Sodium-ion batteries are promising alternatives to lithium-ion batteries for large-scale energy storage, enabling lower-cost and safer energy storage systems. O3-type layered oxides are considered mainstream cathodes materials for practical sodium-ion batteries owing to their high theoretical capacity and scalable production, drawing wide attention from both academia and industry. Nevertheless, their limited capacity within 2.0–4.0 V restricts market competitiveness.

Raising the voltage causes lattice oxygen instability, irreversible phase transitions, and electrolyte decomposition, resulting in structural degradation and rapid performance fading, which blocks their commercial application.

To address these issues, a research team led by Prof. ZHANG Xian-Ming from Taiyuan University of Technology has recently proposed an integrated design concept based on solid-solution reactions and anionic redox chemistry. They successfully developed a low-cost, high-capacity, long-life, and air-stable 4.3 V-class O3-type layered oxide cathode material, NaNi_0.35Fe_0.2Mg_0.05Mn_0.3Ti_0.1O₂ (FMT), fundamentally addressing the two critical problems of irreversible P3→O1 phase transition and lattice oxygen release at high voltages.

The team’s findings were published in Science Bulletin .

Sub-headline: HIT (Shenzhen) researchers develop FedPD to enhance personalized cross-architecture collaboration

Researchers from Harbin Institute of Technology (Shenzhen) proposed FedPD, a personalized federated learning method based on partial distillation. By assessing knowledge relevance for selective transfer, FedPD enables efficient collaboration among clients with diverse model architectures while significantly improving performance on heterogeneous data.

Privacy-preserving feature selection allows identifying more important features while ensuring data privacy, thus enhancing data quality. Secure multiparty computation (MPC) is a cryptographic method that allows effective data processing without a trusted third party. However, most MPC-based feature selection schemes overlook the correlation between features and perform poorly for model training when handling datasets containing both numerical and categorical attributes.

This study rationally engineered the esterase Aes72 based on its resolved crystal structure and quantum mechanical calculations, yielding a mutant with substantially enhanced degradation activity of polyurethane. The results provide an efficient enzymatic resource and mechanistic foundation for the biological recycling of polyurethane.

The National Institutes of Health has renewed support for Artificial Intelligence for Alzheimer’s Disease, or AI4AD. The new $12.6 million award to advance the project’s next phase, AI4AD2, brings its total investment in AI4AD to $30.7 million. Led by Paul M. Thompson, PhD, associate director of the USC Mark and Mary Stevens Neuroimaging and Informatics Institute (Stevens INI) at the Keck School of Medicine of USC, the multi-institutional initiative will develop artificial intelligence (AI) tools to uncover the biological causes of Alzheimer’s and related dementias, improve predictions of disease progression, and help develop more precise treatment options. AI4AD2 unites 10 investigators and 23 co-investigators from 10 institutions in pursuit of four interconnected research goals. The consortium will analyze large-scale datasets, including whole-genome sequencing, brain imaging, cognitive testing, and other biological data, to advance the diagnosis and treatment of dementia. This work builds on the original AI4AD initiative launched in 2020, which developed AI tools to detect Alzheimer’s-related patterns in brain scans and showed how machine learning can link imaging findings to underlying genetic risk. AI4AD2 will also develop new “genomic language models,” a type of AI inspired by the same broad family of technology used in language-based artificial intelligence systems. Instead of analyzing words, these models will analyze genomic sequences to identify combinations of DNA changes associated with Alzheimer’s disease, disease progression, and key biomarkers. The project will train and evaluate these methods using data from over 58,000 participants across 57 cohorts. In practical terms, that involves teaching AI to search vast genetic datasets for patterns that traditional methods could not identify.