Materials Science and Engineering, R20(1997)37-124 R Reports: A Review Journal Rapid vapor-phase densification of refractory composites AlliedSignal, Inc, Mail Stop CTC-1, 101 Columbia Road, Morristown, NJ 07962, USA Received 5 March 1996; accepted 3 January 1997 Abstract The status of vapor-phase routes for the rapid densification of high-tem primarily ccramic-matrix composites, is revicwed. Conventional densification of composites such as carbon- carbon and Sic-Sic is accomplished by isothermal, isobaric chemical vapor infiltration(CVi), either alone or in combination with liquid resin impregnation and thermal annealing These are multi-step processes which ake from several hundred to thousands of hours at high temperature. In this paper we review approaches designed to significantly reduce the processing time and the number of steps required for densification, while producing materials with the desired properties. We describe techniques such as inductively-heated thermal- gradient isobaric CVI, radiantly-heated isothermal and thermal-gradient forced-fow CVI, liquid-immersion thermal-gradient CVI and plasma-enhanced CVi. Different heating methods, such as radiative and inductive, and both hot-wall and cold-wall reactors are compared. Available material properties of composites produced by these techniques are given. o 1997 Elsevier Science S.A. filtration: Fiber; Forced flow; High-temperature composites; Inductive heating; Isobaric Isothermal fication: SiC. 地二 Microwave; Plasma; Preform; Pulsed pressure; Radiant heating; Rapid densi- 1. Preface Composite materials, such as carbon-carbon(C-C), offer advantages of light weight and excel lent mechanical and thermal properties, especially for high-temperature applications, e. g. aircraft brake pads, uncooled engine and other airplane parts and leading-edge sections used in rockets [1-4]. Other applications which are receiving attention are in heat-conducting substrates for electronic chips and in sports equipment, for example for tennis, golf and bicycling. Composites generally consist of fibers surrounded by a matrix. Compared to their monolithic counterparts, composites are much tougher mechanically and allow tailoring of and thermal conductivity to a much larger extent Composites also have a much more forgiving failure mode under stress than monolithics, as illustrated in Fig. 1. There are generally three categories of composite materials: polymer-matrix composites, metal-matrix composites and ceramic-matrix com- posites. In this review, we are concerned with ceramic-matrix composites(CMCs), fabricated mainly around carbon and SiC fibers; examples are C-C and Sic-SiC composites. These materials can ithstand the highest use temperatures. Other fibers studied or used to different extents include glass silica, alumina, alumino-silicates, silica-titania, silicon nitride, zirconia, yttrium-aluminum garnets boron-coated tungsten and boron nitride [5]. One of the most common fabrication methods of such osite structures is densification of a porous body, the preform, having the desired shape and isting solely or principally of fibers. The fibers may be continuous or chopped. The preform may Corresponding author. Tel:(201)455-4938 Fax:(201)455-3008 or (201)455-3942. E-mail: GOLECKIGRESEARCH COM 927-796x/97/$17.00 e 1997 Elsevier Science S.A. All rights reserved PS0927-796X(97)00003X
Materials Science ana’ Engineering, R20 (1997) 37-124 Rapid vapor-phase densification of refractory composites I. Golecki * AlliedSignal, Inc., Mail Stop CTC-I, IO1 Columbia Road, Morristown, NJ 07962, USA Received 5 March 1996; accepted 3 January 1997 Abstract The status of vapor-phase routes for the rapid densification of high-temperature composite materials, primarily ceramic-matrix composites, is reviewed. Conventional densification of composites such as carboncarbon and Sic-Sic is accomplished by isothermal, isobaric chemical vapor infiltration (CVI), either alone or in combination with liquid resin impregnation and thermal annealing. These are multi-step processes which take from several hundred to thousands of hours at high temperature. In this paper we review approaches designed to significantly reduce the processing time and the number of steps required for densification, while producing materials with the desired properties. We describe techniques such as inductively-heated thermalgradient isobaric CVI, radiantly-heated isothermal and thermal-gradient forced-flow CVI, liquid-immersion thermal-gradient CVI and plasma-enhanced CVI. Different heating methods, such as radiative and inductive, and both hot-wall and cold-wall reactors are compared. Available material properties of composites produced by these techniques are given. 0 1997 Elsevier Science S.A. Keywords: Carbon-carbon composites; Ceramic-matrix composites; Chemical vapor deposition; Chemical vapor infiltration; Composites; Densification; Fiber; Forced flow; High-temperature composites; Inductive heating; Isobaric; Isothermal; Liquid immersion; Matrix: Microwave; Plasma; Preform; Pulsed pressure; Radiant heating; Rapid densification; Sic-Sic; Thermal gradient 1. Preface Composite materials, such as carbon-carbon (C-C), offer advantages of light weight and excellent mechanical and thermal properties, especially for high-temperature applications, e.g. aircraft brake pads, uncooled engine and other airplane parts and leading-edge sections used in rockets [ l-41. Other applications which are receiving attention are in heat-conducting substrates for electronic chips and in sports equipment, for example for tennis, golf and bicycling. Composites generally consist of fibers surrounded by a matrix. Compared to their monolithic counterparts, composites are much tougher mechanically and allow tailoring of the thermal properties, such as the thermal expansion coefficient and thermal conductivity to a much larger extent. Composites also have a much more forgiving failure mode under stress than monolithics, as illustrated in Fig. 1. There are generally three categories of composite materials: polymer-matrix composites, metal-matrix composites and ceramic-matrix composites. In this review, we are concerned with ceramic-matrix composites (CMCs) , fabricated mainly around carbon and Sic fibers: examples are C-C and Sic-Sic composites. These materials can withstand the highest use temperatures. Other fibers studied or used to different extents include glass, silica, alumina, alumino-silicates, silica-titania, silicon nitride, zirconia, yttrium-aluminum garnets, boron-coated tungsten and boron nitride [ 51. One of the most common fabrication methods of such composite structures is densification of a porous body, the preform, having the desired shape and consisting solely or principally of fibers. The fibers may be continuous or chopped. The preform may * Corresponding author. Tel: (201) 4554938. Fax: (201) 455-3008 or (201) 455-3942. E-mail: GOLECKIQRESEARCH. ALLLED.COM. 0927-796X/97/$17.00 8 1997 Elsevier Science S.A. All rights reserved. PZZSO927-796X(97)00003-X
Monolithic Fiber-reinforced ceramIc Displacement or Surai ig. 1. Typical stress-strain curves for a monolithic ceramic, exhibiting brittle fracture, and a tougher fiber-reinforced composite ceramic, exhibiting more extensive displacement due to fiber pullout bc fabricated using techniques such as weaving of continuous fibers, needle-punching of fibrous mats, or mixing of chopped fibers with resins and powders, followed by thermal treatment in the 200-1000 C range lo evaporate organic binders or residues [1]. The geometrical density of such preforms varies widely in the range 10-80% of the theoretical value. In this review we describe the densification of such porous preform structures by means of chemical vapor deposition(CvD) and chemical vapor infiltration(CVI). In particular, we concentrate on those vapor infiltration methods which result in a significantly shorter densification time than achieved by conventional routes. Cvd and Cvi involve flowing one or several streams of precursor vapors containing the desired element or compound over and around the porous part, while keeping that part at a temperature sufficient to decompose the precursor. Temperatures can vary, for example, in the 600-1500'C range, depending on the particular hemistry and system. Total pressures are generally in the range 10to 10 Torr. Under the appropriate conditions, the vapor decomposes to produce the desired element or compound in the desired micros- tructure within the pores of the part, thus increasing its density. Minimum final density values are he desired mechanical and thermal compared to other densification methods, such as multiple cycles of liquid resin impregnation and high-temperature treatment. CVi allows penetration of the desired atoms or molecules into the smallest pores of the preform and does not require post-densification treatment to remove organics. CVI produces uniform and conformal coatings around each accessible fiber and surface in the preform. The final shape of a part densified by CVi is closest to the desired shape(so-called near-net or net shape) so that only minimal or no post-densification machining is required. On the other hand, liquid impreg nation usually results in major shrinkage and microcracking during the required thermal annealing (curing and pyrolysis) cycles. After densification of the composite part by any of the above methods, additional heat or surface treatments may be required, for example, to improve the physical properties of the part or its resistance to environmental attack(e.g. oxidation), depending upon the specific applications. These additional treatments are not covered in this review In CVI, the deposition rate usually increases moderately with increasing precursor partial pressure, p and exponentially with increasing substrate temperature, T, as p"exp(-AH/kr), where n is the pressure exponent, AH is the activation energy (24eV molecule" for carbon)andk=1.3805x10-16 ergk-l=8.614X10-Sevk-is Boltzmann's constant Pressure in this context signifies thepressure in the reactor chamber. A common application of Cvi involves densification of porous carbon sub- strates, thicker than 2.5 cm, where a large number of such substrates may be placed in an enclosure uniformly heated to a temperature in the range 1000-1100"C and exposed to a reactant gas, e.g methane atp-5-100 Torr [3]. This approach is known as hot-wall CVI and its major drawback is an extremely long CVi time of 600-2000 h to achieve the desired density. Furthermore, the process must usually be interrupted several times to permit grinding of the exterior surfaces of the substrates in order to open the pores and allow further infiltration. For practical reasons, it is desirable to reduce the
38 I. Golecki / Rapid vapor-phase densijication of refractory composites Displacement or Strain Fig. 1. Typical stress-strain curves for a monolithic ceramic, exhibiting brittle fracture, and a tougher fiber-reinforced composite ceramic, exhibiting more extensive displacement due to fiber pullout. be fabricated using techniques such as weaving of continuous fibers, needle-punching of fibrous mats, or mixing of chopped fibers with resins and powders, followed by thermal treatment in the 200-1000 “C range to evaporate organic binders or residues [ 11. The geometrical density of such preforms varies widely in the range lO-80% of the theoretical value. In this review we describe the densification of such porous preform structures by means of chemical vapor deposition (CVD) and chemical vapor infiltration (CVI) . In particular, we concentrate on those vapor infiltration methods which result in a significantly shorter densification time than achieved by conventional routes. CVD and CVI involve flowing one or several streams of precursor vapors containing the desired element or compound over and around the porous part, while keeping that part at a temperature sufficient to decompose the precursor. Temperatures can vary, for example, in the 600-l 500 “C range, depending on the particular chemistry and system. Total pressures are generally in the range 10m3 to lo3 Torr. Under the appropriate conditions, the vapor decomposes to produce the desired element or compound in the desired microstructure within the pores of the part, thus increasing its density. Minimum final density values are necessary for achieving the desired mechanical and thermal properties. CVI has several advantages compared to other densification methods, such as multiple cycles of liquid resin impregnation and high-temperature treatment. CVI allows penetration of the desired atoms or molecules into the smallest pores of the preform and does not require post-densification treatment to remove organics. CVI produces uniform and conformal coatings around each accessible fiber and surface in the preform. The final shape of a part densified by CVI is closest to the desired shape (so-called near-net or net shape), so that only minimal or no post-densification machining is required. On the other hand, liquid impregnation usually results in major shrinkage and microcracking during the required thermal annealing (curing and pyrolysis) cycles. After densification of the composite part by any of the above methods, additional heat or surface treatments may be required, for example, to improve the physical properties of the part or its resistance to environmental attack (e.g. oxidation), depending upon the specific applications. These additional treatments are not covered in this review. In CVI, the deposition rate usually increases moderately with increasing precursor partial pressure, p and exponentially with increasing substrate temperature, T, as p” exp( - AH/kT), where n is the pressure exponent, AH is the activation energy (2-4 eV molecule -‘forcarbon) andk= 1.3805X lo-l6 erg K - ’ = 8.614 X 10m5 eV K- ’ is Boltzmann’s constant. Pressure in this context signifies the pressure in the reactor chamber. A common application of CVI involves densitication of porous carbon substrates, thicker than 2.5 cm, where a large number of such substrates may be placed in an enclosure uniformly heated to a temperature in the range 1000-l 100 “C and exposed to a reactant gas, e.g. methane at p = 5-l 00 Torr [ 31. This approach is known as hot-wall CVI and its major drawback is an extremely long CVI time of 6UO-2OUO h to achieve the desired density. Furthermore, the process must usually be interrupted several times to permit grinding of the exterior surfaces of the substrates in order to open the pores and allow further infiltration. For practical reasons, it is desirable to reduce the
I Golecki/Ropid vapor-phase densification of refractory composites processing time. However, increasing the precursor pressure and/or temperature beyond certain ranges may produce deleterious effects, such as: (1) homogeneous nucleation of powders(soot)in the gas phase instead of carbon deposition inside the pores of the substrate,(2) surface crusting and pore plugging before the desired density is reached, and (3)undesirable microstructure of the material Furthermore, in previous studies, the progress of the densification was not readily measurable, except by direct weighing of the parts in small-scale laboratory reactors. The lack of in-situ monitoring may result in non-optimal time and other process conditions In this review, the present status of vapor-phase routes to the rapid densification of high-temp ature composite materials is described. Approaches are reviewed for reducing the processing time by a factor of up to 1000 and the number of densification cycles to one, while producing materials with the desired properties. Techniques are described such as inductively-heated thermal-gradient isobaric CVI, radiantly- heated isothermal and thermal-gradient forced-fiow CVi, liquid-immersion thermal dient CVI, and plasma-enhanced CVI. Different heating methods, for example, radiative and inductive, and both hot-wall and cold-wall reactors are compared. Emphasis is placed on those tech iques which have demonstrated rapid densification of functional components or that show potential for the same. The different types of densification reactors are described and available material properties of composites produced by these techniques are given. This review focuses primarily on experimental results and the reader is directed to listed references for modeling and simulation studies. First, we provide a brief overview of the techniques of CVD and CVI 2. Introduction to chemical vapor deposition <6, The basic concept of forming a thin solid coating on a substrate by chemical vapor deposition P] is illustrated in Fig. 2 [8, 7, 9-11]. A fat Si substrate is placed on a resistance heater in a vacuum chamber and heated by means of infrared (r) radiation emanating from the heater. The walls of the chamber are water cooled and are therefore at or only slightly above room temperature, Gases, such as methylsilane, Si(CH3)H3(the precursor) and hydrogen, H2, are made to flow over and around the Si substrate. The flow rates are usually controlled by means of electronic mass flow controllers for each gas line and the pressure can be controlled independently by means of a throttle valve, which varies the flow conductance to the pumps. If the surface of the substrate is above the decomposition temperature of the precursor gas, a solid coating will be deposited on the substrate surface. Under Quartz Tube Pyrometer Grid Heater Throttle valve Fig. 2. A cold-wall, plasma-enhanced chemical vapor deposition reactor [7]
I. Golecki / Rapid vapor-phase densijcation of refractory composites 39 processing time. However, increasing the precursor pressure and/or temperature beyond certain ranges may produce deleterious effects, such as: ( 1) homogeneous nucleation of powders (soot) in the gas phase instead of carbon deposition inside the pores of the substrate, (2) surface crusting and pore plugging before the desired density is reached, and (3) undesirable microstructure of the material. Furthermore, in previous studies, the progress of the densification was not readily measurable, except by direct weighing of the parts in small-scale laboratory reactors. The lack of in-situ monitoring may result in non-optimal time and other process conditions. In this review, the present status of vapor-phase routes to the rapid densification of high-temperature composite materials is described. Approaches are reviewed for reducing the processing time by a factor of up to 1000 and the number of densification cycles to one, while producing materials with the desired properties. Techniques are described such as inductively-heated thermal-gradient isobaric CVI, radiantly-heated isothermal and thermal-gradient forced-flow CVI, liquid-immersion thermalgradient CVI, and plasma-enhanced CVI. Different heating methods, for example, radiative and inductive, and both hot-wall and cold-wall reactors are compared. Emphasis is placed on those techniques which have demonstrated rapid densification of functional components or that show potential for the same. The different types of densification reactors are described and available material properties of composites produced by these techniques are given. This review focuses primarily on experimental results and the reader is directed to listed references for modeling and simulation studies. First, we provide a brief overview of the techniques of CVD and CVI. 2. Introduction to chemical vapor deposition The basic concept of forming a thin solid coating on a substrate by chemical vapor deposition [ 61 is illustrated in Fig. 2 [8,7,9-l 11. A flat Si substrate is placed on a resistance heater in a vacuum chamber and heated by means of infrared (IR) radiation emanating from the heater. The walls of the chamber are water cooled and are therefore at or only slightly above room temperature. Gases, such as methylsilane, Si(CH3)H3 (the precursor) and hydrogen, Hz, are made to flow over and around the Si substrate. The flow rates are usually controlled by means of electronic mass flow controllers for each gas line and the pressure can be controlled independently by means of a throttle valve, which varies the flow conductance to the pumps. If the surface of the substrate is above the decomposition temperature of the precursor gas, a solid coating will be deposited on the substrate surface. Under . Quartz Tube Fig. 2. A cold-wall, plasma-enhanced chemical vapor deposition reactor [7]
. Golecki/Rapid vapor-phase densification of refractory composites Excitation Coil Glow Discharge Mass Flow Regulator R.F. Generator Fig 3. A hot-wall, a c plasma-enhanced, carbon chemical vapor deposition and infiltration reactor [14] appropriate temperature(e. g. 750-900C), pressure and flow-rate conditions, single-crystalline, epitaxial Sic thin films are obtained on(100)-oriented, single-crystalline Si substrates [8, 7,9-11.A coating will generally also be deposited on the heater surface. This particular configuration is known as cold-wall CVD. Other means of heating a substrate in a cold-wall reactor include:(a) placing the substrate on an electrically conducting susceptor hcated by a water-cooled induction coil and which radiates IR energy onto the substrate;(b) direct Joule heating of the substrate by time-alternatin induced currents if the substrate is sufficiently electrically conducting;(c) direct Joule heating of the substrate by contacting it to a direct-current (d.c. )or an alternating-current(ac. power supply and applying a voltage across it(used, for example, in the coating of fibers [12]); and(d)direct optical (e.g. IR) heating of the substrate by illuminating it through a properly cooled window with high- intensity tungsten-halogen lamps placed outside the vacuum chamber [13]. Cold-wall CVD can be used to coat one substrate at a time or multiple substrates simultaneously. An advantage of cold-wall reactors is reduced autodoping (i.e. impurity incorporation) from heated parts other than the substrates Another common configuration is the hot-wall CVd reactor, illustrated in Fig. 3[14, 15]. Here, generally several substrates are located in a uniform temperature(isothermal) zone of a furmace. A cylindrical tube furmace or a rectangular fumace is generally used. The interior furnace walls are heated by resistance heaters and emit iR radiation into the central region of the furnace. In another arrangement, a water-cooled coil driven by an a c, power supply may be used to inductively heat a hollow cylindrical (i.e. annular) graphite or other electrically conducting susceptor, which is located inside the coil and which surrounds the center of and emits IR radiation into the furnace. Historically, hot-wall CVD has been used to coat a very large number of substrates in the same run. The most important quantity describing a Cvd process is the deposition rate of the coating, measured, for example in um h. Values of deposition rates may vary from =10-to 10 um h The most important parameter influencing the deposition rate for a given material system is the substrate temperature. Fig. 4(a)shows SiC deposition rates on Si measured in the chamber of Fig. 2 [9, 11] Under these conditions, the deposition rate increases approximately exponentially with inverse absolute temperature. Over a wider temperature range, most cvd processes exhibit a dependence of the deposition rate on temperature as shown in Fig. 4(b). The concept of the gas-phase boundary layer Fig. 5), although a simplification of the actual process, is useful in understanding the behavior of deposition rate in CVD. Simply stated, to produce a solid coating on the heated substrate:(1)the aneous precursor has to be transported from the center of the gas stream to the boundary layer, (2) the precursor then diffuses across the boundary layer to reach the surface of the substrate, and(3)the precursor decomposes on the surface of the substrate to form the solid coating. The last step involves additional sub-processes, including adsorption of precursor-derived moieties on the surface, desorption
40 I. Golecki / Rapid vapor-phase densijcation of refractory composites Furnace /Mass Flow Regnlatorll R.F. Generator 1 7 Fig. 3. A hot-wall, ax. plasma-enhanced, carbon chemical vapor deposition and infiltration reactor [ 141, appropriate temperature (e.g. 750-900 “C), pressure and flow-rate conditions, single-crystalline, epitaxial Sic thin films are obtained on ( lOO)-oriented, single-crystalline Si substrates [8,7,9-l 11. A coating will generally also be deposited on the heater surface. This particular configuration is known as cold-wall CVD. Other means of heating a substrate in a cold-wail reactor include: (a) placing the substrate on an electrically conducting susceptor heated by a water-cooled induction coil and which radiates lR energy onto the substrate; (b) direct Joule heating of the substrate by time-alternating induced currents if the substrate is sufficiently electrically conducting; (c) direct Joule heating of the substrate by contacting it to a direct-current (dc.) or an alternating-current (a.c.) power supply and applying a voltage across it (used, for example, in the coating of fibers [ 121) ; and (d) direct optical (e.g. JR) heating of the substrate by illuminating it through a properly cooled window with highintensity tungsten-halogen lamps placed outside the vacuum chamber ] 131. Cold-wall CVD can be used to coat one substrate at a time or multiple substrates simultaneously. An advantage of cold-wall reactors is reduced autodoping (i.e. impurity incorporation) from heated parts other than the substrates. Another common configuration is the hot-wall CVD reactor, illustrated in Fig. 3 [ 14,151. Here, generally several substrates are located in a uniform temperature (isothermal) zone of a furnace. A cylindrical tube furnace or a rectangular furnace is generally used. The interior furnace walls are heated by resistance heaters and emit IR radiation into the central region of the furnace. In another arrangement, a water-cooled coil driven by an a.c. power supply may be used to inductively heat a hollow cylindrical (i.e. annular) graphite or other electrically conducting susceptor, which is located inside the coil and which surrounds the center of and emits JR radiation into the furnace. Historically, hot-wall CVD has been used to coat a very large number of substrates in the same run. The most important quantity describing a CVD process is the deposition rate of the coating, measured, for example in pm h- ‘. Values of deposition rates may vary from = 10m2 to lo3 p,m h-‘. The most important parameter influencing the deposition rate for a given material system is the substrate temperature. Fig. 4(a) shows Sic deposition rates on Si measured in the chamber of Fig. 2 [9,11]. Under these conditions, the deposition rate increases approximately exponentially with inverse absolute temperature. Over a wider temperature range, most CVD processes exhibit a dependence of the deposition rate on temperature as shown in Fig. 4(b). The concept of the gas-phase boundary layer (Fig. 5)) although a simplification of the actual process, is useful in understanding the behavior of deposition rate in CVD. Simply stated, to produce a solid coating on the heated substrate: (1) the gaseous precursor has to be transported from the center of the gas stream to the boundary layer, (2) the precursor then diffuses across the boundary layer to reach the surface of the substrate, and (3) the precursor decomposes on the surface of the substrate to form the solid coating. The last step involves additional sub-processes, including adsorption of precursor-derived moieties on the surface, desorption
. Golecki/ Rapid vapor-phase densification of refractory composites T(C) 900 800 700 Pn-=100W 0.g 1000/T(1/K) Gas-phase i Gas-phase Surface nucleation imass transport chemical kinetics 1/T(K Fig. 4.(a)Dependence of Sic deposition rate from methylsilane and hydrogen on temperature and plasma power [9].(b)Three regimes in chemical vapor deposition Axis as Flow d Boundary Layer N Fig. 5. Boundary layer fow in chemical vapor deposition of other moieties from the surface, surface diffusion and chemical reactions. The deposition rate, rof a film on a substrate can be expressed [6] by means of Eq. (1) r=CE[k。hg/(k。+hg】/N where Cg is the precursor concentration in the gas phase, ks is the rate constant for heterogeneous decomposition of the precursor into the film on the surface of the substrate, hg is the gas-phase mass transfcr cocfficicnt of precursor to the substrate and Ns is a normalizing constant. The deposition rate is proportional to the concentration of the precursor in the gas phase; the precursor may be, but often is not, the input chemical introduced into the reactor. The gas-phase mass-transfer coefficient can be expressed as hg=D/db, where D is the gas-phase diffusivity and dD is the thickness of the boundary
I. Golecki/ Rapid vapor-phase densijication of refractory composites 41 (a) T (“C) 900 800 700 t’ ’ I I I I 4 P,=looW 1.24 eV 0.8 0.9 1.0 1000/T (I/K) Gas-phase i Gas-phase f Surface nucleation i mass transports chemical kinetics (b> l/T (K-l) Fig. 4. (a) Dependence of Sic deposition rate from methylsilane and hydrogen on temperature and plasma power [9]. (b) Three regimes in chemical vapor deposition. Axis __-._.-_-.-.-.-.---_-.-~-~-. z - Gas Flow t Boundary Layer Substrate Fig. 5. Boundary layer flow in chemical vapor deposition. of other moieties from the surface, surface diffusion and chemical reactions. The deposition rate, Y of a film on a substrate can be expressed [ 61 by means of Eq. ( 1) : r=C,[k,h,l(k,+h,)]lN, (1) where C, is the precursor concentration in the gas phase, k, is the rate constant for heterogeneous decomposition of the precursor into the film on the surface of the substrate, h, is the gas-phase masstransfer coefficient of precursor to the substrate and iV, is a normalizing constant. The deposition rate is proportional to the concentration of the precursor in the gas phase; the precursor may be, but often is not, the input chemical introduced into the reactor. The gas-phase mass-transfer coefficient can be expressed as h, = D/d,, where D is the gas-phase diffusivity and &, is the thickness of the boundary
1. Golecki/Rapid vapor-phase densification of refractory composites Effect c on the thickness uniformity of a CVD coating in the surface-controlled regime Required temperature uniformity(K) Average temperature uniformity,△r/r(%) (K) △H=1ev AH=2ev △H=4eV 483 1273 layer (D and d, are further discussed below ) When ks is much smaller than hg, rak CR /Ns and the process is surface-reaction controlled. Under these conditions, the reaction rate constant ks and hence the deposition rate r usually increase rapidly with temperature according to the Arrhenius law, ks=kso exp(-AH/ KT), where AH is the activation energy for the controlling surface reaction and T is the absolute temperature in K. For example, the overall decomposition of methylsilane under the appropriate conditions can be written as Si(CH3)H3(gas)+SiC(solid )+3H2(gas) even though the details of the process may be more complex; the reaction rate constant may be that of the aforementioned first-order reaction or of a different decomposition reaction of a species derived from Si(CH3)H3. For the above overall decomposition reaction (Eq. (2))one can write [Sic]=Kp(Si(CH3)H)/P(H2), where [SiC] is the concentration of SiC, p is the respective partial pressure and Khas an Arrhenius temperature dependence. At relatively low temperatures, where this regime predominates, the chemical reaction kinetics on the surface of the substrate are the slowest, rate-controlling process and the supply of precursor from the gas phase is essentially unlimited. The pre-exponential factor kso depends on the partial pressure of the precursor, p, generally as a power law, kso ap(with an exponent, n, often equal to or less than unity), and on other deposition conditions but not on temperature. At higher temperature, when k, is much larger than hg, r=hgCg/N, and the process is gas-phase diffusion controlled. Under conditions, transport of the precursor through the gas phase to the substrate becomes the limiting factor, while the surface chemical kinetics are relatively much more rapid. In this diffusion-controlled regime, the deposition rate has a much weaker, power-law depend- ence on temperature, r=Ar, where the exponent b is of the order of 0.5-1, due to the corresponding dependence of the gas-phase diffusivity. At higher temperatures, past the diffusion-controlled regime, the deposition rate decreases with increasing temperature, due to competing reactions not described by Eq. (1): homogeneous gas-phase nucleation or powder formation and sometimes etching of the surface of the film [16, 17]. Many CVd processes are operated in the surface-reaction controlled regime. In this regime, good uniformity of the surface temperature is required to obtain a film or coating of uniform thickness, as exemplified in Table 1. The higher the activation energy of the process and the lower the deposition temperature, the tighter is the required temperature uniformity. Operating at higher temperatures results in higher deposition rates and is, in principle, advantageous from the uniformity point of view, but often is more difficult practically. Plasma-enhanced CVD, where some or all of the gases flow through an electrical discharge region, results in a lowering of the activation energy and an increase in the deposition rate [9], which are both generally desirable. In some cases it is advantageous to operate in the gas-phase diffusion controlled regime, where the deposition rate is high and its temperature dependence is weak. However, gas-phase nucleation of particulates ctching of the growing film, both of which are very undesirable, often take place to some extent simultaneously with film deposition. Generally, lower temperatures, lower pressures, increaseddilution
42 I. Golecki / Rapid vapor-phase densification of refractory composites Table 1 Effect of substrate temperature uniformity on the thickness uniformity of a CVD coating in the surface-controlled regime Desired thickness Required temperature uniformity (K) uniformity, Ar/r (S) Average temperature W 1 5 10 5 AH=leV 1.4 6.8 13.3 4.0 AH-2eV 0.7 3.4 6.6 2.0 LW=~ eV 0.3 1.7 3.3 1.0 \- 1273 1273 1273 973 layer (D and db are further discussed below), When k, is much smaller than h,, r’= k,C,IN, and the process is surface-reaction controlled. Under these conditions, the reaction rate constant k, and hence the deposition rate r usually increase rapidly with temperature according to the Arrhenius law, k, = k,, exp( - AHIkT), where AH is the activation energy for the controlling surface reaction and T is the absolute temperature in K. For example, the overall decomposition of methylsilane under the appropriate conditions can be written as Si(CH3)Hg (gas) +SiC (solid) +3H2 (gas) (2) even though the details of the process may be more complex; the reaction rate constant may be that of the aforementioned first-order reaction or of a different decomposition reaction of a species derived from Si( CHs) H3. For the above overall decomposition reaction (Eq. (2) ) one can write [Sic] = K’p( Si( CH,)H,) lp3( Hz), where [Sic] is the concentration of Sic, p is the respective partial pressure and K’ has an Arrhenius temperature dependence. At relatively low temperatures, where this regime predominates, the chemical reaction kinetics on the surface of the substrate are the slowest, rate-controlling process and the supply of precursor from the gas phase is essentially unlimited, The pre-exponential factor k,, depends on the partial pressure of the precursor, p, generally as a power law, k,, ap” (with an exponent, n, often equal to or less than unity), and on other deposition conditions, but not on temperature. At higher temperature, when k, is much larger than h,, r= h&IN, and the process is gas-phase diffusion controlled. Under these conditions, transport of the precursor through the gas phase to the substrate becomes the limiting factor, while the surface chemical kinetics are relatively much more rapid. In this diffusion-controlled regime, the deposition rate has a much weaker, power-law dependence on temperature, I= AT b, where the exponent b is of the order of 0.5-l) due to the corresponding dependence of the gas-phase diffusivity. At higher temperatures, past the diffusion-controlled regime, the deposition rate decreases with increasing temperature, due to competing reactions not described by Eq. (1) : homogeneous gas-phase nucleation or powder formation and sometimes etching of the surface of the film [ 16,171. Many CVD processes are operated in the surface-reaction controlled regime. In this regime, good uniformity of the surface temperature is required to obtain a film or coating of uniform thickness, as exemplified in Table 1. The higher the activation energy of the process and the lower the deposition temperature, the tighter is the required temperature uniformity. Operating at higher temperatures results in higher deposition rates and is, in principle, advantageous from the uniformity point of view, but often is more difficult practically. Plasma-enhanced CVD, where some or all of the gases flow through an electrical discharge region, results in a lowering of the activation energy and an increase in the deposition rate [ 91, which are both generally desirable. In some cases, it is advantageous to operate in the gas-phase diffusion controlled regime, where the deposition rate is high and its temperature dependence is weak. However, gas-phase nucleation of particulates and/or etching of the growing film, both of which are very undesirable, often take place to some extent simultaneously with film deposition. Generally, lower temperatures, lower pressures, increaseddilution
1. Galecki/ Rapid vapor-phase densification of refractory composite and higher flow rates (i.e. milder C VD conditions )minimize undesirable processes at the expense of growth rate. The choice of precursor may also influence the deposition rate and the properties of the deposited coating [18] We next briefly review some fundamental quantities of the kinetic theory of gases and of gas flow relevant to CVD and CVi[19-26]. The equation of state of an ideal gas isp V=NkTorn=N/V=p/kT, where n is the number of molecules per unit volume. Common pressure units are Pascal (Pa) Nm, Torr, atm and lb inch -(psi): I atm=760 Torr(at sea level)=1.013X10 Pa=14.70 psi 1 Torr=133.3 Pa=1.333 X10 dyne cm-2 1 Pa=10 dyne cm-2. The average molecular velocity for a Maxwell distribution is =(8kT/m).5=(8RT/M)o 5, where R=8.314 J K-1 mol-=62.364 Torr I-- is the universal gas constant, m is the weight of one molecule and M is the molecular weight; m=M/NA, where NA=6.023 X102 molecules mol" is Avogadro's number. The molecular mean free path is A-(kT/po)(1/y/2), where o is the molecular diameter Note that a varies inversely with pressure. At T=273 K(0 C)and with o in A and p in Torr, a (cm)=6.363x10-2/po. In a mixture containing two gases, the mean free path of molecules of gas 1 colliding with molecules of gas 2 is M1,2=4KT/[(o1+02)P2(1+M1/M2).]. When A>a,where a is a characteristic dimension of the system, the gas molecules collide mostly with the walls of the system and the flow is in the Knudsen or molecular regime. The system could be the reactor chamber or a pore inside a porous substrate. When A/ 3=(kT/Tm).(2/30p). Note that the diffusivity, like the mean free path, is inversely proportional to pressure; it increases with temperature according to a power law with an exponent=1.5. More involved calculations show that the diffusivity depends on T with n closer to 1.7. At T= 273 K, and with p in Torr, o in a and M in g mol-,DF(cm2s-)=5102/0pV/M. The diffusivity of a binary gas mixture is Dr1,=(1n2+A2n1)/3(n,+n2). These expressions for diffusivity are valid in the laminar flow regime. In the molecular flow regime, the Knudsen diffusivity is Dk=a/ 3=(a/3)(8RT/M)o, where a is the characteristic dimension of the system. The Knudsen diffusivity does not depend on pressure. The overall diffusivity, Dx is calculated from 1/Dx=1/DF+1/Dk.The viscosity of an ideal gas is n=0.499nm A-(0.998/0)(MKT/NA)0. Note that the viscosity does not depend on pressure and that it increases with temperature as r9, where q=0.5; more involved calculations show that q=0.7. By contrast, the viscosity of liquids decreases with temperature. At T=273 K and with o in a and M in g mol-l, n(upoise)=448.3 /M/0; 1 poise =1 g cm-ls-1 Viscosity in this sense has meaning in the laminar flow regime. Properties of several common gases re given in Table 2. The following basic quantities relating to flow of gases are relevant for both CVD and cVI. Gas flow through a tube of diameter a is classified as being laminar if Re2100, the flow is classified as turbulent. The physical significance of the Reynolds number is the ratio between inertia and viscosity in the gas. The above fow criterion can also be expressed by means of the gas Basic gas-phase quantities for selected gases at 273K[19] Gas Molecular diameter Mean free N2 67 88 88 4.18
I. Golecki / Rapid vapor-phase densij?cation of refractory composites 43 and higher flow rates (i.e. milder CVD conditions) minimize undesirable processes at the expense of growth rate. The choice of precursor may also influence the deposition rate and the properties of the deposited coating [ 181. We next briefly review some fundamental quantities of the kinetic theory of gases and of gas flow relevant to CVD and CVI [ 19-261. The equation of state of an ideal gas ispV=NkTor II = NIV=plkT, where it is the number of molecules per unit volume. Common pressure units are Pascal (Pa) = N m-‘, Torr, atm and lb inchm2 (psi) : 1 atm = 760 Torr (at sea level) = 1.013 X lo5 Pa = 14.70 psi; 1 Torr= 133.3 Pa= 1.333 X lo3 dyne cmm2; 1 Pa= 10 dyne cme2. The average molecular velocity for a Maxwell distribution is =(~~T/~z)~.~=(~RT/~TM)~.~, where R=8.314 J K-i mol - ’ = 62.364 Torr 1 K-i mol- ’ is the universal gas constant, m is the weight of one molecule and M is the molecular weight; rn=M/N*, where NA= 6.023 X 1O23 molecules mol-’ is Avogadro’s number. The molecular mean free path is h = (kT/pc?) ( 1/7rd2), where (T is the molecular diameter, Note that h varies inversely with pressure. At T= 273 K (0 “C) and with g in A and p in Torr, A (cm) = 6.363 X 10v2/pd. In a mixture containing two gases, the mean free path of molecules of gas 1 colliding with molecules of gas 2 is h1,2 = 4kT/[ r( ul + a2) 2p2( 1 + M1 /M2) OS 1. When /\. > a, where a is a characteristic dimension of the system, the gas molecules collide mostly with the walls of the system and the flow is in the -Knudsen or molecular regime. The system could be the reactor chamber or a pore inside a porous substrate. When h -=z a, the molecules collide mostly with each other, and the flow is viscous, hydrodynamic or laminar. For intermediate values of A, the flow is said to be in the transition regime. The Fickian diffusivity of a single-component, ideal gas is Dr= h / 3 = ( k3T3/ ?m) o.5( 2/32p). Note that the diffusivity, like the mean free path, is inversely proportional to pressure; it increases with temperature according to a power law with an exponent = 1.5. More involved calculationseshow that the diffusivity depends on T” with n closer to 1.7. At T= 273 K, and withpinTorr, ainAandMingmol-‘, DF (cm2 s-‘) =5102/2pJM. The diffusivity of a binary gas mixture is D F1,2 = (hi n2 + h2 nl) /3 ( nl + n2). These expressions for diffusivity are valid in the laminar flow regime. In the molecular flow regime, the Knudsen diffusivity is DK = a / 3 = (a/3) (8RT17rM)“.5 , where a is the characteristic dimension of the system. The Knudsen diffusivity does not depend on pressure. The overall diffusivity, Ds is calculated from 1 /D,: = 1 lD, + 1 /D,. The viscosity of an ideal gas is 7~ = 0.499nm h = ( 0.998/n’.5d) (MkTlNA)o.5. Note that the viscosity does not depend on pressure and that it increases with temperature as T 4, where 4 = 0.5; more involved calculations show that q-0.7. By contrast, the viscosity of liquids decreases with temperature. At T= 273 K and with u in A and M in g mol- ‘, 77 (p,poise) =448,3JMlc?; 1 poise= 1 g cm-’ s-l, Viscosity in this sense has meaning in the laminar flow regime. Properties of several common gases are given in Table 2. The following basic quantities relating to flow of gases are relevant for both CVD and CVI. Gas flow through a tube of diameter a is classified as being laminar if Re 2100, the flow is classified as turbulent. The physical significance of the Reynolds number is the ratio between inertia and viscosity in the gas. The above flow criterion can also be expressed by means of the gas Table 2 Basic gas-phase quantities for selected gases at 273 K [ 191 Gas MFlecular diameter Mean free path at (4 1 Torr (km) N2 3.78 44 H2 2.68 88 CHa 4.18 36 Fick diffusivity at 1 Torr (cm’ s-l) 67 502 73 Viscosity (i.wise) 166 88 103
l. Golecki/ Rapid vapor-phase densification of refractory composites throughput or mass Hlow-rate @=pdv/dt=pvav ra/4 with Re=(4M/RT m)(e/a)or 2=1.337 naRe/M. For N2 at 273 K, Q=7.928x10-2aRe, where Q is in Torr ls-I and a is in cm; if g is expressed in standard cm'per minute(sccm), where 1 Torr I s=78.95 sccm,then 0=6.26aRe Laminar flow of gas through a long circular tube has a parabolic velocity profile [20 ,where the velocity, U(r), is at maximum on the axis of the tube and decreases to zero with increasing radial distance towards the wall: u(r)=(a2-4)(dp/dz )/167. The maximum gas velocity is Imax=16ma(dp/dz). The throughput Q for a tube of length L is equal to(a /256nL)(P12-P22) (ma/128mL)(P1-P2), where P, and P2 are the pressures at the nlet and outlet and is he average pressure; this is Poiseuille's law. The flow conductance of the tube, C-g/ (P1-P2)=(a"/128nL), is proportional to the fourth power of the diameter and to the average pressure in the tube. The temperature dependence enters only through the viscosity. This equation is valid when L>a. For N2 at 273 K, C(Is")=197.la/L, with p in Torr, a and L in cm.A more accurate treatment, taking into account the fraction of molecules which are specularly reflected when striking the wall, is given in Ref. [20]: those molecules have a non-zero velocity vector in the flow direction. The mathematical effect is to add a pressure-independent term to the flow conductance. this term becoming important at relatively low pressures In the molecular flow regime, the conductance of a long, cylindrical tube does not depend on pressure and varies with the diameter cubed, c=(a/)(RT18M)0.In the molecular regime, the conductance increases with temperature, whereas in the laminar regime it decreases with temperature With C in 1s".M in g mol. and a and L in cm. C=3.811(a/L)(T/M). 5: for N2 at 273 K, C( s-)=11.90a/L A combined equation, allowing calculation of the conductance through the different fow regimes, is given in Ref. [20]. A relatively broad minimum in the conductance occurs when a=5.47n(/M)o.with a in Torr cm. For N2 at 273 K, a=4.01X10-3TorT cm,or A=1.097a, so that the minimum occurs when the flow is in the transition regime The thickness of the boundary layer, db, in the laminar flow regime in CVD( Fig. 5)[6] is given by db-5(nL/pvmax)o., where L is the axial distance from the inlet. Using the Reynolds number,the thickness of the boundary layer is d,=5L//Re. The flow velocity at the surface of the substrate is zero and increases to its steady-state value across the thickness of the boundary layer a detailed description of the series of chemical reactions leading from a gaseous precursor to a solid film on a substrate is generally quite complex. The conditions during CVD are usually farremoved from thermodynamic equilibrium. Only a few systems have been studied in relative depth, mostly those leading to deposition of films which are important in semiconductor manufacturing, such as Si and GaAs [16, 17]. In addition, an accurate description of the CVd mechanism requires a numerical solution of the appropriate flow equations for the particular reactor configuration. Although significant progress has been made in the combined solutions of the chemical and flow equations, quantitative modeling of CVD processes in general cases is rather limited. Porting of a process from one reactor to another reactor of different geometry and/or size is generally not trivial. Therefore reliable data must be acquired experimentally. Flow visualization techniques have been used to model flows in CVD reactors. Gas-phase chemistry measurements can be performed in situ, using mass spectrometry or gas chromatography with differential pumping, and a variety of passive and active optical measure- ments can be employed, which are particularly useful when a plasma discharge is used. In general, specially constructed and dedicated reactors are needed to obtain accurate in-situ measurements of the gas phase close to the substrate and of the substrate and growing film. Measurement of deposition rates is usually carried out on fat substrates by measuring the thickness in situ, during deposition, or ex situ, after the deposition has been completed. Several in-situ measurement techniques exist for optically transparent or semi-transparent films, including optical transmissivity, reflectivity and ellipsometry
44 I. GoEecki/Rapid vapor-phase densi$cation of refracro? composites throughput or mass flow-rate Q =pdV/dt=pv,,ra2/4 with Re= (4hi/RT T) (Q/a) or Q = 1.337 X 104qaRelM. For Nz at 273 K, Q = 7,928 X 10e2aRe, where Q is in Torr 1 s- ’ and a is in cm; if Q is expressed in standard cm3 per minute (seem), where 1 Torr 1 s- ’ = 78.95 seem, then Q = 626aRe. Laminar flow of gas through a long circular tube has a parabolic velocity profile [20], where the velocity, v(r) , is at maximum on the axis of the tube and decreases to zero with increasing radial distance towards the wall: U(T) = ( a2 - 412) (dpldz) / 1677. The maximum gas velocity is u,,= 16qa2(dpldz). The throughput Q for a tube of length L is equal to (n-a”/256TL) (~~‘--p~~) = ( lra4/128qL) (pl -p2), wherep, andp, are the pressures at the inlet and outlet and is the average pressure; this is Poiseuille’s law. The flow conductance of the tube, C=Q/ (pl -p2) = ( Ta4/ 128qL) , is proportional to the fourth power of the diameter and to the average pressure in the tube. The temperature dependence enters only through the viscosity. This equation is valid when LB a. For N2 at 273 K, C (1 s-‘) = 197.1a4 /L, with p in Ton; a and L in cm. A more accurate treatment, taking into account the fraction of molecules which are specularly reflected when striking the wall, is given in Ref. [20] ; those molecules have a non-zero velocity vector in the flow direction. The mathematical effect is to add a pressure-independent term to the flow conductance, this term becoming important at relatively low pressures. In the molecular flow regime, the conductance of a long, cylindrical tube does not depend on pressure and varies with the diameter cubed, C = ( a3 / L) ( n-RT/ 18M) o.5. In the molecular regime, the conductance increases with temperature, whereas in the laminar regime it decreases with temperature. WithCinls-‘,Mingmol-‘, andaandLincm,C=3.811(a3/L)(T/M)0~5;forN2at273K,C(l s- ‘) = 1 1.90a3/L. A combined equation, allowing calculation of the conductance through the different flow regimes, is given in Ref. [20]. A relatively broad minimum in the conductance occurs when a=5.47q(T/1IY)~.~ with a in Torr cm. For N2 at 273 K, a=4,01 X 10m3 Torr cm, or h = l.O97a, so that the minimum occurs when the flow is in the transition regime. The thickness of the boundary layer, d,,, in the laminar flow regime in CVD (Fig. 5) [ 6 ] is given by db = 5( qL/pumax) o.5, where L is the axial distance from the inlet. Using the Reynolds number, the thickness of the boundary layer is db = 5LIJRe. The flow velocity at the surface of the substrate is zero and increases to its steady-state value across the thickness of the boundary layer. A detailed description of the series of chemical reactions leading from a gaseous precursor to a solid film on a substrate is generally quite complex. The conditions during CVD are usually far removed from thermodynamic equilibrium. Only a few systems have been studied in relative depth, mostly those leading to deposition of films which are important in semiconductor manufacturing, such as Si and GaAs [ 16,171. In addition, an accurate description of the CVD mechanism requires a numerical solution of the appropriate flow equations for the particular reactor configuration. Although significant progress has been made in the combined solutions of the chemical and flow equations, quantitative modeling of CVD processes in general cases is rather limited. Porting of a process from one reactor to another reactor of different geometry and/or size is generally not trivial. Therefore, reliable data must be acquired experimentally. Flow visualization techniques have been used to model flows in CVD reactors. Gas-phase chemistry measurements can be performed in situ, using mass spectrometry or gas chromatography with differential pumping, and a variety of passive and active optical measurements can be employed, which are particularly useful when a plasma discharge is used. In general, specially constructed and dedicated reactors are needed to obtain accurate in-situ measurements of the gas phase close to the substrate and of the substrate and growing film. Measurement of deposition rates is usually carried out on flat substrates by measuring the thickness in situ, during deposition, or ex situ, after the deposition has been completed. Several in-situ measurement techniques exist for optically transparent or semi-transparent films, including optical transmissivity, reflectivity and ellipsometry
1. Golecki/Rapid vapor-phase densification of refractory composites In addition, a large array of other experimental techniques is used to measure film thicknesses ex situ, for example stylus profilometry, scanning electron microscopy and Rutherford backscattering(if the volumetric density is known). Knowledge of the substrate temperature is clearly of paramount impor- tance in CVD. In a hot-wall reactor, the temperature of the substrate is close to the average temperature of the central hot zone, which can be measured accurately by means of thermocouples placed inside or firmly attached to test specimens. In a cold-wall reactor, however, accurate measurement of substrate temperatures during CVD is more difficult, since generally direct contact to the substrate is not possible and large thermal gradients exist between the substrate and the surrounding components in the reactor Pyrometry can be used if the emissivity can be measured independently [8, 7, 9-11] and if it does not change appreciably during film growth. Indirect optical techniques using absorptivity and pyrometric interferometry [27, 28] have also been used in specific instances. CVD reactors employing"intelli- gent, in-situ control of the growth rate are in relatively early stages of development The main applications of CVD are in the fabrication of microelectronic circuits. The typical substrates are flat semiconductor wafers, although often the morphology consists of trenches with an aspect ratio of 10: 1 or higher. CVD excels in providing conformal coating (i. e a coating of uniform thickness which intimately follows the shape and morphology of the substrate) of such complex structures, since the deposition rate depends primarily on temperature. Therein lies the principal advantage of CVD with respect to directed coating methods, such as sputtering, physical evaporation and molecular beam epitaxy(MBE). In the latter techniques, which(except for MBE) generally do not employ substrate heating, geometric shadowing effects dominate and uniform coverage of complex surface shapes becomes much more difficult. Other applications of CVD include optical coatings and mechanical protective coatings. There are a variety of reactor designs for specific applications of CVD [16,17,29 3. Introduction to chemical vapor infiltration The two principal differences between chemical vapor infiltration(CVI) and chemical vapor deposition(CVD) are that in CVi (1) the surfaces being coated consist of complex-shaped pores located in the interior of a porous solid( the preform), and(2) the absolute quantity of materia deposited is much larger than in CVD. The purpose of CVd is to deposit a functional thin coating on a substrate with the same or a different composition; e. g. the coating may serve as the active part of an electronic device, as an intermediate diffusion barrier(allowing flow of electrical carriers but not phys g layer. Generally, the coating pro by Cvd adds less than 1% of weight to the substrate and the deposition time is of the order of a few minutes to a few hours. The primary purpose of CVi, on the other hand, is to increase the density of porous body consisting of an array of fibers( with or without additional solids) by 100 to 900%,i.e by a factor of 2 to 10, in order to obtain a material with certain desirable properties. The material deposited by CVi forms part or all of the matrix of the fiber-matrix composite and thus may be a significant fraction of the total mass of the final composite. The infiltration time in conventional CVI methods is therefore much longer than the deposition time in CVD In CVI, in order to deposit a solid coating on the surface of a pore located deep inside a porous preform, an additional gas-phase diffusion step needs to occur after the second step in CVd on a solid (non-porous) substrate, described in Section 2, namely diffusion from the surface of the preform into the interior pore. Such diffusion may be driven by a concentration gradient in isobaric CVi or by a pressure gradient in forced-flow CVI Porous preforms generally have a complex pore size distribution, which may consist of several median size ranges, depending on the architecture of the preform. For
I. Golecki/ Rapid vapor-phase densification of refractory composites 45 In addition, a large array of other experimental techniques is used to measure film thicknesses ex situ, for example stylus profilometry, scanning electron microscopy and Rutherford backscattering (if the volumetric density is known). Knowledge of the substrate temperature is clearly of paramount importance in CVD. In a hot-wall reactor, the temperature of the substrate is close to the average temperature of the central hot zone, which can be measured accurately by means of thermocouples placed inside or firmly attached to test specimens. In acold-wall reactor, however, accurate measurement of substrate temperatures during CVD is more difficult, since generally direct contact to the substrate is not possible and large thermal gradients exist between the substrate and the surrounding components in the reactor. Pyrometry can be used if the emissivity can be measured independently [8,7,9-l 1 ] and if it does not change appreciably during film growth. Indirect optical techniques using absorptivity and pyrometric interferometry [ 27,281 have also been used in specific instances. CVD reactors employing “intelligent” in-situ control of the growth rate are in relatively early stages of development. The main applications of CVD are in the fabrication of microelectronic circuits. The typical substrates are flat semiconductor wafers, although often the morphology consists of trenches with an aspect ratio of 10: 1 or higher. CVD excels in providing conformal coating (i.e. a coating of uniform thickness which intimately follows the shape and morphology of the substrate) of such complex structures, since the deposition rate depends primarily on temperature. Therein lies the principal advantage of CVD with respect to directed coating methods, such as sputtering, physical evaporation and molecular beam epitaxy (MBE). In the latter techniques, which (except for WE) generally do not employ substrate heating, geometric shadowing effects dominate and uniform coverage of complex surface shapes becomes much more difficult. Other applications of CVD include optical coatings and mechanical protective coatings. There are a variety of reactor designs for specific applications of CVD [ 16,17,29]. 3. Introduction to chemical vapor infiltration The two principal differences between chemical vapor infiltration (CVI) and chemical vapor deposition (CVD) are that in CVI ( 1) the surfaces being coated consist of complex-shaped pores located in the interior of a porous solid (the preform), and (2) the absolute quantity of material deposited is much larger than in CVD. The purpose of CVD is to deposit a functional thin coating on a substrate with the same or a different composition; e.g. the coating may serve as the active part of an electronic device, as an intermediate diffusion barrier (allowing flow of electrical carriers but not physical intermixing of atoms) or as a protective or passivating layer. Generally, the coating produced by CVD adds less than 1% of weight to the substrate and the deposition time is of the order of a few minutes to a few hours. The primary purpose of CVI, on the other hand, is to increase the density of a porous body consisting of an array of fibers (with or without additional solids) by 100 to 900%, i.e. by a factor of 2 to 10, in order to obtain a material with certain desirable properties. The material deposited by CVI forms part or all of the matrix of the fiber-matrix composite and thus may be a significant fraction of the total mass of the final composite. The infiltration time in conventional CVI methods is therefore much longer than the deposition time in CVD. In CVI, in order to deposit a solid coating on the surface of a pore located deep inside a porous preform, an additional gas-phase diffusion step needs to occur after the second step in CVD on a solid (non-porous) substrate, described in Section 2, namely diffusion from the surface of the preform into the interior pore. Such diffusion may be driven by a concentration gradient in isobaric CVI or by a pressure gradient in forced-flow CVI. Porous preforms generally have a complex pore size distribution, which may consist of several median size ranges, depending on the architecture of the preform. For
1. Golecki/ Rapid vapor-phase densification of refractory composites Width Fig. 6. Plain-woven(left)and five-hamess satin-woven(right)layer architectures of continuous fibrous preforms, showing the warp and weft tows example, in continuous fiber-reinforced composites, the fibers, which may have a diameter of 5-15 [301(e arranged in bundles(also called tows or yarns), with 500-3000 or more fibers per bundle μum,ar (see Fig. 6). The typical pore sizes between individual fibers are the smallest, of the order of 1-10 pm. Pores between fiber bundles are much larger, typically 50-500 um and pores between layers of cloth are similar or larger in size. Because flow conductances vary as the diameter to the third or fourth power, the partial pressure of a precursor inside a small pore in a preform may be different than its value in the reaction chamber. Further, in CVi, the characteristic dimensions(s) for the reactor (1-500 cm) and for the interior of the preform(1-500 um) differ by many orders of magnitude. For example, for a mean free path of 34 um( methane molecules at 1000C and 5 Torr), and assuming e sane pressure in the pore as in the reactor, the flow would be laminar outside the preform(A<su) and laminar, mixed or molecular inside a pore in the preform. Clearly, both the gas-phase diffusion mechanism and the Cvd deposition mechanism may also be different in the reactor and inside a pore Thus, while on the surface of the preform, the Cvd process may be in the surface-reaction controlled regime, inside a pore located in the interior of the preform, there may be significant depletion of the precursor due to slow gas-phase diffusion in the molecular regime and the process may be gas-phase diffusion controlled. This state of affairs may lead to non-uniform densification observed in isothermal isobaric CVI of thick preforms, where the outer surface of the preform has a higher density than the Interior regions. a helpful concept in CVi is the Thiele modulus [311, 0=(k L/Da)., which was originally developed in the field of catalysis. Consider a straight, cylindrical pore of initial diameter a and length L, as depicted in Fig. 7, exposed to a gaseous mixture undergoing decomposition at a uniform temperature T. As solids deposit on the inside walls of the pore, the pore diameter, a, will decrease unction of time, t and of the axial position, z. The Thiele modulus gives the relative importance of the chemical reaction rate vS the gas-phase diffusion in this process; ks is the first-order reaction rate Fig. 7. One-dimensional pore geometry used in explaining the Thiele modulus
46 I. Golecki / Rapid vapor-phase densijication of refractory composites eft, !11 Width Fig. 6. Plain-woven (left) and five-harness satin-woven (right) layer architectures of continuous fibrous preforms, showing the warp and weft tows. example, in continuous fiber-reinforced composites, the fibers, which may have a diameter of 5-1.5 km, are arranged in bundles (also called tows or yarns), with 500-3000 or more fibers per bundle [ 301 (see Fig. 6). The typical pore sizes between individual fibers are the smallest, of the order of l-10 pm. Pores between fiber bundles are much larger, typically 50-500 p,rn and pores between layers of cloth are similar or larger in size. Because flow conductances vary as the diameter to the third or fourth power, the partial pressure of a precursor inside a small pore in a preform may be different than its value in the reaction chamber. Further, in CVI, the characteristic dimensions(s) for the reactor (l-500 cm) and for the interior of the preform ( l-500 tJ,m) differ by many orders of magnitude. For example, for a mean free path of 34 pm (methane molecules at 1000 “C and 5 Torr), and assuming the same pressure in the pore as in the reactor, the flow would be laminar outside the preform (A -=x LZ) and laminar, mixed or molecular inside a pore in the preform. Clearly, both the gas-phase diffusion mechanism and the CVD deposition mechanism may also be different in the reactor and inside a pore. Thus, while on the surface of the preform, the CVD process may be in the surface-reaction controlled regime, inside a pore located in the interior of the preform, there may be significant depletion of the precursor due to slow gas-phase diffusion in the molecular regime and the process may be gas-phase diffusion controlled. This state of affairs may lead to non-uniform densification observed in isothermal, isobaric CVI of thick preforms, where the outer surface of the preform has a higher density than the interior regions. A helpful concept in CVI is the Thiele modulus [ 3 1 ] , 0 = ( kSL2/Da) ‘.‘, which was originally developed in the field of catalysis. Consider a straight, cylindrical pore of initial diameter a and length L, as depicted in Fig. 7, exposed to a gaseous mixture undergoing decomposition at a uniform temperature T. As solids deposit on the inside walls of the pore, the pore diameter, a, will decrease as a function of time, t and of the axial position, z. The Thiele modulus gives the relative importance of the chemical reaction rate vs. the gas-phase diffusion in this process; k, is the first-order reaction rate a (diameter) Fig. 7. One-dimensional pore geometry used in explaining the Thiele modulus