Microstructure and Fracture Mechanism Investigation of Porous Silicon Nitride–Zirconia–Graphene Composite Using Multi-Scale and In-Situ Microscopy

Silicon nitride–zirconia–graphene composites with high graphene content (5 wt.% and 30 wt.%) were sintered by gas pressure sintering (GPS). The effect of the multilayer graphene (MLG) content on microstructure and fracture mechanism is investigated by multi-scale and in-situ microscopy. Multi-scale microscopy confirms that the phases disperse evenly in the microstructure without obvious agglomeration. The MLG flakes well dispersed between ceramic matrix grains slow down the phase transformation from α to β-Si3N4, subsequent needle-like growth of β-Si3N4 rods and the densification due to the reduction in sintering additives particularly in the case with 30 wt.% MLG. The size distribution of Si3N4 phase shifts towards a larger size range with the increase in graphene content from 5 to 30 wt.%, while a higher graphene content (30 wt.%) hinders the growth of the ZrO2 phase. The composite with 30 wt.% MLG has a porosity of 47%, the one with 5 wt.% exhibits a porosity of approximately 30%. Both Si3N4/MLG composites show potential resistance to contact or indentation damage. Crack initiation and propagation, densification of the porous microstructure, and shift of ceramic phases are observed using in-situ transmission electron microscopy. The crack propagates through the ceramic/MLG interface and through both the ceramic and the non-ceramic components in the composite with low graphene content. However, the crack prefers to bypass ceramic phases in the composite with 30 wt.% MLG.


Introduction
Materials for high temperature applications usually require tailored mechanical properties (e.g., fracture toughness, bending strength), good resistance to thermal shock, creep resistance, high thermal conductivity, as well as good tribological and wear properties [1][2][3][4]. Ceramic materials have been extensively investigated in the previous decades due to their high temperature performance in general. Silicon nitride (Si 3 N 4 ) ceramics have the potential to meet the requirements mentioned above (e.g., low coefficient of thermal expansion (CTE), good thermal conductivity and high strength, resulting in a higher thermal shock ogy reported in this study also provides a potential unique approach to understand the microstructure and mechanical behavior correlation for other complex ceramic systems.

Silicon Nitride-Zirconia-Graphene Composite Preparation
A commercial alpha silicon nitride powder (UBE Corp., Ube, Japan, particle size: 0.6 µm, specific surface area: 4.8 m 2 /g) was used as matrix material. The base powder consisted of 90 wt.% α-Si 3 N 4 , 4 wt.% Al 2 O 3 (Alcoa, A16, Pittsburgh, PA, USA) and 6 wt.% Y 2 O 3 (H.C. Starck, grade C, Goslar, Germany). It was mixed by attrition milling (Union Process, type 01-HD/HDDM, Akron, OH, USA) equipped with zirconia agitator discs and ZrO 2 grinding media (3 vol% Y 2 O 3 stabilized, diameter of 1 mm) in a 750 cm 3 zirconia tank. The milling process was performed at a high rotation speed of 3000 min −1 for 5 h in ethanol [33]. ZrO 2 particles were incorporated into the Si 3 N 4 during the milling, from the abrasion of zirconia balls under controlled conditions. The contribution of ZrO 2 was adjusted between 30 and 42 wt.%. Commercial graphite powder (Aldrich, St. Louis, MO, USA, grain size: 1 µm) was milled intensively in ethanol for 10 h using the same attrition milling system, and subsequently added into powder mixture. The final powder mixture (with 5 wt.% and 30 wt.% MLG) was dried and sieved with a filter with a mesh size of 150 µm. Polyethylene glycol (PEG, 10 wt.%) surfactant and deionized water were added to the powder mixture before sintering. Samples with the dimension 5 mm × 5 mm × 50 mm were pressed by dry pressing at 220 MPa. The GPS was applied to form the final composites in nitrogen atmosphere at 1700 • C and 20 MPa for 3 h.

Multi-Scale and In-Situ Microscopy
High surface quality of composites was achieved by grinding, standard polishing, and ion polishing before performing scanning electron microscopy (SEM) studies. SEM imaging was performed at an operating voltage of 3 kV using a Carl Zeiss NVision 40 tool (Oberkochen, Germany), applying an Energy selective Backscattered (EsB) detector. The sample for X-ray microscopy (XRM) was firstly grinded with a relative flat surface, then a square pillar was prepared with a length of about 3 µm using focused ion beam (FIB) milling. An XRM study was carried out at the U41-XM beamline of the electron storage ring BESSY II, Helmholtz-Zentrum Berlin (Berlin, Germany). The used photon energy was 800 eV. The sample was tilted from −65 • to +65 • with 1 • steps. The tomography was reconstructed by Tomo3D [34] (Almeria and Madrid, Spain), and rendered and sliced using the Tomviz software [35] (Ithaca, NY, USA). Standard lift-out lamellae with a thickness of about 200 nm for transmission electron microscopy (TEM) study were prepared using FIB milling, after local carbon and Pt deposition. TEM (Carl Zeiss Libra 200 Cs, Oberkochen, Germany, with an acceleration voltage of 200 kV) was used to study of the microstructure of the sintered composites. Energy-dispersive X-ray spectroscopy (EDX) was performed on the samples using a detector of Oxford Instruments attached to the TEM. A quantitative analysis of the microstructure from SEM and TEM images was performed in Fiji [36]. SEM images with magnification of 5000× and 10,000×, and TEM images with magnification of 20,000× and 30,000× were used. Si 3 N 4 was treated as rod shape with round cross section, and ZrO 2 was treated as globular shape in the analysis. Si 3 N 4 phases with high aspect ratio larger than 3 were used to calculate the length and diameter of the cross section, while the rest was only used to calculate the diameter of the cross-section. The porosity was measured both by water intrusion porosimetry [21] and mercury intrusion porosimetry [37]. The density was measured applying the Archimedes method. Vickers hardness measurement (hardness tester LECO 700AT, St. Joseph, MI, USA) was performed at loads from 9.81 to 150 N, the dwelling time was 10 s in all cases. The sample for in-situ TEM experiment was H-bar sample with a thickness of about 800 nm prepared by FIB milling. A wedge indenter with a piezo control in a TEM holder was used to perform the in-situ test.

Results and Discussion
A square pillar from the synthesized composite with 30 wt.% MLG was prepared for 3D microstructure studies using XRM. X-ray computed tomography (XCT) data are shown in Video S1 (supplementary materials), two extracted slices are shown in Figure 1. The ZrO 2 phases were easily differentiated by the contrast, while only partial Si 3 N 4 phases were distinguished from the mixed Si 3 N 4 phase and MLG flakes. The size of major ZrO 2 phases is less than 1 µm, the size of Si 3 N 4 phases is less than 0.5 µm. The size of MLG flakes could not be unambiguously determined by XRM. Empty space (open pores) is clearly observed as well from the 3D tomography data as shown in Video S1 (Supplementary Materials).
The sample for in-situ TEM experiment was H-bar sample with a thickness of about 800 nm prepared by FIB milling. A wedge indenter with a piezo control in a TEM holder was used to perform the in-situ test.

Results and Discussion
A square pillar from the synthesized composite with 30 wt.% MLG was prepared for 3D microstructure studies using XRM. X-ray computed tomography (XCT) data are shown in Video S1 (supplementary materials), two extracted slices are shown in Figure 1. The ZrO2 phases were easily differentiated by the contrast, while only partial Si3N4 phases were distinguished from the mixed Si3N4 phase and MLG flakes. The size of major ZrO2 phases is less than 1 µm, the size of Si3N4 phases is less than 0.5 µm. The size of MLG flakes could not be unambiguously determined by XRM. Empty space (open pores) is clearly observed as well from the 3D tomography data as shown in Video S1 (Supplementary Materials). The SEM images on the grinded and ion-polished samples detected using an EsB detector show clear compositional contrast, Si3N4, ZrO2 and MLG flakes are easily distinguished (indicated by arrows in Figure 2). After the GPS process, spheroid ZrO2 particles (mostly c-ZrO2, as indicated by the XRD data in Figures S1 and S2) and thin MLG platelets were successfully incorporated into the Si3N4 matrix for both composites. MLG platelets were embedded and entangled among Si3N4 and ZrO2 phases. Slight agglomeration of ceramic particles was also observed. Hexagonal β-Si3N4 phases (rod-like) were commonly observed in the sintered composite with 5 wt.% MLG, while approximately 2.5 wt.% of α-Si3N4 phases apart from the major β-Si3N4 phases still remained in the sintered composite with 30 wt.% MLG (XRD data in Figures S1 and S2). Rod-like β-Si3N4 phases with a high aspect ratio represent the majority in the microstructure in the sample with 5 wt.% MLG addition, while β-Si3N4 phases with a low aspect ratio are more common in the sample with 30 wt.% MLG addition (Figures 2 and  3). Due to the high porosity, gas phase reactions will also influence the grain growth. In the α-β phase transformation, the liquid phase is crucial in the whole process of the dissolution of the fine α-phase starting powder and subsequent precipitation of the β- The SEM images on the grinded and ion-polished samples detected using an EsB detector show clear compositional contrast, Si 3 N 4 , ZrO 2 and MLG flakes are easily distinguished (indicated by arrows in Figure 2). After the GPS process, spheroid ZrO 2 particles (mostly c-ZrO 2 , as indicated by the XRD data in Figures S1 and S2) and thin MLG platelets were successfully incorporated into the Si 3 N 4 matrix for both composites. MLG platelets were embedded and entangled among Si 3 N 4 and ZrO 2 phases. Slight agglomeration of ceramic particles was also observed. Hexagonal β-Si 3 N 4 phases (rod-like) were commonly observed in the sintered composite with 5 wt.% MLG, while approximately 2.5 wt.% of α-Si 3 N 4 phases apart from the major β-Si 3 N 4 phases still remained in the sintered composite with 30 wt.% MLG (XRD data in Figures S1 and S2). Rod-like β-Si 3 N 4 phases with a high aspect ratio represent the majority in the microstructure in the sample with 5 wt.% MLG addition, while β-Si 3 N 4 phases with a low aspect ratio are more common in the sample with 30 wt.% MLG addition (Figures 2 and 3). Due to the high porosity, gas phase reactions will also influence the grain growth. In the α-β phase transformation, the liquid phase is crucial in the whole process of the dissolution of the fine α-phase starting powder and subsequent precipitation of the β-phase [2]. High content addition of MLG could react with the liquid phase, in particular by the reduction in sintering additives [38][39][40]. Therefore, it slowed down both the densification and the phase transformation from α to β-Si 3 N 4 . The high content of MLG plates slowed down the subsequent needle-like growth of β-Si 3 N 4 rod as well ( Figure 2) by serving as barrier layer. The Si 3 N 4 and ZrO 2 phases were quantitatively analyzed using the SEM images, the results are summarized in Figure 4a,b. Compared with the sample with 5 wt.% MLG addition, the size distribution of Si 3 N 4 phase shifts towards a larger size range for the sample with 30 wt.% MLG addition. The size (diameter) mainly ranges from 100 to 300 nm with 5 wt.% MLG addition and from 100 to 350 nm with 30 wt.% MLG addition, respectively. On the contrary, a higher MLG content (30 wt.% MLG) hinders the growth of ZrO 2 phase (Figure 4b), in which the high content of graphene acts as barrier layer. For 5 wt.% MLG addition, the average diameter and length of hexagonal Si 3 N 4 phases were 221 ± 9 nm and 1496 ± 39 nm, and the average size of spheroid ZrO 2 particles was 867 ± 27 nm. For the sample with 30 wt.% MLG addition, the average diameter and length of hexagonal Si 3 N 4 phases was 249 ± 9 nm and 1363 ± 31 nm, respectively, and the average size of the spheroid ZrO 2 phases was 601 ± 19 nm. The average size of Si 3 N 4 phases observed in this study is smaller than in typical monolithic Si 3 N 4 ceramics sintered at similar conditions [17] because of the addition of MLG resulting in the reduction in the liquid phase and a high residual porosity. The volume ratio of Si 3 N 4 phase to ZrO 2 phase observed by image analysis was about 2.65:1 in the sample with 5 wt.% MLG, and about 2:1 in the sample with 30 wt.% MLG. Although the addition of ZrO 2 into the initial Si 3 N 4 powder can noticeably facilitate the densification process and decrease the sintering temperature [10], open pores were apparently observed in both composites. The interface between ZrO 2 and Si 3 N 4 was continuous without any apparent cracks. Since the open pores are closely associated with graphene platelets, it is expected that the porosity increases with the increase in the graphene content. The porosity data, measured by both water and mercury intrusion porosimetry, are given in Table 1. A porosity close to 50% was observed for the composite with 30 wt.% MLG, which proves that MLG fillers in Si 3 N 4 -ZrO 2 ceramics makes the densification of the composite extremely difficult. Even an addition of 5 wt.% MLG can generate a porous microstructure with about 30% porosity. Correspondingly, the densities are 2.71 g/cm 3 and 1.84 g/cm 3 .
phase [2]. High content addition of MLG could react with the liquid phase, in particular by the reduction in sintering additives [38][39][40]. Therefore, it slowed down both the densification and the phase transformation from α to β-Si3N4. The high content of MLG plates slowed down the subsequent needle-like growth of β-Si3N4 rod as well ( Figure 2) by serving as barrier layer. The Si3N4 and ZrO2 phases were quantitatively analyzed using the SEM images, the results are summarized in Figure 4a,b. Compared with the sample with 5 wt.% MLG addition, the size distribution of Si3N4 phase shifts towards a larger size range for the sample with 30 wt.% MLG addition. The size (diameter) mainly ranges from 100 to 300 nm with 5 wt.% MLG addition and from 100 to 350 nm with 30 wt.% MLG addition, respectively. On the contrary, a higher MLG content (30 wt.% MLG) hinders the growth of ZrO2 phase (Figure 4b), in which the high content of graphene acts as barrier layer. For 5 wt.% MLG addition, the average diameter and length of hexagonal Si3N4 phases were 221 ± 9 nm and 1496 ± 39 nm, and the average size of spheroid ZrO2 particles was 867 ± 27 nm. For the sample with 30 wt.% MLG addition, the average diameter and length of hexagonal Si3N4 phases was 249 ± 9 nm and 1363 ± 31 nm, respectively, and the average size of the spheroid ZrO2 phases was 601 ± 19 nm. The average size of Si3N4 phases observed in this study is smaller than in typical monolithic Si3N4 ceramics sintered at similar conditions [17] because of the addition of MLG resulting in the reduction in the liquid phase and a high residual porosity. The volume ratio of Si3N4 phase to ZrO2 phase observed by image analysis was about 2.65:1 in the sample with 5 wt.% MLG, and about 2:1 in the sample with 30 wt.% MLG. Although the addition of ZrO2 into the initial Si3N4 powder can noticeably facilitate the densification process and decrease the sintering temperature [10], open pores were apparently observed in both composites. The interface between ZrO2 and Si3N4 was continuous without any apparent cracks. Since the open pores are closely associated with graphene platelets, it is expected that the porosity increases with the increase in the graphene content. The porosity data, measured by both water and mercury intrusion porosimetry, are given in Table 1. A porosity close to 50% was observed for the composite with 30 wt.% MLG, which proves that MLG fillers in Si3N4-ZrO2 ceramics makes the densification of the composite extremely difficult. Even an addition of 5 wt.% MLG can generate a porous microstructure with about 30% porosity. Correspondingly, the densities are 2.71 g/cm 3 and 1.84 g/cm 3 .   distributes around both ZrO2 and Si3N4; this is probably the remainder of the precipitated liquid phase. The elemental analysis reveals a low content of oxygen (2 to 4 at%) in the MLG flakes. There is no size difference observed in the TEM study for both composites. The thickness ranges mainly between 5 and 30 nm (thickness of about 20 nm (~ 65 carbon layers) is also confirmed by XRD, Figures S1 and S2), while the length ranges mainly between 100 and 300 nm. The quantitative size distribution of MLG flakes is summarized in Figure 4c,d. The data reveal that the carbon is mostly in the form of thin graphite.    Figure 5a shows a representative optical microscopy image of a Vickers indentation site in dense Si3N4 without MLG. Classical radial cracking (extended cracks as indicated by red arrows) is clearly observed in the micrograph. Figure 5b,c show representative optical microscopy images of the Vickers indentation sites in Si3N4/MLG composites. No classical radial cracks occur in the porous composites, indicating that the Si3N4/MLG composites have potential resistance to contact or indentation damage [41]. The high porosity and shear-weak second phases (MLG) could play important roles in redistributing stress under confined shear in indentation (contact loading), resulting in the suppression of macroscopic (long) cracks. Since high porosity could deteriorate the mechanical properties of the sintered composites, potential approaches (e.g., optimizing MLG content, achieving high density, forming sandwich structure with alternating low and high MLG content layers [42,43]) can be applied to compensate the deteriorated mechanical properties. The morphology of the composites and the elemental distribution were investigated by TEM (Figure 3). MLG flakes distributed and embedded in the Si 3 N 4 -based matrix were clearly identified based on scanning TEM images for both composites. A certain amount of ZrO 2 is located between the Si 3 N 4 grains. The cross-section study of multilayered graphene flakes reveals their main presence between the rod-like Si 3 N 4 particles. Compared with the composite with 30 wt.% MLG, more hexagonal β-Si 3 N 4 grains were observed in the composite with 5 wt.% MLG. ZrO 2 grains show an average size of less than 1 µm in both composites, while the size of silicon nitride rods is about 300 nm in diameter and 800 to 1200 nm in length. As shown in Figure 3, Si, N, C, Zr, Y, Al and O are the major elements detected by EDX in the TEM. Corresponding elemental mappings clearly indicate Si 3 N 4 , ZrO 2 and MLG flakes in both 5 wt.% (Figure 3a) and 30 wt.% (Figure 3b) composites. Y distributes homogenously inside ZrO 2 , indicating the high degree of stabilization, which results in the formation of the cubic phase as proven by XRD. No transformation of ZrO 2 is observed indicating that ZrO 2 has no toughening effect. Al distributes around both ZrO 2 and Si 3 N 4 ; this is probably the remainder of the precipitated liquid phase. The elemental analysis reveals a low content of oxygen (2 to 4 at%) in the MLG flakes. There is no size difference observed in the TEM study for both composites. The thickness ranges mainly between 5 and 30 nm (thickness of about 20 nm (~65 carbon layers) is also confirmed by XRD, Figures S1 and S2), while the length ranges mainly between 100 and 300 nm. The quantitative size distribution of MLG flakes is summarized in Figure 4c,d. The data reveal that the carbon is mostly in the form of thin graphite. Figure 5a shows a representative optical microscopy image of a Vickers indentation site in dense Si 3 N 4 without MLG. Classical radial cracking (extended cracks as indicated by red arrows) is clearly observed in the micrograph. Figure 5b,c show representative optical microscopy images of the Vickers indentation sites in Si 3 N 4 /MLG composites. No classical radial cracks occur in the porous composites, indicating that the Si 3 N 4 /MLG composites have potential resistance to contact or indentation damage [41]. The high porosity and shear-weak second phases (MLG) could play important roles in redistributing stress under confined shear in indentation (contact loading), resulting in the suppression of macroscopic (long) cracks. Since high porosity could deteriorate the mechanical properties of the sintered composites, potential approaches (e.g., optimizing MLG content, achieving high density, forming sandwich structure with alternating low and high MLG content layers [42,43]) can be applied to compensate the deteriorated mechanical properties.   Figure 5b,c show representative optical microscopy images of the Vickers indentation sites in Si3N4/MLG composites. No classical radial cracks occur in the porous composites, indicating that the Si3N4/MLG composites have potential resistance to contact or indentation damage [41]. The high porosity and shear-weak second phases (MLG) could play important roles in redistributing stress under confined shear in indentation (contact loading), resulting in the suppression of macroscopic (long) cracks. Since high porosity could deteriorate the mechanical properties of the sintered composites, potential approaches (e.g., optimizing MLG content, achieving high density, forming sandwich structure with alternating low and high MLG content layers [42,43]) can be applied to compensate the deteriorated mechanical properties. In order to gain an in-depth understanding of the mechanical behavior and the fracture mechanism of sintered composites, in-situ wedge indentation tests were performed as shown in Figures 6 and 7. The corresponding videos from experiments are presented in Videos S2 and S3 (Supplementary Materials). Representative TEM images In order to gain an in-depth understanding of the mechanical behavior and the fracture mechanism of sintered composites, in-situ wedge indentation tests were performed as shown in Figures 6 and 7. The corresponding videos from experiments are presented in Videos S2 and S3 (Supplementary Materials). Representative TEM images selected from an in-situ experiment for a composite with 5 wt.% MLG are shown in Figure 6. Rod-like β-Si 3 N 4 phases with high aspect ratio account for the majority of Si 3 N 4 phase. The positions of major cracks observed in the experiment are highlighted by dashed lines (Figure 6a). Cracks initiated in multiple locations, and propagated along weak interfaces (Figure 6b). Cracks turned towards different directions at the ceramic/MLG interface, within the ceramic phase and within the MLG platelets. Pulled out MLGs are visible in the partially fractured interface, fracture within ceramic phases is frequently observed as well (Figure 6c). The fracture surface typically looks sharp and straight in ceramic phases (Figure 6c-f). As shown in Figure 6f, a small part of the composite was completely delaminated and removed at the end of the experiment. Apart from the cracking process, shift of ceramic grains is also observed. Due to the high porosity, densification process occurs commonly in a relatively homogeneous pace. In the composite with 30 wt.% MLG, β-Si 3 N 4 grains with low aspect ratio were mainly obtained after sintering process (Figure 7). A small amount (about 2.5%) of α-Si 3 N 4 phase remained. The cracks propagated in a much faster pace after initiation. The cracks penetrated easily through the composite mainly within the MLG phase. A long crack within the MLG was quickly observed (step 23, Figure 7c), and subsequently a large fracture interface was formed two steps later (step 25, Figure 7d). Pulled-out MLG components are commonly visible at the fracture surface (Figure 7d). Shift of ceramic grains and a densification process were observed, but with a much faster pace. It seems that the cracks prefer to bypass the ceramic phases in the composite with 30 wt.% MLG. Such a crack propagation behavior could be caused by a high content of the weak carbon phase which forms a three-dimensional network.
The microstructure has a strong influence on the fracture behavior of the Si 3 N 4 /MLG composites. Contact loading (indentation) with highly concentrated loads act on a very small region, resulting in an intensely confined shear. Therefore, highly heterogeneous ceramic matrix composites with shear-weak graphite and high porosity here result in a considerable redistribution of stress in the region below the indentation. The behavior can be described by a distributed shear anelasticity in the form of microstructure-localized shear-sliding along numerous interfaces [41]. The high porosity (28% and 47% for 5 wt.% MLG content and 30 wt.% MLG content, respectively) improves the shear-deformability of the composites even more. This type of dispersed damage caused by significant redistributed stress consequently prevents the formation of long macro cracks (classical radial cracks), as observed in the homogenous Si 3 N 4 ceramics (Figure 5a). Considering the fine scale of ceramic grains (about 200 nm in diameter and about 1µm in length for Si 3 N 4 , less than 1 µm in diameter for ZrO 2 ) and of the carbon-based MLG as reinforcing component (tens of nm in thickness and less than 1 µm in size), it is not surprising that no macro toughening is observed, which requires this kind of carbon-based reinforcing component with larger scale (e.g., carbon fiber with a length of hundreds of micrometers [41]). The fine scale of the reinforcing component (MLG) in the composites studied here results in a small toughening zone relative to the crack size.
Although both Si 3 N 4 /MLG (5 wt.% and 30 wt.%) composites are characterized by a resistance against contact damage ( Figure 5), the content of MLG plays an important role in the micro fracture behavior. In the composite with 30 wt.% MLG, the high content of the weak carbon component forms a three-dimensional network. As summarized in Table S1, a higher volume ratio (about 2:1 for Si 3 N 4 :ZrO 2 compared to 2.65:1) of spheroid shape of ZrO 2 with 260 nm smaller in average diameter, lower aspect ratio (5.47:1 compared to 6.77:1) of Si 3 N 4 phase, and high porosity of about 50% facilitate the crack propagation along the three-dimensional network of the weak carbon component. The typical crack paths are sketched in Figure 8b. Since such a three-dimensional network of MLG is not formed in the composite with 5 wt.% MLG, crack deflection and crack-bridging occur locally (see Figure 8a). The cracks propagate into Si 3 N 4 phase as well. The high porosity in the composites causes a redistribution of stress that shifts ceramic and carbon components, and consequently, leads to the direction change of cracks and new crack initiation. Nanomaterials 2021, 11, x FOR PEER REVIEW 9 of 13

Conclusions
In summary, porous silicon nitride-zirconia-graphene composites with high graphene content (5 wt.% and 30 wt.%) were sintered by GPS. Multi-scale microscopy confirmed that all components had been dispersed evenly in the composite without obvious agglomeration. Graphene caused a reduction in the sintering additives, and therefore an increased porosity. MLG fillers in Si3N4-ZrO2 ceramics hampered the densification of the composites, observable by porosity values of about 30% (5 wt.% MLG) and about 50% (30 wt.% MLG). A quantitative analysis on SEM and TEM images revealed that the size distribution of the Si3N4 phase shifts towards a larger size range with increased graphene content. The higher porosity in the composite with the higher graphene content (30 wt.%) hinders the growth of the ZrO2 phase. The average diameters of Si3N4 grains were 221 ± 9 nm (5 wt.% MLG) and 249 ± 9 nm (30 wt.% MLG), and the average sizes of spheroid ZrO2 particles were 867 ± 27 nm (5 wt.% MLG) and 601 ± 19 nm (30 wt.% MLG). The volume ratio of Si3N4 and ZrO2 phase was about 2.65:1 in the composite with 5 wt.% MLG, and about 2:1 in the composite with 30 wt.% MLG. The MLG flakes well dispersed between ceramic grains slowed down the phase transformation from α to β-Si3N4 and the subsequent longitudinal growth of β-Si3N4 rods due to the interaction with the sintering additives, particularly visible for the composite with 30 wt.% MLG. Both Si3N4/MLG composites show a potential resistance against contact or indentation damage due to significant redistributed stress. Crack initiation and propagation, densification of the porous material and shift of ceramic components were observed in insitu experiments. The cracks prefer to bypass ceramic components in the composite with 30 wt.% MLG, which is mainly caused by the formed three-dimensional network with a high content of weak carbon components. Without such a three-dimensional network of MLG in the composite with 5% MLG, crack deflection and crack-bridging occur locally. The cracks propagate into the Si3N4 phase as well. Redistribution of stress moves the ceramic phases and the carbon phase, which leads to the change of crack direction and to new crack initiation. Multi-scale and in-situ microscopy as a combined methodology reported in this study also provides a potential unique approach to understand the microstructure and mechanical behavior correlation for complex ceramic systems.

Conclusions
In summary, porous silicon nitride-zirconia-graphene composites with high graphene content (5 wt.% and 30 wt.%) were sintered by GPS. Multi-scale microscopy confirmed that all components had been dispersed evenly in the composite without obvious agglomeration. Graphene caused a reduction in the sintering additives, and therefore an increased porosity. MLG fillers in Si 3 N 4 -ZrO 2 ceramics hampered the densification of the composites, observable by porosity values of about 30% (5 wt.% MLG) and about 50% (30 wt.% MLG). A quantitative analysis on SEM and TEM images revealed that the size distribution of the Si 3 N 4 phase shifts towards a larger size range with increased graphene content. The higher porosity in the composite with the higher graphene content (30 wt.%) hinders the growth of the ZrO 2 phase. The average diameters of Si 3 N 4 grains were 221 ± 9 nm (5 wt.% MLG) and 249 ± 9 nm (30 wt.% MLG), and the average sizes of spheroid ZrO 2 particles were 867 ± 27 nm (5 wt.% MLG) and 601 ± 19 nm (30 wt.% MLG). The volume ratio of Si 3 N 4 and ZrO 2 phase was about 2.65:1 in the composite with 5 wt.% MLG, and about 2:1 in the composite with 30 wt.% MLG. The MLG flakes well dispersed between ceramic grains slowed down the phase transformation from α to β-Si 3 N 4 and the subsequent longitudinal growth of β-Si 3 N 4 rods due to the interaction with the sintering additives, particularly visible for the composite with 30 wt.% MLG. Both Si 3 N 4 /MLG composites show a potential resistance against contact or indentation damage due to significant redistributed stress. Crack initiation and propagation, densification of the porous material and shift of ceramic components were observed in in-situ experiments. The cracks prefer to bypass ceramic components in the composite with 30 wt.% MLG, which is mainly caused by the formed three-dimensional network with a high content of weak carbon components. Without such a three-dimensional network of MLG in the composite with 5% MLG, crack deflection and crack-bridging occur locally. The cracks propagate into the Si 3 N 4 phase as well. Redistribution of stress moves the ceramic phases and the carbon phase, which leads to the change of crack direction and to new crack initiation. Multi-scale and in-situ microscopy as a combined methodology reported in this study also provides a potential unique approach to understand the microstructure and mechanical behavior correlation for complex ceramic systems.
Supplementary Materials: The following are available online at https://www.mdpi.com/2079-499 1/11/2/285/s1, Figure S1: XRD data from the sintered silicon nitride-zirconia-graphene composite with 5 wt.% MLG, Figure S2: XRD data from the sintered silicon nitride-zirconia-graphene composite with 30 wt.% MLG, Table S1: Microstructure comparison for silicon-zirconia-graphene composites, Video S1: Tomography, Video S2: In-situ TEM for the composite with 5 wt.% MLG, Video S3: In-situ TEM for the composite with 30 wt.% MLG. Funding: This research was funded by the Hungarian National Research Development and Innovation Office (projects NK-FIH NN 127723 and NKFIH-NNE 129976), and DFG in Germany (project number 397380564). This research was funded also by the "Research Centre of Advanced Materials and Technologies for Recent and Future Applications PROMATECH", ITMS 26220220186, supported by the Operational Program "Research and Development" financed through European Regional Development Fund.

Data Availability Statement:
The data presented in this study are available on request from the corresponding author. This data is not publicly available due to excessive size and complex format.