{"id":"c60736e0-e457-4de7-8362-3cedfaaec1ee","arxiv_id":"2501.04215","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"A single WTe2 device encodes strain magnitude and direction through two independent nonlinear Hall voltage signals that respond differently to strain.","lead":"Researchers show that a tiny device made of the layered material WTe2 can sense both the strength and the direction of a stretch, using two independent electrical signals at once. This could lead to electronic skin that perceives complex touches with a single sensor.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The reported strain magnitudes rely on an unvalidated beam-bending formula; without a local strain measurement, the encoded strain vectors are not quantitatively secure.","rationale":"The paper has real strengths: multi-device reproducibility (Fig. S6), the use of intensive normalized fields to remove longitudinal resistivity effects, and a plausible quantum-geometry scaling analysis. However, the central claim is quantitative: a single device encodes the magnitude and direction of the strain vector. Every magnitude label in the paper and in the ANN training set comes from Eq. 10, which converts the bending height h of the PI substrate into a nominal surface strain. No measurement verifies that this nominal value is actually experienced by the WTe2 crystal. The macroscopic resistance repeatability in Fig. S4g confirms that the setup is stable, not that the strain calibration is accurate. This is precisely the load-bearing assumption: if the local strain differs in magnitude or direction, the ANN outputs cannot be trusted and the reported 0.2%/0.4%/0.51% values are not the applied strain. The proposed Raman test would settle this. The reader's conditional verdict is appropriate: the concept is promising, but the quantitative strain encoding claim needs this validation before acceptance. The simulated NJU demonstration and missing error bars are secondary; they affect the strength of the demonstration, not the core calibration validity.","tokens_in":17434,"tokens_out":10744,"duration_ms":113603,"concrete_test":"Conduct in-situ polarized micro-Raman spectroscopy on the WTe2 device while the PI substrate is bent to the same heights h used for the 0.2%, 0.4%, and 0.51% nominal strains (Fig. S4f). Use the strain-induced shifts of the Ag phonon modes with known deformation potentials of Td-WTe2 to extract the local strain tensor in the flake, including its direction relative to the a-axis. Compare the extracted magnitude and angle with Eq. 10 for at least three bending heights and all three in-plane orientations (0 degrees, 45 degrees, 90 degrees). If the local strain magnitude deviates by more than about 20% from the nominal value or the direction by more than about 10 degrees, the reported strain-vector calibration and the ANN-encoded strain maps are not quantitatively reliable; a corrected calibration would be required before the central claim can be accepted.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim that a single WTe2 device encodes both magnitude and direction of a strain vector depends on the assumption that the nominal strain epsilon from Eq. 10 equals the actual local strain tensor in the WTe2 flake. Eq. 10 is a global beam-bending formula for the top surface of the PI substrate; it assumes rigid, slip-free, uniform strain transfer through the van der Waals interface, with no shear lag, clamping, or electrode-induced pinning. The only experimental check offered is the macroscopic four-probe resistance repeatability in Fig. S4g, which does not measure the crystal strain. If the local strain is smaller, non-uniform, or rotated relative to the nominal direction, then every reported strain magnitude (0.2%, 0.4%, 0.51%) and the ANN training labels are systematically wrong; the apparent 'independent' responses of the two harmonic channels could then be responses to an uncontrolled strain state rather than to the intended strain vector.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript proposes and experimentally demonstrates a skin-inspired in-sensor encoding scheme in which a single few-layer Td-WTe2 device uses its second-order and third-order nonlinear Hall responses to encode the magnitude and direction of an applied strain vector. The authors measure linear harmonic responses under various strain magnitudes and directions, extract scaling-law coefficients to argue that the strain response originates from Berry curvature and Berry-connection polarizability, and train an artificial neural network to decode the strain vector from the two harmonic signals. They also demonstrate the encoding of the embossed characters 'NJU' as a proof of concept.","tokens_in":17556,"tokens_out":7067,"duration_ms":66799,"significance":"If the strain calibration and error analysis are strengthened, this would be a notable advance: it uses intrinsic quantum geometric responses in a topological semimetal for multi-parameter tactile sensing, offering a path beyond conventional piezoresistive strain sensors. The experiment is direct and reproducible across several devices (Fig. S6), and the proof-of-concept ANN decoding, despite being trained on interpolated data, demonstrates the encoding concept. The machine-checked data and the availability of multiple device results are strengths.","major_comments":[{"comment":"The strain magnitude applied to the WTe2 flake is derived from a global beam-bending formula for the polyimide substrate, assuming perfect, slip-free, uniform strain transfer across the van der Waals interface. No local strain measurement on the device (e.g., Raman or X-ray) is reported; Fig. S4g only demonstrates repeatability of the four-probe resistance under cycling, which does not calibrate the absolute strain. If shear lag, clamping, or electrode pinning causes the local strain to differ from the nominal value, then all reported strain magnitudes (0.2%, 0.4%, 0.51%) and the ANN target labels are systematically wrong. The authors should provide a local strain calibration or explicitly quantify the uncertainty in the strain transfer.","section":"Section 4, Eq. (10)"},{"comment":"These central figures present the extracted second- and third-order NLH signals as functions of strain magnitude and temperature without any error bars, although the raw voltage signals in Fig. S9 do include error bars. The claim that the two channels respond 'independently' and 'robustly' to strain vectors is based on visual inspection of these curves. Please provide error bars on the extracted intensive quantities (propagated from at least forward/backward scans or repeated devices) and, if possible, a quantitative test of independence (e.g., comparing the strain-response slopes).","section":"Fig. 2h,i and Fig. 3a-d"},{"comment":"The scaling laws E_xy^{2ω}/E_xx^2 = ξ_{2ω} σ^2 + η_{2ω} and E_xy^{3ω}/E_xx^3 = ξ_{3ω} σ^2 + η_{3ω} are fitted, and the coefficients η_{2ω} and η_{3ω} are identified as 'intrinsic' and interpreted as evidence that the strain responses 'primarily stem from variations in the distribution of BC and BPT'. However, this attribution is not directly tested: the same scaling-law fit could be consistent with strain-induced changes in the Fermi surface or band structure that affect other intrinsic contributions. The model in Fig. S1 uses generic parameters (ω=0.05 eV, v=0.1 eV·Å, Δ=0.1 eV, β/a=1.1) and yields only qualitative distributions, not a quantitative comparison with the experimental η values. Please provide a concrete, falsifiable prediction (e.g., a predicted ratio η_{2ω}(θ)/η_{2ω}(0) as a function of strain angle, with parameter ranges) and compare it with the measured values.","section":"Section 2.3"},{"comment":"The ANN is trained on 4290 samples generated by cubic interpolation from 39 measured nonlinear Hall voltage signals, and the validation set is 21 experimental data. This raises the risk that the reported decoding accuracy is inflated by the interpolation. The five strain combinations used in the 'NJU' demonstration are exactly among the training values, so the demonstration does not test generalization to unseen strain states. Please report the ANN error on held-out experimental data without interpolation, and clarify the role of interpolation in the reported accuracy.","section":"Section 4, 'Construction of ANN'"}],"minor_comments":[{"comment":"The tensor notation is inconsistent: the main text defines the measured response as j_y^{2ω}=χ_yxx^{2ω} E_x E_x, but Eq. (7) gives χ_xyy^(2) = -(τ/2)∫ f0 ∂_y Ω_z, which is a different tensor component. Please unify the notation and explain how the symmetry argument for χ_xyy applies to the measured χ_yxx.","section":"Main text and Eq. (7)"},{"comment":"In the ANN description, the sigmoid activation function for the output layer would constrain outputs to [0,1]; since the output includes an angle θ (which should be in [0°,360°) or [0,π)), please specify how the output layer encodes the angle.","section":"Section 4, 'Construction of ANN'"},{"comment":"The Supporting Information table with the scaling-law coefficients is labeled 'Table 1' but is referred to as Table S1 in the main text; please correct the cross-reference.","section":"Supporting Information, Table 1"},{"comment":"The model parameters in Eq. (3) are presented without discussion of their relevance to Td-WTe2; please add a sentence justifying the parameter choices or note that the model is illustrative only.","section":"Section 4, '2D tilted Dirac cone model'"},{"comment":"The sentence 'the strain responses of the second-order and third-order NLH signals primarily stem from the variations in the distribution of BC and BPT modulated by strain' appears before the scaling-law analysis is fully described; consider moving it after the discussion of Table S1.","section":"Section 2.3"}],"recommendation":"major_revision","confidential_remarks":"This is a well-executed experiment with a creative concept, but the quantitative strain calibration and the missing error bars are the main risks to the central claim. The editor may wish to ask the authors for a local strain measurement or, at minimum, an explicit uncertainty analysis of the strain transfer. The mechanism attribution is plausible but not yet decisive; a quantitative comparison between the model and the experimental scaling-law coefficients would significantly strengthen the paper."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: the genuinely new thing here is using two independent nonlinear Hall channels in a single WTe2 device to encode both magnitude and direction of strain. That is a useful idea and the data are consistent with it. The strain-tuning of the second-order NLH in WTe2 was already reported by Ye et al. (ref 46) — the paper cites it but only for skew scattering — so the novelty is incremental, not the 'first time' claimed in the ToC.\n\nWhat the paper does well: clean linear harmonic responses, repeatable across several devices, and the two harmonics show visibly different strain-angle dependences — enough to justify the claim of independent channels. The temperature dependence and the scaling-law analysis are standard and handled adequately; they do support the intrinsic (BC/BPT) contribution being the one that varies with strain. The model calculation is simple but useful for motivating the effect. I also note the paper itself includes a limitation: the NJU demonstration is called 'simulation results' in Section 4 yet the abstract presents it as demonstrated; that needs to be aligned.\n\nSoft spots: first, the strain magnitude is computed from the beam-bending formula Eq. 10, with no independent measurement of the local strain in the flake. Slip or shear lag would scale all the reported strain values and could mix the angle labels. That does not kill the proof-of-concept, but it does mean the quoted 0.2%/0.4%/0.51% values and the ANN training labels are only as good as the transfer assumption. Second, the main strain-response figures (2h,i; 3a,b) lack error bars; the reader has to trust the reproducibility claims from Fig S6 alone. Third, the scanning of 'NJU' is not a physical scan; it is a simulation using measured harmonic signals. That is fine as an extrapolation, but should be labeled as such in the title/abstract.\n\nThe circularity concern is mild: the eta coefficients are extracted from scaling-law fits and then interpreted as intrinsic BC/BPT contributions. That is how the field usually does it, and the fitted values are consistent with prior reports, so I do not see it as a fatal flaw.\n\nWho is it for: people working on nonlinear Hall effects, strain engineering of 2D materials, and flexible quantum-material sensors. It deserves a serious referee: send it out with a request for error bars, a local strain check or explicit caveat, and a clearer statement of prior work.","headline":"Plausible proof-of-concept that two nonlinear Hall channels in WTe2 can encode strain vectors, but the quantitative strain calibration and the 'NJU' demo are weaker than the abstract implies.","tokens_in":18183,"tokens_out":1793,"would_cite":false,"duration_ms":17554,"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":"This paper claims that a single few-layer Td-WTe2 device can encode a strain vector's magnitude and direction by reading two independent nonlinear Hall signals governed by quantum geometry.","keywords":["quantum geometry","nonlinear Hall effect","strain engineering","Td-WTe2","Berry curvature","Berry-connection polarizability tensor","in-sensor encoding","flexible electronics"],"falsifier":"Measure the strain directly on the WTe2 flake during bending—for example by tracking a Raman-active mode of WTe2 or by performing micro-X-ray diffraction over the channel—and compare it with Eq. (10). If the flake's actual strain differs from the bending-height estimate, or if the extracted scaling-law intercepts $\\eta^{2\\omega}$ and $\\eta^{3\\omega}$ do not follow the strain-angle pattern predicted by the strained tilted-Dirac model, the central claim that quantum geometry encodes the strain vector would be disproved.","tokens_in":17204,"feed_emoji":"🖐️","tokens_out":7603,"duration_ms":68576,"temperature":0.7,"pith_summary":"Human skin encodes both the size and the direction of a deformation in a single touch; this paper claims that a single electronic device can do the same. The device is a few-layer flake of the topological semimetal Td-WTe2 on a flexible polyimide strip. When the strip is bent, strain changes the momentum-space distribution of two quantum-geometric quantities—the Berry curvature and the Berry-connection polarizability tensor—and these changes are read out as the second-order and third-order nonlinear Hall voltages. Because the two harmonic responses move independently with strain angle and magnitude, one device can simultaneously sense and encode the full strain vector, which ordinary piezoresistive sensors cannot do. A neural network trained on the two harmonic voltages reconstructs the strain pattern of an embossed character, demonstrating in-sensor encoding.","feed_headline":"One WTe2 flake encodes strain magnitude and direction together","feed_subtitle":"Reading two nonlinear Hall harmonics at once gives one sensor the two independent signals it needs to mimic touch.","key_machinery":"The machinery is the strain-tunable quantum geometry of the electronic wave functions, probed by nonlinear Hall conductivities. The second-order signal is governed by the Berry-curvature dipole $\\partial_x\\Omega_z$; the third-order signal is governed by the Berry-connection polarizability tensor $G_{ab}(\\mathbf{k})=2\\sum_{n\\neq0}\\mathrm{Re}[(\\mathcal{A}_a)_{0n}(\\mathcal{A}_b)_{n0}]/(\\varepsilon_0-\\varepsilon_n)$, through terms like $\\partial_x^2 G_{xy}$. Strain enters a tilted-Dirac model by deforming reciprocal space as $k'=(1-\\mathcal{E}^T)k$ and by adding a pseudogauge field from electron-phonon coupling, which redistributes $\\Omega_z$ and $G$ while preserving their independence; the measured harmonic voltages then serve as two independent readout channels for the strain vector.","core_discovery":"The central claim is that strain vectors can be encoded inside the sensor itself: in Td-WTe2, the strain dependence of the second-order nonlinear Hall conductivity $\\chi^{2\\omega}_{yxx}\\propto -\\int_k f_0\\,\\partial_x\\Omega_z$ and that of the third-order channel $\\chi^{3\\omega}_{yxxx}\\propto -\\int_k f_0\\,\\partial_x^2 G_{xy}+\\frac{1}{2}\\int_k \\partial_\\varepsilon^2 f_0\\, v_x v_y G_{xx}$ are independent, so the pair $(V^{2\\omega}_{xy},V^{3\\omega}_{xy})$ carries both the magnitude and direction of the applied strain. The authors establish that these strain responses are robust across devices and that a scaling-law analysis $E^{2\\omega}_{xy}/E_{xx}^2=\\xi^{2\\omega}\\sigma^2+\\eta^{2\\omega}$, $E^{3\\omega}_{xy}/E_{xx}^3=\\xi^{3\\omega}\\sigma^2+\\eta^{3\\omega}$ shows the strain mainly changes the intercepts $\\eta$ associated with the intrinsic Berry-curvature and BPT distributions rather than the scattering slopes $\\xi$. On this basis a single device, read at the second and third harmonics, is trained through an artificial neural network to output strain magnitude and angle, and the outputs reproduce the strain map of an embossed three-letter pattern.","pith_inferences":["Extension: the same two-channel readout could be applied to other quantum materials with large Berry curvature and BPT, but the paper only demonstrates Td-WTe2.","Extension: the ANN was trained on interpolated data from 39 measured signals; a deployed sensor would need to handle hysteresis, drift, and temperature variation, which the paper does not quantify outside the measured range.","Extension: a direct local strain measurement on the flake, rather than the substrate bending formula, would test whether the reported strain magnitudes are quantitatively correct; this is an implicit next step."],"forward_implications":["A single WTe2 device can replace arrays of unidirectional strain gauges for tasks that need both angle and magnitude, because the two harmonic channels respond independently.","The encoding is intrinsic to quantum geometry, not a piezoresistive artifact: the scaling-law intercepts, tied to Berry curvature and BPT, carry most of the strain response.","Because the response survives over a temperature range up to at least 260 K, the sensing scheme is not restricted to cryogenic operation.","The independently varying second- and third-harmonic channels provide a natural two-dimensional code for a strain vector, which the trained ANN decodes with accuracy above 90%.","The nonlinear Hall channels respond to strain in all tested directions (0°, 45°, 90°), whereas the fundamental longitudinal voltage is sensitive mainly near 90°, so the harmonic readout extends directional coverage."],"supporting_citations":[{"why":"reports the nonlinear Hall effect from Berry curvature in WTe2, providing the experimental foundation for using harmonic Hall channels as geometric probes.","marker":"[43]"},{"why":"supplies the normalized second-order nonlinear Hall measurement and scaling-law analysis used to separate intrinsic and scattering contributions.","marker":"[44]"},{"why":"establishes the third-order nonlinear Hall channel and its relation to the Berry-connection polarizability tensor.","marker":"[45]"},{"why":"derives the Berry-curvature-dipole formula that underlies the second-order nonlinear Hall conductivity.","marker":"[49]"},{"why":"defines the Berry-connection polarizability tensor used in the third-order response.","marker":"[51]"},{"why":"connects the third-order nonlinear Hall effect to BPT in WTe2-type semimetals.","marker":"[52]"},{"why":"provides the scaling-law relation between nonlinear Hall coefficients and conductivity used in the temperature analysis.","marker":"[54]"},{"why":"gives the strain-induced gauge-field coupling that the model uses to describe how strain changes the band geometry.","marker":"[56]"},{"why":"supplies the form of the strain gauge field and its effect on the Dirac Hamiltonian.","marker":"[57]"},{"why":"describes the homemade strain setup and the geometric relation used to convert bending height to strain magnitude.","marker":"[58]"}],"fun_headline_variants":["Strain vector from one WTe2 flake via two nonlinear Hall harmonics","Single WTe2 device encodes strain magnitude and direction via nonlinear Hall","Quantum geometry enables strain-vector encoding in a single WTe2 sensor","Two nonlinear Hall harmonics from one flake give full strain vector","Skin-like strain sensing with quantum geometry in a single topological semimetal"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing assumption is that the strain computed from the bending height of the polyimide substrate is exactly the uniform strain felt by the WTe2 flake; if the flake slips, clamps unevenly, or sees inhomogeneous strain, the reported strain magnitudes and the mapping learned by the ANN would not describe the device.","fun_headline_variants_meta":{"raw":{"variants":["Strain vector from one WTe2 flake via two nonlinear Hall harmonics","Single WTe2 device encodes strain magnitude and direction via nonlinear Hall","Quantum geometry enables strain-vector encoding in a single WTe2 sensor","Two nonlinear Hall harmonics from one flake give full strain vector","Skin-like strain sensing with quantum geometry in a single topological semimetal"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000833,"raw_usage":{"total_tokens":3683,"prompt_tokens":1041,"completion_tokens":2642,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":657,"completion_tokens_details":{"reasoning_tokens":2548}},"tokens_in":657,"tokens_out":2642,"duration_ms":17478,"temperature":1.0,"reasoning_tokens":2548,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T21:39:07.624816+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure the strain directly on the WTe2 flake during bending—for example by tracking a Raman-active mode of WTe2 or by performing micro-X-ray diffraction over the channel—and compare it with Eq. (10). If the flake's actual strain differs from the bending-height estimate, or if the extracted scaling-law intercepts $\\eta^{2\\omega}$ and $\\eta^{3\\omega}$ do not follow the strain-angle pattern predicted by the strained tilted-Dirac model, the central claim that quantum geometry encodes the strain vector would be disproved.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"connects the third-order nonlinear Hall effect to BPT in WTe2-type semimetals."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"reports the nonlinear Hall effect from Berry curvature in WTe2, providing the experimental foundation for using harmonic Hall channels as geometric probes."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"supplies the normalized second-order nonlinear Hall measurement and scaling-law analysis used to separate intrinsic and scattering contributions."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"establishes the third-order nonlinear Hall channel and its relation to the Berry-connection polarizability tensor."},{"cited_title":"Sodemann, L","cited_arxiv_id":null,"evidence_quote":"derives the Berry-curvature-dipole formula that underlies the second-order nonlinear Hall conductivity."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"defines the Berry-connection polarizability tensor used in the third-order response."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"provides the scaling-law relation between nonlinear Hall coefficients and conductivity used in the temperature analysis."},{"cited_title":"Suzuura, T","cited_arxiv_id":null,"evidence_quote":"gives the strain-induced gauge-field coupling that the model uses to describe how strain changes the band geometry."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"supplies the form of the strain gauge field and its effect on the Dirac Hamiltonian."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"describes the homemade strain setup and the geometric relation used to convert bending height to strain magnitude."}],"review_version":1}