{"id":"7caac43c-c409-43ea-aad4-86be529993f1","arxiv_id":"2505.11728","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"Wide-field NV magnetometry shows room-temperature magnetic stray fields localized at the edges of WS2 flakes, with an inferred edge magnetization tilted slightly out of the flake plane.","lead":"Researchers used nitrogen-vacancy diamond sensors to image magnetic fields around thin flakes of the semiconductor WS2 and found fields concentrated at the flake edges. The edge fields grow with an applied magnetic field, suggesting a weak edge magnetic response with potential for tiny spintronic devices.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The claimed 'slightly tilted' edge magnetization (Model ZC) is supported only by an unreported fit parameter: no canting angle, uncertainty, or quantitative model comparison is given, so the orientation claim is not established.","rationale":"I read the paper in good faith and find the core experimental observation plausible: edge-localized stray fields that grow with applied field, with opposite signs on different edges, appear in two diamond orientations and in more than one flake. The reader's identified weakest assumption, that enhanced NV fluorescence at hBN-capped edges could masquerade as magnetic signal, is a real concern for Flake 3, but the central edge-magnetism claim rests primarily on Flakes 1 and 2, which were not hBN-capped. The more load-bearing gap is the unpublished canting angle used to claim Model ZC over Model Z. This is an internal support gap, not an external disagreement: the paper introduces a fitting parameter, uses it to claim a tilted magnetization axis, and never reports the fitted value or its uncertainty. That makes the abstract's quantitative orientation claim unfalsifiable from the presented data. The concern is addressable by reporting the fit and model comparison, so it does not warrant rejection; the reader's CONDITIONAL verdict remains appropriate. My read therefore leaves the verdict unchanged, while shifting the emphasis from the optical-artifact concern to the unquantified tilt parameter.","tokens_in":16351,"tokens_out":10426,"duration_ms":123343,"concrete_test":"Re-analyze the line-cut profiles in Figure 4 for Flakes 1 and 2 using the same COMSOL forward model described in Section S4, fitting the magnetization orientation angle theta_c in the yz plane while holding all other parameters fixed. Report the best-fit theta_c, its 95% confidence interval (from residual bootstrap or covariance), and the delta-chi-squared or delta-AIC relative to Model Z (theta_c = 0). If the confidence interval includes 0 degrees or the evidence ratio is weak (e.g., delta-AIC < 2), the 'slightly tilted' conclusion should be downgraded to an unproven hypothesis.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The direct observation of edge-localized Bstr maps that grow approximately linearly with Bapp and reverse sign across edges is credible and supported by two diamond orientations. The load-bearing weakness is the quantitative interpretation of the tilt. In Section 2.1 and Figure 4, the authors state that Models P, PE, and N 'significantly depart' from the data and that Model ZC best reproduces the profiles, with the magnetization angle in the yz plane introduced as a fitting parameter. However, the fitted angle is never reported, no confidence interval is given, and no statistical or residual comparison (e.g., chi-squared, AIC) is shown between Model Z (angle fixed at 0 degrees) and Model ZC. Because ZC adds a free parameter to address 'minor inconsistencies', such as the slightly asymmetric dispersive profile in Figure 4b, it will trivially improve the fit even if the true canting is zero. The abstract and conclusion nevertheless assert a 'slightly tilted' axis and 'spin canting' based on this unquantified fit. If the best-fit angle is consistent with 0 degrees or the improvement is insignificant, the tilt claim collapses; the remaining evidence would still support edge-localized moments but not the specific canting interpretation. The manuscript itself flags this weakness only vaguely in the Conclusion ('the exact nature of the observed tilt in the ZC model remains unclear') without supplying the missing number.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript reports room-temperature wide-field nitrogen-vacancy (NV) magnetometry of exfoliated WS2 flakes (pristine and Fe-implanted, 45–160 nm thick) transferred onto diamond substrates. The authors observe stray magnetic fields localized at flake edges, with amplitudes up to ±4.7 µT that scale approximately linearly with an applied field of 4.4–220 mT. They compare their measured field profiles with finite-element simulations for five magnetization models and conclude that the data favor edge-localized magnetization tilted slightly from the flake normal (Model ZC), which they interpret as spin canting in antiferromagnetically coupled edge states. The paper also reports that Fe-implanted flakes show edge magnetism similar to pristine flakes, with no evidence of uniform Fe-induced magnetization.","tokens_in":16680,"tokens_out":4366,"duration_ms":48487,"significance":"If the results hold, the paper provides a direct, spatially resolved observation of edge-localized magnetic stray fields in a non-van der Waals magnetic TMD at room temperature, using a technique capable of reaching the µT scale. The measured edge-localized signal is independent of the magnetization model, and the sign reversal across edges plus the growth with applied field are internally consistent and are shown for two different diamond orientations and for pristine and Fe-implanted flakes. These are genuine strengths. However, the headline orientation claim—'slightly tilted' edge magnetization—rests on a fitted tilt angle that is never reported, and the model comparison is qualitative. The optical artifact concern for the hBN-capped flake also deserves a concrete control. The central observation is likely salvageable, but the interpretation as spin canting needs substantially stronger quantitative support before publication.","major_comments":[{"comment":"The claim that the edge magnetization is tilted from the flake normal is not quantitatively supported. The manuscript introduces the angle between the magnetization and the z-axis in the yz plane as a fitting parameter for Model ZC, but the best-fit value, its uncertainty, and the improvement over Model Z (angle fixed at 0°) are never reported. Because ZC contains Z as a special case and adds a parameter while addressing 'minor inconsistencies,' the better visual agreement cannot by itself establish a nonzero canting angle. Please report the fitted tilt angles and confidence intervals for Flakes 1, 2, and the Fe-implanted flake, and provide a quantitative model comparison, e.g., reduced χ² or AIC, across all five models. Without these numbers the abstract's 'slightly tilted' and 'spin canting' conclusions are premature; the direct edge-localized-field observation would remain valid even if the tilt turns out to be statistically indistinguishable from zero.","section":"§2.1, Figure 4 (Model ZC)"},{"comment":"The magnetic interpretation of the ODMR frequency shifts assumes that edge-enhanced fluorescence does not bias the amplitude-weighted dip analysis. For Flake 3 the paper reports roughly twice the fluorescence at WS2/hBN edges (Figure S3.3a), attributed to other quantum emitters or waveguiding effects. If the fluorescence modulation changes the ODMR contrast or lineshape, the Bstr maps computed from Eq. (1) could partially reflect an optical artifact rather than a magnetic stray field. Please provide a control test—for example, comparing ODMR contrast and linewidth on and off the edges, or fitting with a fixed-contrast model—to demonstrate that the edge-localized signal for Flake 3 is magnetic in origin. This is especially important because Flake 3 is used for the thickness-dependence claim, though the core edge-magnetism observation also relies on Flakes 1 and 2.","section":"§2.1, Eq. (1), and SI Figure S3.3"},{"comment":"The model interpretation depends on several ad hoc assumptions: a uniformly magnetized edge bar of 200 nm × 300 nm cross-section, a Gaussian convolution width of 325 nm, the NV sensing depth, and a separate fitted volume magnetization for each flake. The authors state that the bar cross-section has negligible influence and that the consistent signal width supports the sub-resolution assumption, but no sensitivity analysis is shown. Please include a table of all model parameters (bar dimensions, standoff, convolution width, fitted magnetization and tilt for each flake) and a brief robustness check showing that the inferred magnetization orientation, particularly the canting angle, is stable to these choices. Without this, the fitted tilt could be an artifact of the assumed geometry rather than a physical property of the edges.","section":"§2.1 and SI S4 (COMSOL model)"}],"minor_comments":[{"comment":"The caption for Figure 3(c,f) describes Model ZC as assuming 'edge magnetization along the z-axis normal to the image plane,' which contradicts the definition of ZC as a canted model. Please correct the caption to reflect that ZC allows a tilt in the yz plane.","section":"Figure 3 caption"},{"comment":"The conclusion states that 'the exact nature of the observed tilt in the ZC model remains unclear,' which undercuts the abstract's stronger statement that the tilt is 'consistent with spin canting.' Please reconcile these statements, either by tempering the abstract or by providing the quantitative analysis that supports the canting interpretation.","section":"Conclusion"},{"comment":"There are a few typographical errors in the supporting information: 'struggle' should be 'straggle' in the SRIM range discussion, and 'Falke 3' appears instead of 'Flake 3.' These should be corrected.","section":"SI S1"},{"comment":"The sign convention in Eq. (1), specifically the use of 'plus' for Bapp > 102.5 mT near the ground-state level anti-crossing, is described only briefly. Please state explicitly how the sign of Bapp is assigned in the two regimes and how the ambiguity near the anti-crossing is handled.","section":"Methods, Eq. (1)"}],"recommendation":"major_revision","confidential_remarks":null},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: the central observation in this paper is real and worth publishing. You get wide-field NV magnetometry maps showing stray fields localized at the edges of WS2 flakes, linear in applied field up to ±4.7 µT, with opposite signs on opposite edges, reproduced on two different diamond orientations and several flakes. As far as the cited literature goes, this is the first direct local imaging of edge magnetization in WS2, and it lines up reasonably with prior bulk magnetometry and theory. Good credit also for the five-model comparison and for testing a Fe-implanted flake, which strengthens the edge-origin story.\n\nWhere the paper goes soft is the 'slightly tilted' magnetization. Model ZC introduces a canting angle in the yz plane as a fitting parameter, and the text says it best reproduces the profiles. But the fitted angle is never given, no confidence interval is reported, and there is no quantitative comparison between Model Z and Model ZC. A free parameter will trivially improve a fit, so the claim that the magnetization is 'slightly tilted'—and the abstract/conclusion language about 'spin canting'—is not supported by what is shown. The manuscript itself admits the tilt's nature is unclear, which makes the missing number more frustrating. This is fixable: report the best-fit angle and uncertainty, or soften the claim. I would also flag the 'ferromagnetism' label in the conclusion; the data are hysteresis-free and linear, so 'weakly ferromagnetic or paramagnetic' is the more honest descriptor.\n\nThe hBN fluorescence artifact for Flake 3 is a real but secondary concern. Since the edge signal shows up in uncapped flakes as well, it does not undermine the core observation, but it does confound the thickness comparison for that flake. A control (e.g., measuring an hBN-only edge or using a different analysis) would address it, but this is not a load-bearing flaw.\n\nOverall, this is a solid experimental paper for the 2D magnetism and NV magnetometry communities. The main observation deserves publication; the tilt claim needs either numbers or a rewrite. I would send it to peer review and recommend conditional acceptance, with the fit-parameter reporting and language adjustments as the main revision items.","headline":"Edge-localized stray-field imaging in WS2 is a credible first, but the spin-canting tilt claim rests on a fitted angle the paper never reports; that needs fixing before the orientation conclusion can stand.","tokens_in":17246,"tokens_out":2614,"would_cite":true,"duration_ms":25677,"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":"Room-temperature edge-localized magnetization in WS2 flakes is imaged directly with nitrogen-vacancy magnetometry, with stray fields up to ±4.7 µT that scale linearly with applied field and are best explained by a slightly tilted edge…","keywords":["nitrogen-vacancy magnetometry","WS2","transition metal dichalcogenides","edge magnetism","stray field imaging","spin canting","2D spintronics","optically detected magnetic resonance"],"falsifier":"A decisive control would be scanning-NV imaging at ~50 nm resolution over the same flakes: a true dipolar edge field should persist with the same sign pattern and profile shape, whereas an optical artifact should track the fluorescence hotspot map (which is brightest at WS2/hBN edges). A second control, imaging a non-magnetic hBN flake on the same diamond under identical conditions, should show no edge-localized stray field; if it does, the extraction is contaminated.","tokens_in":16154,"feed_emoji":"🧲","tokens_out":14235,"duration_ms":119284,"temperature":0.7,"pith_summary":"This paper reports the first direct imaging of magnetic stray fields localized at the edges of WS2 flakes at room temperature, using nitrogen-vacancy (NV) centers in diamond as the magnetic sensor. It claims that the edge fields, up to ±4.7 µT, scale linearly with an applied magnetic field from 4.4 to 220 mT, and that comparing five magnetization models favors an edge magnetization slightly tilted from the flake normal, an effect attributed to spin canting in antiferromagnetically coupled edge states. The result matters because it localizes in space the weak ferromagnetism previously seen only in bulk measurements of WS2 nanosheets, and it identifies flake edges, not Fe dopants, as the source of the magnetic response. A sympathetic reader would take this as evidence that WS2 flakes are a viable platform for edge-controlled 2D spintronics.","feed_headline":"WS2 flake edges carry room-temperature magnetism","feed_subtitle":"Diamond sensors map edge stray fields up to 4.7 µT growing linearly with field, pointing to edge-controlled spintronics.","key_machinery":"The central mechanism is the conversion of a local magnetic field into a measurable shift of the NV spin resonance frequencies: $B_{\\rm str} = (f_+ \\pm f_-)/(2\\gamma_{\\rm NV}) - B_{\\rm app}$, with $\\gamma_{\\rm NV}=28$ GHz/T, applied pixel-by-pixel to amplitude-weighted Lorentzian fits of the optically detected magnetic resonance dips. The edge-magnetism identification is carried by a five-model comparison: Model P (paramagnetic flake), Model PE (paramagnetic edges), Model N (moments perpendicular to the edge faces), Model Z (moments along the out-of-plane $z$-axis), and Model ZC (out-of-plane with a fitted canting angle in the $yz$-plane). For each model, finite-element magnetostatic simulations are fit to line cuts of the measured stray field after convolution with a ~325 nm Gaussian kernel representing the diffraction-limited resolution; only the $z$-aligned and canting models reproduce the sign-alternating edge profiles observed on both diamond orientations. The ZC model adds one fitting parameter and resolves the minor asymmetries, which the authors attribute to spin canting in antiferromagnetically coupled edge states.","core_discovery":"On the paper's own terms, the authors establish that exfoliated WS2 flakes (45–160 nm thick) produce stray magnetic fields concentrated at their edges, imaged at room temperature by optically detected magnetic resonance of shallow nitrogen-vacancy centers in diamond. The stray-field amplitude is linear in the applied field between 4.4 and 220 mT, reaching ±4.7 µT at 63.2 mT for the 160-nm flake, and the spatial pattern changes from an 'absorptive' to a 'dispersive' shape depending on diamond orientation, as expected for a dipolar field projected onto the NV axis. Finite-element magnetostatic simulations of five magnetization geometries—whole-flake paramagnetism, edge paramagnetism, edge moments normal to the side faces, out-of-plane edge moments, and out-of-plane edge moments with a fitted canting angle (Model ZC)—single out the last as the best description of the measured profiles on both (100) and (110) diamonds. Because Fe-implanted flakes show the same edge signal without any uniform magnetization across the flake, the authors conclude that the magnetism originates from the edges or intrinsic defects rather than from the implanted ions.","pith_inferences":["Because the canting angle in Model ZC is fitted and the paper admits its physical origin is unclear, a field-angle-resolved study should reveal whether the tilt is fixed by the crystal or follows the applied field direction; without that, spin canting is one of several possible explanations.","If edge magnetization is controlled by the filling of edge electronic states, as the cited theory suggests, gated WS2 devices should show a gate-tunable edge stray field; observing such modulation would be a direct test and a practical switch for edge spintronics.","The ~325 nm diffraction-limited resolution cannot distinguish a true one-dimensional edge spin chain from a wider magnetized strip (the simulations use a 200 nm × 300 nm bar), so a scanning-NV probe is the natural next measurement to locate the magnetization at the atomic edge.","The enhanced fluorescence at WS2/hBN edges suggests that contrast artifacts could mimic or distort magnetic maps; a systematic comparison of stray-field maps with fluorescence amplitude maps across many flakes would quantify how much of the apparent signal is optical."],"forward_implications":["WS2 flakes can serve as room-temperature, edge-defined sources of stray magnetic field, so flake shape and edge chemistry control the magnetic landscape for nearby spins.","Because the edge signal scales linearly with applied field rather than showing hysteresis, practical spintronic devices would need a control field or another way to stabilize the edge moment.","The similar edge signal in pristine and Fe-implanted flakes indicates that doping is not required for edge magnetism, steering future work toward edge termination and defect density.","The weak thickness dependence means even 45-nm-thin flakes give measurable edge fields, making monolayer or few-layer WS2 a plausible next target for this technique."],"supporting_citations":[{"why":"Bulk magnetometry that first indicated room-temperature ferromagnetism in WS2 nanosheets, the phenomenon this paper localizes to edges.","marker":"[26]"},{"why":"SQUID data showing ferromagnetic-like hysteresis in WS2 nanosheets but not bulk, supporting an edge/defect origin.","marker":"[27]"},{"why":"Ab initio prediction of spin-polarized zigzag edge states in WS2, the theoretical target the edge-magnetization models build on.","marker":"[28]"},{"why":"Comparison of Fe-doped MoS2 and Fe-doped WS2 showing no Fe-induced ferromagnetism in WS2, which supports the absence of doping-driven magnetization here.","marker":"[29]"},{"why":"Supplies the NV optically detected magnetic resonance imaging scheme and Equation (1) connecting resonance frequencies to stray field.","marker":"[30]"},{"why":"Establishes the wide-field NV magnetometry approach and the diamond NV-layer samples used to image the flakes.","marker":"[31]"},{"why":"Provides the amplitude-weighted mean method for evaluating ODMR dips at each pixel, the fitting step on which the magnetic interpretation depends.","marker":"[59]"}],"fun_headline_variants":["Room-temp edge magnetism in WS2 mapped by diamond sensors","WS2 edges show stray fields up to 4.7 μT at room temp","Diamond sensor spots edge magnetism in WS2 flakes","Edge-localized magnetic fields in WS2 imaged at room temp","NV magnetometry reveals edge magnetism in WS2 flakes"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the edge-localized shifts in the nitrogen-vacancy resonance are magnetic in origin; if the two-fold higher fluorescence at WS2/hBN edges biases the resonance-dip fitting or changes its contrast, the apparent stray-field maps could be partly optical artifacts rather than real magnetic fields.","fun_headline_variants_meta":{"raw":{"variants":["Room-temp edge magnetism in WS2 mapped by diamond sensors","WS2 edges show stray fields up to 4.7 μT at room temp","Diamond sensor spots edge magnetism in WS2 flakes","Edge-localized magnetic fields in WS2 imaged at room temp","NV magnetometry reveals edge magnetism in WS2 flakes"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000706,"raw_usage":{"total_tokens":3210,"prompt_tokens":1003,"completion_tokens":2207,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":619,"completion_tokens_details":{"reasoning_tokens":2119}},"tokens_in":619,"tokens_out":2207,"duration_ms":13617,"temperature":1.0,"reasoning_tokens":2119,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T20:48:23.410349+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A decisive control would be scanning-NV imaging at ~50 nm resolution over the same flakes: a true dipolar edge field should persist with the same sign pattern and profile shape, whereas an optical artifact should track the fluorescence hotspot map (which is brightest at WS2/hBN edges). A second control, imaging a non-magnetic hBN flake on the same diamond under identical conditions, should show no edge-localized stray field; if it does, the extraction is contaminated.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Bulk magnetometry that first indicated room-temperature ferromagnetism in WS2 nanosheets, the phenomenon this paper localizes to edges."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"SQUID data showing ferromagnetic-like hysteresis in WS2 nanosheets but not bulk, supporting an edge/defect origin."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Ab initio prediction of spin-polarized zigzag edge states in WS2, the theoretical target the edge-magnetization models build on."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Comparison of Fe-doped MoS2 and Fe-doped WS2 showing no Fe-induced ferromagnetism in WS2, which supports the absence of doping-driven magnetization here."},{"cited_title":"Fescenko, A","cited_arxiv_id":null,"evidence_quote":"Supplies the NV optically detected magnetic resonance imaging scheme and Equation (1) connecting resonance frequencies to stray field."},{"cited_title":"Lamichhane, K","cited_arxiv_id":null,"evidence_quote":"Establishes the wide-field NV magnetometry approach and the diamond NV-layer samples used to image the flakes."},{"cited_title":"Ultra-fast detection of the center frequency of a spectral line from amplitude-weighted average","cited_arxiv_id":"2303.12544","evidence_quote":"Provides the amplitude-weighted mean method for evaluating ODMR dips at each pixel, the fitting step on which the magnetic interpretation depends."}],"review_version":1}