A cheap polymer membrane usually used only as a backing layer can sort crude oil molecules once heavy hydrocarbons partly block its pores, KAIST and Georgia Tech researchers report in Nature. Modelling suggests a third less distillation energy, but the four-week trial was laboratory scale.
Beijing researchers have made an injectable gel that flows through an endoscope, sets in about 30 seconds at body temperature and sticks only on the side facing tissue. It closed stomach fistulas in rats and pigs, but has not been tested in people.
A Chinese and German team used machine learning to design ductile irons made entirely from contaminated scrap. In laboratory tests the best recipes were stronger or more ductile than earlier recycled iron and about 8 per cent cheaper to make, but they have not yet been run in a commercial foundry.
A new EWI Cologne study for TU Darmstadt and the DESERTEC Foundation models iron powder as a seasonal energy store and finds savings of up to 18 billion euros a year in a European system. The numbers are scenarios, and the technology is still at demonstration scale.
Kyoto University has sent data at 14.6 Gbit/s from a car on a real road to a receiver 140 metres away using the 300 GHz band. It is a research trial at 20 km/h with a fixed antenna, but it moves terahertz radio out of the laboratory.
Ulm-based NVision and Ulm University have shown a single designed molecule working as a light-connected spin qubit. The physics is peer reviewed; the roadmap to on-chip quantum computers is the company’s own.
KAUST researchers have made a retrofit coating that reflects desert sun and draws heat out of a live cable. Modelling suggests up to 18 per cent more daytime capacity in Riyadh, but it has not yet been tried on an operating transmission line.
A DGIST team, part-funded by a Korean magnet maker, says a new diffusion recipe delivers comparable magnet performance with 80 per cent less dysprosium. The result is laboratory scale, and large magnets are the next test.
A Fraunhofer team in Potsdam is building a robot platform that produces enzyme variants without living cells, then trains AI on the results. The bottleneck it attacks is real, but the platform is early and the spin-off does not exist yet.
An HKUST team has stretched room-temperature spin coherence in a perovskite to about five nanoseconds, roughly 300 times longer than a conventional single well. The devices built on it are proofs of concept, not products.
