Thermodynamic and experimental approach to ceramic materials: gas - solid/liquid equilibria

Massimiliano Valle - ICRA S.p.A., Via Lioni 8, S.Paolo d'Argon (BG) Stefano Poli - Dipartimento di Scienze della Terra, Via Botticelli 23, 20133 Milano INTRODUCTION The knowledge of the equilibrium state under a set of environmental variables (e.g. T, oxygen fugacity) is a reference condition for a correct description of any phenomenon and process in material science. Even though local equilibria and metastability are very common in industrial processes, phase diagrams contain informations relevant for interpretation and prediction of the evolution of solids and gases both in experimental investigations and in industrial applications. It is far beyond the scope of this brief review to provide a comprehensive thermodynamic treatment for ceramic materials, and to overview the usage of thermodyamics in the interpretation of experimental data to construct reliable models for the production of ceramics. Therefore we will focus here on the definition and the consequences of equilibria between condensed matter and volatiles. Gas-solid interactions are very important in ceramic systems, because, similarly to rocks, redox states, phase stabilities, crystallization paths, ceramic textures as well as kinetic properties are directly influenced by partial pressures of volatile species. We should point out that the technology of experimental and industrial processes allow us to control some of these pressures and the definition of mobile and immobile components (open vs. close systems) is a user choice. Gas-solid reactions in ceramic systems are of increasing importance in advanced engineering application. For example, the development and the applications of nitride and carbide ceramics is strictly related to partial pressures of oxygen, nitrogen, hydrogen, CO and CO2, as these materials are much less stable than traditional oxide ceramics to routine oxidizing service conditions. We can profit of gas-solid equilibria to promote purification of raw materials, or for coating substrates, in chemical vapour deposition. These technologies are largely used in a number of industrial processes and represent an important innovation in the synthesis of ceramic materials. The advancement of gas-solid reactions is controlled both by kinetic and by thermodynamic factors. We will focus here only on the thermodynamic constraints, as these are usually sufficient to define the mechanisms and the direction of phase transformations in multicomponent systems. Thermodynamic data on many oxides, nitrides, carbides and other high-temperature phases are currently available in extensive compilations both of material science source (JANAF Tables, Metals Handbook, Thermoynamic Properties of Inorganic Substance) and of more "geological" source (e.g. Thermocalc Database). Phase equilibria are then predictable through computation (e.g. ChemSage, or Perplex) and a first order estimate of phase relationships is readily available. However, the large uncertainties in thermodynamic properties even for pure compounds, the presence of solid solutions, and unpredictable complexities often lead to unsatisfactory results. A careful experimental investigation for specific problems is therefore mandatory. Thermodynamic calculations and experimental results are commonly displayed in graphical representations. A phase diagram is a graphical representation of the values of thermodynamic variables when equilibrium is established among the phases and the Gibbs energy of the system is minimized. Geologists are used to think of phase diagrams mostly as plots of temperature versus pressure, or as binary or ternary chemographies as a function of P or T. However because in the synthetic systems investigated in material science, the composition of volatile species can be

carefully fixed and monitored, the chemical potential of volatile components is proficuously used as an explicit and independent variable. VOLATILES, PHASE DIAGRAMS AND ELLINGHAM DIAGRAMS The basic thermodynamic principle which postulates that entropy, S, in a closed heterogeneous system must be maximized at equilibrium (Callen, 1985), translates, via Legendre transformation, to the criterion of the minimization of the Gibbs free energy G: dG = -SdT + VdP + idni Because the Gibbs free energy is an extensive property, we can write the Gibbs energy change of reaction, Gr: Gr = G +RT ln K where G is the Gibbs free energy for pure compounds at pressure and temperature of interest, and K is the equilibrium constant of the reaction. K is defined as K = pcini / rcini, where p stands for product, r for reactant, c for the "concentration" of i, n i for the reaction coefficient of i. Therefore, for a volatile-bearing reaction: Cu2O + 2 SO2 + 3/2 O2 = 2 CuSO4 we can write that: ln K = -2 ln pSO2 - 3/2 ln pO2 = constant at T, P. Because in most technological applications the total pressure approaches 1 bar, the very common P-T phase diagrams used in Earth sciences, are rarely applied in material science. On the contrary, as ceramic systems are often open systems with respect to gases, the chemical potential of volatile compounds is an independent variable and can be explicitely used as a phase diagram variable. Therefore, when the partial pressures of volatile species (ln px) are used on both axis of an isothermal phase diagram (e.g. Figure 1a), phase boundaries are straight lines with a constant slope which follows the stoichiometry of the reaction. As an example plotting ln pSO2 versus ln pO2, the slope of reaction (A) will be (-3/2)/2 = -3/4. Such diagrams, commonly referred to as - in the geological literature (Kozhinskii, 1959) are called "predominance diagrams" in material science. Note that, for simplicity, numerical values are often transformed from ln to log10. When the temperature is not kept constant, but used as an explicit variable on a phase diagram axis, it is convenient to plot the quantity RTln px. Because at equilibrium G = - RT ln K, a plot of RTln px versus T is a plot of the Gibbs free energy versus T. Such "predominance", - T, diagrams are also called Ellingham diagrams in material science literature, and an example of its usage will be given in Figure 8. FROM PHASE DIAGRAMS TO VOLATILITY DIAGRAMS Volatility diagrams are useful phase diagram projections to be used when phase relationships involve at least one condensed species and more than one volatile species. Volatility diagrams are not phase diagrams in a strict sense as only the volatile species with the highest partial pressure are (A)

shown at each point on univariant lines (Lou et al. 1985, Heuer and Lou 1990). Lines on volatility diagrams represent graphical solutions to the thermodynamics of various gas - condensed phase reactions, and thus they bound "fields" where various condensed phases are stable with a "predominant" volatile. In the case of oxides, volatility diagrams are constructed with log pO2 as abscissa and log pMOx as ordinate. The reactions required to construct the volatility diagram for the Si-O system (Figure 1) at 1900 K (above the melting point and ignoring possible crystallization of SiO 2 to cristobalite) are reported in Table1. 1 2 3 4 5 6 7 8 Si(l) + O2 SiO2(l) Si(l) Si(g) SiO2(l) Si(g) + O2 2 Si(l) + O2 2 SiO(g) 2 SiO2(l) 2 SiO(g) + O2 SiO2(l) + Si(l) 2 SiO(g) Si(l) + O2 SiO2(g) SiO2(l) SiO2(g) Tab.1. Equilibria in the system Si-O. In this system the volatile species are Si, SiO, SiO2 so that we can build three diagrams with pSiOx as a variable, where x can be 0, 1, 2. Some of these reactions are independent of pSi, pSiO, pSiO2 and therefore they have a vertical slope (e.g. reaction 1), or they are independent of pO2 and have an horizontal slope (e.g. reaction 2, 8), other involve both gaseous species and they plot with a slope which depends on gaseous species molar ratio. The intersection of the three diagrams provides a direct inspection of the most representative gaseous species stable over condensed phases (lines with the highest partial pressures of volatile components). The point T indicates the maximum possible pMOx in the system, pSiO in Figure 1, which causes an active oxidation (effective loss of material from condensed phases, Figure 2), and the value of pO2 needed to prevent reduction of SiO2 to Si or oxidation of Si to SiO2 (passive oxidation, i.e. precipitation of volatile components). Combining a number of such isothermal volatility diagrams, we obtain a master diagram covering the temperature range of interest (Figures 3a-c), to be used in the case of systems both open and closed to O2. Reactions which occur in closed systems, at vacuum conditions or in the presence of inert gases, proceed following the mass-balance constraints imposed by reaction stoichiometries; partial pressures of gaseous species are therefore linearly dependent. Because of absence of free oxygen in the system, liquid silicon does not transform to SiO(g) and forms gaseous silicon at temperatures higher than 1900 K. SiO 2 evaporates via Eq.(5) at any temperature on diagram in Figure 3a. For each mole of SiO 2, 1 mol of SiO(g) and mol of O2 are produced such that pSiO = 2 pO2. This relation (dashed line on diagram) is the so called isomolar line which is the expression of the mass balance criterion. It also defines the maximum pSiO over SiO2(l) at equilibrium (point of intersection of the isomolar with a maximum equilibrium pressure lines). Only conditions to the right of this line are realistically attainable in open systems. It is well know that reducing gases can cause severe weight loss of metals at high temperatures; volatility diagrams can predict these behaviours, after considering vapour pressure of various gaseous species involved in high temperature reactions. For example, in the case of hydrogen in the system Si-O (Figure 3c), pSiO has to maintain a molar criterion with H2O because one of the reactions working in the system is now: log K (1900 K) 15,68 -4,92 20,60 13,86 -17,50 -1,82 8,29 7,30

Si(l) + H2O SiO(g) + H2 and pSiO must be equal to pH2O. So, if we process the system at 1900 K, pH2=1bar with H2O/H2 =10-3, data in the JANAF tables predict a pO2= 10-7,5, and pSiO must be equal to 102. This value of pSiO is much higher and produces an increase of evaporation rate. These diagrams can be used to predict the behaviours of materials at extreme conditions at high temperature or reducing atmosphere, and they are also useful to determine innovative techniques to produce ceramic materials. In the same way we can approach systems like Si-C (Figure 4), Si-N, Si-C-N-O and Si-N-O in order to develop the synthesis conditions and stability fields of silicon nitrides and carbides (from Heuer and Lou, 1990). APPLICATIONS Chemical Vapour Deposition - CVD CVD has been proved to be one of most attractive techniques because materials are directly obtained by thermal decomposition of vapour. CVD is useful to produce highly dense and pure materials, especially carbides, nitrides and oxides. Via CVD it is possible to depose thin layers of condensed phases (coating) on ceramics or other materials (i.e. ZrO2 on metals, to improve thermal properties). For example CVD ?-SiC coatings can be produced on induction-heated graphite substrates into a quartz bell jar under atmospheric pressure, fluxing compounds like SiH4, SiCl4, (CH3)2SiCl3. Changing experimental parameters like substrate temperature, concentration and ratio of precursor gases, time of exposure, it is possible to control the microstructure of the coating layer and the mechanical properties of the composed material. In order to determine the conditions effective for deposition on a substrate we should know the smoke conditions in a multicomponent system (Figure 3b and 5). In the diagram Si-O the smoke field in inert or environmental conditions is defined at pSiO higher than isomolar or isobaric points, and lower than T conditions; the break of this equilibrium (typically on a low temperature substrate) determines the drop of solid phase onto the substrate. Such fields can be maximized either in extent or in pSiO, e.g. using a H2-bearing gas. Both strategies have benefits and drawbacks, e.g. higher pSiO will lead to a more efficient drop and deposition, but as the smoke field is more restricted, the correct conditions will be hardly attained. Since sintering processes, powders production, and CVD of carbides/nitrides are commonly performed in the presence of graphite (substrate or furnace resistance), the carbon activity must also be considered as a major "technological" variable. This variable can change substantially the stability of solid phases and open the possibility of multilayer deposition (Figure 6, from Wada et al. 1988). In the system Si-O-C-N, for a synthesis performed in a graphite furnace (i.e. ac=1), SiC can be deposed at temperature higher than 1374C with 1 atm N2, whereas temperatures higher than 1536C are required if pN2 = 10 atm. If CVD is performed in non-graphite furnace (ac=10-4) SiC field is suppressed. Transition from passive to active oxidation to promote the volatilization of protective SiO2 thin layers in silicon, silicon nitride and carbide powders. Technical nano-powders are extremely reactive and potentially unstable even at low temperature so that they frequently require passivation (or they naturally passivate during early stages of the

sintering process) to preserve their characteristics. For example Si3N4 powders are oxidated in low temperature processes in order to produce a protective thin oxide layer or to enable the control of slurry properties in aqueous and non-aqueous suspensions. Experimental studies devoted to the definition of variables and parameters responsible for the production and removal of thin layers are very important in industrial applications in order to save time, money and energy or to improve properties of ceramics materials. As an example, the nitridation kinetics of Si powder occurs in two distinct stages. The initial stage involves a devitrification of the silica layer via reaction: SiO2 + Si 2 SiO(g) The second stage produce the oxynitride sinoite according to equation: 2 SiO(g) + N2 Si2N2O + 1/2 O2 Both reactions are rapidly suppressed as the oxygen liberated competes with nitrogen to reoxidize the exposed silicon surface. The kinetics of this process is slow because these reactions occur at low pSiO and because there is no way to remove oxygen from the system. In order to prevent the reoxidation of silicon, it is possible to mix H2 into N2. The oxygen present in the system reacts with hydrogen and, in this case, possible reactions are: 2 Si + SiO2 + N2 SiO(g) + Si2N2O 9 Si + 3 SiO2 + 4 N2 6 SiO(g) + 2 Si3N4 These reactions occur at high pSiO and have rapid kinetics. When pO2 is very low, other reactions which imply the passive oxidation of silicon, may occur: Si + SiO(g) + N2 Si2N2O 3 Si + 2 N2 Si3N4 In presence of hydrogen the extent of reaction is greatly enhanced at lower temperature removing the passivating layer which appears to be an excellent diffusion barrier. Figure 7 shows clearly these two distinct stages consuming nitrogen (which means nitridation of Si) and the effect of hydrogen on reaction kinetics. Corrosion of SiC or Si3N4 in gaseous environments. The growing interest in the use of silicon carbide and nitride for structural, electronic, and thermal applications requires a thorough knowledge of material properties and characteristics (Ziegler et al. 1987). The oxidation behavior is one of the most important aspects. Although thermodynamically unstable in oxidizing environments, SiC can form a protective oxide layer that effectively inhibits further oxidation. Kinetics of oxidation is diffusion controlled through the protective layer. Volatility diagrams in simple chemical systems define the conditions which govern the formation or disruption of this layer. However, more complex processes can be introduced both in the development stage and in applications. As an example, Figure 8 shows the evolution of a reaction sequence for Si3N4 in an alkali-containing atmosphere (Sun et al. 1994).

Selective active oxidation. An interesting application of Ellingham diagrams deals with the purification of crystalline phases by reduction and volatilization of selected compounds, at controlled oxygen partial pressures. For example, mullite can be purified from the residual SiO2-rich glass phase which concentrates impurities of Na2O, K2O Fe2O3 and other minor compounds, by a thermal treatment at high T, under low partial pressure of oxygen (Souto and Guitian, 1999). Profiting of the computed Ellingham diagram, pO2 should be set below the stability limits of silica and metal oxides in the glassy phase but above the stability limit of mullite in order to destabilize glass but without altering mullite. A continuous monitoring and adjustment of oxygen partial pressure in industrial process has major benefits: it permits the use of alkali to promote the appearance of a liquid phase at low temperature and therefore to increase the kinetics of mullite formation, but also to remove the liquid at the end of sintering process. Moreover, it does not cause pollution as it happens when purification is achieved using HF or NaOH solutions. REFERENCES Barsoum M., Kangutkr P., Koczack M.J., 1991. Nitridation kinetics and thermodynamics of Silicon powder compacts. J. Am. Ceram. Soc. 74: 1248-53. Callen H.B., 1985. Thermodyamics and an introduction to thermostatistics. Wiley. Heuer A.H., Lou V.L.H., 1990. Volatility diagrams for silica, silicon nitride, and silicon carbide and their application at high-temperature decomposition and oxidation. J. Am. Ceram. Soc. 73: 27852802. Korzhinskii D.S., 1959. Physicochemical basis of the analysis of the paragenesis of minerals. Chapman and Hall, London Lou V.L.K., Mitchell T.E. and Heuer A.H., 1985. Graphical displays of thermodynamics of hightemperature gas-solid reactions and their application to oxidation of metals and evaporation of oxides. J. Am. Ceram. Soc. 68: 49-58. Souto A. and Guitian F., 1999. Purification of mullite by reduction and volatilization of impurities. J. Am. Ceram. Soc. 82: 2060-64. Sun T., Pickrell G.R., and Brown J.J., 1994. Corrosion kinetics of silicon nitride in dry air containing sodium nitride vapors. J. Am. Ceram. Soc. 77: 3209-14. Wada H., Wang M-J., and Tien T-Y., 1988. Stability of Phases in the Si-C-N-O System. J. Am. Ceram. Soc. 71: 837-40 Ziegler G., Heinrich J., Wotting G., 1987. Review - Relationships between processing, microstructure and properties of dense and reaction-bonded silicon nitride. J. Mater. Sci. 22: 30413086.

FIGURE CAPTIONS Figure 1. Schematic sequence for the constuction of volatility diagrams. From Heuer and Lou 1990. Figure 2. Reactions which govern active and passive oxidation in the systems Si-O-H, Si-O-C, and Si-O-N. Figure 3. Evolution of silica and silicon compounds processed in a controlled atmosphere at hightemperature conditions. a) behaviour of silica in a closed system where oxygen is neither added nor removed by reactions with additional volatile components. b) behaviour of silicon in a system open to oxygen (oxygen externally buffered), but without interaction with additional volatiles. c) behaviour of silicon in a system open to oxygen, in the presence of hydrogen. Further volatile species, e.g. H2O, result from equilibria established between O2 and H2. Figure 4. Volatility diagram for the system Si-O-C. From Heuer and Lou 1990. Figure 5. Schematic process for smoke generation in the system Si-O-H as a basis for chemical vapour deposition (CVD). Figure 6. Phase relationships in the system Si-N-O-C according to Waada et al. 1988. Figure 7. Nitrogen consumption during nitridation process of silicon powders (From Barsoum et al. 1991). Negative peaks correspond to advancement of nitridation reactions. Figure 8. Oxidation of silicon nitride. After Sun et al. 1994. Figure 9. Use of an Ellingham diagram for purfication of mullite from alkali-bearing liquids. After Souto and Guitian 1999.