ELSEVIE Surface and Coatings Technology 120-121(1999)250-258 Recent advances in forced-flow, thermal-gradient Cvi for refractory composites Kent J. Probst b. Theodore M. Besmann a, * David P Stinton a. Richard A. Lowden a Timothy J. Anderson b, Thomas L. Starr c Metals and Ceramics Division, Oak Ridge National Laboratory, Oak Ridge, TN37831-6063, US.A b Department of Chemical Engineering, University of Florida, Gainesville, FL 32611, USA c Department of Chemical Engineering, University of Louisville, Louisville, KY 40292, US.A Abstract Chemical vapor infiltration(CVI)is simply chemical vapor deposition (CVD)on the internal surfaces of a porous preform and has been used to produce a variety of developmental and application materials. The greatest use of CVi is to infiltrate continuous-filament preforms taking advantage of the relatively low-stress CVD process. In CVI, reactants are introduced in the porous preform via either diffusion or forced convection and the cvd precursors deposit the appropriate phase(s). As infiltration roceeds, the deposit on the internal surfaces becomes thicker. Thus, after some length of time, the growing surfaces meet bonding the preform and fill much of the free volume with deposited matrix. The forced-flow/thermal-gradient technique(FCvi)developed at Oak Ridge National Laboratory overcomes the problems of slow diffusion and restricted permeability, and has demonstrated a capability to produce thick-walled, simple-shaped, Sic-matrix components in times of the order of hours. A model has beer developed for the process that predicts flow, thermal and density profiles as a function of time. The results have been compared with an initial set of experiments and indicate qualitative agreement. It is expected that improved property relationships, such as permeability and thermal conductivity as a function of density, will allow the model to closely represent the fcvi process and be useful in fabrication and product optimization. o 1999 Published by Elsevier Science S.A. All rights Keywords: Chemical vapor infiltration; Continuous fiber ceramic composites; CVI; FCVI; Forced-flow/thermal-gradient technique: Refractory mposites; Silicon carbide 1. Introductio ups). The earliest report of CVI for ceramics was a 1964 patent for infiltrating fibrous alumina with chromium infiltration(CVI)is carbides [2] vapor deposition (CvD)on the inter of a In CVI, reactants are introduced in the porous pre- porous preform and has been used to pr a variety form via either diffusion or forced convection and the of developmental and application materials. The greatest CVd precursors deposit the appropriate phase(s).As use of Cvi is to infiltrate continuous-filament preforms infiltration proceeds, the deposit on the internal surfaces taking advantage of the relatively low-stress CVd becomes thicker. Thus, after some length of time, the growing surfaces meet, bonding the preform and filling Chemical vapor infiltration originated in efforts to much of the free volume with deposited matrix [ 3] densify porous graphite bodies by infiltration with The major advantage CVI has over competing densi- carbon [l]. The technique has developed commercially fication processes is the low thermal and mechanical such that half of the carbon/carbon composites currently stress to which the relatively sensitive fibers are sub- produced world-wide are made by CVI (the remainder jected. Chemical vapor infiltration can occur at temper- are fabricated by curing polymer-impregnated fiber lay atures much more moderate than the melting point of the deposit and, therefore, usually well below the sin- a* Corresponding author. Tel. +1-423-574-6852: tering temperature. In addition, the process imparts little x:+1-423-5746918 mechanical stress to the preform as compared with more E-mail address: tmb@ornl. gov(T.M. Besmann) traditional techniques such as hot pressing H33 see front matter o 1999 Published by Elsevier Science S.A. All nights reserved. 72(99)00459-4
Surface and Coatings Technology 120–121 (1999) 250–258 www.elsevier.nl/locate/surfcoat Recent advances in forced-flow, thermal-gradient CVI for refractory composites Kent J. Probst b, Theodore M. Besmann a,*, David P. Stinton a, Richard A. Lowden a, Timothy J. Anderson b, Thomas L. Starr c a Metals and Ceramics Division, Oak Ridge National Laboratory, Oak Ridge, TN 37831-6063, USA b Department of Chemical Engineering, University of Florida, Gainesville, FL 32611, USA c Department of Chemical Engineering, University of Louisville, Louisville, KY 40292, USA Abstract Chemical vapor infiltration (CVI) is simply chemical vapor deposition (CVD) on the internal surfaces of a porous preform and has been used to produce a variety of developmental and application materials. The greatest use of CVI is to infiltrate continuous-filament preforms taking advantage of the relatively low-stress CVD process. In CVI, reactants are introduced in the porous preform via either diffusion or forced convection and the CVD precursors deposit the appropriate phase(s). As infiltration proceeds, the deposit on the internal surfaces becomes thicker. Thus, after some length of time, the growing surfaces meet bonding the preform and fill much of the free volume with deposited matrix. The forced-flow/thermal-gradient technique (FCVI) developed at Oak Ridge National Laboratory overcomes the problems of slow diffusion and restricted permeability, and has demonstrated a capability to produce thick-walled, simple-shaped, SiC-matrix components in times of the order of hours. A model has been developed for the process that predicts flow, thermal and density profiles as a function of time. The results have been compared with an initial set of experiments and indicate qualitative agreement. It is expected that improved property relationships, such as permeability and thermal conductivity as a function of density, will allow the model to closely represent the FCVI process and be useful in fabrication and product optimization. © 1999 Published by Elsevier Science S.A. All rights reserved. Keywords: Chemical vapor infiltration; Continuous fiber ceramic composites; CVI; FCVI; Forced-flow/thermal-gradient technique; Refractory composites; Silicon carbide 1. Introduction ups). The earliest report of CVI for ceramics was a 1964 patent for infiltrating fibrous alumina with chromium Chemical vapor infiltration (CVI) is simply chemical carbides [2]. vapor deposition (CVD) on the internal surfaces of a In CVI, reactants are introduced in the porous preporous preform and has been used to produce a variety form via either diffusion or forced convection and the of developmental and application materials. The greatest CVD precursors deposit the appropriate phase(s). As use of CVI is to infiltrate continuous-filament preforms infiltration proceeds, the deposit on the internal surfaces taking advantage of the relatively low-stress CVD becomes thicker. Thus, after some length of time, the process. growing surfaces meet, bonding the preform and filling Chemical vapor infiltration originated in efforts to much of the free volume with deposited matrix [3]. densify porous graphite bodies by infiltration with The major advantage CVI has over competing densicarbon [1]. The technique has developed commercially fication processes is the low thermal and mechanical such that half of the carbon/carbon composites currently stress to which the relatively sensitive fibers are subproduced world-wide are made by CVI (the remainder jected. Chemical vapor infiltration can occur at temperare fabricated by curing polymer-impregnated fiber lay- atures much more moderate than the melting point of the deposit and, therefore, usually well below the sintering temperature. In addition, the process imparts little * Corresponding author. Tel.: +1-423-574-6852; mechanical stress to the preform as compared with more fax: +1-423-574-6918. E-mail address: tmb@ornl.gov (T.M. Besmann) traditional techniques such as hot pressing. 0257-8972/99/$ – see front matter © 1999 Published by Elsevier Science S.A. All rights reserved. PII: S0257-8972(99)00459-4
diffusion for species transport. It generally operates at of the deposition of SiC, deposit thickness profiles shE K.J. Probst et al./ Surface and Coatings Technology 120-121(1999)250-258 The most widely used commercial process pressure can be clearly seen in the review by Nasla isothermal/isobaric CVI (ICVI)which depends only on [7]. Assuming a first-order reaction for the simulation reduced pressure(1-10 kPa) for deposition rate control. smaller gradients at lower temperature and pressure This diffusion-dependent process is slow, requiring sev Several cvi efforts have described non-isothermal eral week-long infiltration times. It is commercially non-isobaric processes. In these systems, there are addi attractive, however, because large numbers of parts of tional complexities associated with heat transfer and varying dimensions are easily accommodated in a forced convection as infiltration proceeds and the pore single reactor. structure of the material changes. To accurately describe The forced-flow/thermal-gradient technique(FCVI) infiltration behavior, descriptions of the complexities developed at Oak Ridge National Laboratory (ORNL) associated with these transport processes should be overcomes the problems of slow diffusion and restricted combined with the complex microstructure evolution permeability, and has demonstrated a capability to and chemical kinetics. Regardless of these complexities produce thick-walled, simple-shaped, SiC-matrix com- and the current inability to fully describe them, under ponents in times of the order of hours [4-6 standing the general relationships between the relevant Chemical vapor infiltration is a maturing technology kinetic processes has led to the practical solution of with specialty manufacturing in place and the emergence many CVi problems. Simple reactant gas depletion of broader markets imminent. A number of recent within the CVi reactor or poisoning of the deposition advances have made the technology more attractive and process by reaction byproducts(e. g, HCl in the Cvi of provided a broader base for applications. These include SiC from MTS)also reduces the deposition rate [12] obtaining an understanding of fiber/matrix interface In the thermal gradient process, controlling the temper- hemical and thermomechanical issues [7], the ability ature difference prevents the gas entrance surfaces from to design preforms for specific applications [8], and the becoming sealed until after the interior of the component development of process models for use in optimization. has reached an acceptable density [6]. Fig. I illustrates This paper will review some current efforts in FCvi the competition between temperature and depletion/HCI process modeling and the development of a technologi- formation in the infiltration of SiC from MTS. Depletion cally useful geometry, a ceramic composite tube pre- is defined as the percentage of input MTS consumed in pared by FCV depositing SiC, and the resultant production of HCl via the re 2. Fundamentals of Cvi CH3SiCl3→SiC+3HCl (1) During CVI, the primary objective is to maximize The rates computed for the curves use the more complex the rate of matrix deposition and minimize density kinetic expressions that include the ' poisoning'effect of gradients. Unfortunately, there is an inl herent competi HCl [12], and we can see the strong effect of depletion ion between the deposition reaction and the mass (Fig. 1). It is this type of competitive relationship that transport of the gaseous reactant and product species permits high-density composites to be prepared by the The most common ceramic matrix is SiC, and its depos non-isothermal, non-isobaric processes ition via the thermal decomposition of methyltrichlorosi lane (MTS)will be considered the model system for the current discussion [9-l1 3. Modeling Cvi Deposition reactions that are too rapid usually result in severe density gradients where there is essentially The modeling of Cvi involves the mathematical complete densification near the external surfaces and description of transport and reaction phenomena within much lower densities in the interior regions. a simulation domain. Fundamental processes to be Alternatively, exceptionally slow deposition reactions modeled include heat transfer by conduction, convection require an uneconomically long time to densify a part. and radiation; transport and reaction of gaseous reactant Control of temperature is critical because chemical species; and pressure-driven gas flow Differential equa- reactions exhibit Arrhenius behavior such that the rates tions representing these phenomena can be written in increase exponentially with temperature. Therefore, rela- the following steady-state form: tively low temperatures slow the deposition rate substan- v(pud)=v(vo)+S, tially more than does diffusion. In addition, low pressure decreases concentration and the diffusive flux to the where o is temperature, pressure or concentration, u is deposition surface. The deposition down the length of the gas velocity, p and I are constants, and s is a source a pore provides a simple model of this problem, and the term. Using the finite volume method of Patankar [13] effect of the CVi parameters of temperature and total the discretized version of this equation is solved over
K.J. Probst et al. / Surface and Coatings Technology 120–121 (1999) 250–258 251 The most widely used commercial process is pressure can be clearly seen in the review by Naslain isothermal/isobaric CVI (ICVI) which depends only on [7]. Assuming a first-order reaction for the simulation diffusion for species transport. It generally operates at of the deposition of SiC, deposit thickness profiles show reduced pressure (1–10 kPa) for deposition rate control. smaller gradients at lower temperature and pressure. This diffusion-dependent process is slow, requiring sev- Several CVI efforts have described non-isothermal, eral week-long infiltration times. It is commercially non-isobaric processes. In these systems, there are addiattractive, however, because large numbers of parts of tional complexities associated with heat transfer and varying dimensions are easily accommodated in a forced convection as infiltration proceeds and the pore single reactor. structure of the material changes. To accurately describe The forced-flow/thermal-gradient technique (FCVI) infiltration behavior, descriptions of the complexities developed at Oak Ridge National Laboratory (ORNL) associated with these transport processes should be overcomes the problems of slow diffusion and restricted combined with the complex microstructure evolution permeability, and has demonstrated a capability to and chemical kinetics. Regardless of these complexities produce thick-walled, simple-shaped, SiC-matrix com- and the current inability to fully describe them, underponents in times of the order of hours [4–6 ]. standing the general relationships between the relevant Chemical vapor infiltration is a maturing technology kinetic processes has led to the practical solution of with specialty manufacturing in place and the emergence many CVI problems. Simple reactant gas depletion of broader markets imminent. A number of recent within the CVI reactor or poisoning of the deposition advances have made the technology more attractive and process by reaction byproducts (e.g., HCl in the CVI of provided a broader base for applications. These include SiC from MTS) also reduces the deposition rate [12]. obtaining an understanding of fiber/matrix interface In the thermal gradient process, controlling the temperchemical and thermomechanical issues [7], the ability ature difference prevents the gas entrance surfaces from to design preforms for specific applications [8], and the becoming sealed until after the interior of the component development of process models for use in optimization. has reached an acceptable density [6]. Fig. 1 illustrates This paper will review some current efforts in FCVI the competition between temperature and depletion/HCl process modeling and the development of a technologi- formation in the infiltration of SiC from MTS. Depletion cally useful geometry, a ceramic composite tube pre- is defined as the percentage of input MTS consumed in pared by FCVI. depositing SiC, and the resultant production of HCl via the reaction: 2. Fundamentals of CVI CH3 SiCl3 H2 SiC+3HCl. (1) The rates computed for the curves use the more complex During CVI, the primary objective is to maximize kinetic expressions that include the ‘poisoning’ effect of the rate of matrix deposition and minimize density HCl [12], and we can see the strong effect of depletion gradients. Unfortunately, there is an inherent competi- (Fig. 1). It is this type of competitive relationship that tion between the deposition reaction and the mass permits high-density composites to be prepared by the transport of the gaseous reactant and product species. non-isothermal, non-isobaric processes. The most common ceramic matrix is SiC, and its deposition via the thermal decomposition of methyltrichlorosilane (MTS) will be considered the model system for the current discussion [9–11]. 3. Modeling CVI Deposition reactions that are too rapid usually result The modeling of CVI involves the mathematical in severe density gradients where there is essentially complete densification near the external surfaces and description of transport and reaction phenomena within much lower densities in the interior regions. a simulation domain. Fundamental processes to be Alternatively, exceptionally slow deposition reactions modeled include heat transfer by conduction, convection require an uneconomically long time to densify a part. and radiation; transport and reaction of gaseous reactant Control of temperature is critical because chemical species; and pressure-driven gas flow. Differential equareactions exhibit Arrhenius behavior such that the rates tions representing these phenomena can be written in the following steady-state form: increase exponentially with temperature. Therefore, relatively low temperatures slow the deposition rate substan- V(ruw)=V·(CVw)+S, (2) tially more than does diffusion. In addition, low pressure decreases concentration and the diffusive flux to the where w is temperature, pressure or concentration, u is deposition surface. The deposition down the length of the gas velocity, r and C are constants, and S is a source a pore provides a simple model of this problem, and the term. Using the finite volume method of Patankar [13], effect of the CVI parameters of temperature and total the discretized version of this equation is solved over
K.J. Probst et al./ Surface and Coatings Technology 120-121(1999)250-258 20%Depletion ≌ 1373 1323 @1273K 1273 Fig 1. Plot of computed Sic deposition rate versus temperature or percentage depletion of 20 mol% MTS in hydrogen, illustrating the compet- the simulation domain that is divided into control radiating surfaces by standard formulas based on the volume elements nodal temperatures. To obtain a self-consistent solution, several iterations of the heat transfer equation are 3.1. Heat transfer required to incorporate the non-linear behavior of the radiant energy exchange The heat transfer equation contains both diffusion and convection components and a source term such 3. 2. Mass transport and reaction (Cp7)=V·(kV7)+ The flow of the carrier gas and the concentrations of reacting species are determined by differential equations where Cp is the heat capacity of the flowing gas and K in the form of Eq (2). For pressure-driven gas flow: is the thermal conductivity of the material. The source term S contains any heat generated or absorbed by a volume element such as the heat from chemical reac- tions. This source term can also be used to account for thermal radiation where k is the Darcy permeability for the material of Calculation of the diffusive and convective contribu- each volume element, V is the gas molar volume, P is tions to the heat balance for each volume element is the total pressure, and u is the gas viscosity. This straightforward given the flow rate and heat capacity of formulation of gas transport does not include source or the gas, the thermal conductivities of the materials and convective (inertial)terms. It will not be accurate for the thermal boundary conditions. The heat-flux terms high-velocity gas flow in open reactors but is suitable for each volume element depend only on these quantities for the pressure-driven gas flow through semi-permeable and on the temperatures of adjacent volume elements materials the case for FCVI. The transport equa- Since the radiation contribution may depend on the tion for the reacting species includes convection, diffu temperatures of non-adjacent volume elements, it cannot sion and source terms be included as a flux term in the same manner. Instead the radiation contribution is calculated and included 卩(uC)=VDVC)+S a source term Based on the discretization of the simula where Ci is the species concentration and D is the tion domain, a ray-tracing program calculates the view effective diffusion coefficient for species i. Both MTS factors of the control volume surfaces. During the and HCl concentrations are included in the transport solution of the heat transfer equation, the view factors rates. The SiC matrix is deposited on the fiber surfaces are used to calculate the energy exchange between the throughout the preform volume. The matrix deposition
252 K.J. Probst et al. / Surface and Coatings Technology 120–121 (1999) 250–258 Fig. 1. Plot of computed SiC deposition rate versus temperature or percentage depletion of 20 mol% MTS in hydrogen, illustrating the competing effects. the simulation domain that is divided into control radiating surfaces by standard formulas based on the volume elements. nodal temperatures. To obtain a self-consistent solution, several iterations of the heat transfer equation are 3.1. Heat transfer required to incorporate the non-linear behavior of the radiant energy exchange. The heat transfer equation contains both diffusion and convection components and a source term such 3.2. Mass transport and reaction that: The flow of the carrier gas and the concentrations of V(Cp uT)=V· (KVT)+S, (3) reacting species are determined by differential equations where Cp is the heat capacity of the flowing gas and K in the form of Eq. (2). For pressure-driven gas flow: is the thermal conductivity of the material. The source term S contains any heat generated or absorbed by a V· A k mV VP B =0, (4) volume element such as the heat from chemical reactions. This source term can also be used to account for where k is the Darcy permeability for the material of thermal radiation. Calculation of the diffusive and convective contribu- each volume element, V is the gas molar volume, P is the total pressure, and m is the gas viscosity. This tions to the heat balance for each volume element is formulation of gas transport does not include source or straightforward given the flow rate and heat capacity of convective (inertial ) terms. It will not be accurate for the gas, the thermal conductivities of the materials, and high-velocity gas flow in open reactors but is suitable the thermal boundary conditions. The heat-flux terms for the pressure-driven gas flow through semi-permeable for each volume element depend only on these quantities and on the temperatures of adjacent volume elements. materials, as is the case for FCVI. The transport equaSince the radiation contribution may depend on the tion for the reacting species includes convection, diffusion and source terms: temperatures of non-adjacent volume elements, it cannot be included as a flux term in the same manner. Instead, V(uCi )=V·(Deff i VCi )+S, (5) the radiation contribution is calculated and included as a source term. Based on the discretization of the simula- where Ci is the species concentration and Deff i is the tion domain, a ray-tracing program calculates the view effective diffusion coefficient for species i. Both MTS factors of the control volume surfaces. During the and HCl concentrations are included in the transport solution of the heat transfer equation, the view factors rates. The SiC matrix is deposited on the fiber surfaces are used to calculate the energy exchange between the throughout the preform volume. The matrix deposition
K.J. Probst et al./ Surface and Coatings Technology 120-121(1999)250-258 253 rate depends on the temperature and concentrations of ized into four radial and 29 axial volume elements. The both species reason for only four radial elements through the wall The coupled systems of differential equations for thickness even though there are large thermal and temperature, pressure and chemical species concen- concentration gradients is the inho tration are solved in the steady state. For a selected time material. The 1.6 mm radial elements capture a maxi- increment, the local reaction rate is used to calculate a mum of three cloth layers, thus smaller elements would new density for each preform volume element. A new not capture sufficient material to make average property steady-state solution is then calculated and the density values meaningful incremented again, producing a series of snapshots'of The transport properties for all materials in the model the densification process domain are defined in separate material files. The ther- mal conductivities for the graphite, graphite felt, stain- less steel and ceramic fiberboard are functions of 4. Modeling tubular geometry for FCVI temperature. The gas mixture's heat capacity, viscosity and thermal conductivity, and the binary diffusivities of The model has been applied to a system for prepara- MTS and HCl in hydrogen, are also functions of tempe tion of composite tubes by FCVI. The preform is made ature. The thermal conductivity, permeability, effective up of concentric layers of Nextel@ 312(alu diffusivity and surface area of the preform are functions icate fiber, 3M Company, Minneapolis, MN) fibrous of density For the calculations used here, the initial tubular sleeves. The model calculations are performed in cylindrical geometry due to its ease of use and to be significantly lower than that determined from the symmetry. The FCVI experimental reactor configuration total fiber surface(850 cm-1 versus 1500 cm)due to used is shown in Fig. 2. A graphite coating chamber fiber tow contacts that limit the available area for matrix radiatively heats the fibrous tubular preform exterior deposition and its interior is cooled with a water-cooled line. The Several key boundary conditions are applied to the MTS carried in hydrogen is injected inside the preform model. A constant temperature of 50.C is assumed The gas mixture infiltrates through the preform thickne along the centerline of the water-cooled line. Both tota and exhausts at atmospheric pressure. Ceramic fiber- molar flux and MTS mole fraction are specified at the board is used to seal the preform ends. Graphite felt hydrogen/MTS inlet and atmospheric pressure is fixed insulation is placed on both ends of the preform to at the gas exhaust educe axial heat loss Figs 3-6 describe densification utilizing the CVI The reactor configuration shown in Fig. 2 is discret- model. The baseline FCVI parameters used are ized into a two-dimensional array of finite volume elements. Neglect of the circumferential direction coating chamber mid-line temperature: 1200.C reduces the cylindrical discretization to only radial and hydrogen/ MTS molar feed ratio axial components. The preform is 37 cm in length and total flow(STP) 6I min 6. 4 mm in thickness with an inside diameter of 5.1 cm and has a variable fiber volume. The array size chosen The initial temperature profile produced by the model for the model domain is 35 radial volume elements by for r the uninfiltrated preform is displayed in Fig 3. The 49 axial volume elements. The preform itself is discret- model utilized boundary values provided by measure Heating element Coating Chambe (1200°C) Exhaust MTS/H 「 Cooling line Graphite Felt 50°c cale Fiberboard Fig 2 Schematic view of the tubular FCVi reactor illustrating the gas flow paths, heat source and centerline cooling
K.J. Probst et al. / Surface and Coatings Technology 120–121 (1999) 250–258 253 rate depends on the temperature and concentrations of ized into four radial and 29 axial volume elements. The both species. reason for only four radial elements through the wall The coupled systems of differential equations for thickness even though there are large thermal and temperature, pressure and chemical species concen- concentration gradients is the inhomogeneity of the tration are solved in the steady state. For a selected time material. The 1.6 mm radial elements capture a maxiincrement, the local reaction rate is used to calculate a mum of three cloth layers, thus smaller elements would new density for each preform volume element. A new not capture sufficient material to make average property steady-state solution is then calculated and the density values meaningful. incremented again, producing a series of ‘snapshots’ of The transport properties for all materials in the model the densification process. domain are defined in separate material files. The thermal conductivities for the graphite, graphite felt, stainless steel and ceramic fiberboard are functions of 4. Modeling tubular geometry for FCVI temperature. The gas mixture’s heat capacity, viscosity and thermal conductivity, and the binary diffusivities of The model has been applied to a system for prepara- MTS and HCl in hydrogen, are also functions of tempertion of composite tubes by FCVI. The preform is made ature. The thermal conductivity, permeability, effective up of concentric layers of Nextel@ 312 (alumino-borosil- diffusivity and surface area of the preform are functions icate fiber; 3M Company, Minneapolis, MN ) fibrous of density. For the calculations used here, the initial tubular sleeves. The model calculations are performed specific surface of the preform was arbitrarily assumed in cylindrical geometry due to its ease of use and to be significantly lower than that determined from the symmetry. The FCVI experimental reactor configuration total fiber surface (850 cm−1 versus 1500 cm−1) due to used is shown in Fig. 2. A graphite coating chamber fiber tow contacts that limit the available area for matrix radiatively heats the fibrous tubular preform exterior deposition. and its interior is cooled with a water-cooled line. The Several key boundary conditions are applied to the MTS carried in hydrogen is injected inside the preform. model. A constant temperature of 50°C is assumed The gas mixture infiltrates through the preform thickness along the centerline of the water-cooled line. Both total and exhausts at atmospheric pressure. Ceramic fiber- molar flux and MTS mole fraction are specified at the board is used to seal the preform ends. Graphite felt hydrogen/MTS inlet and atmospheric pressure is fixed insulation is placed on both ends of the preform to at the gas exhaust. reduce axial heat loss. Figs. 3–6 describe densification utilizing the CVI The reactor configuration shown in Fig. 2 is discret- model. The baseline FCVI parameters used are: ized into a two-dimensional array of finite volume elements. Neglect of the circumferential direction coating chamber mid-line temperature: 1200°C reduces the cylindrical discretization to only radial and hydrogen/MTS molar feed ratio: 5 axial components. The preform is 37 cm in length and total flow (STP): 6 l min−1 6.4 mm in thickness with an inside diameter of 5.1 cm, and has a variable fiber volume. The array size chosen The initial temperature profile produced by the model for the model domain is 35 radial volume elements by for the uninfiltrated preform is displayed in Fig. 3. The 49 axial volume elements. The preform itself is discret- model utilized boundary values provided by measureFig. 2. Schematic view of the tubular FCVI reactor illustrating the gas flow paths, heat source and centerline cooling
K.J. Probst et al./ Surface and Coatings Technology 120-121(1999)250-258 ■1100-1150 10501100 日10001050 口9501000 Normalized Thickness Normalized Length Fig 3 Computed thermal profile within the composite tube wall before the onset of infiltration. AT(C) 80 Infiltration Time(hours) Fig 4. Computed average thermal gradient through the wall of the tube as a function of infiltration time. ■700%750% 口5509600% Normalized Thickness 0.75 Fig. 5. Density profile within the tube wall after 12 h of infiltration. The values are relative to a full density value of 100%6, with an initial value of 35% mass volume of fiber ments of the temperature of the exterior graphite reactor a slightly higher radial gradient in the tubular preform chamber at three axial positions. The radial temperature region directly above the injector gradient is quite uniform along the preform length with Fig 4 shows the computed transient radial temp
254 K.J. Probst et al. / Surface and Coatings Technology 120–121 (1999) 250–258 Fig. 3. Computed thermal profile within the composite tube wall before the onset of infiltration. Fig. 4. Computed average thermal gradient through the wall of the tube as a function of infiltration time. Fig. 5. Density profile within the tube wall after 12 h of infiltration. The values are relative to a full density value of 100%, with an initial value of 35% mass volume of fiber. ments of the temperature of the exterior graphite reactor a slightly higher radial gradient in the tubular preform chamber at three axial positions. The radial temperature region directly above the injector. gradient is quite uniform along the preform length with Fig. 4 shows the computed transient radial temper-
K.J. Probst et al./ Surface and Coatings Technology 120-121(1999)250-258 255 ■85.0%90.0% ■80.0%85.0% 口70.0%7509 Normalized 0.75 Normalized Lengt Fig. 6. Density profile within the tube wall near completion (32 h of infiltration) ature gradient profile at a nominal axial position. As nated as reactants are directed to less infiltrated areas infiltration proceeds SiC matrix is deposited throughout due to their higher permeability the tubular fibrous preform, increasing the preform Utilizing the baseline process process parameters, an averag Fig.5. The density near the mid-line is greatest due to fied above. Shown are the results of compuai g p a? a thermal conductivity and reducing the radial temper- mass volume for the tubular components can be pre dicted as a function of infiltration time. Fig. 7 contain The computed mass volume(percent density) profile plots of the average density of an infiltrated tube as f the composite after 12 h of infiltration is shown in function of infiltration time under the conditions sp the initially higher temperatures in this volume causing assuming a fiber loading of 35 vol% for the tw more rapid SiC deposition. Also seen is an axial asymme- spe eCific area values, 1500 cm-1 and 850 cm/ o initial try due to the introduction of the reactant flow into the volume of the tube at the left side, as seen in Fig. 2 After 32 h(Fig. 6), near completion, the density distri- 5. Experiment bution has become more fat. however the effect of the on-uniform axial temperature profile and the introduc- Several tubular preforms have been infiltrated by tion of the precursor flow is still evident. As is the case FCVI utilizing the standard conditions described above in FCvi. such non-uniformities will eventually be elimi These tubular preforms were fabricated by pulling con- r50P% Model results -Initial specific area 1500 cm Initial specific area 850 cm Infiltration Time Fig. 7. Comparison of computed and experimental percentage mass volume( theoretical density) as a function of infiltration time
K.J. Probst et al. / Surface and Coatings Technology 120–121 (1999) 250–258 255 Fig. 6. Density profile within the tube wall near completion (32 h of infiltration). ature gradient profile at a nominal axial position. As nated as reactants are directed to less infiltrated areas infiltration proceeds SiC matrix is deposited throughout due to their higher permeability. the tubular fibrous preform, increasing the preform Utilizing the baseline process parameters, an average thermal conductivity and reducing the radial temper- mass volume for the tubular components can be preature gradient. dicted as a function of infiltration time. Fig. 7 contains The computed mass volume (percent density) profile plots of the average density of an infiltrated tube as a of the composite after 12 h of infiltration is shown in function of infiltration time under the conditions speciFig. 5. The density near the mid-line is greatest due to fied above. Shown are the results of computations the initially higher temperatures in this volume causing assuming a fiber loading of 35 vol% for the two initial more rapid SiC deposition. Also seen is an axial asymme- specific area values, 1500 cm−1 and 850 cm−1. try due to the introduction of the reactant flow into the volume of the tube at the left side, as seen in Fig. 2. After 32 h (Fig. 6), near completion, the density distri- 5. Experiment bution has become more flat; however, the effect of the non-uniform axial temperature profile and the introduc- Several tubular preforms have been infiltrated by tion of the precursor flow is still evident. As is the case FCVI utilizing the standard conditions described above. in FCVI, such non-uniformities will eventually be elimi- These tubular preforms were fabricated by pulling conFig. 7. Comparison of computed and experimental percentage mass volume (theoretical density) as a function of infiltration time
K.J. Probst et al./ Surface and Coatings Technology 120-121(1999)250-258 centric Nextel@ 312 braided tubular sleeves (2.75 tows cm )over a mandrel and rigidizing by resin impregnation and curing. The tube preform specific tions are inside diameter 5.08 outside diameter: 6.64 35.6cm fiber volume 34.4% Initial temperature measurements were obtained using optical pyrometry at three positions along the length of the reactor chamber: inlet side, 1087C; mid line. 1203C. and outlet side. 1104.C. Infiltration was allowed to proceed for specific time periods and then the reactant flow was replaced with inert gas and the system allowed to cool to room temperature 6. Results The results of four experimental runs are Table I and Fig. 7. The mass volume is determined from the geometric volume of the component and the weight gain. It is apparent that mass volume, or void fraction, changes non-linearly with infiltration time. The initial temperature gradient through the preform that limits and an infiltrated tube ettel 312 cloth sleeve, a rigidized prefor lower slope in the mass volume is due to the steep deposition of Sic due to the lower temperatures of the inner volume. As infiltration proceeds, the through thickness thermal conductivity of the component dark areas are void volume. The oval Nextel 312 fibers increases with the density, increasing the temperature of are apparent at higher magnification Fig 9(b)]. The lume and allowing more rapid infiltration gray phase is some minor amounts of residual carbon Fig. 8 is a photograph of a Nextel 312 cloth sleeve, a that resulted from the pyrolysis of the phenolic resin rigidized preform and a densified tubular composite. An used to rigidize the preform asymmetry in axial infiltration is seen in the partially infiltrated tubes. These have apparently higher densities near the precursor inlet side due to the tendency for the 7. Di reactants to initially flow out of the tube through the area of the wall closest to the inlet tube. as infiltration In this initial work, the comparison between experi progressed(for the tubes with longer infiltration times), ment and the model of mass volume infiltrated as a this effect was not as apparent. function of time shows reasonable agreement when the Fig 9 shows a typical microstructure for a component smaller value for initial specific area is used(Fig. 7) infiltrated to a relatively high density. The lower-magni- The model and experiment indicate an early exponential ation image[Fig 9(a)] reveals the fiber bundles within increase in density which, after 10 to 12 h, undergoes a the woven material. The white phase is deposited SiC, transition to a slower, monotonic increase that is typi- which is seen to infiltrate and overcoat the bundles. The cally seen in CVI [ll]. It is at this point that the fine FCVI experimental runs and modeling Identification time Observed average mass volume(%) Predicted average mass volume(%) CVI 1211 6 69.6 CvI 1218
256 K.J. Probst et al. / Surface and Coatings Technology 120–121 (1999) 250–258 centric Nextel@ 312 braided tubular sleeves (2.75 tows cm−1) over a mandrel and rigidizing by resin impregnation and curing. The tube preform specifications are: inside diameter: 5.08 cm outside diameter: 6.64 cm length: 35.6 cm fiber volume: 34.4% Initial temperature measurements were obtained using optical pyrometry at three positions along the length of the reactor chamber: inlet side, 1087°C; midline, 1203°C; and outlet side, 1104°C. Infiltration was allowed to proceed for specific time periods and then the reactant flow was replaced with inert gas and the system allowed to cool to room temperature. 6. Results The results of four experimental runs are given in Table 1 and Fig. 7. The mass volume is determined from the geometric volume of the component and the weight gain. It is apparent that mass volume, or void fraction, changes non-linearly with infiltration time. The initial lower slope in the mass volume is due to the steep Fig. 8. Photograph of a Nextel@ 312 cloth sleeve, a rigidized preform temperature gradient through the preform that limits and an infiltrated tube. deposition of SiC due to the lower temperatures of the inner volume. As infiltration proceeds, the throughthickness thermal conductivity of the component dark areas are void volume. The oval Nextel@ 312 fibers increases with the density, increasing the temperature of are apparent at higher magnification [Fig. 9(b)]. The the inner volume and allowing more rapid infiltration. gray phase is some minor amounts of residual carbon Fig. 8 is a photograph of a Nextel@ 312 cloth sleeve, a that resulted from the pyrolysis of the phenolic resin rigidized preform and a densified tubular composite. An used to rigidize the preform. asymmetry in axial infiltration is seen in the partially infiltrated tubes. These have apparently higher densities near the precursor inlet side due to the tendency for the 7. Discussion reactants to initially flow out of the tube through the area of the wall closest to the inlet tube. As infiltration In this initial work, the comparison between experiprogressed (for the tubes with longer infiltration times), ment and the model of mass volume infiltrated as a this effect was not as apparent. function of time shows reasonable agreement when the Fig. 9 shows a typical microstructure for a component smaller value for initial specific area is used (Fig. 7). infiltrated to a relatively high density. The lower-magni- The model and experiment indicate an early exponential fication image [Fig. 9(a)] reveals the fiber bundles within increase in density which, after 10 to 12 h, undergoes a the woven material. The white phase is deposited SiC, transition to a slower, monotonic increase that is typiwhich is seen to infiltrate and overcoat the bundles. The cally seen in CVI [11]. It is at this point that the fine Table 1 Results of FCVI experimental runs and modeling Identification Run time (h) Observed average mass volume (%) Predicted average mass volume (%) CVI 1211 6 45.6 39.7 CVI 1223 12 67.6 54.0 CVI 1217 24 74.5 69.6 CVI 1218 32 78.0 79.5
K.J. Probst et al./ Surface and Coatings Technology 120-121(1999)250-258 Fig 9. Optical micrographs of a polished cross-section of an infiltrated composite at (a) low and (b) high magnification. porosity within the fiber bundles is either fully infiltrated 8. Conclusions or the bundle surface is sufficiently sealed so that the igh-surface-area bundles are no longer available for The modeling efforts for the tubular FCVI system deposition. Since the mass deposition rate is propor- utilizing observed temperature boundaries have indi ional to surface area, the result is a sharp decline in cated a non-uniform, initial densification profile. The the rate of mass gain. Thus, the model accurately profile, however, becomes significantly more uniform presents the onset of this transition ith time, as expected, in the self-correcting FCV The predicted axial asymmetry due to the precursor process. inlet being positioned at one end of the preform appears The initial experiments have shown that densification to be qualitatively observed in the partially infiltrated generally follows the trends indicated by the modeling tubes. This is an effect of the precursor flow predomi- results. Discrepancies in the relative values highlight the nantly following a path through the preform wall closest need to obtain improved property data as a function to the inlet tube. As the density of that region grows of densit disproportionately compared with the downstream As has been the case with the development of FCVI volume, its permeability also decreases disproportion- densified plates, it is expected that the production of ately. The result is that, with time, a greater flux of high-density tubes should be possible. The current work precursor will be directed to the lower-density, down- is encouraging for the relatively high densities already stream regions, eventually obtaining uniform achieved and for the semi-quantitative agreement with densification the model. Increased accuracy in the model should result he model results are governed by the from improved property modules and will allow deriva properties input to the computations. These tion of processing conditions for more efficient fabrica- critical relationships such as the permeability ar tion of high-density tub mal conductivity of the material as a function of density and orientation. Currently, there is little information vith regard to these variables; thus the relationships were largely based on earlier work with other fiber systems [14-16]. The observed results support the need Acknowledgements the objects of the current work which has been one of The authors would like to thank t.s. geer for the The model and experimental results both indicate metallographic work, R.E. Dearien for help in manu that the mass volume can be increased with greater script preparation, and A. A Wereszczak and J w. Klett infiltration times( Fig. 7). The mass volume is, however, for their valuable comments. The research was spon ultimately limited by the fiber architecture to approxi- sored by the US Department of Energy, Office of Fossil mately 90%[14-16]. This limitation is a feature of the Energy, Advanced Research and Technology pore-size distribution and structure that causes some Development Materials Program, under contract porosity to be closed before it can be filled DEAC05-96OR22464
K.J. Probst et al. / Surface and Coatings Technology 120–121 (1999) 250–258 257 Fig. 9. Optical micrographs of a polished cross-section of an infiltrated composite at (a) low and (b) high magnification. porosity within the fiber bundles is either fully infiltrated 8. Conclusions or the bundle surface is sufficiently sealed so that the high-surface-area bundles are no longer available for The modeling efforts for the tubular FCVI system deposition. Since the mass deposition rate is propor- utilizing observed temperature boundaries have inditional to surface area, the result is a sharp decline in cated a non-uniform, initial densification profile. The the rate of mass gain. Thus, the model accurately profile, however, becomes significantly more uniform represents the onset of this transition. with time, as expected, in the self-correcting FCVI The predicted axial asymmetry due to the precursor process. inlet being positioned at one end of the preform appears The initial experiments have shown that densification to be qualitatively observed in the partially infiltrated generally follows the trends indicated by the modeling tubes. This is an effect of the precursor flow predomi- results. Discrepancies in the relative values highlight the nantly following a path through the preform wall closest need to obtain improved property data as a function to the inlet tube. As the density of that region grows of density. disproportionately compared with the downstream As has been the case with the development of FCVIvolume, its permeability also decreases disproportion- densified plates, it is expected that the production of ately. The result is that, with time, a greater flux of high-density tubes should be possible. The current work precursor will be directed to the lower-density, down- is encouraging for the relatively high densities already stream regions, eventually obtaining uniform achieved and for the semi-quantitative agreement with densification. the model. Increased accuracy in the model should result The model results are governed by the material from improved property modules and will allow derivaproperties input to the computations. These include tion of processing conditions for more efficient fabricacritical relationships such as the permeability and ther- tion of high-density tubes. mal conductivity of the material as a function of density and orientation. Currently, there is little information with regard to these variables; thus the relationships were largely based on earlier work with other fiber systems [14–16]. The observed results support the need Acknowledgements to improve these relationships, which has been one of the objects of the current work. The authors would like to thank T.S. Geer for the The model and experimental results both indicate metallographic work, R.E. Dearien for help in manuthat the mass volume can be increased with greater script preparation, and A.A. Wereszczak and J.W. Klett infiltration times (Fig. 7). The mass volume is, however, for their valuable comments. The research was sponultimately limited by the fiber architecture to approxi- sored by the US Department of Energy, Office of Fossil mately 90% [14–16]. This limitation is a feature of the Energy, Advanced Research and Technology pore-size distribution and structure that causes some Development Materials Program, under contract porosity to be closed before it can be filled. DE-AC05-96OR22464
J. Probst et al./ Surface and Coatings Technology 120-121(1999)250-258 References [10]TL. Starr, D.Y. Chiang, T M. Besmann, D P. Stinton, J.C. McLaughlin, W M. Matlin, in: K.v. Logan, Z.A. Munir, R. M. [IR. L. Bickerdike, A R.G. Brown, G. Hughes, H. Ranson, in: S Spriggs (Eds ) Ceramic Transactions 79 vol. 79, American Mrosowski, M.C. Studebaker, P L. Walker(Eds ) Proc. Fifth Ceramic Society, Westerville, OH, 1996, P. 127. Confon Carbon vol I, Pergamon Press, New York, 1962, p. 575. [11]TL. Starr, in: T.M. Besmann, M D. Allendorf, McD.Robinson, [2]WC. Jenkins, US Patent No. 3 160517, 1964. R K. Ulrich( Eds ) Proceedings of the Thirteenth Conference on [3]TM. Besmann, B.W. Sheldon, R.A. Lowden, D P Stinton, Sci- CVD, The Electrochemical Society, Pennington, NJ, 1996, p 541 ence253(1991)1104 [12]T M. Besmann, B w. Sheldon, T.S. Moss Ill,, M D. Kaster [4]WJ. Lackey, A.J. Caputo, US Patent No 4 580524, 1986 J.Am. Ceram.Soc.75(10(1992)2899 [5]TM. Besmann, D P Stinton, R.A. Lowden, MRS Bull. XIlI(In) [13]SV. Patankar, Numerical Heat Transfer and Fluid Flow, Hemi- 1988)45 sphere Publishing Corporation, New York, 1980. [6D. P. Stinton, T M. Besmann, R.A. Lowden, Am. Ceram Soc. []T M. Besmann, W.M. Matlin, D P Stinton, in: A F. Hepp, PN Bull.67(2)(1988)350. Kumta, J.J. Sullivan, G.S. Fischman, A E. Kaloyeros(Eds ) [7]R.R. Naslain, Solid State ionics 101-103(1997)959 Covalent Ceramics Ill- Science and Technology of Non-Oxides, [8]A. Muhlratzer, P. Agatonovic H. Koberle, K. Wildenrotter, Ind MRS Symposium Proceedings vol. 410, Materials Research Soci- Ceran.16(2)(1996)111 ety, Pittsburgh, PA, 1996, p. 441 [9]D.Y. Chiang, T L. Starr, in: W.S. Johnson(Ed ) Proceedings of [15]T L. Starr, N. Hablutzel, J Am Ceram Soc. 81(5)(1998)1298 the American Society for Composites: Eleventh Technical Confer- [16]S.B. Lee, S.R. Stock, M.D. Butts, T.L. Starr, T.M. Breunig, J H ence, Technomic Publishing, Lancaster, PA, 1996, p. 381 Kinney, J Mater Res. 13(5)(1998)1209
258 K.J. Probst et al. / Surface and Coatings Technology 120–121 (1999) 250–258 References [10] T.L. Starr, D.Y. Chiang, T.M. Besmann, D.P. Stinton, J.C. McLaughlin, W.M. Matlin, in: K.V. Logan, Z.A. Munir, R.M. Spriggs (Eds.), Ceramic Transactions 79 vol. 79, American [1] R.L. Bickerdike, A.R.G. Brown, G. Hughes, H. Ranson, in: S. Ceramic Society, Westerville, OH, 1996, p. 127. Mrosowski, M.C. Studebaker, P.L. Walker (Eds.), Proc. Fifth [11] T.L. Starr, in: T.M. Besmann, M.D. Allendorf, McD. Robinson, Conf. on Carbon vol. I, Pergamon Press, New York, 1962, p. 575. R.K. Ulrich (Eds.), Proceedings of the Thirteenth Conference on [2] W.C. Jenkins, US Patent No. 3 160 517, 1964. CVD, The Electrochemical Society, Pennington, NJ, 1996, p. 541. [3] T.M. Besmann, B.W. Sheldon, R.A. Lowden, D.P. Stinton, Sci- [12] T.M. Besmann, B.W. Sheldon, T.S. Moss III,, M.D. Kaster, ence 253 (1991) 1104. J. Am. Ceram. Soc. 75 (10) (1992) 2899. [4] W.J. Lackey, A.J. Caputo, US Patent No. 4 580 524, 1986. [13] S.V. Patankar, Numerical Heat Transfer and Fluid Flow, Hemi- [5] T.M. Besmann, D.P. Stinton, R.A. Lowden, MRS Bull. XIII (11) sphere Publishing Corporation, New York, 1980. (1988) 45. [14] T.M. Besmann, W.M. Matlin, D.P. Stinton, in: A.F. Hepp, P.N. [6] D.P. Stinton, T.M. Besmann, R.A. Lowden, Am. Ceram. Soc. Bull. 67 (2) (1988) 350. Kumta, J.J. Sullivan, G.S. Fischman, A.E. Kaloyeros (Eds.), Covalent Ceramics III – Science and Technology of Non-Oxides, [7] R.R. Naslain, Solid State Ionics 101–103 (1997) 959. MRS Symposium Proceedings vol. 410, Materials Research Soci- [8] A. Mu¨hlratzer, P. Agatonovic¸, H. Ko¨berle, K. Wildenrotter, Ind. Ceram. 16 (2) (1996) 111. ety, Pittsburgh, PA, 1996, p. 441. [9] D.Y. Chiang, T.L. Starr, in: W.S. Johnson (Ed.), Proceedings of [15] T.L. Starr, N. Hablutzel, J. Am. Ceram. Soc. 81 (5) (1998) 1298. the American Society for Composites: Eleventh Technical Confer- [16] S.-B. Lee, S.R. Stock, M.D. Butts, T.L. Starr, T.M. Breunig, J.H. ence, Technomic Publishing, Lancaster, PA, 1996, p. 381. Kinney, J. Mater. Res. 13 (5) (1998) 1209