Mechanism Identification and Kinetics Analysis of Thermal Degradation for Carbon Fiber/Epoxy Resin

For carbon fiber epoxy resin used in aerostructure, thermal degradation mechanism and kinetics play an important role in the evaluation of thermal response and combustion characteristics. However, the thermal decomposition process and mechanism are difficult to unify strictly due to the complexity of the components from different suppliers. In the present study, a product of carbon fiber epoxy resin made by AVIC (Aviation Industry Corporation of China) composite corporation is examined to identify its thermal degradation mechanism and pyrolysis products by measurements, including simultaneous thermal analysis, Fourier transform infrared spectroscopy and mass spectrometry, establish the kinetic model by Kissinger/Friedman/Ozawa/Coats-Redfern methods. The results show thermal degradation occurs in three steps under the inert atmosphere, but in four steps under air atmosphere, respectively. The first two steps in both environments are almost the same, including drying, carbon dioxide escape and decomposition of the epoxy resin. In the third step of inert atmosphere, phenol is formed, methane decreases, carbon monoxide basically disappears and carbon dioxide production increases. However, in air, thermal oxidation of the carbonaceous residues and intermolecular carbonization are observed. Furthermore, thermal degradation reaction mechanism submits to the F4 model. These results provide fundamental and comprehensive support for the application of carbon fiber epoxy resin in aircraft industry.


Introduction
Attributed to high resistance to corrosion, high specific strength and stiffness of FRP (Fiber Reinforced Polymer) material, its utilization in infrastructures and manufacturing is more and more popular. For example, concrete slabs with GFRP(Glass Fiber Reinforced Polymer) bars and hallow composite reinforcing system [1] and a novel prefabricated FRP composite jacket with an easy-fit and self-locking mechanical joining system [2] have been developed. In addition, carbon fiber epoxy resin has been widely used in the design and manufacture of aircraft load-bearing and functional structures. However, thermal behavior of FRP is completely different from that of aluminum, because it can soften, decompose and burn to release fumes, especially in fire, leading to threaten the safety of occupants. As a result, both the Federal Aviation Administration (FAA) and the European Aviation Safety Agency (EASA) require us to consider fire protection, flame retardancy, combustion product toxicity, and other thermal issues of composite aircraft structures in airworthiness certification [3]. For this reason, the generic material combustion computational model, ThermaKin [4][5][6][7] and a toxicity assessment model [8][9][10] have been developed by FAA. In addition, Tranchard [11] developed a material thermal response model based on SEMCEF for fire environments and validated it with the T700/M21 composite material used in Airbus 350. Li [12,13] also developed a thermal response model for composites in the fire environment based on ABAQUS.
Thermal response and combustion characteristics of carbon fiber epoxy resin in the fire environment mainly depends on the thermal decomposition of resin matrix. For this and thorough insight into thermal behavior of CFRP used in aerostructures is presented, and detailed data can support thermal response modeling, combustion modeling and toxicity assessment of material in fire.

Materials and Methods
Samples of carbon fiber reinforced epoxy resin matrix composite material used for aerostructures were provided by AVIC composite corporation (Beijing, China), in which matrix is mainly DGEBA. The preparation process was prepreg cutting-lay-up-autoclave molding-machining. The size of sample was 4 mm × 4 mm × 1.2 mm with the lamina thickness of prepreg of 0.15 mm, shown as Figure 1. Stacking sequence is [0/45/90/135]s with the fiber volume of 54% and the density of 1.55 g/cm 3 .
Thermal analysis experiments were performed using the simultaneous thermal analyzer (NETZSCH STA 449 F3, NETZSCH, Selb, Germany), FTIR spectrometer (BRUKER VERTEX 80, BRUKER,Billerica, MA, USA) and mass spectrometry (NETZSCH QMS403C, NETZSCH, Selb, Germany) to investigate the mass changes (TG, Thermogravimetric), derivative of mass changes (DTG, Derivative Thermogravimetric) and gas composition (FTIR/MS), referring to the test standards [37][38][39][40]. The sample used in the simultaneous thermal analyzer was heated from room temperature to 1100 °C at the heating rates of 5, 10, 20 and 40 °C/min, under the atmosphere of air and inert gas, respectively, with the purge flow rate set to 20 mL/min. The primary aim of the TG analyses was to know the thermal degradation characteristics of materials, and to obtain adequate data to perform kinetic analysis in order to obtain the Arrhenius parameters. IR spectra were recorded in the spectral range of 4500-650 cm −1 with a resolution of 4 cm −1 and an average number of scans of 16. The intensities of 17 ions (m/z = 14, 16,17,18,26,28,30,42,44,46, 56, 60, 64 and 94) were determined according to the previous work [16,41].  Thermal analysis experiments were performed using the simultaneous thermal analyzer (NETZSCH STA 449 F3, NETZSCH, Selb, Germany), FTIR spectrometer (BRUKER VERTEX 80, BRUKER, Billerica, MA, USA) and mass spectrometry (NETZSCH QMS403C, NETZSCH, Selb, Germany) to investigate the mass changes (TG, Thermogravimetric), derivative of mass changes (DTG, Derivative Thermogravimetric) and gas composition (FTIR/MS), referring to the test standards [37][38][39][40]. The sample used in the simultaneous thermal analyzer was heated from room temperature to 1100 • C at the heating rates of 5, 10, 20 and 40 • C/min, under the atmosphere of air and inert gas, respectively, with the purge flow rate set to 20 mL/min. The primary aim of the TG analyses was to know the thermal degradation characteristics of materials, and to obtain adequate data to perform kinetic analysis in order to obtain the Arrhenius parameters. IR spectra were recorded in the spectral range of 4500-650 cm −1 with a resolution of 4 cm −1 and an average number of scans of 16. The intensities of 17 ions (m/z = 14, 16,17,18,26,28,30,42,44,46,56,60,64 and 94) were determined according to the previous work [16,41].  Figure 2 shows the TG and DTG curves of carbon fiber reinforced epoxy resin matrix composite heated from room temperature to 1100 • C at 20 • C/min under air and inert atmosphere, respectively. Under inert atmosphere, the TG curve shows three mass loss stages, corresponding to three DTG curve peaks. The first stage of weight loss ranged from room temperature to 286 • C with the mass loss of 0.69%, and the maximum weight loss rate was 0.17%/min with the peak temperature of 244 • C. At this stage, physical escape of lowmolecular-weight gases occurs, with the drying effect caused by the temperature rise from room temperature, where water was evaporated rapidly, justified by mass spectrometry in Section 3.3.2. The same drying process is also observed in thermos-gravimetric analysis of epoxy resins by Ahamad [28] and Shao [42].  Figure 3 shows the TG and DTG curves at different heating rates under iner atmosphere. Two inflection points appeared in the TG curves for different heati under inert atmosphere, which corresponded to the two peaks of the DTG curve temperature rise rate increases, the weight loss curve shifts to the higher temper gion [46]. However, in air atmosphere, the effect of the temperature rise rate on loss behavior appears more complex due to the oxidative decomposition of carb The lower the temperature rise rate is, the longer the time is to reach the same t ture, and so the more sufficient the reaction is. As a result, the two lower heating and 10 °C/min) make the temperature at which the weight loss rate reached the pe lower, so that two more peaks could be seen below 1000 °C. However, under the rates of 20 and 40 °C/min, the degradation was not completed within 1100 °C corresponding peaks were observed.  The second stage ranged from 286 to 515 • C with the mass loss of 23.29%, where the maximum weight loss rate was 4.99%/min at the peak temperature of 410 • C. This range is close to the pyrolysis interval of epoxy resin, where the ether bond is broken to generate BPA (Bisphenol-A), and the phenol comes from the further decomposition of BPA and/or the breakage of the C-C bond connecting the benzene ring after the breakage of unilateral ether bond on the epoxy resin [43], which is an obvious degradation behavior of the resin matrix. The weight loss of 4.54% in the third stage occurred in a wide temperature range from 515 to 1100 • C, which was relatively gentle, and the maximum loss rate was 0.64%/min corresponding to the peak temperature of 537.4 • C. According to the literature [44], the pyrolysis temperature of BPA epoxy resin is higher than 660 • C, so this stage was considered as further pyrolysis of the polymer [28].

Kinetic Model
Under the air atmosphere, the specimen was almost completely degraded with the mass loss of 90.64% reached at 1100 • C. The existence of four peaks in the DTG curve indicated that the weight loss process was divided into four stages, and the peak temperature was shown in Figure 1. The characteristics of the first weight loss stage in air atmosphere was almost consistent with that in inert gas, which was a typical drying stage. The peak temperature in the second stage was only slightly lower, but the pyrolysis interval was also similar to that under the inert atmosphere, indicating that the degradation of the resin matrix also occurred at this stage. Therefore, it showed that the reaction atmosphere has less influence on the degradation reaction of the resin matrix. However, the percentage of weight loss and weight loss rate in the inert atmosphere were greater.
The temperature range of the third stage was from 497 to 710 • C with the mass loss of 18.38%, the maximum weight loss rate of 2.53%/min and the peak temperature of 684 • C.
Thermal oxidation of intermediate transient carbonaceous residues [16], which is a process of intermolecular cross-linking, carbonization, and further removal of small molecules, occurred at this stage [29]. The weight loss in the fourth stage ranged from 710 to 1100 • C with the mass loss of 52.65%; the maximum weight loss rate was 2.55%/min, and the peak temperature was 745.5 • C. This was due to the reaction of the carbon fiber reinforced material with oxygen and the nearly complete oxidative decomposition of the carbon fiber caused by prolonged heating, which was supported by literature [45]. Figure 3 shows the TG and DTG curves at different heating rates under inert and air atmosphere. Two inflection points appeared in the TG curves for different heating rates under inert atmosphere, which corresponded to the two peaks of the DTG curve. As the temperature rise rate increases, the weight loss curve shifts to the higher temperature region [46]. However, in air atmosphere, the effect of the temperature rise rate on weightloss behavior appears more complex due to the oxidative decomposition of carbon fiber. The lower the temperature rise rate is, the longer the time is to reach the same temperature, and so the more sufficient the reaction is. As a result, the two lower heating rates (5 and 10 • C/min) make the temperature at which the weight loss rate reached the peak value lower, so that two more peaks could be seen below 1000 • C. However, under the heating rates of 20 and 40 • C/min, the degradation was not completed within 1100 • C, and no corresponding peaks were observed.  Figure 3 shows the TG and DTG curves at different heating rates under inert and air atmosphere. Two inflection points appeared in the TG curves for different heating rates under inert atmosphere, which corresponded to the two peaks of the DTG curve. As the temperature rise rate increases, the weight loss curve shifts to the higher temperature region [46]. However, in air atmosphere, the effect of the temperature rise rate on weightloss behavior appears more complex due to the oxidative decomposition of carbon fiber. The lower the temperature rise rate is, the longer the time is to reach the same temperature, and so the more sufficient the reaction is. As a result, the two lower heating rates (5 and 10 °C/min) make the temperature at which the weight loss rate reached the peak value lower, so that two more peaks could be seen below 1000 °C. However, under the heating rates of 20 and 40 °C/min, the degradation was not completed within 1100 °C, and no corresponding peaks were observed. The different methods for solving the kinetic constants of reactions are based on the same basic theoretical formula derived from the Arrhenius formula, as shown in the following equation: The different methods for solving the kinetic constants of reactions are based on the same basic theoretical formula derived from the Arrhenius formula, as shown in the following equation:

Kinetic Model
, M e are the initial mass, the mass of the specimen at the current moment, and the final mass at the end of the reaction, respectively. f (α) is a function of the kinetic mechanism, which is a function of the conversion degree, depending on the specific degradation mechanism. R is the ideal gas constant, with a value of 8.314 J/(mol K).
In this paper, the kinetic parameters of the reaction were solved by the Kissinger [47] differential method, the Friedman [20] differential method and the Ozawa [22] integral method, respectively. These three methods did not require the degradation reaction mechanism of the material to obtain the activation energy. As to calculate the reaction order, it can be assumed that the mechanism function can be expressed as follows: f (α) = (1 − α) n . The Kissinger method [47] assumes the maximum reaction rate at the peak temperature. The Friedman method [20] assumes that the reaction kinetic constant A, E A and n are independent of the heating rate. While the Ozawa method [22] assumes that the conversion degree at the peak temperature at different heating rates is a constant (iso-conversion method). Additionally, the Coats-Redfern method was used to determine the reaction mechanism function. The data obtained under inert atmosphere are shown in the following Table 1. Differentiation on both sides of (1) yields the following equation: where T m is the peak temperature from the DTG curve, α m , is the peak conversion corresponding to T m , which is obtained from the TG curve, and β = dT dt is the constant heating rate during the experiment. Kissinger considers n(1 − α m ) n−1 is independent of β and its value is approximately equal to 1.
The activation energy can be obtained from the above equation by logarithm and derivation: The thermal weight loss process of carbon fiber reinforced epoxy resin composites under inert atmosphere has two main stages. As shown in Figure 4, the intercept and slope data are obtained.
ing rate during the experiment. Kissinger considers (1 ) m n α − is independent of β and its value is approximately equal to 1.
The activation energy can be obtained from the above equation by logarithm and derivation: The thermal weight loss process of carbon fiber reinforced epoxy resin composites under inert atmosphere has two main stages. As shown in Figure 4, the intercept and slope data are obtained. The relationship between the pre-exponential factor A, and the slope and intercept of the curve is as follows: When calculating the reaction order n, consider the mechanism function to be ex- The relationship between the pre-exponential factor A, and the slope and intercept of the curve is as follows: When calculating the reaction order n, consider the mechanism function to be expressed as f (α) = (1 − α) n and n is calculated by the following equation: The results are shown in the Table 2 below: Table 2. Kinetic parameters by Kissinger method.

Friedman Method
Taking the logarithm of (1) gives the following equation where ln β dα dT = ln dα dt . The relationship between ln dα dt and 1/T at different conversion degrees was calculated as shown in Figure 5.
The relationship between ln   Figure 6 shows a line chart of the activation energy with conversion degree, which illustrates that the activation energy is not a constant, but changes as the thermal decomposition reaction proceeds and the conversion degree increases, showing an overall upward trend. Therefore, the thermal decomposition reaction of carbon fiber reinforced epoxy composite under inert gas is a multi-step. Furthermore, the predominant factor was found to be 1.34287 × 10 14 .  Figure 6 shows a line chart of the activation energy with conversion degree, which illustrates that the activation energy is not a constant, but changes as the thermal decomposition reaction proceeds and the conversion degree increases, showing an overall upward trend. Therefore, the thermal decomposition reaction of carbon fiber reinforced epoxy composite under inert gas is a multi-step. Furthermore, the predominant factor was found to be 1.34287 × 10 14 .

Ozawa Method
The activation energy EA was calculated by the following formula: The relationship between log β and 1/T was calculated for different conversion degrees as shown in the following Figure 7:

Ozawa Method
The activation energy E A was calculated by the following formula: The relationship between log β and 1/T was calculated for different conversion degrees as shown in the following Figure 7: Polymers 2021, 13, 569 The relationship between log β and 1/T was calculated for different conversion degrees as shown in the following Figure 7:  The pre-exponential factor was calculated by the following formula [48,49]: At the same conversion degree, yielding the values of the corresponding pre-exponential factor for different heating rates, as shown in Figure 8. And the mean values were finally obtained. The activation energies obtained by Ozawa's method were similar to those obtained by Friedman's method, which verified the accuracy of the results. The pre-exponential factor was calculated by the following formula [48,49]: At the same conversion degree, yielding the values of the corresponding pre-exponential factor for different heating rates, as shown in Figure 8. And the mean values were finally obtained. The activation energies obtained by Ozawa's method were similar to those obtained by Friedman's method, which verified the accuracy of the results.

Coats-Redfern Method
Coats and Redfern assumed the function of the kinetic mechanism is ( ) ( ) 1 n f α α = − , and derived the following expression: where α in the logarithmic term of the left equation is the result of integrating ( ) 1/f α to α and then ignoring higher-order terms under the condition of low conversion rate after polynomial expansion. In order to study the kinetic model of thermal decomposition, assuming that the mechanism function is unknown, the alternative term to α in the logarithmic term is the integral form ( ) F α of ( )

1/f α , and considering that 2
A RT E approaches zero, the formula is as follows: Different mechanism models were established by this method, as shown in Table 3. Table 3. Thirteen kinds of commonly used pyrolysis kinetics models.

Coats-Redfern Method
Coats and Redfern assumed the function of the kinetic mechanism is f (α) = (1 − α) n , and derived the following expression: where α in the logarithmic term of the left equation is the result of integrating 1/ f (α) to α and then ignoring higher-order terms under the condition of low conversion rate after polynomial expansion. In order to study the kinetic model of thermal decomposition, assuming that the mechanism function is unknown, the alternative term to α in the logarithmic term is the integral form F(α) of 1/ f (α), and considering that 2RT E A approaches zero, the formula is as follows: Different mechanism models were established by this method, as shown in Table 3. Table 3. Thirteen kinds of commonly used pyrolysis kinetics models.

F(α) Reaction Mechanism
One-dimensional random nuclear reaction Avrami random nucleation n = 4 Phase boundary reaction, cylindrical symmetry Phase boundary reaction, spherical symmetry Three-dimensional diffusion, spherical symmetry, Ginstling-Brounshtein equation The ln F(α) T 2 − 1/T curves for the conversion rates from 0.2 to 0.7 at different heating rates were made to obtain the activation energy values. Pearson's correlation coefficients were calculated to measure the correlation between the variables, as shown in Table 4. Comparing the activation energy values in Table 4 with those calculated by the three kinetic methods in Table 1. The average E A of F 4 is the closest to the previous calculation, which was 211.691 kJ/mol. It is concluded that the pyrolysis reaction of carbon fiber reinforced epoxy resin composite in the inert atmosphere is the closest to that predicted by the F 4 model. The expression of the mechanism function is: (1 − α) 4 . After averaging the activation energies and pre-exponential factors of the four heating rates, the reaction mechanism model was obtained as follows

FTIR Analysis
The Gram-Schmidt spectrum under inert atmosphere in Figure 2 represented the FTIR absorption intensity of the escaping gas during pyrolysis of the material, showing that there were only two peaks at 425 • C and 557 • C, respectively, which were delayed compared to the corresponding DTG curve. This was due to the time delay between gas detection and gas generation in the FTIR spectrometer. The larger peak at 425 • C indicated that the amount of gas escaping at this stage was higher, and had a higher infrared extinction coefficient. Figure 9 shows the infrared spectral data of the gas at the characteristic temperatures determined by the DTG curve. At the characteristic temperature of 244 • C, which was the first stage of mass loss, the products were carbon dioxide (CO 2 : 2300-2400 cm −1 ) and water (H 2 O: 3500-4000 cm −1 ). The same phenomenon is observed by Ahamad [28] in the study of urea-formaldehyde/epoxy resin blends. Thus, this stage was not only a drying stage, but also a physical escape of small molecule gases. Obvious degradation behavior of the resin matrix occurred in the second stage, corresponding to the infrared spectral data at 410 • C. Since the stretching vibration of saturated hydrocarbon C-H is usually close to the frequency absorption of 3000 cm −1 , it is speculated that there was methane gas (CH 4 ) according to the three peaks in the range of 2900-3170 cm −1 : 3045 cm −1 , 3016 cm −1 and 2970 cm −1 . In addition to water and carbon dioxide, carbon monoxide (CO: 2000-2200 cm −1 with double peaks at 2048 cm −1 and 2070 cm −1 ) and phenol (main peaks of C 6 H 5 OH: 3650, 1608, 1510, 1260 and 1176 cm −1 ) were also observed. The above products were considered to be the chain scission and depolymerization reactions of BPA epoxy resins [31], where CO and CO 2 are mainly derived from the ether group (R-O-R'), methyl group (-CH 3 ), methylene group (-CH 2 ) and other carbon-containing functional groups. At 537 • C, carbon monoxide almost disappeared and turned into carbon dioxide, while the production of the other gas products decreased compared with that at 410 • C.   Figure 9. Spectra collected at 244, 410 and 537 °C of the degradation products of composites released during TGA (20 °C/min-inert atmosphere).

MS Analysis
The following gas molecules were detected by mass spectrometry at a heating rate of According to the order of magnitude of ion current intensity, the classification analysis was carried out from high to low. As shown in Figure 10, the left ordinate shows the mass spectra of gas products with a mass-to-charge ratio of 28, which was supposed to be CO and/or C2H4. Additionally, there were three peaks in the curve, which respectively occurred at 125 °C, 437 °C and 920 °C. The peak with the largest ion current intensity appeared between 300 and 500 °C, corresponding to the main decomposition peak in the DTG analysis, which was presumed to be the precipitation of CO gas, also observed by Tranchard [16] and Ahamad [28] and related to the certain amount of oxygen-containing groups in the polymer. The mass spectrum of phenol with the mass-to-charge ratio of 94
According to the order of magnitude of ion current intensity, the classification analysis was carried out from high to low. As shown in Figure 10, the left ordinate shows the mass spectra of gas products with a mass-to-charge ratio of 28, which was supposed to be CO and/or C 2 H 4 . Additionally, there were three peaks in the curve, which respectively occurred at 125 • C, 437 • C and 920 • C. The peak with the largest ion current intensity appeared between 300 and 500 • C, corresponding to the main decomposition peak in the DTG analysis, which was presumed to be the precipitation of CO gas, also observed by Tranchard [16] and Ahamad [28] and related to the certain amount of oxygen-containing groups in the polymer. The mass spectrum of phenol with the mass-to-charge ratio of 94 was shown on the right ordinate. Its precipitation range was 300-750 • C, which covered the second and third mass loss stages, also corresponded to the peaks of infrared spectrum at 410 • C and 537 • C. For the detection result of water molecules in Figure 11, a trough appeared before 250 °C, which corresponded to the first peak in the DTG curve, verifying that this was the drying stage. In addition, due to the pyrolysis reaction of the resin, the peak of water molecules appeared at 406 °C, which was ahead of the precipitation of CO, indicating that the decomposition of the epoxy resin would preferentially produce water molecules, but mainly precipitated CO. At this stage, the water produced by the thermal decomposition of epoxy resin mainly come from the breaking of the bonds of -OH functional groups.
For the gas mass-to-charge ratio of 44, CO2, N2O and C3H8 were considered. However, the peak of the curve occurred at 410 °C, ahead of 435 °C, which was the precipitation temperature of the CO peak. Considering the sequence of reactions, CO mainly comes from the reforming of the C-O group in the bisphenol A epoxy resin, and then combined with oxygen ions to form CO2. Therefore, it was speculated that the m/z = 44 mass spectrum curve was mainly represented by the C3H8 gas molecule. Since the reaction was carried out under an inert atmosphere, oxygen ions were supplied only by the material itself, which was also verified in the order of magnitude. It was speculated that the relatively small increasement of the amplitude of this curve afterwards might correspond to the formation of CO2.
The peak temperature of C2H2 gas molecules was close to the temperature of the third peak in the DTG curve, which was supposed to be the product of the second step of decomposition. It can be verified that the decomposition of carbon fiber reinforced epoxy resin matrix composite under inert atmosphere exhibits at least two steps (a major step and a secondary step).  For the detection result of water molecules in Figure 11, a trough appeared before 250 • C, which corresponded to the first peak in the DTG curve, verifying that this was the drying stage. In addition, due to the pyrolysis reaction of the resin, the peak of water molecules appeared at 406 • C, which was ahead of the precipitation of CO, indicating that the decomposition of the epoxy resin would preferentially produce water molecules, but mainly precipitated CO. At this stage, the water produced by the thermal decomposition of epoxy resin mainly come from the breaking of the bonds of -OH functional groups. For the detection result of water molecules in Figure 11, a trough appeared before 250 °C, which corresponded to the first peak in the DTG curve, verifying that this was the drying stage. In addition, due to the pyrolysis reaction of the resin, the peak of water molecules appeared at 406 °C, which was ahead of the precipitation of CO, indicating that the decomposition of the epoxy resin would preferentially produce water molecules, but mainly precipitated CO. At this stage, the water produced by the thermal decomposition of epoxy resin mainly come from the breaking of the bonds of -OH functional groups.
For the gas mass-to-charge ratio of 44, CO2, N2O and C3H8 were considered. However, the peak of the curve occurred at 410 °C, ahead of 435 °C, which was the precipitation temperature of the CO peak. Considering the sequence of reactions, CO mainly comes from the reforming of the C-O group in the bisphenol A epoxy resin, and then combined with oxygen ions to form CO2. Therefore, it was speculated that the m/z = 44 mass spectrum curve was mainly represented by the C3H8 gas molecule. Since the reaction was carried out under an inert atmosphere, oxygen ions were supplied only by the material itself, which was also verified in the order of magnitude. It was speculated that the relatively small increasement of the amplitude of this curve afterwards might correspond to the formation of CO2.
The peak temperature of C2H2 gas molecules was close to the temperature of the third peak in the DTG curve, which was supposed to be the product of the second step of decomposition. It can be verified that the decomposition of carbon fiber reinforced epoxy resin matrix composite under inert atmosphere exhibits at least two steps (a major step and a secondary step).   For the gas mass-to-charge ratio of 44, CO 2 , N 2 O and C 3 H 8 were considered. However, the peak of the curve occurred at 410 • C, ahead of 435 • C, which was the precipitation temperature of the CO peak. Considering the sequence of reactions, CO mainly comes from the reforming of the C-O group in the bisphenol A epoxy resin, and then combined with oxygen ions to form CO 2 . Therefore, it was speculated that the m/z = 44 mass spectrum curve was mainly represented by the C 3 H 8 gas molecule. Since the reaction was carried out under an inert atmosphere, oxygen ions were supplied only by the material itself, which was also verified in the order of magnitude. It was speculated that the relatively small increasement of the amplitude of this curve afterwards might correspond to the formation of CO 2 .
The peak temperature of C 2 H 2 gas molecules was close to the temperature of the third peak in the DTG curve, which was supposed to be the product of the second step of decomposition. It can be verified that the decomposition of carbon fiber reinforced epoxy resin matrix composite under inert atmosphere exhibits at least two steps (a major step and a secondary step).

Effect of the Heating Rate on Gas Phase Products
The increase of the heating rate leads to the phenomenon of thermal hysteresis, which is specifically manifested as the peak temperatures of CO, H 2 O, CO 2 and CH 4 ion flow all move toward the high temperature region with the increase of the heating rate. The production of H 2 O, CO 2 and CH 4 all increase with the increase of heating rate as shown in Figure 12, which indicates that faster temperature rise promotes the reaction. The production of CO had a relatively obvious weak value at the low heating rate, but had little change at the heating rates of 10 • C, 20 • C and 40 • C/min. The increase of the heating rate leads to the phenomenon of thermal hysteresis, which is specifically manifested as the peak temperatures of CO, H2O, CO2 and CH4 ion flow all move toward the high temperature region with the increase of the heating rate. The production of H2O, CO2 and CH4 all increase with the increase of heating rate as shown in Figure 12, which indicates that faster temperature rise promotes the reaction. The production of CO had a relatively obvious weak value at the low heating rate, but had little change at the heating rates of 10 °C, 20 °C and 40 °C/min.

Conclusions
Thermal degradation of a new product of carbon fiber reinforced epoxy resin composite, namely BA3202, occurs in three steps under inert atmosphere, but in four steps under air atmosphere, respectively. The first two steps were almost the same. The first step is the physical changes, mainly water drying and carbon dioxide escape. The second step is the decomposition of the epoxy resin matrix, and the products are mainly methane, water, carbon monoxide, carbon dioxide and phenol. The third step in the inert atmosphere is the further pyrolysis of the polymer, in which phenol is formed, methane de-

Conclusions
Thermal degradation of a new product of carbon fiber reinforced epoxy resin composite, namely BA3202, occurs in three steps under inert atmosphere, but in four steps under