SCIENCE CHINA Materials•2026•DOI: 10.1007/s40843-025-3686-6
Fused silica (SiO2) exhibits exceptional thermal stability and dielectric properties, making it an attractive material for aerospace and military applications. However, its relatively poor mechanical performance has limited its widespread practical utilization. This study proposed an innovative approach to fabricate SiO2-hexagonal boron nitride (hBN) composite ceramics via spark plasma sintering (SPS), leveraging the high-temperature phase transformation of cubic boron nitride (cBN) to introduce randomly oriented hBN as a reinforcing phase within the SiO2 matrix. The randomly oriented hBN nanoplates allow cracks to propagate along stronger grain boundaries, rather than along weaker interlayers of hBN, significantly improving the overall strength and fracture toughness of the composite. The maximum flexural strength and fracture toughness achieved are 183.4 MPa and 2.06 MPa m1/2 respectively, which are 3.6 times and 4 times that of fused SiO2. Concurrently, the composites exhibit low dielectric constants (ε = 3.58–3.69) and dielectric losses (tan δ < 0.0087) at 1 MHz. This work successfully enhanced the mechanical performance of fused SiO2 while preserving its excellent dielectric characteristics, opening new possibilities for its potential applications in advanced structural and functional fields.
SCIENCE CHINA Materials•2026•DOI: 10.1007/s40843-026-4186-7
The development of ultrahigh-temperature technologies, such as nuclear reactors, rocket nozzles, scramjet propulsion systems, and hypersonic vehicles, demands materials with melting points (Tm) exceeding 3273 K. However, the highest reported Tm among non-radioactive oxides is 3125 K (MgO), limiting progress. Existing predictive models for oxide Tm suffer from a trade-off between physical insight and practical utility: thermodynamic approaches require complex calculations, Lindemann's criterion relies on elusive parameters like Debye temperature, and semi-empirical correlations lack transferability. Machine learning models offer predictive power but lack interpretability and reliable extrapolation. Here, we simplify a previously derived semi-empirical model based on bond-breaking probability, which links Tm to effective potential barrier Ueff, proportional to bond length (d), bond density (Nb), and bond ionicity (fi). By analyzing a dataset of 68 oxides, we establish a simplified linear relationship between Tm and Ueff, expressed as Tm = 0.052Ueff + 122.6 K, with Ueff in kJ/mol. This model achieves a mean absolute error of 76 K and a coefficient of determination (R²) of 0.97, outperforming existing empirical models. The model's physical transparency and simplicity enable rapid screening of novel oxides, guiding the design of materials with Tm exceeding 3273 K. Our findings provide a practical tool for accelerating the discovery of ultrahigh-temperature oxides, addressing a critical bottleneck in next-generation thermal protection systems.