A study on the oxidation kinetics and mechanism of three-dimensional (3D) carbon fiber braid coated by gradient SiC

Materials Chemistry and Physics 93 (2005) 164–169

A study on the oxidation kinetics and mechanism of three-dimensional (3D) carbon fiber braid coated by gradient SiC Pengzhao Gao a,∗ , Hanning Xiao a , Hongjie Wang b , Zhihao Jin b b

a College of Materials Science and Engineering, Hunan University, Changsha 410082, PR China State Key Laboratory for Mechanical Behavior of Materials, Xi’an Jiaotong University, Xi’an 710049, PR China

Received 4 October 2004; received in revised form 2 March 2005; accepted 7 March 2005

Abstract The oxidation kinetics and mechanism of gradient SiC coating/3D carbon fiber braid (coated braid, CB) in static air were studied through isothermal oxidation-weight loss, non-isothermal thermogravimetric (TG) and differential thermogravimetric (DTG) also. The results showed that the oxidation process of CB in isothermal condition was reaction-controlled in the first step and, gas diffusion and reaction-controlled in the second step. In non-isothermal condition, the characteristic of oxidation process was self-catalytic. The oxidation mechanism was one-dimensional diffusion. The kinetic parameters were: log A (min−1 ) = 10.12 and Ea (kJ mol−1 ) = 224.96. © 2005 Elsevier B.V. All rights reserved. Keywords: Gradient SiC coating; Oxidation kinetics and mechanism; TG–DTG; Self-catalytic

1. Introduction Carbon fiber reinforced composites (ceramic, glass, and polymers) have received much attention in recent years, since their mechanical strength, thermal, and chemical properties enable them to be widely used in many industries. Recently, attention has been focused on 3D carbon fiber braid reinforced composites in order to meet mechanical and thermal properties requirements along the thickness of the composites [1]. However, carbon fiber exhibits a poor oxidation resistance even at temperatures as low as 700 K. Consequently, carbon fiber reinforced composites are restricted to inert atmospheres in order to prevent degradation of the carbon fiber [2]. Once the strong interfacial bond between fiber and matrix has been debonded by oxidation, a significant reduction in the mechanical properties of the composite is unavoidable [3,4]. Thus, the oxidation protection of carbon fiber is one of the major factors concerning its use in composite. Ceramic coatings, used as a barrier to suppress oxygen diffusion, have been studied [2,5–7]. A new challenge is to ∗

Corresponding author. Tel.: +86 731 8822269; fax: +86 731 8821483. E-mail address: [email protected] (P. Gao).

0254-0584/$ – see front matter © 2005 Elsevier B.V. All rights reserved. doi:10.1016/j.matchemphys.2005.03.003

find an effective coating to protect the fiber while being thin enough so as not to crack (due to the thermal mismatch between coating and fiber). Gradient coating as a new technique in development could satisfy the above demands [8]. However, little work has been done on the oxidation protection of 3D carbon fiber braid because it is difficult to prepare uniform coating on each fiber of the braid. In addition, the oxidation properties of ceramic coating/carbon fiber has not been studied carefully, which could help to develop more effective oxidation protection for the carbon fiber. Thermogravimetric analysis (TGA) is one of the quantitative analysis techniques used to study the oxidation kinetics and mechanism of solid material. In the present work, the appearance and composition of CB were studied through Xray diffraction (XRD) and SEM; the oxidation kinetics and mechanism of CB were studied through isothermal oxidationweight loss and non-isothermal TG–DTG.

2. Experimental procedures 2.1. Experimental materials The 3D carbon fiber braid (raw braid), which was cut into blocks of 20 mm × 20 mm × 20 mm, was PAN-based carbon

P. Gao et al. / Materials Chemistry and Physics 93 (2005) 164–169

165

fiber (diameter about 10 ␮m) from Lanzhou Carbon Plant, PR China. 2.2. Preparation and characterization of CB The SiC coating/3D carbon fiber braid was prepared by carbothermic reduction reaction of SiO2 coating/3D carbon fiber braid [9] under a high-purity argon atmosphere at 1823 K, 30 min. In the present work, two kinds of CB, which coating thickness differences was 0.38 and 0.55 ␮m, were prepared. The serial numbers were CB-B and CB-C, respectively. The serial number of A stood for raw braid. The CB was examined by X-ray diffraction (XRD, Damax-2000). The micrographs and chemical composition of CB (through line scan program) were examined by scanning electron microscopy (SEM, JSM-56102Y).

Fig. 1. XRD spectrum of CB.

3. Results

The SEM micrographs of CB (different position of blocks of CB-B and C have the similar micro-pattern) are shown in Fig. 2. Fig. 2A exhibited the fracture of a single coated fiber of CB, and Fig. 2B exhibited the micro-area pattern of CB. It can be seen that coating was uniform and adhered well with fiber surface (Fig. 2B). There was no crack on coating surface and coating/fiber interface (Fig. 2A and B). Thus, in this work, the uniform coating on each fiber of 3D braid was prepared and coating adhered well with fiber surface. The line scan spectrums of CB-B and CB-C are shown in Fig. 3 (scanning line is shown in Fig. 2A). In both spectrums, the content of SiC decreased gradually from the edge of fiber to the core; Contrarily the content of carbon increased gradually from the edge to the core. Thus, the gradient SiC coating on fiber surface was obtained and the thickness was about 0.38 (CB-B) and 0.55 (CB-C) ␮m, respectively. According to the results in Figs. 2 and 3, it was easy to see that the uniform gradient coating (SiC) on each fiber of braid was prepared and coating adhered well with fiber surface.

3.1. Characterization of CB

3.2. Isothermal oxidation properties of CB

The XRD spectrum of CB is shown in Fig. 1. It was obvious that the main component of coating was ␣-SiC. Carbon was amorphous, graphite-like structure, which had the extended 2D molecular chains [10].

The isothermal oxidation-weight loss curves of A, CB-B and CB-C are shown in Fig. 4. It was easy to see that the oxidation process of A (raw braid) began from 673 K; but that of CB was from 873 K; raw braid was burnt out at 873 K,

2.3. The oxidation kinetics and mechanism experiments A muffle furnace was used for the isothermal oxidationweight loss experiment. Static air was employed and the temperature accuracy was estimated as ±5 K. The tests were carried out at temperature ranging from 573 to 1473 K (interval 100 K), 1 h. The weight change was noted by an electricity balance (with a sensitivity of ±0.1 mg). The USATA5000 TG thermal analyzer was employed for the study of oxidation kinetics and mechanism of CB (thermobalance with a sensitivity of ±0.1 mg, at temperature ranging from room temperature to 1273 K under static air, heating rate was 10 K min−1 ).

Fig. 2. The SEM photograph of CB.

166

P. Gao et al. / Materials Chemistry and Physics 93 (2005) 164–169

Fig. 3. Line scan spectrums of CB-B and CB-C.

the quality of CB retained about 26–30% even at 1473 K. This result showed that the oxidation resistance of raw braid was improved clearly. The main reason was that the uniform coating existed on each fiber of braid and coating adhered well with fiber surface. For the coating/carbon materials, the oxidation-weight loss was proportional to time when weight-loss was below 70% [11] (Eq. (1)) mo − m = kt mo

Ea 1 R T

(2)

where A was pre-exponential factor, Ea the apparent activation energy of fiber oxidation reaction in kJ mol−1 , T the absolute temperature in K, R the gas constant in J mol−1 K−1 . Eq. (1) was changed and substituted into Eqs. (2) and (3) was obtained: ln

Serial number A Ea (kJ mol−1 ) r

93.3 0.994

CB-B

CB-C

B1

B2

C1

C2

149.2 0.999

22.45 0.998

175.5 0.997

39.2 0.966

B1, B2, and C1, C2 are oxidation steps.

(1)

where mo was the initial quality of sample, m the quality of sample at time t, k reaction rate constant that was not changed at a fixed temperature. The correlation of reaction rate constant with temperature followed the Arrhenius equation (Eq. (2)): ln k = ln A −

Table 1 The activity energy data of sample A, CB-B and CB-C

mo − m Ea 1 = (ln A + ln t) − mo R T

(3)

In the experimental conditions, the value of (ln A + ln t) was constant. The Arrhenius curve is obatined through plotting ln((mo − m)/mo ) versus 1/T (the data of mo , m, and T were

obtained from Fig. 4 (weight loss curves). And the value of Ea could be obtained from the slope of this curve. The Arrhenius curves of sample A, CB-B and CB-C are shown in Fig. 4, and the Ea value was in Table 1. It was easy to see that both Arrhenius curves of CB-B and CB-C consisted of two different slope lines, thus there were two different Ea values (Ea,CB-B1 , Ea,CB-B2 and Ea,CB-C1 , Ea,CB-C2 ). It meant that there were two different reaction processes with the increase of temperature, as the result in Li’s work [12]. In the first oxidation step, the temperature was low and the amount of active carbon atom (ACA) in carbon fiber surface was small because of the SiC coating existing, thus the reaction rate was low. And the rate that gas diffused through coating was higher than that of reaction. The oxidation process was controlled by chemical reaction. Even through gas diffusion made some contribution to the overall reaction rate (for Ea, CB-B1 is 149.2 kJ mol−1 , Ea, CB-C1 is 175.5 kJ mol−1 ). In the second oxidation step, the oxidation reaction rate increased with the increase of temperature. It was in the vicin-

Fig. 4. The isothermal oxidation-weight loss and Arrhenius curves of samples A, CB-B and CB-C.

P. Gao et al. / Materials Chemistry and Physics 93 (2005) 164–169

167

Fig. 5. TG–DTG curves of A and CB-B.

ity or even exceeding the gas diffusion rate, i.e., the oxidation process of CB was controlled by chemical reaction and gas diffusion together.

Satava–Sestak equation:   0.4567Ea 1 AEa − 2.135 − log[G(a)] = log βR R T

3.3. Non-isothermal oxidation properties of CB

Coats–Redfern equation:   G(a) AR Ea 1 ln 2 = ln − T βEa R T

The TG–DTG curves of A and CB-B are shown in Fig. 5. From TG curves, it was easy to see that the oxidation process of A (raw braid) began from 673 K, but that of CB-B was from 904 K; A was burnt out at 973 K, the quality of CB-B retained about 28% even at 1213 K. We can see that the result was consistent with that of Fig. 4. From DTG curve, it was easy to see that the oxidation reaction rate (da/dT) of CB-B was up to maximum, when the CB-B was burnt off 43.42%. Thus, the characteristic of this reaction was self-catalytic, according to [13]. Initially, the amount of active carbon atom (ACA) on carbon fiber surface was small since the SiC gradient coating existed. The CB had the lowest oxidation rate. As the extended 2D chain structure of carbon fiber should be broken off by the oxidation to form several smaller species when O2 diffused through the coating and reacted with carbon, thereby increasing the ACA [10]. Thus, with the increasing of oxidation-weight loss, the amount of ACA increased, the reaction rate also Increased. So the self-catalytic characteristic was displayed. But there was a critical point, when the value of weight loss was bigger than 43.42%, the amount of species decreased with the increase of weight loss. And the oxidation rate decreased. 3.4. Study of oxidation kinetics and mechanism of CB In the present work, the equations of Achar–Brindley– Sharp–Wendworth [14,15], Satava–Sestak [16] and Coats–Redfern [17] were used to study the oxidation kinetics and mechanism of coated braid in non-isothermal condition. Achar–Brindley–Sharp–Wendworth equation:
   1 da A Ea 1 ln = ln − (4) F (a) dT β R T

(5)

(6)

where β was the heating rate, a the reaction fraction, F(a) and G(a) the differential and integral mechanism function, respectively. The basic data of a, T and da/dT obtained by the TG–DTG curves are listed in Table 2. The integral (G(a)) and differential (F(a)) functions for the most common mechanism used in kinetic study of solid-state decomposition in this work are listed in Table 3 [18–20]. The kinetic analyses were completed by the linear leastsquares method on a computer. According to Eqs. (4)–(6), the values of linear correlation coefficient r and standard deviation S.D. r were obtained from ln[(da/dT)/F(a)], log[G(a)] Table 2 Oxidation decomposition data of CB-B from TG and DTG curves T (K)

a

da/dT (K−1 )

1063.03 1068.01 1072.97 1077.95 1082.99 1087.96 1092.92 1097.89 1103.07 1108.04 1113.04 1118.06 1123.03 1127.99 1132.94 1138.12 1142.94 1148.10 1153.03 1158.03 1162.97

0.346 0.366 0.387 0.411 0.438 0.467 0.499 0.534 0.573 0.613 0.655 0.698 0.740 0.781 0.821 0.860 0.892 0.923 0.948 0.969 0.985

0.368 0.410 0.458 0.505 0.560 0.617 0.670 0.728 0.782 0.823 0.847 0.850 0.838 0.816 0.777 0.714 0.641 0.553 0.464 0.369 0.279

168

P. Gao et al. / Materials Chemistry and Physics 93 (2005) 164–169

Table 3 Mechanism function of thermal decomposition kinetic equation [18–20] Name of function

Type of mechanism

Form of function

No.

F(a)

G(a)

Avrami-Erofeev equatipon

Random nucleation

n=1 n=2 n=3

(1 − a) 2(1 − a) [−ln(1 − a)]1/2 3(1 − a) [−ln(1 − a)]2/3

−ln(1 − a) [−ln(1 − a)]1/2 [−ln(1 − a)]1/3

Parabola law Valensi equation Jander equation Anti–Jander equation G–B equationa Z–L–T equationb

D diffusion, decelerator a–t curve 2D diffusion, decelerator a–t curve 3D diffusion, decelerator a–t curve 3D diffusion 3D diffusion, spherical symmetry 3D diffusion

1/(2a) [−ln(1 − a)]−1 1.5(1 − a)2/3 [1 − (1 − a)1/3 ]−1 1.5(1 + a)2/3 [(1 + a)1/3 − 1]−1 1.5 [(1 − a)−1/3 − 1]−1 1.5(1 − a)4/3 [(1 − a)−1/3 − 1]−1

a2 a + (1 − a)ln(1 − a) [1 − (1 − a)1/3 ]2 [(1 + a)1/3 − 1]2 [1 − 2a/3]−(1 − a)2/3 [(1 − a)−1/3 − 1]2

D1 D2 D3 D4 D5 D6

Phase boundary reaction

Decelerator a–t curve

Cylindrical symmetry Spherical symmetry

(1 − a)1/2 1 − (1 − a)2/3

2[1 − (1 − a)1/2 ] 3[1 − (1 − a)1/3 ]

Reaction order

n=2 n=3 n=4

0.5(1 − a)−1 1/3(1 − a)−2 1/4(1 − a)−3

1 − (1 − a)2 1 − (1 − a)3 1 − (1 − a)4

R2 R3 R4

Chemical reaction

n = 2 decelerator a–t curve

(1 − a)2

(1 − a)−1

F2

n=2 n=3 n=4

2a1/2

a1/2

3a2/3 4a3/4

a1/3 a1/4

M2 M3 M4

Mampl power

a b

P2 P3

Ginstling–Brounstein equation. Zhuralev–Lesokin–Tempelman equation. Table 4 (Continued )

Table 4 Kinetic parameters obtained from different treatment methods Ea (kJ mol−1 )

No.

log A (min−1 )

r

S.D.

Achar–Brindly–Sharp A1 257.753 A2 176.828 A3 149.875 D1 253.361 D2 302.500 D3 362.127 D4 219.348 D5 323.092 D6 479.234 P2 199.200 P3 13.575 R2 23.539 R3 −93.567 R4 −210.673 F2 374.860 M4 56.110 M3 65.504 M2 84.289

11.0351 7.021 5.608 10.399 12.577 14.919 7.670 12.958 20.804 8.093 −0.348 −0.434 −6.143 −11.903 16.920 0.837 1.374 2.375

−0.983 −0.986 −0.987 −0.996 −0.996 −0.989 −0.994 −0.994 −0.972 −0.996 −0.357 −0.263 −0.524 −0.694 −0.957 −0.899 −0.926 −0.957

0.231 0.146 0.118 0.106 0.124 0.255 0.114 0.163 0.561 0.086 0.171 0.416 0.732 1.050 0.545 0.131 0.128 0.122

Satava–Sestak A1 153.913 A2 76.956 A3 51.304 D1 214.369 D2 240.166 D3 272.748 D4 191.617 D5 250.889 D6 346.989 P2 128.361 P3 136.374 R2 75.761

6.294 2.882 1.820 8.619 9.644 10.663 6.468 9.541 14.482 5.020 5.419 2.738

−0.977 −0.977 −0.977 −0.994 −0.990 −0.984 −0.996 −0.988 −0.968 −0.987 −0.98421 −0.998

0.074 0.037 0.025 0.052 0.074 0.108 0.039 0.084 0.196 0.045 0.0539 0.011

No.

A1 A2 A3

R3 R4 F2 M4 M3 M2

Ea (kJ mol−1 ) 54.787 40.443 111.361 26.796 35.728 53.592

Coats–Redfern A1 143.577 A2 62.6511 A3 35.675 D1 207.152 D2 234.281 D3 268.543 D4 183.226 D5 245.556 D6 346.615 P2 116.707 P3 125.134 R2 61.394 R3 39.338 R4 24.254 F2 98.831 M4 9.902 M3 19.295 M2 38.081

log A (min−1 )

r

S.D.

1.912 1.372 4.893 0.864 1.151 1.800

−0.993 −0.984 −0.870 −0.994 −0.994 −0.994

0.014 0.016 0.138 0.006 0.009 0.013

8.713 4.639 3.157 11.341 12.468 13.602 9.088 12.404 17.636 7.270 7.725 4.80 3.320 2.562 7.009 2.638 2.936 2.183

−0.971 −0.964 −0.952 −0.993 −0.989 −0.982 −0.994 −0.986 −0.965 −0.984 −0.980 −0.996 −0.984 −0.950 −0.832 −0.963 −0.981 −0.988

0.168 0.084 0.055 0.117 0.168 0.247 0.089 0.193 0.451 0.103 0.122 0.025 0.034 0.038 0.317 0.013 0.018 0.028

and ln[G(a)/T2 ] versus 1/T lines directly, the values of Ea and A can be calculated from the slope and intercept of above lines, respectively. The results (Ea , A, r and S.D.) of ln[(da/dT)/F(a)], log[G(a)], ln[G(a)/T2 ] versus 1/T for all possible mechanisms (F(a) and G(a)) are listed in Table 4.

P. Gao et al. / Materials Chemistry and Physics 93 (2005) 164–169

If the following conditions: (1) The values of Ea were between 80 and 250 kJ mol−1 and the values of log A were between 7 and 30 min−1 , which were obtained by the different methods were approximately equal; (2) The linear relevant coefficient r was better (<−0.98); (3) The standard deviation S.D. was small (<0.3). were all satisfied at the same time, it can be concluded that the relevant function was the probable mechanism of oxidation reaction [21,22]. From the data in Table 4, it can be seen clearly that only function D1 satisfied the above-mentioned conditions. Thus the possible mechanism function was function D1 (D1 stood for 1D diffusion, decelerator a–t curve). We conclude that the kinetic equation for the thermal decomposition of CB was da/dT = (A/β)exp(−Ea /RT)(0.5a−1 ). The kinetic parameters were: log A = 10.12 min−1 and Ea = 224.96 kJ mol−1 .

4. Discussion The oxidation mechanism of CB-B was 1D diffusion, decelerator a–t curve. The decelerator a–t curve’ meant that the reaction process was diffusion-controlled mechanism. There was some ACA, which could react with oxygen, existed in coating/fiber interface. When the oxygen diffused through coating and reacted with ACA, one ACA would be consumed. But the extended 2D chain structure of fiber should be broken off by the oxidation to form several smaller species. Every species has become a new ACA. Namely with the increase of oxidation, more and more new ACA were produced, the oxidation rate of fiber increased. The self-catalytic characteristic was observed. As we knew, the amount of new produced ACA was determined by the length of 2D molecular chain and fiber surface area, which all decreased with the increase of oxidation rate. But with the increase of oxidation rate, more and more ACA s was consumed. Namely with the increase of oxidation rate, the amount of new produced ACA decreased and more ACA were consumed. Thus, when the amount of consumed was equal to that of produced ACA. The reaction rate was up to maximum (weight loss = 43.42%). When the weight loss was bigger than 43.42%, the amount of consumed ACA by oxidation was bigger than that of new produced. The oxidation rate decreased with the increase of weight loss. In the whole oxidation process, the rate that carbon and O2 reaction was higher than that of gas diffusion. Thus the reaction process was diffusion-controlled mechanism. In a word, the self-catalytic characteristic of coated braid oxidation reaction could be explained through the oxidation mechanism (1D diffusion, decelerator a–t curve)

169

5. Conclusion 1. A uniform SiC coating on each fiber of 3D braid was prepared and the coating adhered well with fiber surface. The coating composition was gradient change and the thickness was 0.38 ␮m (CB-B) and 0.55 ␮m (CB-C), respectively. 2. The oxidation process of CB in isothermal oxidation condition was reaction-controlled in the first step, gas diffusion and reaction-controlled in the second step. 3. In non-isothermal oxidation condition, the oxidation process of CB exhibited a self-catalytic characteristic; the reaction rate was up to maximum when the CB was burnt off 43.42%. 4. The oxidation mechanism of the CB was 1D diffusion, decelerator a–t curve. The kinetic parameters were: log A (min−1 ) = 10.12 and Ea (kJ mol−1 ) = 224.96.

Acknowledgement This work was supported by the National Natural Science Foundation of PR China (No. 90305001).

References [1] Y. Xu, L. Cheng, L. Zhong, H. Yin, X. Yin, Mater. Sci. Eng. A (300) (2001) 196–202. [2] T. Shimoo, K. Okamura, T. Akizuk, M. Takmure, J. Mater. Sci. 30 (1995) 3387–3394. [3] F. Lamouroux, S. Bertrand, R. Pailler, R. Naslain, M. Cataldi, Compos. Sci. Technol. 59 (1999) 1073–1085. [4] T.A. Bullions, J.E. McGrath, A.C. Loos, Compos. Sci. Technol. 63 (2003) 1737–1748. [5] S. Lu, B. Rand, K.D. Bartle, J. Mater. Sci. 34 (1999) 571–578. [6] Y. Deslandes, F.N. Sabir, J. Mater. Sci. Lett. 9 (1990) 200–202. [7] S. Chu, H. Wang, G. He, J. Inorg. Mater. 8 (3) (1993) 327– 333. ¨ C.R.M. Silva, J.C. Bressiani, J. Eur. [8] C.A.A. Cairo, M.L.A. GracEa, Ceram. Soc. 21 (2001) 325–329. [9] P. Gao, H. Wang, Z. Jin, J. Inorg. Mater. 18 (4) (2003) 849– 854. [10] P. Gao, H. Wang, Z. Jin, Study of oxidation properties and decomposition kinetics of three-dimensional (3D) braided carbon fiber, Thermochim. Acta 414 (2004) 59–63. [11] R. Luo, J. Cheng, T. Wang, Carbon 40 (2002) 1965–1972. [12] H. Li, X. Zeng, X. Zhu, Carbon (China) 3 (1999) 2–6. [13] L.R. Zhao, B.Z. Jang, J. Mater. Sci. 32 (1997) 2811–2819. [14] B.N. Achar, G.W. Brindley, J.H. Sharp, Proc. Int. Clay Conf., Jerusalem 1 (1996) 67. [15] J.H. Sharp, S.A. Wendworth, Anal. Chem. 41 (14) (1969) 2060–2062. [16] V. Satava, J. Sestak, J. Therm. Anal. 8 (3) (1975) 477–489. [17] A.W. Coats, J.P. Redfern, Nature 201 (4914) (1964) 68–69. [18] X. Gao, D. Dollimore, Thermochem. Acta 215 (1993) 47–55. [19] M.A. Gabal, Thermochem. Acta 40 (2003) 199–208. [20] R.M. Mahfouz, M.A.S. Monshi, N.M. Abd El-Salam, Thermochem. Acta 383 (2002) 95–101. [21] T. Zhang, R. Hu, F. Li, Thermochem. Acta 244 (1994) 177– 184. [22] R. Hu, Z. Yang, Thermochem. Acta 123 (1988) 135–142.