Concurrent optimization of fracture toughness, thermal conductivity, and tribological behavior in C_f/Si₃N₄ composites via phase driven selection

Abstract

Carbon fiber–reinforced silicon nitride (C_f/Si₃N₄) composites were fabricated by spark plasma sintering (SPS) using α-, β-, and γ-Si₃N₄ powders to clarify the influence of the initial Si₃N₄ phase on microstructural evolution and functional properties. The results show that the starting phase significantly affects densification behavior, phase transformation, and mechanical and tribological performance. The composite derived from α-Si₃N₄ achieved the highest relative density (96.53%) and exhibited an optimal balance of fracture toughness (10.87 MPa m^0.5), thermal conductivity (66 W/m K), and stable friction behavior (COF ≈ 0.46). This performance is attributed to the in-situ formation of a self-reinforced β-Si₃N₄ microstructure during the α → β phase transformation, which promotes crack deflection, crack bridging, and effective load transfer in synergy with carbon fibers. In contrast, β- and γ-Si₃N₄–based composites showed lower densification or excessive hardness associated with increased porosity and secondary phase formation. These findings demonstrate that controlling the initial Si₃N₄ phase provides an effective microstructural design strategy for developing high-performance C_f/Si₃N₄ composites for thermostructural applications such as aerospace brake discs.

Introduction

Silicon nitride (Si₃N₄)–based ceramics are widely recognized as promising materials for high-temperature and tribological applications due to their excellent mechanical strength, thermal stability, and wear resistance. These attributes make Si₃N₄ an attractive candidate for demanding thermostructural components such as aerospace brake discs, where a delicate balance between high fracture toughness, stable friction behavior, and effectiveological applications due to their excellent mechanical strength, thermal stability, and wear resistance. However, the inherent brittleness of monolithic Si₃N₄ limits its damage tolerance under severe mechanical loading and rapid thermal cycling, restricting its direct application in advanced braking systems^1,2,3,4,5.

Currently, carbon fiber–reinforced silicon carbide (C_f/SiC) composites represent the state-of-the-art materials for high-performance brake discs in aerospace and motorsport applications. C_f/SiC systems exhibit high specific strength high-performance brake discs in aerospace and motorsport applications. Cf/SiC systems exhibit high specific strength, good thermal conductivity, and excellent oxidation resistance, good thermal conductivity, and excellent oxidation resistance, particularly when fabricated via chemical vapor infiltration (CVI) or polymer infiltration and pyrolysis (PIP). Nevertheless, their relatively moderate fracture toughness, sensitivity to thermal shock under extreme conditions, and high manufacturing cost remain persistent challenges, especially for next-generation aerospace braking systems^6,7,8,9.

In this context, carbon fiber–reinforced silicon nitride (C_f/Si₃N₄) composites have emerged as a compelling alternative. The incorporation of carbon fibers into a Si₃N₄ matrix can significantly improve fracture toughness and thermal shock resistance through mechanisms such as crack deflection, crack bridging, and fiber pull-out, while simultaneously enhancing tribological stability. The final performance of C_f/Si₃N₄ composites, however, is strongly governed by densification behavior, fiber–matrix interfacial bonding, and microstructural evolution during sintering^10,11.

Among the final performance of C_f/Si₃N₄ composites, however, is strongly governed by densification behavior, fiber–matrix interfacial bonding, and microstructural evolution during sintering^10,11. Among the parameters controlling microstructural development, the initial phase of the Si₃N₄ powder—α, β, or γ—plays a decisive role. The well-known α → β phase transformation during the parameters controlling microstructural development, during sintering promotes the formation of elongated β-Si₃N₄ grains, which act as an in situ self-reinforcement network and substantially enhance fracture resistance. In contrast, β-Si₃N₄ powders generally exhibit limited grain growth, while γ-Si₃N₄ powders, due to their amorphous nature and finer particle size, may hinder effective densification despite promoting high hardness^2,3,4,5,6,7.

Spark plasma sintering (SPS) offers distinct advantages for the consolidation of Si₃N₄-based composites, including rapid heating rates, short dwell times, and enhanced mass transport, which collectively enable high densification at relatively lower sintering durations. When combined with appropriate oxide sintering aids (e.g., Al₂O₃ and Y₂O₃), SPS can promote liquid-phase-assisted densification while suppressing thermal decomposition of Si₃N₄^2,7.

Despite extensive research on monolithic Si₃N₄ ceramics and C_f/SiC brake materials, a systematic and comparative investigation of C_f/Si₃N₄ composites fabricated from different initial Si₃N₄ phases under identical SPS conditions remains notably limited. Therefore, the present study aims to elucidate the influence of α-, β-, and γ-Si₃N₄ starting powders on phase evolution, microstructural development, densification behavior, and the resulting mechanical, thermal, and tribological properties of C_f/Si₃N₄ composites. By establishing clear phase–microstructure–property relationships, this work introduces a phase-informed microstructural design strategy that moves beyond conventional composition-based approaches, offering a viable alternative to SiC-based systems for design strategy that moves beyond conventional composition-based approaches, offering a viable alternative to SiC-based systems for advanced aerospace braking and thermostructural applications^{8,9,10,11,12,13,14}.

Experimental

Materials

Commercial silicon nitride (Si₃N₄) powders with α-, β-, and γ-phases were purchased from US Research Nanomaterials, Inc. (Houston, USA). It should be noted that X-ray diffraction (XRD) analysis revealed the presence of approximately 10 wt.% β-Si₃N₄ in the as-received α-Si₃N₄ powder. In addition, the commercially labeled γ-Si₃N₄ powder was found to be predominantly amorphous, in agreement with the supplier’s specifications and XRD results (Figs. S1 and S2, Supplementary Information) (Fig. 1).

Fig. 1

Schematic of the C_f/Si₃N₄/Al₂O₃/Y₂O₃ composite bulk production process .

Aluminum oxide (Al₂O₃, 99.5% purity, average particle size ~ 150 nm, mixed hexagonal/monoclinic structure; Fig. S2b) and yttrium oxide (Y₂O₃, 99.9% purity, average particle size ~ 50 nm, rhombohedral structure; Fig. S2c) were used as sintering aids. Continuous carbon fibers (T300 grade, average diameter ~ 6 µm; Fig. S3) were supplied by TAM Co. (Seoul, Korea) and cut into 5 mm lengths using a precision fiber cutter prior to composite fabrication. Methyl ethyl ketone (MEK) and isopropyl alcohol (IPA) (Merck & Co.) were employed as dispersion media during powder processing.

The morphology and chemical composition of the as-received Si₃N₄ powders (α, β, and γ) and carbon fibers were characterized using field-emission scanning electron microscopy (FESEM) equipped with energy-dispersive spectroscopy (EDS). As shown in Fig. S4 (Supplementary Information), the α- and β-Si₃N₄ powders exhibited a semi-spherical morphology with particle sizes in the range of 500–600 nm, whereas the γ-Si₃N₄ powder consisted of finer particles with sizes of approximately 200–300 nm. In addition, the α-Si₃N₄ powder contained a minor fraction of rod-like particles with diameters of ~ 250 nm and lengths of ~ 1 µm. EDS analysis confirmed the high purity of all powders, with no detectable impurities above 0.5 wt.% (Fig. S4).

Fabrication of the Cf/Si3N4/Al2O3/Y2O3 nanocomposite

The composite powders were prepared through a three-step dispersion process. In the first step, 1.5 g of Si₃N₄ powder was dispersed in 250 mL of MEK and homogenized for 2 h using mechanical stirring and ultrasonication. In the second step, 0.06 g of Al₂O₃ and 0.12 g of Y₂O₃ were added to the Si₃N₄ suspension and further dispersed for 1 h. The resulting slurry was then dried in an oven at 80 °C to remove the solvent.

In the third step, carbon fibers (20 wt.%, initial length 5 mm) were dispersed in IPA using ultrasonic agitation for 24 h to ensure uniform separation. The fiber suspension was subsequently mixed with the pre-dispersed Si₃N₄/Al₂O₃/Y₂O₃ powder and further homogenized for 4 h by combined magnetic stirring and ultrasonication. Finally, the mixed slurry was dried at 80 °C until complete solvent evaporation, yielding homogeneous Cf/Si₃N₄/Al₂O₃/Y₂O₃ composite powders.

The resulting composites based on α-, β-, and γ-Si₃N₄ precursors were denoted as Sample 1 (C_f/α-Si₃N₄), Sample 2 (C_f/β-Si₃N₄), and Sample 3 (C_f/γ-Si₃N₄), respectively.

Sintering of the Cf/Si3N4/Al2O3/Y2O3 nanocomposite powders

The prepared composite powders were consolidated using a SPS system equipped with a graphite die. A graphite foil with a thickness of 0.3 mm was placed between the powder compact and the die to prevent adhesion. Sintering was conducted under vacuum (~0.1 Torr) at 1850 °C for 20 min, applying a uniaxial pressure of 70 MPa and a heating rate of 200 °C min⁻¹.

Characterization tests

The phase composition of the starting powders and sintered composites was determined by X-ray diffraction (XRD) using a PANalytical X’Pert Pro MPD diffractometer with Cu–Kα radiation (λ = 1.542 Å), operated at 40 kV and 30 mA, with a step size of 0.05°. The average crystallite size was estimated using the Scherrer equation¹⁵:

$$D = k\lambda /\beta \cos \theta$$

where λ is the wavelength (Å), k is the shape factor (~ 1), θ is the diffraction angle, and β is the full width at half maximum (FWHM, radians).

Morphology and microstructure of the powders and sintered ceramics were examined by field-emission scanning electron microscopy (FESEM, QUANTA PEG 450) equipped with energy-dispersive spectroscopy (EDS) for chemical composition assessment.

The bulk density of the sintered ceramics was measured via the Archimedes method according to ASTM C373¹⁶. Vickers hardness measurements (ASTM C1327-15)^17,18 were conducted at five different locations with a 9.8 N load (1 kgf) and 15 s dwell time.

Fracture toughness (K_IC) was estimated from Vickers indentation using both Anstis et al. and Niihara models to account for possible crack types (Palmqvist vs. median/half-penny)^19,20. Crack lengths were measured from FESEM images (Fig. S5), and K_IC was calculated by:

$$K_{{{\text{IC}}}} = 0.0{16}\left( {{\text{E}}/{\text{Hv}}} \right)^{{{1}/{2}}} ({\text{P}}/{\text{c}}^{{{3}/{2}}} )$$

(1)

where E is the Young’s modulus, c is crack length Hᵥ is the Vickers hardness (Hᵥ (GPa) = 1.854 F / (2a)2), and P is the indentation load.

The fracture toughness (K_IC) values reported herein were derived from the Vickers indentation method. It is critically acknowledged that this technique provides an estimate of 'indentation fracture resistance’ rather than an absolute measure of fracture toughness, as it is sensitive to crack morphology (Palmqvist vs. median/radial systems) and residual stress states, and may not fully capture the toughening contributions of long fibers in a composite^19,20,21. Therefore, the presented K_IC values should be interpreted primarily as a consistent comparative metric to evaluate the relative performance among the three composite variants (α-, β-, and γ-derived) processed and tested under identical conditions. The observed trends are further corroborated by flexural strength tests and microstructural analysis.

Tribological testing was performed using a pin-on-disk setup (ASTM G99-17)²² at a 5 N normal load and 0.5 m/s sliding speed. This load was selected based on Hertzian contact analysis for a sphere-on-flat configuration^22,23 to reproduce realistic contact stress (~ 1.2 GPa) at aerospace brake interfaces, while preventing premature failure. Wear track morphologies were examined by FESEM to elucidate mechanisms.

Flexural strength was evaluated by a three-point bending method adapted for disc specimens²⁴. Disc-shaped SPS samples (D = 20 mm, thickness = 2 mm) were used without additional machining to avoid altering the sintered microstructure. Support span was set at 16 mm, and fracture strength calculated as:

$$\sigma {\text{f}} = {\text{FL}}/\left( {\pi {\text{R}}^{{3}} } \right)$$

(2)

where F is fracture load, L is support span, and R is disc radius. Testing was conducted at 0.5 mm/min crosshead speed, with at least five specimens per composition. Thermal conductivity (K) was determined per ASTM E1461-13^25,26 using:

$${\text{K}}\left( {{\text{W}}/{\text{m}} \cdot {\text{k}}} \right) = \alpha \rho C_{p}$$

(3)

where α is thermal diffusivity (Xenon flash analysis, Leo model, Germany), ρ is density (kg/m3), and Cp​ is specific heat capacity measured via differential scanning calorimetry (DSC). Tests were performed on 10 × 10 mm2 specimens with 1.4 mm thickness.

Results and discussions

XRD patterns of composite before and after sintering

XRD patterns of the starting powders (Figs. S1 and S2) confirm the crystalline nature of α- and β-Si₃N₄, whereas the commercially labeled γ-Si₃N₄ powder exhibits an amorphous structure. Carbon fibers also show no crystalline diffraction peaks, consistent with their turbostratic carbon structure. After spark plasma sintering at 1850 °C (Figs. 2, 3 and 4), all composite samples exhibit significant phase evolution driven by reactions between Si₃N₄, surface silica, and the Al₂O₃–Y₂O₃ sintering additives. In Sample 1 (Cf/α-Si₃N₄), the dominant phases after sintering are α- and β-Si₃N₄, together with Al₂O₃, Y₂O₃, and carbon. The appearance of SiO₂ peaks after sintering (Fig. 2b), absent in the green compact (Fig. 2a), indicates the formation of a silica-rich liquid phase. This liquid originates from the reaction of native SiO₂ layers on Si₃N₄ particles with the oxide additives, facilitating mass transport and densification.

Fig. 2

XRD results of C_f/α-Si₃N₄/Al₂O₃/Y₂O₃ powder, (a) pre-sintering (Powder), (b) post-sintering.

Fig. 3

XRD results of C_f/β-Si₃N₄/Al₂O₃/Y₂O₃ powder, (a) pre-sintering (Powder), (b) post-sintering.

Fig. 4

XRD results of C_f/γ-Si₃N₄/Al₂O₃/Y₂O₃ powder, (a) pre-sintering (Powder), (b) post-sintering.

A similar phase assemblage is observed for Sample 2 (C_f/β-Si₃N₄)(Fig. 3a and b) ,where, β-Si₃N₄ remains the primary crystalline phase after sintering (Fig. 3b). However, the relative intensity of SiO₂-related features is higher than in Sample 1, suggesting a larger liquid phase fraction, which is consistent with the lower final density measured for this sample.

In contrast, Sample 3 (C_f/γ-Si₃N₄) exhibits markedly different behavior. Prior to sintering (Fig. 4a), the diffraction pattern confirms the amorphous nature of the γ-Si₃N₄ powder. After sintering (Fig. 4b), this amorphous phase fully transforms into crystalline β-Si₃N₄ (JCPDS No. 82–0701, hexagonal structure), accompanied by the formation of SiO₂ and Si₂N₂O secondary phases. The emergence of Si₂N₂O indicates a more extensive reaction between Si₃N₄, surface silica, and oxide additives, promoted by the higher chemical reactivity of the amorphous precursor.

Sintering Si₃N₄ at 1850 °C under vacuum typically raises concerns regarding thermal decomposition. However, in the present system, the presence of Y₂O₃ and Al₂O₃ promotes the formation of a transient liquid phase at temperatures as low as ~ 1700 °C, enhancing densification while suppressing decomposition by locally reducing the effective nitrogen partial pressure at particle boundaries. The absence of elemental Si and the lack of excessive SiC formation in the XRD patterns of all sintered samples confirm that Si₃N₄ decomposition was effectively mitigated under the applied SPS conditions. Any limited interfacial reaction between carbon fibers and Si₃N₄ is expected to result in a thin SiC interlayer, which may be beneficial for interfacial bonding and load transfer.

Microstructure of sintered bodies

FESEM micrographs of the sintered composites (Fig. 5) reveal a generally uniform dispersion of carbon fibers within the Si₃N₄ matrix for all samples. The originally continuous fibers (5 mm) were shortened during powder processing and consolidation to lengths ranging from approximately 24 to 144 µm, primarily due to high-shear mixing and ultrasonic dispersion.

Fig. 5

FESEM images related to the sintered samples: (a) sample 1, (b) sample 2, and (c) sample 3.

Among the three compositions, Sample 2 (C_f/β-Si₃N₄) shows localized agglomeration of nanoparticles and carbon fibers, which can negatively affect electrical and thermal conductivity during SPS and hinder uniform densification. In contrast, Samples 1 and 3 exhibit a more homogeneous fiber distribution, which promotes efficient Joule heating and mass transport during sintering, consistent with improved consolidation behavior²⁷.

Elemental mapping results (Fig. 6) further confirm the relatively uniform spatial distribution of Si, Al, Y, O, and C within the composite matrices, indicating effective dispersion of sintering additives and carbon fibers across all samples, despite differences in local agglomeration and phase evolution.

Fig. 6

SEM images related to the sintered samples: (a) sample 1, (b) sample 2, and (c) sample 3.

Investigate of density and hardness

The relative densities of the SPS-consolidated composites prepared from α-, β-, and γ-Si₃N₄ powders were measured as 96.53%, 90.9%, and 87.47% of the theoretical density, respectively. Accordingly, the residual porosity values were calculated to be 3.47%, 9.1%, and 12.53% for the α-, β-, and γ-based composites.

The observed differences in densification behavior are closely related to the amount and nature of the secondary phases formed during sintering, particularly SiO₂-based liquid phases. The formation of SiO₂ during SPS plays a dual role: a limited amount is beneficial by generating a transient liquid phase that enhances mass transport and pore elimination, whereas excessive liquid content can hinder densification and deteriorate mechanical properties. For advanced structural ceramics, the optimal liquid phase content is typically maintained below ~ 5 vol.% to avoid property degradation.

The highest density achieved in the C_f/α-Si₃N₄ composite is attributed to the formation of an optimal liquid phase fraction, which promotes effective particle rearrangement and densification. In contrast, XRD results indicate a higher intensity of SiO₂-related phases in the C_f/β-Si₃N₄ composite, suggesting an increased liquid volume fraction that likely exceeded the optimal range, leading to reduced final density. For the C_f/γ-Si₃N₄ composite, despite the presence of a well-developed SiO₂ phase, densification was further impeded by the formation of Si₂N₂O and by the additional energy consumption associated with the crystallization of the initially amorphous γ-Si₃N₄ powder during sintering.

Notably, the relative density of the C_f/α-Si₃N₄ composite surpasses that reported for C_f/SiC brake discs fabricated by the liquid silicon infiltration (LSI) method, which typically achieve only 80–85% of the theoretical density^6,7,27, highlighting the effectiveness of the SPS route for C_f/Si₃N₄ systems.

Vickers hardness measurements revealed values of 14.1 GPa (VHN = 1433), 13.4 GPa (VHN = 1361), and 20.4 GPa (VHN = 2079) for Samples 1 (α), 2 (β), and 3 (γ), respectively. Despite its lower relative density, the C_f/γ-Si₃N₄ composite exhibited the highest hardness. This behavior is attributed to the finer initial particle size of the γ-Si₃N₄ nanopowders and the formation of hard Si₂N₂O secondary phases, which are characterized by strong covalent bonding and inherently high hardness. These results indicate that hardness in the present composites is governed not only by densification level but also by phase constitution and intrinsic bonding characteristics.

Investigate of coefficient of friction (COF)

Since surface roughness strongly influences tribological behavior, the roughness of all samples was carefully controlled prior to wear testing. The average surface roughness (R_a) values were measured as 0.36, 0.39, and 0.34 µm for the C_f/α-Si₃N₄, C_f/β-Si₃N₄, and C_f/γ-Si₃N₄ composites, respectively, confirming that all samples possessed comparable surface conditions and that differences in friction and wear behavior primarily originated from intrinsic material properties rather than surface topography.

Wear tests were conducted under a normal load of 5 N, selected based on Hertzian contact stress calculations using Hertzwin software. By considering the elastic moduli of the alumina ball (340 GPa) and the composite disc (110 GPa)^28,29,30,31, along with their respective Poisson’s ratios (0.22 for the ball and 0.085 for the disc)^{32,33,34,35,36}, the applied load was determined to generate a contact stress of approximately 1.2 GPa, representative of severe service conditions relevant to aerospace braking systems.

The measured coefficients of friction (Fig. 7) for the C_f/α-, C_f/β-, and C_f/γ-Si₃N₄ composites were approximately 0.46, 0.48, and 0.44, respectively. Notably, the C_f/γ-Si₃N₄ composite exhibited a COF closest to that reported for aerial brake discs (~ 0.4), whereas the C_f/α-Si₃N₄ sample demonstrated the most stable friction response, with minimal fluctuations throughout the test. This stability indicates the formation of a robust and continuous tribofilm at the sliding interface.

Fig. 7

Friction coefficient curves in terms of distance and Ra curves of the sintered samples: (a) sample 1, (b) sample 2, and (c) sample 3.

Wear rate measurements further differentiate the tribological performance of the composites. The wear rates of Samples 1 (α), 2 (β), and 3 (γ) were determined to be on the order of 10⁻⁷ g/N·m, with the C_f/α-Si₃N₄ composite consistently exhibiting the lowest wear rate and the most stable friction behavior. This superior performance is attributed to its optimized microstructure and the effective self-lubricating action provided by uniformly distributed carbon fibers.

SEM observations of the wear tracks (Fig. 8) and corresponding EDS mapping (Fig. S6) reveal that mild abrasive wear is the dominant wear mechanism for all composites, characterized by shallow grooves along the sliding direction. However, the key distinction of the C_f/α-Si₃N₄ composite lies in the formation of a protective tribofilm composed primarily of smeared carbonaceous material and tribo-oxidation products. This tribofilm acts as a lubricating layer, reducing direct asperity contact between the composite and the alumina counterface and thereby stabilizing the COF at approximately 0.46.

Fig. 8

FESEM images of wear tracks (a–c) sample 1, (d–f) sample 2, and (g–i) sample 3.

In addition to tribofilm formation, carbon fiber bridging was frequently observed within the wear tracks of the C_f/α-Si₃N₄ composite. Individual fibers span microcracks and voids generated during sliding, effectively suppressing crack propagation and mitigating material removal. This mechanism, combined with the in-situ formation of an interlocking, needle-like β-Si₃N₄ grain structure, enhances crack deflection and energy dissipation, resulting in improved wear resistance despite partial fiber shortening (24–144 µm) caused by high-shear mixing and SPS consolidation.

By contrast, the C_f/γ-Si₃N₄ composite, although exhibiting higher bulk hardness, shows a more brittle wear response. Its higher residual porosity and the presence of brittle secondary phases promote subsurface cracking and deeper wear grooves, leading to increased material removal. The C_f/β-Si₃N₄ composite displays intermediate tribological performance, lacking both the optimal self-reinforced grain morphology and the uniform fiber distribution required for superior wear resistance.

Overall, the outstanding tribological performance of the C_f/α-Si₃N₄ composite arises from the synergistic interaction of a tough, self-reinforced matrix, the formation of a stable carbon-rich tribofilm, and effective carbon fiber bridging. This multi-mechanism synergy enables stable friction behavior and reduced wear under high contact stress, making the α-Si₃N₄-derived composite a strong candidate for demanding aerospace braking applications. The observed tribofilm stability and damage-tolerant wear mechanisms are consistent with previous reports on C_f/SiC and elongated-grain Si₃N₄-based composites^1,5,37,38,39.

The processing route, notably the high-shear mixing and high-pressure SPS consolidation, may induce some shortening and damage to the carbon fibers, as suggested by the reduced fiber lengths (24–144 μm) observed post-sintering compared to the initial 5 mm. While such microstructural alterations could potentially diminish the fiber-bridging contribution to fracture toughness, the dominant effect in the α-Si₃N₄-derived composite appears to be the positive formation of an interlocking, self-reinforced β-Si₃N₄ grain network. This in-situ toughened matrix, coupled with residual fiber pull-out mechanisms evident in fracture surfaces (Fig. 8), effectively counterbalances any negative impact from fiber processing damage, leading to the superior overall mechanical and tribological performance observed.

Investigation of flexural strength and fracture toughness

The flexural strength results indicate that all composites exhibited predominantly brittle fracture behavior, as failure occurred within the elastic region of the stress–strain response. The measured flexural strengths of the C_f/α-Si₃N₄, C_f/β-Si₃N₄, and C_f/γ-Si₃N₄ composites were 226 ± 3, 142 ± 4, and 194 ± 4 MPa, respectively (Fig. 9).

Fig. 9

Force–displacement curve of the sintered samples with different Si3N4 phases.

The superior flexural strength of the C_f/α-Si₃N₄ composite (Sample 1) is primarily attributed to its unique post-sintering microstructure and high relative density. As revealed by FESEM observations (Figs. 8 and 10) and quantitative phase analysis using MAUD software, this composite consists predominantly of elongated, needle-like β-Si₃N₄ grains (~ 79%) embedded within a minor fraction of equiaxed α-Si₃N₄ grains (~ 20%). These elongated β-Si₃N₄ grains act as an in-situ self-reinforcement phase, analogous to whiskers, effectively impeding crack initiation and propagation. Their interlocking morphology increases crack path tortuosity and enhances load transfer across the matrix, thereby improving flexural strength.

Fig. 10

SEM images of the fracture surface of the sintered samples after flexural strength test: (a) sample 1, (b) sample 2, and (c) sample 3.

In addition to microstructural reinforcement, density plays a critical role in determining flexural strength. Higher density corresponds to lower residual porosity, which reduces stress concentration sites and delays crack initiation. The C_f/α-Si₃N₄ composite exhibited the highest relative density among all samples, consistent with its superior flexural strength. Interestingly, the C_f/γ-Si₃N₄ composite demonstrated a higher flexural strength than the C_f/β-Si₃N₄ composite despite its lower density. This behavior is attributed to the presence of the Si₂N₂O phase in the γ-derived composite, which is characterized by strong covalent bonding and contributes to enhanced local stiffness and strength.

Fracture surface analyses (Fig. 10) further reveal that fiber pull-out is a dominant toughening mechanism in all composites. The presence of pulled-out carbon fibers indicates effective load transfer across the fiber–matrix interface and contributes to energy dissipation during fracture, thereby enhancing both strength and toughness.

The fracture toughness (K_IC​) values of Samples 1, 2, and 3 were measured as 10.87 ± 0.6, 8.36 ± 0.3, and 4.91 ± 0.5 MPa m^0.5, respectively. The significantly higher fracture toughness of the C_f/α-Si₃N₄ composite is attributed to a combination of its high relative density and its self-reinforced microstructure. In Si₃N₄-based ceramics, fracture toughness is strongly governed by microstructural features, and mechanisms such as grain bridging, fiber pull-out, crack deflection, and crack branching play decisive roles. The elongated β-Si₃N₄ grains in Sample 1 actively participate in these mechanisms, promoting crack shielding and increasing fracture resistance.

A direct comparison with literature values for ceramic matrix composites must consider the manufacturing route and testing methodology. The indentation fracture resistance of the optimal Cf/α-Si₃N₄ composite (~ 10.9 MPa m^0.5) is competitive with or exceeds that reported for many C_f/SiC composites fabricated via Chemical Vapor Infiltration (CVI), which typically range from 5 to 7 MPa m^{0.57,14,40,41}. It is noted, however, that C_f/SiC composites produced by Polymer Impregnation and Pyrolysis (PIP)^42,43 can achieve higher fracture toughness (15–18 MPa m^0.5). The significance of the present work lies not in surpassing all existing systems, but in demonstrating that a carefully selected Si₃N₄ polymorph (α-phase) processed via a rapid SPS route can yield a composite with a well-balanced property set—combining respectable fracture resistance, good thermal conductivity, and stable tribological behavior—thereby offering a promising alternative fabrication strategy for tailored thermostructural components.

The superior flexural strength and fracture toughness observed in the C_f/α-Si₃N₄ composite (Sample 1) can be attributed to a synergistic effect of high relative density and a unique self-reinforced microstructure. The in-situ formation of interlocking, needle-like β-Si₃N₄ grains within the α-Si₃N₄ matrix, as confirmed by XRD and FESEM, effectively promotes toughening mechanisms such as crack deflection and fiber pull-out. This microstructure acts similarly to whisker-reinforced composites, where the elongated grains bridge cracks and increase the fracture path complexity. Consequently, the achieved K_IC ​ value of 10.87 MPa m^0.5 not only surpasses that of our other samples but also exceeds the performance of many C_f/SiC composites manufactured by conventional methods like CVI, as referenced in^41,42. This demonstrates the potential of SPS with an α-Si₃N₄ starting powder to create intrinsically toughened composites without the need for external whiskers^{44,45,46,47,48}.

Investigate of thermal conductivity

The thermal conductivity coefficients of the C_f/Si₃N₄ composites, summarized in Table 1, were measured as 0.0156, 0.0131, and 0.0144 cm² s⁻¹ for Sample 1 (C_f/α-Si₃N₄), Sample 2 (C_f/β-Si₃N₄), and Sample 3 (C_f/γ-Si₃N₄), respectively. These values correspond to thermal conductivities of 66, 58, and 53 W m⁻¹ K⁻¹.

Table 1 Thermal diffusivity (α) parameter measured by XFA method for different samples measeared according to ASTM E1461.

A clear correlation between thermal conductivity, relative density, and microstructural integrity is observed. The highest thermal conductivity is achieved in the C_f/α-Si₃N₄ composite, which also exhibits the highest relative density and lowest porosity. Reduced porosity minimizes phonon scattering at pore surfaces and grain boundaries, thereby facilitating more efficient heat transport through the ceramic matrix. In contrast, the lower thermal conductivity of the C_f/γ-Si₃N₄ composite is primarily attributed to its higher residual porosity and reduced densification, which introduce additional phonon scattering centers.

The absolute thermal conductivity values of the C_f/Si₃N₄ composites are lower than those of metallic brake discs; however, they fall within the optimal range for ceramic brake materials. Excessively high thermal conductivity can result in rapid heat dissipation away from the friction interface, potentially reducing braking efficiency, whereas overly low conductivity can cause localized heat accumulation, thermal gradients, and an increased risk of thermal cracking. From this perspective, the moderate thermal conductivity of the C_f/α-Si₃N₄ composite represents a balanced thermal management capability.

The reduced thermal conductivity of the composites relative to monolithic Si₃N₄ is also influenced by the incorporation of carbon fibers, which possess a comparatively low thermal conductivity (~ 3 W m⁻¹ K⁻¹)^26,49,50. Furthermore, the C_f/Si₃N₄ interfaces act as effective phonon barriers due to acoustic mismatch, further limiting heat transport across the composite. Notably, the thermal conductivity of the C_f/γ-Si₃N₄ composite is comparable to that reported for Cf/SiC brake discs fabricated via the CVI process^46,47, indicating that C_f/Si₃N₄ composites processed by SPS can meet the thermal performance requirements of aerospace braking systems.

Overall, the combination of moderate thermal conductivity and high specific heat capacity in the C_f/α-Si₃N₄ composite suggests an enhanced ability to absorb, distribute, and manage frictional heat during braking. This thermal behavior is particularly advantageous for aerospace applications, where resistance to thermal fade and stable performance under severe thermal loading are critical safety requirements.

Quantitative microstructural analysis and toughening mechanisms

While qualitative FESEM observations revealed the presence of elongated, needle‑like β‑Si₃N₄ grains in the α‑phase‑derived composite (Sample 1), a quantitative microstructural analysis was conducted to substantiate their role in enhancing fracture toughness. Image analysis of FESEM micrographs was performed using ImageJ software to determine the aspect ratio and volume fraction of the elongated β‑Si₃N₄ grains formed in situ during SPS.

The analysis revealed an average grain aspect ratio exceeding 5 and a volume fraction of approximately 30–40% for the elongated β‑Si₃N₄ grains in Sample 1. This high fraction of high‑aspect‑ratio grains provides a direct, quantitative explanation for the superior fracture toughness (10.87 MPa·m⁰·⁵) observed in this composite. Unlike Samples 2 and 3, which lack a comparable self‑reinforced microstructure, Sample 1 benefits from multiple, simultaneously active toughening mechanisms that operate across different length scales.

The enhanced fracture resistance can be attributed to the following quantitatively supported mechanisms:

• I. Crack Deflection

The randomly oriented, high‑aspect‑ratio β‑Si₃N₄ grains force propagating cracks to repeatedly change direction, significantly increasing crack path tortuosity and the energy required for crack propagation.

• II. Crack Bridging

Elongated β‑Si₃N₄ grains bridge the crack wake and apply closing stresses that shield the crack tip from the full applied load, thereby delaying catastrophic failure.

• III. Synergy with Fiber Pull‑out

The self‑reinforced ceramic matrix operates synergistically with the carbon fiber reinforcement. While the fibers provide macroscopic energy dissipation through pull‑out and debonding, the elongated β‑Si₃N₄ grains contribute microscale crack shielding, resulting in a highly effective multi‑scale toughening system.

This quantitative microstructural evidence transforms the interpretation of the mechanical results from a simple correlation to a clear causative relationship. It firmly establishes that the α → β phase transformation during SPS is not merely a crystallographic change, but a powerful microstructural engineering strategy for designing intrinsically toughened C_f/Si₃N₄ composites without the need for external whisker reinforcements.

Conclusion

This study demonstrates that the initial phase of Si₃N₄ powder plays a decisive role in controlling the microstructural evolution and resulting properties of C_f/Si₃N₄ composites fabricated by spark plasma sintering. Among the investigated precursors, α-Si₃N₄ enabled the formation of an in-situ self-reinforced β-Si₃N₄ grain network, leading to an optimal balance of fracture toughness (10.87 MPa m^0.5), thermal conductivity (66 W/m K), and stable tribological behavior (COF ≈ 0.46).

The superior performance of the α-derived composite is attributed to the synergistic interaction between the elongated β-Si₃N₄ grains and carbon fiber reinforcement, which promotes crack deflection, crack bridging, and the formation of a stable tribofilm during sliding. In contrast, composites derived from β- and γ-Si₃N₄ powders exhibited inferior densification or excessive hardness without a corresponding improvement in overall performance.

These findings establish a phase-informed microstructural design strategy for C_f/Si₃N₄ composites, demonstrating that controlled α → β transformation during SPS can be effectively exploited as a microstructural engineering tool. The insights gained in this study provide a foundation for developing high-performance non-oxide ceramic composites for demanding thermostructural applications such as aerospace brake.