A new hydrogel electrolyte controls water reactivity and stabilizes both electrodes in aqueous zinc-iodine batteries. Laboratory cells operated from minus 40 to 90 °C Celsius, while a 120 mAh pouch cell retained about 83% capacity after 2,300 cycles.

How deep is the deepest ocean—and how precisely can we actually measure it? A Japanese research team has mapped the Challenger Deep in the Mariana Trench using the EM124 multibeam echosounder aboard the research vessel Hakuho-maru.

Osaka Metropolitan University researchers developed an AI-driven transfer-learning method that refines volume size factor predictions from first-principles calculations, enabling more reliable prediction of properties in previously unexplored alloy systems.

Researchers at the Nano Life Science Institute (WPI-NanoLSI) at Kanazawa University, Kyoto University, SOKENDAI, and the National Institute for Physiological Sciences have revealed how CaMKIIα—a key brain protein involved in learning and memory—organizes itself into chain-like structures. Their study, published in Science Advances, shows how activation causes these protein structures to grow and how a mutation associated with neurodevelopmental disorders changes their organization, findings that could deepen our understanding of memory formation and neurological disease.　

 As new memories form the connections between brain cells can become stronger. A protein called CaMKIIα plays an important part in this process and is found in especially large amounts at the points where brain cells communicate. Scientists have long known that more of this protein gathers at active connections, but they have not been able to see clearly how the individual protein molecules come together.

A research team from NVIDIA and Thinking Machines Lab has explored whether the internal responses of artificial-intelligence models can serve as order parameters for physical systems. The study uses a three-dimensional convolutional neural network to analyze crystalline structures under controlled vibrational disorder. By treating the model output as a continuous measure of structural order, the work shows that body-centered cubic crystals lose their specific lattice identity before they lose generic solid-like order. This result suggests a new way to extend classical order-parameter ideas using machine-learned structural signals.

Materials phases are usually described by order parameters. For example, density is used to distinguish liquid and gas phases, while magnetization is used to distinguish ferromagnetic and paramagnetic phases. These order parameters are experimentally meaningful variables: they summarize collective order in a material rather than the behavior of a single atom or molecule.