《复合材料 Composites》课程教学资源（学习资料）第五章 陶瓷基复合材料_Effect of static pre-loading on fracture toughness of Nicalon fibre glass matrix composite

AERIALS SHENGE ENGINEERIG SEVIER Materials Science and Engineering A 367(2004)17-23 www.elsevier.com/locate/msea Effect of static pre-loading on fracture toughness of Nicalon fibre glass matrix composite I Dlouhy a,*,Z. Chlupa, K.K. Chawla b, R. Kulkarni b, M. Koopman, A.R. Boccaccinic b Department of Materials Science and Engineering, University of Alabama, Birmingham,AL, USA Department of Materials, Imperial College London, London Sw72BP UK Received 13 May 2003; received in revised form 18 August 2003 Abstract Static fatigue of commercially available unidirectional Sic-fibre reinforced glass matrix composite has been followed by measuring the fracture toughness of samples statically pre-loaded at different stresses and times. Pre-load levels were just below the loads corresponding fracture strength, the(bending) pre-stress being on level 300, 310 and 315 MPa. Three hold times, 25, 50 and 75 h, were applied at each of the stresses. For fracture toughness(Kle)determination, a chevron-notch(CN) three-point bend test was applied The method has been found to be useful to accurately detect and quantify very subtle microdamage on fibre/matrix interfaces caused by static pre-stressing. Despite of the high scatter of the measured Kle values, a decrease in fracture toughness with increasing pre-stress and hold time has been found. A parameter is proposed that may be used to quantify the cumulative and combined effect of pre-stress and hold time on fracture toughness of composit C 2003 Elsevier B V. All rights reserved Keywords: Nicalon"fibre; Glass matrix composite; Fracture toughness; Chevron-notch; Static fatigue 1. Introduction the process zone in the wake of the crack [4-6]. Thus, the crack growth resistance rises as the crack propagates. In Glass and glass-ceramic matrix composites are a these materials. it is difficult to define an intrinsic fracture tively new family of high-temperature capability, light struc- toughness(Klc)as a material parameter due to the character- tural materials exhibiting quasi-ductile fracture behaviour istic rising crack growth resistance curve [5]. Nevertheless, [1]. Interest in these silicate matrix composites arises from an exact method of quantifying the fracture behaviour, in their relative ease of fabrication and potential use at tem- particular fracture toughness, would be an important contri- peratures up to about 1200C, characteristics which allow bution in the development of fibre reinforced brittle matrix their possible applications as critical components, for exam- composites, especially in the assessment of their possible ple, in the aerospace sector [2]. As reviewed recently [3], structural degradation in service other applications of these materials include thermal protec- Many ceramic materials when subjected to a constant load tion systems, metallurgy and glass production, precision and undergo a progressive weakening with time-a phenomenon microengineering, special machinery, vacuum pumps, auto- known as static fatigue. This results in failure under a static motive components as well as in chemical plants. In many load that is lower thannormal,, after some time. This sug- of these applications the materials are likely to work under gests that the defect population is evolving with time fatigue and impact loading conditions [1-3 The chevron-notched (CN) specimen technique is In fibre reinforced ceramic or glass matrix composites, well-established method used to determine the fracture fibre bridging and fibre pull-out mechanisms are the major toughness and the work of fracture of brittle materials arces of toughening [4]. Both these mechanisms in 5,7-9, including particle reinforced glass matrix com- posites [10]. As indicated elsewhere [111 however, the ng author.Tel:+420-5-32-290-342; echnique has not received wide application to measure fax:+420-5-41-218-657 the fracture toughness of fibre reinforced brittle matrix E-mail address: dlouhy@ipm.cz(. Dlouhy ) composites. Other specimen configurations, for examp 0921-5093/s-see front matter 2003 Elsevier B V. All rights reserved doi:10.1016msea.2003.09.058

Materials Science and Engineering A 367 (2004) 17–23 Effect of static pre-loading on fracture toughness of Nicalon® fibre glass matrix composite I. Dlouhy a,∗, Z. Chlup a, K.K. Chawla b, R. Kulkarni b, M. Koopman b, A.R. Boccaccini c a Institute of Physics of Materials ASCR, Žižkova 22, 61662 Brno, Czech Republic b Department of Materials Science and Engineering, University of Alabama, Birmingham, AL, USA c Department of Materials, Imperial College London, London SW7 2BP, UK Received 13 May 2003; received in revised form 18 August 2003 Abstract Static fatigue of commercially available unidirectional SiC-fibre reinforced glass matrix composite has been followed by measuring the fracture toughness of samples statically pre-loaded at different stresses and times. Pre-load levels were just below the loads corresponding to fracture strength, the (bending) pre-stress being on level 300, 310 and 315 MPa. Three hold times, 25, 50 and 75 h, were applied at each of the stresses. For fracture toughness (KIc) determination, a chevron-notch (CN) three-point bend test was applied. The method has been found to be useful to accurately detect and quantify very subtle microdamage on fibre/matrix interfaces caused by static pre-stressing. Despite of the high scatter of the measured KIc values, a decrease in fracture toughness with increasing pre-stress and hold time has been found. A parameter is proposed that may be used to quantify the cumulative and combined effect of pre-stress and hold time on fracture toughness of composites. © 2003 Elsevier B.V. All rights reserved. Keywords: Nicalon® fibre; Glass matrix composite; Fracture toughness; Chevron-notch; Static fatigue 1. Introduction Glass and glass–ceramic matrix composites are a rela￾tively new family of high-temperature capability, light struc￾tural materials exhibiting quasi-ductile fracture behaviour [1]. Interest in these silicate matrix composites arises from their relative ease of fabrication and potential use at tem￾peratures up to about 1200 ◦C, characteristics which allow their possible applications as critical components, for exam￾ple, in the aerospace sector [2]. As reviewed recently [3], other applications of these materials include thermal protec￾tion systems, metallurgy and glass production, precision and microengineering, special machinery, vacuum pumps, auto￾motive components as well as in chemical plants. In many of these applications the materials are likely to work under fatigue and impact loading conditions [1–3]. In fibre reinforced ceramic or glass matrix composites, fibre bridging and fibre pull-out mechanisms are the major sources of toughening [4]. Both these mechanisms increase ∗ Corresponding author. Tel.: +420-5-32-290-342; fax: +420-5-41-218-657. E-mail address: idlouhy@ipm.cz (I. Dlouhy). the process zone in the wake of the crack [4–6]. Thus, the crack growth resistance rises as the crack propagates. In these materials, it is difficult to define an intrinsic fracture toughness (KIc) as a material parameter due to the character￾istic rising crack growth resistance curve [5]. Nevertheless, an exact method of quantifying the fracture behaviour, in particular fracture toughness, would be an important contri￾bution in the development of fibre reinforced brittle matrix composites, especially in the assessment of their possible structural degradation in service. Many ceramic materials when subjected to a constant load undergo a progressive weakening with time—a phenomenon known as static fatigue. This results in failure under a static load that is lower than ‘normal’, after some time. This sug￾gests that the defect population is evolving with time. The chevron-notched (CN) specimen technique is a well-established method used to determine the fracture toughness and the work of fracture of brittle materials [5,7–9], including particle reinforced glass matrix com￾posites [10]. As indicated elsewhere [11], however, the technique has not received wide application to measure the fracture toughness of fibre reinforced brittle matrix composites. Other specimen configurations, for example 0921-5093/$ – see front matter © 2003 Elsevier B.V. All rights reserved. doi:10.1016/j.msea.2003.09.058

I. Dlouhy et al /Materials Science and Engineering A 367(2004)17-23 Table I Properties of the glass matrix, fibres and composite [11, 14, 19 Density(g/cm) Youngs modulus(GPa) Poisson's ratio coefficient(K Matrix DURAN Glass 0.22 Fibre sic nicalon 3.0×10-6 3.1×10-6 raight notched specimens [12-15], have been more exten- Bars of dimensions 4.5 mm x 3.8 mm x 50 mm were cut ively used in these materials. Apart from our own previous from the as-received samples for static fatigue test. The bars research [11, 14, 15], few other reports are available on were loaded in three-point bending prior to notching at se- the applicability of the chevron-notched specimen to fibre lected load levels for holding times of 25, 50, and 75h. The reinforced ceramics and glasses [16, 171 load levels were selected just below the loads correspond The purpose of the present contribution is to analyse ing to the fracture strength, which corresponded to stresses he applicability of the chevron-notch technique for the in the range 300-450 MPa. The separate values are shown assessment of damage produced in a static fatigue test of in Table 2. The(bending)pre-stresses ap have been calcu unidirectional fibre reinforced glass matrix composites. The lated from the applied load and real specimen dimensions fracture behaviour of the composites after static pre-loading by using the standard equation (pre-stressing), under different stresses and hold times, was studied by means of Kle measurements using the p (1) flexural test where F is the applied load, L the support span(14.3 mm), B the thickness(breadth)and W is the height of the specimen 2. Experimental The chevron-notched specimen technique(Fig. 2)was employed for fracture toughness (Kle) determination in The material investigated was a commercially available as-received and pre-stressed samples. Chevron-notches with unidirectional Nicalon(NL202)fibre reinforced borosil- an included angle of 90 were cut using a thin diamond cate duran)glass matrix composite fabricated by Schott wheel(of width less than 0.2 mm), three chevron-notches Glass(Mainz, Germany). The same composite has been were produced in each pre-stressed bar. A three-point bend investigated in previous research works [11, 14]. Nicalon method with a span of 16 mm was employed at a constant fibre consists of SiC, SiO2, and some free carbon [18]. Prop- cross-head speed of 0. I mm/min. Graphs of load versus erties of the matrix, fibre and composite are given in Table 1. deflection were recorded and the maximum force was de- The composite was prepared by the sol-gel-slurry method termined from each trace. The fracture toughness value [19). The samples were received in the form of rectangular was calculated from the maximum load(Fmax)and the test bars of nominal dimensions 4.5mm x3.8 mm x 100 mm. corresponding minimum value of geometrical compliance The density of the composites was 2.4 g/cm and the fibre function(Ymin). The calculation of the function Ymin for volume fraction was 0.4. Fairly regular fibre distribution chevron-notch bend bars was based on the use of Bluhm's (Fig. 1)and the absence of porosity were found by mi- slice model [20]. The detailed procedure used for the pu structural investigations, as shown elsewhere [11, 14] poses of this investigation has been described elsewhere [9] The fracture toughness was calculated from the standard Omar 5 F/2 Fig..Microstructure of the Nicalon" fibre reinforced glass matrix com Fig. 2. Geometry of chevron-notch specimen to determine fracture tough- posite in a section perpendicular to the fibres

18 I. Dlouhy et al. / Materials Science and Engineering A 367 (2004) 17–23 Table 1 Properties of the glass matrix, fibres and composite [11,14,19] Density (g/cm3) Young’s modulus (GPa) Poisson’s ratio Thermal expansion coefficient (K−1) Matrix DURAN® Glass 2.23 63 0.22 3.25 × 10−6 Fibre SiC Nicalon® 2.55 198 0.20 3.0 × 10−6 Composite 2.40 135 0.21 3.1 × 10−6 straight notched specimens [12–15], have been more exten￾sively used in these materials. Apart from our own previous research [11,14,15], few other reports are available on the applicability of the chevron-notched specimen to fibre reinforced ceramics and glasses [16,17]. The purpose of the present contribution is to analyse the applicability of the chevron-notch technique for the assessment of damage produced in a static fatigue test of unidirectional fibre reinforced glass matrix composites. The fracture behaviour of the composites after static pre-loading (pre-stressing), under different stresses and hold times, was studied by means of KIc measurements using the chevron-notch flexural test. 2. Experimental The material investigated was a commercially available unidirectional Nicalon® (NL202) fibre reinforced borosili￾cate (DURAN®) glass matrix composite fabricated by Schott Glass (Mainz, Germany). The same composite has been investigated in previous research works [11,14]. Nicalon® fibre consists of SiC, SiO2, and some free carbon [18]. Prop￾erties of the matrix, fibre and composite are given in Table 1. The composite was prepared by the sol–gel-slurry method [19]. The samples were received in the form of rectangular test bars of nominal dimensions 4.5 mm×3.8 mm×100 mm. The density of the composites was 2.4 g/cm3 and the fibre volume fraction was 0.4. Fairly regular fibre distribution (Fig. 1) and the absence of porosity were found by mi￾crostructural investigations, as shown elsewhere [11,14]. Fig. 1. Microstructure of the Nicalon® fibre reinforced glass matrix com￾posite in a section perpendicular to the fibres. Bars of dimensions 4.5 mm × 3.8 mm × 50 mm were cut from the as-received samples for static fatigue test. The bars were loaded in three-point bending prior to notching at se￾lected load levels for holding times of 25, 50, and 75 h. The load levels were selected just below the loads correspond￾ing to the fracture strength, which corresponded to stresses in the range 300–450 MPa. The separate values are shown in Table 2. The (bending) pre-stresses σp have been calcu￾lated from the applied load and real specimen dimensions by using the standard equation: σp = 3FL 2BW2 (1) where F is the applied load, L the support span (14.3 mm), B the thickness (breadth) and W is the height of the specimen. The chevron-notched specimen technique (Fig. 2) was employed for fracture toughness (KIc) determination in as-received and pre-stressed samples. Chevron-notches with an included angle of 90◦ were cut using a thin diamond wheel (of width less than 0.2 mm), three chevron-notches were produced in each pre-stressed bar. A three-point bend method with a span of 16 mm was employed at a constant cross-head speed of 0.1 mm/min. Graphs of load versus deflection were recorded and the maximum force was de￾termined from each trace. The fracture toughness value was calculated from the maximum load (Fmax) and the corresponding minimum value of geometrical compliance function (Y∗ min). The calculation of the function Y∗ min for chevron-notch bend bars was based on the use of Bluhm’s slice model [20]. The detailed procedure used for the pur￾poses of this investigation has been described elsewhere [9]. The fracture toughness was calculated from the standard equation: KIc = FmaxY∗ min BW1/2 (2) Fig. 2. Geometry of chevron-notch specimen to determine fracture tough￾ness

I. Dlouhy et al /Materials Science and Engineering A 367(2004) Table 2 Primary data of fracture toughness determination on samples that had been subjected to different static fatigue conditions Sample Youngs Pre-load Pre-stress Hold Change in Youngs Fracture toughness number modulus(GPa) time(h) modul Kle(MPam/2 135.3 6.6±1.2 97 9829 6 0000000050 555 .2±1. 3.2±1.8 Failed in less than 24 h 17 138.6 Failed in less than 4 h Failed immediate where B, W have the meaning given above. The chevron-notch depth ao necessary for calculation of min was measured from optical micrographs of fractured specimens An acoustic emission(AE)technique was used during the tests. Traces of the cumulative number of counts(AE events) were obtained in the same time scale as the load versus time plots. Small AE transducer(having circular contact surface of diameter 4 mm) enabled direct application on specimen side surface. One channel acoustic emission kit was used for sampling the events during the experimental set followed ◇ in this investigation. This technique allowed for an accurate detection of the microcrack initiation at the chevron-notch which corresponded to a sharp increase in the number of Ae 8 events. Valid measurements for computing Kle were those in which this increase of Ae events coincided with the end of the linear part of the force versus time trace Scanning electron microscopy(SEM) was used to analyse Pre-stress [MPa] 3. Results A summary of pre-loading parameters and other measured data are given in Table 2. Fracture toughness values given here show the mean value and standard deviation from three tests. The Klc values measured on the materials with the 266-28. MPam /2 essIng(300 MPa, 25 h), in the range lowest level of pre-stre 留g巴号 vious data obtained by the same chevron-notched specimen technique on similar composites [111 Fracture toughness data obtained by using chevron-notch three-point bend tests are shown in Fig. 3. The values are shown here in relation to the level of pre-stress applied dur- Pre-stress [MPa ing static fatigue test. In Fig. 3(a)all data that have been generated for all pre-stresses and hold times applied are plot- Fig3.(a) Effect of pre-stress on fracture toughness of the investigated ted. A set of fracture toughness values obtained for hold composite for all pre-stresses and hold times applied. The dashed curve time 75 h and different pre-stresses is shown in Fig. 3(b) fracture toughness of the investigated composite for the highest applied The data are characterised by high scatter, which will be dis- hold time(75 h). The full curve represents an exponential fit through the cussed further below. Because only three values have been data corresponding to the hold time

I. Dlouhy et al. / Materials Science and Engineering A 367 (2004) 17–23 19 Table 2 Primary data of fracture toughness determination on samples that had been subjected to different static fatigue conditions Sample number Young’s modulus (GPa) Pre-load (N) Pre-stress (MPa) Hold time (h) Change in Young’s modulus Fracture toughness KIc (MPa m1/2) 7 133.8 977.0 300 25 No 28.8 ± 3.3 8 135.3 972.1 300 25 No 28.7 ± 2.3 9 135.6 987.9 300 25 No 26.6 ± 1.2 3 136.4 967.1 300 50 No 30.9 ± 2.4 4 143.9 977.0 300 50 No 26.1 ± 2.7 16 138.5 1004.5 310 50 No 25.7 ± 1.2 5 134.9 982.9 300 75 No 27.3 ± 2.2 6 135.1 977.0 300 75 No 25.1 ± 1.5 15 135.7 1009.6 310 75 No 24.2 ± 1.0 18 135.0 1025.9 315 75 No 23.2 ± 1.8 2 133.2 1074.7 330 Failed in less than 24 h 17 138.6 1031.6 320 Failed in less than 4 h 1 136.6 1433.0 450 Failed immediately where B, W have the meaning given above. The chevron-notch depth a0 necessary for calculation of Y∗ min was measured from optical micrographs of fractured specimens. An acoustic emission (AE) technique was used during the tests. Traces of the cumulative number of counts (AE events) were obtained in the same time scale as the load versus time plots. Small AE transducer (having circular contact surface of diameter 4 mm) enabled direct application on specimen side surface. One channel acoustic emission kit was used for sampling the events during the experimental set followed in this investigation. This technique allowed for an accurate detection of the microcrack initiation at the chevron-notch, which corresponded to a sharp increase in the number of AE events. Valid measurements for computing KIc were those in which this increase of AE events coincided with the end of the linear part of the force versus time trace. Scanning electron microscopy (SEM) was used to analyse the fracture surfaces. 3. Results A summary of pre-loading parameters and other measured data are given in Table 2. Fracture toughness values given here show the mean value and standard deviation from three tests. The KIc values measured on the materials with the lowest level of pre-stressing (300 MPa, 25 h), in the range 26.6–28.8 MPa m1/2, are in very good agreement with pre￾vious data obtained by the same chevron-notched specimen technique on similar composites [11]. Fracture toughness data obtained by using chevron-notch three-point bend tests are shown in Fig. 3. The values are shown here in relation to the level of pre-stress applied dur￾ing static fatigue test. In Fig. 3(a) all data that have been generated for all pre-stresses and hold times applied are plot￾ted. A set of fracture toughness values obtained for hold time 75 h and different pre-stresses is shown in Fig. 3(b). The data are characterised by high scatter, which will be dis￾cussed further below. Because only three values have been Fig. 3. (a) Effect of pre-stress on fracture toughness of the investigated composite for all pre-stresses and hold times applied. The dashed curve represents an exponential fit through all data. (b) Effect of pre-stress on fracture toughness of the investigated composite for the highest applied hold time (75 h). The full curve represents an exponential fit through the data corresponding to the hold time

1. Dlouhy et al /Materials Science and Engineering A 367(2004)17-23 available for each set of conditions no statistical analysis for different pre-stressed samples has been carried out in this in- 73 vestigation. The experimental values have been fitted by an exponential curve, shown in Fig 3(a)by a dashed line. Sim- ilarly, for data obtained at one particular hold time(75 h), he exponential fit is shown in Fig. 3(b) by a full curve For comparison, the curve from Fig. 3(a) is also plotted Fig 3(b)as a dashed line Taking into account all data sets for different pre-loading conditions. the decrease of fracture toughness with increas- ng static pre-stress values can be clearly seen. A similar trend is observed if one selects a constant holding time Deflection [um or example, for 75 h the negative effect of the pre-stress is evident from Fig. 3(b). Note that the loads applied during Fig. 5. Load deflection traces obtained from three-point bend test for low pre-stressing were slightly below the load corresponding to 315 MPa for 75 h) showing also the cumulative traces of AE events.The arrows indicate different fracture stages, as discussed in the text. Another plot shows the dependence of fracture toughness on hold time at a given pre-stress, as seen in Fig. 4.An exponential fit for all the data set is shown in this figure by traces. Some typical curves are shown in Fig. 5, for a he solid curve. In order to show the effect of hold time at specimen with low pre-stressing(the lowest load and hold a given pre-stress level the exponential fit was obtained for time applied) on the left side and for specimen with high data collected at 310 MPa. This fit is plotted by a dashed pre-stressing(the highest load applied) on the right side curve in the same Fig. 4 Analysis of the deflection traces indicates that in both cases Also, in this case there is a decrease in fracture tough- the crack developed at about two-third of maximum load ess values with increasing hold time at all pre-stress levels. at the chevron-notch tip and it propagated in a controlled When selecting one particular pre-stress value (310 MPa), manner up to maximum load after which unstable fracture a qualitatively similar trend is seen despite the significant occurred. The acoustic emission technique indicates that scatter of fracture toughness data individual microfracture events started at about two-third It can thus be concluded that fracture toughness decreased of the maximum load, as demonstrated by an increase in with increasing severity of the pre-stressing conditions, e.g. the cumulative number of AE events(marked by arrow I in increasing pre-stress and hold time. It also follows from Fig. 5). This increase in the AE is so high that crack propa Figs. 3 and 4 that both parameters investigated, i.e. pre-stress gation through the glass matrix and fibre/matrix debonding and hold time, have a similar effect on fracture toughness and fracture are thought to be responsible for this signif values, described empirically by an exponential dependence. cant effect, as discussed in a previous paper [ 15]. a detailed To further analyse the data additional observations were analysis of the traces in Fig. 5 indicates that deviation fro made, in particular a detailed analysis of load deflection linearity occurs at this stage. An increase in the number of AE events observed at the end of the increasing part of the load-deflection traces(arrows "2 in Fig. 5)indicates that the actual crack has developed at the chevron-notch tip. In these cases, the measurements of Klc can be taken as valid since the unstable fracture(at maximum force)occurs from a propagating crack perpendicular to the fibre axis [15]. No qualitative differences were observed when comparing the behaviour of samples that had been under different static pre-loading levels, as can be seen from comparison of both load deflection traces and curves of cumulative number of E events in Fig. 5 It can be speculated that the only irreversible microstruc tural damage that can be introduced in the sample during static fatigue is localised microcracking in the vicinity of the fibre matrix interface boundary. A larger extent of such Hold time [hI kind of damage could lead to decrease of load transfer from points.Full cure different pre-stress levels distinguished by different the Young's modulus. However, it follows from the anam Fig. 4. effe time at the different pre-stresses on fracture tough- the matrix to the fibre and this effect is likely to decrea ness. Dat presents an exponential fit for all data and dashed sis of the linear part of the load deflection traces(e.g.in urve for samples pre-stressed at 310 MPa only Fig. 5)that no important differences in modulus should be

20 I. Dlouhy et al. / Materials Science and Engineering A 367 (2004) 17–23 available for each set of conditions no statistical analysis for different pre-stressed samples has been carried out in this in￾vestigation. The experimental values have been fitted by an exponential curve, shown in Fig. 3(a) by a dashed line. Sim￾ilarly, for data obtained at one particular hold time (75 h), the exponential fit is shown in Fig. 3(b) by a full curve. For comparison, the curve from Fig. 3(a) is also plotted in Fig. 3(b) as a dashed line. Taking into account all data sets for different pre-loading conditions, the decrease of fracture toughness with increas￾ing static pre-stress values can be clearly seen. A similar trend is observed if one selects a constant holding time. For example, for 75 h the negative effect of the pre-stress is evident from Fig. 3(b). Note that the loads applied during pre-stressing were slightly below the load corresponding to specimen fracture. Another plot shows the dependence of fracture toughness on hold time at a given pre-stress, as seen in Fig. 4. An exponential fit for all the data set is shown in this figure by the solid curve. In order to show the effect of hold time at a given pre-stress level the exponential fit was obtained for data collected at 310 MPa. This fit is plotted by a dashed curve in the same Fig. 4. Also, in this case there is a decrease in fracture tough￾ness values with increasing hold time at all pre-stress levels. When selecting one particular pre-stress value (310 MPa), a qualitatively similar trend is seen despite the significant scatter of fracture toughness data. It can thus be concluded that fracture toughness decreased with increasing severity of the pre-stressing conditions, e.g. increasing pre-stress and hold time. It also follows from Figs. 3 and 4 that both parameters investigated, i.e. pre-stress and hold time, have a similar effect on fracture toughness values, described empirically by an exponential dependence. To further analyse the data additional observations were made, in particular a detailed analysis of load deflection Fig. 4. Effect of hold time at the different pre-stresses on fracture tough￾ness. Data for the different pre-stress levels distinguished by different points. Full curve represents an exponential fit for all data and dashed curve for samples pre-stressed at 310 MPa only. Fig. 5. Load deflection traces obtained from three-point bend test for low and high pre-stressed samples (no. 7 3 at 300 MPa for 25 h, no. 18 1 at 315 MPa for 75 h) showing also the cumulative traces of AE events. The arrows indicate different fracture stages, as discussed in the text. traces. Some typical curves are shown in Fig. 5, for a specimen with low pre-stressing (the lowest load and hold time applied) on the left side and for specimen with high pre-stressing (the highest load applied) on the right side. Analysis of the deflection traces indicates that in both cases the crack developed at about two-third of maximum load at the chevron-notch tip and it propagated in a controlled manner up to maximum load after which unstable fracture occurred. The acoustic emission technique indicates that individual microfracture events started at about two-third of the maximum load, as demonstrated by an increase in the cumulative number of AE events (marked by arrow 1 in Fig. 5). This increase in the AE is so high that crack propa￾gation through the glass matrix and fibre/matrix debonding and fracture are thought to be responsible for this signifi- cant effect, as discussed in a previous paper [15]. A detailed analysis of the traces in Fig. 5 indicates that deviation from linearity occurs at this stage. An increase in the number of AE events observed at the end of the increasing part of the load–deflection traces (arrows ‘2’ in Fig. 5) indicates that the actual crack has developed at the chevron-notch tip. In these cases, the measurements of KIc can be taken as valid, since the unstable fracture (at maximum force) occurs from a propagating crack perpendicular to the fibre axis [15]. No qualitative differences were observed when comparing the behaviour of samples that had been under different static pre-loading levels, as can be seen from comparison of both load deflection traces and curves of cumulative number of AE events in Fig. 5. It can be speculated that the only irreversible microstruc￾tural damage that can be introduced in the sample during static fatigue is localised microcracking in the vicinity of the fibre matrix interface boundary. A larger extent of such kind of damage could lead to decrease of load transfer from the matrix to the fibre and this effect is likely to decrease the Young’s modulus. However, it follows from the analy￾sis of the linear part of the load deflection traces (e.g. in Fig. 5) that no important differences in modulus should be

I. Dlouhy et al/Materials Science and Engineering A 367(2004) 0.5mm 0.5mm ture surfaces of the as-received(a) and high pre-stressed samples(b)(no. 18-1 at 315 MPa for 75 h) showing extensive and uniform fibre expected. Also the data shown in Table 2 reveal that there conditions before reaching the flexural strength value. Sim- fracture toughness values measured by the chevron-notch fore the crack initiation, i.e. just before the stage indicated test seem to be independent of Youngs modulus, at least for in Fig. 5 by arrow 1 is reached the data set generated in this investigation A large amount of such microdamage leads to decrease of Typical fracture surfaces of chevron-notched samples are load transfer from the matrix to the fibre. If composite does shown in Fig. 6. Fig. 6(a) shows the fracture surface of not tend to fracture (i.e. no additional load is applied)this specimen in as-received condition, as characterised in an effect is likely to decrease the Youngs modulus. This was arlier study [21] whereas a fracture surface of a composite not the case for the specimens pre-stressed at conditions de- after pre-stressing at 315 MPa for 75 h is shown in Fig. 6(b). scribed in our investigation. There was no change in modu- It follows from fracture surface analyses of as-received, low lus with pre-stress and hold time. The Young's modulus ap- and high pre-stressed specimens that no substantial change in peared to be independent of applied pre-stress level and hold fracture surface morphology occurred. Besides the extensive time(Table 2), at least under the accuracy of the method fibre pull-out, typical of this kind of composites [1l], the employed for its determination(from load-deflection data) crack propagation from the chevron-notch tip was observed The only characteristic that has been found to be sensi The fracture surface of the chevron-notched specimens was tive to pre-stressing condition was fracture toughness. This uniform in all cases in the sense that the matrix fracture has is due to the high susceptibility of the chevron-notch tech taken place mainly in one plane(mode I crack) nique to the material behaviour in the small volume at the chevron-notch tip. This has been convincingly shown re- cently when investigating the effect of stress relief treatment 4. Discussion on fracture toughness in pure borosilicate glass [23].Thus the chevron-notch fracture toughness has been found in the Generally, in the stress-strain response of ceramic ma- present investigation as a very accurate method enabling to trix composites, the matrix microcracking occurs relatively reveal microdamage on the fibre matrix interface caused by early in the loading cycle [4]. When a mechanical stress static pre-stressing is applied to a composite, a multiplicity of several fracture The high sensitivity of the technique to very subtle dif- phenomena can occur in the material, including matrix mi- ferences in microstructural state at the chevron-notch tip cocracking, fibre-matrix debonding, delamination and fibre can be also taken as one of the possible causes for the high failure. The effects of this damage have been demonstrated scatter of data observed in our investigation. The three-point in many papers for most of the common continuous fibre bending test used in pre-stressing the samples before cut- reinforced glass and ceramic matrix composites [4-6, 22]- ting the chevron-notches produces an inhomogeneous stress These effects can also be evaluated quantitatively by differ- field. The maximum stress is located under the tip of the ent methods such as change in flexural strength or Youngs loading device and decreases with increasing distance from modulus or fracture toughness the section with this maximum stress. The distribution of In this investigation, the only irreversible microstructural microdamage in the vicinity of the fibre matrix interface damage that can be introduced in the composite during static does not need to be regular throughout the tested bar volume pre-stressing is a localised microcracking in the vicinity of and therefore the scatter in fracture resistance characteristic the fibre matrix interface boundary. This could be the dar can be taken as a natural feature of the material tested It age corresponding to the microstructural state close to the must be also noted that the typical scatter in Kle values

I. Dlouhy et al. / Materials Science and Engineering A 367 (2004) 17–23 21 Fig. 6. Fracture surfaces of the as-received (a) and high pre-stressed samples (b) (no. 18 1 at 315 MPa for 75 h) showing extensive and uniform fibre pull-out. expected. Also the data shown in Table 2 reveal that there was no change in modulus after static pre-stressing. Thus the fracture toughness values measured by the chevron-notch test seem to be independent of Young’s modulus, at least for the data set generated in this investigation. Typical fracture surfaces of chevron-notched samples are shown in Fig. 6. Fig. 6(a) shows the fracture surface of specimen in as-received condition, as characterised in an earlier study [21] whereas a fracture surface of a composite after pre-stressing at 315 MPa for 75 h is shown in Fig. 6(b). It follows from fracture surface analyses of as-received, low and high pre-stressed specimens that no substantial change in fracture surface morphology occurred. Besides the extensive fibre pull-out, typical of this kind of composites [11], the crack propagation from the chevron-notch tip was observed. The fracture surface of the chevron-notched specimens was uniform in all cases in the sense that the matrix fracture has taken place mainly in one plane (mode I crack). 4. Discussion Generally, in the stress–strain response of ceramic ma￾trix composites, the matrix microcracking occurs relatively early in the loading cycle [4]. When a mechanical stress is applied to a composite, a multiplicity of several fracture phenomena can occur in the material, including matrix mi￾crocracking, fibre-matrix debonding, delamination and fibre failure. The effects of this damage have been demonstrated in many papers for most of the common continuous fibre reinforced glass and ceramic matrix composites [4–6,22]. These effects can also be evaluated quantitatively by differ￾ent methods such as change in flexural strength or Young’s modulus or fracture toughness. In this investigation, the only irreversible microstructural damage that can be introduced in the composite during static pre-stressing is a localised microcracking in the vicinity of the fibre matrix interface boundary. This could be the dam￾age corresponding to the microstructural state close to the conditions before reaching the flexural strength value. Sim￾ilar conditions are present at the chevron-notch tip just be￾fore the crack initiation, i.e. just before the stage indicated in Fig. 5 by arrow 1 is reached. A large amount of such microdamage leads to decrease of load transfer from the matrix to the fibre. If composite does not tend to fracture (i.e. no additional load is applied) this effect is likely to decrease the Young’s modulus. This was not the case for the specimens pre-stressed at conditions de￾scribed in our investigation. There was no change in modu￾lus with pre-stress and hold time. The Young’s modulus ap￾peared to be independent of applied pre-stress level and hold time (Table 2), at least under the accuracy of the method employed for its determination (from load–deflection data). The only characteristic that has been found to be sensi￾tive to pre-stressing condition was fracture toughness. This is due to the high susceptibility of the chevron-notch tech￾nique to the material behaviour in the small volume at the chevron-notch tip. This has been convincingly shown re￾cently when investigating the effect of stress relief treatment on fracture toughness in pure borosilicate glass [23]. Thus, the chevron-notch fracture toughness has been found in the present investigation as a very accurate method enabling to reveal microdamage on the fibre matrix interface caused by static pre-stressing. The high sensitivity of the technique to very subtle dif￾ferences in microstructural state at the chevron-notch tip can be also taken as one of the possible causes for the high scatter of data observed in our investigation. The three-point bending test used in pre-stressing the samples before cut￾ting the chevron-notches produces an inhomogeneous stress field. The maximum stress is located under the tip of the loading device and decreases with increasing distance from the section with this maximum stress. The distribution of microdamage in the vicinity of the fibre matrix interface does not need to be regular throughout the tested bar volume and therefore the scatter in fracture resistance characteristic can be taken as a natural feature of the material tested. It must be also noted that the typical scatter in KIc values