{"id":"36991b92-7026-4623-9730-d418a7100055","arxiv_id":"2607.26003","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"Element-specific X-ray microscopy shows CrSBr magnetism flips with layer parity, and a 90° twisted CrSBr bilayer loses its in-plane easy axis, pointing out-of-plane instead.","lead":"This paper images magnetic order layer-by-layer in the 2D magnet CrSBr using X-ray microscopy, showing odd/even layer-count effects and that a 90° twisted bilayer no longer keeps the monolayer easy axis. It matters because twist engineering could give spintronics a new knob for controlling magnetic anisotropy in atomically thin devices.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The 90°-twisted c-axis easy-axis claim rests on an absence-of-signal argument (zero XMCD at grazing incidence, undecomposed XLD) with no direct out-of-plane probe.","rationale":"The reader's weakest assumption identifies the same load-bearing concern: the new easy-axis direction is inferred from the absence of XMCD and from XLD azimuthal invariance, without ruling out compensated in-plane or multi-domain states and without separating XMLD from XNLD. This is the most consequential gap because the paper's headline claim depends entirely on it. The proposed high-incidence XMCD measurement directly probes the out-of-plane component and would settle whether the easy axis is truly c-axis. The internal U inconsistency and the near-zero MAE further weaken the DFT support but are secondary; even if the DFT were corrected, the experimental inference would remain indirect. Therefore the verdict should remain CONDITIONAL: the claim is plausible but requires the direct out-of-plane measurement before it is established.","tokens_in":17729,"tokens_out":11023,"duration_ms":108933,"concrete_test":"Repeat the azimuthal XMCD-PEEM series on the same 90° twisted bilayer sample with the X-ray incidence angle increased to θ≈60° (or normal incidence if available), so that an out-of-plane moment projects onto the X-ray propagation direction with factor sin60°≈0.87 instead of 0.28. If the twisted region develops XMCD contrast at this geometry, the c-axis claim is directly confirmed; if it remains zero at all azimuths, an out-of-plane collinear easy axis is falsified.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim—that the orthogonally twisted region has an out-of-plane (c-axis) easy axis—is inferred, not measured. The evidence is (i) zero XMCD contrast at φ=0°,45°,90° and (ii) azimuth-invariant XLD ordering between φ=45° and 90°. Both are absence/non-decomposition arguments. XMCD was acquired at grazing incidence θ=16° (Methods), so the projection of an out-of-plane moment onto the X-ray direction is only sin16°≈0.28; 'negligible' contrast is never quantified against a noise floor or error bar. Zero net XMCD is equally consistent with compensated in-plane AFM order, multi-domain in-plane order, or a canted/non-collinear state—the latter explicitly conceded in the text ('we cannot exclude a non-collinear canted configuration at higher fields'). The XLD azimuthal invariance is ambiguous because XLD = XNLD (crystal field) + XMLD (magnetic); no XNLD baseline was measured in the twisted region above T_N (the 200 K XNLD control in Fig. 1f,g is for the untwisted multilayer). If XNLD dominates, the invariance carries no magnetic information. The DFT 'corroboration' is also weak: the stated MAE is ~0.002 meV/Cr (Table S2), and the supplement states U=0 was used for the twisted systems despite the Methods stating U_eff=3 eV—an internal inconsistency. Thus the central claim lacks a direct test of an out-of-plane moment.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript reports an X-ray magnetic circular/linear dichroism (XMCD/XLD) photoemission electron microscopy (PEEM) study of exfoliated CrSBr from multilayer down to monolayer, and of a 90°-twisted bilayer. In untwisted samples, the authors observe layer-parity-dependent net magnetization (odd layers magnetic, even layers compensated), consistent with A-type antiferromagnetic order, and show that in-plane field pulses switch odd layers while even layers remain stable. For the 90°-twisted bilayer, they find suppressed XMCD contrast at all azimuthal angles and azimuth-invariant XLD peak ordering, and interpret this as evidence that the easy axis of the twisted region is no longer in-plane but aligns along the out-of-plane c-axis, corroborated by DFT. The central claim is therefore that orthogonal twisting fundamentally changes the magnetic ground state of CrSBr.","tokens_in":18157,"tokens_out":3680,"duration_ms":33020,"significance":"If the twist-induced c-axis easy-axis claim is correct, the work is significant for moiré magnetism and for designing CrSBr-based spintronics. The strengths of the paper are the element-specific, layer-resolved XMCD/XLD-PEEM data down to the monolayer, the clear layer-parity demonstration, the field-switching behavior, and the fact that the DFT prediction is not fit to the XMCD maps and independently agrees with Ref. [35]. The untwisted results appear self-consistent and are likely a useful contribution. However, the central twist claim currently rests on absence-of-signal and non-decomposed XLD arguments, with no direct out-of-plane magnetic probe and an internal DFT parameter inconsistency; as presented, the evidence does not yet establish the c-axis easy axis.","major_comments":[{"comment":"The central conclusion is supported only by the absence of XMCD in the twisted region at φ = 0°, 45°, 90° (Fig. 4b–d). With the grazing incidence θ = 16° stated in Methods, the projection of a c-axis moment onto the X-ray direction is sin16° ≈ 0.28, and the paper does not quantify the minimum detectable XMCD asymmetry or provide error bars on the 'negligible' contrast. Zero net XMCD is equally consistent with compensated in-plane AFM order, multi-domain in-plane order, or a canted/non-collinear state — the last is explicitly conceded ('we cannot exclude a non-collinear canted configuration'). A direct out-of-plane probe (e.g., normal-incidence geometry or field along c) or a quantified upper bound is required to support the c-axis easy-axis claim.","section":"Magnetic Ground State in 90° Twisted Bilayer CrSBr; Methods"},{"comment":"The azimuthal-invariance XLD argument does not distinguish XNLD (crystal-field) from XMLD (magnetic). The only XNLD control above T_N (Fig. 1f,g) is for the untwisted multilayer, not for the twisted region. Without measuring XLD above T_N in the twisted region, or decomposing XLD(φ) into XNLD and XMLD contributions, the invariance cannot rule out an in-plane Néel vector. The conclusion 'invariant peak structure under rotation points to a structural reconstruction' is therefore an inference from a non-decomposed observable.","section":"Twist Angle Dependent XLD-PEEM and Fig. S10 (main text reference)"},{"comment":"There is an internal inconsistency in the DFT protocol. Methods state an effective Hubbard U_eff = 3 eV for Cr 3d electrons, while the supplement states that U = 0 was used consistently for bilayer and twisted systems because the QuantumATK MEA framework reproduces the VASP benchmark at U = 0. The manuscript must state clearly which calculations used which U. Moreover, the 90°-twist MAE in Table S2 is only ~0.002 meV/Cr (also quoted as 2.291 µeV/Cr), which is close to the 10^-7 eV total-energy convergence criterion quoted in Methods; the sign of such a tiny MAE is numerically fragile and the c-axis prediction needs convergence tests with respect to k-points, strain, and U.","section":"Computational Methods; Supplemental Table S2 and Fig. S3"},{"comment":"The narrative of a structural reconstruction in the twisted region is introduced to explain the XLD invariance, but no microscopic structural characterization (e.g., STM or TEM) is presented. The 4×3×1 commensurate supercell is a modelling assumption for a nominally incommensurate 90° twist. If the structural reconstruction is load-bearing for the claim that the easy axis changes, it needs independent support or the argument should be decoupled from it.","section":"Magnetic Ground State in 90° Twisted Bilayer CrSBr"}],"minor_comments":[{"comment":"The main text refers to 'Fig. S10' for the twist-angle-dependent XLD maps, but the supplement labels this figure as Fig. S9. Please correct the cross-reference.","section":"Figure references"},{"comment":"The author contributions mention 'A.L.N.K' as growing crystals, but this individual does not appear in the author list. Presumably A. L. N. Kondusamy from Ref. [23] is meant; clarify or correct.","section":"Author contributions"},{"comment":"Typos and formatting: 'tertalayer' in the Supplement should be 'tetralayer'; 'Crd-orbital' in the main text should be 'Cr d-orbital'; use consistent accents for 'Néel'.","section":"Throughout"},{"comment":"The phrase 'start voltage close to 0 to capture the secondary electrons cascade' is vague; please specify the electron kinetic-energy acceptance window or detector settings used for imaging.","section":"Methods - PEEM"},{"comment":"For the statement that the bottom layer and twisted region 'remain unaffected' by ±100 mT, provide a quantitative measure of contrast change (e.g., mean and standard deviation of asymmetry in those regions) rather than visual inspection alone.","section":"Fig. 4e,f"}],"recommendation":"major_revision","confidential_remarks":"The untwisted layer-parity results are solid and likely publishable, but the twist-dependent c-axis easy-axis claim is the headline result and currently rests on indirect evidence. Given that the DFT prediction independently agrees with a prior calculation, the claim may well be correct; however, the experimental support needs either a direct out-of-plane sensitive measurement or a quantified treatment of the XMCD upper limit and XNLD/XMLD decomposition. I recommend major revision rather than rejection because the issues are addressable within the manuscript's scope."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Two things to know. The layer-parity part is the real advance: element-specific XMCD/XLD-PEEM imaging of CrSBr from multilayer down to monolayer, with clear odd/even net-magnetization contrast and field-switching behavior. That is solid, well-controlled work. The twist part is a different story: the claim that a 90° twist reorients the easy axis out-of-plane is argued from absence of signal — zero XMCD at all azimuths and an azimuth-invariant XLD — not from a direct measurement of an out-of-plane moment. That is the load-bearing gap.\n\nWhat the paper does well: the depth-weighted PEEM signal is handled carefully; the layer-parity assignment is consistent with A-type AFM; the graphene encapsulation control is a nice touch. The azimuthal series in the twisted sample is a clever way to isolate the two monolayers, and the observation of persistent zero XMCD in the twisted region at 0, 45, and 90 degrees is striking. If that is real, it already distinguishes the twisted region from both monolayer responses. The DFT is less convincing: the MAE for the twisted case is ~0.002 meV/Cr, essentially zero, and the Methods say U_eff=3 eV while the supplement says U=0 was used for the twisted systems. The authors justify this choice, but it is still an internal inconsistency that needs to be resolved in revision.\n\nThe main soft spot is the XLD reasoning. The paper claims azimuth-invariant XLD rules out an in-plane Néel vector, but XLD contains both XNLD (crystal field) and XMLD (magnetic). No decomposition is given, and the XNLD control at 200 K was measured only on the untwisted multilayer, not in the twisted region. So the invariance could be dominated by crystal-field effects and carry no magnetic information. Also, zero XMCD at 16° grazing incidence is equally consistent with compensated in-plane AFM domains or a canted state — the paper itself concedes it cannot exclude non-collinear canted configuration. None of this is fatal to the layer-parity results, but it means the central twist claim is not established.\n\nWho should read this: anyone doing X-ray spectro-microscopy of vdW magnets, and the CrSBr community. The layer-parity data will be a useful reference. I would not cite the twist conclusion until it is backed by a direct out-of-plane probe (normal-incidence XMCD, for instance) or quantitative XLD decomposition with error bars.\n\nRecommendation: send it to peer review. The experimental work is substantial and the layer-parity part deserves publication. The referee should push hard on the twist claim and the DFT inconsistency; if the authors cannot supply direct evidence, the twist part should be reframed as suggestive rather than a conclusion.","headline":"The layer-parity XMCD-PEEM results are a genuine advance; the 90°-twist easy-axis reorientation is a plausible but unsupported inference that needs a direct out-of-plane probe.","tokens_in":18699,"tokens_out":3353,"would_cite":false,"duration_ms":33043,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"In CrSBr, rotating one monolayer 90° relative to another changes the magnetic easy axis itself, turning the twisted interface into a new magnetic ground state rather than a simple superposition of the two layers.","keywords":["CrSBr","van der Waals antiferromagnet","layer parity","twisted bilayer","XMCD-PEEM","magnetic anisotropy","moiré magnetism","Néel vector"],"falsifier":"Measure the same 90° twisted CrSBr bilayer with an X-ray geometry sensitive to out-of-plane moments—for example, normal-incidence XMCD at the Cr L3 edge—and check whether a perpendicular contrast appears that is absent at grazing incidence; equivalently, measure XLD as a function of azimuth on an unencapsulated twisted sample to disentangle XNLD from XMLD, since the c-axis conclusion requires the magnetic part to dominate the azimuthal invariance.","tokens_in":17632,"feed_emoji":"🧲","tokens_out":6068,"duration_ms":52980,"temperature":0.7,"pith_summary":"The paper aims to establish that the magnetic order of atomically thin CrSBr is layer-parity-controlled—odd layer counts carry a net moment along the crystallographic b-axis while even counts are compensated A-type antiferromagnets—and that a 90° twisted bilayer does not simply superpose two in-plane ferromagnetic orders. In the twisted region, X-ray magnetic circular dichroism is absent at every azimuthal angle, the linear dichroism does not rotate with the sample, and density-functional calculations put the easy axis along the out-of-plane c-axis. The authors interpret these observations as a structural reconstruction at the twisted interface that reshapes orbital hybridization and spin-orbit anisotropy pathways. The result matters because it shows, with element-specific nanoscale imaging, that twist angle can qualitatively reorient magnetism in a van der Waals antiferromagnet.","feed_headline":"90° twist sends CrSBr magnetism out of plane","feed_subtitle":"Element-specific imaging shows a twisted bilayer loses its in-plane easy axis, opening a new knob for moiré magnetism.","key_machinery":"The central mechanism is the combination of layer-resolved XMCD/XMLD-PEEM with azimuthal sample rotation. XMCD measures the projection of Cr magnetization onto the X-ray propagation direction, so for a given azimuth one monolayer's in-plane moment is visible while the orthogonal layer is silent; the persistent absence of XMCD in the twisted region therefore means no in-plane moment is available at any projection. XLD tracks both crystal-field orbital anisotropy and magnetic linear dichroism, and its rotation invariance is used to exclude an in-plane Néel vector. The accompanying density-functional calculations of magnetic anisotropy energy on a twisted supercell that relaxes toward a nearly","core_discovery":"Using element-specific X-ray magnetic circular and linear dichroism imaged with photoemission electron microscopy, the authors resolve the Cr spin structure of CrSBr from monolayer to tetralayer and into a 90° twisted bilayer. In untwisted films, the net magnetization is set by surface termination: odd layer counts give a net moment along the crystallographic b-axis, even counts are nearly compensated, consistent with A-type antiferromagnetic interlayer coupling. In the orthogonally twisted bilayer, XMCD contrast is absent at all azimuthal angles, which rules out preserved in-plane easy axes or a simple superposition of orthogonal moments in the twisted region. Azimuth-invariant XLD is inter","pith_inferences":["As an extension, a decisive test would be a measurement with out-of-plane magnetic sensitivity—for example, normal-incidence XMCD or field-angle-resolved XMLD—because the paper's c-axis conclusion currently rests on the absence of in-plane contrast, which is consistent with but does not uniquely prove out-of-plane alignment.","The authors' own caveat about non-collinear canted configurations at higher fields suggests that field-dependent studies above 100 mT could map the anisotropy landscape of the twisted interface and reveal a field-driven reorientation.","If the c-axis easy axis is confirmed, the 90° twisted region could serve as a perpendicular Néel order parameter at a buried interface, making twisted CrSBr a plausible building block for antiferromagnetic spintronic devices.","The layer-parity imaging result is likely generalizable to other A-type van der Waals antiferromagnets, where the depth-weighted PEEM signal could be used to identify the topmost layer's magnetization and count layers non-destructively."],"forward_implications":["Layer parity is a directly imaged property of CrSBr: monolayer and other odd layer stacks have switchable net magnetization, while even stacks remain compensated up to applied fields of at least ±100 mT.","Graphene encapsulation measurably enhances the Cr XMCD signal by suppressing surface degradation, providing a practical protocol for element-specific studies of other air-sensitive van der Waals magnets.","The 90° twisted interface is magnetically distinct from its constituent monolayers and does not respond to in-plane fields that switch the top layer, indicating that twist angle can be used to engineer magnetic anisotropy in CrSBr heterostructures.","The theoretical prediction of an out-of-plane easy axis in the twisted region gives a concrete, falsifiable target for complementary probes."],"fun_headline_variants":["90° twist flips CrSBr easy axis out of plane","Twisted CrSBr bilayer: magnetism goes vertical","Element-specific imaging shows twist-induced out-of-plane spins","CrSBr: 90° twist reorients magnetic easy axis","Moiré magnetism: twisting CrSBr by 90° sends spins out of plane"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The central claim that the twisted bilayer's easy axis points out of plane rests on the assumption that the azimuth-invariant XLD is dominated by magnetic XMLD rather than crystal-field XNLD, and that the vanishing XMCD does not arise from multi-domain or compensated in-plane order; the authors explicitly leave open a non-collinear canted configuration at higher fields.","fun_headline_variants_meta":{"raw":{"variants":["90° twist flips CrSBr easy axis out of plane","Twisted CrSBr bilayer: magnetism goes vertical","Element-specific imaging shows twist-induced out-of-plane spins","CrSBr: 90° twist reorients magnetic easy axis","Moiré magnetism: twisting CrSBr by 90° sends spins out of plane"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000162,"raw_usage":{"total_tokens":1058,"prompt_tokens":711,"completion_tokens":347,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":455,"completion_tokens_details":{"reasoning_tokens":266}},"tokens_in":455,"tokens_out":347,"duration_ms":3802,"temperature":1.0,"reasoning_tokens":266,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-01T00:52:01.867249+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure the same 90° twisted CrSBr bilayer with an X-ray geometry sensitive to out-of-plane moments—for example, normal-incidence XMCD at the Cr L3 edge—and check whether a perpendicular contrast appears that is absent at grazing incidence; equivalently, measure XLD as a function of azimuth on an unencapsulated twisted sample to disentangle XNLD from XMLD, since the c-axis conclusion requires the magnetic part to dominate the azimuthal invariance.","supporting_citations":[],"review_version":1}