Physics News — September 2026
2026年9月の最新研究・発見（全53件・更新終了） 最終更新 / Last updated: — 査読論文・一次ソース付き / Peer-reviewed, with primary sources —
📰 2026年9月 のニュース / September 2026 （全53件・更新終了）
2026年9月（September 2026）に発表・注目された基礎物理学の最新ニュースと研究解説。一次ソース（DOI・arXiv・機関発表）付きで月内の項目を掲載しています。暗黒物質・宇宙論・素粒子・超伝導・量子情報など分野横断で掲載しています。諸事情により、毎月の更新はこの月をもちまして終了いたします。Recent physics news and research explanations from September 2026, with primary sources — dark matter, cosmology, particle physics, superconductivity, quantum information and more. For various reasons, monthly updates end with this September 2026 edition.
⏱️ / 「世界で最も正確な時計」が誕生——シンガポール国立大学（NUS）量子技術センター（CQT）のMurray Barrettらが、電場で捕らえた1個のルテチウム176イオン（¹⁷⁶Lu⁺）の848 nm時計遷移を使う光時計で、周波数を小数点以下19桁まで評価し、不確かさ1×10⁻¹⁹（2台の系統不確かさは1.2×10⁻¹⁹と1.3×10⁻¹⁹で、2025年にNISTのアルミニウムイオン時計が出した記録5.5×10⁻¹⁹のおよそ4分の1）を達成。2台の時計を約200時間の相関分光で直接比べ、5.7×10⁻¹⁹の精度で一致を確かめた——これまでで最も精密な時計比較。ルテチウムの時計遷移は温度や磁場の変化をほとんど感じないのが強みで、2030年以降に見込まれる「秒」の再定義の有力候補に。次は可搬型への小型化をめざす（K. J. Arnold・M. Lee・M. D. Barrettら、CQT／NUS、Nature・9月23日掲載）
Optical atomic clocks, which count the oscillations of laser light locked to an electronic transition in an atom or ion, now outperform the caesium clocks that have defined the SI second since 1967 by more than two orders of magnitude, and the international metrology community is weighing data from them towards a redefinition of the second expected in or after 2030. The record holders of recent years have been clocks based on ytterbium, strontium and aluminium; in 2025 NIST reported an aluminium-ion clock with a systematic uncertainty of 5.5×10−19. At the Centre for Quantum Technologies (CQT) of the National University of Singapore, Murray Barrett’s group has spent more than a decade on an element that, to its knowledge, no other group uses for timekeeping: lutetium.
In a paper published in Nature on 23 September 2026, Kyle Arnold, Michael Lee and colleagues report two clocks, each built around a single trapped 176Lu+ ion whose clock transition is probed by a laser at 848 nm. Lutetium’s advantage lies in the atom itself: its clock transition is barely shifted by changes in temperature — that is, by the thermal radiation of the surroundings — or by stray magnetic fields, two effects that limit other optical clocks, and the group invented a scheme it calls hyperfine averaging to define the transition. According to CQT, the clock’s frequency has been evaluated to the 19th digit with an uncertainty of 1×10−19; the paper gives total systematic uncertainties of 1.2×10−19 and 1.3×10−19 for the two clocks, roughly a quarter of the 2025 aluminium record. An uncertainty of 1×10−19 corresponds to drifting by about one second in some 300 billion years. Barrett says the clock would stay accurate whether it were run in Death Valley or on the Antarctic plateau.
To back the claim, the team compared the two clocks directly with correlation spectroscopy over 200 hours of measurement: both ions are put into superpositions and left to evolve freely, and even after the probe laser has completely dephased their oscillations remain correlated, so the phase difference they accumulate reveals any frequency difference. The clocks agreed within 5.7×10−19, the most precise clock comparison reported so far. The authors count only 12 accuracy evaluations of optical clocks in the past nine years, just five of them backed by high-accuracy comparisons between clocks of the same species — two by the CQT group — whereas the international roadmap calls for comparisons between institutes. At this level that is hard: a 10−19 clock feels the gravitational slowing of time over millimetres of height. The comparison could resolve a 5-mm height difference between the two clocks on one table, so the team measured the ions’ relative height to better than a millimetre, and differences in gravity between distant sites are not yet known well enough for comparisons at this level. The next step is to miniaturize the laboratory clock into a transportable system.
🧵 / 引き離すと「弦」が切れて新しい粒子対が生まれる——デューク大学量子センターのChristopher Monroeらが、レーザーで相互作用を制御した13個の捕獲イオンの鎖に、弦の切断（ストリング・ブレイキング）を表す簡略化した格子ゲージ理論の模型を書き込み、非平衡状態から時間発展させて、粒子・反粒子対に相当する実効的な電荷が現れる様子を観測。クォークを引き離すと間の「弦」にエネルギーがたまり、ついには切れて新しいクォーク・反クォーク対ができるという量子色力学（QCD）の閉じ込めに対応する現象で、捕獲イオン型の量子シミュレーターによる初期の実証例。この規模ではまだ古典計算で検証でき、結果は一致した。Googleの超伝導回路、QuEraの中性原子に並ぶ3つ目の方式での実現で、将来は初期宇宙の物質生成の研究にもつながると期待される（A. De・Z. Davoudi・A. V. Gorshkov・C. Monroeら、デューク大学／メリーランド大学ほか、Nature Physics・9月23日掲載）
Quarks are never seen in isolation. In quantum chromodynamics (QCD) the gluon field between a quark and an antiquark forms a narrow flux tube, or string, whose energy grows with the distance between them. Pull them far enough apart and it becomes energetically cheaper for the string to break by creating a new quark–antiquark pair from the vacuum, so that an attempt to separate two particles ends with two (or more) bound hadrons. This string breaking, a direct consequence of confinement, happens in high-energy collisions such as those at the Large Hadron Collider and would have mattered in the early Universe, but following it in real time is extremely hard for classical computers because the quantum state quickly becomes highly entangled. Encoding lattice gauge theories in programmable quantum simulators offers a way in.
A team led by the Duke Quantum Center, with researchers from the University of Maryland, the University of Oxford, Caltech, Cornell University and KU Leuven, encoded a simplified lattice-gauge-theory model of string breaking in a chain of 13 trapped atomic ions, using precisely controlled laser beams to set how the ions interact and thereby the energy stored in the ‘string’. The researchers prepared the chain in an out-of-equilibrium state and followed its evolution, detecting the appearance of effective charges — the analogues of newly created particle–antiparticle pairs — and reconstructing the dynamics of the breaking string. At 13 ions the problem can still be solved classically, which made it possible to check the simulator: the classical calculations agreed with the experimental results.
First author Arinjoy De, a former PhD student in Christopher Monroe’s group who now works at QuEra Computing, says the work opens experimental routes to studying confinement in a controlled laboratory setting. Teams led by Google and by QuEra have recently reproduced related string-breaking models with superconducting circuits and with neutral atoms, respectively; with trapped ions added, Monroe notes, the three leading quantum-computing platforms now provide a benchmark for one another, each with its own strengths and limitations. If such simulations can be scaled beyond the reach of classical supercomputers, they could also offer, in the words of the University of Maryland’s Zohreh Davoudi, insights into the out-of-equilibrium physics of the early Universe. The work was published in Nature Physics on 23 September 2026.
🔊 / 「音の粒」が消える瞬間をリアルタイムで捉えた——スタンフォード大学のAmir Safavi-Naeiniらが、チップ製造技術でつくった微小な機械共振器（約2ミリ秒鳴り続ける“極小の音叉”）を超伝導量子ビットと組み合わせ、共振器にある1個の音の量子（フォノン）を壊さずに何百回も読み出して、エネルギーが1から0へ突然跳ぶ「量子跳躍」を初めてリアルタイムで直接観測。量子跳躍の直接観測は捕獲イオン（1986年）、光子（2007年）に続くもので、膨大な数の原子の集団運動である機械振動にまで広がった。量子計算で誤りとなる励起の消失を起きた瞬間に検出する技術や、細胞内のタンパク質を見分ける超高感度センサーへの応用が期待される（T. Makihara・E. Szakiel・A. H. Safavi-Naeiniら、スタンフォード大学、Science、9月17日発表）
Since Niels Bohr’s model of the atom, physicists have described quantum systems as changing energy in discrete jumps, but watching an individual jump took until 1986, when three groups saw single trapped ions switch abruptly between bright and dark states. In 2007 Serge Haroche’s group extended the feat to light, recording the birth and death of single microwave photons in a superconducting cavity. Mechanical vibration is a harder case: a quantum of sound, or phonon, is not a property of one particle but the coordinated motion of an enormous number of atoms, and earlier experiments had found only indirect evidence of phonon jumps.
A team led by Amir Safavi-Naeini at Stanford University, with Takuma Makihara and Erik Szakiel as co-first authors, now reports in Science the first direct, real-time observation of quantum jumps of sound. Their mechanical resonator, made with chip-fabrication techniques, works like a microscopic tuning fork that keeps ringing for about two milliseconds — the equivalent of a full-size tuning fork ringing for several hours. Paired with a superconducting qubit that serves as the detector, the resonator can be checked again and again without destroying its fragile quantum state: during the two-millisecond ringdown the qubit takes hundreds of readings of whether the resonator holds one phonon or none. In individual runs the team can therefore pinpoint the moment at which the single phonon disappears and the energy jumps from 1 to 0, rather than only seeing an average decay.
Beyond completing a story that began a century ago, the result matters for technology. In many quantum-computing architectures the loss of a single excitation is precisely an error, so catching such jumps as they happen is a step towards error correction in mechanical and hybrid quantum processors. The compact resonator–qubit combination could also act as an extremely sensitive sensor: Safavi-Naeini’s team is working with Michael Roukes’ group at Caltech to use the platform to detect and identify proteins inside cells, and the researchers see longer-term benefits for everyday acoustic devices such as smartphone components. Stanford announced the work on 17 September 2026 (DOI: 10.1126/science.aeh7535).
🌑 / 暗黒物質は「暗黒の原子核」かもしれない——イタリアのグランサッソ国立研究所で2013〜2019年に運転された液体アルゴン検出器DarkSide-50の国際共同実験が、多数の「暗黒核子」が結合した超重の複合暗黒物質（核状超重暗黒物質、UHDM）を、2相式液体アルゴン時間射影検出器（TPC）で初めて探索。こうした粒子は約3.3マイクロ秒で検出器を通り抜ける間に原子核と何度も散乱するため、多重散乱の信号に最適化した解析を、地球の岩盤によるエネルギー損失も考慮して532日分の地下アルゴンのデータに適用した。信号は見つからず、暗黒核子の質量10・50・100・500 GeV/c²のそれぞれについて、暗黒物質の質量と散乱断面積の広い範囲を90％信頼度で除外。構成粒子の質量で感度が変わることを示し、暗黒物質の「内部構造」を調べる道を開いた（DarkSide Collaboration、Phys. Rev. D 114, 052001、arXiv:2602.09421、Phys.org 9月28日紹介）
Most direct searches for dark matter look for a single, faint recoil of a nucleus struck by a weakly interacting massive particle (WIMP). But if the dark sector has no long-range force analogous to electromagnetism to keep ‘dark nucleons’ apart, big-bang dark nucleosynthesis could have assembled enormous composite dark nuclei containing up to about 1020 constituents. Such nuclear ultraheavy dark matter (UHDM) would be rare, because each particle is so massive, but it could interact strongly enough to scatter many times on a single pass through a detector, leaving a track of energy deposits rather than one isolated recoil. Its finite size matters too: like an ordinary nucleus, a large dark nucleus has a form factor that suppresses hard collisions and favours many soft ones.
The DarkSide Collaboration has now carried out the first search for nuclear UHDM in a dual-phase liquid-argon time-projection chamber (TPC), using DarkSide-50, which operated at Italy’s Gran Sasso National Laboratory (LNGS) from 2013 to 2019. Its active volume, a cylinder 36 cm across and 36 cm tall viewed by 38 photomultiplier tubes, sat inside a liquid-scintillator neutron veto and a water-Cherenkov muon veto. Because a UHDM particle would cross the chamber in about 3.3 μs, faster than ionization electrons drift to the top, the analysis used only the prompt scintillation signal (S1), summing the light of all deposits in a window of 100–8000 photoelectrons, roughly tens of keV to about 1 MeV. The team modelled the energy that candidates lose in the rock above the laboratory with the Verne code and applied selections optimized for multiple-scatter topologies to the 532-day campaign with low-radioactivity underground argon. The data are described by known radiogenic backgrounds alone.
The result is a set of 90%-confidence exclusion regions in the plane of UHDM mass and per-nucleon scattering cross section, computed separately for dark-nucleon masses of 10, 50, 100 and 500 GeV/c2. The regions are bounded from above by the geometric cross section of the dark nucleus, beyond which contact interactions are unphysical, from below by the cross section at which a particle is expected to scatter on every pass, and on the low-mass side by the overburden. DEAP-3600 and LZ have previously used multiple-scatter signatures to search for ultraheavy dark matter; by modelling UHDM explicitly as a composite particle and systematically varying the mass of its constituents — an approach not previously explored in liquid-argon TPCs — the DarkSide-50 analysis shows that experimental sensitivity depends on the dark matter’s internal composition, so future detectors may probe not only whether such particles exist but what they are made of. The paper appears as Phys. Rev. D 114, 052001 and was featured by Phys.org on 28 September 2026; the collaboration is now building the much larger DarkSide-20k detector at Gran Sasso.
Keywords: dark matter, 暗黒物質, ultraheavy dark matter, 超重暗黒物質, UHDM, nuclear dark matter, 核状暗黒物質, composite dark matter, 複合暗黒物質, dark nucleosynthesis
🌍 / 1日の長さを数十年かけて変えていたのは「内核の重力」だった——アルバータ大学のHuifeng ZhangとMathieu Dumberryが、1日の長さ（LOD）が数十年規模で数ミリ秒変動する原因を検証。核とマントルが角運動量をやりとりする3つの仕組み——内核とマントルの重力的結合、核–マントル境界での電磁的結合と地形的結合——を約60年分の観測と突き合わせた結果、1970年代初めから2021年までの変動の大部分は内核がマントルを重力で引くトルクで説明でき、電磁的・地形的な結合は逆向きに働いて変動を抑えていることがわかった。内核が約10年の時間スケールで粘性的に変形していることも示唆され、1日の長さの記録が地球深部を探る手段になる。一方、約6年周期の変動は未解明のまま（H. Zhang & M. Dumberry、アルバータ大学、Nature・9月23日掲載）
A day is not exactly 24 hours, and its length (LOD) is not constant. Winds and ocean currents change it seasonally by fractions of a millisecond, tidal friction from the Moon slowly lengthens it over millions of years, and on timescales of decades it wanders by a few milliseconds for reasons that have long been debated. Because Earth’s total angular momentum must be conserved, those decadal changes are thought to come from exchanges between the mantle and the core: when the core spins up, the mantle slows down, and vice versa. Three coupling mechanisms have been proposed: electromagnetic coupling across the core–mantle boundary, where the electrically conducting outer core meets the mantle; topographic coupling, in which core flow pushes against bumps on that boundary; and a gravitational torque between the slightly aspherical inner core and density variations in the mantle.
Huifeng Zhang and Mathieu Dumberry of the University of Alberta tested the three mechanisms against roughly six decades of observations. They combined a previously reconstructed history of the inner core’s rotation relative to the mantle, inferred from seismic waves, with models of flow in the fluid core inferred from changes in the geomagnetic field, and fitted statistical models with different combinations of torques to the measured LOD record. The fit is best when the gravitational torque from the inner core dominates. Between the early 1970s and 2021, core processes changed the length of the day by a few milliseconds; the inner core’s gravitational pull on the mantle accounts for most of that change, while the torques at the core–mantle boundary — friction-like electromagnetic and topographic drag — act in the opposite direction and limit how much the day can lengthen or shorten.
The analysis also implies that the solid inner core deforms viscously on a timescale of about ten years, a surprisingly dynamic picture of Earth’s deepest layer, and it turns the length of the day into a probe of the shape and softness of the inner core and of dense structures at the base of the mantle — regions no one can visit. Not everything is explained: faster variations on 10–30-year timescales probably involve additional processes, and a separate oscillation of about six years in the LOD record remains an open problem that Zhang is now investigating. The paper, ‘Gravitational torque drives multidecadal variations in length of day’, was published in Nature on 23 September 2026.
解析からは、固体の内核が約10年の時間スケールで粘性的に変形していることも示唆された。地球の最深部は意外なほど動的だということになる。さらにこの成果によって、1日の長さの記録は、内核の形や柔らかさ、マントル最下部の高密度な構造といった、誰も行けない領域を探る手段になる。すべてが解決したわけではない。10〜30年程度の比較的速い変動にはほかの過程も関わっているとみられ、1日の長さに見られる約6年周期の振動は未解明のままで、Zhangが現在取り組んでいる。論文「Gravitational torque drives multidecadal variations in length of day」は2026年9月23日のNatureに掲載された。
Many processes in quantum matter seem to obey a speed limit set only by fundamental constants and temperature. In the strange-metal phase of cuprate superconductors and other materials, the electrical resistivity grows linearly with temperature, and the scattering rate inferred from it sits close to kBT/ħ across materials that otherwise have little in common. Jan Zaanen called this Planckian dissipation, and the associated timescale τPl = ħ/(kBT) — about 25 femtoseconds at room temperature — has since appeared in bounds on transport, chaos and thermalization. Yet whether quantum mechanics actually forbids a system from thermalizing faster than τPl remained a conjecture, and there are even apparent counterexamples, such as collision models in which a system relaxes arbitrarily fast by colliding quickly enough with identical copies that are already in equilibrium.
Paolo Abiuso (IQOQI Vienna, Austrian Academy of Sciences), Alberto Rolandi (IQOQI Vienna and TU Wien), John Calsamiglia, Pavel Sekatski and Martí Perarnau-Llobet now prove such a bound in a model-independent way, using tools from quantum information geometry and metrology. Their framework is deliberately adversarial: it imagines a ‘thermalization machine’ with complete control over the process, required only to output, for a non-trivial class of Hamiltonians, states close to the corresponding thermal (Gibbs) ensemble. Roughly speaking, a thermal state carries information about the Hamiltonian that generated it, so a device that thermalizes quickly must in effect estimate the Hamiltonian quickly, and quantum mechanics limits how fast that can be done. Combining the two constraints, the authors show that at finite temperature the thermalization time obeys τ ≥ τPl/2, and that at low temperature the bound takes the form τ ≥ ħ/Δ, where Δ is the spectral gap — closely connected to the quantum adiabatic theorem.
The bound is half the conjectured Planckian time, so it confirms the scaling rather than an exact prefactor, and it applies to arbitrary quantum processes that must produce near-thermal states across a class of Hamiltonians rather than to a single fixed one. It thereby places the Planckian timescale, so far an empirical regularity and a conjecture, on the footing of a theorem of quantum mechanics, with relevance to strange-metal transport, quantum chaos and the fast preparation of thermal states on quantum devices. The paper was published in Nature Physics on 24 September 2026 (arXiv:2506.19188).
In 1954 Robert Dicke predicted that N excited atoms packed within a wavelength of light do not radiate independently but can emit a collective burst — superradiance — whose peak intensity scales as N2 rather than N. The same physics has a downside. In densely packed quantum devices, vacuum fluctuations of the electromagnetic field mediate long-range dissipative interactions that cannot be switched off, so decay becomes correlated and a many-body quantum state can decay much faster than the sum of its individual decay rates. That makes a basic question pressing: what is the maximum decay rate of a large quantum system, and how does it scale with the number of particles? Answering it directly is hopeless for large N, because the Hilbert space grows exponentially.
Wai-Keong Mok, Avishi Poddar, Eric Sierra, Cosimo Rusconi, John Preskill and Ana Asenjo-Garcia (Caltech, the AWS Center for Quantum Computing and Columbia University) recast the problem: finding the maximal decay rate over all possible states is equivalent to finding the ground-state energy of a 2-local Hamiltonian. That is hard in general, but ideas from Hamiltonian complexity theory — quantum approximation theory and semidefinite-programming relaxations — yield rigorous upper and lower bounds. The bounds are universal: they hold for a broad class of Markovian many-body systems and depend only on global, spectral properties of the decoherence matrix, whose dimension grows only linearly with system size, rather than on microscopic details. For many physically relevant systems they are asymptotically tight and so become exact scaling laws; when decay is delocalized, the maximal rate grows as N times the largest eigenvalue of the decoherence matrix. In this language Dicke’s N2 splits into two factors: one N from permutation symmetry, the other from the delocalized dominant decay channel together with a non-vanishing density of excitations.
For arrays of atoms in free space, the scaling depends only on the dimensionality of the array and the range of the interactions, and can be expressed through the system’s optical depth; it determines, for example, how strong collective resonance fluorescence and the superradiant bursts of incoherently pumped atoms can become. The authors argue that these laws set fundamental limits on the decay rates of all quantum states of such systems and may ultimately constrain the scalability of quantum processors and simulators built from dense atom arrays, while also bounding applications that rely on collective emission, such as new light sources and quantum metrology. The paper appeared in Nature Physics on 25 September 2026 together with a Research Briefing (arXiv:2406.00722).
Photonic quantum technologies have already gone to space: satellites now perform space-to-ground quantum key distribution, distribute entanglement and test fundamental physics. But quantum states of light had never been used in orbit as a computational resource. There are practical reasons to want that. Satellites collect far more raw data than their downlinks can carry, so processing on board — ‘edge computing’ — and sending only the result saves bandwidth and energy, and the team argues that photon interference can enhance the machine-learning models used for such processing at fixed hardware resources. A quantum processor for space, however, has to cope with heat, vibration, electromagnetic interference and other disturbances, within tight limits on size and power.
A team led by Philip Walther at the University of Vienna, with partners including the Institute for Photonics and Nanotechnologies of Italy’s National Research Council (CNR-IFN), which made the photonic chip, reports a programmable quantum photonic processor operating on a nanosatellite in a Sun-synchronous orbit at about 510 km. Launched on 23 June 2025 on SpaceX’s Transporter-14 rideshare mission after two years of development, the payload is hosted by D-Orbit’s ION orbital transfer vehicle. It brings together the two building blocks of linear-optical quantum computing — a programmable six-mode universal photonic integrated circuit and photon–photon interference — with a free-space delay line that synchronizes the photons’ arrival and a control module containing an ARM processor and an FPGA. By programming distinct unitary transformations on the chip and tuning the two photons into indistinguishability, the researchers observed two-photon quantum interference in orbit, establishing on-board generation, manipulation and detection of non-classical light.
The demonstration is small — two photons in six modes is a scale long surpassed on the ground — and the preprint, posted in September 2026 and covered in the press late in the month, has not yet been peer reviewed; press coverage also notes technical issues during operation. Its significance lies in showing that the delicate interference on which photonic quantum computing relies survives the space environment in a device of about three litres. The researchers describe it as a first step towards classifying and compressing data directly on photonic hardware in orbit and downlinking only the inference result, and a space-based platform could also allow tests of quantum physics in a new environment.
After its final physics run, the Large Hadron Collider was switched off in mid-2026 to begin the third Long Shutdown (LS3), a multi-year programme to turn it into the High-Luminosity LHC (HiLumi LHC). The aim is to deliver far more collisions — about ten times the integrated luminosity foreseen in the LHC’s original design — for precision studies of the Higgs boson and searches for rare processes. The centrepiece is the replacement of the inner triplets, groups of three superconducting quadrupole magnets on both sides of the main experiments that squeeze the beams just before they collide: the tighter the focus, the higher the luminosity. In the LHC alone, about 1.2 km of magnets and other components are to be removed and replaced.
Since 7 September, teams have been dismantling the sections of the machine on either side of ATLAS and CMS to take out the 28 superconducting magnets, including the inner triplets, that are due to be replaced; these magnets have been in the LHC since its original construction. The first magnet interconnection was cut in mid-September, and CERN Director-General Mark Thomson visited LHC Point 1, home of the ATLAS experiment, to mark the start of the operation. The new inner triplets, the result of many years of research and development, are a major step beyond the niobium–titanium magnets now in the machine: they use niobium–tin (Nb3Sn) superconductor and a much larger aperture and are reported to be about 40% stronger. Around ALICE and LHCb the existing magnets will stay but will still be upgraded so that those experiments, too, benefit from the higher luminosity.
According to Jean-Philippe Tock, head of the LS3 Coordination Team, the first quadrupole of the new triplets should arrive in the tunnel at the start of 2029, and in total 16 cryostats and 28 cryo-assemblies will be installed. The HiLumi magnet programme is international: the final-focus quadrupoles are produced by CERN together with US Department of Energy laboratories, while the beam-separation dipoles beside them are being built by Japan’s KEK. While no beams circulate, thousands of researchers continue to analyse the vast data sets accumulated in the LHC era — the source of results such as this month’s measurement of entanglement between Z bosons — and to upgrade the detectors for the high-luminosity phase.
Boron, element 5, is famous for structural complexity. Its well-known forms — α- and β-rhombohedral boron and high-pressure phases such as the partly ionic γ-B28 — are built from B12 icosahedra; they are very hard, have high melting points and are semiconductors, but they are brittle and conduct electricity poorly. A boron allotrope combining high electrical conductivity with the ability to deform plastically has remained out of reach, partly because the usual one-step high-pressure, high-temperature synthesis pushes boron atoms into dense, tightly packed crystals.
Researchers led by Xiao-Ji Weng and Xiang-Feng Zhou at Yanshan University in Qinhuangdao, China, took a detour. They first made the sodium–boron compound Na4B60 — known since 1970 but little studied because good crystals are hard to grow — under high pressure, which yields larger crystals, and then heated it to about 900 °C under vacuum to drive out almost all of the sodium. What remains is Imma-B60, an open-framework allotrope in which B12 icosahedra are linked by triangular B3 units, with empty space where the sodium atoms used to be. The new phase has a narrow band gap of less than 0.2 eV and an electrical conductivity of about 9×102 S m−1, seven orders of magnitude higher than that of β-boron. It is also plastic: compressed nanopillars deform substantially without fracturing — Phys.org reports about 32% compressive strain — and once deformed the material does not spring back.
Artem Oganov, whose team identified the high-pressure γ-B28 phase in 2009, told Nature that the precursor-then-degassing strategy is not entirely new but is perhaps its most elegant and modern application, with potentially far-reaching consequences. The authors suggest that the route could open access to other unconventional framework materials — and to new allotropes of other elements — and provide a basis for mechanically resilient, functional inorganic materials, for instance in flexible electronics. The paper was published in Nature Chemistry on 23 September 2026.
🌊 / いびつな形の空洞で、波がカオスにならず一本の輪に集まる——双曲型メタマテリアルでつくった空洞で、波のカオスが抑えられ、安定した幾何学的な模様「波のアトラクター」が生じることを実証。等方的な通常の材料でできた不規則な空洞では波はカオス的に広がるが、双曲型の媒質では波は決まった角度の方向にしか進めず、反射しても媒質の軸に対する角度が保たれるため、非対称な空洞では閉じた軌道（アトラクター）に集束する。成層した海や回転する流体の内部波で知られてきた現象で、弾性波を使った実験でアトラクターの相転移、対称性に由来する特徴、欠陥への頑健性を示し、材料と空洞形状の対称性が同時に破れると、完全に線形な系でも頑健でキラルな波の組織化が生じることを明らかにした。小型で多機能な信号処理・センシング素子への応用が期待される（S. Yves・E. M. Renzi・S. A. Mannら、Nature Physics・9月28日掲載）
Waves trapped in an irregularly shaped cavity normally behave chaotically: in ordinary, isotropic materials, reflections from the walls scramble the wave pattern, much as a ball bounces unpredictably on a stadium-shaped billiard table. Geophysical fluids are a striking exception. In a density-stratified or rotating fluid, internal waves obey a hyperbolic wave equation: at a given frequency they can travel only along directions at a fixed angle to gravity or to the rotation axis, and on reflection they keep that angle with respect to the fluid’s axis rather than mirroring off the wall. In a non-symmetric container this geometric rule makes rays converge, after many reflections, onto a closed loop — a wave attractor — a phenomenon studied in laboratory tanks since the 1990s and relevant to energy focusing in oceans and planetary interiors.
Writing in Nature Physics, S. Yves, E. M. Renzi, S. A. Mann and colleagues show that cavities made from hyperbolic metamaterials behave in the same way. These are artificial structures whose extreme anisotropy gives them a hyperbolic rather than elliptic dispersion relation, so that waves of a given frequency are confined to fixed propagation directions. Odd-shaped cavities of this kind suppress chaotic wave dynamics and instead produce stable, geometric wave patterns. The team demonstrated the effect experimentally with elastodynamic (mechanical) waves in a hyperbolic metamaterial and revealed attractor phase transitions, features dictated by symmetry, and robustness against defects. Because symmetry is broken simultaneously in the material and in the cavity geometry, the attractors organize the waves in a robust, chiral way — even though the system is fully linear.
The results show that wave-attractor physics, familiar from oceanography and fluid dynamics, extends across natural and artificial hyperbolic media. The authors see possibilities for compact, multifunctional devices for wave-based signal processing and sensing that merge functions usually associated with large, wavelength-scale structures with those of deeply subwavelength cavities. Hyperbolic responses also occur naturally in some crystals — for phonon polaritons in hexagonal boron nitride, for example — and in engineered photonic and acoustic structures, which suggests that similar focusing could be explored in other wave systems. The paper was published in Nature Physics on 28 September 2026.
S. Yves、E. M. Renzi、S. A. MannらはNature Physicsに発表した論文で、双曲型メタマテリアルでつくった空洞が同じように振る舞うことを示した。双曲型メタマテリアルは極端な異方性をもつよう設計された人工構造で、分散関係が楕円型ではなく双曲型になり、ある周波数の波は決まった方向にしか伝わらない。こうした素材でつくったいびつな形の空洞では、波のカオス的な振る舞いが抑えられ、代わりに安定した幾何学的な波の模様が現れる。研究チームは、双曲型メタマテリアル中の弾性波（力学的な波）でこれを実験的に実証し、アトラクターの相転移、対称性に由来する特徴、欠陥に対する頑健性を明らかにした。材料と空洞形状の両方で同時に対称性が破れているため、アトラクターは波を頑健でキラルなかたちに組織化する——系は完全に線形であるにもかかわらずだ。
Quantum entanglement — correlations between particles that cannot be described as properties of each one separately — has been confirmed countless times with photons, atoms and superconducting circuits, all at low energies. In recent years the ATLAS and CMS experiments at CERN’s Large Hadron Collider (LHC) carried the question to the opposite extreme and observed spin entanglement between top quarks, the heaviest known fermions. Whether the same holds for the massive carriers of the weak force had not been measured. The decay of a Higgs boson into two Z bosons is a natural place to look: the Higgs has spin 0, so the spins of the two Z bosons must combine to zero, and because the Higgs is lighter than two Z bosons one of them is always ‘off shell’ (Z*). Each Z is a spin-1 particle with three helicity states — a ‘qutrit’ rather than a qubit — and the Standard Model predicts that the pair emerges in a coherent superposition of the combinations |+−⟩, |00⟩ and |−+⟩ that cannot be factorized.
The ATLAS Collaboration studied the ‘golden channel’ H → ZZ* → four charged leptons (electrons or muons), a clean, fully reconstructible final state with little background. Because the weak interaction is chiral, the directions in which the leptons fly out in the rest frame of each Z act as analysers of that boson’s spin, so angular correlations between the two lepton pairs give access to the elements of the two-boson spin-density matrix. Using the full Run 2 data set (140 fb−1 at 13 TeV) and the 2022–2024 part of Run 3 (164 fb−1 at 13.6 TeV), the team followed two routes. First, it measured two off-diagonal, entanglement-sensitive coefficients, C2,1,2,−1 = −0.71 ± 0.45 and C2,2,2,−2 = 0.08 ± 0.44, compatible with the Standard Model expectations of −0.97 and 0.64; by the Peres–Horodecki criterion a non-zero value of either one signals entanglement, but with today’s sample the statistical uncertainties are still large. Second, it used the full shape of the angular distribution to test the entangled Standard Model state against the only separable alternative allowed by the analysis assumptions — both Z bosons longitudinally polarized, |00⟩. That non-entangled hypothesis is rejected at 4.7 standard deviations (4.9σ expected).
ATLAS describes this as strong evidence for, and the first measurement of, quantum entanglement between massive vector bosons at the electroweak scale. Two caveats apply. The significance is below the conventional 5σ discovery threshold, and the hypothesis test relies on Standard Model inputs: a CP-even spin-0 Higgs boson and the standard chiral couplings of the Z to leptons. The collaboration also stresses that the analysis establishes the non-separability of the ZZ* spin state; it is not a Bell-inequality test and does not address alternatives to quantum mechanics. The CMS Collaboration reached a similar conclusion in a preliminary analysis of Run 2 and Run 3 data (CMS-PAS-HIG-25-011), and CERN presented the two results together on 17 September 2026. Oxford’s Alan Barr, a co-author who was among the first to propose collider-based entanglement tests, and colleagues note that it is one of the highest-energy confirmations of entanglement to date. The much larger data sets of the High-Luminosity LHC should turn such measurements into a new precision probe of the Higgs boson’s interactions. The ATLAS paper is published as Physical Review Letters 137, 111804 (arXiv:2603.26463).
Keywords: quantum entanglement, 量子もつれ, Z boson, Zボソン, Higgs boson, ヒッグス粒子, H→ZZ*→4ℓ, golden channel, spin density matrix, スピン密度行列
🌀 / OpenAIが「ナビエ–ストークス方程式の解は有限時間で爆発しうる」とする証明を公開（未査読）——クレイ数学研究所のミレニアム懸賞問題の一つに、AIが決着をつけたと主張。9月8日、同社は約1万体のAIエージェントを約88時間協調させて得たという166ページの論文原稿「Finite Time Blowup for Navier–Stokes」と、証明支援系Lean 4による形式化を公開した。主張は「任意の正の粘性について、滑らかな初期値と滑らかな外力のもとで、運動エネルギーが有界な大域的に滑らかな解が存在しない例がある（全空間ℝ³および周期境界ℝ³/ℤ³）」というもので、クレイ研究所の公式問題文の選択肢(C)・(D)（解の破綻）にあたる。外力なしの爆発を示したものではなく、査読と独立検証はこれからで、クレイ研究所は査読誌への掲載と学界での検証を経るまで判断しない立場。一方、ニューヨーク大学のTristan Buckmasterとハーバード大学のLevent Alpögeが進めていた関連研究とのクレジットをめぐる論争も起き、OpenAIは利用者データからの学習を否定している。物理的には、空気の場合この特異点は渦の太さが約70 nm（分子の平均自由行程程度）まで細くなったところで現れ、連続体近似そのものが破綻する領域にあたるという（OpenAI 9月8日公開、Nature 9月17日・18日報道）
The Navier–Stokes equations, written down in the nineteenth century, describe how viscous fluids such as water and air flow, and they underpin everything from aircraft design to weather forecasting. Yet it has never been proved that, in three dimensions, smooth solutions of the incompressible equations stay smooth for all time, or whether the velocity can instead become infinite at some point after a finite time — a ‘blow-up’ or singularity. The question has been open for roughly 90 years and is one of the seven Millennium Prize Problems of the Clay Mathematics Institute, each carrying a US$1-million award. The official problem statement offers four alternatives: (A) and (B) assert global existence and smoothness, in the whole space ℝ3 or with periodic boundary conditions; (C) and (D) assert breakdown, that is, the existence of smooth initial data and a smooth external force for which no smooth solution exists for all time.
On 8 September 2026 OpenAI announced that a group of about 10,000 coordinating AI agents, running a next-generation internal model for roughly 88 hours, had produced a proof of breakdown. The company released a 166-page manuscript, Finite Time Blowup for Navier–Stokes, together with a formalization in the proof assistant Lean 4. The claim is that for every positive viscosity there exist smooth initial data and smooth forcing for which no global smooth solution with uniformly bounded kinetic energy exists in ℝ3, and likewise no global smooth solution on the periodic torus — alternatives (C) and (D). In the main theorem the fluid starts at rest and is stirred by a smooth force that acts only in a bounded region and for a finite time; the velocity then becomes unbounded at a finite time while the kinetic energy remains bounded. The mechanism is a vortex that spirals inward while being stretched ever longer and thinner. A companion manuscript treats the inviscid Euler equations. The work builds on a long line of human mathematics that the manuscript cites, including Terence Tao’s blow-up for an averaged version of the equations and the Buckmaster–Vicol non-uniqueness results for weak solutions.
The status is that of a claim under scrutiny. The manuscript has not been peer reviewed, and the Clay Mathematics Institute says it will consider a solution only after publication in a refereed journal and further vetting by the community. The theorem also concerns forced flows: it does not show that a fluid left to itself can blow up, which remains open. The announcement immediately triggered a dispute over credit. According to Nature, Tristan Buckmaster (New York University), who with Levent Alpöge (Harvard University) had been attacking a related problem with AI tools from OpenAI and Anthropic, wrote publicly the day before that the company had taken up the problem after hearing of their work and that its models might have learned from his ChatGPT sessions. OpenAI told Nature that no user inputs after 3 July could have influenced the system, that it began work on 1 September and that it had not seen the pair’s work before it was made public. Specialists quoted by Nature say that much of the credit belongs to Buckmaster and Alpöge and to Diego Córdoba and Luis Martínez Zoroa in Madrid, and 25 Fields medallists have signed an open letter warning that AI tools blur attribution.
What would a confirmed blow-up mean for physics? George Karniadakis (Brown University) estimates that for air the singularity would appear when the vortex has thinned to about 70 nanometres — comparable to the distance an air molecule travels between collisions. At that scale the fluid is no longer a continuum, the founding assumption of the equations, so the infinity signals the limit of a model, not a physical infinity. Charles Fefferman (Princeton University) points out that the compressible version of the equations was already known to develop singularities, and the equations are known to fail for rarefied gases; descriptions based on the Boltzmann equation or on molecular dynamics take over there.
2026年9月8日、OpenAIは、次世代の社内モデルを動かす約1万体のAIエージェントを約88時間協調させ、破綻の証明を得たと発表した。同社は166ページの論文原稿「Finite Time Blowup for Navier–Stokes」と、証明支援系Lean 4による形式化を公開した。主張は、任意の正の粘性について、ℝ3では運動エネルギーが一様に有界な大域的に滑らかな解が存在せず、周期的なトーラス上でも大域的に滑らかな解が存在しないような、滑らかな初期値と滑らかな外力が存在するというもので、選択肢(C)と(D)にあたる。主定理では、静止した流体を、空間的にも時間的にも限られた範囲でだけ働く滑らかな外力でかき混ぜると、運動エネルギーは有界のまま、有限の時刻に速度が非有界になる。その仕組みは、内側へ巻き込みながら細く長く引き伸ばされていく渦である。粘性のないオイラー方程式を扱う姉妹原稿も公開された。原稿は、Terence Taoによる「平均化した」方程式での爆発解の構成や、Buckmaster–Vicolによる弱解の非一意性など、人間の数学者による長年の蓄積を引用している。
Why the Universe contains matter but almost no antimatter is unexplained, and one of the sharpest ways to look for a difference between the two is to compare protons and antiprotons held in ultrahigh-vacuum Penning traps — a test of CPT invariance. The BASE collaboration has measured the magnetic moments of both particles, to 0.3 and 1.6 parts per billion respectively, without finding a discrepancy. Such experiments are possible only at CERN’s Antimatter Factory (AMF), the world’s sole source of low-energy antiprotons, and that is now the problem: magnetic-field fluctuations caused by the operation of the facility limit the precision. The way forward is to take the antiprotons somewhere quieter.
For that purpose the team led by Christian Smorra and BASE spokesperson Stefan Ulmer (Heinrich Heine University Düsseldorf) built BASE-STEP, a cryogenic, open and transportable Penning-trap system with its own superconducting magnet. After a rehearsal with protons in October 2024, on 24 March 2026 they extracted 92 antiprotons from the AMF, disconnected the trap, loaded it onto a truck and drove it for about half an hour over a 7.5-km route around the CERN site. The paper now published in Nature, with Marcel Leonhardt as lead author, documents that not a single particle was lost and that the trap vacuum showed no measurable degradation. It also reports a storage record: the antiprotons, injected on 7 March, were kept in the mobile trap for more than a month, through particle-separation and merging studies and transport-preparation tests; after the drive they remained safely stored, and only one annihilated, during an extraction test on day 27.
Smorra calls the drive a successful dress rehearsal. The intended destination is a high-precision laboratory being set up at Düsseldorf, where comparisons of protons and antiprotons are expected to become at least 100 times more precise than is possible at CERN. Getting there would take at least 12 hours, during which the superconducting magnet must stay below 8.2 K; because a liquid-helium reserve would not last, the team is adding a cryocooler powered by a generator on the truck. Techniques for transferring the antiprotons from the transport trap into a stationary precision experiment also still have to be developed. In an accompanying News & Views, Nancy Paul and Davide Gamba write that the advance might one day let laboratories outside CERN use antimatter in their own experiments. BASE is an international collaboration that includes groups at RIKEN and the University of Tokyo.
That all bodies fall at the same rate — the universality of free fall, a cornerstone of Einstein’s theory of gravity — has been tested with ordinary atoms, with neutrons and, recently, with antihydrogen. All of these are built from first-generation particles. Whether gravity acts in exactly the same way on the heavier second and third generations of matter has never been tested directly. Muonium, a hydrogen-like atom made of a positive antimuon and an electron, is an ideal probe: it is electrically neutral, purely leptonic (so the strong interaction plays no part) and almost all of its mass sits in a second-generation antiparticle. The obstacles are the muon’s lifetime of only 2.2 microseconds and the fact that existing muonium sources emit a diffuse thermal cloud flying off in all directions at widely varying speeds.
Anna Soter’s group at ETH Zurich and the Paul Scherrer Institute (PSI), with lead author Jesse Zhang, has now produced the cold, intense beam that such an experiment needs. Antimuons from PSI’s accelerator, which delivers the world’s most intense continuous muon beams, are stopped in a thin layer of superfluid helium cooled close to absolute zero. There an antimuon picks up an electron and becomes muonium. The quantum liquid tolerates no impurities: the atom has a positive chemical potential in the helium and is pushed out, and at the surface this energy is converted into motion, so that the atoms shoot vertically out of the liquid — an ‘atomic cannon’, in Zhang’s words. The trick works only because muonium can travel through the superfluid without collisions and reach the surface within its brief life. Cryogenic silicon detectors that reconstruct the atoms’ paths from their decay products show a beam with a mean velocity of about 2,180 m/s and a spread of less than 150 m/s — far narrower than a thermal distribution, hence ‘superthermal’ — while the yield matches that of the most intense diffuse sources. The scheme goes back to an idea by PSI’s David Taqqu, a co-author.
Next comes an atom interferometer, in which Earth’s gravity shifts the interference pattern formed by the muonium matter waves; the beam, which now emerges vertically, will also have to be turned to the horizontal. Soter hopes to test the method with the new beam this year and to carry out the gravity measurement in two to three years, aiming at a per-cent-level determination of muonium’s gravitational acceleration — the first direct test of the weak equivalence principle with second-generation (anti)matter. The same beam should allow 1S–2S laser spectroscopy at the sub-kilohertz level, sharpening the muon mass and tests of bound-state quantum electrodynamics. A deviation from ordinary free fall would be a major surprise and could hint at a fifth force, but Soter says she is open-minded and simply wants to make the measurement for the first time. The work, supported by the Swiss research network Muoniverse, appeared in Nature Physics on 14 September 2026 with a News & Views by Jian Tang and Shihan Zhao.
The nucleus of thorium-229 has an excited state at an exceptionally low energy of about 8.4 eV, which puts the transition within reach of vacuum-ultraviolet lasers at a wavelength of 148 nm. That makes it the basis for a nuclear clock: a nucleus is far smaller than an atom and shielded by its electrons, so its transition frequency is unusually immune to external perturbations, and thorium nuclei can be embedded in huge numbers in a transparent crystal to make a robust solid-state device. Such a clock would also be a sensitive probe of possible variations of the fundamental constants. In 2024 several groups finally excited the transition with laser light in thorium-doped crystals. But those experiments relied on pulsed laser systems, in which only a small fraction of the photons is resonant with the extremely narrow nuclear line, and they detected the excitation through the nuclear fluorescence, which decays over a time of order ten minutes.
A team from TU Wien (Thorsten Schumm) and Germany’s national metrology institute PTB (Ekkehard Peik), with the Max Born Institute in Berlin, has now shown that the nucleus can be driven with a continuous-wave (CW), narrow-bandwidth, solid-state laser source at 148 nm delivering less than a nanowatt, and that the resonance can be seen directly in absorption — as light missing from the beam transmitted through a thorium-doped calcium fluoride (CaF2) crystal — rather than in fluorescence. Removing the slow fluorescence decay from the detection process allows fast signal acquisition, which is what a working clock needs in order to steer its laser onto the nuclear resonance. The spectra also distinguish two different thorium centres in the crystal, that is, thorium ions sitting in different local environments, and the team measured the isomeric shift between them — the small displacement of the nuclear line caused by the different electron density at the nucleus, familiar from Mössbauer spectroscopy.
The result moves the solid-state nuclear clock from proof-of-principle spectroscopy with pulsed light towards a continuously operated frequency standard built from compact solid-state lasers; the authors present it as a path to a far more efficient way of building such a clock. The group has since posted a preprint on a thorium-229 optical nuclear clock operated with a feedback loop (arXiv:2606.04997). Much remains to be done before nuclear clocks compete with the best optical atomic clocks: the available vacuum-ultraviolet power is still tiny, and the dependence of the line on the crystal site, which the two centres illustrate, has to be understood and controlled. Published in Nature on 16 September 2026 (arXiv:2604.16640).
💽 / 超伝導体そのものが「記憶」する——ケンブリッジ大学キャベンディッシュ研究所のAlexander G. Eatonらが、スピン三重項（p波）超伝導の有力候補である二テルル化ウラン（UTe₂）のバルク単結晶で、電流パルスだけで書き込み・消去できる超伝導メモリー効果を発見。磁場をかけて2つの異なる超伝導相（SC1とSC2）が共存する兆候を示す中間領域に置くと、直流電流の急な変調によって臨界電流密度J_cが高い準安定状態へ移り、その状態は少なくとも数時間保たれ、十分大きな擾乱を与えるまで元に戻らない。切り替えは刺激の強さと持続時間で制御できる。チームはこれを、乱れた非平衡のガラス的な渦糸状態が急冷（クエンチ）されて強いピン止めを得るためと解釈。磁性体や半導体との界面を必要としない、超伝導秩序そのものに根ざした機能で、低消費電力素子や極低温ニューロモルフィック・メモリーへの応用可能性を示す（Z. Wu・A. G. Eatonら、ケンブリッジ大学・カレル大学ほか、Nature・9月16日掲載、arXiv:2603.02450）
Electronics built from superconductors would dissipate far less energy than semiconductor circuits, and superconducting diodes and switches have advanced rapidly in recent years. A read/write memory has been harder to achieve. Existing schemes mostly combine superconductors with ferromagnetic layers, or trap magnetic flux in lithographically patterned loops that must be addressed with magnetic fields — the memory is engineered from outside rather than being a property of the superconductor. Uranium ditelluride (UTe2) is an unusual starting point: it is a leading candidate for spin-triplet (p-wave) superconductivity and one of the very few materials, alongside CeRh2As2, UPt3 and URhGe, that host more than one superconducting phase.
Zheyu Wu, Alexander G. Eaton and colleagues at the University of Cambridge’s Cavendish Laboratory, with collaborators including Charles University in Prague, measured current–voltage characteristics of bulk single crystals in a magnetic field tuned to an intermediate regime where two distinct superconducting phases, SC1 and SC2, show signs of coexisting. Deep inside either phase the curves look like those of any type-II superconductor in a field. In between, abrupt changes of the applied direct current leave the crystal in a metastable state with an enhanced critical current density. The material stays there on subsequent measurements — for at least several hours — and returns to its original behaviour only after a sufficiently large perturbation resets it. By choosing the amplitude and duration of current ramps and pulses, the researchers could switch selectively between the low and high critical-current states: the crystal ‘remembers’ which one it is in, and it is written, read and erased purely electrically.
The team attributes the effect to the vortex matter of the superconductor: a sudden perturbation quenches the magnetic flux lines into a disordered, out-of-equilibrium, glass-like configuration that is pinned more strongly and therefore carries a larger supercurrent, whereas gentle changes anneal them back. Because no magnetic or semiconducting interface is involved, the memory is intrinsic to the superconducting order itself. This is a proof of principle in a uranium compound that superconducts only at very low temperature and in an applied magnetic field, not a device; but the authors argue that it shows the vortex matter of an unconventional superconductor can be manipulated electrically to encode and process information, with possible uses in low-energy electronics, cryogenic neuromorphic memory and quantum-computing hardware. Published in Nature on 16 September 2026 (arXiv:2603.02450).
ケンブリッジ大学キャベンディッシュ研究所のZheyu Wu、Alexander G. Eatonらは、プラハのカレル大学などとの共同研究で、バルク単結晶の電流–電圧特性を、2つの異なる超伝導相SC1とSC2が共存する兆候を示す中間領域に磁場を合わせて測定した。どちらか一方の相の内部では、特性は磁場中の第二種超伝導体としてごく普通のものだ。ところが中間領域では、印加する直流電流を急に変化させると、結晶は臨界電流密度が高まった準安定状態に移る。物質はその後の測定でもこの状態にとどまり——少なくとも数時間——十分大きな擾乱でリセットされるまで元の振る舞いに戻らない。電流の掃引やパルスの大きさと長さを選ぶことで、研究チームは臨界電流の低い状態と高い状態を選択的に切り替えられた。結晶はどちらの状態にあるかを「覚えて」おり、書き込み・読み出し・消去がすべて電気的に行える。
General relativity predicts that a rotating mass drags spacetime around with it. For Earth this frame dragging (the Lense–Thirring effect) is minute — only a few parts in ten billion of the planet’s own rotation rate — and it has so far been measured only with satellites: to about 20% with the laser-ranged LAGEOS satellites in 2004, to about 18% by Gravity Probe B in 2011, whose four superconducting gyroscopes were limited by unexpected torques discovered after launch, and to about 2% in 2019 once the LARES satellite was added. These results depend on elaborate models of Earth’s gravitational field and on averaging over many orbits, and all of them used classical probes.
Kai-Isaak Ellers, Marios Christodoulou, Keith Schwab (Caltech) and K. Birgitta Whaley (University of California, Berkeley) propose to do the measurement in a laboratory with a quantum sensor: a loop of superfluid helium-4 interrupted by a single Josephson junction. A superfluid is described by a single macroscopic wavefunction, and rotation of the loop — or the off-diagonal terms that frame dragging adds to the spacetime metric — shifts its quantum phase around the loop, much as a magnetic flux shifts the phase of electrons in a SQUID; here the neutral helium atoms couple to gravity through their mass. The junction turns that phase into a measurable change in the superfluid flow. The authors derive how the frame-dragging, geodetic and Thomas contributions enter the signal and lay out a procedure for measuring them, then compute the thermal noise floor. With Josephson junctions made from nanoporous two-dimensional materials, which are expected to work at millikelvin temperatures far below the superfluid transition at 2.17 K, and with the lowest-loss mechanical materials, they find a noise spectral density of 5 × 10−17 rad s−1 Hz−1/2 at 10 mK — enough to resolve the frame-dragging rate to 0.2% within one second, a rotational sensitivity equivalent to one revolution in four billion years.
This is a theoretical proposal, not a measurement: the required junctions are ‘near-future’ technology, and an experiment would have to separate the relativistic signal from Earth’s vastly larger rotation, since the device senses the total rotational flux through the loop. The authors compare their scheme with GINGER, a proposed Earth-based measurement using ring-laser gyroscopes. If it can be realized, a local measurement would avoid averaging over many satellite orbits, could be improved step by step in the laboratory, and would couple general relativity directly to a macroscopic quantum phenomenon. Published in Physical Review Letters 137, 121404 on 17 September 2026 and highlighted in a Physics Synopsis (arXiv:2510.20772).
Since 1996, when a femtosecond laser pulse was shown to quench the magnetization of nickel in less than a picosecond, ultrafast demagnetization has been a route towards magnetic switching far faster than magnetic fields allow. Thermally driven magnetization dynamics, however, has a built-in speed limit: as a conventional ferromagnet approaches its Curie temperature — or is heated close to it by a stronger laser pulse — its response becomes slower, not faster. This critical slowing down is a key obstacle for ultrafast magnetization-switching devices.
Zhao-Hua Cheng’s group at the Institute of Physics of the Chinese Academy of Sciences (first author Yu-Han Gao) reports the opposite behaviour in superlattices of CaRuO3/SrTiO3, which are moderately correlated, weak itinerant ferromagnets. In their time-resolved experiments the demagnetization rate increases with rising temperature, with increasing pump fluence and with an applied magnetic field — an anomalous acceleration where ordinary ferromagnets slow down. As the title of the Letter indicates, the authors attribute the effect to spin fluctuations, which dominate the magnetism of weak itinerant ferromagnets and here govern how fast the magnetization collapses after the laser pulse.
In an accompanying Viewpoint in Physics, Daniel Steil (University of Göttingen) writes that the behaviour runs contrary to expectations and holds promise for high-speed spintronics. It is a materials-physics result rather than a device: whether the mechanism can be carried over to magnets that order at room temperature remains to be seen. But it identifies a class of materials — weak itinerant ferromagnets with strong spin fluctuations — in which the usual thermal speed limit does not apply. Published in Physical Review Letters 137, 126903 on 16 September 2026; the work also involves the University of Chinese Academy of Sciences and the Songshan Lake Materials Laboratory.
⏳ / 量子インターネットの「時間合わせ」に必要な長寿命メモリー——スペイン・光科学研究所（ICFO）のHugues de Riedmattenらが、希土類イオンを添加した結晶Pr³⁺:Y₂SiO₅による吸収型量子メモリーに、原子周波数コム（AFC）スピン波プロトコルとXY4スピン再位相化（リフェージング）パルス列を組み合わせ、通信波長帯の光子で到来を告知（ヘラルド）された単一光子を最長180マイクロ秒保存したうえで、告知光子と保存光子のあいだの非古典的な相関（相互相関値は最大4.6±0.4）を確認。古典光領域ではAFCエコーを約3ミリ秒まで観測した。スピン再位相化を組み込んだ吸収型メモリーでの単一光子保存はこれまで実現しておらず、離れたノード間でもつれ交換のタイミングを合わせる量子中継器の実現に向けた一歩（A. E. Rodríguez-Moldes・F. Appas・H. de Riedmattenら、ICFO、Phys. Rev. Lett. 137, 120803・9月16日掲載、arXiv:2609.02579）
Photons are lost exponentially in optical fibre, so a continental-scale quantum network will need quantum repeaters: the link is divided into segments, entanglement is created across each one and then extended by entanglement swapping. Every segment requires a quantum memory that holds one photon of a pair while its partner — ideally at a telecom wavelength — travels to a distant measuring station that ‘heralds’ success, and neighbouring memories have to be synchronized. The memory must therefore store a photon at least as long as light and the classical signals need to cross the segment: hundreds of microseconds for tens of kilometres. Rare-earth-doped crystals are attractive absorptive memories because they allow massive multiplexing, high efficiency and integration on chips, but long storage requires transferring the excitation to nuclear-spin (hyperfine) states and protecting it with sequences of radio-frequency pulses — a ‘spin rephasing’ that has been implemented in a variety of solid-state systems. Storing genuine single photons in an absorptive memory that includes it had remained elusive.
Hugues de Riedmatten’s group at ICFO near Barcelona, with Alberto Rodríguez-Moldes and Félicien Appas as joint lead authors, has now done it. A cavity-enhanced parametric down-conversion source produces photon pairs; one photon, at a telecom wavelength, heralds its partner, which is absorbed in a Pr3+:Y2SiO5 crystal using the atomic-frequency-comb (AFC) spin-wave protocol. While the excitation sits in the hyperfine states, an XY4 sequence of pulses repeatedly refocuses the spins and shields them from environmental perturbations. With classical light the team observed AFC echoes after storage times of up to about 3 ms. With true heralded single photons it demonstrated non-classical correlations between the heralding and the stored photons for storage times of up to 180 μs, with measured cross-correlation values as high as 4.6 ± 0.4.
In 180 μs light travels several tens of kilometres through fibre, so the storage time is already in the range needed to wait for a heralding signal from a distant node — which is why the accompanying Physics Synopsis presents the device as a way to synchronize a long-distance quantum network. The demonstration concerns storage and retrieval of single photons, not yet entanglement between remote memories, and the gap between the 3 ms reached with classical pulses and the 180 μs reached with single photons suggests that noise and efficiency, rather than spin coherence alone, currently limit the quantum regime. Published in Physical Review Letters 137, 120803 on 16 September 2026 (arXiv:2609.02579).
バルセロナ近郊の光科学研究所（ICFO）のHugues de Riedmattenのグループ（Alberto Rodríguez-MoldesとFélicien Appasが共同筆頭著者）がこれを達成した。共振器で増強したパラメトリック下方変換光源が光子対を生成し、通信波長帯の光子が相方の到来を告知する。相方の光子は、原子周波数コム（AFC）スピン波プロトコルによってPr3+:Y2SiO5結晶に吸収される。励起が超微細準位にあるあいだ、XY4パルス列がスピンを繰り返し再収束させ、環境からの擾乱から守る。古典光を使った測定では、約3ミリ秒までの保存時間でAFCエコーが観測された。告知された本物の単一光子では、最長180マイクロ秒の保存時間まで、告知光子と保存光子のあいだの非古典的な相関が実証され、相互相関値は最大で4.6±0.4に達した。
Heisenberg’s uncertainty principle of 1927 was cast into its general textbook form by Robertson in 1929 — the product of the standard deviations of two observables A and B is bounded from below by half the magnitude of the expectation value of their commutator — and sharpened by Schrödinger in 1930, who added a second term built from the covariance of the two observables. The resulting Robertson–Schrödinger relation has been a workhorse of quantum theory for almost a century.
Gen Kimura (now at Tohoku University), Aina Mayumi (Shibaura Institute of Technology), Hiromichi Ohno (Shinshu University), Jaeha Lee (Institute of Industrial Science, University of Tokyo) and Dariusz Chruściński (Nicolaus Copernicus University, Toruń) prove a universal strengthening of the relation. To the familiar commutator and covariance terms they add a further non-negative contribution that stems purely from the non-commutativity of the two observables. It is the expectation value of a positive observable — the squared modulus of the commutator — so, like the original bound, it involves only directly measurable quantities. The extra term is a genuinely quantum trade-off that had been overlooked, and it becomes more pronounced as the state becomes more mixed. For two-level systems (qubits) the inequality turns into an exact equality for every state and every pair of observables, which means that in this case the bound is tight in the strongest possible sense and cannot be improved further.
The result also rigorously proves, and extends, an uncertainty relation that members of the same group had conjectured in 2024 (Phys. Rev. A 110, 062215). It is a mathematical theorem rather than an experiment, but because every term is an expectation value the authors also discuss possible experimental tests. Tighter uncertainty relations are more than a matter of principle: they set the ultimate limits for quantum measurement and metrology and are widely used as tools in quantum information. Published open access in npj Quantum Information on 21 September 2026 (arXiv:2504.20404).
Einstein’s equivalence principle says that for an observer in free fall, gravity locally disappears: someone falling freely in a lift feels weightless. The principle has passed extraordinarily precise tests with ordinary matter, but it was unclear how to test it with quantum objects, which behave as waves and can effectively travel along more than one path at once. Although quantum particles have long been used to measure gravity, the specific quantum phase that a freely falling wave is predicted to accumulate had never been measured directly.
An international team led by Ron Folman (Ben-Gurion University of the Negev), with Wolfgang Schleich (University of Ulm) providing the quantum theory of the apparatus and Vlatko Vedral and Sir Roger Penrose (Oxford) among the co-authors, built what they call a Quantum Galileo Interferometer. Clouds of rubidium atoms cooled to just above absolute zero were manipulated close to the surface of an atom chip. Microwave pulses put each atom into a superposition of two paths. Tiny wires on the chip generated precisely controlled magnetic fields: one part of the atomic wave responded to the field and was held stationary relative to the laboratory by an upward force exactly cancelling gravity, while the other part was kicked upward by a magnetic pulse and then switched into a state almost unaffected by the field, so that it fell freely along a ballistic trajectory like a thrown ball. A final magnetic pulse brought the two parts back together, and their interference revealed the tiny phase difference accumulated while one fell and the other stood still.
The measured phase is exactly the one predicted when the equivalence principle is applied to a quantum wave, giving what the team describes as the first direct measurement of the predicted quantum phase of a freely falling object. The result does not unite quantum mechanics with gravity, nor does it show that gravity itself is quantum; it shows that the equivalence principle remains consistent with quantum mechanics in the regime tested. Nor does it overturn Penrose’s proposal that quantum superpositions of sufficiently massive objects held for long enough could collapse: the masses and timescales here are far too small for that. The team hopes the technique will be a stepping stone to experiments with much heavier objects such as nanodiamonds, which are now under way in the same group. Participating institutions include Ben-Gurion University, Oxford, Southampton, the German Aerospace Center’s Institute of Quantum Technologies and Universität Ulm, and Texas A&M University. Published in Science Advances on 2 September 2026.
Ron Folman（ベングリオン大学）率いる国際チームは、ウルム大学のWolfgang Schleichが装置の量子論的理解を担い、オックスフォード大学のVlatko VedralとSir Roger Penroseも共著者に加わって、「量子ガリレオ干渉計」と名付けた装置を開発した。絶対零度のすぐ上まで冷却したルビジウム原子の雲を原子チップの表面近くで操作し、マイクロ波パルスで各原子を2つの経路の重ね合わせ状態に置く。チップ上の微小な配線が精密に制御された磁場を発生させ、原子の波の一方はこの磁場に応答して、重力を正確に打ち消す上向きの力で実験室（地球）に対し静止させられる。もう一方は磁気パルスで上方へ打ち出されたのち、磁場の影響をほとんど受けない状態に切り替えられ、投げ上げたボールのように弾道軌道で自由落下する。最後にもう一度磁気パルスで両者を再結合させると干渉が起こり、一方が落下し他方が静止しているあいだに生じた微小な位相差が読み取れる。
On 9 October 2022 the gamma-ray burst GRB 221009A, quickly nicknamed the “BOAT” (brightest of all time), was detected at a redshift of about 0.15, roughly two billion light-years away. China’s LHAASO observatory recorded thousands of photons above 500 GeV from it, and the Carpet air-shower array at the Baksan Observatory in the Russian Caucasus reported an even more extreme event: a photon of about 251 TeV in its 2022 partial dataset, which the collaboration’s recently completed full analysis now puts at 300 (+43/−38) TeV. The absence of muons in the shower argues that this was a genuine photon rather than a misidentified hadron. The problem is that, according to conventional physics, such a photon should never have arrived: over intergalactic distances, photons of these energies collide with photons of the cosmic microwave background (CMB) and are converted into electron–positron pairs.
Giorgio Galanti (INAF, Milan) and Marco Roncadelli (INFN Pavia and INAF Brera Observatory) systematically compared the possible explanations. Within standard propagation models the detection is in strong tension with expectations. Scenarios in which photons oscillate into axion-like particles (ALPs) — light pseudoscalars that could travel unimpeded and convert back into photons near the Galaxy — are, on their own, strongly disfavoured. Instead, the photon is compatible with specific frameworks of Lorentz-invariance violation (LIV), in which a modified photon dispersion relation raises the threshold for pair production on the CMB. Requiring the photon to survive pins down the LIV energy scale: it must lie below 1.22 (+0.19/−0.22) × 1021 GeV in the linear case, about a hundred times the Planck energy, and below 2.03 (+0.17/−0.22) × 1013 GeV in the quadratic case, both at 95% confidence.
Finally, the authors outline new scenarios in which standard photon–ALP oscillations are combined with LIV-modified photon propagation. These give a consistent interpretation of the highest-energy photons from GRB 221009A detected by both LHAASO and Carpet. The claim rests on a single Carpet event whose association with the burst is itself a matter of discussion, and Lorentz-invariance violation would be a radical departure from special relativity, so independent confirmation by future detectors will be essential; but the analysis shows how a single extreme photon can probe quantum-gravity-motivated physics at energy scales far beyond the reach of any accelerator. Published open-access in Physical Review Letters on 8 September 2026.
The 1998 discovery that the expansion of the Universe is accelerating, recognized with the 2011 Nobel Prize in Physics, rests on Type Ia supernovae serving as standardizable candles. As reported earlier on this page, a Yonsei University team argues that the standardized brightness of these explosions drifts with the age of their progenitor stars, and that correcting for this “progenitor-age bias” removes the evidence for acceleration; a Southampton-led group replied in June 2026 that the supernova data are robust to host-galaxy age evolution. The dispute has now drawn in one of cosmology’s most persistent sceptics of dark energy.
Subir Sarkar of Oxford’s Rudolf Peierls Centre for Theoretical Physics, with Animesh Sah and Mohamed Rameez of the Tata Institute of Fundamental Research in Mumbai, re-examined Pantheon+, a compilation of more than 1,700 Type Ia supernovae, applying the proposed progenitor-age correction. Once the correction is included, the data no longer favour a uniformly accelerating universe; taken as a whole, they indicate that cosmic expansion is slowing. The team also tested whether the apparent acceleration is the same in every direction, as the standard cosmological model assumes. They report that the inferred acceleration is directed mainly along the direction of our local motion, the direction marked by the hot spot in the cosmic microwave background, and dies away with distance — a dipole-like pattern that Sarkar and Rameez have argued for since 2019. Because an effect of the quantum vacuum should not depend on direction, they conclude that this anisotropy rules out dark energy regardless of whether the brightness correction is applied, and that the correction then turns the remaining isotropic component into a deceleration.
The conclusion is contested. In the same issue of Monthly Notices of the Royal Astronomical Society, a paper by Phil Wiseman, Brodie Popovic, Mark Sullivan and colleagues, including Adam Riess, Dan Scolnic, Brian Schmidt and Oxford’s Maria Vincenzi, concludes that the observations continue to support an accelerating universe once host-galaxy properties are handled consistently; Vincenzi says that this analysis by experts in supernova environments gives further confidence in the cosmological framework built up over three decades. Both camps agree on what comes next: the Rubin Observatory’s Legacy Survey of Space and Time is expected to deliver hundreds of thousands of well-characterized supernovae, enough to test whether the acceleration is real, whether it varies with direction, and what role, if any, dark energy plays.
この結論には異論がある。Monthly Notices of the Royal Astronomical Societyの同じ号には、Phil Wiseman・Brodie Popovic・Mark Sullivanらに加え、Adam Riess、Dan Scolnic、Brian Schmidt、そしてオックスフォード大のMaria Vincenziが名を連ねる論文が掲載され、母銀河の性質を整合的に扱えば観測は依然として加速する宇宙を支持すると結論している。Vincenziは、超新星の環境を熟知した専門家によるこの解析が、30年かけて築かれた宇宙論の枠組みへの信頼をさらに高めるものだと述べる。次に何が来るかについては両陣営の見解が一致している。ルービン天文台のLSST（時空間レガシーサーベイ）が数十万個の高品質な超新星をもたらし、加速が本物か、方向によって変わるか、暗黒エネルギーが実際どんな役割を果たしているのかを検証するのに十分なデータになると期待されているのだ。
Neutral hydrogen, the most abundant element in the Universe, emits a faint radio line at a wavelength of 21 centimetres; as the Universe expands, the line from distant gas is stretched to longer wavelengths, so its observed wavelength tells astronomers how far back in cosmic history they are looking. Hydrogen intensity mapping exploits this by measuring the combined 21-cm glow from enormous volumes of space without resolving individual galaxies. It has long been seen as an efficient way to chart large-scale structure, but the signal is buried beneath far brighter foreground emission from our Galaxy and from radio sources, human-made interference and instrumental effects. Until now, reliable detections at these distances have typically relied on cross-correlating the radio maps with optical galaxy surveys.
A team led by Sourabh Paul, with Laura Wolz (University of Manchester) and colleagues at the University of the Western Cape, has now made a direct detection using observations from South Africa’s 64-dish MeerKAT array alone. Analysing about 96 hours of data on a field called DEEP2 — observations not originally designed for intensity mapping — they extracted the hydrogen signal at two epochs, redshifts of about 0.32 and 0.44, corresponding to emission that travelled roughly four to five billion years to reach Earth. The measurements trace the distribution of hydrogen on scales of several million light-years, comparable to the separation between the Milky Way and Andromeda, and two independent analysis methods gave consistent results.
Paul calls the detection an exciting milestone showing that intensity mapping is becoming a practical tool for cosmology; Wolz notes that extracting the signal from data not designed for the purpose demonstrates the scientific value of MeerKAT and points the way to the SKA-Mid telescope, for which MeerKAT is a precursor. Longer integrations, larger sky areas and better control of systematics should let future surveys map neutral hydrogen across vast volumes, probing how galaxies formed, how dark matter shapes the cosmic web and how dark energy drives the expansion. Published in The Astrophysical Journal Letters (vol. 1005, L56).
🔺 / 3つの粒子が束縛した「トリオン」のホール効果を初観測——コーネル大学のJie ShanとKin Fai Makらが、厚さ1.5〜2 nmの六方晶窒化ホウ素（hBN）で隔てた単層WSe₂（正孔層）と二層MoSe₂（電子層）からなるクーロン結合した電子–正孔二層で、励起子（電子–正孔対）にさらに電子または正孔が1個束縛した3体の複合粒子「トリオン」の平衡流体をつくり、垂直磁場下でトリオンがローレンツ力を受けて横方向にそれる「トリオン・ホール効果」を実証。1.5 K・4 Tでの磁気輸送とクーロン引きずり（クーロンドラッグ）測定を組み合わせると、負に帯電したトリオンが存在する領域では、電子層を開放にしたまま測った正孔層に「電子型」の負のホール応答が現れる——正孔層の電流をトリオンが担い、引きずられた電子層の電流を自由電子の逆流が相殺するためだ。効果はトリオンが解離する高温（数K以上）や高ドープ密度で消え、束縛エネルギーは低密度極限で約0.7〜1.2 meV。トリオンが可動な荷電キャリアとして振る舞う直接の証拠で、トリオン版の量子ホール状態の実現に一歩（R. Chaturvedi・P. X. Nguyen・P. Knüppel・渡邊賢司・谷口尚〔NIMS〕・K. F. Mak・J. Shan、Nature Physics・9月10日掲載・オープンアクセス）
In Coulomb-coupled electron–hole double layers, an electron-doped semiconductor and a hole-doped semiconductor are separated by an insulating spacer so thin that electrons in one layer bind to holes in the other, forming equilibrium excitons without any magnetic field. Such systems, built from atomically thin transition-metal dichalcogenides, have revealed excitonic insulators and perfect Coulomb drag in recent years. Adding extra electrons or holes to the exciton fluid creates trions, bound states of two electrons and one hole or vice versa; an equilibrium trion liquid was reported in 2025 when one layer holds twice the carrier density of the other. Because trions are both stable and charged, they should feel the Lorentz force in a perpendicular magnetic field, so a trion Hall effect is expected — but it had never been observed.
Raghav Chaturvedi, Phuong X. Nguyen, Kin Fai Mak and Jie Shan (Cornell University and the Max Planck Institute for the Structure and Dynamics of Matter), with Patrick Knüppel and with Kenji Watanabe and Takashi Taniguchi of NIMS, studied dual-gated devices in which a monolayer of WSe2 (holes) and a bilayer of MoSe2 (electrons, behaving effectively as a monolayer) are separated by 1.5–2 nm of hexagonal boron nitride. At 1.5 K and 4 T, magnetotransport and helicity-resolved optical spectroscopy map out the phase diagram: under a field, the narrow trion regions of the zero-field diagram expand into two triangular regions hosting mixtures of excitons and positive or negative trions, bounded below by the lines where one density is twice the other and above by trion dissociation at high doping. Because the trions are spin singlets, they become unstable once the valley Zeeman splitting exceeds their binding energy, which sets a field window of order 10 T.
The Hall measurements deliver the signature. Inside the trion regions the Hall drag resistance is large, positive for positive trions and negative for negative ones, and vanishes outside. Most strikingly, in the negative-trion region the ordinary Hall resistance measured on the hole layer, with the electron layer in open circuit, turns negative — electron-like — below a few kelvin. A model with a 4×4 conductivity tensor explains why: the current in the hole layer is carried by trions, the current they drag in the electron layer must be balanced by a counterflow of free electrons, and the Hall resistance becomes −(B/e)(1/nT + 2/ne), with nT and ne the trion and free-electron densities. The effect disappears smoothly as trions dissociate with rising temperature; the crossover, about 8 K (0.7 meV) in the low-density limit, and an optical estimate of about 1.2 meV give the trion binding energy. Resistance oscillations seen in the positive-trion region are not trion Landau levels — the extracted mass of 0.65 electron masses is far below the expected trion mass of 1.4–1.6 — but recurring transitions between hole quantum Hall states and the trion phase. Realizing trion quantum Hall states will require cleaner samples in which the binding energy exceeds the cyclotron energy. Published open-access in Nature Physics on 10 September 2026.
🐦 / 「追う者と逃げる者」が入り混じった群れは、大きくなるとカオスに沈む——ザールラント大学のChul-Ung Woo・Heiko Riegerとソウル市立大学のJae Dong Nohが、群れ運動（フロッキング）の代表的モデルであるVicsekモデルに、作用・反作用が釣り合わない非相反的な向きそろえ相互作用をする2種の粒子を導入して解析。個々の粒子に固有のキラリティ（回転の癖）がなくても群れ全体が一斉に回転するカイラル秩序は小さな群れでのみ生き残り、大きな群れでは不安定な方向の数が系のサイズに比例して増える「広がる時空カオス」に置き換わる。両者を隔てるのはカイラル軌道の旋回半径で決まる有限波長の不安定性で、系がこの長さを超えるとカオスが展開する——非相反性が集団振動やアクティブ乱流を生むことは知られていたが、流体力学的乱流以外の系でも広がる時空カオスを引き起こすことを示した。同日掲載のC. Myin・S. Saha・B. Mahaultの論文も非相反な群れでの創発的カイラル秩序の崩壊と欠陥カオスを報告し、Physics誌がViewpoint「非相反性が群れをカオスに送り込む」で両論文を特集（C.-U. Woo・J. D. Noh・H. Rieger、Phys. Rev. Lett. 137, 118301／C. Myin・S. Saha・B. Mahault、Phys. Rev. Lett. 137, 118302・ともに9月8日掲載、Physics 19, 115）
Flocks of birds, schools of fish and swarms of self-propelled colloids are studied with models of active matter, whose best-known example is the Vicsek model: each particle moves at constant speed and aligns its direction with its neighbours, and above a critical density the whole population spontaneously moves together. Real populations, however, often mix species that do not treat one another symmetrically — a predator turns toward its prey while the prey turns away — and such nonreciprocal interactions violate the action–reaction principle. They are known to produce collective oscillations in which the group rotates even though no individual particle has any intrinsic chirality, and, in active hydrodynamics, turbulent regimes. Whether nonreciprocity can also drive genuine spatiotemporal chaos outside hydrodynamic turbulence had remained unclear.
Chul-Ung Woo and Heiko Rieger of Saarland University and Jae Dong Noh of the University of Seoul answer this with the two-species Vicsek model, in which particles of one species align with the other in a nonreciprocal way. They find that the two states coexist and are separated by system size: chiral order, with the mixed flock circling coherently, survives only in small flocks, while large flocks fall into extensive spatiotemporal chaos, in which the number of unstable directions grows in proportion to the size of the system. The boundary is a finite-wavelength instability whose scale is set by the rotation radius of the chiral orbits; once the system is larger than this length, the coherent rotation breaks up and chaos unfolds.
In a companion Letter published the same day, Charlotte Myin, Suropriya Saha and Benoît Mahault report the breakdown of emergent chiral order and defect chaos in nonreciprocal flocks, and Physics magazine highlights both papers in a Viewpoint titled “Nonreciprocity Sends Flocks into Chaos.” Beyond the statistical physics of non-equilibrium systems, the results bear on mixed biological populations and on robot swarms whose agents follow asymmetric rules: coherent collective motion in such systems may be a small-system phenomenon, destined to become disorderly as the group grows. Both papers appeared in Physical Review Letters on 8 September 2026.
Companion paper / 姉妹論文：
C. Myin, S. Saha & B. Mahault, “Breakdown of Emergent Chiral Order and Defect Chaos in Nonreciprocal Flocks,” Phys. Rev. Lett. 137, 118302 (8 September 2026)（PRL 137巻11号の目次 / table of contents）
Protons and neutrons are made of up and down quarks; the unstable strange quark can turn into an up quark, and the strength of that transition is the element |Vus| of the Cabibbo–Kobayashi–Maskawa (CKM) matrix, whose unitarity is a cornerstone test of the Standard Model. Current determinations from kaon and tau decays show tensions that could hint at new physics. Semileptonic hyperon decays such as Λ → pe−ν̄e offer an independent probe, but they have been largely untapped for 35 years: with an undetected neutrino and no handle on the hyperon’s spin, earlier experiments lacked the kinematic information needed to pin down the relevant form factors. The most precise measurement of the axial-vector coupling dates from a 1990 Fermilab hyperon-beam experiment with about 37,000 candidate events.
The BESIII Collaboration at the Beijing Electron–Positron Collider now revives the decay with a technique that turns quantum mechanics into a measuring tool. In 10.087 billion J/ψ decays, the J/ψ resonance produces polarized, spin-entangled ΛΛ̄ pairs. Reconstructing the antihyperon’s decay Λ̄ → p̄π+ “tags” the event, fixes the Λ momentum from the known beam energy, and lets the team reconstruct the semileptonic decay’s full kinematics, including the momentum transfer, event by event. Because the pair’s polarization and spin correlations were already measured precisely in earlier BESIII work, only three physics couplings remain free in a seven-dimensional likelihood fit. From 1,854 ± 49 signal events the team obtains the first absolute branching fraction for the electron mode, (8.16 ± 0.22 ± 0.15) × 10−4, the axial-vector coupling, a weak-magnetism coupling of 0.89 ± 0.38 — consistent with the conserved-vector-current prediction of 0.97, whereas the 1990 value was 0.15 ± 0.30 — and the first measurement of the weak-electricity coupling in Λ decay, −0.19+0.65−0.63 ± 0.18, consistent with zero and hence with the absence of second-class currents.
Combining the branching fraction with recent lattice-QCD form factors yields |Vus| = 0.2339 ± 0.0041, a model-independent determination consistent with CKM unitarity at the 1.4σ level; the SU(3)-symmetric alternative gives 0.2194. The team also extracts, for the first time in any baryon semileptonic decay, the theory-free product |Vus|·√(f12+3g12) = 0.4496 (±0.0061 stat., ±0.0041 syst.). Although the Fermilab dataset was about 20 times larger, the entanglement-based analysis reaches comparable precision on weak magnetism and accesses a parameter that was previously unmeasurable, and it applies directly to other hyperon decays at BESIII, at PANDA and at proposed super tau–charm factories — a path toward a |Vus| determination competitive with kaon decays. Published open-access in Nature 657, 92–97 on 2 September 2026, with a News & Views by Matt Kenzie (Cambridge).
Superradiance, predicted by Dicke in 1954, is one of the most striking collective quantum effects: when many emitters radiate into a common mode and nothing records which emitter produced a given photon, the emission pathways interfere constructively and the peak rate scales as N2 rather than N. It underlies superradiant lasers and has been seen in atomic gases and waveguides. In 2025, Jones and Formaggio proposed extending it from photons to neutrinos: a Bose–Einstein condensate of about one million radioactive 83Rb atoms, each of which decays by nuclear electron capture into a 83Kr atom plus a neutrino, was predicted to emit neutrinos superradiantly, shortening the 86-day half-life to roughly 2.5 minutes — a nearly 50,000-fold enhancement, and the first “neutrino laser” (Phys. Rev. Lett. 135, 111801).
Wolfgang Ketterle’s group at MIT (Yu-Kun Lu and Hanzhen Lin) has now examined the idea in two complementary papers and finds that it fails for reasons rooted in quantum mechanics itself. The first paper considers the most favourable case — every neutrino emitted into a single mode, no recoil, no decoherence — and shows that collective enhancement requires successive decays to walk the system up a ladder of collective states. But the 83Kr product is a fermion: the Pauli exclusion principle forbids two krypton atoms from occupying the same quantum state, so the chain stops after the very first decay and the maximum emission rate is proportional only to N, exactly as for independent emitters. The second paper grants krypton hypothetical bosonic statistics and asks whether a condensate in free space can support single-mode emission at all. The emission cone is set by the condensate geometry and the emitted wavelength; for the highly relativistic neutrinos of nuclear decay that wavelength is picometres, giving a cooperativity C ∼ 10−12, so that even a million atoms yield NC ∼ 10−6, far below the superradiance threshold NC ≫ 1. Worse, momentum conservation gives each recoiling krypton atom a velocity of a few thousand metres per second, so it crosses the condensate in under a microsecond — far shorter than the build-up time of collective emission — and its motion encodes which-atom information that destroys the coherence.
The verdict is not an engineering limitation but a fundamental one, and the two studies distil the ingredients any collective quantum emitter must have: indistinguishable emission pathways, sufficient cooperativity, and coherence that survives long enough for collective dynamics to develop. As the accompanying Viewpoint by Ana Maria Rey, James Thompson and Haoqing Zhang (JILA) puts it, understanding why an appealing proposal fails can be as instructive as showing that it works. Both papers appeared in Physical Review Letters 137 (101804 and 101805) on 2 September 2026.
1954年にディッケが予言した超放射は、最も鮮やかな集団的量子効果の一つだ。多数の放出体が共通のモードに放射し、どの放出体がその光子を出したかの情報が残らないとき、放出経路が建設的に干渉して最大放出率はNではなくN2に比例する。超放射レーザーの原理であり、原子気体や導波路で観測されてきた。2025年、JonesとFormaggioはこれを光子からニュートリノへ拡張する提案をした。放射性の83Rb原子約100万個のボース・アインシュタイン凝縮体（BEC）では、各原子が原子核の電子捕獲によって83Kr原子とニュートリノに崩壊する。このニュートリノ放出が超放射的になれば、86日の半減期が約2.5分に縮まる——放出率にして約5万倍の増強であり、史上初の「ニュートリノ・レーザー」になるというのだ（Phys. Rev. Lett. 135, 111801）。
🧲 / 砂粒大の「浮いた磁石」で、細胞並みに重い暗黒物質を初めて探索——ライス大学とライデン大学のPOLONAISE実験が、絶対零度近くまで冷やした超伝導トラップにミリグラム級の永久磁石を磁気浮上させ（力感度0.07 fN/√Hz、1 TeV/c程度の微小な力積まで分解）、約1カ月分の静かな夜間データで、通過する超重暗黒物質粒子が与える「一撃」を探索。候補事象は見つからず、質量10⁶〜10¹⁵ GeV/c²（9桁）にわたって、軽い媒介粒子（30 meV/c²未満）を介した暗黒物質–中性子結合をα_n＝3.2×10⁻⁹まで95%信頼度で除外し、複合暗黒物質の暗黒物質–中性子断面積に最も強い制限を与えた。浮上センサーの探索質量域を光浮上より7桁重い側へ拡張（D. G. Uitenbroek・D. W. P. Amaral・J. Qin・J. Langendorff・A. Gingerich・T. H. Oosterkamp・C. D. Tunnell、arXiv:2608.20464、ライス大学9月4日発表）※査読前
Most dark-matter detectors, including the tonne-scale liquid-xenon experiment at the top of this page, are built to catch particles with masses not far from those of atomic nuclei. But some theories propose ultraheavy, composite dark matter, single particles potentially as massive as a living cell, which would pass through a large detector so rarely, and deposit their energy so differently, that the usual searches are blind to them. A team led by Christopher Tunnell (Rice University) with Tjerk Oosterkamp’s group at Leiden University takes a different approach: a permanent magnet about the size of a grain of sand, levitated above a superconductor cooled to just above absolute zero. Because the magnet touches nothing, friction is almost eliminated and its position can be tracked to about one-hundredth the width of an atom, making it a force sensor of 0.07 fN/√Hz able to resolve impulses as small as about 1 TeV/c.
The same apparatus, called POLONAISE, was used in 2025 for the first levitated-magnet search for ultralight wave-like dark matter, which would tug on the magnet rhythmically. Ultraheavy dark matter calls for the opposite strategy: instead of listening for a steady oscillation, the team waited for a sudden “knock” from a single particle passing through. They collected about a month of data, concentrated on quiet overnight periods, and treated every candidate impulse as a possible dark-matter event. None survived, which allows the team to set optimum-interval upper limits: for dark-matter masses from 106 to 1015 GeV/c2 — nine orders of magnitude — interacting with neutrons through a new light mediator (lighter than 30 meV/c2), couplings as small as αn = 3.2 × 10−9 are excluded at 95% confidence, and the result sets the leading constraints on the dark-matter–neutron cross section for composite dark matter.
The search extends levitated sensing about seven orders of magnitude in mass beyond what optically levitated particles have reached — roughly ten million times heavier than previous levitation searches. The result was released at the 2026 International Conference on Particle Physics and Cosmology and posted as a preprint (arXiv:2608.20464); it has not yet been peer reviewed. Next steps are to cool the magnet further, run for longer and levitate several magnets at once so that a genuine dark-matter kick can be distinguished from vibrations. Large underground detectors remain essential at lower masses; levitated sensors open a complementary window on the heaviest candidates.
Keywords: ultraheavy dark matter, 超重暗黒物質, composite dark matter, 複合暗黒物質, magnetic levitation, 磁気浮上, levitated sensor, 浮上センサー, superconducting trap, 超伝導トラップ
☀️ / 「なぜ他の恒星ではコロナ質量放出（CME）がほとんど見えないのか」をレーザー実験室で再現——エコール・ポリテクニーク（LULI）とライプニッツ天体物理学研究所ポツダムなどの国際チームが、高出力レーザーでテフロン標的を照射して生じるプラズマ流に、コイルで横方向の強磁場をかけ、磁気圧とプラズマ圧の比などの無次元量を恒星と揃えて実験。10テスラでは妨げられず流れたプラズマが、30テスラではキンク不安定性で曲がって分裂し、約10 mmで停止・減速した。若く活動的な恒星（磁場が太陽の10〜100倍）で観測されるCMEが弱いことと整合し、典型的なCMEエネルギーでは約100ガウス（0.01 T）程度の磁場で放出が抑制されると外挿——2018年の数値シミュレーションの予言に初の直接的な実験的証拠（S. N. Chen・J. D. Alvarado-Gómez・J. Fuchsら、Phys. Rev. Lett. 137, 105201・9月4日掲載、Physics誌Focus）
Coronal mass ejections (CMEs) — eruptions of magnetized plasma launched when the Sun’s twisted magnetic field suddenly releases its energy — drive geomagnetic storms and aurorae on Earth. Yet on other stars they have been sighted only rarely, and the few candidates are far weaker than the solar record would predict. One suspected reason is that the stars targeted by such searches are young, fast-rotating and carry surface fields 10 to 100 times stronger than the Sun’s. Simulations by Julian Alvarado-Gómez and colleagues in 2018 suggested that an overlying large-scale field of that strength could confine a CME and stop it from escaping, but there was no direct evidence.
An international team led by Alvarado-Gómez (Leibniz Institute for Astrophysics Potsdam) and plasma physicist Julien Fuchs (École Polytechnique) has now recreated the launch of a stellar CME at the LULI laser facility in France. Laser pulses fired at a flat Teflon target generate a hot plasma stream flowing away from the surface; a pair of coils imposes a strong magnetic field transverse to the flow, compressing it into a flat outward-moving sheet. Key dimensionless quantities, such as the ratio of magnetic to plasma pressure, were tuned to match astrophysical values so that the laboratory system evolves like a stellar eruption. At 10 tesla the plasma flowed unimpeded over the duration of the experiment; at 30 tesla a clear transition appeared: the flow bent and split, slowed markedly and halted altogether after travelling about 10 mm. A numerical model reproducing the shots traced the behaviour to kink-like instabilities that break the flow into streamlets and drain its energy.
The observations agree with the astrophysical simulations and, extrapolated to stellar scales, imply that a background field of roughly 100 gauss (0.01 T) — ordinary for a young star — would suppress a CME of typical energy, naturally explaining the weak, slow stellar events seen so far. Eric Blackman (Rochester) notes that real stars are more complicated, with CME strength correlated to field strength, a case for future work. The same platform can now be used to test other predictions, such as the high-energy radiation a confined CME should produce in the stellar corona. Published as S. N. Chen et al., Physical Review Letters 137, 105201 (4 September 2026), highlighted in a Physics Focus story.
Keywords: coronal mass ejection, コロナ質量放出, CME, stellar magnetic field, 恒星磁場, laboratory astrophysics, 実験室天体物理学, laser plasma, レーザープラズマ, LULI
🧠 / 原子と光子でつくった「スピングラス」が、AIの連想記憶を超える——スタンフォード大学のBenjamin Levらが、多モード光共振器内の超低温原子集団を光子で全結合させた量子光学スピングラスを、不完全な入力から記憶全体を呼び出す連想記憶として動作させることに成功。従来のホップフィールド・ネットワークでは記憶を詰め込みすぎるとフラストレーションで指数関数的に多くの「偽の記憶」が生じて呼び出しが破綻するが、駆動散逸的な量子光学ダイナミクスの下ではその偽の記憶が信頼できる記憶として使える。16スピンのネットワークでヘブ学習によるホップフィールドの容量を最大7倍上回り、さらに原子の運動が結合を動的に変える「短期的シナプス可塑性」に似た効果で容量が増大——量子光学系での学習の前段階を実現（B. P. Marsh・D. Atri Schuller・Y. Ji・H. S. Hunt・S. Ganguli・S. Gopalakrishnan・J. Keeling・B. L. Lev、Science・9月3日掲載）
The Hopfield network, recognised by the 2024 Nobel Prize in Physics, stores memories in the couplings of an all-to-all-connected set of spins and recalls them by relaxing toward an energy minimum — the way a face is recognised from a blurred photograph. Its weakness is capacity: store too many patterns and frustration among the couplings creates an exponential number of spurious minima, the network becomes a spin glass (Parisi’s 2021 Nobel topic), and recall collapses. A standard 16-neuron Hopfield network reliably holds only two or three memories.
Benjamin Lev’s group at Stanford builds spin glasses from ultracold atomic gases trapped between the mirrors of a multimode optical cavity, where photons bouncing inside the cavity mediate programmable all-to-all interactions between the atomic ensembles; in 2025 the group used this platform to observe replica symmetry breaking directly (Science 389, 1122). Now, with Surya Ganguli (Stanford), Sarang Gopalakrishnan (Princeton) and Jonathan Keeling (St Andrews), they show that the same system works as an associative memory whose capacity exceeds the Hopfield limit. Because the atom–photon network is driven and dissipative, its dynamics are not those of an equilibrium Hopfield model: the glassy minima that would be useless “false memories” under equilibrium recall become reliable memories. In a 16-spin network the storage capacity surpasses that of a Hebbian-trained Hopfield network by up to seven-fold, with one configuration holding 25 memories. A second discovery is that the motion of the atomic ensembles in the cavity light dynamically reshapes the connectivity, a polaronic “elasticity” that mimics short-term synaptic plasticity in biological neural networks and roughly doubles the capacity.
The experiment is lab-scale (vacuum chambers, laser traps, networks of up to 20 spins), but it demonstrates a precursor to learning in a quantum-optical system, in which the physics itself performs the memory computation. The authors see routes to larger networks, to exploiting entanglement, and potentially to lower-energy AI hardware; the platform is equally a probe of how physical, and quantum, systems compute. Published in Science on 3 September 2026 (DOI: 10.1126/science.aec3917); preprint arXiv:2509.12202.
Classical computers separate logic from storage: a processor works on a few registers while dynamic random-access memory holds everything else. Superconducting quantum processors have no such thing — every qubit is simultaneously a logic and a storage element, each with its own control lines. Yet three-dimensional superconducting microwave cavities can hold quantum states for far longer than transmon qubits, which makes them natural candidates for a quantum memory that could be optimised separately from the logic and shared among many logical qubits. The concept of quantum random-access memory was formalised by Giovannetti, Lloyd and Maccone in 2008, but a controllable, multimode realisation in a superconducting architecture had remained out of reach.
David Schuster’s group (Stanford/SLAC), with collaborators at the University of Chicago, Fermilab, NYU and Rutgers, now reports an eight-bit cascaded random-access quantum memory. A single transmon classically addressesseven storage modes of a multimode cavity, selecting which mode to write to or read from, while a buffer layer inserted between the processor and the storage cavity isolates the memory modes from the processor’s nonlinearities — the cascade that gives the device its name. Randomized benchmarking shows arbitrary random access with an average infidelity below 1.5% per mode, and the team characterises the many-body interactions among modes that dominate the remaining error budget, giving a clear target for the next generation.
The payoff is architectural. Because one transmon can serve many memory modes, the number of control lines per logical qubit drops substantially, and because the memory holds several modes side by side, transversal operations can be performed within the memory module itself. The authors present the device as a scalable unit cell for fault-tolerant superconducting architectures in which logic and storage are engineered independently. Published in Nature Physics on 8 September 2026.
David Schusterのグループ（スタンフォード大学／SLAC）は、シカゴ大学、フェルミ研究所、ニューヨーク大学、ラトガース大学の共同研究者とともに、8ビットのカスケード型ランダムアクセス量子メモリを報告した。1個のトランズモンが多モード共振器の7つの記憶モードを古典的にアドレス指定し、どのモードに書き込み、どのモードから読み出すかを選ぶ。プロセッサと蓄積共振器の間に挟んだバッファ層が、記憶モードをプロセッサの非線形性から隔離する——この「カスケード」構造が装置の名の由来だ。ランダム化ベンチマークにより、任意のランダムアクセスをモードあたり平均1.5%未満の不忠実度で実証し、残る誤りの主因であるモード間の多体相互作用を特定して、次世代に向けた明確な目標を示した。
Quantum error correction spreads a logical qubit over many physical qubits and uses the outcomes of repeated parity checks (the “syndrome”) to infer, on a classical computer, which errors occurred and how to undo them. Larger codes suppress logical errors but cost more qubits, so a complementary strategy has attracted attention: post-selection, discarding only those runs whose correction is judged unreliable. The catch is that simple, generic selection criteria throw away many perfectly good results, while precise criteria demand prohibitive classical computation.
Nobuyuki Yoshioka (ICEPP, University of Tokyo) and Kento Tsubouchi (UTokyo School of Engineering), with Haipeng Xie and Ying Li of the Graduate School of China Academy of Engineering Physics, propose Argument Reweighting (AR): a decoding pass is run as usual to obtain a candidate set of error locations; the decoder’s weights are then perturbed so that this particular candidate becomes less likely, and the same syndrome is decoded again. If the answer survives the “shake”, it is accepted; if it changes, the run is flagged as low-confidence and discarded. The method needs no extra qubits, no circuit changes and only one or two additional classical decoding passes, and it applies to essentially any code and decoder. Tested across many standard quantum LDPC codes, it performs best on large codes: for the gross code, a member of the bivariate-bicycle (BB) family favoured for superconducting and neutral-atom hardware, rejecting only about one run in 70,000 lowers the logical error rate by more than a factor of 60.
Because post-selecting high-quality quantum resources is also central to magic-state preparation for fault-tolerant computing, Bell-state distillation for quantum networks and key distillation in quantum cryptography, the authors expect the principle to carry over to those tasks. Published online in Physical Review Letters on 8 September 2026 as “Simple, Efficient, and Generic Post-Selection Decoding for qLDPC Codes”; supported in part by IBM Quantum and Google Quantum AI.
Can a remote party prove that it really is where it claims to be? Classical “distance-bounding” schemes time a challenge–response exchange against the speed of light, but they are provably insecure against an untrusted prover: classical information can be copied and relayed by colluding adversaries who produce the right answer without any extra delay. Encoding part of the challenge in a qubit, which cannot be copied, was proposed as the way out, and theory shows that attacking such quantum position verification (QPV) requires auxiliary quantum resources — entanglement or quantum gates — that grow with the size of the classical input. But an experiment faces brutal constraints: excess latency of a millisecond translates into a position error of hundreds of kilometres, the prover must evaluate a randomly chosen Boolean function on tens of bits almost instantly, and losses must stay within a few decibels.
A team at the University of Science and Technology of China led by Guang-Can Guo, Zheng-Fu Han, Shuang Wang and Zhen-Qiang Yin now realises a complete QPV protocol. Instead of single photons they use phase-randomized weak coherent states, for which they derive a finite-size security bound that treats the vacuum, single-photon and multiphoton components separately (assuming the adversary attacks multiphoton pulses perfectly). Two verifiers, 2 km apart with the prover roughly midway, each send 20 classical bits over anti-resonant hollow-core fibre — light travelling at nearly its vacuum speed — while one verifier sends a polarization qubit through ultralow-loss fibre. The prover evaluates a uniformly random 40-bit Boolean function with an FPGA and a 1-terabit DDR lookup table in about 117 ns, measures the qubit in the basis this selects and returns the result. A Sagnac polarization encoder gives a 0.27% error rate, a high-voltage-driven optical switch and superconducting detectors give 70% overall efficiency, and the system runs at 2 MHz.
In five trials of 107 rounds each — a few seconds per verification — the prover’s score exceeded the security threshold, with a total failure probability below 5 × 10−10. The prover’s maximum excess response delay of 247.8 ns bounds its position to within 74.3 m, comparable to a building or secure zone. To cheat, an adversary would need roughly five bits of auxiliary quantum resources per round sustained at the 2-MHz rate — about 10 Mbit/s of entanglement, more than two orders of magnitude beyond current capabilities. The authors envision position-based authentication for transactions, access control and the tracking of high-value assets. Published open-access in Nature Physics on 3 September 2026.
Bell tests, from Freedman and Clauser in 1972 to the loophole-free superconducting-circuit experiment of 2023, certify that nature is not classical by showing that two separated players win a cooperative game more often than any classical strategy allows. Over the past decade a second family of tests has emerged in which a quantum computer performs a computational task, such as random-circuit sampling, that classical computers cannot practically reproduce. Those demonstrations have three known limitations: their claims rest on unproven complexity-theoretic assumptions, they are fragile against hardware noise, and their outputs are inefficient to verify, which raises doubts about how they scale.
Marcello Benedetti, Gabriel Marin-Sanchez, Jordi Weggemans and Matthias Rosenkranz of Quantinuum, with Harry Buhrman (Quantinuum, University of Amsterdam and QuSoft), propose a test that avoids all three. It is a game based on complement sampling — an efficiently verifiable problem known to achieve the largest possible separation between quantum and classical computation when both the input and the output are samples from probability distributions. When the inputs are restricted to instances inspired by the Bernstein–Vazirani problem, the game admits an exponentially large violation of classicality without relying on any computational-hardness assumption, in the same unconditional spirit as a Bell inequality but for a computation. The team ran the game on Quantinuum’s System Model H2 trapped-ion computers, executing thousands of different circuits on up to 55 qubits; the observed scores can be explained only by a systematic adoption of a quantum strategy.
The result is thus an efficient and scalable certification of the quantum nature of the hardware itself, complementary to sampling-based advantage claims: as devices grow, the same game can be re-run and re-verified. Published open-access in Nature Communications on 5 September 2026; preprint arXiv:2511.11008.
Keywords: violation of classicality, 古典性の破れ, non-local game, 非局所ゲーム, Bell inequality, ベル不等式, complement sampling, 補集合サンプリング, Bernstein-Vazirani, quantum advantage
🤖 / 核融合プラズマを20ミリ秒ごとにAIが操る——プリンストン・プラズマ物理研究所（PPPL）とプリンストン大学のEgemen Kolemenらが、複数の機械学習モデルをトカマクの制御系に直接組み込むモジュール型AI制御フレームワーク「PACMAN」（Prediction And Control using MAchiNe learning）を開発し、DIII-Dトカマクの5つの実験で実証。計測値の収集と検証、AIモデルによる予測、コントローラによる指令計算、安全制限の適用という4段階を約20ミリ秒で一巡し、これを繰り返す。強化学習モデルによる加熱系の完全制御、プラズマ端のエネルギーバースト予測、高速粒子が駆動する波の検出と制御、密度・回転の目標追従に加え、ティアリングモード不安定性を約200ミリ秒前に予測して発生そのものを回避。6基のジャイロトロン（マイクロ波加熱）のミラー角度と出力も同時に最適化した。安全限界は人間が設定し、AIの提案に関わらず強制される（H. Farre Kaga・A. Rothsteinら、量子科学技術研究開発機構〔QST〕のTakuma Wakatsukiも参加、Nuclear Fusion掲載・PPPL 9月2日発表）
In a tokamak, a plasma hotter than the Sun’s core is held by magnetic fields, and keeping it hot, dense and stable requires constant adjustment of heating systems, magnets and gas injectors. Instabilities can grow in a few milliseconds — far faster than any human operator, who responds on the order of seconds — and the first-principles codes that simulate a plasma take days or months to run. Machine-learning models are, in practice, the only way to model a plasma on millisecond timescales, but most attempts to use them for control were built from scratch, without a common architecture that lets several models work together.
Researchers at the DOE’s Princeton Plasma Physics Laboratory and Princeton University led by Egemen Kolemen, with co-lead authors Hiro Farre Kaga and Andy Rothstein, have built such an architecture. PACMAN (Prediction And Control using MAchiNe learning) works like an assembly line with four stations: it gathers the tokamak’s real-time measurements (temperatures, densities, magnetic signals), checks and packages them, lets AI models read whatever they need and predict what the plasma is doing or about to do, has controllers turn those predictions into commands, and finally resolves conflicts between controllers and enforces strict hardware safety limits before sending commands to the machine. The whole loop typically runs in about 20 milliseconds and repeats continuously. In five experiments on the DIII-D National Fusion Facility tokamak in San Diego, PACMAN let a reinforcement-learning model take complete control of the heating systems, predicted sudden energy bursts from the plasma edge, detected and controlled waves driven by fast particles, steered density and rotation to preset targets, and — most strikingly — predicted a tearing-mode instability about 200 milliseconds in advance and changed the plasma to avoid it altogether, where conventional controllers can only react once it has begun. It also steered all six of DIII-D’s gyrotrons at once, retargeting their mirrors and adjusting their power in real time.
Because each model and controller is an independent module, installing the second model took a couple of days instead of months, allowing week-by-week iteration on a research machine. The framework does not remove people from the loop: humans set the goals and the safety limits, which apply regardless of what an AI recommends, and physicists review each shot. The authors note it is not suited to sub-millisecond events such as vertical displacements, and its modularity is meant to carry over to tokamaks of other sizes and shapes, including machines not yet designed. Co-authors include Takuma Wakatsuki of Japan’s National Institutes for Quantum Science and Technology (QST). Published in Nuclear Fusion; PPPL press release 2 September 2026.
米エネルギー省プリンストン・プラズマ物理研究所（PPPL）とプリンストン大学のEgemen Kolemenが率いるチーム（共同筆頭著者はHiro Farre KagaとAndy Rothstein）は、そのアーキテクチャを作り上げた。PACMAN（Prediction And Control using MAchiNe learning）は4つの工程を持つ組立ラインのように働く。トカマクのリアルタイム計測値（温度、密度、磁気信号）を集めて検証・パッケージ化し、AIモデルが必要なものを読み取ってプラズマの現状と直近の挙動を予測し、コントローラがその予測を指令に変換し、最後にコントローラ間の衝突を解消してハードウェアの厳格な安全制限を適用したうえで装置に送る。このループ全体が通常約20ミリ秒で回り、繰り返し実行される。サンディエゴのDIII-D国立核融合施設トカマクでの5つの実験で、PACMANは強化学習モデルに加熱系の完全な制御を委ね、プラズマ端からの突発的なエネルギーバーストを予測し、高速粒子が駆動する波を検出・制御し、密度と回転を事前に設定した目標に追従させた。そして最も際立った成果として、ティアリングモード不安定性を約200ミリ秒前に予測し、プラズマを変化させて発生そのものを回避した——従来の制御器は発生後にしか反応できない。DIII-Dの6基のジャイロトロンも同時に操り、ミラーの向きと出力をリアルタイムで調整した。
For most of a century, physicists have tried to identify dark matter, the invisible substance thought to make up roughly 85% of the matter in the Universe. The LUX-ZEPLIN (LZ) experiment — a two-phase time projection chamber holding 10 tonnes of ultrapure liquid xenon (seven active tonnes), operating nearly a mile (1.6 km) underground at the Sanford Underground Research Facility in Lead, South Dakota — is the world’s most sensitive detector optimised for WIMPs (weakly interacting massive particles). It looks for the faint flashes of light and ionisation charge produced when a passing particle kicks a xenon nucleus. The collaboration of about 250 scientists and engineers from 39 institutions is managed by Lawrence Berkeley National Laboratory.
On 1 September 2026, at the TeV Particle Astrophysics (TeVPA) conference in Tendo, Japan, LZ reported a search over an exposure of 2.84 tonne-years in which the nuclear-recoil energy window was deliberately extended up to about 270 keV. That extension targets effective-field-theory and inelastic dark-matter models in which high-energy recoils make up a larger share of the expected spectrum than in the standard spin-independent case. The team observed one event consistent with a nuclear recoil of 248 ± 23 (stat) ± 23 (sys) keV, in a region where the known background expectation is low. A profile-likelihood-ratio test finds tension with the background-only hypothesis at a global significance of 2.6σ after accounting for look-elsewhere effects, with a maximum local significance of 3.4σ across the models tested — roughly a 0.5% chance that the event comes from a known background.
This is emphatically not a discovery claim: the conventional threshold is 5σ, and a single unexplained event can either fade statistically as data accumulate or become the first member of a pattern. But the interpretation is intriguing. If the event was caused by a WIMP, that particle would have a mass of at least 200 GeV/c2 — more than 200 times the proton mass — and would point to a type of WIMP–nucleon interaction beyond the simplest model, which also sets a target for collider and indirect searches. LZ already holds the world’s largest dark-matter dataset and continues to accumulate WIMP-search data toward a planned 1,000 live days, which should settle whether the anomaly grows or disappears. The paper was posted on the collaboration’s website, is being released on arXiv and will be submitted to Physical Review Letters.
NASA’s Nancy Grace Roman Space Telescope lifted off at 7:26 a.m. EDT on Sunday, 30 August 2026, aboard a SpaceX Falcon Heavy from Launch Complex 39A at Kennedy Space Center, Florida. About 70 minutes after launch the Deep Space Network took over communications, and one hour 23 minutes after liftoff the team confirmed deployment of the solar panels and the lower instrument sun shade. Roman is now on a roughly three-month, million-mile (1.5 million km) journey to a halo orbit around the Sun–Earth L2 point, the same neighbourhood as the James Webb Space Telescope; instrument checkouts begin en route, and the launch came about nine months ahead of the May 2027 readiness date NASA had committed to after its pandemic-era replan.
Roman pairs a Hubble-class 2.4-m mirror with a Wide Field Instrument whose field of view is at least 100 times larger than Hubble’s, plus a technology-demonstration Coronagraph Instrument to which the space agencies of Japan (JAXA), Europe and France and Germany’s Max Planck Institute all contributed. Its core surveys are designed to measure the growth of cosmic structure and the expansion history of the Universe — and hence the behaviour of dark energy — through weak gravitational lensing, baryon acoustic oscillations and Type Ia supernovae, to map dark matter, and to find thousands of exoplanets by gravitational microlensing toward the Galactic bulge. Over its lifetime it should record light from about a billion galaxies and return on the order of a terabyte of data per day.
The observatory is named for Nancy Grace Roman, NASA’s first chief of astronomy, often called the “mother of Hubble” for championing that telescope. For this site’s recurring themes, Roman is a direct successor to the supernova and BAO measurements at the heart of the still-unresolved cosmic-acceleration debate covered elsewhere on this page: its supernova and lensing surveys are exactly the kind of large, homogeneous datasets that both sides of that dispute say will be needed to settle it. Science operations are expected to begin in early 2027 once commissioning is complete.
Keywords: Nancy Grace Roman Space Telescope, ローマン宇宙望遠鏡, NASA, Falcon Heavy, ファルコンヘビー, Sun-Earth L2, ラグランジュ点L2, dark energy, 暗黒エネルギー, dark matter
⚛️ / 探していた粒子は現れず、代わりに「2つの新しい構造」が見つかった——ジェファーソン研究所のGlueX実験が、高強度の偏光光子ビームを陽子標的に当てる光生成過程 γp→φπ⁺π⁻p で、2006年にBaBarが報告したエキゾチック候補Y(2175)を探索。Y(2175)は観測されず生成断面積に上限を設定した一方、約2.24 GeVにY(2240)（5σ超）、約1.82 GeVにX(1830)（3σ）という2つの構造を検出。ストレンジクォーク対とグルーオン励起を含むハイブリッド中間子やテトラクォークなど「XYZ状態」の正体に新データ（GlueX国際共同実験、Phys. Rev. Lett. 136巻25号、9月2日報道）
Since the early 2000s, experiments have turned up a growing zoo of hadrons whose quantum numbers and decays do not fit the original quark model of quark–antiquark mesons and three-quark baryons. Physicists lump them together as XYZ states; candidate explanations include hybrid mesons in which excited gluon fields contribute directly to the structure, four-quark tetraquarks, and molecule-like bound states of ordinary hadrons. In 2006 the BaBar experiment at SLAC reported one such candidate in the strangeonium (strange–antistrange) sector, Y(2175) at about 2.16 GeV, later confirmed by BES in China and Belle in Japan — but only ever in electron–positron annihilation.
The GlueX experiment in Hall D of Jefferson Lab was built precisely to hunt hybrid mesons. A thin diamond radiator converts the electron beam of the Continuous Electron Beam Accelerator Facility (CEBAF) into a polarised photon beam of unmatched intensity at these energies; millions of photons per second strike a liquid-hydrogen target and a large-acceptance spectrometer records the debris. Searching the reaction γp → φπ+π−p for Y(2175) produced by photoproduction, the collaboration found no signal at the expected mass and set an upper limit on the photoproduction cross section. Instead, two other structures appeared nearby: Y(2240) at roughly 2.24 GeV with a significance above 5σ, and X(1830) at roughly 1.82 GeV at about 3σ.
“We went searching for a confirmed XYZ candidate with a photon beam but instead found two other structures,” said Jefferson Lab staff scientist Malte Albrecht. Because these are the first XYZ-like signals seen in photoproduction, the production mechanism itself becomes a new discriminator: theorists can now ask which exotic quark–gluon configurations would appear in photon–proton collisions but not in e+e− annihilation, and which measurements would distinguish a hybrid from a tetraquark or a hadronic molecule. GlueX spokesperson Justin Stevens (William & Mary) says the result “opens the door for a whole new set of hadron spectroscopy measurements”, with far more data still to be analysed. Published in Physical Review Letters 136 (issue 25); announced by Jefferson Lab on 2 September 2026.
General relativity sets no lower limit on the mass of a black hole. In 1993, numerical simulations (Matthew Choptuik’s) of a collapsing scalar field revealed that black-hole formation has a sharp threshold with the hallmarks of a phase transition: tune the initial energy just below a critical value and the field disperses, leaving flat spacetime; tune it just above and a black hole forms, with a mass that scales as a universal power law. Exactly at the threshold sits a strange, unstable discretely self-similar solution in which the curvature of spacetime repeats itself in a regular pattern across space and time — a “spacetime crystal”, in the words of Daniel Grumiller of TU Wien, who compares it to water at 0 °C that will freeze into a regular ice lattice at the slightest nudge. Such conditions may have existed in the chaotic early Universe, making critical collapse relevant to primordial black holes. Yet for three decades no one could write down that solution analytically.
Christian Ecker (Goethe University Frankfurt), Florian Ecker and Daniel Grumiller (TU Wien) have now done so by a counter-intuitive detour: instead of working in four spacetime dimensions, they write the Einstein–Klein–Gordon equations in D dimensions and take the limit D → ∞. Remarkably, the equations become tractable there, and the team derives an exact, pen-and-paper formula for the discretely self-similar critical solution. The result can then be systematically improved with further approximation methods to move back toward lower dimensions, including our own four-dimensional spacetime. “Our technique turns out to be remarkably stable,” says Florian Ecker — a new method for black-hole phenomena that previously could only be studied numerically.
The work does not claim that our Universe has infinitely many dimensions; the large-D expansion is a mathematical tool, in the same spirit as large-N methods in gauge theory. What it offers is analytic control over one of the most delicate phenomena in classical gravity, and hence a way to probe how microscopic black holes form without relying entirely on supercomputer simulations. Published in Physical Review Letters 136 (issue 19); TU Wien press release, reported 30 August 2026.
Christian Ecker（フランクフルト・ゲーテ大学）、Florian EckerとDaniel Grumiller（ウィーン工科大学）は、直感に反する回り道でこれを成し遂げた。4次元時空で計算する代わりに、アインシュタイン–クライン・ゴルドン方程式をD次元で書き、極限D→∞を取ったのである。驚くべきことに、そこでは方程式が扱いやすくなり、チームは離散的自己相似な臨界解の厳密な公式を「紙と鉛筆」で導出した。得られた結果は、追加の近似法で系統的に改善して、私たちの4次元時空を含む低次元へと戻していくことができる。「この手法は驚くほど安定であることがわかった」とFlorian Eckerは言う。これまで数値計算でしか調べられなかったブラックホール現象に対する、新しい解析手法である。
Keywords: critical collapse, 臨界崩壊, Choptuik, チョプチュイク, discrete self-similarity, 離散的自己相似性, spacetime crystal, 時空の結晶, large D expansion, large-D展開
🍎 / ニュートンの逆二乗則は、数億光年離れた銀河団のあいだでも成り立っていた——アタカマ宇宙論望遠鏡（ACT）が観測した宇宙マイクロ波背景放射の「運動学的スニヤエフ・ゼルドビッチ効果」を用い、数十万個の銀河団の相対運動から宇宙論的スケールで重力の距離依存性を測定。史上最大スケールの重力の法則の検証で、結果はニュートン／アインシュタインの予言とほぼ一致。MONDなど「重力を修正して暗黒物質をなくす」モデルが予言する緩やかな減衰は見られず、暗黒物質の存在を補強する（P. A. Gallardo〔ペンシルベニア大学〕ら40名超、Phys. Rev. Lett. 136巻15号、ペンシルベニア大発表・8月24日報道）
Stars in the outskirts of galaxies, and galaxies inside clusters, move far too fast for the visible matter to hold them together. There are two ways out of this “massive discrepancy in the cosmic ledger”, as University of Pennsylvania cosmologist Patricio Gallardo puts it: either the Universe contains large amounts of unseen dark matter, or the law of gravity itself must be modified on large scales, as in MOND (Modified Newtonian Dynamics) and its relatives. The two options make different predictions for how the strength of gravity falls off with distance — and that is something one can measure, provided one can watch enormous masses pulling on each other across enormous distances.
Gallardo and more than 40 collaborators did exactly that with the Atacama Cosmology Telescope (ACT), an instrument developed largely by Penn researchers led by Mark Devlin. Their probe is the kinematic Sunyaev–Zeldovich (kSZ) effect: when the cosmic microwave background — light released about 380,000 years after the Big Bang — passes through the hot gas of a moving galaxy cluster, the cluster’s line-of-sight velocity imprints a tiny Doppler shift on it. Stacking these shifts over hundreds of thousands of galaxy clusters separated by tens to hundreds of millions of light-years gives the clusters’ pairwise infall velocities, which are set by how strongly gravity acts between them as a function of separation. This is the largest-scale test of the gravitational force law performed to date.
The result: the strength of gravity decreases with distance almost exactly as Newton’s inverse-square law — incorporated into Einstein’s general relativity — predicts. MOND-like models would have required a gentler decline, which the data do not show. “It is remarkable that the law of the inverse of the squares, proposed by Newton in the 17th century, is still holding its ground in the 21st century,” Gallardo says. Because modifying gravity does not account for the missing mass in these observations, the measurement strengthens the case for dark matter and the standard cosmological model, while saying nothing about what dark matter is made of. Future CMB measurements combined with larger galaxy surveys should sharpen the test further. Published in Physical Review Letters 136 (issue 15); announced by the University of Pennsylvania, reported 24 August 2026.
Hydrogen-rich superhydrides can superconduct near room temperature, but only under pressures above about 150 GPa. The class of metal clathrate superhydrides — metal atoms caged inside hydrogen frameworks — came to prominence in 2019, when LaH10 was shown to superconduct at a then-record 250–260 K. Yanming Ma’s group at Jilin University subsequently predicted that adding scandium should push the transition higher, and, according to the accompanying Physics synopsis, the group’s synthesised LaSc2H24 showed a critical temperature of 298 K at 260 GPa. What the original prediction could not explain was why scandium helps.
Using a more sophisticated model, Zefang Wang, Wenbo Zhao, Yuan Ma, Hanyu Liu and Yanming Ma now trace the difference to the electronic structure. In LaH10, an anisotropy in the band structure opens two distinct superconducting channels with two different gaps and two different transition temperatures, and this two-gap structure caps the overall Tc. In LaSc2H24 the scandium 3d orbitals act as bridges connecting the hydrogen 1s orbitals, providing a continuous pathway for electron transport that makes the band structure more isotropic and unifies the two gaps into one. The scandium atoms also distort and soften the H–H bonds, strengthening the lattice vibrations (phonons) that mediate the pairing and raising Tc further.
The lesson is a design rule rather than a single material: isotropising the electronic structure of a clathrate superhydride is a route to higher transition temperatures. The team’s next target is the related compound La2ScH36, which they predict could go higher still. The catch remains the megabar pressure, which keeps these materials in the realm of diamond-anvil-cell physics rather than technology — as with all superhydride results, independent replication of the experimental Tc is essential. Published in Physical Review B 114, 154501 on 1 September 2026, highlighted as a synopsis in Physics.
得られた教訓は一つの物質というより設計指針である。クラスレート型超水素化物の電子構造を等方化することが、より高い転移温度への道になる。チームの次の標的は、さらに高いTcが予測される関連化合物La₂ScH₃₆だ。難点は依然としてメガバール級の圧力で、これらの物質は技術というよりダイヤモンドアンビルセル物理の領域にとどまる——そして超水素化物の常として、実験的Tcの独立な再現が不可欠である。Physical Review B 114, 154501に2026年9月1日掲載、Physics誌のSynopsisで紹介。
Bilayer nickelates — oxides in which nickel–oxygen double perovskite layers alternate with rare-earth rock-salt layers — became the newest family of high-temperature superconductors in 2023, when bulk crystals were reported to superconduct at up to about 80 K under high pressure. In 2024, thin films whose in-plane lattice was compressed by epitaxial strain from the substrate were found to superconduct above 40 K at ambient pressure. Two routes to the same physics, apparently — but nobody had compared them in a single sample, so how the high-pressure and ambient-pressure superconducting states are related remained unknown.
A team from the University of Tokyo, RIKEN’s Center for Emergent Matter Science and Tohoku University — Motoki Osada, Atsushi Tsukazaki, Yoshinori Tokura, Ryotaro Arita and colleagues — grew strained La2LnNi2O7 films (Ln = rare earth) by pulsed-laser deposition followed by ozone treatment, mapped how the transition temperature varies systematically with the rare-earth element and lattice, characterised the ambient-pressure state in pulsed magnetic fields up to 59 tesla, and then squeezed the films in a cubic-anvil press up to 20 GPa. The ambient-pressure Tc of about 40 K rose steadily with pressure, reaching 67 K in La2SmNi2O7 and up to 73 K in La2NdNi2O7 at 16 GPa — approaching the regime reported for bulk crystals. Under pressure the resistivity also showed a two-step drop, a characteristic new transition observed here for the first time.
By comparing the pressure dependence with the rare-earth dependence at ambient pressure, the group finds that what matters for a high Tc is not simply the size of the unit cell but the ratio of in-plane to out-of-plane lattice constants and the electronic structure that follows from it — a knob that both epitaxial strain and hydrostatic pressure turn, in different ways. This is the first experimental demonstration connecting the two superconducting regimes in the same material system, and it points toward designing higher-Tc nickelates by controlling crystal structure and electronic states. Published online in Nature Materials on 18 August 2026; press release 20 August.
Quantum networks promise communication that cannot be eavesdropped on, but quantum information is fragile and, because an unknown quantum state cannot be copied, anything lost in transit is lost for good. Quantum teleportation — transferring an unknown state using entanglement shared in advance plus a classical message — is the workaround, but it is usually done one state at a time, which makes transmitting anything as complex as an image impractically slow. Earlier attempts to send states in parallel relied on multiplexing several channels into the properties of a single beam, at the cost of complicated decoding at the receiver and little ability for the sender to manipulate channels individually.
Jietai Jing’s group at East China Normal University now demonstrates a scheme with far greater bandwidth. A spatial light modulator patterns the transverse profile of a laser beam into a 10 × 10 grid of 100 distinct spatial modes, each of which can be controlled independently to form a pixelated image. To teleport the image, the sender mixes the information-carrying light field with one half of a pair of entangled light fields shared beforehand with the receiver — the interaction scrambles the transmitted information — and when the receiver combines the scrambled field with the other half of the entangled resource, the full 100-pixel image is reconstructed all at once. Because each channel is an independent spatial mode, the layout is reconfigurable and the sender can address any subset of pixels.
The demonstration is a multimode implementation rather than a record for fidelity or distance, but it addresses the bandwidth bottleneck directly: the number of quantum states teleported in one shot scales with the number of spatial modes, which spatial light modulators can push well beyond 100. That makes it a natural building block for image-level quantum communication and for scaling up quantum networks. Published in Physical Review Letters 137, 080801 on 20 August 2026 and highlighted in a Physics synopsis.
A typical pure state of a quantum many-body system is nothing like a thermal state as a whole, yet if you can only observe and manipulate its particles one at a time it is almost indistinguishable from thermal equilibrium — a fact at the heart of how isolated quantum systems thermalise. Thermodynamics offers a sharp way to phrase this: no work can be extracted from a thermal state, and with local operations alone almost none can be extracted from a typical quantum state either, whereas an agent who can manipulate the whole system freely can exploit its correlations to extract work proportional to the system size. What happens in between — when each particle can be operated on and measured individually and the outcomes shared by classical communication, the setting known as LOCC — was not understood.
Toshihiro Yada, Nobuyuki Yoshioka and Takahiro Sagawa (University of Tokyo) formalise the thermodynamic work extractable under LOCC, a task made hard by the adaptive protocols LOCC allows, in which later operations depend on earlier measurement results. Their key result is an inequality bounding the extractable work from above by the geometric entanglement of the state, a measure of how far it lies from any unentangled state. From this it follows that states with very strong multipartite entanglement — Haar-random states, for instance — yield no extensive work even under LOCC and are therefore thermodynamically equivalent to thermal states in this operational sense. Conversely, states that look thermal locally but have comparatively simple entanglement structure do allow work proportional to system size once classical communication is added.
The upshot is a classification of quantum many-body states by how “thermal” they are under realistic, intermediate levels of control, keyed to the structure of their entanglement rather than to their local appearance. Beyond the foundations of quantum statistical mechanics, the framework quantifies how much of the information and correlations in a quantum state can actually be cashed out as work with a given set of operations — a design principle for quantum heat engines that exploit precise control of many-body systems. Published online in Physical Review Letters on 21 August 2026; press release 24 August.
Keywords: quantum thermodynamics, 量子熱力学, extractable work, 取り出せる仕事, LOCC, local operations and classical communication, 局所操作と古典通信, thermal state, 熱平衡状態, thermalization
🎭 / 「単一ギャップ」に見えた超伝導は、実は2つの超伝導秩序が強く結合した"デュエット"だった——ヘブライ大学のチームが、数原子層のNbSe₂とTaS₂に対する高感度トンネル分光で、単一の超伝導秩序では再現できなかったエネルギースペクトルの形状と磁場応答を、強く相互作用する2つの超伝導秩序（2バンド超伝導）を含むモデルで説明。バルクNbSe₂では3つの秩序が絡み合っている可能性も示唆し、量子デバイス材料としての2次元超伝導体の設計に指針を与える（S. Simon・M. Klang・O. Millo・H. Steinberg、Phys. Rev. Lett. 137巻1号、2026年8月31日報道）
Niobium diselenide (NbSe2) is one of the most thoroughly studied two-dimensional superconductors, and when thinned to a few atomic layers it appeared to behave simply: a single superconducting energy gap, the quantity that encodes how electrons pair to flow without resistance. Yet the detailed shape of its superconducting spectrum had never been fully reproduced by single-gap theory — a small discrepancy that hinted the simple picture was incomplete.
Using highly sensitive tunnelling spectroscopy on few-layer NbSe2 and the closely related TaS2, Shahar Simon, Maya Klang, Oded Millo and Hadar Steinberg of the Hebrew University of Jerusalem show that each material in fact hosts two distinct superconducting orders that interact so strongly they masquerade as one — “like listening to what sounds like a single singer, only to discover it’s a perfectly synchronised duet.” A two-band model reproduces the measured spectra far more accurately than the conventional one and also accounts for the materials’ response to applied magnetic fields. The analysis further suggests that bulk NbSe2 may contain three interacting superconducting orders.
The finding resolves a long-standing puzzle and matters beyond bookkeeping: transition-metal dichalcogenide superconductors are candidates for Josephson devices, hybrid qubits and topological platforms, and knowing that their pairing lives on multiple coupled bands changes how one should engineer interfaces, proximity effects and field responses. Published in Physical Review Letters 137 (issue 1); Hebrew University press release, reported 31 August 2026.
The wavefunction is the central object of quantum mechanics, yet it cannot be observed directly. Inside molecules, the electron wavefunctions — the molecular orbitals — determine how a molecule absorbs light and reacts, so a complete three-dimensional picture of one would be enormously informative. Photoemission orbital tomography gets part of the way: the momentum distribution of electrons photo-emitted from oriented molecules on a surface is essentially the squared magnitude of the orbital’s Fourier transform, which yields one half of the information (the amplitude) without disturbing the state; the missing phase must then be recovered by a reconstruction algorithm. Extending this to full 3D has so far required long measurement campaigns at synchrotron facilities, which has made it nearly impossible to image changing wavefunctions as movies.
An interdisciplinary team at the University of Göttingen led by Stefan Mathias and G. S. Matthijs Jansen, with first author Wiebke Bennecke and mathematician D. Russell Luke, now reconstructs the complete 3D molecular orbital of a nanometre-sized organic molecule, resolving features smaller than the spacing between its carbon atoms, entirely in the laboratory. Two ingredients make this possible: a reconstruction algorithm redesigned from the ground up that produces reliable 3D images from far less experimental data, and a table-top, high-power soft-X-ray/extreme-ultraviolet source that delivers ultrashort, femtosecond pulses.
Because that light source is pulsed, the technique is inherently time-resolved. “This might mean that stroboscopic videography becomes a reality,” says Bennecke — watching not just the shape of a wavefunction but how it changes with femtosecond resolution as a molecule responds to light, charge or chemical change, and ultimately learning to steer those interactions at the level of a few atoms. Published in Nature Communications 17 (2026); University of Göttingen press release, reported 24 August 2026.
ゲッティンゲン大学のStefan MathiasとG. S. Matthijs Jansenが率い、筆頭著者Wiebke Benneckeと数学者D. Russell Lukeを含む学際チームは今回、ナノメートルサイズの有機分子の完全な3次元分子軌道を、炭素原子間の間隔より小さい構造まで分解して、すべて実験室内で再構成した。それを可能にしたのは2つの要素だ。根本から設計し直され、はるかに少ない実験データから信頼できる3次元像を生成する再構成アルゴリズムと、フェムト秒の超短パルスを出す卓上型の高出力軟X線／極端紫外光源である。
The popular picture of quantum hacking is factoring large integers, but roughly 90% of deployed public-key cryptography today instead rests on the elliptic-curve discrete-logarithm problem — the mathematics behind web security, software authentication, electronic passports and the transaction signatures of Bitcoin and Ethereum. A classical computer would need a billion years or more to break it; Shor-type quantum algorithms could do it in principle, but earlier resource estimates put the requirement at millions of physical qubits, comfortably beyond any foreseeable machine.
Ryan Babbush of Google Quantum AI and colleagues — including Craig Gidney, Hartmut Neven, Ethereum researcher Justin Drake and Stanford cryptographer Dan Boneh — have re-evaluated that estimate using a new hacking algorithm and recent advances in quantum error correction. For security reasons they withhold the algorithm’s details, instead supplying a zero-knowledge proof of its efficacy (a cryptographer has since reconstructed a full version from the published hints, arXiv:2606.02235). Their conclusion: a superconducting quantum computer with about 500,000 physical qubits and roughly 80 million gates could break elliptic-curve cryptography in under ten minutes — a twenty-fold reduction in qubit count from prior estimates, and a runtime short enough to threaten cryptocurrency transactions during the brief window in which a public key is exposed. A separate estimate by the company Oratomic for slower but lower-error reconfigurable neutral-atom qubits comes in at about 26,000 physical qubits, at the cost of a runtime of several days (arXiv:2603.28627).
Quantum computers today have passed 1,000 qubits and their error rates keep falling, so “Q-Day” depends on hardware as much as on algorithms; Scott Aaronson (UT Austin) calls the results “reasonable” and says they appear to shorten the timeline. The authors urge migration to post-quantum cryptography “without delay”, and Stephanie Simmons (Simon Fraser University) adds a sobering note: the publicly known state of the art “is unlikely to be the actual state of the art.” Published in PRX Quantum 7, 031001 on 21 August 2026; featured as a Physics Focus story.
For 21 years NASA’s Neil Gehrels Swift Observatory has been the workhorse of time-domain astronomy: when its Burst Alert Telescope catches a gamma-ray burst or other sudden outburst, the spacecraft slews within minutes and alerts observatories worldwide, a capability central to the study of gamma-ray bursts, tidal disruption events, magnetar flares and the electromagnetic counterparts of gravitational-wave sources. Swift sits in low Earth orbit and has no propulsion of its own, so atmospheric drag slowly pulls it down — and the Sun’s recent high activity has heated and expanded the upper atmosphere, accelerating the decay far beyond what was planned.
In September 2025 NASA contracted Katalyst Space to attempt something never done before for a science satellite: fly a servicing spacecraft, LINK, to Swift, capture it and boost it to a higher orbit, all within a year. LINK launched in July 2026 on a Northrop Grumman Pegasus rocket, but developed attitude-control problems, and on 19 August NASA and Katalyst announced that the capture and reboost would not be attempted; LINK will still perform rendezvous and proximity operations near Swift to test in-space servicing technologies. Meanwhile Swift had been placed in a low-drag configuration, with the Ultraviolet/Optical Telescope and X-ray Telescope switched off in February and the Burst Alert Telescope idled in April to save power and reorient the solar panels.
With the rescue off the table, NASA switched the UVOT and XRT back on on 26 August and expects to return the Burst Alert Telescope to data collection within a few weeks, collecting as much science as possible. The critical threshold is an altitude of about 300 km (185 miles): below it, operating the telescope becomes difficult and the orbit decays faster still. Earlier estimates had Swift staying above that line until October; the team now expects to reach it within one to two months. Part of Swift’s rapid-response role can be taken up by newer missions such as the Einstein Probe and SVOM, but its loss will be felt across multi-messenger astronomy.
The 1998 discovery of cosmic acceleration rests on Type Ia supernovae (SNe Ia) being standardizable candles — on the assumption that, once corrected for light-curve shape and colour, their intrinsic brightness does not drift with cosmic time. In late 2025 a Yonsei University team reported that this assumption fails: standardized SN Ia magnitude correlates with the stellar-population age of the progenitor, and because host ages evolve systematically with redshift, an uncorrected age bias would mimic exactly the dimming that acceleration was inferred from. After applying their age correction, the supernova sample became consistent with a non-accelerating, rapidly evolving dark-energy (w0waCDM) universe — and, strikingly, with the independent, supernova-free constraints DESI derives from baryon acoustic oscillations (BAO).
In June 2026 an international group led by the University of Southampton (including two Nobel laureates) published a rebuttal in MNRAS, arguing that the Yonsei analysis omitted the standard host-galaxy stellar-mass correction, and that once it is applied no residual age dependence survives; they also argued that progenitor ages evolve with redshift far less steeply than host-galaxy ages do.
The Yonsei team has now answered that rebuttal in MNRAS. Their central technical point is that the Southampton host-age–Hubble-residual slope is severely underestimated because their combined SN sample spans an unusually wide redshift range (0.04 < z < 0.42), over which the mean host age itself evolves by roughly 3 Gyr; supernovae with very different host ages are therefore assigned similar Hubble residuals before regression, artificially flattening the inferred age–residual relation. They further argue that the rebuttal emphasised the reduced age evolution while overlooking the accompanying increase in the age-bias slope — when both effects are treated consistently, the resulting correction, and hence the cosmological conclusion, is close to that of their original study. The dispute is not settled, and neither paper is a proof; what has changed is that the question of whether cosmic acceleration is an artefact of supernova standardization is once again live. Larger, homogeneous samples from the Vera C. Rubin Observatory are widely expected to provide the next test.
今回、延世大チームがこの反論にMNRAS誌上で再反論した。技術的な核心は、サウサンプトン側が導いた「母銀河年齢とハッブル残差の傾き」が大幅に過小評価されているという指摘である。彼らの統合サンプルは0.04 < z < 0.42という異例に広い赤方偏移範囲にまたがり、その間に母銀河の平均年齢自体が約30億年も進化する。そのため、母銀河年齢が大きく異なる超新星に回帰前からほぼ同じハッブル残差が割り当てられ、年齢と残差の関係が人為的に平坦化されてしまうという。さらに、反論は「年齢進化が小さくなること」だけを強調し、それに伴って年齢バイアスの傾きが大きくなることを見落としている——両者を整合的に扱えば、補正量も宇宙論的結論も当初の研究とほとんど変わらない、というのが延世大側の主張だ。決着はついておらず、どちらの論文も証明ではない。変わったのは、「宇宙加速膨張は超新星の標準光源化に伴うアーティファクトではないか」という問いが再び生きた問題になった、という点である。ヴェラ・ルービン天文台による大規模で均質なサンプルが次の検証の場になると期待されている。
Keywords: Type Ia supernova, Ia型超新星, cosmic acceleration, 宇宙加速膨張, dark energy, 暗黒エネルギー, progenitor age bias, 前駆天体年齢バイアス, standardizable candle, 標準光源
🕳️ / 宇宙に満ちる「重力波のハム音」は、130億年以上前のダークスターの残骸が奏でているのかもしれない——暗黒物質を熱源とする原始星が崩壊してできた超大質量ブラックホールの種が、パルサータイミングアレイのナノヘルツ重力波背景を支配しうると解析。逆にPTAの測定値を「初期種ブラックホールの存在量の上限」に読み替えられることを示す（Ghodla・Ilie、コルゲート大学、Phys. Rev. D Letter掲載）
In 2023, pulsar timing arrays (PTAs) reported a nanohertz stochastic gravitational-wave background — a faint, all-sky “hum” detected by watching millisecond pulsars for tiny, correlated timing deviations. The conventional explanation is a cosmic population of inspiralling supermassive black-hole binaries. But that explanation inherits an unsolved problem: how the first supermassive black holes grew so large so early, a puzzle sharpened by JWST’s over-massive early black holes.
Sohan Ghodla and Cosmin Ilie (Colgate University) connect the two. They follow the cosmological evolution of black holes grown from two candidate classes of early, heavy seed: direct-collapse black holes, and the remnants of supermassive Dark Stars — hypothetical primordial stars whose luminosity is powered by dark-matter (WIMP) annihilation heating rather than nuclear fusion, which stay cool and extended while accreting and can reach ~106 solar masses before collapsing. Modelling the host halos, merger rates and resulting strain spectrum, they find that the Dark Star channel can reach and even dominate the measured PTA background, whereas the much rarer direct-collapse population yields a substantially weaker signal.
The more useful consequence runs the other way: because overproducing seeds would overproduce the hum, the PTA measurement becomes an upper limit on the abundance of supermassive black-hole seeds formed at redshift z > 10. In their models, seed number densities in the 10−2–10−1 Mpc−3 range would already overshoot the observed background, with the exact threshold depending on the mass of the dark-matter halos in which the seeds formed. The framework also confirms that binaries with total mass above 109 solar masses generate the overwhelming majority of the PTA signal. Published as a Letter in Physical Review D on 17 August 2026; testing the picture will require both improved PTA sensitivity and better censuses of early black holes.
Scaling quantum computers beyond a single chip requires distributed entanglement between physically separated modules. In practice that has meant active schemes: repeated measurement, heralding and feedback, all of which add control overhead and latency. An alternative was proposed more than twenty years ago — engineer the qubits’ environment so that dissipation itself drives them into an entangled steady state — but it had never been realized for remote qubits.
A team at the Institute of Science and Technology Austria (ISTA) has now done it. A Josephson parametric converter generates microwave photons in a two-mode squeezed state; its two correlated outputs travel roughly 50 cm of coaxial line to two separate transmon qubits. That correlated field acts as a non-local reservoir — a “quantum bath” — whose dissipation continuously pushes the pair toward an entangled steady state close to a Bell state, with no measurement and no feedback. The state was verified by quantum-state tomography, reconstructing the two-qubit density matrix from many measurement bases.
The team reports that the entanglement is preserved about as well as with conventional continuous and discrete active schemes, while the measured concurrence remains modest, reaching no more than roughly 10% — limited by imperfect waveguide couplings, line losses and an unexpectedly large mismatch between the two qubits’ decay rates, not by the principle. The authors argue the approach should scale well precisely because the squeezing is generated over a bandwidth broader than the qubit linewidth, which suits multi-mode, longer-distance and higher-rate networks. Published in Physical Review X (ISTA press release 14 July 2026; recirculated by ScienceDaily on 31 August 2026); the ISTA group is separately pursuing microwave-to-optical links for connecting distant processors over fibre.