{"id":"84ec434c-4de6-4bbf-b798-4ee10c05e44d","arxiv_id":"2607.14712","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"Momentum-resolved EELS shows a highly anisotropic, linear exciton dispersion in multilayer CrSBr, with slope 7.02 eV Å along ΓY and a nearly flat dispersion along ΓX.","lead":"Electron-beam measurements in the layered magnet CrSBr show that its excitons move with a very steep, linear energy-momentum relation along one crystal axis and almost none along the perpendicular axis. The result gives a concrete material platform for steering exciton transport in low-symmetry semiconductors.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Unresolved X/X* doublet makes the single-Gaussian peak extraction vulnerable to q-dependent spectral weight transfer, so the linear slope may be an artifact; a two-peak fit would settle it.","rationale":"The reader identified the unresolved X/X* doublet as the weakest assumption, and my reading confirms that this is the most load-bearing threat to the central claim. The experimental result is presented as a direct observation of a linear exciton dispersion with a record slope, but the slope is extracted from single-Gaussian centroids in spectra where two resonances separated by only ~16 meV are not resolved. Over the tiny momentum window used for the linear fit, the reported energy shift is only ~49 meV, so even a modest q-dependent redistribution of spectral weight between X and X* could generate an apparent linear term. The BSE calculations provide independent evidence that a linear dispersion is theoretically expected, which is genuinely supportive, but they do not model the measured EELS lineshape or the X* contribution; therefore they cannot by themselves validate the experimental peak extraction. I do not see a reason to reject the paper on this basis—the test could well confirm the dispersion—but the central quantitative claim should not be treated as fully established until the two-peak extraction is performed. Since the reader's verdict is already CONDITIONAL and my concern matches theirs, the verdict remains unchanged.","tokens_in":8367,"tokens_out":3653,"duration_ms":37075,"concrete_test":"Fit the raw q-EELS spectra along ΓY for |q| ≤ 0.01 Å⁻¹ with a two-peak model whose relative separation and total linewidths are fixed to the reflectance-derived values (X: 1.360 eV, 16 meV; X*: 1.376 eV, 20 meV), allowing only the two amplitudes and a common energy shift to vary with q, and convolve with the measured 18 meV instrumental response. Compare the extracted common shift (true 1s dispersion) with the single-Gaussian centroids used in Fig. 3. If the two-peak common shift is flat to within ±10 meV over |q|<0.007 Å⁻¹ while the single-Gaussian centroids show the reported ~49 meV rise, the linear slope is an artifact of q-dependent spectral weight transfer. A complementary synthetic check: generate spectra from two non-dispersing peaks with q-dependent amplitudes chosen to reproduce the observed map intensity, run the paper's single-Gaussian pipeline, and see whether a 7.02 eV Å slo","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central quantitative result—the 7.02 eV Å linear slope along ΓY in Fig. 3(b)—rests on single-Gaussian peak positions extracted from q-EELS spectra. The same sample's zero-momentum optical response (Fig. 1b) requires two b-polarized resonances: 1s(X) at 1.360 eV and X* at 1.376 eV, with linewidths 16 and 20 meV, while the EELS energy resolution is ~18 meV. X and X* are therefore not individually resolved in the q-EELS maps. The reported linear range |q|<0.007 Å⁻¹ corresponds to a total energy shift of ~49 meV—about three times the X/X* splitting. If the relative oscillator strength or energy of X* changes with q, even by a small amount, the position of a single fitted Gaussian will shift without any change in the true 1s exciton dispersion. The manuscript does not report any q-dependent two-peak fit, nor does it address how the X* component behaves with momentum. The BSE comparison cannot rule this out: only the 1s dispersion is shown, with no calculated q-dependent lineshape or spectral weight for X*. Because the headline slope and the 'largest reported group velocity' claim both depend on this extraction, the doublet ambiguity is load-bearing.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript reports momentum-resolved EELS measurements of the exciton dispersion in multilayer CrSBr. Using defocus-engineered q-EELS, the authors map the lowest exciton band along ΓY and ΓX and observe a strong in-plane anisotropy: a linear dispersion along ΓY with slope 7.02 eV Å within |q|<0.007 Å^{-1} and an essentially dispersionless response along ΓX. Finite-momentum BSE calculations reproduce the linear ΓY dispersion, and the authors attribute the effect to the long-range electron–hole exchange interaction enhanced by strong out-of-plane confinement and governed by the b-axis transition dipole. They also compare measurements at 100 K (AFM) and 300 K (PM) and report no resolvable change across the magnetic transition.","tokens_in":8789,"tokens_out":4821,"duration_ms":47809,"significance":"If the measurement is robust, this would be a valuable direct observation of a strongly anisotropic, linear exciton dispersion in a van der Waals magnetic semiconductor, with a group velocity larger than several previously studied low-dimensional systems. The BSE support is independent and not fitted to the measured slope, which strengthens the intrinsic-origin interpretation, and the analytical model in Eq. (2) provides a plausible microscopic mechanism. The magnetic-phase comparison is a useful additional piece of information. The main quantitative claim, however, rests on peak positions extracted from spectra that may contain two unresolved resonances, so the experimental evidence needs additional analysis before the central conclusion is fully established.","major_comments":[{"comment":"The headline slope of 7.02 eV Å is obtained from single-Gaussian peak positions, but the same sample's zero-momentum response (Fig. 1b, Eq. 1) requires two b-polarized oscillators: 1s(X) at 1.360 eV and X* at 1.376 eV, with widths 16 and 20 meV, separated by only ~16 meV. The EELS energy resolution is ~18 meV, so X and X* are not individually resolved in the q-EELS maps. The manuscript states that X* is only 'a lineshape modulation,' but no q-dependent two-peak fit is reported and no calculated finite-q lineshape or spectral weight for X* is shown. If the relative strength or energy of X* changes with q, the centroid of a single fitted Gaussian can shift by tens of meV without any true 1s dispersion. Since the reported linear range |q|<0.007 Å^{-1} corresponds to a total shift of ~49 meV—about three times the X–X* splitting—this is a plausible artifact. Please provide a two-peak fit of t","section":"§3, Fig. 3(b); Fig. 1(b)"},{"comment":"The linear slope 7.02 eV Å is reported without a statistical uncertainty, and the experimental peak positions in Fig. 3(b) are shown without error bars. The linear-fit range is stated in the abstract as |q|<0.007 Å^{-1}, but the main-text description and Fig. 3(b) do not precisely define the fitted range, the number of points, or the residuals. The comparison with slopes in other materials ('largest reported group velocity') depends directly on this number. Please report the fit range, the slope uncertainty, and the peak-position uncertainties; also show how the result changes if the range is varied or if a quadratic term is included.","section":"§3, Fig. 3(b), Eq. (4)"}],"minor_comments":[{"comment":"The color scale and normalization of the momentum–energy maps are not defined in the main text. Please state what quantity is plotted (e.g., normalized loss intensity), the normalization procedure, and the color mapping.","section":"Fig. 2(c,d)"},{"comment":"The Gaussian fitting procedure is not specified: no background model, energy window, or fit-function details are given. This information should appear in the main text or in the Supplemental Material, along with examples of the fits and residuals.","section":"§3, Fig. 3(a)"},{"comment":"The claim that the dispersion is 'essentially unchanged' across the magnetic transition is supported only by qualitative comparison to supplemental figures. A quantitative bound on any change in slope or energy (e.g., within the uncertainty of the 100 K measurement) would make this claim more convincing.","section":"§4, Supplemental Figs. S3–S5"},{"comment":"The statement that X* is 'a lineshape modulation rather than a distinct spectral peak' is presented without a criterion. Since the two oscillators overlap heavily, a constrained two-oscillator fit or a discussion of the fit degeneracy would clarify why this interpretation is preferred.","section":"§2, Eq. (1)"},{"comment":"The Supplemental Material is referenced for computational details and additional figures, but is not included in the submitted manuscript. Please ensure all referenced supplementary material is available and contains the full BSE parameters, convergence tests, and the 300 K data analysis.","section":"General"}],"recommendation":"major_revision","confidential_remarks":"The core experimental result is plausible and the BSE support is independent, but the unresolved X/X* doublet makes the single-Gaussian peak extraction a load-bearing concern. If the authors can demonstrate with two-peak fits or equivalent analysis that the linear slope is not an artifact of q-dependent spectral-weight transfer, I would support publication."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The new thing here is real: a direct q-EELS measurement of the exciton dispersion in CrSBr, showing a sharp linear upturn along ΓY and a flat band along ΓX. That anisotropy is exactly what the field has been looking for in this material, and the defocus-engineered setup gives momentum resolution good enough to see the small-q regime. The BSE calculation is independent — not fitted to the measured slope — and it reproduces the linear dispersion. That is genuine support, not hand-waving. The across-magnetic-phase comparison is a nice extra, even if it is qualitative. I also appreciate that the analytical model is taken from earlier work and used for interpretation; the paper does not pretend the model is new.\n\nThe soft spot is the one the stress test flags. The reflectance fit at q=0 already needs two b-polarized resonances, X and X*, separated by 16 meV, with linewidths 16–20 meV. The EELS energy resolution is 18 meV. So the q-EELS spectra cannot resolve them, yet the peak position vs q is extracted with a single Gaussian at every q. The reported linear range |q|<0.007 Å⁻¹ spans about 49 meV of shift — three times the X/X* splitting. If even a few percent of the X* spectral weight moves with q, the single-Gaussian centroid shifts without any true 1s dispersion. The paper never addresses this, and the BSE comparison doesn't help because only the 1s branch is shown. This is a load-bearing ambiguity, not a cosmetic one.\n\nThere are smaller issues: the slope is quoted as 7.02 eV Å with no uncertainty, the linear fit range is described as |q|<0.007 Å⁻¹ but I did not see how that range was chosen, and the spectra are normalized over 1.25–1.55 eV per q, which can distort lineshapes when background varies. No data or code is shipped, which makes the missing error bars harder to forgive.\n\nWho gets value from this? 2D exciton and q-EELS people will want to read it; it is a strong candidate measurement of anisotropic exciton dispersion in a van der Waals magnet. But the quantitative claim — \"largest reported slope,\" the group velocity — rests on the single-peak extraction. The fix is straightforward: a two-peak fit with the X* parameters either held fixed or allowed to evolve smoothly with q, plus an error estimate on the slope. If that survives, this is a very nice paper. As is, it deserves a serious referee but needs revision before I would cite the slope.","headline":"First momentum-resolved look at CrSBr excitons shows a beautifully anisotropic dispersion, but the headline slope depends on single-Gaussian fits to an unresolved doublet, so the quantitative claim needs a robustness check before I'd trust it.","tokens_in":9200,"tokens_out":1144,"would_cite":true,"duration_ms":13207,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"CrSBr excitons show an intrinsically anisotropic dispersion: linear along the b-axis at 7.02 eV·Å and flat along the a-axis, traced to long-range electron–hole exchange.","keywords":["exciton dispersion","momentum-resolved EELS","CrSBr","anisotropic exciton transport","electron-hole exchange","Bethe-Salpeter equation","van der Waals magnetic semiconductor","layered semiconductor"],"falsifier":"Repeat the q-EELS measurement at an energy resolution better than ~10 meV, or fit each momentum-resolved spectrum with two Gaussians for the X and X* resonances; if the isolated X peak does not reproduce the 7.02 eV·Å linear slope, then the single-exciton analysis is the source of the effect. Alternatively, map the full 2D dispersion and check whether the linear term obeys the predicted cos²θ_q angular dependence.","tokens_in":8329,"feed_emoji":"⚡","tokens_out":6830,"duration_ms":63863,"temperature":0.7,"pith_summary":"This paper sets out to measure how the energy of an exciton in the layered magnetic semiconductor CrSBr changes with its center-of-mass momentum, and to explain why. Using momentum-resolved electron energy-loss spectroscopy with engineered defocus, the authors find a linear dispersion along the crystalline b-axis with a slope of 7.02 eV·Å, one of the largest reported, and almost no dispersion along the a-axis. They attribute this anisotropy to the long-range electron–hole exchange interaction, strengthened by strong out-of-plane confinement and directed by the transition dipole alignment along b, and they support it with finite-momentum Bethe–Salpeter calculations. If correct, CrSBr becomes a model system for studying anisotropic exciton transport and a testbed for engineering directional energy flow in low-symmetry layered semiconductors.","feed_headline":"CrSBr excitons show record 7.02 eV·Å linear dispersion","feed_subtitle":"The b-axis slope is the highest seen in low-dimensional materials, pointing to fast directional exciton flow.","key_machinery":"The central instrument is a defocus-engineered momentum-resolved electron energy-loss spectroscopy (q-EELS) configuration in a scanning transmission electron microscope: by defocusing the sample, electrons at different incident angles land at different positions, yielding a momentum resolution of ~4×10⁻⁴ Å⁻¹ near the Brillouin zone center. The paper pairs this with finite-momentum Bethe–Salpeter equation (first-principles exciton) calculations and an analytical expression E(q) = E0 + A|q|cos²θ_q + parabolic terms, where the linear term arises from long-range electron–hole exchange and cos²θ_q encodes the transition dipole's directional selection. Along ΓY (the dipole axis) the linear term do","core_discovery":"The paper reports that momentum-resolved electron energy-loss spectroscopy on multilayer CrSBr reveals an exciton dispersion that is linear along the crystalline b-axis (ΓY) with a slope of 7.02 eV·Å within |q| < 0.007 Å⁻¹, while remaining nearly dispersionless along the a-axis (ΓX). Finite-momentum Bethe–Salpeter calculations reproduce the linear branch, confirming its intrinsic origin. The authors identify the mechanism as the long-range electron–hole exchange interaction, enhanced by strong out-of-plane confinement of the exciton wavefunction and governed by the directional selection rules of the b-axis transition dipole moment. Comparative measurements across the magnetic phase transitio","pith_inferences":["The angular factor cos²θ_q in the analytical model predicts that the linear slope should vanish continuously as the momentum direction rotates from ΓY to ΓX; a full 2D dispersion map would directly test this prediction and check whether additional directional exchange components exist.","If the linear dispersion is truly exchange-driven, its slope should depend on the degree of out-of-plane confinement; measuring the dispersion as a function of layer thickness (from monolayer to bulk-like flakes) would isolate this dependence and could guide tuning of exciton transport.","The apparent insensitivity to magnetic order suggests the exciton center-of-mass motion is decoupled from spin degrees of freedom at these momenta, but applying an external magnetic field to drive a spin-flop transition could reveal any hidden spin–exciton coupling that static measurements miss.","A spatially resolved photoluminescence experiment along the b-axis could look for signatures of the large group velocity, such as long diffusion lengths or ballistic propagation, providing a transport-level confirmation of the measured dispersion."],"forward_implications":["CrSBr emerges as a model system for anisotropic exciton dynamics in low-symmetry layered semiconductors, with an intrinsic momentum scale set by the b-axis dipole.","The record slope of 7.02 eV·Å corresponds to an unusually large exciton group velocity along the b-axis, suggesting fast directional energy transport over mesoscopic length scales.","The near-flat dispersion along the a-axis implies exciton propagation is strongly suppressed in that direction, making CrSBr a platform for quasi-one-dimensional exciton behavior.","Because the dispersion is essentially unchanged between paramagnetic and antiferromagnetic phases, magnetic ordering does not appreciably renormalize the exciton center-of-mass motion, simplifying the interpretation of exciton dynamics in magnetic van der Waals semiconductors.","The combination of defocus-engineered q-EELS and Bethe–Salpeter calculations establishes a general route for measuring small-momentum exciton dispersions in anisotropic layered materials."],"fun_headline_variants":["Anisotropic exciton dispersion directly observed in CrSBr","CrSBr excitons: linear b-axis dispersion, flat a-axis","CrSBr's excitons: 7.02 eV·Å slope along b-axis","Momentum-resolved excitons reveal CrSBr's anisotropic flow","Direct measurement of anisotropic exciton motion in CrSBr"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The analysis assumes the energy-loss feature is a single 1s exciton peak at every momentum, even though reflectance fitting shows a second, weaker exciton (X*) only ~16 meV away; if that second resonance shifts or changes weight with momentum, the Gaussian-extracted peak positions could create an apparent linear dispersion rather than the true 1s branch.","fun_headline_variants_meta":{"raw":{"variants":["Anisotropic exciton dispersion directly observed in CrSBr","CrSBr excitons: linear b-axis dispersion, flat a-axis","CrSBr's excitons: 7.02 eV·Å slope along b-axis","Momentum-resolved excitons reveal CrSBr's anisotropic flow","Direct measurement of anisotropic exciton motion in CrSBr"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001121,"raw_usage":{"total_tokens":4491,"prompt_tokens":726,"completion_tokens":3765,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":470,"completion_tokens_details":{"reasoning_tokens":3670}},"tokens_in":470,"tokens_out":3765,"duration_ms":25284,"temperature":1.0,"reasoning_tokens":3670,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-02T01:15:59.309963+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Repeat the q-EELS measurement at an energy resolution better than ~10 meV, or fit each momentum-resolved spectrum with two Gaussians for the X and X* resonances; if the isolated X peak does not reproduce the 7.02 eV·Å linear slope, then the single-exciton analysis is the source of the effect. Alternatively, map the full 2D dispersion and check whether the linear term obeys the predicted cos²θ_q angular dependence.","supporting_citations":[],"review_version":1}