14 Journal of the American Ceramic SocietyJacobson et al. Vol. 82. No 6 In summary, at low oxygen contents, such as would be found deep within a composite, SiC is expected to oxidize and effe tively getter oxygen. At higher oxygen contents, SiC and BN ding to borosilicate glass formation As time progresses, residual water vapor volatilizes the B2O3 TIme, h (3) Oxidation in Higli-) Measurements of Recession Fig. 6. TGA kinetics of CvD Sic coated with 2 Figure 1(b)illustrates the effect of a high-water-vapor con- Dosition-temperature(1900C)CVD BN tent and lower-temperature gas-oxidizing gas stream. Under 70ppmH2Oat900°C. hese conditions. volatilization dominated. and the bn was om composite. This was most pro- nced for the bn deposited at lower t weight loss is observed, which is very likely due to water- cal parameter was the depth of recession. These recession dis- apor-enhanced boron volatilization from the borosilicate tances were measured with single -tow, BN-coated CVD glass. Thus, the two types of bn behave qualitatively ver minicomposites with a range of interphase thicknesses, as de different rates due to the different types of BN osses-but at scribed in the experimental section. After heat treatment les were An important implication for composites is that, as boron is determine the distance of bn recession due to volatilization. 19 emoved from the borosilicate liquid, the viscosity of the bo- Figure 9 shows an example of a polished longitudinal section rosilicate liquid increases. 6 This would be expected to occur of the minicamp described in Table I. The identity of each for composites where the BN-interphase oxidation product is phase is delineated in Fig. 9. EDS measurements are taken to xposed to the H,o-containing environment, i.e., with a matrix determine the depth of recession. The micrograph in Fig. 9 lustrates the transition from bn to oxidized bN. The unaf crack that extends to the BN coating. The liquid could even- fected BN interphase could be discerned from a stronger nI- tually freeze to form a solid glass as the Sio, content of the lass increases, leading to stress concentrations placed on the fibers and premature composite failure. This was believed to be observed for static and cyclic tensile stress experiments per- deposit type of BN contains ted. The oxidized BN interphase is a region of glass, EDS measurements show contains oxygen and limite formed at -900C in air after -2 h 12, nitrogen. Although too small for an accurate EDS measure ment, it is likely that this region is a borosilicate glass. Because (2) Oxidation in Low-Water-Vapor Environments- Layered Sic/BN/SiC Model Compounds rom bn to glass, the interfaces are not as clean as would be The layered compounds are closer to the actual desired for recession distance measurements. this leads to and are examined in the same low-water-vapor/oxyg scatter in the data ronment. Figure 2 illustrates the preparation of this sa Recession data are shown as the filled circles on Figs three regions at which the data in Fig. 7(a(c)are take (c).9 In each case, the actual coating thickness is listed indicate oxidation effects inside a composite the filled circle. There are several important points about Figure 7(a)is-4100 um into the sample data. Note, the time dependence is not linear, but rather reces- and the thin intermediate layer of SiC are qu There is sion rate decreases with time. In fact, it appears that often the a very limited amount of oxygen. Some of th hannels seal off, ending the recession processes. This is con- to be dissolved in the bn and some appears te Sic/bn sistent with the data of Brun and Luthra. I Recession is only measured at two temperatures, but there does not seem to be a Figure 7(b)is-2000 um into the sample. Although the Bn strong dependence on temperature. There is, however, a de- is intact, there is considerably more oxygen present. The thi pendence on water-vapor content in the gas stream. This sug- intermediate layer of SiC is almost entirely oxidized, although gests the importance of various H-B-0 gaseous species it appears to have some boron on it. Thermochemically, SiC should oxidize at a low Po,. Figure 8 shows the oxygen poter (4 Oxidation in High-Water-Vapor Environments- tials set by Sio,/SiC, SiO,/SiC, BN/B,O3, and BN/B,O3 Mechanism of recession CThe underline indicates unit activity. These are calculated We did a free-energy minimization calculation to deter from the janaf thermochemical datas as follows nine the vapor species in equilibrium with B2O3 and 1% and 10% H,O/O2. This calculation indicated HBO(g), H3 BO3(g), Si+O,= SiO (3) and H, B, O(g)are the dominant species. Henceforth, we shall SiC +O2= Sio2+ C refer to these with the general term H, B, o- g) The mechanism of recession is illustrated by Fig. 11. First, An analogous set of calculations is done for BN/B,O3. These assume there is a layer of B,O, on the BN. Then, the following calculations indicate that SiC oxidizes at a lower oxygen eps tential than BN, and, when the supply of oxygen is limited (1)H,O diffuses through the boundary layer. ffectively getters the oxygen so that bn does not oxidize. This (2)H2O diffuses through the annular-shaped channel consistent with the results and model of Sheldon et al. 9 fo (3) H,O reacts with the B,Oa(e) at the base of the channel BN-coated SiC fibers to form HBO(g), H3 BO3(g), and H3.(g) Figure 7(c) is taken adjacent to the exposed face of the (4)HBO2(g), H3BO3(g), and H B.(g) diffuse out of the sample, where the oxygen potential is high. Note, there is a channel micrograph--where the oxygen potential is highest--shows (5)HBO2(g), H3 BO3 (8), and H3 B3 O(g) diffuse out substantial amount of oxide formation. The right side of the ough the boundary layer 10 um of Sio2. As discussed in the previous section, this is (6) The cross section of the annular channel gradually de- far more than would be expected from simple oxidation of pure reases because of boron-enhanced oxidation of the walls SiC. However, this can be explained by boron-enhanced oxi- shown in Fig. 11 dation of SiC to borosilicate and subsequent volatilization of In order to describe this process, we need to determine which BO from the borosilicate. which leaves behind a thick sio )2 step is rate-controlling. It is well-known that boundary-layer control would lead to a linear dependence of recession distanceweight loss is observed, which is very likely due to water￾vapor-enhanced boron volatilization from the borosilicate glass. Thus, the two types of BN behave qualitatively very similar—showing both weight gains and weight losses—but at different rates due to the different types of BN. An important implication for composites is that, as boron is removed from the borosilicate liquid, the viscosity of the bo￾rosilicate liquid increases.16 This would be expected to occur for composites where the BN-interphase oxidation product is exposed to the H2O-containing environment, i.e., with a matrix crack that extends to the BN coating. The liquid could even￾tually freeze to form a solid glass as the SiO2 content of the glass increases, leading to stress concentrations placed on the fibers and premature composite failure. This was believed to be observed for static and cyclic tensile stress experiments per￾formed at ∼900°C in air after ∼2 h.12,17 (2) Oxidation in Low-Water-Vapor Environments— Layered SiC/BN/SiC Model Compounds The layered compounds are closer to the actual composite and are examined in the same low-water-vapor/oxygen envi￾ronment. Figure 2 illustrates the preparation of this sample and three regions at which the data in Fig. 7(a)–(c) are taken. These indicate oxidation effects inside a composite. Figure 7(a) is ∼4100 mm into the sample and shows the BN and the thin intermediate layer of SiC are quite intact. There is a very limited amount of oxygen. Some of the oxygen appears to be dissolved in the BN and some appears to be at the SiC/BN interfaces. Figure 7(b) is ∼2000 mm into the sample. Although the BN is intact, there is considerably more oxygen present. The thin intermediate layer of SiC is almost entirely oxidized, although it appears to have some boron on it. Thermochemically, SiC should oxidize at a low pO2 . Figure 8 shows the oxygen poten￾tials set by SiO2/SiC, SiO2/SiC, BN/B2O3, and BN/B2O3. (The underline indicates unit activity.) These are calculated from the JANAF thermochemical data18 as follows: Si + O2 4 SiO2 (3) SiC + O2 4 SiO2 + C (4) An analogous set of calculations is done for BN/B2O3. These calculations indicate that SiC oxidizes at a lower oxygen po￾tential than BN, and, when the supply of oxygen is limited, effectively getters the oxygen so that BN does not oxidize. This is consistent with the results and model of Sheldon et al.9 for BN-coated SiC fibers. Figure 7(c) is taken adjacent to the exposed face of the sample, where the oxygen potential is high. Note, there is a substantial amount of oxide formation. The right side of the micrograph—where the oxygen potential is highest—shows ∼10 mm of SiO2. As discussed in the previous section, this is far more than would be expected from simple oxidation of pure SiC. However, this can be explained by boron-enhanced oxi￾dation of SiC to borosilicate and subsequent volatilization of B2O3 from the borosilicate, which leaves behind a thick SiO2 layer. In summary, at low oxygen contents, such as would be found deep within a composite, SiC is expected to oxidize and effec￾tively getter oxygen. At higher oxygen contents, SiC and BN oxidize simultaneously, leading to borosilicate glass formation. As time progresses, residual water vapor volatilizes the B2O3 from this glass. (3) Oxidation in High-Water-Vapor Environments— Measurements of Recession Figure 1(b) illustrates the effect of a high-water-vapor con￾tent and lower-temperature gas-oxidizing gas stream. Under these conditions, volatilization dominated, and the BN was essentially removed from the composite. This was most pro￾nounced for the BN deposited at lower temperatures. The criti￾cal parameter was the depth of recession. These recession dis￾tances were measured with single-tow, BN-coated CVD minicomposites with a range of interphase thicknesses, as de￾scribed in the experimental section. After heat treatment, samples were mounted in epoxy and longitudinally polished to determine the distance of BN recession due to volatilization.19 Figure 9 shows an example of a polished longitudinal section of the minicomposite described in Table I. The identity of each phase is delineated in Fig. 9. EDS measurements are taken to determine the depth of recession. The micrograph in Fig. 9 illustrates the transition from BN to oxidized BN. The unaf￾fected BN interphase could be discerned from a stronger ni￾trogen peak, because this type of BN contains oxygen as￾deposited. The oxidized BN interphase is a region of glass, which EDS measurements show contains oxygen and limited nitrogen. Although too small for an accurate EDS measure￾ment, it is likely that this region is a borosilicate glass. Because of the fluidity of the glass products and the gradual transition from BN to glass, the interfaces are not as clean as would be desired for recession distance measurements. This leads to scatter in the data. Recession data are shown as the filled circles on Figs. 10(a)– (c).19 In each case, the actual coating thickness is listed above the filled circle. There are several important points about these data. Note, the time dependence is not linear, but rather reces￾sion rate decreases with time. In fact, it appears that often the channels seal off, ending the recession processes. This is con￾sistent with the data of Brun and Luthra.11 Recession is only measured at two temperatures, but there does not seem to be a strong dependence on temperature. There is, however, a de￾pendence on water-vapor content in the gas stream. This sug￾gests the importance of various H-B-O gaseous species. (4) Oxidation in High-Water-Vapor Environments— Mechanism of Recession We did a free-energy minimization20 calculation to deter￾mine the vapor species in equilibrium with B2O3 and 1% and 10% H2O/O2. This calculation indicated HBO2(g), H3BO3(g), and H3B3O6(g) are the dominant species. Henceforth, we shall refer to these with the general term HxByOz(g). The mechanism of recession is illustrated by Fig. 11. First, assume there is a layer of B2O3 on the BN. Then, the following steps occur: (1) H2O diffuses through the boundary layer. (2) H2O diffuses through the annular-shaped channel. (3) H2O reacts with the B2O3(,) at the base of the channel to form HBO2(g), H3BO3(g), and H3B3O6(g). (4) HBO2(g), H3BO3(g), and H3B3O6(g) diffuse out of the channel. (5) HBO2(g), H3BO3(g), and H3B3O6(g) diffuse out through the boundary layer. (6) The cross section of the annular channel gradually de￾creases because of boron-enhanced oxidation of the walls, as shown in Fig. 11. In order to describe this process, we need to determine which step is rate-controlling. It is well-known that boundary-layer control would lead to a linear dependence of recession distance Fig. 6. TGA kinetics of CVD SiC coupon coated with 2 mm of high-deposition-temperature (∼1900°C) CVD BN and oxidized in O2/ 70 ppm H2O at 900°C. 1476 Journal of the American Ceramic Society—Jacobson et al. Vol. 82, No. 6