A team led by Guoyin Yin at Wuhan University and the Shanghai Artificial Intelligence Laboratory recently proposed a modular machine learning framework using LoRA fine-tuning. This framework can not only accurately predict single organic reactions but also achieve the prediction of "one model handling multiple types of reactions." Even using only natural language to describe chemical reactions, the prediction accuracy is comparable to machine learning models based on expert experience. The article was published as an open access Research Article in CCS Chemistry, the flagship journal of the Chinese Chemical Society.

The research team has successfully achieved the following breakthroughs in the well-known anthracene [4+4] photocycloaddition reaction (*1):

１．Dual physical control using “light” and “heat”

　The team established a two-parameter control system in which both the reaction rate and the initiation/termination of the reaction can be freely modulated by optical intensity and temperature.

２．World-first visualization of intermediate molecular structures

　By drastically slowing down the ultrafast reaction—normally completed within 10⁻⁸–10⁻⁶ s—the team succeeded in directly visualizing intermediate states using single-crystal X-ray diffraction (SCXRD) (*2).

This achievement represents the first direct structural observation of the anthracene [4+4] reaction pathway.

３．Directional dependence between incident light and molecular orientation

　The researchers proved that the reaction efficiency depends on the angle between the incident light and the molecular transition dipole moment μ (*6).

　By changing light direction, the reaction could be switched “on” or “off,” revealing orientation as the third regulatory factor.

Researchers at the Department of Energy’s Oak Ridge National Laboratory have developed an innovative energy storage system design that introduces a safer, more efficient method for electrical charge transfer. This study advances fundamental understanding of how functional groups impact the mechanical and electrochemical performance of highly charged polyelectrolyte membranes — thin, charged polymer sheets that help control the movement of ions in energy devices.

Cell division is an essential process for all life on earth, yet the exact mechanisms by which cells divide during early embryonic development have remained elusive – particularly for egg-laying species. Scientists from the Brugués group at the Cluster of Excellence Physics of Life (PoL) at TUD Dresden University of Technology have revealed a novel mechanism that explains how early embryonic cells may divide without forming a complete contractile ring, traditionally seen as essential for this process. The findings, published in Nature, challenge the long-standing textbook view of cell division, revealing how parts of the cytoskeleton, and material properties of the cell interior (or cytoplasm) cooperate to drive division through a ‘ratchet’ mechanism.

Terahertz (THz) radiation underpins many next-generation technologies and advances in materials science, but current THz spectroscopy methods cannot deliver both high spectral and spatial resolution simultaneously. Now, researchers have addressed this challenge by developing a novel methodology called spatially resolved asynchronous-sampling THz spectroscopy (SPRATS). Their system combines two existing THz measurement techniques to achieve micrometer-level near-field imaging and ultra-high spectral resolution, enabling the characterization of advanced THz resonant structures. SPRATS represents a transformative advancement in terahertz spectroscopy, bridging critical gaps between far-field and near-field methodologies while establishing new benchmarks for resolution and accuracy. This technology opens possibilities across multiple disciplines, including materials science, biomedical sensing, integrated photonics, and fundamental physics research.

A new holographic computation method could significantly advance augmented‑reality head‑up displays (AR‑HUDs) for vehicles. Current HUDs are limited to flat, fixed‑distance projections and rely on computationally heavy algorithms that struggle to generate high‑quality 3D holograms in real time. Research published in Advanced Photonics Nexus demonstrates a matrix‑multiplication‑based diffraction approach that reduces computation time by about 58 percent and sharply cutts memory usage. A prototype projected multiple holographic images at different depths—from 0.1 to 1.5 meters—aligning seamlessly with real‑world objects. The technique supports extreme size differences between input and output images and unifies near‑ and far‑field holography within a single framework. This advance paves the way for compact, wide‑field AR‑HUDs that can overlay critical driving information directly onto the environment, bringing next‑generation smart windshields closer to reality.