Solid-state electrolytes are widely considered the "holy grail" for safer, more energy-dense batteries, promising to replace flammable liquid electrolytes in everything from smartphones to electric vehicles. However, a major hurdle remains: achieving high ionic conductivity—the ease with which lithium ions move—while maintaining long-term stability.

As the electric vehicle (EV) market surges, the biggest anxiety for owners and manufacturers remains the battery. How long will it last? Is it safe? Accurately predicting a battery's State of Health (SOH) is notoriously difficult under real-world driving conditions. Now, a research collaboration including Jilin University and major automaker China FAW Group has developed a novel deep learning model that cracks this code, achieving prediction errors of less than 1% even in dynamic environments.

As the global demand for sustainable energy solutions intensifies, the efficiency of devices like metal-air batteries and fuel cells hinges on a critical chemical process: the oxygen reduction reaction (ORR). Historically, platinum group metals have been the gold standard for catalyzing this reaction, but their scarcity and high cost remain significant barriers to widespread commercialization.

 Researchers have developed an integrated gray wolf optimization algorithm-based hybrid estimation framework that combines sample entropy, localized voltage area, and fuzzy entropy to accurately estimate the state of health of bipolar lead-acid batteries. Partial charging profiles are utilized to extract and validate battery health feature attributes based on gray relational grades. The proposed hybrid models utilize two pairs of battery health attributes: localized voltage area paired with either fuzzy entropy or sample entropy. The average mean absolute error and average root mean squared error values are below 1.02 percent and 1.5 percent respectively for the localized voltage area and fuzzy entropy health attribute pair.

 Researchers have developed a hybrid surrogate model for iso-octanol oxidation to iso-octanal that integrates data-driven approaches with chemical equations grounded in mass transfer, heat transfer, momentum transfer, and reaction engineering. By establishing a precise mechanistic model based on an Aspen Plus generated database, the team overcame the challenge of scarce oxidation experimental data caused by long operating cycles and hydrogen safety concerns. Compared to direct process simulation and multi-objective optimization methods, surrogate models exhibit computational speeds exceeding 400 times those of traditional methods. The optimization results reveal significant reductions in both primary energy demand and greenhouse gas emissions.

Researchers have developed several data-mechanism hybrid driven methods to improve key variables prediction in process industry. Based on random forest, extreme gradient boosting, and artificial neural network, these methods were validated through benzene-toluene-xylene distillation and steam methane reforming cases. Under noise intensity of 10 to 20 percent and sample sizes of 100 to 400, the coefficient of determination improved by up to 5.2 percent for random forest, 17.7 percent for extreme gradient boosting, and 36.2 percent for artificial neural network compared to data-driven models alone.

Due to their molecular-like ability to form directional bonds and self-assemble into complex architectures, patchy particles represent a promising frontier in the design of novel functional colloids. However, developing efficient strategies for synthesizing such intricate structures remains a significant challenge. In this study, we present a new multistep approach to creating two distinct patches on silica particles using metallic layers of controlled thickness as sacrificial masks. Selective dissolution of these masks enables sequential functionalization of predefined surface areas, resulting in bi-patchy particles with two clearly differentiated functional patches, as confirmed by fluorescence microscopy.