Kim and Moorehead [5.112] found a similar dependence upon water vapor pressure for the oxidation of HP-Si3N4 containing 6 wt.% Y2O3 and 1.5 wt.% Al2O3. In the low-pressure region, however, the magnitude of the weight loss for the HP material was about three times greater than that of the CVD material. This difference was attributed to the greater surface area for reaction in the HP material, because it had a much smaller grain size compared to the CVD material. At the higher water vapor pressures, the grain boundaries were not preferentially attacked and thus the two materials exhibited similar weight losses.

Other Nitrides

The crystalline solution series of materials of alumina dissolved into β-silicon nitride (Si6-xAlxOxN8-x) make up a truly interesting series of materials. The hopes were that these materials would yield properties that were the best of the two end members. One improvement over silicon nitride is the oxidation resistance with increasing amounts of alumina while maintaining the relatively low thermal expansion characteristics of the pure silicon nitride phase. Weight gain behaviors have been reported to be parabolic with mullite being the oxide that formed on the surface. Singhal and Lange [5.125] reported that mullite formed only in those compositions containing more than 20 wt.% alumina and that, above 40 wt.% alumina, additional unidentified phases occurred. Chartier et al. [5.126] prepared sialon crystalline solutions with x=0.4. Because 14.05 wt.% yttria was added to the original mix, the final pressureless sintered samples contained β’-Y2Si2O7 and a glassy phase as grain boundary phases. Oxidation was parabolic and very slow below 1380°C. Above this temperature, more rapid oxidation occurred with departures from parabolic behavior. A thin alumino-silicate film formed first, but as metal cation migration occurred (predominately yttrium) reaction with this film formed more complex silicates. This film was dense below 1400°C, and gradually became porous and nonprotective as the temperature was increased. Visual observation indicated a light

Copyright © 2004 by Marcel Dekker, Inc.

gray zone under the surface scale that Chartier et al. reported to be attributable to selective oxidation of the grain boundary phase.

Wang et al. [5.127] investigated the oxidation in air of rareearth aluminum oxynitrides with the ideal formula LnAl12O18N (Ln=La, Ce, Pr, Nd, Sm, and Gd) at temperatures ranging from 700 to 1500°C. Noticeable oxidation started at about 700°C, and increased with temperature. The final reaction products depended upon the particular rare-earth, but progressed through several intermediate stages. At temperatures around 1000°C, the products were LnAl11O18 and α-alumina. LnAlO3 also formed with or without the disappearance of LnAl11O18 at higher temperatures depending upon the rare-earth. In ceriumcontaining materials, CeO2 formed at 900°C. At temperatures as low as 700°C, lattice parameter changes (increasing c/a ratio) were noted for the oxynitrides, which were attributed to the initiation of oxidation.

AlN is an important material in the electronic ceramics industry and is an example of when a small amount of oxidation is beneficial to the application. In this case, the formation of a thin (1–2 µm) protective coating of AlON is formed and is used to improve the adhesion of copper films. Suryanarayana [5.128] found the oxidation of AlN powders between 600 and 1000°C in flowing air to follow a linear rate law initially, and then a parabolic law as the oxide layer thickness became sufficient to require diffusion for further growth. In contrast, Abid et al. [5.129] found that the oxide layer that formed on polycrystalline AlN in air at 1200°C was α-Al2O3, whereas below 800°C, no oxidation was observed. Dutta et al. [5.130] reported that oxidation of sintered polycrystalline AlN between 20 and 200°C progressed from individual α-Al2O3 particles of 2–3 nm in size to a 50-nm-thick film after 150 hr at 200°C. They also commented that their data were consistent with the formation of an oxynitride layer, but believed α-Al2O3 to be the oxide formed at low temperature based on thermodynamic calculations. Others have shown that an oxynitride formed as an intermediate preceding alumina formation at high temperatures [5.131], and

Copyright © 2004 by Marcel Dekker, Inc.

McCauley and Corbin [5.132] reported that a region of ALON stability occurred between Al2O3 and AlN at temperatures between 1800 and 2050°C in flowing nitrogen.

The oxidation of TaN to Ta2O5 was reported to commence at about 450°C by Montintin and Desmaison-Brut [5.133]. As the temperature is raised, the initially powdered reaction product densified; however, the high volume expansion of Ta2O5 generated stresses in the coating that caused failure and spalling at high stress regions. Between 590 and 770°C in oxygen, the kinetics of the reaction were characterized by a sigmoidal rate law associated with the formation of the nonprotective Ta2O5.

Silicon Carbide

Oxidation. The oxidation of green hexagonal powdered SiC has been described by Ervin [5.134]. Ervin stated that oxidation at low oxygen pressures took place with the formation of SiO gas, while at atmospheric pressure under flowing air, SiO2 formed. The rate-controlling step was thought to be the growth of an ordered lattice of SiO2 by solid diffusion. The following reactions are representative of the oxidation of silicon carbide:

(5.45)

(5.46)

Jorgensen et al. [5.135] proposed that the rate-controlling step in the growth of the SiO2 layer formed on powdered SiC may be the diffusion of either oxygen ions or silicon ions. They ruled out the diffusion of molecular O2, CO2, and CO based upon their experimentally determined activation energies being too large for molecular diffusion. Harris [5.136] studied the oxidation of crystals of 6H-αSiC and determined that the rate of oxidation on the (0001) carbon face was approximately seven times greater than that on the silicon face at 1060°C for 70 hr. The thin oxide layer on the (0001) silicon face grew according to a linear rate law at all temperatures, whereas

Copyright © 2004 by Marcel Dekker, Inc.

the thick oxide on the carbon face initially grew with linear kinetics but then changed to parabolic when the thickness became greater than 250 nm. At high temperatures and/or long times during oxidation of powdered samples, the oxidation rate changed from parabolic to linear presumably because the growth of the linearly controlled face overtook that of the parabolically controlled face. This change in oxidation rate at high temperatures has been attributed to a change in the oxide layer from amorphous to crystalline by Ervin [5.134] and Jorgensen et al. [5.137], and suggested by Costello and Tressler [5.138].

The desorption of CO gas formed at the SiC/SiO2 interface has been reported to be the rate-controlling step by Singhal [5.139]; however, Hinze et al. [5.140] and many others have reported that it is the inward diffusion of oxygen through the surface layer of SiO2. Spear et al. [5.141] ruled out the diffusion of CO as rate-controlling based upon their experiments that exhibited a dependence of the oxidation rate upon the partial pressure of oxygen and the almost identical activation energies obtained for the oxidation of SiC and Si metal. Fergus and Worrell [5.142] have concluded that the various contradictions in reported kinetics were attributable to a change in the diffusing species from molecular to ionic oxygen at about 1400°C. This was based upon two observations: one being that the activation energy for the growth of amorphous silica on CVD SiC increased above 1400°C, and the other being that the activation energy for the growth of cristobalite increased, but at the higher temperature of 1600°C. Decreases in oxidation rates at low temperatures have been attributed to sufficiently long times to allow crystallization of the silica scale.

In an analysis of the various possible rate-controlling steps, Luthra [5.85] concluded that a mixed interface reaction/diffusion process was the limiting feature in the oxidation of SiC. This was based upon the following facts:

1.Oxidation rate is lower than for pure silicon

2.Presence of gas bubbles in the oxide layer

Copyright © 2004 by Marcel Dekker, Inc.

3.Oxidation rate of single crystals dependent upon crystallographic orientation

4.Higher activation energy than for pure silicon (although Spear et al. [5.141] reported similar energies)

Because all of the above, except the presence of gas bubbles, are consistent with interface reaction control and the fact that bubbles are present, a mixed controlled process was concluded. Luthra suggested that mixed control should yield a rate law more complex than the generally observed linear or parabolic laws.

For pure monolithic CVD SiC and Si3N4, Fox [5.143] reported oxidation rates for 100 hr at temperatures between 1200 and 1500°C in flowing dry oxygen to be similar. In silicon nitride, any additives present will affect the oxidation rate. In general, increased levels of additives or impurities result in higher oxidation rates. These higher oxidation rates are attributable to the migration of the additive to the oxidized layer, thus lowering the viscosity, which increases the diffusion of the oxidant to the SiC/SiO2 interface. Fergus and Worrell [5.142] reported that 0.5 wt.% boron in sintered α-SiC did not, however, significantly affect the oxidation rate.

Understandably, the active oxidation of SiC has not been investigated quite as thoroughly as passive oxidation; however, it should be remembered that active oxidation to SiO gas can occur at any temperature if the oxygen partial pressure at the SiC surface falls below some critical value. Not all data reported in the literature agree. The variations reported for this transition are attributable to the differences in the SiC materials tested and in the experimental conditions used. The partial pressure of oxygen at the transition from passive-to-active oxidation decreases with an increase in the total gas flow through the system [5.127]. This is the result of a decreasing gaseous boundary layer thickness with increasing velocity. The total gas pressure of the system can also affect results as suggested by Narushima et al. [5.144], because molecular gas flow exists at low pressures and viscous gas flow exists at higher pressures,

Copyright © 2004 by Marcel Dekker, Inc.

thus changing the gas diffusion phenomena. Because the ratecontrolling mechanism in active oxidation is the oxygen diffusion through the gaseous boundary layer, the characteristics of the gaseous boundary layer play a major role in the oxidation. If experiments were conducted at very high flow rates and very low total pressures, as was the case for the work of Rosner and Allendorf [5.145], the rate-controlling step may no longer be oxygen diffusion through the gaseous boundary layer, but the kinetics of gas arrival and removal from the surface. The active- to-passive transition as determined by several investigators and compiled by Vaughn and Maahs [5.102] is shown in Fig. 5.14. The variations for the reported transition are attributable to differences in the gas flow rates of the tests and possibly differences in the SiC material tested.

FIGURE 5.14 Literature data for active-to-passive oxidation transition for SiC. (From Ref. 5.102; reprinted with permission of The American Ceramic Society, www.ceramics.org. Copyright ©1990. All rights reserved.)

Copyright © 2004 by Marcel Dekker, Inc.

The oxidation of SiC fibers and whiskers is about as diverse as it is for other forms of SiC. Not only is the corrosive degradation of fibers and whiskers complexed by their chemistry (containing impurities of C, and SiO2) and structure (containing more than one polymorph), but their surface area-to-volume ratio greatly enhances reaction rates when compared to an equal weight of some other form. SiC fibers manufactured from polycarbosilane polymer precursors generally contain excess carbon, silica, and some combined nitrogen. Jaskowiak and DiCarlo [5.146] reported the weight loss behavior of SiC fibers at temperatures ranging from 1000 to 2200°C at argon pressures of 0.1 and 138 MPa and under vacuum (10-9 MPa). Although the high external pressure delayed the onset of weight loss from about 1250 to 1550°C, active oxidation occurred through the formation of SiO. Wang et al. [5.147] measured the oxide layer thickness on SiC whiskers for the low temperature (600, 700, and 800°C) linear region at times less than 4 hr to be between 2 and 10 nm as determined by XPS analysis and X-ray photoelectron spectroscopy.

The wide variation of oxidation rates and activation energies reported in the literature is a result of one or a combination of many factors, including:

1.Decrease in reactive area with advancing oxidation (taken into account by some but not all)

2.Differences in materials studied (α, β, or amorphous)

3.Density and porosity variations

4.Variation and amount of preexistent surface oxide

5.Differences in oxide layer formed (crystalline, amorphous, or liquid)

6.Amounts and type of additives and impurities

Reaction in Other Atmospheres. McKee and Chatterji [5.22] reported no oxidation of SiC when exposed to gaseous environments of pure H2, pure N2, or H2–10%H2S at 900°C. No evidence of sulfide formation was found in the hydrogen- H2S mixture. In a mixture of N2–2%SO2, which resulted in a partial pressure of oxygen of 10-10 atm, active oxidation was

Copyright © 2004 by Marcel Dekker, Inc.

observed. With the addition of 5% CH4 to the mixture of N2– 2%SO2, an initial (first hour) rapid weight loss was noted, presumably as a result of the formation of the volatile SiS.

Reaction of SiC in gas mixtures of 5% H2/H2O/Ar when at 1300°C was predicted by Jacobson et al. [5.148] to fall within one of three regions: passive oxidation, active oxidation, or selective carbon removal depending upon the water content of the mixture. Gas phase diffusion (i.e., water transport to the SiC) was reported to be the rate-controlling step in the active oxidation region (oxygen partial pressures of 10-22 to 10-26 atm). In the carbon removal region (oxygen partial pressures less than 10-26 atm), iron impurities were found to react with the free silicon present to form iron silicides.

Maeda et al. [5.149] investigated the oxidation of several different SiC materials in flowing humid air containing 1–40 vol.% water vapor at a temperature of 1300°C for 100 hr. They found that water vapor greatly accelerated the oxidation of SiC, and that a linear relationship existed between percent water vapor and weight gain. Lu et al. [5.150] found that oxidation rates of SiC thin films were increased by 10–20 times in wet oxygen when compared to dry oxygen at temperatures of 950– 1100°C. The active oxidation (i.e., weight loss) of SiC was reported to occur in 1 atm hydrogen containing water vapor at pressures of 10-6 to 10-3 MPa between 1400 and 1527°C by Readey [5.151]. At high water vapor pressures, a reaction product of SiO2 was formed; however, active oxidation continued, because this SiO2 was reduced to SiO by the hydrogen present. The reactions that took place can be represented by the following equations:

(5.47)

(5.48)

In addition to water vapor, alkali vapors have been shown by Pareek and Shores [5.152] to enhance oxidation rates. They studied the oxidation of α-SiC in flowing gas mixtures of dry

Copyright © 2004 by Marcel Dekker, Inc.

CO2–O2 (9:1 ratio) containing small quantities of K2CO3 and K vapors at 1300–1400°C for times of up to 42 hr. Water vapor was added in some tests; however, the vapor species in those cases was KOH. Pareek and Shores found, at low potassium levels, that the oxidation to SiO2 followed a parabolic rate law; at higher potassium levels, the growth followed a linear law; and when low levels of water vapor were also present (i.e., KOH vapors), the growth kinetics were intermediate between parabolic and linear, indicative of a possible transition from one rate law to another. At moderate to high levels of potassium in the presence of water vapor, the kinetics of oxidation again followed a linear rate law. The increased oxidation in atmospheres containing potassium vapors was suggested to be a result of the enhanced mobility of the oxidant through the oxide layer containing dissolved potassium, although the reported activation energy of 225–463 kJ/mol was much higher than expected for oxygen diffusion through silica, which is about 115 kJ/mol. The scales were determined to be composed of cristobalite under most test conditions. At higher potassium levels and higher temperatures, the scale was sufficiently fluid to flow from the samples.

Federer [5.153] studied the effects of a vaporized solution of water containing 1 wt.% NaCl in air upon sintered α-SiC under a mechanical load at 1200°C. He reported that a molten reaction layer of sodium silicate formed causing premature failure under load within an average time of about 150 hr. The same material, when exposed to a 1200°C in air and the same loading conditions, could sustain the stress without failure for at least 1500 hr. In a similar test, Federer [5.154] exposed several types of SiC to a flowing atmosphere containing sodium sulfate and water vapor in air at 1200°C. In these tests, Federer reported that the reaction layer contained tridymite embedded in a sodium silicate liquid. Enhanced oxygen diffusion through this liquid allowed continued corrosion to take place. No discussion was given for the effects of SO3 gas upon the corrosion as Jacobson and coworkers did [5.69, -5.72], other than to state that sodium sulfate vapor reacted with silica under low partial pressures of SO3.

Copyright © 2004 by Marcel Dekker, Inc.