Availableonlineatwww.sciencedirect.com SCIENCE DIRECTO composites Part A: applied science ELSEVIER Composites: Part A 34(2003)613-622 www.elsevier.com/locate/composite Mesostructure of whipoX all oxide cmcs M. SchmuckerA. Grafmuller h. schneide German Aerospace Center(DLR), Institute of Materials Research, D51147 KOln, Germany Received 29 October 2002; revised 26 February 2003: accepted 14 March 2003 Abstract Wound highly porous oxide matrix(WHIPOX) ceramic matrix composites consist of oxide fibers(mullite- or alumina-type) which are embedded in mullite- or alumina-rich matrices, respectively. In the ideal case the fiber distribution is homogeneous; in reality, however, fabrication(winding)-induced matrix agglomerations do occur. As knowledge on the homogeneity of the material is crucial for the prediction f the mechanical behavior a technique to describe the mesostructure of WhiPOX quantitatively has been developed by means of optical microscopy(transmitted light). The technique makes use of light conductivity and opacity of fibers and matrix, respectively. Three dimensional plots of the matrix agglomerations were obtained by tomographic methods using 25 individual slices of 1.5 mm thickness for mesostructurally. The study showed that delamination- induced failure of WhIPOX is essentially controlled by localized interlaminate matrix agglomerations. Compression of WhiPOX plates in the pre-sintering moist stage helps to achieve a better homogeneity and thus improved near strength of WHIPOX components C 2003 Elsevier science ltd. all rights reserved Keywords: A Ceramic-matrix composites(CMCs); C. Statistical properties/methods; D. Optical microscopy 1. Introduction matrix concept. WHIPOX CMCs exhibit excellent mech anical, thermomechanical, and thermal properties [4-71 Oxide ceramics have a high potential WHIPOX CMCs are made up of oxide fibers(ce protection systems(TPS)in combustion chambers of ga Nextel, 3M, fibers of type 610, 650 or 720)which are turbine engines and for re-entry space vehicles. A promising embedded in a porous mullite or alumina matrix. Adjacent way to achieve the required tough and damage tolerant matrix infiltrated fiber bundles typically combine to form ceramics is the reinforcement of the ceramic substrate bodies laminates with the occurrence of relatively large fiber free matrix agglomerations(Fig. 1). The fractography of by continuous ceramic fibers [1]. The non-brittle quasI- WHIPOX components and structures has shown that plastic behavior of fiber-reinforced ceramics is due to a relatively weak bonding between fibers and the matrix thus delamination is the predominant failure mechanism allowing crack deflection and fiber pull-out [2]. To achieve a ggesting that interlaminate matrix agglomerations have weak fiber/matrix bonding either suitable fiber coatings have a controlling influence on the material's mechanical behavior. An improved design of the meso- and micro- phases)or, in an alternative approach, a highly porous matrix structure of WHlPoX therefore requires reliable and may be used. Materials making use of the porous matrix statistically robust information on the amount and the concept exhibit only few local fiber/matrix contacts and spatial orientation of fiber free areas which, however, is not hence the matrix acts like a weak fiber/matrix interphase [3] able ye WHIPOX (Wound highly porous oxide matrix) Different methods have been presented for the quanti- ites have been designed according to this porou tative microstructural analysis of fiber-reinforced materials such as metal or polymer matrix composites. These Corresponding author. Tel: +49-22-03-601-2462: fax: +49-22-03 procedures all are focused on analyzes of the fiber 696-480 ns using Dirichlet tessalation [8, 9], fiber cell E-mail address: martin. schmuecker@dlr. de(M. Schmucker) analysis [10], or radial fiber-fiber distance distributions [9] 1359-835X/03/S- see front matter 2003 Elsevier Science Ltd. All rights reserved. doi:10.1016/s1359-835X03)00100-3
Mesostructure of WHIPOX all oxide CMCs M. Schmu¨cker*, A. Grafmu¨ller, H. Schneider German Aerospace Center (DLR), Institute of Materials Research, D-51147 Ko¨ln, Germany Received 29 October 2002; revised 26 February 2003; accepted 14 March 2003 Abstract Wound highly porous oxide matrix (WHIPOX) ceramic matrix composites consist of oxide fibers (mullite- or alumina-type) which are embedded in mullite- or alumina-rich matrices, respectively. In the ideal case the fiber distribution is homogeneous; in reality, however, fabrication (winding)-induced matrix agglomerations do occur. As knowledge on the homogeneity of the material is crucial for the prediction of the mechanical behavior a technique to describe the mesostructure of WHIPOX quantitatively has been developed by means of optical microscopy (transmitted light). The technique makes use of light conductivity and opacity of fibers and matrix, respectively. Threedimensional plots of the matrix agglomerations were obtained by tomographic methods using <25 individual slices of 1.5 mm thickness for each sample. Samples from different sites of a WHIPOX plate, and samples which have been differently pressed prior to sintering were examined mesostructurally. The study showed that delamination-induced failure of WHIPOX is essentially controlled by localized interlaminate matrix agglomerations. Compression of WHIPOX plates in the pre-sintering moist stage helps to achieve a better homogeneity and thus improved shear strength of WHIPOX components. q 2003 Elsevier Science Ltd. All rights reserved. Keywords: A. Ceramic-matrix composites (CMCs); C. Statistical properties/methods; D. Optical microscopy 1. Introduction Oxide ceramics have a high potential for thermal protection systems (TPS) in combustion chambers of gas turbine engines and for re-entry space vehicles. A promising way to achieve the required tough and damage tolerant ceramics is the reinforcement of the ceramic substrate bodies by continuous ceramic fibers [1]. The non-brittle quasiplastic behavior of fiber-reinforced ceramics is due to a relatively weak bonding between fibers and the matrix thus allowing crack deflection and fiber pull-out [2]. To achieve a weak fiber/matrix bonding either suitable fiber coatings have to be employed (cleavable, porous, low toughness interphases) or, in an alternative approach, a highly porous matrix may be used. Materials making use of the porous matrix concept exhibit only few local fiber/matrix contacts and hence the matrix acts like a weak fiber/matrix interphase [3]. WHIPOXe (Wound highly porous oxide matrix) composites have been designed according to this porous matrix concept. WHIPOX CMCs exhibit excellent mechanical, thermomechanical, and thermal properties [4–7]. WHIPOX CMCs are made up of oxide fibers (commercial Nextel, 3M, fibers of type 610, 650 or 720) which are embedded in a porous mullite or alumina matrix. Adjacent matrix infiltrated fiber bundles typically combine to form laminates with the occurrence of relatively large fiber free matrix agglomerations (Fig. 1). The fractography of WHIPOX components and structures has shown that delamination is the predominant failure mechanism suggesting that interlaminate matrix agglomerations have a controlling influence on the material’s mechanical behavior. An improved design of the meso- and microstructure of WHIPOX therefore requires reliable and statistically robust information on the amount and the spatial orientation of fiber free areas which, however, is not available yet. Different methods have been presented for the quantitative microstructural analysis of fiber-reinforced materials such as metal or polymer matrix composites. These procedures all are focused on analyzes of the fiber distributions using Dirichlet tessalation [8,9], fiber cell analysis [10], or radial fiber–fiber distance distributions [9]. 1359-835X/03/$ - see front matter q 2003 Elsevier Science Ltd. All rights reserved. doi:10.1016/S1359-835X(03)00100-3 Composites: Part A 34 (2003) 613–622 www.elsevier.com/locate/compositesa * Corresponding author. Tel.: þ49-22-03-601-2462; fax: þ49-22-03- 696-480. E-mail address: martin.schmuecker@dlr.de (M. Schmu¨cker)
M. Schmiicker et al /Composites: Part A 34(2003)613-622 In order to achieve higher density and a better homogeneity the green plates were pressed between two aluminium oxide plates. In the final step the green plates were sintered in air at 1300°C. xE 2. 2. Preparation of test se Two series of samples were prepared from the middle of wn In microscopic analysis were taken from series a, whereas the samples corresponding to series b were used to determine the mechanical characteristics, particularly the shear strength. As not all samples from row a were used for microscopy, it could be established that, although the mechanical properties can vary significantly from sample to sample in the direction perpendicular to the winding direction (i.e. form sample 1-12 in Fig. 2), they are reasonably coherent in corresponding samples from a and b This allows the assignment of microscopical and mechan ical data relating adjacent samples. Fig 1. Scanning electron micrograph (overview)of WHlPox all oxide of Nextel 610(3M) alumina fibers and an Al2O3-rich free areas(matrix agglomerations with macropores) from each other alumino-silicate matrix Samples cut from the same plate but from different positions, one from the edge (la)and one The fiber distribution alone, however, does not provide the from the middle(6a)(see Fig. 2), were investigated in order full information on the mesostructure of porous all-oxide clarify whether location dependent differences exist. In ceramics as direct data on interlaminate fiber free areas addition, the effects of green bodies compression were cannot be obtained. A suitable technique to describe the vestigated by comparing a mildly(10%)and a strongly mesostructure of WHIPOX oxide fiber/oxide matrix com- compressed(25%)material posites has recently been developed at DLR [ll]. Fiber fre areas were recorded by means of optical microscopy 2.3. Mechanical testing (transmitted light) utilizing the light conductivity and opacity of fibers and matrix, respectively. Mechanical tests were performed by 3-point bendin sing a 20 mm support distance. Bending bar width and thickness was 10 and 3 mm, respectively. Load deflection 2. Experimental methods curves show a non-brittle fracture behavior [6]. The interlaminar shear strength (ILSS)was calculated accordin 2.1. Processing of WHIPOX CMCs slices WHIPOX CMCs are produced by a continuously work ing winding technique. The fibers first run through a tube furnace where the sizing is burned off. The rovings are then infiltrated with water-based matrix slurry and pre-dried in a continuously working microwave (Agni, Thermal and Materials Technology, Aachen, Germany). The infiltrated yarns are wound onto a plastic mandrel(diameter, 200 mm) Series a>123 4516 in a closed box of constant high air humidity(H2O-saturated air). The winding process i.e. the traverse and rotational speed, is computer controlled, using the PC program Series b> CADWIND(Materials SA; Brussels, Belgium) which llows different fiber orientations and winding patterns. The investigated samples all display winding angles a of 15, i.e. the angle between the fiber rovings and the winding normal is 15 and hence the angle between intersecting fiber Fig. 2. Sketch of sample positions within a WHIPOX plate. Series b specimens were tested mechanically while selected series a specimens were bundles is 30(also see Fig. 10). To obtain fat plates used for microscopic analyzes. For that, an individual series a specimen bar the material is removed from the mandrel and straightened was cut into 20 slices of 1.5 mm thickness
The fiber distribution alone, however, does not provide the full information on the mesostructure of porous all-oxide ceramics as direct data on interlaminate fiber free areas cannot be obtained. A suitable technique to describe the mesostructure of WHIPOX oxide fiber/oxide matrix composites has recently been developed at DLR [11]. Fiber free areas were recorded by means of optical microscopy (transmitted light) utilizing the light conductivity and opacity of fibers and matrix, respectively. 2. Experimental methods 2.1. Processing of WHIPOX CMCs WHIPOX CMCs are produced by a continuously working winding technique. The fibers first run through a tube furnace where the sizing is burned off. The rovings are then infiltrated with water-based matrix slurry and pre-dried in a continuously working microwave (Agni, Thermal and Materials Technology, Aachen, Germany). The infiltrated yarns are wound onto a plastic mandrel (diameter, 200 mm) in a closed box of constant high air humidity (H2O-saturated air). The winding process i.e. the traverse and rotational speed, is computer controlled, using the PC program CADWIND (Materials S.A; Brussels, Belgium) which allows different fiber orientations and winding patterns. The investigated samples all display winding angles a of 158, i.e. the angle between the fiber rovings and the winding normal is 158 and hence the angle between intersecting fiber bundles is 308 (also see Fig. 10). To obtain flat plates the material is removed from the mandrel and straightened. In order to achieve higher density and a better homogeneity the green plates were pressed between two aluminium oxide plates. In the final step the green plates were sintered in air at 1300 8C. 2.2. Preparation of test samples Two series of samples were prepared from the middle of each plate, as shown in Fig. 2. The samples used for microscopic analysis were taken from series a, whereas the samples corresponding to series b were used to determine the mechanical characteristics, particularly the shear strength. As not all samples from row a were used for microscopy, it could be established that, although the mechanical properties can vary significantly from sample to sample in the direction perpendicular to the winding direction (i.e. form sample 1–12 in Fig. 2), they are reasonably coherent in corresponding samples from a and b. This allows the assignment of microscopical and mechanical data relating adjacent samples. All WHIPOX materials investigated in this study consist of Nextel 610 (3M) alumina fibers and an Al2O3-rich alumino-silicate matrix. Samples cut from the same plate, but from different positions, one from the edge (1a) and one from the middle (6a) (see Fig. 2), were investigated in order to clarify whether location dependent differences exist. In addition, the effects of green bodies compression were investigated by comparing a mildly (<10%) and a strongly compressed (<25%) material. 2.3. Mechanical testing Mechanical tests were performed by 3-point bending using a 20 mm support distance. Bending bar width and thickness was 10 and 3 mm, respectively. Load deflection curves show a non-brittle fracture behavior [6]. The interlaminar shear strength (ILSS) was calculated according Fig. 1. Scanning electron micrograph (overview) of WHIPOX all oxide CMC. Matrix infiltrated fiber bundles may form laminae separated by fiber free areas (matrix agglomerations with macropores) from each other. Fig. 2. Sketch of sample positions within a WHIPOX plate. Series b specimens were tested mechanically while selected series a specimens were used for microscopic analyzes. For that, an individual series a specimen bar was cut into 20 slices of 1.5 mm thickness. 614 M. Schmu¨cker et al. / Composites: Part A 34 (2003) 613–622
M. Schmiicker et al /Composites: Part A 34(2003)613-622 to the European Prestandard ENV 658/5 to be 3L/(4bh) with L= load, b= sample width and h= sample height e 2. 4. Microscopy and image processing Prior to ceramography the porous WHIPOX material was infiltrated with low-viscous epoxy resin. Subsequently, the samples were cut into 1.5-2.0 mm thick slices and prepared for optical inspections by grinding and polishing with standard methods For the microscopic analysis, an optical microscope was used in the transmission mode with low magnification (50 X). This method has the advantage that Fig 3. Optical micrographs of WHIPOX taken in reflection (left) and the fibers'light transmission produces a strong contrast, ansmission mode (right), respectively. The fibers' light conductivity making fibers appear bright and larger fiber free areas dark yields very high contrasts in the transmission mode (Fig. 3), thus facilitating semi-automatic image processing Microscopy ss an contrast reduce size to30% Assemble pictures of one slice Further adjust brightness and contrast Apply Gaussian Blur Change colour mode to black and white Analysis with “ Scion Image Fig. 4. Flow chart showing steps of image processing and analysis
to the European Prestandard ENV 658/5 to be 3L=ð4bhÞ with L ¼ load, b ¼ sample width and h ¼ sample height. 2.4. Microscopy and image processing Prior to ceramography the porous WHIPOX material was infiltrated with low-viscous epoxy resin. Subsequently, the samples were cut into 1.5–2.0 mm thick slices and prepared for optical inspections by grinding and polishing with standard methods. For the microscopic analysis, an optical microscope was used in the transmission mode with low magnification (50 £ ). This method has the advantage that the fibers’ light transmission produces a strong contrast, making fibers appear bright and larger fiber free areas dark (Fig. 3), thus facilitating semi-automatic image processing. Fig. 3. Optical micrographs of WHIPOX taken in reflection (left) and transmission mode (right), respectively. The fibers’ light conductivity yields very high contrasts in the transmission mode. Fig. 4. Flow chart showing steps of image processing and analysis. M. Schmu¨cker et al. / Composites: Part A 34 (2003) 613–622 615
616 M. Schmiicker et al /Composites: Part A 34(2003)613-622 Low magnification was used to cover the entire area of each flaws per x-row is projected onto the z-axis. These projections slice 3-4 micrographs per slice were recorded, put together are then summed up(Fig. 5) and may indicate which plane is d processed by means of a commercial image processing most susceptible to delamination. The fiber density within software( Corel Photopaint) to a black/white image, the lamellae was analyzed by means of fiber cell-size showing all fiber free areas in black. The pictures were analysis. Fiber cells are the areas around individual fiber then analyzed using Scion Image'(Scion Corp. software. centers containing all points closer to the respective fiber centers than to any others [10 The smaller these areas are, 2.5. Image analysis the greater is the fiber density. For the fiber density analys transmission mode pictures with higher magnification Size and position of all fiber free regions larger than 75 (200x) were taken (Fig. 6a) and the fiber cell were tels(5800 um)are recorded. The discriminating size constructed by means of a self-designed program(Fig. 6b) large in comparison to the size of the fibers( 100 um),as too much detail would only distract. From this data the total percentage of flaw area and the 3 Results and discussion distribution of flaw sizes were determined. Fig 4 gives a fow chart showing the individual steps of image processing and 3.1. Distribution of mesostructural flaws analysis In order to estimate the probability of delamination, he faw distribution in the materials was analyzed as to Ideally. WHIPOX all oxide CMCs consist of fibers whether there are planes in which there is a larger than embedded homogeneously in a matrix of high but homo- average area of fiber free material. For this, the amount of geneously distributed microporosity. Starting from this Fig. 5. Analysis of interlaminate matrix agglomerations: The amount of matrix-free areas is projected along x onto the z-axis and summed-up for all slices
Low magnification was used to cover the entire area of each slice. 3–4 micrographs per slice were recorded, put together and processed by means of a commercial image processing software (Corele Photopaint) to a black/white image, showing all fiber free areas in black. The pictures were then analyzed using ‘Scion Image’ (Scion Corp.) software. 2.5. Image analysis Size and position of all fiber free regions larger than 75 pixels (5800 mm2 ) are recorded. The discriminating size is large in comparison to the size of the fibers (<100 mm2 ), as too much detail would only distract. From this data the total percentage of flaw area and the distribution of flaw sizes were determined. Fig. 4 gives a flow chart showing the individual steps of image processing and analysis. In order to estimate the probability of delamination, the flaw distribution in the materials was analyzed as to whether there are planes in which there is a larger than average area of fiber free material. For this, the amount of flaws per x-row is projected onto the z-axis. These projections are then summed up (Fig. 5) and may indicate which plane is most susceptible to delamination. The fiber density within the lamellae was analyzed by means of fiber cell-size analysis. Fiber cells are the areas around individual fiber centers containing all points closer to the respective fiber centers than to any others [10]. The smaller these areas are, the greater is the fiber density. For the fiber density analysis transmission mode pictures with higher magnification (200 £ ) were taken (Fig. 6a) and the fiber cell were constructed by means of a self-designed program (Fig. 6b). 3. Results and discussion 3.1. Distribution of mesostructural flaws Ideally, WHIPOX all oxide CMCs consist of fibers embedded homogeneously in a matrix of high but homogeneously distributed microporosity. Starting from this Fig. 5. Analysis of interlaminate matrix agglomerations: The amount of matrix-free areas is projected along x onto the z-axis and summed-up for all slices. 616 M. Schmu¨cker et al. / Composites: Part A 34 (2003) 613–622
M. Schmiicker et al /Composites: Part A 34 (2003)613-622 617 For a three-dimensional analysis, faws clearly dis- 经势 tinguishable in consecutive slices were marked and the coordinates of their centers were determined. Although the accuracy of this method is limited it gives a clear impression of the spatial distribution of flaw areas. The results are shown in Fig. 8: faw types i and iii their position on the plane perpendicular to the winding direction (x-z-plane) in going through the material winding direction (y-direction). Closer inspection reveals a continuous change in x-direction but a constant z-position Type ii flaws(Fig. 8b), on the other hand, do not change their x-z-positions in going from one slice to the other. e The angle between faw alignment and the winding ection (y-axis) is determined by the faw centers Fig.6.Construction of fiber cell areas(after [0): Fiber cell x contains all x-displacement and the known distance from one slice to points closer to fiber x than to any other. the next which is about 1.5 mm the angles thus found for a idealized microstructure, all deviations from a homogeneous total number of 13 type i and type iii flaws range between 6 fiber distribution leading locally to matrix enrichments, are and 20 Considering the accuracy of the measurements, it considered as fabrication-induced 'flaws,. Such matrix can be assumed that type ii and type iii flaws spread through agglomerations typically contain large macropores[ll] the sample by the winding angle of 1 The flaws occurring in the investigated WhIPOX CMCs can be grouped into three general types that differ in size, 3.2. Origin offiaws shape, and spatial distribution(Fig. 7) Type i flaws. As the rovings are wound at specific angle, Type i: Relatively large matrix agglomerations with huge misalignment of adjacent rovings will occur at their crossing pores(Fig. 7a) points: Flattening and compressing in the moist stage either Type ii: Narrow, lense-shaped matrix accumulations produces distorted laminae or wedge-shaped breakoff arise existing throughout the entire height of samples and as sketched in Fig. 9. Voids formed in these break off dividing the individual laminae from each other. This regions can be partially filled by matrix slurry pressed out type of defect is especially observed in samples from the from the surrounding infiltrated fiber bundles. Since these middle of the WHIPOX plates kind of flaws are associated with the crossing points of the Type iii: Matrix agglomerations which, however, are not fiber bundles they move across the material at the same ompletely fiber free. Thereby, the number of individual angle as the rovings are wound. For this type of flaws, fibers dispersed within this flaw area corresponds to the therefore, one should expect a change of position of 15,as number of fibers in the fiber roving used for the cmc one moves through the sample from one slice to the next. fabrication This actually has been observed Fig. 7. Different types of faws identified in the investigated WHIPOX CMCs. Type i(a): bulky matrix agglomerations with macropores; type ii(b): interlaminate matrix accumulations of moderate thickness existing throughout the entire height(z-direction, see Fig. 5); type iii(c): spread out fiber roving with high amounts of interlaminate matrix
idealized microstructure, all deviations from a homogeneous fiber distribution leading locally to matrix enrichments, are considered as fabrication-induced ‘flaws’. Such matrix agglomerations typically contain large macropores [11]. The flaws occurring in the investigated WHIPOX CMCs can be grouped into three general types that differ in size, shape, and spatial distribution (Fig. 7): Type i: Relatively large matrix agglomerations with huge pores (Fig. 7a). Type ii: Narrow, lense-shaped matrix accumulations existing throughout the entire height of samples and dividing the individual laminae from each other. This type of defect is especially observed in samples from the middle of the WHIPOX plates. Type iii: Matrix agglomerations which, however, are not completely fiber free. Thereby, the number of individual fibers dispersed within this flaw area corresponds to the number of fibers in the fiber roving used for the CMC fabrication. For a three-dimensional analysis, flaws clearly distinguishable in consecutive slices were marked and the coordinates of their centers were determined. Although the accuracy of this method is limited it gives a clear impression of the spatial distribution of flaw areas. The results are shown in Fig. 8: flaw types i and iii (Fig. 8a and c) change their position on the plane perpendicular to the winding direction (x–z-plane) in going through the material in winding direction (y-direction). Closer inspection reveals a continuous change in x-direction but a constant z-position. Type ii flaws (Fig. 8b), on the other hand, do not change their x–z-positions in going from one slice to the other. The angle between flaw alignment and the winding direction (y-axis) is determined by the flaw centers’ x-displacement and the known distance from one slice to the next which is about 1.5 mm. The angles thus found for a total number of 13 type i and type iii flaws range between 68 and 208 Considering the accuracy of the measurements, it can be assumed that type ii and type iii flaws spread through the sample by the winding angle of 158. 3.2. Origin of flaws Type i flaws. As the rovings are wound at specific angle, misalignment of adjacent rovings will occur at their crossing points: Flattening and compressing in the moist stage either produces distorted laminae or wedge-shaped breakoffs arise as sketched in Fig. 9. Voids formed in these break off regions can be partially filled by matrix slurry pressed out from the surrounding infiltrated fiber bundles. Since these kind of flaws are associated with the crossing points of the fiber bundles they move across the material at the same angle as the rovings are wound. For this type of flaws, therefore, one should expect a change of position of 158, as one moves through the sample from one slice to the next. This actually has been observed. Fig. 7. Different types of flaws identified in the investigated WHIPOX CMCs. Type i (a): bulky matrix agglomerations with macropores; type ii (b): interlaminate matrix accumulations of moderate thickness existing throughout the entire height (z-direction, see Fig. 5); type iii (c): spread out fiber roving with high amounts of interlaminate matrix. Fig. 6. Construction of fiber cell areas (after [10]): Fiber cell x contains all points closer to fiber x than to any other. M. Schmu¨cker et al. / Composites: Part A 34 (2003) 613–622 617