Enhanced Mechanical Property of RMI C/SiC Composites with Optimized Porous Carbon Structure by Heat Treatment

C/SiC composites were fabricated by reactive melt inltration (RMI) using porous C/C preforms as the skeleton, followed by the inltration of molten silicon. A convenient technique of heat treatment in the range of 1500 to 2400 o C was applied to modify the porous carbon structure. The effects of heat treatment on the porous C/C preforms and the as-synthesized C/SiC composites were investigated in detail. The results show that the optimal porous carbon structure could be obtained after heat treatment at 1500 o C. After 1500 o C heat treatment, the median pore size and porosity of the porous C/C preform were 9.3 µm and 13.65% respectively, which were in favor of the subsequent inltration of molten Si. The C/SiC composites with the optimized porous carbon structure showed a dense and uniform morphology without obvious cracks. Their bending strength could be up to 276 MPa, which was 28% higher than that of the C/SiC composites with untreated C/C preforms. However, with the increasing heat treatment temperature (2400 o C), the bending strength of the as obtained composites began to decrease because of the degradation of in-situ ber strength. The optimized C/SiC composites exhibited a typical pseudo-plastic fracture behavior with obvious ber pull-out. The improved mechanical property could be ascribed to the lower porosity of composites, the higher in-situ strength of bers and the reduced matrix cracks.


Introduction
The demand for advanced aeroengines with higher speed leads to the further increase in the gas inlet temperature, which stimulates the development of structural materials with outstanding high-temperature properties [1]. Ceramic matrix composites (CMCs) have been recognized as one of the most promising candidates due to their low density, high speci c strength, and the improved high-temperature stability [2]. In the CMC family, C/SiC composites have attracted much attention because of their superior hightemperature strength retention and thermophysical properties [3,4]. Up to date, several different processing routes have been developed to fabricate C/SiC composites and the dominant routes include polymer in ltration and pyrolysis (PIP), chemical vapor in ltration (CVI) and reactive melt in ltration (RMI). Among these preparation routes, RMI is ideal for the industrial production and application owing to the low-cost and e cient densi cation process [5,6].
In RMI process, porous carbon/carbon (C/C) preforms are rstly prepared by the pyrolysis of resin mixture, followed by the in ltration and reaction of molten Si to fabricate the C/SiC composites. It is well known that the properties of the as-fabricated RMI C/SiC composites highly depend on the porous carbon structure of the preforms (the pore size, the porosity, the wettability with molten Si, et al.) [7,8]. Up to now, some effort has been devoted to modify the sizes, amount and distribution of the pores in the porous C/C preforms by varying the initial carbonaceous precursor and con rm the optimal pore size and volume fraction. Researchers in Shanghai Institute of Ceramics prepared a resin mixture of furfuryl resin, poreforming agent and solvent as the precursor, and controlled the pore distribution of porous carbon by varying the pore-forming agent [9]. The results of Margiotta et al. indicated that the ideal median pore diameter of porous carbon was around 1 µm, which was bene cial to the fully in ltration of molten Si and the formation of dense SiC [10]. Singh et al. investigated the effect of initial pore volume fraction on the subsequent in ltration and reaction process, which demonstrated that the pore volume fraction of 0.5 could effectively avoid the chocking-off phenomena and the formation of cracks [11]. Wang et al. found that a simple method of heat treatment could effectively increase the apparent porosity and average pore size of the porous carbon, inferring the potential of heat treatment to modify the porous C/C preforms without introducing other impurities [12]. However, it is still necessary to systematically investigate the effects of heat treatment on the porous carbon structure and clarify the correlation between the annealing temperature and the mechanical property of the as-fabricated RMI C/SiC composites, which is important for the performance optimization of RMI C/SiC composites.
In the present work, the porous C/C preforms were prepared by in ltrating and pyrolyzing a furfuryl resin mixture and then the porous carbon structure in the preforms was modi ed by heat treatment at different temperatures (1500 o C, 1800 o C and 2400 o C). After molten Si in ltration, the C/SiC composites with different porous carbon structure were prepared. The microstructure of the preforms and composites were investigated systematically. The mechanical properties of C/SiC composites were tested and the toughening and strengthening mechanisms of the improved composites were analyzed.

Material and Processing
The preparation of RMI C/SiC composites consists of four steps, 1) weaving and coating bers, 2) impregnation and pyrolysis of the resin mixture, 3) heat treatment, 4) melt Si in ltration. A schematic of the sample preparation in this work using the RMI process is shown in Fig. 1.
PAN-based carbon bers (T700, Toray, Japan) were used as raw materials to fabricate the 2.5D preforms of C/SiC composites. The 2.5D preforms were formed by repeatedly stacking ber web layers, non-woven layers as 0°/90°/0°/90° pierced with carbon ber bundles in Z direction. At rst, for protecting the bers and improving the mechanical properties of composites, a pyrolytic carbon (PyC) interface with a thickness of ~ 800 nm was deposited on the ber surface by the technique of chemical vapor in ltration (CVI) using propane as precursor at 1000 o C with a reduced pressure of 5 KPa and the deposition time was about 250 h. After the interface deposition, the 2.5D preforms were impregnated with the furfuryl resin mixture (60 wt.% furfuryl resin, 20 wt.% glycols, 5 wt.% paratoluene sulfonic acid) in vacuum followed by curing in a drying oven at 200 o C for 4 h. Then the preforms were pyrolyzed at 900 o C for 1 h in argon atmosphere to obtain the porous C/C preforms which were marked as C0. In order to modify their porosity and structure, the porous C/C preforms were annealed at different temperatures (1500 o C, 1800 o C and 2400 o C) for 1 h in argon atmosphere, respectively. The porous C/C preforms after annealing were denoted by C1500, C1800 and C2400, according to the annealing temperatures. Finally, molten Si was in ltrated into these porous C/C preforms (C0, C1500, C1800 and C2400) and reacted with the pyrolysis carbon at 1700 o C for 1 h in vacuum to form SiC matrix. The as-synthesized C/SiC composites were denoted by CS0, CS1500, CS1800 and CS2400, respectively.

Test and Characterization
The carbon yield of furfuryl was measured by thermogravimetry (TG). The test was carried out in argon from 25 to 1500 o C using a Netzsch STA-449F3 thermal analysis system with a heating rate of 10 o C /min. Mercury porosimetry (Auto Pore IV9510, Micromeritics Instrument Ltd, USA) was used to examine the pore size distribution of porous C/C preforms. Bulk density and open porosity of the C/SiC composites were obtained by Archimedes method.
Morphology and element distribution of the samples were characterized by scanning electron microcopy (SEM, Zeiss Supra 55, Germany), equipped with energy dispersive spectroscope (EDS). Crystalline structure and phase composition were investigated by X-ray diffraction (XRD, Rigaku D/MAX 2200 pc, Japan) using Cu K α radiation.
Bending strength of the C/SiC composites was measured using a three-point bending test (Instron 5565, USA) with a span of 40 mm. At least ve samples with a dimension of 60 mm × 9 mm × 4 mm were used for the test. The loading rate was 0.5 mm/min and the loading direction was perpendicular to the nonwoven layers.

Effects of heat treatment on the porous C/C preforms
It is known that the morphology and structure of the porous C/C preforms can be modi ed by appropriate heat treatment [12,13]. To investigate the effect of annealing on morphology of preforms, SEM images of porous C/C preforms with and without heat treatment were shown in Fig. 2. It was shown that the untreated C/C preform (C0) possessed a rather dense structure and the ber web layer was rather smooth without many pores or particulates [ Fig Table 1 Pore characterization data of C/C preforms The pore size distribution of the different C/C preforms was obtained by the technique of mercury porosimetry and the results were shown in Fig. 3 and Table 1. It was indicated that after the heat treatment, the porosity of the C/C preform was increased from 8.6-23.4% while the median pore size was increased from 6.1 to 11.7 µm. The sizes of most pores in C0 were between 1 and 10 µm, while the pore size distributions of C1500 and C1800 preforms were similar, ranging from 2 to 20 µm. However, the pore sizes showed a signi cant increase as the annealing temperature further increased. In C2400, 70% of the total pore volume was contributed from the pores with a size of > 10 µm. From Fig. 3 and Table 1, it can be suggested that the heat treatment for 1 h leaded to the higher porosity and a larger pore size of the C/C preforms.
It is known that carbon matrix skeletons of C/C preforms are formed by the pyrolysis of the furfuryl resin mixture. Based on the results in Fig. 3 and Table 1, it was inferred that the pore structure change of C/C preforms during heat treatment might be attributed to the further pyrolysis from the remaining carbonized products of the furfuryl resin mixture. In order to con rm this assumption and investigate the detailed pyrolysis process of the furfuryl resin mixture during heat treatment, TG was tested from room temperature to 1500 o C and the corresponding curve was presented in Fig. 4. A dramatic decline appeared between 150 and 500 o C, inferring the main pyrolysis process of the furfuryl resin mixture. In the present work, C/C preforms were synthesized by pyrolysis of the furfuryl resin mixture at 900 o C for 1 h. However, in Fig. 4, a trend of persisting declination could be observed up to 1500 o C. The result indicated that the furfuryl resin mixture and its carbonized products were still pyrolyzed at 1500 o C or even higher temperature, so the porosity and pore sizes of the C/C preforms might further increase during heat treatment. Therefore, it is necessary to anneal the C/C preforms to obtain an optimized pore structure.
Besides the porosity and pore size distribution, the carbon microstructure might be modi ed during heat treatment because of the graphitization of turbostratic carbon [14]. It is well known that the porous carbon pyrolyzed from resin shows a typical three-dimensional disordered network structure [15]. The disordered network structure could be changed into an ordered graphite structure during heating [16]. To identify the effect of heat treatment on the microstructure of the porous carbon matrix, XRD was used to examine its crystallinity and the result was shown in Fig. 5. With the increasing annealing temperature, the peak of C(002) became sharper and its full width at half maximum (FWHM) decreased gradually, which indicated the increased order degree of the carbon structure.

Effects of different porous carbon structure on the RMI process
The C/C performs were annealed at different temperatures to modify their porous carbon structure, followed by being in ltrated with molten Si to fabricate the RMI C/SiC composites. According to the various annealing temperatures, the as-synthesized C/SiC samples were indicated as CS1500, CS1800 and CS2400, respectively. For comparison, the RMI C/SiC sample using the untreated C/C preform was also prepared (CS0).
Quantitative XRD tests were employed to investigate the phase composition and relative amount of C/SiC composites. Corresponding results were exhibited in Fig. 6. All the C/SiC samples consisted of β-SiC, amorphous carbon and free Si peaks. The Si phase was from the unreacted residual Si during the in ltration. The intensity of Si and SiC peaks were used to determine their relative amounts (the relative amount of C could not be obtained because of its amorphism) and the results were listed in Table 2. There were less residual Si (SiSiC < 0.1) in the CS0, CS1500 and CS1800, but signi cantly increased residual Si (SiSiC ~ 0.5) could be detected in CS2400. The results indicated that the relatively large pore size of C2400 might cause the reaction of Si and C to be incomplete and then the melt Si to be left between the generated SiC phase. However the average pore size of C0 was so small that unreacted residual C might be remained after the reaction. Therefore, the C1500 and C1800 might be more suitable for RMI process.  Figure 7 exhibited the cross-section SEM images of the C/SiC composites. The composites were composed of black, grey and light grey phases. Corresponding EDS analyses indicated that the black, grey and light grey phases were C, SiC and Si, respectively. The stacking structure of ber web layers and non-woven layers were clear in the low-magni cation SEM images [ Fig. 7(a), (c), (e) and (g)] and the SiC and Si were mainly distributed in the ber web layers because the non-woven layers were too compact for the in ltration of molten Si. Meanwhile, the micropores presented in the ber web layers were bene cial for the Si in ltration via capillary force. It was worth noting that the volume ratio of the in ltrated Si in the composites increased when the annealing temperature was up to 2400 o C [ Fig. 7 (g)]. In addition, the detailed images con rmed that there were large amounts of residual Si and cracks in the CS2400 [ Fig. 7(h)]. However, no obvious residual Si and cracks can be found in CS0, CS1500 and CS1800.
The above results indicates that the porous carbon structure of C/C preforms could be modi ed by heat treatment, and further affects the fabrication of C/SiC composites with different microstructure via Si in ltration. To illuminate the correlation between porous carbon structure and the microstructure of C/SiC composites, it is essential to understand the effects of porous carbon structure on the RMI process. It is an exothermic reaction that occurred in the molten Si in ltration: The in ltration effect during RMI process is highly dependent on the pore sizes of C/C preforms. On the one hand, smaller pores might lead to the incomplete in ltration of molten Si due to the "chock-off" phenomenon [ Fig. 7(b)]. This phenomenon is attributed to the fact that, in the in ltration process, the diffusion channels of molten Si became narrower and even completely closed, and then the Si ow would stop. Meanwhile, for a preform with smaller pore sizes, the reaction rate between Si and C was faster than the in ltration rate of Si, and the reaction was mild and limited. The reaction became nearly isothermal since the chemical reaction heat could be released outward in time. On the other hand, larger pores might result in the super uous unreacted Si formed in the as-obtained C/SiC composites due to the limitation of atom diffusion in reaction [ Fig. 7(h)]. Besides, for a preform with larger pore sizes, the in ltration rate of Si was higher than the reaction rate between Si and C. So molten Si would rapidly in ltrate and ll the pores of preforms. The reaction was violent and the reaction heat was hard to be released, which resulted in the rather high local temperature. Therefore, some microcracks would generate in the ceramic matrix because of the remarkable thermo-stress concentration [ Fig. 7(h)] [17].
The porosity and density data of the porous C/C preforms and C/SiC composites were summarized in Table 1 and Table 2. It was shown that the porosity of C/C preforms increased from 8.59-13.65% and 12.83% after annealing at 1500 o C and 1800 o C, respectively. When the annealing temperature was raised to 2400 o C, the porosity of C/C preform was up to 23.38% (2.7 times higher than that before annealing). The increase of porosity could be ascribed to the further pyrolysis from the remaining carbonized product of the furfuryl resin mixture, which was con rmed in the TG test (Fig. 4). An appropriate porous C/C preform is the basis for the fabrication of RMI C/SiC composites with optimized microstructure.
Compared with CS0, CS1500 and CS1800, CS2400 had larger porosity and lower density. Given the fact that the density of SiC is higher than that of Si and C, the result infers that in CS2400, less SiC could be generated and some large pores could not be lled completely due to the large pore size and porosity of C2400.
Besides the pore structure of carbon preforms, the effect of wettability and reactivity between C and Si ,which was attribute to the heat treatment, on the in ltration process is worth noting.
The wettability of solid and liquid is highly related to their atomic or molecular structure [18]. Some previous researches indicated that the increasing structural order degree of carbon materials would result in the increasing wettability between liquid Si and solid carbon [19][20][21]. According to the Washburn model [22], the in ltration of liquid Si into porous C could be described by the following equation where, is the in ltration rate, h is the in ltration depth, R is the capillary radius, η is the viscosity, σ is the surface tension, and θ is the equilibrium contact angle. Washburn equation indicates a negative correlation between equilibrium contact angle θ and in ltration speed/in ltration depth. Therefore, molten Si with a better wettability with carbon (a smaller contact angle θ) will lead to an improved in ltration effect [23].
With the increasing structural order degree of carbon after annealing, the wettability was improved and the in ltration process was enhanced. In addition, the reactivity between the porous carbon substrate and the molten Si might be strengthened by heat treatment [24].
However, based on the results shown in Fig. 7 and Table 2, it can be inferred that, compared with the graphitization degree and reactivity of carbon matrix, the porous structure of carbon preforms is the main factor of the melt in ltration process. The C1500 and C1800 are more suitable for Si in ltration and forming dense C/SiC composites without cracks.

Mechanical properties of the C/SiC composites with various porous carbon structure
The results of mechanical properties of C/SiC composites were given in Table 3. It was obvious that the average bending strength and modulus of CS1500 and CS1800 were higher than those of CS0. As the annealing temperature of the C/C preforms increased to 2400 o C, the strength and modulus of asobtained C/SiC composites decreased. The bending strength of CS2400 was 34% lower than that of CS1500.
The typical bending stress-strain curves for CS0, CS1500, CS1800 and CS2400 were shown in Fig. 8. It was obvious that CS1500 had the highest bending strength and toughness. All the four curves were composed of three stages (1) At rst, the bending stress linearly increased with the bending strain, during which elastic deformation occurred. At this stage, no new cracks were formed in the matrix and the existing microcracks did not propagate. (2) When the bending strength reached a critical value, the curves were deviated from linearity to nonlinearity. The new cracks generated and the existing cracks began to grow. The increasing crack density with the rise of load resulted in the gradually declining stiffness of the composites till the rst bundle of bers broke and the curves reached maximum stress. (3) After the peak stress, the curves fell slowly in a step-like mode instead of catastrophic failure, indicating a typical pseudo-plastic fracture behavior.  Figure 9 presented the typical macroscopic and microscopic fracture morphology of the C/SiC composites after three-point bending tests. The average pull-out length of the bers in CS1500 was about 5 mm and the delamination between different layers can be observed [ Fig. 9 (b)]. In CS1800, the ber pullout length was shorter than that of CS1500 and no obvious delamination can be found [ Fig. 9(c)]. While compared with CS1500 and CS1800, CS2400 showed a much shorter pull-out length (~ 2 mm) without delamination [ Fig. 9(d)]. The ber pull-out length of CS0 (~ 2.5 mm) was a little longer than that of CS2400 [ Fig. 9(d)].
During the loading process, the stress would be transferred from the matrix to the bers by the interface.
The cracks in matrix would be de ected when they encountered the weak PyC interface and then propagated along the direction paralleled to the ber axis [25]. When the stress exceeded the in-situ strength of a ber and the interface was weak enough, the ber would be fractured and pulled out, resulting in the increasing fracture energy and the improved bending strength and toughness. According to Fig. 9, in comparison with other composites (CS0, CS1800 and CS2400), the longer pull-out length of bers and the obvious delamination can be observed in CS1500, suggesting a better toughness. The result was in consistent with the stress-train curves in Fig. 8.
From results above, it can be indicated that after appropriate heat treatment for the C/C preforms, the porous carbon structure was optimized and the strength of the as-synthesized C/SiC composites increased (CS1500 and CS1800). But the excessive annealing temperature for C/C preforms will lead to a decreased strength of the C/SiC composites (CS2400). Generally, the mechanical properties of composites can be in uenced by various factors, such as density, matrix microstructure, bonding strength of interface, ber properties, and preform architecture [26]. In the present work, when the annealing temperature was up to 2400 ℃, the pore size and porosity of C/C preforms increased signi cantly. Higher local thermal stress was generated in the as-synthesized C/SiC composites owing to the release of massive reaction heat during RMI process. In addition, the porosity of CS2400 was quite higher than that of other composites and the pores might be the crack initiation. Hence, the existing cracks and high porosity of CS2400 were detrimental for its mechanical properties. Meanwhile, as the main load-carrying component, the continuous carbon bers with a volume fraction of 40% played an important role in the mechanical properties of C/SiC composites. Excessive annealing temperature might result in the degraded ber in-situ strength [27].
To clarify the effect of heat treatment on the in-situ strengths of carbon bers, the ber fracture surfaces of the different C/SiC composites after bending test were inspected by SEM to compare their in-situ strengths (Fig. 10). The in-situ strength of ber can be calculated using the following equation [28] where S is the in-situ ber strength, A m is a constant related to ber type, and r m is the mirror radius of the fractured ber.
In general, the cross-section of a fractured ber consists of four parts source, mirror, mist, hackle [29]. The outer mirror radius (r m ) is the distance from the origin to the onset of hackle. However, the morphology of fracture surface in the CS0 and CS1500 were rather rough that the measurement of mirror radius was di cult [ Fig. 10(a-b)]. This phenomenon infers that the in-situ ber strength of CS0 and CS1500 were quite high. The fracture surface of bers in CS1800 and CS2400 showed clearer mirror radiuses, which were about 1 µm and 1.54 µm, respectively [ Fig. 10(c-d)]. The larger mirror radius suggested the degradation of in-situ ber strength in CS2400 [28]. Overall, the strength of RMI C/SiC composites was determined by the combined effects of the porosity, the existing cracks and the in-situ strength of bers.

Conclusions
RMI C/SiC composites with enhanced mechanical property were prepared by optimizing the porous carbon structure using a simple method of heat treatment. The effects of heat treatment on the structure of porous carbon preforms, RMI process and mechanical properties of C/SiC composites were investigated in detail. The porosity and average pore size of the porous carbon preforms increased as the heat treatment temperature increased. The optimal pore structure of porous carbon preforms could be obtained after heat treatment at 1500 o C, which is bene cial to in ltrate molten Si. The microstructure of C/SiC composites with optimal pore structure are dense and nearly without cracks. Therefore, owing to the lower porosity, the good in-situ strength of bers and the less matrix cracks, the as-prepared C/SiC composites present a superior mechanical property and a pseudo-plastic fracture behavior. The pull-out of bers and the delamination of composites could be observed during the fracture of samples.