You Zhou

The room‐temperature plastic deformation behavior of single crystals of high‐purity β‐silicon nitride has been investigated by micropillar compression tests as a function of loading‐axis orientation and specimen size. Prism < c > slip ({010}[0001] slip) is identified to be the only operative slip system under uniaxial compression at room temperature. The critical resolved shear stress (CRSS) value for prism < c > slip does not depend much on the specimen size and loading‐axis orientation, with the average CRSS value of 1.66 ± 0.22 GPa. The [0001] dislocations on the prism slip planes exhibit a strong tendency to align along their screw orientation, indicating the existence of a high Peierls barrier for the [0001] dislocation in the screw orientation. The generalized stacking fault energy calculations by first‐principles density functional theory indicate that the [0001] dislocations prefer to glide as undissociated dislocations on the prism slip plane with the largest interplanar spacing, where the large empty channels (imaged as regular hexagons in the [0001] projection) are not cut. The values of fracture toughness K_IC are evaluated to be 2.0–2.3 MPa·m^1/2 without apparent orientation dependence by single‐cantilever bend tests of chevron‐notched single‐crystal specimens. In addition, grain boundary sliding is found to occur at a relatively low shear stress of about 0.9–3.3 GPa during compression tests on micropillar specimens unintentionally containing a grain boundary.

Pressureless sintering of Si ₃ N ₄ doped with 2 mol% Y ₂ O ₃ and 5 mol% MgO was carried out at temperatures ranging from 1700°C to 1770°C in three kinds of packing powders consisting of Si ₃ N ₄ and BN coupled with or without addition of Y ₂ O ₃ , MgO, and SiO ₂ . Effects of packing powders on weight change, densification, and grain growth during sintering and properties (thermal conductivity, fracture toughness, and bending strength) of the sintered samples were studied. It was found that adding Y ₂ O ₃ and MgO into the packing powder resulted in improvement of densification, grain growth, thermal conductivity, fracture toughness, and bending strength of the sintered samples. However, further addition of SiO ₂ into the packing powder could lead to suppressed grain growth and decrease in thermal conductivity, fracture toughness, and bending strength. The sample sintered at 1770°C in the packing powder containing Y ₂ O ₃ and MgO was fully densified, and it possessed well‐balanced properties: thermal conductivity of 76 W m ⁻¹ K ⁻¹ , fracture toughness of 7.82 MPa m ^1/2 , and bending strength of 932 MPa.

For the purpose of systematically studying the effects of rare‐earth (RE) oxide additives on the microstructure and properties of silicon nitride ceramics, all the 16 members of the RE family (RE = La, Ce, Pr, Nd, Sm, Eu, Gd, Tb, Dy, Ho, Y, Er, Tm, Yb, Lu, and Sc) were employed in parallel in the present study. An amount of 2 mol% of each RE oxide, in combination of 5 mol% of MgO, was used as additives for a high purity silicon powder in preparing Si ₃ N ₄ ceramics by the sintering of reaction‐bonded silicon nitride method. Full nitridation was achieved for all the 16 samples. A post‐sintering at 1900°C for 3 h resulted in complete densification, except for the Eu‐bearing sample whose densification was hindered by the large weight loss occurring during sintering. The effects of the 16 RE oxides on the microstructural characteristics (grain diameter, grain length, and grain aspect ratio) and the materials properties (thermal conductivity, bending strength, and fracture toughness) were examined for the sintered samples. While a clear correlation between the grain growth behavior or properties and the RE ionic size was often not found, a rough trend can occasionally be observed. It was found that some materials achieved remarkable properties. For example, the material doped with Sc ₂ O ₃ attained a high thermal conductivity of 122 W m ⁻¹ K ⁻¹ ; while fracture toughness of the material doped with Sm ₂ O ₃ was as high as 10.36 MPa m ^1/2 ; the Si ₃ N ₄ doped with Tm ₂ O ₃ showed a good balance between the thermal conductivity, fracture strength and fracture toughness (118 W m ⁻¹ K ⁻¹ , 748 MPa, and 9.23 MPa m ^1/2 , respectively).

Large silicon nitride (Si ₃ N ₄ ) grain can enhance thermal conductivity and fracture toughness in the Si ₃ N ₄ ceramics. In this study, in order to use pores as a grain growth trigger and to achieve final densification, a limited amount of pores were introduced into the green compact. The pores were introduced by using four types of pore formers with diameters of 2, 20, 30, and 100 µm, and the pore volume was in the range of 3–18 vol%. The presence of pores promoted grain growth regardless of their diameter and volume, because no collisions occurred among the grains in the pores, and larger pores formed larger elongated grains. However, their large volume and diameter inhibited densification. Based on this tradeoff, the pore formers with a diameter of 30 µm exhibited superior properties, and fracture toughness and thermal conductivity increased to 10.2 MPa·m ^1/2 and 118.2 W/(m·K), respectively, as the pore volume increased to 5 vol%. These results revealed that the pores in the compacts before sintering could improve the thermal conductivity and fracture toughness if they promoted grain growth during the sintering procedure and finally disappeared.

Artificial intelligence (AI) models such as a convolutional neural network (CNN) are powerful tools for predicting the properties of materials from their microstructural images, etc. It is, however, critically essential to understand how the AI models use images and information to predict the target properties. In this study, we tried to gain insight into the inner workings of two AI models trained to predict bending strength (BS) and thermal conductivity (TC) of silicon nitride ceramics. Focusing on the intermediate feature representation of the microstructural images in the networks, the high‐dimensional data points corresponding to sample images were mapped onto a two‐dimensional plane using t‐distributed stochastic neighbor embedding (t‐SNE). The maps demonstrated that the AI models predicted BS and TC primarily based on the porosity and grain sizes of the samples. The result indicates that t‐SNE is a useful technique for making the basis of models' predictions more understandable and well founded.