Cavendish’s apparatus turns toward millicharged particles while memristor noise and charge-density waves expose new clues; closed instruments make the limits of reproducibility visible too.
Science··Evening
Cavendish apparatus turns toward a tiny charge
A team from Fermilab, Stanford and the University of Delaware has calculated that hypothetical particles carrying a very small electric charge could be hunted with an updated form of the classic Cavendish experiment. The work published in Physical Review Letters targets the mass range between MeV and GeV. In the proposed setup, ordinary charged particles stay outside a Faraday cage while millicharged ones can pass through and leave a measurable electric field inside; that departure from Coulomb’s and Gauss’s laws is the signature being sought. The calculation says such particles could be caught even if they make up less than one part in a trillion of dark matter. Harikrishnan Ramani is the lead author, with Asher Berlin, Zachary Bogorad and Peter W. Graham, and Kent Irwin’s group at Stanford taking part. No measurement exists yet: the team is building a prototype and expects results within a few years. A two-century-old laboratory geometry is thus being turned into a search window for candidates that carry only a faint electric charge.[1]
Noise and pairing become measurement clues
A KAIST group led by Kyung Min Kim, with Do Hoon Kim as first author, reports in Advanced Materials that a memristor’s current noise changes in magnitude and behaviour as its resistance state changes, and builds a programmable probabilistic neuron on that. Presetting the resistance lets one circuit encode hertz-range human activity signals and kilohertz-range speech, at reported accuracies of 94.8 per cent and 95.0 per cent. The group treats the noise as a tunable information-processing resource rather than an error to be suppressed. The announcement gives the two recognition accuracies without a comparison device or an energy figure. Separately, Sheng-Chih Lin and Alfred Zong, working at UC Berkeley and Stanford, tracked layered 1T-TiSe2 with ultrafast extreme-ultraviolet absorption spectroscopy and found excitonic correlations surviving above the roughly 200 kelvin transition, with susceptibility climbing as the material approaches it. The data give a dynamical measure of the instability rather than proof a condensate formed; the work appears in Nature Physics.[2], [3]
Closed tools also limit checking
An article in BioScience argues that ecology and environmental science increasingly rest on proprietary artificial intelligence, satellite imagery, online platforms and sensors whose inner workings are closed. Researchers cannot reach the training data, the algorithms or the testing, which the authors say puts independent checking of findings at risk. Ivan Jarić of the University of Paris-Saclay leads the article with Karen Anderson of the University of Exeter and Michael Bertram of the Swedish University of Agricultural Sciences and Stockholm University. They link the dependence to publication pressure and the urgency of environmental crises, and propose open-source alternatives, benchmarking of proprietary tools against known cases, fuller documentation, sustained human oversight and regulatory routes to platform data. The piece is a review rather than new measurement, and the authors concede that some closed systems will resist every one of these fixes. The Cavendish search, memristor noise work and charge-density-wave spectroscopy show how a faint signal becomes available only through a suitable apparatus, while the closed-instrument review recalls that inaccessible training data, algorithms and testing steps can block the same confidence. Hidden signatures, tunable noise and pre-transition pairing depend on what measurement can open; a trade-secret tool makes checking harder.[4], [1], [2], [3]
