This paper intends to develop and illustrate by a few examples the principles of thermodynamic analysis which can be applied at an early stage to assist in selecting the following:

• the material to be deposited for a well-defined application,

• the nature of reactants

• the range of experimental parameters.

Specific emphasis will be given to the main problems relating to data selection: assessments, availability of coherent data, sensitivity of results to data accuracy, etc.

Finally, the validity of a thermodynamic approach at equilibrium will be discussed with a view to optimizing a dynamic non-equilibrium process.

Tailoring the chemical and physical nature of CVD deposits for specific applications requires understanding and controlling the chemical processes occurring in a deposition reactor. This implies understanding the combined equilibrium, kinetic, and transport processes. The occurring processes produce composition gradients in the reactor which can cause variations in the deposition chemistry. Derived physical and mathematical models of possible mechanisms provide critical tools for manipulating and controlling the chemical and physical nature of CVD deposits. Examples of silicon boride deposition from SiH4, BC13, and H2 mixtures are used to illustrate thermochemical modeling techniques which utilize the concept of partial equilibrium. Predictions of limits on the chemical behavior, and comparison of these limits with experimental observations are used to hypotheses deposition mechanisms.

Numerical modeling of gas phase chemical kinetics was used to investigate reactions following silane decomposition to suggest experimental conditions and to interpret concentrations measurements. The effects of concentration and temperature gradients on kinetic pathways and reactive intermediates were studied. These processes are relevant to silicon-based chemical vapor deposition (CVD), to flame-driven gas phase particle nucleation, and to laser-induced processes for materials fabrication.

A unified approach to model the general CVD process, utilizing the versatility and flexibility of bond graph methods, is presented. The bond graph representations for important CVD sub-processes are derived. The application of these elements is illustrated through hypothetical examples of CVD processes. It is also shown that several different CVD processes can be handled by a single bond graph computer program.

A non-iterative method is presented for constructing CVD phase diagrams which include the systematic effects of chemical element segregation in mass-transfer controlled flow reactors. Element segregation is shown to substantially shift predicted deposit phase boundaries when the vapor/deposit interface equilibrium is calculated using the local element fractions instead of the feed gas elemental composition, as is usually done. The mass transfer analysis developed here accounts for both Fick and Soret multicomponent diffusion acting across a non-isothermal boundary layer. The gas phase is assumed to be chemically frozen, with local thermochemical equilibrium imposed only at the vapor/deposit interface (as in a cold wall reactor with a hot substrate). As a specific example, this model is applied to the chemical vapor deposition of titanium diboride from TiCl4(g), BC13(g) and H2(g) in an Ar(g) carrier gas for ceramic composite material applications. These calculations, which account for boron and titanium transport via 17 chemical species, are illustrated for a long cylindrical reactor with a resistively heated coaxial fiber deposition substrate and coaxial annular flow. However, the method presented here is general in that both the chemical system and CVD reactor geometry can be changed to any other system of interest, provided: i) adequate themochemical and thermophysical data are available, ii) the deposition rate is vapor transport controlled, and iii) convective-diffusion heat and mass transfer coefficients are estimable.

A simple model, based on physical microstructure, that predicts the anisotropic transport properties of cloth lay-up composites is discussed. This anisotropic material is modeled as a layer structure consisting of two, or more, isotropic components. Composite transport coefficients perpendicular and parallel to the layers are given by “series” and “parallel” combining laws involving the isotropic coefficients of these component materials and their volume fractions. These component material properties are determined using microstructural information and fundamental material properties. Model calculations are compared to experimental measurements for gas permeability.

Forced-flow thermal-gradient chemical vapor infiltration (FCVI) has demonstrated excellent potential for fabrication of high strength, high toughness ceramic composites. Extension of this process to large and complex shapes is facilitated by use of a computer model to optimize process conditions and hardware for rapid, uniform infiltration.

A 3-D model has been developed using a “finite volume” formulation. A steady-state solution for heat conduction and Darcy's law permeation produces temperature and gas flow distributions within the fiber preform. These are used to generated matrix deposition rates within each volume element. By “marching” through time, a complete simulation of the densification process can be obtained.

The model is demonstrated for a FCVI system with cylindrical symmetry and compared to experimental results obtained at the Oak Ridge National Laboratory. The model results suggest a self-optimizing feature of the force flow/thermal gradient CVI process that produces uniform density in the final composite over a range of infiltration conditions. This matches experimental observation where good uniformity has been achieved over a wide range of gas flows, pressure and temperature.

A model for chemical vapor infiltration (CVI) under pressure and temperature gradients is applied to the study of deposition of SiC from the pyrolysis of CH3SiCl3 within a 3-D woven fibrous preform. The model considers the infiltration of reactants into a preform with temperature gradients by applying a pressure gradient between the vapor inlet and outlet; it also takes into account the variation in concentration of the vapor precursor. A quasi-steady state approach has been adopted to simulate the matrix deposition in a 3-D unit cell. The density distribution, consolidation profile, and total fabrication period have been theoretically predicted.

A model is presented for CVI of a fibrous preform under isothermal and thermal gradient conditions. Transport and geometric properties during densification are calculated from a novel structural model, allowing prediction of deposition profiles and experimentally observed trends without the use of empiricism.

A mathematical model is developed to describe mass transport, reaction., and structure evolution during densification of porous media of initially fibrous or cellular structure. The use of Monte Carlo simulation procedures for determining the variation of the local (average) structural properties of the porous structure and of the effective diffusion coefficients in the porous medium with the porosity is also discussed, and results are presented for Knudsen diffusion in cellular or fibrous media. The model is used to theoretically study the transient behavior of the densification process during preparation of SiC/SiC ceramic matrix composites by chemical vapor infiltration (CVI) of porous preforms. Particular emphasis is placed on the investigation of the effects of pressure pulsing on the density gradients in the densifying structure.

The interaction of borazine (N3B3H6) with a Re(0001) surface has been studied by low energy electron diffraction (LEED), thermal desorption spectroscopy (TDS), Auger electron spectroscopy (AES) and electron energy loss spectroscopy (ELS). It is shown that chemisorbed borazine dissociates into atomic N, B and H on Re(0001) at sample temperatures > 450K. However, borazine is found to decompose to a boron-nitrogen compound when irradiated with a 2 KeY e–beam.

The early growth of chemically vapor deposited TiN and TiC coatings on pyrolytic graphite was studied in the kinetic- and mass transport-controlled regimes. While steady-state growth of these coatings results in columnar grains, such morphologies do not originate at the substrate/coating interface. Rather, TiC deposition begins on the substrate as fine grains less than 100 nm in diameter. Early TiN growth occurs in layers of 50 nm grains. In both cases, early fine-grained growth occurs at a lower rate than the linear, steady rate observed for columnar growth. A laser scattering technique has been developed as a tool for characterizing early growth through surface roughness. This noncontact method can be used as an in-situ diagnostic to detect changes in the surface of the growing deposit.

Surface morphologies created by the chemical vapor deposition (CVD) of silicon carbide were examined with light scattering. Silicon carbide was deposited from methyltrichlorosilane under various conditions to create different surfaces. A helium neon laser was used, and the scattered light was measured over a range of scattering angles. These measurements are compared with scanning electron microscopy (SEM) observations and profilometer measurements of the same surfaces. In theory the scattered light contains all of the information needed to provide a statistical description of a given surface; however, a complete vector theory for the scattering phenomena is too complex to provide any simple basis for experimental analysis. The application and limits of existing descriptions of scattering from a rough surface are discussed.

The formation of LPCVD tungsten by means of the reduction of WF6 with Si, H2 and SiH4 is monitored in situ using a wavelength adjustable reflectometer. The initial self stopping growth of W by Si reduction is strongly dependant on surface status [1]. SEM observations together with Auger depth profiling and weight measurements support a growth model of islands that grow laterally and vertically until islands touch. After the self stopping Si reduction the W layer was increased in thickness by either the h or SiH4 reduction. The surface roughness calculated from the reflectance appears to increase linearly with thickness in the case of H2 reduction. Typical rms roughness was found to be 7% of layer thickness in the H2 reduction case. The reflectance of H2 reduced W layers could be improved by interrupting the growth process with a renucleation step using SiH4. Selective deposition and in situ growth rate measurements can be monitored when the deposition is carried out on a grating of SiO2 as a mask. Precleaning of the reactor with an NF3 plasma results in a strong retardation of the H2 reduction reaction.

A technique has been developed for studying the early stages of heteroepitactic nucleation and growth of oxide thin films by TEM using direct deposition onto electron-transparent ceramic substrates. Observations of α-Fe2O3 island growth on four orientations of α-Al2O3 – (0001), {1102}, {1010}, and {1120} – are reported. Moir6 fringes and selected-area diffraction are used to show heteroepitactic growth. Island morphology is compared to known growth habits of hematite crystals. Preferential nucleation of islands at low-energy sites (e.g. surface steps) is also demonstrated.

A novel technique for probing chemical vapor deposition reaction mechanisms is presented. A conventional hot-wall Pyrex reactor is coupled to a molecular beam apparatus. Preliminary results of the decomposition of an organometallic precursor to AIN, [AI(CH3)2NH2]3, indicate a decomposition temperature between 200 and 270°C. The mass spectrum of the precursor at 100°C provides evidence for the existence of a trimer-dimer equilibrium of the precursor at this temperature

The object of this study is the decomposition mechanism of Aluminium-Tri- Isopropoxide and its influence on the deposition of A12O3 by MOCVD. DSC experiments reveal that the decomposition occurs in two steps: formation of aluminum hydroxide, followed by dehydration.

Deposition of A12O3 takes place by diffusion of AI(OH)3 to the substrate surface, followed by dehydration to Al2O3. This is supported by the insignificant amount of carbon found in the coating analyzed by XPS. A Langmuir type adsorption of the reactive components explains the change in reaction order of ATI from one at atmospheric pressure to two at 3 torr.

Using a performant spontaneous Raman equipment, with intracavity laser excitation, the temperatures of tungsten hexafluoride and hydrogen have been measured in a cell reproducing the industrial conditions of C.V.D. tungsten deposition, in the vicinity of a heated graphite substrate. For tungsten hexafluoride, a vibrational temperature, much lower than that of the substrate, is determined. For hydrogen, rotational temperatures much higher than that of the substrate are found. After a discussion of the sources of experimental errors, possible interpretations of the results are given.

Laser photolysis, optical emission, and laser-induced fluorescence (LIF) were used to investigate laser driven decomposition processes in the gas phase pertaining to the systems: SiH4 → Si (s) and SiH4 → NH3 → Si3N4 (s). These processes are important to silicon/ silicon-nitride chemical vapor deposition, flame-driven gas phase silicon-particle nucleation, and laser-induced processes for materials fabrication. UV laser photolysis was used to generate SiHx and NHx species from silane and ammonia. A number of photofragments were identified by emission from excited states. The rate of reaction of NH2 with silane was measured using LIF to detect NH2 as a function of time following photolysis of ammonia

“Silicon carbide” is a generic term for a host of different materials produced by several process routes which yield a variety of microstructures and associated property characteristics. The route of chemical vapor deposition (CVD) is used primarily to deposit SiC for wear- and corrosion- resistant coatings and for diffusion barriers to and from the underlying substrate. Presently this technique is also being used to deposit monocrystalline semiconductor thin films of SiC and to infiltrate various high temperature woven fabrics with this material. The following sections describe the results of thermodynamic calculations to define SiC CVD diagrams using various precursor gas mixtures, discuss various CVD techniques and detail the results of deformation, infiltration and thin film deposition studies that have been recently conducted on vapor deposited SiC.