There is a class of materials that two-dimensional physics has mostly had to admire from a distance: the metal oxides. Oxides host high-temperature superconductivity, Mott physics, multiferroicity, and colossal magnetoresistance, and they underpin much of electronics, catalysis, and energy conversion. But most oxides are not layered. Their bonding is ionic or covalent in three dimensions, so the trick that opened the 2D era — peeling a monolayer off a bulk crystal — does not work. The thinnest films that could be made were unit-cell thick, several atomic layers at minimum, and they mostly behaved like bulk. A genuine single-atomic-layer oxide, free-standing and crystalline, remained out of reach.
A paper published in Nature Materials on October 29, 2026, by Meng Gao, Deping Guo, Ang Li, Renhong Wang, and colleagues — with Wu Zhou, Wei Ji, and Stephen J. Pennycook among the senior authors, at the University of Chinese Academy of Sciences, Renmin University of China, and Shanghai Jiao Tong University — reports that the limit has now been crossed, and crossed by subtraction rather than growth. The team starts from a layered metal oxyhalide, a crystal that is exfoliable, and uses a focused electron beam inside a scanning transmission electron microscope to strip the halogen atoms away. What remains relaxes into a free-standing oxide monolayer. Applied to chromium, erbium, and bismuth oxyhalides, the method yields monolayers of CrO, ErO, and BiO, plus lateral heterostructures and twisted bilayers. The bilayers, the authors report, show pronounced excitonic responses with an unusually large exciton binding energy. For a field that has spent a decade learning to read excitons in dichalcogenides and perovskites, a new, atomically thin oxide family with strong excitonic character and a predicted magnetic order is a platform worth taking seriously — with the usual caveats about what has been measured and what has only been calculated.
Why oxides resisted the monolayer limit
The difficulty is structural, and it is worth stating plainly because it explains why this result took so long. In a van der Waals crystal such as MoS₂ or graphite, the layers are held together by weak forces, so mechanical exfoliation isolates a monolayer cleanly. In a multinary layered oxide — a Bi-cuprate, an Aurivillius phase, a Ruddlesden–Popper perovskite — the thinnest exfoliable block is a full unit cell, multiple atomic layers thick, and it largely reproduces bulk behavior.
Growth methods filled part of the gap. Molecular beam epitaxy, liquid-metal printing, and intercalative oxidation have all produced ultrathin oxides, but the films typically suffer from substrate coupling, chemical instability, or degraded crystallinity. Freestanding SrTiO₃ and BiFeO₃ films at single-unit-cell thickness — still thicker than one atomic layer — required sacrificial buffer layers and still showed poor crystallinity. Earlier electron-beam experiments produced atomically thin patches, but only small ones confined within nanopores. What was missing was a route to a large, clean, free-standing monolayer of an oxide with no layered bulk analogue at all.
Subtraction instead of growth
The new route inverts the problem. Instead of trying to grow an oxide one layer thick, the team starts with a layered oxyhalide — a compound of metal, oxygen, and halogen — and removes the halogen. The model system is CrOCl, chromium oxychloride, grown as a monolayer by chemical vapor deposition on mica and transferred onto a transmission electron microscopy grid. In CrOCl, chromium sits at the center of distorted octahedra formed by chlorine and oxygen atoms. When an atomic-sized electron probe, operated at 60 keV in an aberration-corrected microscope, is scanned over a suspended CrOCl monolayer, defects appear rapidly and coalesce into a new crystal with a square lattice.
The identification of that crystal is careful work. The new lattice is dimmer in annular dark-field imaging than the parent, consistent with atom loss. Electron energy loss spectroscopy (EELS) shows chromium L₂,₃ and oxygen K edges with no detectable chlorine signal: the dechlorination is complete. Atomic-scale EELS mapping resolves individual chromium and oxygen sites, and quantitative analysis gives a Cr:O ratio of 1:1, confirmed by atom counting. Four-dimensional STEM ptychography visualizes both the heavy chromium and the light oxygen atoms, with oxygen centered within chromium squares. The product is monolayer CrO, a square lattice with a lattice constant of 0.405 nm measured experimentally (0.404 nm in theory), stabilized by a slight alternating buckling of about 0.17 ångströms. The single-crystalline domains exceed 130 nm across — more than 10,000 square nanometres — and the paper reports controllable patterning at the micrometre scale, since the beam writes where it scans.
The mechanism matters as much as the product, because it is what makes the method general rather than a one-material accident. Simulations show that knock-on displacement — the beam physically knocking atoms out — cannot on its own eject chlorine from a suspended CrOCl monolayer. Ionization is the dominant mechanism: beam-induced excitation transiently populates an unoccupied state near the Brillouin-zone center with pronounced antibonding character for Cr–Cl bonds, weakening them. The effect is much stronger for Cr–Cl than for Cr–O bonds, consistent with chlorine vacancies costing about 2.2 eV less to form than oxygen vacancies in density functional theory (DFT). Once the chlorine is gone, the lattice expands anisotropically — the a-axis grows from 0.320 nm to 0.404 nm while the b-axis elongates by only 3.8% — flattening the corrugated Cr–O layer into a nearly planar square lattice. The authors argue the same ionization-driven selectivity should extend across the oxyhalide family, and a high-throughput DFT screen identified 27 candidate freestanding monolayers with 1:1 stoichiometry and tetragonal symmetry.
Three blocks of the periodic table
The generality claim is tested, not just asserted. From ErOCl, the team obtained monolayer ErO with an orthorhombic lattice (a = 0.38 nm, b = 0.61 nm) and an Er:O ratio close to 1:1, again with no detectable chlorine edge. From BiOCl, a p-block oxyhalide, the behavior was markedly different: monolayer BiOCl is highly beam-sensitive and turns amorphous, but bilayer BiOCl under 200 keV irradiation transforms into a square-lattice phase (a = b = 0.38 nm) hosting a central Bi–Bi dimer in each unit cell. DFT identifies it as a chlorine-intercalated Bi₂O₂–M bilayer, Bi₄O₄Cl, in which residual intercalated chlorine donates charge, shifting the bismuth valence from +2 toward +3 and stabilizing the bilayer through coupled charge doping and stacking reconstruction. That is a second mechanism hiding inside the first: where complete dehalogenation fails, partial dehalogenation with residual halogen intercalation can stabilize a metastable ultrathin oxide instead.
Together the three materials span the d-block (chromium), the f-block (erbium), and the p-block (bismuth). That breadth is the paper’s real claim to generality. It is also, honestly, the point at which the evidence thins: three materials demonstrate a method, but the 27 predicted monolayers remain predictions, and the accessible set is limited, as the authors note, by the availability of high-quality oxyhalide precursors. The method’s reach will be measured by the fourth, fifth, and sixth materials, not the first three.
A semiconductor, a predicted magnet, and a moiré
Monolayer CrO, the representative system, is characterized in unusual depth. DFT gives an indirect bandgap of 2.11 eV; valence EELS measures 2.23 ± 0.18 eV, consistent within uncertainty. The monolayer is dynamically stable — its phonon dispersion shows no imaginary frequencies once the buckling is included — and mechanically flexible: some regions ripple substantially without losing crystallinity, a striking contrast with the brittleness of bulk oxides.
Magnetically, the calculations predict a Néel antiferromagnetic ground state with out-of-plane moments, and the published paper estimates the ordering temperature at below 230 K. That number deserves a careful sentence. It is a theoretical estimate from an anisotropic Heisenberg model fitted to DFT, not a measured magnetic transition. No magnetometry, neutron scattering, or magneto-optical measurement of monolayer CrO’s ordering temperature is reported. The estimate is high enough to be interesting — above liquid-nitrogen temperature — but until an experiment confirms it, 230 K is a prediction with a model’s error bars, not a property of a material. Readers should also note that an earlier preprint of this work estimated 116 K; the published figure revises that estimate upward, which is itself a reminder of how sensitive such estimates are to model details. The honest statement is: monolayer CrO is predicted to be a semiconducting antiferromagnet, and its ordering temperature is an open experimental question.
The structural control extends beyond monolayers. Because the beam converts only where it scans, the team fabricated coherent lateral heterostructures in which CrOCl and CrO coexist side by side within the same crystal, with an atomically sharp interface along the CrOCl y-axis, free of misfit dislocations — coherence made possible by the nearly matched lattice constants along that axis (0.390 vs. 0.405 nm). And by converting pre-twisted CrOCl bilayers, they produced twisted CrO bilayers at final angles of 22.9°, 36.8°, 28.1°, and 41.4°, forming short-period moiré superlattices (√13×√13, √5×√5, √17×√17, and √73×√73 respectively). The √5×√5 bilayer appears in three distinct stacking polymorphs, and the √13×√13 bilayer shows a square-octagon (4-8) tiling pattern — a decorated square lattice theoretically linked to topological phases. This is oxide twistronics at the atomic-layer limit, built with the same tool that made the monolayers.
Where the excitons come in
For this publication, the most consequential claim is the optical one. The fabricated bilayers — aligned and twisted — exhibit pronounced excitonic responses with an unusually large exciton binding energy, according to the study, opening opportunities for exploring emergent excitonic phenomena in low-dimensional oxides. The paper’s framing is that the combination is new: a free-standing oxide monolayer that is simultaneously a semiconductor, a predicted antiferromagnet, and a host of strongly bound excitons, with twist angle and stacking registry as tuning knobs.
Two qualifications belong here. First, the published summary does not attach a single numerical binding energy to the bilayer excitons, and the preprint version of the work contains no excitonic characterization section; the excitonic results appear to have been developed for the published paper. This article therefore reports the claim at the level the study states it — pronounced response, unusually large binding energy — rather than inventing a number. Second, the excitonic response of a material whose monolayer is a predicted antiferromagnet raises an obvious next question that the paper does not yet answer: how do the excitons couple to the magnetic order? In NiPS₃, the magnetic exciton’s identity was settled only by strain-resolved spectroscopy that distinguished local from hopping excitations. CrO offers a parallel opportunity — an exciton in a square-lattice antiferromagnetic oxide, tunable by twist — but the coupling experiments, like the magnetic-ordering measurement, remain to be done.
What it is not
Three distinctions keep this result in its proper place. First, it is not a scalable manufacturing process, not yet. The domains exceed 10,000 square nanometres and the patterning reaches the micrometre scale, but the fabrication happens inside an electron microscope, one beam-written region at a time, on material transferred to a TEM grid. “Prospects for scalability” is the paper’s phrase, and it is a prospect, not a demonstration. Any claim that atom-thin oxides are now manufacturable would be reading past the evidence.
Second, it is not a device result. No transistor, photodetector, or emitter is demonstrated. The monolayer’s bandgap is measured by EELS, not by a transport or photoluminescence device; the magnetism is predicted, not measured; the excitonic response is reported for bilayers, not engineered into an application. The significance is as a materials platform: new phases at the atomic limit, with properties that bulk oxides do not offer.
Third, it is not a claim that all oxides can now be thinned. The method requires a layered oxyhalide precursor, and its success depends on the ionization selectivity between metal–halogen and metal–oxygen bonds. Bismuth already shows the failure mode — monolayer BiOCl amorphizes, and only the bilayer survives as a stabilized intercalated phase. The periodic-table screen predicts candidates; the precursors and the beam physics will decide which ones are real.
What to watch
The immediate experiments write themselves. Confirm the Néel order in monolayer CrO and measure its ordering temperature — the 230 K prediction is the most falsifiable number in the paper, and the most consequential if confirmed, since it would put a 2D antiferromagnetic oxide within reach of nitrogen cooling. Measure the bilayer excitons spectroscopically: energies, linewidths, twist-angle dependence, and any signature of coupling to the magnetic order. And extend the materials set beyond the first three, particularly toward predicted monolayers whose precursors can actually be grown.
The longer consequence is conceptual. Two-dimensional physics has largely been built on materials that nature made layered. This work shows that a subtractive, beam-driven conversion can create atomically thin crystals with no layered bulk counterpart — CrO has none; bulk CrO is rock-salt, and the monolayer is best described as half a rock-salt unit cell, stabilized by buckling into a distinct single-atom-thick layer. If the method generalizes as the authors argue, the 2D materials library stops being limited to what exfoliates. For exciton physics specifically, that means a new family of hosts in which strong binding, magnetic order, and moiré geometry can be combined in one atomically thin system. The fabrication is done; the spectroscopy is the next chapter.
Sources and further reading
Gao, M.; Guo, D.; Li, A.; Wang, R.; Jin, Z.; Huang, C.; Lu, Y.; Pan, Y.; Dai, J.; Huang, P.; Liu, K.; Wang, C.; Bao, L.; Zhou, L.; Qiao, J.; Ji, W.; Pennycook, S. J.; Gao, H.-J.; & Zhou, W. “Free-standing two-dimensional metal oxides at the single-atomic-layer limit.” Nature Materials (published October 29, 2026). DOI: 10.1038/s41563-026-02793-2. Domain sizes (>10,000 nm²), micrometre-scale patterning, lateral heterostructures, twisted bilayers, excitonic responses, and the 230 K Néel-order prediction are from the published paper as summarized at publication. Fabrication details (60 keV beam, ~2,000 e·Å⁻²·s⁻¹ dose rate; CrOCl CVD growth on mica; EELS Cr-L₂,₃ redshift 575.4 vs. 578.0 eV; 1:1 Cr:O stoichiometry; 0.404/0.405 nm lattice constant; ~0.17 Å buckling; 2.11 eV DFT bandgap vs. 2.23 ± 0.18 eV valence-EELS bandgap; ErO 0.38 × 0.61 nm orthorhombic lattice; Bi₄O₄Cl bilayer; twist angles 22.9°, 36.8°, 28.1°, 41.4°; 27 predicted monolayers; ~2.2 eV Cl-vs-O vacancy formation-energy difference), the ionization-driven mechanism, and the 116 K preprint estimate are from the preprint version: Gao et al., Research Square (posted November 5, 2025), DOI: 10.21203/rs.3.rs-7893525/v1. A Chinese patent covering this subtractive fabrication method, filed by the University of Chinese Academy of Sciences, is declared as a competing interest in the preprint. This article is analysis of published research, not investment advice.