix in these types of composites is characterised by a porosity level around 30–40%10 which renders the matrix sufficiently weak to enable damage tolerance during loading. Test specimens of two different fibre lay-ups (0/90 and 45) were produced from rectangular plates of 300300 mm. Specimens with 0/90 fibre orientation were obtained from two plates (A and B) processed in different batches; 45 specimens were taken from a third plate (C). The composites were plain weave with a unit cell of approximately 1200900 microns (Fig. 1). The warp and weft fibre tows had average widths of 250 and 340 microns, respectively. Furthermore, they exhibited up to 0.5 deviation from the nominal orientation with respect to the side of the test bars in both the 0/90 and 45 samples. 3.2. Experimental procedure Tensile tests were performed on centre circular notched straight-sided specimens with a length of 100 mm for tests at ambient temperature and 200 mm for tests at high-temperature. Their width was 12.5 mm. The circular centre hole with its axis normal to the LW surface (Fig. 2) was drilled with a carbide drill. The ratio between hole diameter and specimen width (a/w) was one of three nominal values (0.1, 0.25 and 0.4). For practical reasons the actual values sometimes deviated somewhat from the nominal. Fig. 2. Definition of test bar surfaces. Fig. 1. Fibre fabric structure of the composites. Loading direction $. M.-L. Antti et al. / Journal of the European Ceramic Society 24 (2004) 565–578 567
M.L. Antti et al. Journal of the European Ceramic Society 24(2004)565-578 pecimens were heat-treated in a box furnace at tem- the matrix was investigated as a function of heat-treat peratures of 500, 1000 and 1100C for times between 20 ment time and temperature, using a digital microhard- and 3240 h. The holes were drilled before heat-treat- ness indenter(Matsuzawa MXT-a)at a load of 100 g ment but the dimensions used for the stress calculation and a loading time of 15 s were measured afterwards. Both as-processed and ther- Test specimens before and after thermal exposure mally exposed test specimens were tensile tested either at were investigated with X-ray diffraction (XRD). The X- room or elevated temperature using a servohydraulic ray spectra were obtained between 10 and 90 20 in step mechanical testing machine (MTS 810, Eden Prairie, intervals of 0.03 20 at a rate of 1.5 s/step. An automatic MN, USA)equipped with hydraulically actuated grips divergence slit was used with a beam area of 12x 16 mm and a compact furnace with Sic heating elements. The which thus covered both matrix and fibre regions of the tensile tests were performed at a constant cross-head composite. High-temperature XRd has also been per- displacement rate of 10 um/s while the deformation of formed on pure Nextel 720 fibres, in a powder X-ray the specimen was measured over a 25 mm gauge section diffractometer(Philips Pw 1710) with a step of 0.03 20 using a strain gauge extensometer at ambient tempera- at a rate of 8 S/step 4 After crushing the fibres to a tures, and a low-contact force capacitance extensometer powder heating cycles up to 1400C were performed with at elevated temperatures a heating rate of 5C per minute and holding times of 10 The tested specimens were embedded in epoxy, min every 150C to perform appropriate angular scans. polished and examined using optical and scanning elec- Selected samples were also examined using Raman tron microscopy (SEM), along axes both transverse and spectroscopy. A Dilor xY 800 triple stage Raman parallel (LT and TW surfaces) to the loading direction microprobe (Y, Inc, Edison, NJ)and an Innova 308C (see the schematic diagram, Fig. 2). The fracture sur- Argon ion laser(Coherent, Inc, Santa Clara, CA, USA) faces of both fibre orientations were studied in the operating at 514.5 nm with a 300 mW output power SEM, both directly and embedded in epoxy and were used to record Raman spectra from the fibres and polished on the LW surface Density and porosity mea the matrix separately. The laser was focuse d onto areas surements were performed using the Archimedes of interest with an optical objective providing a spatial method as well as image analysis. The microhardness of resolution of 2 um Table I Properties of 0/90 fibre orientation Full section strength [MPa stiffness [GPa As-received A (RT) 0.34 0.26 As-received B (RT) 0.30 0.28 0.29 200 h at C(B) 0.29 100 h at C(B) 0.28 0.29 0.19 3240hatl000°C(B) 0.15 0.1 0.11 0 h at 1100°C(A) 0.11 0.17 0.25 0.15 0.39 l00 h at 1100°C(A)s Asrec test at 1000C(B) As-rec test at 1100C(A)
Specimens were heat-treated in a box furnace at temperatures of 500, 1000 and 1100 C for times between 20 and 3240 h. The holes were drilled before heat-treatment but the dimensions used for the stress calculation were measured afterwards. Both as-processed and thermally exposed test specimens were tensile tested either at room or elevated temperature using a servohydraulic mechanical testing machine (MTS 810, Eden Prairie, MN, USA) equipped with hydraulically actuated grips and a compact furnace with SiC heating elements. The tensile tests were performed at a constant cross-head displacement rate of 10 mm/s while the deformation of the specimen was measured over a 25 mm gauge section using a strain gauge extensometer at ambient temperatures, and a low-contact force capacitance extensometer at elevated temperatures. The tested specimens were embedded in epoxy, polished and examined using optical and scanning electron microscopy (SEM), along axes both transverse and parallel (LT and TW surfaces) to the loading direction (see the schematic diagram, Fig. 2). The fracture surfaces of both fibre orientations were studied in the SEM, both directly and embedded in epoxy and polished on the LW surface. Density and porosity measurements were performed using the Archimedes method as well as image analysis. The microhardness of the matrix was investigated as a function of heat-treatment time and temperature, using a digital microhardness indenter (Matsuzawa MXT-a) at a load of 100 g and a loading time of 15 s. Test specimens before and after thermal exposure were investigated with X-ray diffraction (XRD). The Xray spectra were obtained between 10 and 90 2 in step intervals of 0.03 2 at a rate of 1.5 s/step. An automatic divergence slit was used with a beam area of 1216 mm which thus covered both matrix and fibre regions of the composite. High-temperature XRD has also been performed on pure Nextel 720 fibres, in a powder X-ray diffractometer (Philips PW 1710) with a step of 0.03 2 at a rate of 8 s/step.24 After crushing the fibres to a powder heating cycles up to 1400 C were performed with a heating rate of 5 C per minute and holding times of 10 min every 150 C to perform appropriate angular scans. Selected samples were also examined using Raman spectroscopy. A Dilor XY 800 triple stage Raman microprobe (JY, Inc, Edison, NJ) and an Innova 308C Argon ion laser (Coherent, Inc., Santa Clara, CA, USA) operating at 514.5 nm with a 300 mW output power were used to record Raman spectra from the fibres and the matrix separately. The laser was focused onto areas of interest with an optical objective providing a spatial resolution of 2 mm. Table 1 Properties of 0/90 fibre orientation Sample a/W Net-section strength [MPa] Full section stiffness [GPa] Hardness HV Strain to failure [%] As-received A (RT) 0.10 201 72 – 0.34 0.26 203 70 0.28 0.39 197 62 0.26 As-received B (RT) 0.10 204 68 204 0.34 0.30 179 62.5 0.28 0.42 191 54 0.29 200 h at 500 C (B) 0.10 203a 74 – 0.32 0.29 191 62 0.29 100 h at 1000 C (B) 0.10 180 72 213 0.28 0.29 172 66 0.24 0.42 149 57 0.19 3240 h at 1000 C (B) 0.10 114 80 323 0.15 0.29 107 74 0.12 0.42 102 66 0.11 20 h at 1100 C (A) 0.11 125 81 334 0.17 0.25 113 73 0.15 0.39 114 69 0.12 100 h at 1100 C (A)s 0.10 54.5 90 457 0.06 0.24 37 83 0.04 0.38 39 74 0.04 As-rec test at 1000 C (B) 0.09 210 68 – 0.35 0.29 171 61 0.24 0.40 186 56 0.24 As-rec test at 1100 C (A) 0.09 189 57 – 0.35 0.24 164 63 0.24 0.38 154 53 0.18 a Failed at grips. 568 M.-L. Antti et al. / Journal of the European Ceramic Society 24 (2004) 565–578
M.L. Antti et al. Journal of the European Ceramic Society 24(2004)565-578 4. Results heat-treated at 1000 oC. after 100 hours at 1100 oc the strength fell to less than one third of the as-received 4.. Stress-strain behaviour strength. The 0/90 material lost stiffness when tested at 1100oC, but after heat-treatment the stiffness at room Tables I and 2 summarise the tensile strength, stiff- temperature increased. The strain to failure of the sam- ness, strain to failure and matrix hardness values of the ples with 0/90 fibre orientation was severely decreased tested samples for 0/90 and +45 orientation, respec- after heat-treatment indicating embrittlement of the tively. Each set of tensile bars contained a replicate material samples to assess reproducibility. This was found to be The +45 material increased in strength after heat atisfactory treatment, but the fracture sequences were abrupt and Representative sets of stress-strain curves for samples violent compared to the as-received fracture which was with a width of 12.5 mm and an a/w ratio of 0. 25 are a gradual process, indicating an embrittlement after shown in Fig. 3(0/90 fibre orientation) and Fig. 4 heat-treatment. The stiffness also increased significantly (+45 fibre orientation). It can be seen that the material as a result of heat-treatment. It is to be noted that with with 0/90 fibre orientation lost some strength when thermal exposure the properties of the 0/90 and +45 Table 2 Properties of±4s° fibre orientation@ plate C Net-section strength [MPa] Stiffness [ GPa Strain to failure [% As-received (RT) 0.10 200hat500°C 10 0.29 l00hatl000°C 0.09 66.5 0.24 0.29 0.40 62.5 3240hatl000°C 0.10 0.29 0.40 0.17 100hatl100°C 0.10 0.29 93.5 0.40 0.15 As- rec test at I000°C 0.10 0.20 Failed at grips, because plate C was not perfectly flat which induced bending forces when clamped 装150 目 200hrs500 50 OOhrs10 00,0sD,10,1s Strain % o] Fig 3. Representative stress-strain curves of 0/90 samples
4. Results 4.1. Stress–strain behaviour Tables 1 and 2 summarise the tensile strength, stiff- ness, strain to failure and matrix hardness values of the tested samples for 0/90 and 45 orientation, respectively. Each set of tensile bars contained a replicate samples to assess reproducibility. This was found to be satisfactory. Representative sets of stress–strain curves for samples with a width of 12.5 mm and an a/w ratio of 0.25 are shown in Fig. 3 (0/90 fibre orientation) and Fig. 4 (45 fibre orientation). It can be seen that the material with 0/90 fibre orientation lost some strength when heat-treated at 1000 C; after 100 hours at 1100 C the strength fell to less than one third of the as-received strength. The 0/90 material lost stiffness when tested at 1100 C, but after heat-treatment the stiffness at room temperature increased. The strain to failure of the samples with 0/90 fibre orientation was severely decreased after heat-treatment indicating embrittlement of the material. The 45 material increased in strength after heattreatment, but the fracture sequences were abrupt and violent compared to the as-received fracture which was a gradual process, indicating an embrittlement after heat-treatment. The stiffness also increased significantly as a result of heat-treatment. It is to be noted that with thermal exposure the properties of the 0/90 and 45 Table 2 Properties of 45 fibre orientation (plate C) Sample a/w Net-section strength [MPa] Stiffness [GPa] Strain to failure [%] As-received (RT) 0.10 64 57 0.28 0.24 61 52 0.19 0.43 69 47 0.16 200 h at 500 C 0.10 61 53 0.25 0.29 64.5 51 0.17 100 h at 1000 C 0.09 66.5 56.5 0.24 0.29 66 57 0.14 0.40 62.5 49 0.15 3240 h at 1000 C 0.10 91 62 0.27 0.29 91 63 0.16 0.40 96.5 58 0.17 100 h at 1100 C 0.10 94 69 0.25 0.29 93.5 66 0.16 0.40 87 56 0.15 As-rec test at 1000 C 0.10 68a 48 0.20 0.30 81a 45 0.22 0.40 95 40 0.41 a Failed at grips, because plate C was not perfectly flat which induced bending forces when clamped. Fig. 3. Representative stress–strain curves of 0/90 samples. M.-L. Antti et al. / Journal of the European Ceramic Society 24 (2004) 565–578 569
lic society24(2004)56-578 materials approach each other. Most remarkable was (taken from a parallel study of load cycling)- which the large increase in strength of the +45 material, with indicates that the knee is associated partly with a corresponding increase in strain to failure when tested reduction in stifness and partly with an irreversible strain increment. Indication of shear strain damage Whereas the 0/90 samples fractured by fibre bundle occurring diagonally from the hole in the 0/90 compo- fracture and pull-out (Fig. 5 a), the samples with +4 sites prior to failure was provided by a parallel study fibre orientation fractured with little or no indication of using thermal emission- but also here by diagonal bundle fracture(Fig. 5 b). The fracture is presumably able stepped fracture paths in which individual fibre bundles to occur by matrix failure followed by ply separation failed in a tensile mode but away from the centre plane The 0/90 stress-strain curves exhibited a relatively of the sample(Fig 5a) distinct fall in slope (knee) situated in the room-tem- In Fig. 7 the effect of the heat-treatments on strength perature tests at a net stress level of between 30 and 50 are presented in the form of a Larson-Miller plot. For MPa and a strain of between 0.03 and 0.06%. This net treatments above 500 oc the results(for stress level corresponded to a theoretical intensified follow a uniform trend which makes it possible to pre- stress at the hole of 70-135 MPa Interestingly, the knee dict strength values for heat-treatments at other times stress and strain were somewhat higher at 1000 and and/ or temperatures. The values from Jurf and But 1 100C(60-90 MPa and 0.07-0.15%, respectively). The ner, also included in the figure show good agreement knee is associated with damage in the composite Fig. 6 with the present results. A minor difference is seen for shows the stress-strain curve of a load-cycled sample higher temperatures and or times, where the material in 02 20 l00 hrs at 1000°C 00050,10150,20 Strain o Strain[%] Fig 4. Representative stress-strain curves of +45 samples. Fig. 6. Stress-strain curve of a load-cycled sample Fig. 5. As-received samples after tensile testing, showing extensive fibre bundle pull-out. (a)0/90 fibre orientation; (b)+45 fibre orientation
materials approach each other. Most remarkable was the large increase in strength of the 45 material, with a corresponding increase in strain to failure when tested at 1000 C. Whereas the 0/90 samples fractured by fibre bundle fracture and pull-out (Fig. 5 a), the samples with 45 fibre orientation fractured with little or no indication of bundle fracture (Fig. 5 b). The fracture is presumably able to occur by matrix failure followed by ply separation. The 0/90 stress–strain curves exhibited a relatively distinct fall in slope (knee) situated in the room-temperature tests at a net stress level of between 30 and 50 MPa and a strain of between 0.03 and 0.06%. This net stress level corresponded to a theoretical intensified stress at the hole of 70–135 MPa. Interestingly, the knee stress and strain were somewhat higher at 1000 and 1100 C (60–90 MPa and 0.07–0.15%, respectively). The knee is associated with damage in the composite. Fig. 6 shows the stress–strain curve of a load-cycled sample (taken from a parallel study of load cycling)25 which indicates that the knee is associated partly with a reduction in stiffness and partly with an irreversible strain increment. Indication of shear strain damage occurring diagonally from the hole in the 0/90 composites prior to failure was provided by a parallel study using thermal emission25 but also here by diagonal, stepped fracture paths in which individual fibre bundles failed in a tensile mode but away from the centre plane of the sample (Fig. 5a). In Fig. 7 the effect of the heat-treatments on strength are presented in the form of a Larson–Miller plot. For treatments above 500 C the results (for a given a/w) follow a uniform trend which makes it possible to predict strength values for heat-treatments at other times and/or temperatures. The values from Jurf and Butner,13 also included in the figure show good agreement with the present results. A minor difference is seen for higher temperatures and/or times, where the material in Fig. 5. As-received samples after tensile testing, showing extensive fibre bundle pull-out. (a) 0/90 fibre orientation; (b) 45 fibre orientation. Fig. 4. Representative stress–strain curves of 45 samples. Fig. 6. Stress–strain curve of a load-cycled sample. 570 M.-L. Antti et al. / Journal of the European Ceramic Society 24 (2004) 565–578
M.L. Antti et al. Journal of the European Ceramic Society 24(2004)565-578 this study shows a larger decrease in strength. This can as providing a means of strength prediction via Eq (1) at least partly be explained by the fact that Jurf and Thus the embrittlement of a composite can be expected Butner reported unnotched strength to lead to a reduction of both Ke and co In order to apply Eq.(1)to the present strength 4.2. Notch sensitivity values. these were first orrected for the finite width effect by applying the factor K/3, where K is the stress intensity As indicated earlier, the tensile fracture stress results factor for a hole in a plate with finite width: for the 0/90 materials were analysed in terms of the Waddoups model. The basis of the model is that the K=3.00-3 13(9)+3.66(9)-1.53( fracture stress, oF, of a centre-hole notched infinite plate is given by The corrected stress value is in effect the predicted strength of an infinite plate. The width- corrected value (1) for a nominal a/w=0. 25 are listed in Table 3 and for all a/w they are presented in Fig 8 The corrected values were used with Eq.(1) to find where a is the hole diameter, Ke is a critical stress the Ke and co values for each treatment condition. The ntensity factor and co is the length of each of two three a/w geometries for each condition permitted three cracks on opposite sides of and adjacent to the hole. independent solutions for these constants. In almost all In fibre composites it is assumed that the cracks are cases the values obtained lay very close to each other equivalent to the damage zones adjacent to the holes. It thus justifying the assumption that they were constant is also assumed that both Ke and co are constants, that is within the range of the experiments. Average values for ndependent of hole size, for a given material. The two the various heat treatments are included in Table 3 parameters cannot readily be related to actual physical together with the unnotched strength values estimated processes in the material but provided that they are by inserting the values of the constants into Eq (1)with found experimentally to indeed be constants then they offer a convenient means of comparing materials as well stern 150 -08 210225103 35104 4.510 LM=T(25+logt) [K-h a [ma Fig. 7. Larson-Miller plot of strength of heat-treated samples. ncluding literature data. 3The results are normalised with respect to ig. 8. Net section strength. corrected for finite width. versus hole oom temperature stren diameter for 0/90 fibre orientation. Table 3 Strength values corrected for finite width, and the results of the application of Waddoups model Corrected net-section Kc/] Estimated unnotched strength[MPa strength[MPa Fibre orientation: 0/900 Received A(RT) Sreceived B (RT) 17.5 210 200hat500°C(B) l00 h at 1000°C(B) 14.4 240hatl000°C(B) 20hatl100°C(A) l00hatl100°C(A) For nominal a/w=0.25
this study shows a larger decrease in strength. This can at least partly be explained by the fact that Jurf and Butner reported unnotched strength. 4.2. Notch sensitivity As indicated earlier, the tensile fracture stress results for the 0/90 materials were analysed in terms of the Waddoups model. The basis of the model is that the fracture stress, F, of a centre-hole notched infinite plate is given by: Kc ¼ F ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi a 2 þ c0 r ð1Þ where a is the hole diameter, Kc is a critical stress intensity factor and c0 is the length of each of two cracks on opposite sides of and adjacent to the hole.19 In fibre composites it is assumed that the cracks are equivalent to the damage zones adjacent to the holes. It is also assumed that both Kc and c0 are constants, that is independent of hole size, for a given material. The two parameters cannot readily be related to actual physical processes in the material but provided that they are found experimentally to indeed be constants then they offer a convenient means of comparing materials as well as providing a means of strength prediction via Eq. (1). Thus the embrittlement of a composite can be expected to lead to a reduction of both Kc and c0. In order to apply Eq. (1) to the present strength values, these were first orrected for the finite width effect by applying the factor K/3, where K is the stress intensity factor for a hole in a plate with finite width:19 K ¼ 3:00 3:13 a w þ 3:66 a w 2 1:53 a w 3 ð2Þ The corrected stress value is in effect the predicted strength of an infinite plate. The width-corrected values for a nominal a/w=0.25 are listed in Table 3 and for all a/w they are presented in Fig. 8. The corrected values were used with Eq. (1) to find the Kc and c0 values for each treatment condition. The three a/w geometries for each condition permitted three independent solutions for these constants. In almost all cases the values obtained lay very close to each other thus justifying the assumption that they were constant within the range of the experiments. Average values for the various heat treatments are included in Table 3 together with the unnotched strength values estimated by inserting the values of the constants into Eq. (1) with Fig. 7. Larson–Miller plot of strength of heat-treated samples, including literature data.13 The results are normalised with respect to room temperature strength. Table 3 Strength values corrected for finite width, and the results of the application of Waddoups model21 Sample Corrected net-section strengtha [MPa] KC MPa ffiffiffiffi m p c0 [mm] Estimated unnotched strength [MPa] Fibre orientation: 0/90 As-received A (RT) 163 18.9 2.7 205 As-received B (RT) 140 17.5 2.2 210 200 h at 500 C (B) – – – – 100 h at 1000 C (B) 135 14.4 2.0 182 3240 h at 1000 C (B) 84 8.9 1.7 122 20 h at 1100 C (A) 92 9.8 1.7 134 100 h at 1100 C (A) 30 2.7 0.3 88 a For nominal a/w=0.25. Fig. 8. Net section strength, corrected for finite width, versus hole diameter for 0/90 fibre orientation. M.-L. Antti et al. / Journal of the European Ceramic Society 24 (2004) 565–578 571
M.L. Antti et al. Journal of the European Ceramic Society 24(2004)565-578 a=0. It is seen that the constants indicate a significant 43. Microstructure and fractography embrittlement with increasing treatment time and tem- perature. The embrittlement is particularly noticeable at Fig. 9 shows representative micrographs of the frac- 100C, 100 h. This is seen, not only in the low values tured, as-received 0/90 material. There were numerous of Ke and co but also as a greater notch sensitivity mea- cracks in the matrix perpendicular to the plies(Fig. 9a) sured as the relative loss in fracture strength of notched one example being shown more closely in Fig. 9b. These samples with respect to the unnotched strength. At a/w never penetrated the fibres. Since the density of cracks 0.25 the loss was 65% for the 1100C/100h material was similar in untested material it is supposed that the compared with 20-30% for the other treatments and the majority are shrinkage cracks formed during production aS-received material rather than multiple matrix cracking generated during The constants can in association with Eq (1)be used loading. The shrinkage cracks widened significantly to predict the effect of notch diameter on the strength of during heat-treatment, probably due to shrinkage of the an infinite plate and these predictions are included as matrix(see later). There were large voids in some of the e curves in Fig 8. The closeness of these curves to the fibre-free areas, but also occasionally inside bundles experimental results provides another justification of the These large voids will in the following be denoted mac- applicability of the model ropores. The infiltration into most bundles was effective, Fig 9. As-received material showing shrinkage cracks, and voids(a and b) and examples of complete(c)and incomplete (d) bundle infiltration. Tw urfaces.(a) Overview, optical microscope (b)Shrinkage crack between two fibres (SEM, SED).(c)Successful infiltration in fibre bundle. (SEM SED. (d)Infiltration not complete. (SEM, BED)
a=0. It is seen that the constants indicate a significant embrittlement with increasing treatment time and temperature. The embrittlement is particularly noticeable at 1100 C, 100 h. This is seen, not only in the low values of Kc and c0 but also as a greater notch sensitivity measured as the relative loss in fracture strength of notched samples with respect to the unnotched strength. At a/w =0.25 the loss was 65% for the 1100 C/100h material compared with 20–30% for the other treatments and the as-received material. The constants can in association with Eq. (1) be used to predict the effect of notch diameter on the strength of an infinite plate and these predictions are included as the curves in Fig. 8. The closeness of these curves to the experimental results provides another justification of the applicability of the model. 4.3. Microstructure and fractography Fig. 9 shows representative micrographs of the fractured, as-received 0/90 material. There were numerous cracks in the matrix perpendicular to the plies (Fig. 9a) one example being shown more closely in Fig. 9b. These never penetrated the fibres. Since the density of cracks was similar in untested material it is supposed that the majority are shrinkage cracks formed during production rather than multiple matrix cracking generated during loading. The shrinkage cracks widened significantly during heat-treatment, probably due to shrinkage of the matrix (see later). There were large voids in some of the fibre-free areas, but also occasionally inside bundles. These large voids will in the following be denoted macropores. The infiltration into most bundles was effective, Fig. 9. As-received material showing shrinkage cracks, and voids (a and b) and examples of complete (c) and incomplete (d) bundle infiltration. TW surfaces. (a) Overview, optical microscope. (b) Shrinkage crack between two fibres. (SEM, SEI). (c) Successful infiltration in fibre bundle. (SEM, SEI). (d) Infiltration not complete. (SEM, BEI). 572 M.-L. Antti et al. / Journal of the European Ceramic Society 24 (2004) 565–578
M.L. Antti et al. Journal of the European Ceramic Society 24(2004)565-578 but some of them showed incomplete infiltration(see The fracture surface of the +450 material and after Fig 9c and d) for 100 h at 1100C is shown in Fig Overviews of the microstructure after heat-treatment As mentioned above, at room temperature the samples for 100 h at 1000 and 1 100oC are shown in Figs. 10a separated without bundle breakage, failing by inter- and b, respectively. The heat-treatment caused an laminar shear of the matrix(Fig 5b). After heat-treat- opening of some of the shrinkage cracks and coarsening ment bundle fracture occurred with no fibre bundle of the macropores. Characterisation of the porosity is pull-out giving a brittle impression(Fig. 12) presented in Section 4.4. No fibre bundle fractures were found. The response of +450 material was identical to 4.4. Density and porosity measurements that of the 0/90 material Evidence for shear damage was sought on micro- The density, porosity and microhardness measure- graphs of the LW surfaces of fractured samples of as- ments are summarised in Table 4. The overall dimen received 0/90 material. Damage in the form of crack sions and the weight of the samples and consequently networks in the matrix that could well be attributed to their overall density did not change significantly with shear deformation was observed in zones stretching thermal exposure. Similarly no significant change in the diagonally from the holes in several samples open porosity was detected. However, there was a sig- SEM fractographs of selected fracture surfaces of 0/ nificant change in the nature of the macroporosity; the 90 samples are shown in Fig. 11. The as-received macropores tended to grow larger and some matrix material exhibited considerable fibre pull-out. The cracks opened during exposure(cf Figs. 9, 10a and b) pulled out fibres were smooth and free of attached The macroporosity was therefore characterised quanti matrix consistent with a low ratio of interface debond- tatively on TW sections, defining macropores as all ing energy to fibre fracture energy. After heat-treat- pores with a diameter larger than 22 um or, in the case ment at 1000C, fibre pull-out was still extensive but of elongated pores and cracks, lengths greater than 44 significantly reduced in comparison with that of the as- um. It was found (table 4) that the volume fraction of received material. Moreover traces of matrix material macropores increased with thermal exposure and at sticking to fibre surfaces could be observed as well as 1100oC their average size also increased. An obvious groups of fibres sintered together. Evidence of increased interpretation of these results is that the matrix densified matrix sticking could also be seen in the samples tested by sintering. Instead of leading to overall shrinkage of at high temperature (i.e. after very short thermal expo- the composite which was constrained dimensionally by sure). After heat-treatment for 100 h at 1100C the the fibre skeleton, the densification occurred internally material exhibited negligible fibre pull-out(Fig. 11 c). by growth of existing macropores and matrix cracks Very short pull-out of sintered bundles could be found SEM observation did indicate a reduction in the micro- ig. 11 d) porosity in the matrix from about 30 to 10 vol. but 器 你多;标小 包 Fig. 10. Microstructures of material heat-treated for 100 h ures of 1000C(a)and 1100C(b), respectively (LT surface sections of 0/90 samples).(a) After 100 h at 1000C.(b) After 100 h at 1100
but some of them showed incomplete infiltration (see Fig. 9c and d). Overviews of the microstructure after heat-treatment for 100 h at 1000 and 1100 C are shown in Figs. 10a and b, respectively. The heat-treatment caused an opening of some of the shrinkage cracks and coarsening of the macropores. Characterisation of the porosity is presented in Section 4.4. No fibre bundle fractures were found. The response of 45 material was identical to that of the 0/90 material. Evidence for shear damage was sought on micrographs of the LW surfaces of fractured samples of asreceived 0/90 material. Damage in the form of crack networks in the matrix that could well be attributed to shear deformation was observed in zones stretching diagonally from the holes in several samples. SEM fractographs of selected fracture surfaces of 0/ 90 samples are shown in Fig. 11. The as-received material exhibited considerable fibre pull-out. The pulled out fibres were smooth and free of attached matrix consistent with a low ratio of interface debonding energy to fibre fracture energy. After heat-treatment at 1000 C, fibre pull-out was still extensive but significantly reduced in comparison with that of the asreceived material. Moreover traces of matrix material sticking to fibre surfaces could be observed as well as groups of fibres sintered together. Evidence of increased matrix sticking could also be seen in the samples tested at high temperature (i.e. after very short thermal exposure). After heat-treatment for 100 h at 1100 C the material exhibited negligible fibre pull-out (Fig. 11 c). Very short pull-out of sintered bundles could be found (Fig. 11 d). The fracture surface of the 45 material and after heat-treatment for 100 h at 1100 C is shown in Fig. 12. As mentioned above, at room temperature the samples separated without bundle breakage, failing by interlaminar shear of the matrix (Fig. 5b). After heat-treatment bundle fracture occurred with no fibre bundle pull-out giving a brittle impression (Fig. 12). 4.4. Density and porosity measurements The density, porosity and microhardness measurements are summarised in Table 4. The overall dimensions and the weight of the samples and consequently their overall density did not change significantly with thermal exposure. Similarly no significant change in the open porosity was detected. However, there was a significant change in the nature of the macroporosity; the macropores tended to grow larger and some matrix cracks opened during exposure (cf. Figs. 9, 10a and b). The macroporosity was therefore characterised quantitatively on TW sections, defining macropores as all pores with a diameter larger than 22 mm or, in the case of elongated pores and cracks, lengths greater than 44 mm. It was found (Table 4) that the volume fraction of macropores increased with thermal exposure and at 1100 C their average size also increased. An obvious interpretation of these results is that the matrix densified by sintering. Instead of leading to overall shrinkage of the composite which was constrained dimensionally by the fibre skeleton, the densification occurred internally by growth of existing macropores and matrix cracks. SEM observation did indicate a reduction in the microporosity in the matrix from about 30 to 10 vol.% but Fig. 10. Microstructures of material heat-treated for 100 h at temperatures of 1000 C (a) and 1100 C (b), respectively (LT surface sections of 0/90 samples). (a) After 100 h at 1000 C. (b) After 100 h at 1100 C. M.-L. Antti et al. / Journal of the European Ceramic Society 24 (2004) 565–578 573
M.L. Antti et al. Journal of the European Ceramic Society 24(2004)565-578 this was not possible to confirm with accuracy due the is possible that large voids \ s am therr just beneath the surface extreme fineness of the pores. Densification of the in these cases. In Table 4 are ed average hardnesses matrix was however indicated by an increase in micro- and standard deviations based arness ting these extremely low values. The scatter of the The microhardness of the matrix increased markedly values also increased significantly after thermal expo- with treatment temperature and time at 1000 and sure, indicating that the matrix became inhomogeneous 1 100C(see Table 4). At a few locations on samples This is also indicated in the micrograph in Fig 13 where treated at 1100C and at 1000C for 3240 h very low variations in porosity levels can be seen, the denser hardness values were obtained in association with a areas giving brighter reflection. Several indentations glassy appearance of the matrix and with the formation also indicated matrix embrittlement after heat treatment of cracks from the corners and edges of the indent. by the formation of indentation cracks emanating from Since SEM-EDS analysis of these locations did not indi- indent corners(Fig. 14). It can be noted that the cracks cate any significant deviation in chemical composition it only propagate perpendicular to the fibres and not Fig. Il. Fracture surfaces of as-received and heat-treated samples(a)As-received. (b) Heat-treated 100 h at 1000C.(c) Heat-treated 100 h at 1100C.(d) Heat-treated 100 h at 1100C showing sintered bundle
this was not possible to confirm with accuracy due the extreme fineness of the pores. Densification of the matrix was however indicated by an increase in microhardness. The microhardness of the matrix increased markedly with treatment temperature and time at 1000 and 1100 C (see Table 4). At a few locations on samples treated at 1100 C and at 1000 C for 3240 h very low hardness values were obtained in association with a glassy appearance of the matrix and with the formation of cracks from the corners and edges of the indent. Since SEM–EDS analysis of these locations did not indicate any significant deviation in chemical composition it is possible that large voids lay just beneath the surface in these cases. In Table 4 are listed average hardnesses and standard deviations based on 10 indentations omitting these extremely low values. The scatter of the values also increased significantly after thermal exposure, indicating that the matrix became inhomogeneous. This is also indicated in the micrograph in Fig. 13 where variations in porosity levels can be seen, the denser areas giving brighter reflection. Several indentations also indicated matrix embrittlement after heat treatment by the formation of indentation cracks emanating from indent corners (Fig. 14). It can be noted that the cracks only propagate perpendicular to the fibres and not Fig. 11. Fracture surfaces of as-received and heat-treated samples. (a) As-received. (b) Heat-treated 100 h at 1000 C. (c) Heat-treated 100 h at 1100 C. (d) Heat-treated 100 h at 1100 C showing sintered bundle. 574 M.-L. Antti et al. / Journal of the European Ceramic Society 24 (2004) 565–578