Availableonlineatwww.sciencedirect.com SCIENCE c01052e5 t A: applied science ELSEVIER Composites: Part A 37(2006)23-30 Design and optimisation of glass-celsian composites V.Cannillo*, E. Carlier, T. Manfredini, M. Montorsi, C. Siligardia Dipartimento di ingegneria dei Materiali e dell'Ambiente, Universiry of Modena and Reggio Emilia, via Vignoles, 905, 41100 Modena, Italy eCole nale Superieure de Ceramique industrielle, 87065 Limoges, france Received28O October 2004; revised 13 May 2005: accepted 13 May 2005 Abstract The aim of this paper is to fabricate novel glass matrix composites reinforced by means of celsian particulate. In fact, the attractive features of celsian, such as chemical stability and high mechanical resistance, can be favourably exploited in order to obtain enhanced-performance composites with respect to bulk glass. A design of experiments(doE )procedure has been utilized to optimise the fabrication route of glass-celsian composites. This method allowed the determination of the optimal processing conditions for the obtainment of a fully dense material with a good particulate C 2005 Elsevier Ltd. All rights reserved. Keywords: A. Glasses: A. Discontinuous reinforcement; B. Microstructures; C. Statistical properties/methods: Celsian 1. Introduction to toughen glass by means of a particulate reinforcement 5. 6]. However, to the best of authors'knowledge, it is the he objective of this work is to investigate the feasibility first time that celsian is used as reinforcement for glass or of glass matrix composites reinforced with celsian ceramic matrix composites. The idea is to exploit the good particulate. Celsian is a material characterized by great features of celsian to strengthen the glass matrix and obtain hemical stability, high mechanical resistance, high melting a material that has a significantly lower cost than celsian- point and low dielectric constants, and therefore has matrix composites applications for electromagnetic windows or radome a-Celsian has been synthesized starting from zeolites and pplications at high temperatures, packaging for microelec- a commercial frit has been used for the glass matrix, as tronics, high voltage condensers and other electric insulat- described in the following section. These constituents have ing products [1]. Moreover, celsian was also used as matrix been combined in order to produce composites with volume materials for high temperature structural applications in hot fraction of reinforcement in the range 10-30%. Since it is sections of turbine engines [l], in whiskers reinforced well known that sintering conditions may have a great effect celsian-matrix composites [2], in Sic fiber reinforced on the final microstructure, different processing parameters celsian-matrix composites [3] and in mullite/celsia have been tested in order to optimise the final density of the composites [4]. In this work, the possibility to use celsian composites. In fact, porosity may reduce the mechanical as a reinforcement for glass matrix composites is performance of the composites; it is desirable to obtain fully ivestigated. In fact, glass is a low cost material with dense materials, with a even particulate dispersion and a interesting optical, electrical and thermo-insulating proper- good interface between the two phases. In order to optimise the processing route in the shortest times, a doe (design of material. Recently, several papers illustrated the possibility experiments)procedure has been adopted. DOE is a statistically-based method for determining the relationship Corresponding author. Tel: +39 059 205 6240: fax: +39 059 205 between parameters affecting a process and the 6243 such process. The basic idea is that among all possible combinations of parameters, only some of them are tested in 1359-835x/Ssee front matter O 2005 Elsevier Ltd. All rights reserved. order to significantly reduce test time [7, 8]. This method doi:10.1016/j.compositesa.2005.05.037 enables to individuate the most critical parameters
Design and optimisation of glass–celsian composites V. Cannilloa,*, E. Carlierb , T. Manfredinia , M. Montorsia , C. Siligardia a Dipartimento di Ingegneria dei Materiali e dell’Ambiente, University of Modena and Reggio Emilia, via Vignolese, 905, 41100 Modena, Italy b Ecole Nationale Supe´rieure de Ce´ramique Industrielle, 87065 Limoges, France Received 28 October 2004; revised 13 May 2005; accepted 13 May 2005 Abstract The aim of this paper is to fabricate novel glass matrix composites reinforced by means of celsian particulate. In fact, the attractive features of celsian, such as chemical stability and high mechanical resistance, can be favourably exploited in order to obtain enhanced-performance composites with respect to bulk glass. A design of experiments (DOE ) procedure has been utilized to optimise the fabrication route of glass–celsian composites. This method allowed the determination of the optimal processing conditions for the obtainment of a fully dense material with a good particulate dispersion. q 2005 Elsevier Ltd. All rights reserved. Keywords: A. Glasses; A. Discontinuous reinforcement; B. Microstructures; C. Statistical properties/methods; Celsian 1. Introduction The objective of this work is to investigate the feasibility of glass matrix composites reinforced with celsian particulate. Celsian is a material characterized by great chemical stability, high mechanical resistance, high melting point and low dielectric constants, and therefore has applications for electromagnetic windows or radome applications at high temperatures, packaging for microelectronics, high voltage condensers and other electric insulating products [1]. Moreover, celsian was also used as matrix materials for high temperature structural applications in hot sections of turbine engines [1], in whiskers reinforced celsian–matrix composites [2], in SiC fiber reinforced celsian–matrix composites [3] and in mullite/celsian composites [4]. In this work, the possibility to use celsian as a reinforcement for glass matrix composites is investigated. In fact, glass is a low cost material with interesting optical, electrical and thermo-insulating properties; however, brittleness limits its usage as structural material. Recently, several papers illustrated the possibility to toughen glass by means of a particulate reinforcement [5,6]. However, to the best of authors’ knowledge, it is the first time that celsian is used as reinforcement for glass or ceramic matrix composites. The idea is to exploit the good features of celsian to strengthen the glass matrix and obtain a material that has a significantly lower cost than celsian– matrix composites. a-Celsian has been synthesized starting from zeolites and a commercial frit has been used for the glass matrix, as described in the following section. These constituents have been combined in order to produce composites with volume fraction of reinforcement in the range 10–30%. Since it is well known that sintering conditions may have a great effect on the final microstructure, different processing parameters have been tested in order to optimise the final density of the composites. In fact, porosity may reduce the mechanical performance of the composites; it is desirable to obtain fully dense materials, with a even particulate dispersion and a good interface between the two phases. In order to optimise the processing route in the shortest times, a DOE (design of experiments) procedure has been adopted. DOE is a statistically-based method for determining the relationship between parameters affecting a process and the output of such process. The basic idea is that among all possible combinations of parameters, only some of them are tested in order to significantly reduce test time [7,8]. This method enables to individuate the most critical parameters Composites: Part A 37 (2006) 23–30 www.elsevier.com/locate/compositesa 1359-835X/$ - see front matter q 2005 Elsevier Ltd. All rights reserved. doi:10.1016/j.compositesa.2005.05.037 * Corresponding author. Tel.: C39 059 205 6240; fax: C39 059 205 6243. E-mail address: valeria@unimore.it (V. Cannillo)
V. Cannillo er al. / Composites: Part A 37 (2006 )23-30 controlling the process and therefore makes the optimiz finally a cooling step from 1300 to 800C at 20C/min in the ation of such parameters feasible [9] oven, followed by a cooling in air. To detect and identify the crystalline phases formed during the heat treatments, X-ray diffraction (XRD) was performed on finely ground sp 2. Materials mens Patterns were collected using a powder diffractometer (Philips PW3710) with a Ni-filtered Cu Ko radiation in the Celsian is a barium-aluminosilicate and may be obtained 10-50 20 range, step size 0.020 20 and 1 s data collection in several different ways, such as by hydrothermal time step (see Fig. 1). In the studied samples, only treatments, starting from gel of composition BaO-Al2O monoclinic celsian crystalline phase is formed. nSiO2(1<n<9) under temperatures in the range 150- The so-obtained a-celsian is characterized by a low 450C for a time between I and 25 days [10 by hot- thermal expansion coefficient (2.29X100C-), a rela pressing BAS glass powder derived via the sol-gel method tively high melting point (1760"C)and a good resistance to [11], by heating a ceramic mixture [12] or by devetrification chemical attack of glasses having the stoichiometric celsian composition The glass matrix is a industrial frit(FFA 17 Colorobbia [13. However, these methods are usually time-consuming Italia, Italy) belonging to CaO-Al2O3-B2O3-SiOz system, and exper with the following chemical composition (in wt%): CaO Recently, a low-cost procedure to synthesize a-celsiar 10%,BaO7%,Al2O314%,B2O331%,SiO238%.Th glass has a quite low thermal expansion coefficient(4.50X going ionic substitution of sodium by barium, with a residual 10C). The transition temperature, Tg, of glas amount of sodium of 0.43 mequiv g. In this work, such a measured by DTA corresponds to 583C. procedure [14, 15] has been utilized to produce celsian for the composite reinforcement: 6 wt% of water was added to the zeolite powders which were then pressed at 28 MPa to 3 Composites preparation and optimisation obtain small pellets of 15 mm diameter and 5 mm thickness The pellets were dried in the oven at 105C for 24 h to As mentioned in Section l, the objective is the remove the water. The sintering and crystallization behavior optimisation of the processing conditions of the composites of the powders was studied by performing a heat treatment in The contribution of the various fabrication parameters is an electrical oven (Lenton, mod. EHF 1700)following this assessed by using DOE, in particular with a graeco-latin heating cycle: from 20 to 500C at 5C/min, from 1000C square[7, 8. The studied factors are(a) the temperature and to 1300 at 10C/min, with a 5 h soaking time at 1300C. and b) the time of thermal treatment,(c)applied pressure and W TWO THETA(°) Fig. 1. XRd patterns of celsian
controlling the process and therefore makes the optimization of such parameters feasible [9]. 2. Materials Celsian is a barium–aluminosilicate and may be obtained in several different ways, such as by hydrothermal treatments, starting from gel of composition BaO–Al2O3– nSiO2 (1%n%9) under temperatures in the range 150– 450 8C for a time between 1 and 25 days [10], by hotpressing BAS glass powder derived via the sol–gel method [11], by heating a ceramic mixture [12] or by devetrification of glasses having the stoichiometric celsian composition [13]. However, these methods are usually time-consuming and expensive. Recently, a low-cost procedure to synthesize a-celsian has been proposed [14,15], which employs zeolites undergoing ionic substitution of sodium by barium, with a residual amount of sodium of 0.43 mequiv. gK1 . In this work, such a procedure [14,15] has been utilized to produce celsian for the composite reinforcement: 6 wt% of water was added to the zeolite powders which were then pressed at 28 MPa to obtain small pellets of 15 mm diameter and 5 mm thickness. The pellets were dried in the oven at 105 8C for 24 h to remove the water. The sintering and crystallization behavior of the powders was studied by performing a heat treatment in an electrical oven (Lenton, mod. EHF 1700) following this heating cycle: from 20 to 500 8C at 5 8C/min, from 1000 8C to 1300 at 10 8C/min, with a 5 h soaking time at 1300 8C, and finally a cooling step from 1300 to 800 8C at 20 8C/min in the oven, followed by a cooling in air. To detect and identify the crystalline phases formed during the heat treatments, X-ray diffraction (XRD) was performed on finely ground specimens. Patterns were collected using a powder diffractometer (Philips PW3710) with a Ni-filtered Cu Ka radiation in the 10–508 2q range, step size 0.028 2q and 1 s data collection time step (see Fig. 1). In the studied samples, only monoclinic celsian crystalline phase is formed. The so-obtained a-celsian is characterized by a low thermal expansion coefficient (2.29!10K6 8CK1 ), a relatively high melting point (1760 8C) and a good resistance to chemical attack. The glass matrix is a industrial frit (FFA 17 Colorobbia Italia, Italy) belonging to CaO–Al2O3–B2O3–SiO2 system, with the following chemical composition (in wt%): CaO 10%, BaO 7%, Al2O3 14%, B2O3 31%, SiO2 38%. The glass has a quite low thermal expansion coefficient (4.50! 10K6 8CK1 ). The transition temperature, Tg, of glass measured by DTA corresponds to 583 8C. 3. Composites preparation and optimisation As mentioned in Section 1, the objective is the optimisation of the processing conditions of the composites. The contribution of the various fabrication parameters is assessed by using DOE, in particular with a graeco-latin square [7,8]. The studied factors are (a) the temperature and (b) the time of thermal treatment, (c) applied pressure and Fig. 1. XRD patterns of celsian. 24 V. Cannillo et al. / Composites: Part A 37 (2006) 23–30
V. Cannillo et al. /Composites: Part A 37 (2006 )23-30 (d)volume fraction of reinforcement. The output is analysed Table 2 in terms of shrinkage--in thickness and in diameter-and Processing conditions bulk density. In particular, since composites with different Samples Temperature Time(min)Pressure V(%) volume fraction of reinforcement were produced, the cC) (MPa) relative density (percentage of theoretical value)was determined Therefore, the materials as designed with the graeco-latin square were realized and investigated in relation to the 000000 selected output. Moreover, the composites were characterized by using 7 scanning electron microscope(SEM), image analysis and X-ray diffraction, as described in the following It should be noted that each typestyle corresponds to a 3. 1. Design of experiments actor. le The optimisation route is performed by using the graeco- ·1,2,3: temperature(685,785,885°C); latin square which allows to study four factors(k=4), each I, Il, III time(30/, 1 h, 2 h); of them assuming three possible level(m;=3). In the A, B, C: pressure(28, 42, 56 MPa) present investigation, the four factors, namely the tempera- a, B, r: celsian volume fraction(10, 20, and 30%) ture and time of thermal treatment, the pressure applied to the samples and the volume fraction of second phase, were This table is then transposed onto the experiments design assigned different values. i.e. the levels. table, in which each line correspond to one experiment. Table 2 summarizes the so-defined processing procedures temperature of thermal treatment: 685, 785, 885C time of thermal treatment: 30 min. 1 h. 2 h: 3.2. Composites preparation and characterization applied pressure: 28, 42, 56 MPa volume fraction of reinforcement: 10. 20. 30%O Celsian samples were milled in an agate mortar for 2 h and then sieved below 38 um. The commercial glass frit was The temperatures of the thermal treatment were selecte wet milled into an alumina ball mill with 50% water for 1 h taking into account the glass transition temperature of the and then sieved below 75 um. The so-obtained slurry was matrIx(T=583°C) then dried in a oven for 48 h and the powders were stored in The total number of possible experimental tests is given hermetic containers to avoid any pollution or re y Nr, where Nt=mi=81. In the present investigation humidification only selected tests were performed. The glass and celsian powders mixed with water were put The basic assumption of the approach is that the effects in an alumina ball mill for 20 min in order to homoge r the interactions of the different factors are linearly the mixtures. The slurry was then put in an heat chamber at additive. Under this hypothesis, the number of unknowns u 125C for 48 h to obtain dry composite powders.Such is equal to 9. powders were humidified at 6% of water and then Thus, the graeco-latin square should contain nine maintained in a hermetic container for 24 h to optimise cases,as illustrated in Table 1. This plan is constructed the humidification. by assigning one and only one combination of characters The samples were pressed, with a pressing cycle which to each box, where each combination is obtained by imposed a rise in pressure of 10 MPa/s and a release of circular permutation of the Latin characters located in the pressure at P/2 in order to allow evacuation of air first line of the table and by circular permutation in After drying for 24 h at 125C, the samples were heat the other direction of the Greek characters(see Table 1). treated, with a rise in temperature of 10C/min until the final temperature T, with a soaking time according to Table 2. Cooling was made in the furnace(approximately Graeco-latin square 5C/min)until 400C then in air until room temperature The composites were characterized by using different techniques. In particular, diameter and thickness linear shrinkage percentage were calculated samples before and after the heat treatment. Bulk was determined by means of a helium picnometer(Accu 1. 2. 3: temperature: I, Il, Ill: time: A. B, C: pressure: a, B, r: celsian Pyc 1330 Picnometer Micromeritics). The microstructural volume fraction orphology was investigated by using scanning electron
(d) volume fraction of reinforcement. The output is analysed in terms of shrinkage—in thickness and in diameter—and bulk density. In particular, since composites with different volume fraction of reinforcement were produced, the relative density (percentage of theoretical value) was determined. Therefore, the materials as designed with the graeco-latin square were realized and investigated in relation to the selected output. Moreover, the composites were characterized by using scanning electron microscope (SEM), image analysis and X-ray diffraction, as described in the following. 3.1. Design of experiments The optimisation route is performed by using the graecolatin square, which allows to study four factors (kZ4), each of them assuming three possible level (miZ3). In the present investigation, the four factors, namely the temperature and time of thermal treatment, the pressure applied to the samples and the volume fraction of second phase, were assigned different values, i.e. the levels: † temperature of thermal treatment: 685, 785, 885 8C; † time of thermal treatment: 30 min, 1 h, 2 h; † applied pressure: 28, 42, 56 MPa; † volume fraction of reinforcement: 10, 20, 30%. The temperatures of the thermal treatment were selected taking into account the glass transition temperature of the matrix (TgZ583 8C). The total number of possible experimental tests is given by NT, where NTZmk i Z81. In the present investigation, only selected tests were performed. The basic assumption of the approach is that the effects or the interactions of the different factors are linearly additive. Under this hypothesis, the number of unknowns u is equal to 9. Thus, the graeco-latin square should contain nine cases, as illustrated in Table 1. This plan is constructed by assigning one and only one combination of characters to each box, where each combination is obtained by circular permutation of the Latin characters located in the first line of the table and by circular permutation in the other direction of the Greek characters (see Table 1). It should be noted that each typestyle corresponds to a factor, i.e. † 1, 2, 3: temperature (685, 785, 885 8C); † I, II, III: time (300 , 1 h, 2 h); † A, B, C: pressure (28, 42, 56 MPa); † a, b, g: celsian volume fraction (10, 20, and 30%). This table is then transposed onto the experiments design table, in which each line correspond to one experiment. Table 2 summarizes the so-defined processing procedures. 3.2. Composites preparation and characterization Celsian samples were milled in an agate mortar for 2 h and then sieved below 38 mm. The commercial glass frit was wet milled into an alumina ball mill with 50% water for 1 h and then sieved below 75 mm. The so-obtained slurry was then dried in a oven for 48 h and the powders were stored in hermetic containers to avoid any pollution or rehumidification. The glass and celsian powders mixed with water were put in an alumina ball mill for 20 min in order to homogenise the mixtures. The slurry was then put in an heat chamber at 125 8C for 48 h to obtain dry composite powders. Such powders were humidified at 6% of water and then maintained in a hermetic container for 24 h to optimise the humidification. The samples were pressed, with a pressing cycle which imposed a rise in pressure of 10 MPa/s and a release of pressure at P/2 in order to allow evacuation of air. After drying for 24 h at 125 8C, the samples were heat treated, with a rise in temperature of 10 8C/min until the final temperature T, with a soaking time according to Table 2. Cooling was made in the furnace (approximately 5 8C/min) until 400 8C then in air until room temperature. The composites were characterized by using different techniques. In particular, diameter and thickness linear shrinkage percentage were calculated measuring the samples before and after the heat treatment. Bulk density was determined by means of a helium picnometer (Accu Pyc 1330 Picnometer Micromeritics). The microstructural morphology was investigated by using scanning electron Table 2 Processing conditions Samples Temperature (8C) Time (min) Pressure (MPa) Vf (%) 1 685 30 28 10 2 685 60 42 20 3 685 120 56 30 4 785 30 56 20 5 785 60 28 30 6 785 120 42 10 7 885 30 42 30 8 885 60 56 10 9 885 120 28 20 Table 1 Graeco-latin square I II III 1 Aa Bb Cg 2 Cb Ag Ba 3 Bg Ca Ab 1, 2, 3: temperature; I, II, III: time; A, B, C: pressure; a, b, g: celsian volume fraction. V. Cannillo et al. / Composites: Part A 37 (2006) 23–30 25
V. Cannillo er al. /Composites: Part A 37 (2006 )23-30 microscopy(Philips SEM XL40) on resin embedded images of the same composite sample was determined, in samples, polished with 0.5 um polycrystalline spray order to ensure the reliability of such analysis from a diamond, in order to detect crystals shape and type, porosity statistical point of view. and distribution of the reinforcement within the Mineralogical analysis was performed by means of a Moreover, SEM micrographs were acquired by an image X-ray powder diffractometer(Philips PW3710), with an analysis software to determine porosity, in order to support angular interval 20 between 10 and 50 with a step size of density measurements. An average on 10 microscopy 0.02(1 s per step) by using the Cu Ko. radiation 8u9 T4 T5 TREATMENT u至 TREATMENT 75.00 Fig. 2. Diameter(a)and thickness(b)shrinkage;(c)relative density
microscopy (Philips SEM XL40) on resin embedded samples, polished with 0.5 mm polycrystalline spraydiamond, in order to detect crystals shape and type, porosity and distribution of the reinforcement within the composite. Moreover, SEM micrographs were acquired by an image analysis software to determine porosity, in order to support density measurements. An average on 10 microscopy images of the same composite sample was determined, in order to ensure the reliability of such analysis from a statistical point of view. Mineralogical analysis was performed by means of a X-ray powder diffractometer (Philips PW3710), with an angular interval 2q between 10 and 508 with a step size of 0.028 (1 s per step) by using the Cu Ka radiation. Fig. 2. Diameter (a) and thickness (b) shrinkage; (c) relative density. 26 V. Cannillo et al. / Composites: Part A 37 (2006) 23–30
V. Cannillo et al. /Composites: Part A 37 (2006 )23-30 4. Results and discussion those treatments where the soaking temperature was 685 C namely treatments 1, 2 and 3 Fig. 2(a-c) illustrates the results obtained for each As regards the shrinkage, the effect of temperature, processing treatment in terms of diameter and thickness soaking time, pressure and reinforcement volume fraction shrinkage, and relative density(percentage of the theoretical can be evaluated. In particular, as previously observed, the value). These results show that treatments 7,8 and 9 are temperature of 885 C leads to an expansion in thickness; in unacceptable since the sample present an expansion in fact, such temperature coincides with the temperature of half-ball [16] of the frit used. In this condition, the viscosity Each output parameter, namely the diameter and becomes too low and spherical pores arise. The most thickness shrinkage and relative density, were separately suitable temperature for composites densification seems to analysed in order to investigate the effect of each factor. An be 785C. With reference to the soaking time, 30 min and examination grid was constructed, i.e. an arithmetic mean I h lead to the best results in terms of shrinkage; therefore, a vas calculated on all values obtained in different processing thermal treatment of 30 min was selected in order to treatment where a factor had a specific value. The results are minimize the processing time. Higher pressure seems to plotted in Fig 3 for the shrinkage and the relative density, give a greater shrinkage. No significant conclusion can be respectively. For example, the value for diameter shrinkage drawn out in terms of volume fraction of reinforcement of Fig 3a corresponding to 685C was obtained making an As regards the relative density, it should be noted that the average on all diameter shrinkage output values obtained in values were determined both with helium picnometer and FACTORS EFFECT 20900 08900890 0897089708930895 F075 6857895985301h2h28MPa42MPa56MPa10%20%30% FACTORS I Fig 3 Layout of the effect of the different factors on:(a)the shrinkage:()diameter;()thickness and(b) the relative density
4. Results and discussion Fig. 2(a–c) illustrates the results obtained for each processing treatment in terms of diameter and thickness shrinkage, and relative density (percentage of the theoretical value). These results show that treatments 7, 8 and 9 are unacceptable since the sample present an expansion in thickness. Each output parameter, namely the diameter and thickness shrinkage and relative density, were separately analysed in order to investigate the effect of each factor. An examination grid was constructed, i.e. an arithmetic mean was calculated on all values obtained in different processing treatment where a factor had a specific value. The results are plotted in Fig. 3 for the shrinkage and the relative density, respectively. For example, the value for diameter shrinkage of Fig. 3a corresponding to 685 8C was obtained making an average on all diameter shrinkage output values obtained in those treatments where the soaking temperature was 685 8C, namely treatments 1, 2 and 3. As regards the shrinkage, the effect of temperature, soaking time, pressure and reinforcement volume fraction can be evaluated. In particular, as previously observed, the temperature of 885 8C leads to an expansion in thickness; in fact, such temperature coincides with the temperature of half-ball [16] of the frit used. In this condition, the viscosity becomes too low and spherical pores arise. The most suitable temperature for composites densification seems to be 785 8C. With reference to the soaking time, 30 min and 1 h lead to the best results in terms of shrinkage; therefore, a thermal treatment of 30 min was selected in order to minimize the processing time. Higher pressure seems to give a greater shrinkage. No significant conclusion can be drawn out in terms of volume fraction of reinforcement. As regards the relative density, it should be noted that the values were determined both with helium picnometer and Fig. 3. Layout of the effect of the different factors on: (a) the shrinkage: (C) diameter; (&) thickness and (b) the relative density. V. Cannillo et al. / Composites: Part A 37 (2006) 23–30 27