Effects of β-Si3N4 Seeds on Microstructure and Performance of Si3N4 Ceramics in Semiconductor Package

Among the various ceramic substrate materials, Si3N4 ceramics have demonstrated high thermal conductivity, good thermal shock resistance, and excellent corrosion resistance. As a result, they are well-suited for semiconductor substrates in high-power and harsh conditions encountered in automobiles, high-speed rail, aerospace, and wind power. In this work, Si3N4 ceramics with various ratios of α-Si3N4 and β-Si3N4 in raw powder form were prepared by spark plasma sintering (SPS) at 1650 °C for 30 min under 30 MPa. When the content of β-Si3N4 was lower than 20%, with the increase in β-Si3N4 content, the ceramic grain size changed gradually from 1.5 μm to 1 μm and finally resulted in 2 μm mixed grains. However, As the content of β-Si3N4 seed crystal increased from 20% to 50%, with the increase in β-Si3N4 content, the ceramic grain size changed gradually from 1 μm and 2 μm to 1.5 μm. Therefore, when the content of β-Si3N4 in the raw powder is 20%, the sintered ceramics exhibited a double-peak structure distribution and the best overall performance with a density of 97.5%, fracture toughness of 12.1 MPa·m1/2, and a Vickers hardness of 14.5 GPa. The results of this study are expected to provide a new way of studying the fracture toughness of silicon nitride ceramic substrates.


Introduction
The development of third-generation semiconductors has induced a huge demand for ceramic substrates for power-integrated circuits [1][2][3][4]. Conventionally, Al 2 O 3 has been utilized as the substrate in power-integrated circuits due to its cost-effectiveness. However, Al 2 O 3 has a low toughness (3-5 MPa·m 1/2 ) and thermal conductivity (18-24 W·m −1 ·K −1 ) and cannot effectively satisfy the requirements of high-power demand in energy and other emerging fields [5]. AlN is another ceramic substrate material that shows high thermal conductivity (150-270 W·m −1 ·K −1 ), but it has the drawback of low fracture toughness (typically 3-3.5 MPa·m 1/2 ). When used in harsh working conditions of thermal shock and impact, such as those encountered in automobiles and wind power, AlN substrates are typically used in combination with plastic shock absorbers, which affects device miniaturization [6]. Due to their high theoretical thermal conductivity (320 W·m −1 ·K −1 ) and good fracture toughness (~10 MPa·m 1/2 ) [7], Si 3 N 4 ceramics have been regarded as the most promising alternatives for Al 2 O 3 and AlN. However, the performances of Si 3 N 4 ceramics currently available are far from theoretical optimization, which has limited their further applications.

Method
A mixture comprising 20 g of 1# and 2# Si 3 N 4 was introduced into a beaker as indicated in Table 1. Subsequently, 360 µL of a solution containing PEG 400, ammonium citrate (dispersant), and 600 µL of PAA (adhesive) were added. To facilitate the sintering process, a total of 1.4 g of Y 2 O 3 and 0.6 g of Al 2 O 3 were introduced. The solution's pH of 9.8 was modified and subsequently introduced into a polyurethane ball milling tank, containing zirconia grinding balls in a ball-to-powder weight ratio of 3:1. The wet-mixing process was carried out using pure water as the medium, and ball milling for 24 h. The compositions of various samples are provided in Table 1. Following ball milling, the slurry underwent vacuum drying at 120 • C for 24 h. After that, an 80-mesh sieve was utilized to obtain the Si 3 N 4 composite precursor. The precursor was then subjected to pre-sintering at 800 • C for 2 h in a tube furnace, under an atmosphere of N 2 (OTF-1200X, Hefei Kejing Materials Co., Ltd., China). After pre-sintering, the product was again sieved through an 80-mesh sieve to acquire Si 3 N 4 composite particles. Finally, the as-prepared Si 3 N 4 composite was inserted into a graphite mold and subjected to Spark Plasma Sintering (SPS) at 1650 • C for 30 min while being subjected to a pressure of 30 MPa (SPS-5T-5-III, Shanghai Huachen Technology Co., Ltd., Shanghai, China). As shown in Figure 1.

Method
A mixture comprising 20 g of 1# and 2# Si3N4 was introduced into a beaker as indicated in Table 1. Subsequently, 360 μL of a solution containing PEG 400, ammonium citrate (dispersant), and 600 μL of PAA (adhesive) were added. To facilitate the sintering process, a total of 1.4 g of Y2O3 and 0.6 g of Al2O3 were introduced. The solution's pH of 9.8 was modified and subsequently introduced into a polyurethane ball milling tank, containing zirconia grinding balls in a ball-to-powder weight ratio of 3:1. The wet-mixing process was carried out using pure water as the medium, and ball milling for 24 h. The compositions of various samples are provided in Table 1. Following ball milling, the slurry underwent vacuum drying at 120 °C for 24 h. After that, an 80-mesh sieve was utilized to obtain the Si3N4 composite precursor. The precursor was then subjected to presintering at 800 °C for 2 h in a tube furnace, under an atmosphere of N2 (OTF-1200X, Hefei Kejing Materials Co., Ltd., China). After pre-sintering, the product was again sieved through an 80-mesh sieve to acquire Si3N4 composite particles. Finally, the as-prepared Si3N4 composite was inserted into a graphite mold and subjected to Spark Plasma Sintering (SPS) at 1650 °C for 30 min while being subjected to a pressure of 30 MPa (SPS-5T-5-Ⅲ , Shanghai Huachen Technology Co., Ltd., Shanghai, China). As shown in Figure 1.

Measurement
The sample density was measured by using the Archimedes method. The samples were analyzed using a Bruker D8 X-ray diffractometer equipped with an energy disper-

Measurement
The sample density was measured by using the Archimedes method. The samples were analyzed using a Bruker D8 X-ray diffractometer equipped with an energy dispersive spectrometer (XRD, Karlsruhe, Germany) and a field-emission scanning electron microscope (SEM, FEI-Nova Nano 450, Hillsboro, OR, USA) to ascertain their phase compositions and fracture morphologies, respectively. The hardness of polished Si 3 N 4 samples was tested using a Vickers hardness meter (HVS-50, Shanghai Wanheng Precision Instrument Co., Ltd., Shanghai, China) under a pressure of 1000 N for 15 s. Each sample was exposed to five replicates and the average value was taken as the final result. The fracture toughness of Si 3 N 4 (K IC ) was determined by using the Vickers hardness indentation method. The crack lengths of the samples were also determined. The fracture toughness can be calculated by: where K IC is the fracture toughness (kgf·m −3/2 ), P is the load (kgf), and c is the diagonal length of the indentation crack (mm). 1 kgf·m −3/2 = 0.31 MPa·m 1/2 . where KIC is the fracture toughness (kgf·m −3/2 ), P is the load (kgf), and c is the diag length of the indentation crack (mm). 1 kgf·m −3/2 = 0.31 MPa·m 1/2 .

Phase Composition of Seed Crystal and Microstructure of Samples
XRD analyses were utilized for investigating the phase composition of the raw p ders, as shown in Figure 2. The results indicate that both 1# and 2# samples were c posed of hexagonal α-Si3N4 (JCPDS No.74-0554, a = b = 7.765 nm, c = 5.622 nm, α = 120°, γ = 90°) and hexagonal β-Si3N4 (JCPDS No.72-1308, a = b = 7.608 nm, c = 2.911 α = β = 120°, γ = 90°). No other phases were observed. Among them, the main pha 1# Si3N4 powder was α-Si3N4, while the main phase of 2# Si3N4 powder was β-Si3N4. Figure 3 shows SEM images of raw powder samples of α-rich Si3N4 and β-rich S Raw powder samples of α-rich Si3N4 comprised irregularly shaped particles that wer served to aggregate (size = 0.5-1 μm), as shown in Figure 3a. Combined with the ana of XRD patterns (Figure 2), it can be concluded that the observed rod-like powder w Si3N4. Raw powder samples of β-rich Si3N4 also comprised irregularly shaped part with aggregation (size = 0.5-1 μm), as shown in Figure 3b. Overall, the two powders consistent shapes and sizes, and the particles were found to be well dispersed with n gible agglomeration.   Raw powder samples of α-rich Si 3 N 4 comprised irregularly shaped particles that were observed to aggregate (size = 0.5-1 µm), as shown in Figure 3a. Combined with the analysis of XRD patterns (Figure 2), it can be concluded that the observed rod-like powder was β-Si 3 N 4 . Raw powder samples of β-rich Si 3 N 4 also comprised irregularly shaped particles with aggregation (size = 0.5-1 µm), as shown in Figure 3b. Overall, the two powders had consistent shapes and sizes, and the particles were found to be well dispersed with negligible agglomeration.

Effects of Seed Crystal Composition on Phase Composition and Microstructure of Ceramic Samples
XRD patterns of samples with different contents of β-Si3N4 seed crystal (SN0~SN4) prepared by SPS at 1100 °C exhibited characteristic peaks corresponding to hexagonal α-   Figure 5 illustrates the grain size changes in SN0~SN4 ceramic samples prepared by SPS at 1100 °C. In the cases with few amounts of β-Si3N4 seeds, e.g., sample SN0 ( Figure  5a), the grain size is relatively small; in the cases with a few amounts of β-Si3N4, e.g., sample SN1~SN3 (Figure 5b-d), some grains become smaller while the other grains grew, the grain size variation in SN4 samples is small.     Figure 5 illustrates the grain size changes in SN0~SN4 ceramic samples prepared by SPS at 1100 °C. In the cases with few amounts of β-Si3N4 seeds, e.g., sample SN0 (Figur 5a), the grain size is relatively small; in the cases with a few amounts of β-Si3N4, e.g., sam ple SN1~SN3 (Figure 5b-d), some grains become smaller while the other grains grew, th grain size variation in SN4 samples is small.  Figure 5 illustrates the grain size changes in SN0~SN4 ceramic samples prepared by SPS at 1100 • C. In the cases with few amounts of β-Si 3 N 4 seeds, e.g., sample SN0 (Figure 5a), the grain size is relatively small; in the cases with a few amounts of β-Si 3 N 4 , e.g., sample SN1~SN3 (Figure 5b-d), some grains become smaller while the other grains grew, the grain size variation in SN4 samples is small. No characteristic peaks of Al2O3 and Y2O3 sintering aids were detected. Complete phase transition of α-Si3N4 into β-Si3N4 and transition of Al2O3 and Y2O3 into Y4SiAlO8N solid solution were observed in the composite powder. Since the bond lengths of Si-N and Al-O were 0.174 nm and 0.175 nm, respectively, it is speculated that the Al-O bond may replace the Si-N bond during the sintering process. This resulted in the generation of Y4Si-AlO8N solid solution via reaction with Y2O3, suggesting that Y, Al, and O were all present in the lattice structure of β-Si3N4 giving rise to an interfacial phase. This speculation is consistent with the observations from earlier reports [28,29]. were all present in the lattice structure of β-Si 3 N 4 giving rise to an interfacial phase. This speculation is consistent with the observations from earlier reports [28,29]. The grain size distribution of Si3N4 significantly impacts its fracture toughness. Fig  ure 7 illustrates the fracture morphologies of SN0-SN4 samples prepared by SPS at 165 °C. As can be observed in Figure 7a, the SN0 sample comprised matrix grains (with ~1 μm granular grains), large grains (rod-like grains with a length of about 2 μm, width o about 1 μm), and abundant pores (~2 μm), a uniform distribution of large and small grain was observed. The SN1 sample comprised matrix grains (~1.3 μm granular grains), larg grains (rod-like grains with a length of about 2 μm, width of about 1 μm), and a sma number of pores (~1 μm), as shown in Figure 7b, the size distribution exhibited two peak where again the large and small grains were uniformly distributed. The SN2 sample com prised matrix grains (with ~1.3 μm granular grains and a small amount of rod-like grain with a length of about 2.2 μm, width of about 1 μm), large grains (rod-like grains with length of about 6 μm, width of about 2.3 μm), and a small number of pores (~1 μm), a shown in Figure 7c, the size distribution exhibited two peaks, herein, the large and sma grains were uniformly distributed in an interlocked structure. The SN3 sample comprise matrix grains (with ~1.3 μm granular grains and a small amount of rod-like grains with length of about 1 μm, width of about 0.5 μm), large grains (rod-like grains (length of abou 8 μm, width of about 2.2 μm)), and a small number of pores (~1 μm), as shown in Figur 7d, similar to the case of SN2, large and small grains were uniformly distributed in a interlocked structure. The SN4 sample comprised matrix grains (with granular grains (~ μm) and rod-like grains with a length of about 1 μm, width of about 0.5 μm), large grain (with granular grains (~1.5 μm) and rod-like grains with a length of about 1.5 μm, widt of about 0.5 μm), and a small number of pores (~1 μm), as shown in Figure 5e, similar t the other samples, the large and small grains were uniformly distributed. Additionally sample failure was dominated by intergranular fracture.

Effects of Seed Crystal Composition on Phase Composition and Microstructure of Ceramic Samples
With the increase in β-Si3N4 seed content, the grain distribution gradually change from a single peak distribution (Figure 7a) to a bimodal distribution (Figure 7c). Whe excess β-Si3N4 seeds were introduced, the grain distribution changed from a bimodal di tribution (Figure 7c) back to a unimodal distribution (Figure 7e). The grain size distribution of Si 3 N 4 significantly impacts its fracture toughness. Figure 7 illustrates the fracture morphologies of SN0-SN4 samples prepared by SPS at 1650 • C. As can be observed in Figure 7a Figure 7c, the size distribution exhibited two peaks, herein, the large and small grains were uniformly distributed in an interlocked structure. The SN3 sample comprised matrix grains (with~1.3 µm granular grains and a small amount of rod-like grains with a length of about 1 µm, width of about 0.5 µm), large grains (rod-like grains (length of about 8 µm, width of about 2.2 µm)), and a small number of pores (~1 µm), as shown in Figure 7d, similar to the case of SN2, large and small grains were uniformly distributed in an interlocked structure. The SN4 sample comprised matrix grains (with granular grains (~1 µm) and rod-like grains with a length of about 1 µm, width of about 0.5 µm), large grains (with granular grains (~1.5 µm) and rod-like grains with a length of about 1.5 µm, width of about 0.5 µm), and a small number of pores (~1 µm), as shown in Figure 5e, similar to the other samples, the large and small grains were uniformly distributed. Additionally, sample failure was dominated by intergranular fracture.
With the increase in β-Si 3 N 4 seed content, the grain distribution gradually changed from a single peak distribution (Figure 7a) to a bimodal distribution (Figure 7c). When excess β-Si 3 N 4 seeds were introduced, the grain distribution changed from a bimodal distribution (Figure 7c) back to a unimodal distribution (Figure 7e).
The elemental composition of the SN0-SN4 samples was verified through the utilization of an EDX analysis, as shown in Table 2. The Table demonstrates a  The elemental composition of the SN0-SN4 samples was verified through the utilization of an EDX analysis, as shown in Table 2. The Table demonstrates a uniform distribution of Si, N, Al, Y, and O elements within the silicon nitride matrix. The proportion of each element in SN0~SN4 ceramic samples is similar.

Effect of β-Si 3 N 4 Seed on the Sintering and Grain Growth of Silicon Nitride Ceramics
During the liquid phase sintering process, α-Si 3 N 4 in the original powder gradually dissolves into the liquid phase aid and diffuses. Its reprecipitation mainly occurs in two ways. The first is through uniform nucleation and growth, where small-sized β-Si 3 N 4 particles are formed. The second involves non-uniform nucleation and growth in the high-energy crystal plane (rod-shaped axis) of β-Si 3 N 4 seeds, which causes the β-Si 3 N 4 seeds to grow into rod-shaped crystals.
The influence of crystal seed phase composition on the microstructure of ceramics is shown in Figure 7. As shown in Figure 8a, when 0% β-Si 3 N 4 crystal seeds are introduced, α-Si 3 N 4 mainly transforms into β-Si 3 N 4 following a uniform nucleation growth mode, which results in the formation of uniform grains with fine grain size (Figure 7a). duced, during the dissolution precipitation process, α-Si3N4 exhibits both uniform nucleation and growth as well as non-uniform nucleation. According to the Ostwald ripening theory [30,31], β-Si3N4 particles in close contact with α-Si3N4 precipitate on the surface of the original β-Si3N4 seeds, leading to the formation of large grains. At the same time, a portion of α-Si3N4 precipitates and grows uniformly, forming an interlocking structure with the generated large extension grains (Figure 7c).
As shown in Figure 8c, when an excess amount (50%) of β-Si3N4 crystal seeds is introduced, non-uniform nucleation of β-Si3N4 occurs during the dissolution precipitation process. Due to the introduction of excessive β-Si3N4 crystal seeds into the original powder, the β-Si3N4 crystal seeds come into contact with each other, resulting in a decrease in α-Si3N4 on the surface of the β-Si3N4 crystal seeds. The large particles formed by precipitation on the surface of the β-Si3N4 crystal seeds are reduced, which is similar in size to the β-Si3N4 particles formed by uniform nucleation and growth, resulting in a more uniform size of grain growth (Figure 7e).  (c) 50% of β-Si 3 N 4 seed crystal. ("early stage" refers to the original powder before sintering; "medium term" refers to the microstructure change of silicon nitride ceramics sintered at 1100 • C; "later period" refers to the microstructure change of silicon nitride ceramics sintered at 1650 • C). Figure 8b, when an appropriate amount (20%) of β-Si 3 N 4 seeds is introduced, during the dissolution precipitation process, α-Si 3 N 4 exhibits both uniform nucleation and growth as well as non-uniform nucleation. According to the Ostwald ripening theory [30,31], β-Si 3 N 4 particles in close contact with α-Si 3 N 4 precipitate on the surface of the original β-Si 3 N 4 seeds, leading to the formation of large grains. At the same time, a portion of α-Si 3 N 4 precipitates and grows uniformly, forming an interlocking structure with the generated large extension grains (Figure 7c).

As shown in
As shown in Figure 8c, when an excess amount (50%) of β-Si 3 N 4 crystal seeds is introduced, non-uniform nucleation of β-Si 3 N 4 occurs during the dissolution precipitation process. Due to the introduction of excessive β-Si 3 N 4 crystal seeds into the original powder, the β-Si 3 N 4 crystal seeds come into contact with each other, resulting in a decrease in α-Si 3 N 4 on the surface of the β-Si 3 N 4 crystal seeds. The large particles formed by precipitation on the surface of the β-Si 3 N 4 crystal seeds are reduced, which is similar in size to the β-Si 3 N 4 particles formed by uniform nucleation and growth, resulting in a more uniform size of grain growth (Figure 7e). Figure 9 shows the trends of relative density of SN0~SN4 prepared by SPS at 1650 • C. As the content of β-Si 3 N 4 seed crystal increased from 0% to 20%, the relative density of Si 3 N 4 samples increased from 96.7% to 97.9%. This was suggestive of the fact that grain size distribution with two peaks favors high compactness of ceramics (see fracture morphology, Figure 7). As the content of β-Si 3 N 4 seed crystal increased from 20% to 50%, the relative density of Si 3 N 4 samples decreased from 97.9% to 96.5%, which may be attributed to the degrading grain size distribution with two peaks (see fracture morphology, Figure 7).  Figure 9 shows the trends of relative density of SN0~SN4 prepared by SPS at 1650 °C. As the content of β-Si3N4 seed crystal increased from 0% to 20%, the relative density of Si3N4 samples increased from 96.7% to 97.9%. This was suggestive of the fact that grain size distribution with two peaks favors high compactness of ceramics (see fracture morphology, Figure 7). As the content of β-Si3N4 seed crystal increased from 20% to 50%, the relative density of Si3N4 samples decreased from 97.9% to 96.5%, which may be attributed to the degrading grain size distribution with two peaks (see fracture morphology, Figure  7). Figure 9. Effects of content of β-Si3N4 seed crystal on relative density of Si3N4 prepared by SPS at 1650 °C. Figure 10 shows the Vickers hardness of SN0-SN4 samples prepared by SPS at 1650 °C. As the content of β-Si3N4 seed crystal was increased from 0% to 20%, the Vickers hardness of Si3N4 samples increased from 13.8 GPa to 14.5 GPa, emphasizing the significance of two peaks in the grain size distribution (see fracture morphology, Figure 7). When the content of β-Si3N4 seed crystal was increased from 20% to 50%, the Vickers hardness of Si3N4 samples decreased from 14.5 GPa to 13.4 GPa, which was attributed to degradation in the grain size distribution with two peaks and grain interlocking structure (see fracture morphology, Figure 7). Hence, it can be concluded that the two peaks observed in the grain size distribution have a major influence on the mechanical properties of the samples (see fracture morphology, Figure 4). Among all the samples tested, the SN2 sample exhibited the highest Vickers hardness (14.5 GPa).   Figure 10 shows the Vickers hardness of SN0-SN4 samples prepared by SPS at 1650 • C. As the content of β-Si 3 N 4 seed crystal was increased from 0% to 20%, the Vickers hardness of Si 3 N 4 samples increased from 13.8 GPa to 14.5 GPa, emphasizing the significance of two peaks in the grain size distribution (see fracture morphology, Figure 7). When the content of β-Si 3 N 4 seed crystal was increased from 20% to 50%, the Vickers hardness of Si 3 N 4 samples decreased from 14.5 GPa to 13.4 GPa, which was attributed to degradation in the grain size distribution with two peaks and grain interlocking structure (see fracture morphology, Figure 7). Hence, it can be concluded that the two peaks observed in the grain size distribution have a major influence on the mechanical properties of the samples (see fracture morphology, Figure 4). Among all the samples tested, the SN2 sample exhibited the highest Vickers hardness (14.5 GPa). Figure 11 shows the fracture toughness of SN0-SN4 samples prepared by SPS at 1650 • C. As the content of β-Si 3 N 4 seed crystal was increased from 0% to 20%, the fracture toughness of Si 3 N 4 samples increased from 9.2 MPa·m 1/2 to 12.1 MPa·m 1/2 . This was ascribed to the emergence of intergranular fracture (see fracture morphology, Figure 7). As the content of β-Si 3 N 4 seed crystal was increased from 20% to 50%, the fracture toughness of Si 3 N 4 samples decreased from 12.1 MPa·m 1/2 to 9.5 MPa·m 1/2 (see fracture morphology, Figure 7). The decrease in Vickers hardness may be related to the degradation in the grain size distribution with two peaks and grain interlocking structure. Among all samples examined, the SN2 sample exhibited the highest fracture toughness (12.1 MPa·m 1/2 ). This was attributed to its unique grain size distribution with two peaks and grain interlocking structure.

Effects of Seed Crystal Composition on Sample Performances
Si3N4 samples decreased from 14.5 GPa to 13.4 GPa, which was attributed to d in the grain size distribution with two peaks and grain interlocking structure (s morphology, Figure 7). Hence, it can be concluded that the two peaks obser grain size distribution have a major influence on the mechanical properties of t (see fracture morphology, Figure 4). Among all the samples tested, the SN2 sam ited the highest Vickers hardness (14.5 GPa).   Figure 11 shows the fracture toughness of SN0-SN4 samples prepared by SPS at 1 °C. As the content of β-Si3N4 seed crystal was increased from 0% to 20%, the fract toughness of Si3N4 samples increased from 9.2 MPa·m 1/2 to 12.1 MPa·m 1/2 . This was cribed to the emergence of intergranular fracture (see fracture morphology, Figure 7). the content of β-Si3N4 seed crystal was increased from 20% to 50%, the fracture toughn of Si3N4 samples decreased from 12.1 MPa·m 1/2 to 9.5 MPa·m 1/2 (see fracture morpholo Figure 7). The decrease in Vickers hardness may be related to the degradation in the gr size distribution with two peaks and grain interlocking structure. Among all samples amined, the SN2 sample exhibited the highest fracture toughness (12.1 MPa·m 1/2 ). T was attributed to its unique grain size distribution with two peaks and grain interlock structure. Figure 11. Effects of content of β-Si3N4 seed crystal on fracture toughness of Si3N4 prepared by at 1650 °C.
In the SN0 sample, due to the presence of small and equiaxed β-Si3N4 grains, the cr deflection during fracturing was low, which resulted in low fracture toughness. In SN1 sample, the β-Si3N4 grains consisted of small and equiaxed rod-like grains, which a led to low fracture toughness. However, for the SN2 sample, the grain size distribut exhibited two peaks, where the large and small grains formed an interlocking structu This feature resulted in low porosity and imparted the sample with high relative den (97.9%) and the highest Vickers hardness (14.5 GPa) among all the samples, as shown Figure 7. Herein, the heterogeneous precipitation of α-Si3N4 on the surface of β-Si3N4 s crystals led to the formation of large, rod-like grains, which activated grain bridging extrusion, and ultimately resulted in the self-toughening of Si3N4 samples. The fract toughness of the SN2 sample reached 12.1 MPa·m 1/2 , as shown in Figure 10. With exc β-Si3N4 seed crystals (30-50%), the original β-Si3N4 seed crystals had consistent sizes w β-Si3N4 particles generated by homogeneous phase transition of α-Si3N4, resulting in crack deflection during fracturing, which drastically reduced the fracture toughness.

Conclusions
Si3N4 ceramics with different contents of β-Si3N4 seed crystal were prepared using spark plasma sintering method, and the correlation of the β-Si3N4 content with hardn In the SN0 sample, due to the presence of small and equiaxed β-Si 3 N 4 grains, the crack deflection during fracturing was low, which resulted in low fracture toughness. In the SN1 sample, the β-Si 3 N 4 grains consisted of small and equiaxed rod-like grains, which also led to low fracture toughness. However, for the SN2 sample, the grain size distribution exhibited two peaks, where the large and small grains formed an interlocking structure. This feature resulted in low porosity and imparted the sample with high relative density (97.9%) and the highest Vickers hardness (14.5 GPa) among all the samples, as shown in Figure 7. Herein, the heterogeneous precipitation of α-Si 3 N 4 on the surface of β-Si 3 N 4 seed crystals led to the formation of large, rod-like grains, which activated grain bridging and extrusion, and ultimately resulted in the self-toughening of Si 3 N 4 samples. The fracture toughness of the SN2 sample reached 12.1 MPa·m 1/2 , as shown in Figure 10. With excess β-Si 3 N 4 seed crystals (30-50%), the original β-Si 3 N 4 seed crystals had consistent sizes with β-Si 3 N 4 particles generated by homogeneous phase transition of α-Si 3 N 4 , resulting in low crack deflection during fracturing, which drastically reduced the fracture toughness.

Conclusions
Si 3 N 4 ceramics with different contents of β-Si 3 N 4 seed crystal were prepared using the spark plasma sintering method, and the correlation of the β-Si 3 N 4 content with hardness and fracture toughness was investigated. The grain growth in Si 3 N 4 was found to be controlled by the dissolution-precipitation process. All the as-prepared samples comprised small particles (diameter about 1 µm) and large particles (about 2 µm). Additionally, the large and small grains were observed to be uniformly distributed. Results showed that a suitable content of (≤20%) β-Si 3 N 4 seed crystal could induce the phase transition from α-Si 3 N 4 to rod-like β-Si 3 N 4 , which resulted in the grain size distribution exhibiting two peaks. At this optimized β-Si 3 N 4 seed crystal content of 20%, the best overall mechanical performance was achieved. At higher content of β-Si 3 N 4 seed crystal (>50%), the β-Si 3 N 4 particles showed a tendency to be mutually constrained during sintering, with uniform size distribution and lower performance. Among samples prepared by the spark plasma sintering method (1650 • C, 30 MPa, 30 min), the sample with a β-Si 3 N 4 seed crystal content of 20% exhibited the best overall performance with a density of 97.9%, hardness of 14.5 GPa, and fracture toughness of 12.1 MPa·m 1/2 . The optimized Si 3 N 4 in this work (using 20% β-Si 3 N 4 as seed crystal) presents the highest toughness but the lowest hardness among previous works [32][33][34][35] as shown in Table 3. This work may provide a potential application of ceramics substrate in highspeed rail and new energy vehicles. Further verification of the actual use of silicon nitride ceramics by researchers in the field of automotive development is needed.   Informed Consent Statement: Written informed consent has been obtained from the patient(s) to publish this paper.
Data Availability Statement: All data generated or analyzed during this study are included in this published article.