Science

Five scorched, ultra-short-period rocky worlds are the test cases: models show most end up as bare rock, with global melt fractions capping out at 20% or less. One gas, S2, spikes when the surface is solid rather than molten, giving JWST a possible tracer to tell which interior state a planet is really in before its atmosphere burns away.

— via arXiv — Astro-ph, Mariana Sastre, Tim Lichtenberg, Lisa Dang, Anjali Piette, Haiyang S. Wang, Mercedes L'opez-Morales, Thomas Wilson, Charles-'Edouard Boukar'e, Mahesh Herath, Nicolas Cowan, Md Abdullah Al Zaman, Madyson G. Barber, Casey Brinkman-Traverse, Nicholas Connors, Ian Crossfield, Lina D'Aoust, Oliver Herbort, Leoni Janssen, Mathilde Kervazo, Owen Lammert, Yamila Miguel, Raymond Pierrehumbert, Allona Vazan, Joost P. Wardenier, Sebastian Zieba

Cellphone data from 11 U.S. metro areas shows traveling farther exposes people to more politically mixed places, but visiting more different kinds of spots doesn't cut segregation and can deepen it. Community spots and social businesses skew left-leaning visitors regardless of who lives nearby.

— via arXiv — Physics, Marco Tonin, Michele Tizzoni, Bruno Lepri, Esteban Moro

In roughly 36% of cases where solar flares seem to trigger sudden magnetic-field shifts, the "change" may just be flare heating messing with how light escapes the sun's atmosphere. New inversions and 3D simulations show cold, ejected material and shifting line-formation depths can mimic real magnetic evolution, meaning current methods can't always tell fact from artifact.

— via arXiv — Astro-ph, Dimitri Arramy, Jaime de la Cruz Rodr'iguez, Adur Pastor Yabar, Jorrit Leenaarts

A new likelihood method skips binning time data altogether, building probabilities event by event to pin down the g factor in perturbed angular distribution experiments. Tested against simulated data, it beats traditional binned analysis on low-statistics datasets, though results often come out multimodal rather than the usual tidy Gaussian curve.

— via arXiv — Physics, Franziskus von Spee

Zap a lab-made ice of ethanolamine, water and methanol with electrons and it cooks up formaldehyde chains, formic acid and hints of choline-like molecules. The result: icy grains in interstellar clouds may quietly build precursors to nutrients, though the choline signal is too muddled to confirm.

— via arXiv — Astro-ph, Tara L. Stoib, Alejandro Guerrero-Caicedo, Duncan V. Mifsud, P'eter Herczku, Gerg\H{o} Lakatos, Zuzana Ka\v{n}uchov'a, Zolt'an Juh'asz, B'ela Sulik, Nigel J. Mason, Sergio Ioppolo, Paola Caselli, Barbara Michela Giuliano, Karl E. Duderstadt, Felipe Fantuzzi, Heidy M. Quiti'an-Lara

This is a list of arXiv site tools and toggles, not a news story. No factual claims, findings, or events are present to summarize into a deck.

— via arXiv — Physics, Johann Ostmeyer

An 8-billion-parameter LLM debating in groups can lose to a wrong answer even when 75% of agents start out correct, and math predicting when consensus flips got it wrong every time. The real culprit: how a claim is worded biases each model's verdict before it reads a single message.

— via arXiv — Physics, Makoto Fukushima

An AI trained to spot galaxy features found 710,000 star-forming clumps hiding in 210,000 nearby galaxies, structures once thought exclusive to the distant early universe. Their sizes and inward drift suggest these clumps migrate toward galactic centers, building the bulges at their cores.

— via arXiv — Astro-ph, J"urgen J. Popp, Hugh Dickinson, Stephen Serjeant, Lucy F. Fortson, Vihang Mehta

Pretraining a neural PDE surrogate on one airfoil family can be worth 3.25 times more target data than training from scratch, but that edge shrinks or flips as the target dataset grows to 5,000 samples. The catch: when the new task adds transition modeling instead of matching the old physics, the payoff shows up on a different, less favorable schedule entirely.

— via arXiv — Physics, Pochinapeddi Sai Bhargav, Nithin Somasekharan, Rohit Sunil Kanchi, Sicheng He, Shaowu Pan

Nanoscale versions of the "hat" tile, the 2023 shape that finally solved math's Einstein problem, twist laser light into pinwheel diffraction patterns no ordinary quasicrystal produces. The effect flips when the structure is mirrored, tying the light's handedness directly to the pattern's odd, non-repeating symmetry.

— via ScienceDaily

A quench in the q/2-body SYK model still thermalizes at a Planckian rate, but the rate shrinks as Γ ~ T/q, not grows as conjectured. The catch: the causal Green's function turns thermal instantly, while off-diagonal time correlations stay stubbornly non-thermal, dragging down the overall rate.

— via arXiv — Physics, Alexander Osterkorn, Jan C. Louw

A distant quasar's black hole is stirring hot gas some 300,000 light-years beyond its home galaxy, with turbulence energy roughly 100 times higher than earlier estimates suggest. X-ray satellite XRISM caught the signal, energy on par with billions of supernovae combined.

— via ScienceDaily

A new two-scale simulation method reveals that contact-angle hysteresis on rough surfaces can swing by more than three times depending on which direction the contact line points, spiking sharply near diagonals. The finding helps explain why droplets on textured surfaces can settle into square-like shapes with sides running along those diagonal directions.

— via arXiv — Physics, Zijie Lin, William M Feldman, Braxton Osting

Give a diffusion model enough time and it still won't show you a protein unfolding, at least not with standard sequential sampling. A new method called METALICA swaps parallelism across particles for parallelism across diffusion levels, and at a budget where the old approach finds nothing, it uncovers a second free energy minimum.

— via arXiv — Physics, Alireza Omidi, Jiajun He, J"org Gsponer, Saifuddin Syed

A quantum dot's lopsided electron and hole positions, not some exotic material, is what unlocks a forbidden photon-pair emission in a nanoparticle-mirror cavity. The math shows a working operating principle; whether anyone can actually build and test it is still unresolved.

— via arXiv — Physics, Karol Kawa

A physics model of bendy, self-actuating bodies explains why eels anchored in flowing water suddenly flop from standing upright to lying flat. The theory treats posture as a cost-benefit tradeoff, and it predicts scaling laws for how much of the body stays exposed to the current.

— via arXiv — Physics, Adam Pearl, Ludwig A. Hoffmann, L. Mahadevan

A position-dependent-mass oscillator makes the standard entropy math too hard to solve directly, so the calculation gets punted to an asymptotic limit that behaves like an ordinary harmonic oscillator. Momentum-space entropies, though, come out exact, no approximation needed, and both versions converge to Shannon entropy as expected.

— via arXiv — Physics, Dorcas A. Addo, Daniel Sabi-Takou, Eugene Adjei, Lat'evi M. Lawson

No shortcut survives at scale: for large disordered systems, low-order approximations lose all predictive grip on the final outcome unless a specific response-to-volume threshold is met. Simulations of a 3D random-field Ising model up to size 192 confirm the effect, while a related game shows outcomes can hinge on the order events unfold.

— via arXiv — Physics, Bo Li, Chaoqian Wang

Austrian physicists are wiring an electron microscope directly into a quantum computer built from trapped ions, letting stray electrons entangle with it instead of vanishing as noise. The goal: sharper images of fragile samples like proteins using far fewer electrons, with a working prototype now under construction at TU Wien.

— via ScienceDaily

A new memory chip writes data using less than 0.1 picojoules per bit, built on a standard 300-mm wafer line rather than exotic nanoscale materials. The trick: swapping magnetic spin for orbital angular momentum in a thicker, manufacturable metal layer, hitting 182% tunnel magnetoresistance and nanosecond switching.

— via arXiv — Physics, Dinggui Zeng, Yang Gao, Jinyu Duan, Lei Zhao, Yuhao An, Xing He, Jintao Ke, Yonglong Ga, Shasha Wang, Zhenghui Ji, Muyuan Chen, Hengan Zhou, Xuejie Xie, Enlong Liu, Junlu Gong, Qijun Guo, Yihui Sun, Zejie Zheng, Weiming He, Xiaolei Yang, Fantao Meng, Yaohua Wang, Hongxin Yang, Delin Zhang, Yong Jiang, Wanjun Jiang, Shikun He