{"id":"2c28441e-2bac-49d5-88b9-0574ab4ab162","arxiv_id":"2501.04777","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"Scanning tunneling spectroscopy resolves Hofstadter subbands and self-similar gap structure in twisted bilayer graphene at low magnetic fields.","lead":"Researchers used a scanning tunneling microscope to directly map the fractal Hofstadter energy spectrum in twisted bilayer graphene near its second magic angle. The data show flat moiré bands splitting into discrete subbands at rational magnetic flux, with signatures of self-similarity and interaction-driven changes.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The gate-voltage-to-filling calibration is unverified, and every quantitative claim—gap labels, Chern numbers, and the claimed self-similarity collapse—depends on it.","rationale":"I read the paper in good faith as an experimental claim that STS resolves Hofstadter subbands and their gaps in second-magic-angle TBG, with self-similarity signatures. The strongest claim is concrete and testable, and the raw data plus the LDOS model provide real support. The reader's conditional verdict identifies the gate-to-filling calibration as the weakest assumption, and my independent reading reaches the same conclusion: every quantitative and topological label in the paper passes through the Vg-to-ν conversion. The manuscript does not document a capacitance calibration or any independent density check, and the Streda-formula Chern numbers and the self-similarity rescaling are only meaningful if that conversion is accurate and linear. This is not an internal inconsistency, but it is a missing experimental control that could change the central conclusions if it fails. The concrete test I propose—an independent Hall measurement on the same device—would settle whether the concern lands. If the calibration holds, the conditional verdict should be upgraded to acceptance; if it fails, the gap labels and fractal claim would need revision. Since the reader already assigned CONDITIONAL largely on this basis, my stress-test pass does not move the verdict; it sharpens the necessary check.","tokens_in":10437,"tokens_out":4825,"duration_ms":51426,"concrete_test":"Perform an independent density calibration on the same device (or an identically fabricated device) by measuring Hall resistance R_xy and longitudinal resistance R_xx as a function of Vg at fixed B ≈ 1.5–3 T, and extract n(Vg) from the low-field Hall slope. Compare this with the assumed linear mapping and with the Vg positions of the ν = 4N conductance suppressions used in Fig. 3. If the Hall-derived ν(Vg) differs from the assumed mapping by more than ~0.1 filling at any point, or shows a Vg-dependent offset, the gap labels and the self-similarity scaling collapse must be re-derived before the headline claim can stand.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central argument requires that the back-gate voltage Vg maps linearly to the filling ν, with zero offset and with zero-field integer fillings ν = 4N serving as absolute density anchors. This calibration underlies the gap-labeling formula ν = t Φ/Φ0 + s stated in the main text, the Streda-formula Chern number extraction, and the discrete self-similarity rescalings in Fig. 4 (ν → ¼(Φ0/Φ)(ν − νref) + νref). The paper provides no capacitance calibration, no Hall density measurement, no independent compressibility check, and no statement of the dielectric thickness or gate coupling used to convert Vg to n. A small offset or a few-percent nonlinearity in n(Vg) would shift every inferred filling; for example, a density error of δν ≈ 0.2 would turn the integer gap label (4,4) into a non-integer pair, invalidating the topological assignment. The same error would misalign the scaled panels in Fig. 4 even if the raw spectral peaks are genuine. Because the headline claim of direct spectroscopic observation of Hofstadter subbands rests on correct identification of gap fillings, this unverified calibration is the most consequential weak link in the argument. The raw dI/dV maps and the LDOS model comparison are valuable independent evidence, but they do not by themselves determine the absolute density scale.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript reports scanning tunneling spectroscopy of twisted bilayer graphene near the second magic angle in a magnetic field, claiming the first direct spectroscopic observation of the fractal Hofstadter spectrum in a moiré material. At rational flux Φ/Φ0 = 1/q, the authors observe the fractionalization of flat moiré bands into q Hofstadter subbands, label spectroscopic gaps by topological invariants (t, s) via Streda's formula, and show that energy-versus-density maps near ν = ±4 at different rational fluxes can be collapsed onto each other by a discrete linear scaling transformation. The paper also reports interaction-driven modifications of the Hofstadter spectrum, including vanishing gaps and density-dependent subband splitting. The central experimental evidence is a set of dI/dV(Vs, Vg) maps, supported by LDOS calculations from a Bistritzer–MacDonald-based model with heterostrain and a filling-dependent Hartree potential.","tokens_in":10688,"tokens_out":3264,"duration_ms":35812,"significance":"If the claims hold, this would be a landmark result: the first direct spectroscopic access to Hofstadter subbands and their fractal self-similarity in a moiré system, with the gap labels providing Chern numbers through Streda's formula. The raw data show clear field-dependent subband splitting and nontrivial gap dynamics, and the paper should be credited for making the code available through a Code Ocean capsule, for comparing data to an explicit model, and for reporting an anomalous vanishing gap that the model qualitatively reproduces. However, the absolute density calibration underpinning every quantitative label is not established, and the model comparison depends on an in-preparation companion paper and unspecified interaction parameters. The significance is therefore conditional: the experimental phenomenology appears rich, but the headline claims about topological labels and discrete self-similarity are not yet supported to the standard required.","major_comments":[{"comment":"The mapping from back-gate voltage Vg to filling ν is asserted but never calibrated. The spectral labels and all quantitative claims rely on the formula ν = t Φ/Φ0 + s, on Streda's formula for Chern numbers, and on the self-similarity transformation ν → ¼(Φ0/Φ)(ν − νref) + νref in Fig. 4. The Methods section states only that a p-doped Si back-gate was used; no dielectric thickness, capacitance, Hall density measurement, or compressibility check is given, and the zero-field suppressions at ν = 4N are used as density anchors. An offset or a few-percent nonlinearity in n(Vg) would shift every inferred filling; for example, δν ≈ 0.2 would turn the reported (4,4) gap label into a non-integer pair and invalidate the topological assignment. The paper must provide a capacitance calibration or an independent density measurement, and an uncertainty analysis showing that plausible gate-oxide or offset variations do not change any assigned (t, s).","section":"Spectroscopy of Hofstadter Gaps; Fig. 3"},{"comment":"The LDOS model used to reproduce the anomalous vanishing of the (4,4) gap is not self-contained. The main text specifies only that the model includes heterostrain and a filling-dependent Hartree potential and cites Ref. 33, an in-preparation paper, for the formalism; the functional form of the Hartree term, the screening parameter, and the precise parameter values for Devices A1 and A2 are not given. The statement that calculations without interactions show 'weaker connections' is not quantified. Because this model comparison is the primary evidence that the vanishing gap is a real interaction-enhanced effect, the paper should include the model equations, the parameter values used, and a sensitivity study over the Hartree interaction strength and strain parameters; otherwise the agreement cannot be independently assessed.","section":"Model comparison, Figs. 3b,d"},{"comment":"The discrete self-similarity claim is supported only by visual inspection of the scaled data. No quantitative similarity metric, correlation coefficient, or residual analysis is provided for the collapse in the bottom panels of Figs. 4a,c, and the factor ¼ in the density transformation is not derived in the main text. The violations at non-integer Φ0/Φ (Figs. 4b,d) are likewise identified by eye. Since the same density transformation inherits the uncalibrated ν axis from the previous comment, the collapse could in part reflect a trivial linear rescaling of coordinates. A quantitative measure of feature alignment, with and without the calibrated density axis, is needed to support the self-similarity claim.","section":"Self-Similar Fractal Hofstadter Spectrum; Fig. 4"}],"minor_comments":[{"comment":"The text uses lowercase 'v' in several places where the filling ν is meant (e.g., 'near v = -4' and 'v = -4 V' in the self-similarity section); these should be corrected to ν.","section":"General"},{"comment":"The reported gate voltages for fixed fillings, such as Vg = 16.8 V for ν = +4 and Vg = 3 V for ν = 0, imply a specific zero-bias offset that is never justified; a consistency check between these implied fillings and the uncalibrated Vg axis would help the reader evaluate the density scale.","section":"Fig. 5 caption and related text"},{"comment":"The Code Ocean capsule is mentioned but no capsule identifier or DOI is given, which makes the code availability statement hard to verify; please provide a persistent identifier.","section":"Data and Code Availability"}],"recommendation":"major_revision","confidential_remarks":"The central weakness is the uncalibrated gate-to-density conversion, which is exactly the kind of issue that a careful referee should require to be fixed before publication. In addition, the model comparison relies on Ref. 33, which is not yet public; given that the vanishing-gap claim depends on that model, the editor may wish to insist that the companion paper or a full methods appendix be made available at the time of review. The raw data appear rich and the qualitative phenomenology is compelling, but the topological labels and self-similarity collapse should not be accepted without an explicit density calibration and quantified comparison."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Bottom line: this is the first direct STS look at Hofstadter subbands in a moiré material, and the raw data are good enough to carry the paper. The fractionalization of each flat band into q subbands at Φ/Φ0 = 1/q is visible by eye in the dI/dV maps, and the comparison with the single-particle-plus-Hartree LDOS model reproduces several non-trivial details, including gaps that anomalously vanish near ν = ±4. The discrete self-similarity collapse in Fig. 4 is striking, and the authors are honest that it breaks at non-integer Φ0/Φ, which strengthens the claim.\n\nWhat I would push on before publication is the gate-to-density conversion. Every quantitative label—gap fillings ν = tΦ/Φ0 + s, the Chern numbers via Streda, and the self-similarity scaling ¼(Φ0/Φ)(ν−νref)+νref—assumes a linear, zero-offset relation between Vg and carrier density, with the zero-field integer fillings ν = 4N as absolute anchors. The main text gives no capacitance calibration, no Hall check, no dielectric thickness, no independent density probe. A small offset or few-percent nonlinearity would scramble the topological assignments and misalign the scaled panels while leaving the raw spectral peaks untouched. I think this is a fixable reporting gap—the SI may well contain the calibration—but as written it is the weakest link.\n\nTwo smaller things. The Hartree interaction strength that makes their LDOS model match is not quantified in the main text; I want to know how much it is tuned. And ref. 33, the model used for all the LDOS calculations, is 'in preparation'—I understand the dynamics, but for a paper leaning on that comparison, a preprint or a fuller description in SI would help. The self-similarity test is also visual rather than quantitative; a simple metric on the collapsed vs. raw panels would move it from striking to convincing.\n\nIf the calibration survives scrutiny, this is a significant result for moiré and quantum Hall physics. The observational core—directly resolved Hofstadter subbands and density-dependent spectral evolution—does not depend on the model, only the labels do. I'd send it to a strong referee, with instructions to check Fig. 4's scaling and the Vg conversion. Worth citing once the SI is released.","headline":"First direct STS of Hofstadter subbands in TBG: raw data are strong and the model comparison is convincing, but the unverified gate-to-density calibration underpins every quantitative label.","tokens_in":11273,"tokens_out":2141,"would_cite":true,"duration_ms":20284,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["73.43.-f","73.22.Pr","68.37.Ef"],"model":"deepseek-v4-flash","headline":"Hofstadter's fractal spectrum resolved in twisted bilayer graphene","keywords":["Hofstadter butterfly","twisted bilayer graphene","second magic angle","scanning tunneling spectroscopy","moiré flat bands","self-similarity","Chern numbers","Streda formula"],"falsifier":"A Hall-effect or capacitance measurement on the same device that shows the Vg-to-ν conversion is nonlinear or offset would shift every (t, s) label and break the self-similarity collapse, even if the raw spectral peaks are genuine. Alternatively, counting Hofstadter subbands at a flux where 1/q is not one of the reported values, say q = 7, and finding the number of subbands differs from q would falsify the fractionalization claim.","tokens_in":10231,"feed_emoji":"🦋","tokens_out":5999,"duration_ms":51724,"temperature":0.7,"pith_summary":"This paper reports scanning tunneling spectroscopy of twisted bilayer graphene near the second magic angle in a magnetic field. It claims that at rational magnetic flux per moiré unit cell, Φ/Φ0 = 1/q, each flat moiré band fractionalizes into q discrete Hofstadter subbands, and that the resulting gaps can be labeled by two topological integers through the Streda formula. The authors further find that energy-versus-density maps near fillings ν = ±4 at different rational fluxes are related by a discrete linear scaling transformation, a signature of the spectrum's fractal self-similarity. If correct, this is the first direct spectroscopic observation of the Hofstadter butterfly in a moiré material, and it shows how correlations modify the single-particle fractal spectrum.","feed_headline":"Hofstadter's fractal spectrum resolved in twisted bilayer graphene","feed_subtitle":"At rational flux, each flat band splits into q subbands, with gaps following a scaling law.","key_machinery":"The central object is the Hofstadter subband structure that emerges when a flat moiré band is placed in a perpendicular magnetic field with rational flux per unit cell Φ/Φ0 = 1/q. Each flat band folds into q subbands in a q-times-larger magnetic unit cell, and the gaps between subbands carry the topological labels (t, s), with the filling given by ν = t Φ/Φ0 + s; the integer t is the Chern number read off from the density-versus-field slope via the Streda formula. The argument for self-similarity rests on a discrete scaling transformation—multiplying energy and density by ¼(Φ0/Φ) at integer Φ0/Φ—that maps spectra at different rational fluxes onto each other near ν = ±4. A companion continuum-model LDOS calculation, including heterostrain and a filling-dependent Hartree potential, is used to match the measured dI/dV maps and to explain why certain Hofstadter gaps vanish.","core_discovery":"The paper's central claim is that the flat bands of twisted bilayer graphene near the second magic angle (θ ≈ 0.6°) are an ideal arena in which Hofstadter's butterfly can be resolved directly by tunneling spectroscopy. At rational flux Φ/Φ0 = 1/q, the magnetic unit cell is enlarged q-fold, and each flat moiré band fractionalizes into q Hofstadter subbands; the resulting gaps appear at fillings ν = t Φ/Φ0 + s, where (t, s) are integer topological invariants. The paper further claims that STS maps at Φ/Φ0 = 1/6, 1/5, and 1/4 collapse onto one another when energy and density are rescaled by the linear factor ¼(Φ0/Φ), demonstrating discrete self-similarity near ν = ±4, and that non-integer flux ratios break this scaling. The measurements also show density-dependent changes in the spectrum, such as Hofstadter gaps that appear and disappear with filling, which the authors attribute to electron-electron interactions and model using a Hartree-corrected continuum calculation.","pith_inferences":["If the density-axis calibration (Vg to ν) is accurate, the same (t, s) assignment should predict the quantized Hall conductance of each gap; a transport measurement on the same device would directly test this.","The observed discrete self-similarity suggests that STS energy-density maps could be used as a quantitative local probe of flux density, potentially imaging spatial variations of magnetic field or strain.","Because the Hartree-corrected model reproduces the vanishing gaps, a similar analysis could predict which correlated insulating states survive at finite field in other moiré flat-band systems.","A direct test of the fractionalization claim would be to count Hofstadter subbands at a rational flux not reported here, such as q = 7 at lower field, and verify that the number of subbands equals q."],"forward_implications":["At any rational flux 1/q, each flat moiré band should split into q resolvable subbands, giving a spectroscopic fingerprint for identifying flux ratios in moiré materials.","The gap labels (t, s) assign Chern numbers to each spectroscopic gap, so STS maps can serve as a local, density-resolved probe of topology.","The discrete self-similarity near ν = ±4 means that a single measured spectrum at one rational flux can be rescaled to predict the spectrum at another rational flux, within the single-particle picture.","Correlation effects, captured by a density-dependent Hartree potential, change the Hofstadter spectrum qualitatively, so the fractal structure cannot be fully described by non-interacting models."],"supporting_citations":[{"why":"Supplies the predicted fractal energy spectrum for Bloch electrons in rational magnetic flux, the object under study.","marker":"[1]"},{"why":"Provides the continuum model of twisted bilayer graphene used to predict the flat bands and to compute the Hofstadter spectrum.","marker":"[11]"},{"why":"Predicted multiple flat bands and the topological Hofstadter butterfly near the second magic angle, motivating the experiment.","marker":"[12]"},{"why":"Gives the relation between Hall density and Chern number used to label gaps with (t, s).","marker":"[3]"},{"why":"Defines the bound-charge topological invariant s from the electron count per unit cell in the zero-field limit.","marker":"[31]"},{"why":"Demonstrates the use of the Streda relation in magic-angle twisted bilayer graphene to assign Chern numbers to correlated gaps.","marker":"[32]"},{"why":"Provides the continuum model LDOS calculation method used to simulate the measured dI/dV maps.","marker":"[33]"},{"why":"Shows how strain designs flat bands and is used to include heterostrain effects in the model.","marker":"[34]"}],"fun_headline_variants":["Tunneling spectroscopy resolves Hofstadter's fractal subbands","Hofstadter's butterfly seen in twisted bilayer graphene flat bands","Self-similar Hofstadter spectrum directly imaged in moiré graphene","Moiré electrons display Hofstadter's fractal energy structure","Spectroscopy captures Hofstadter's fractal spectrum in twisted bilayer"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The entire gap-labeling scheme assumes the back-gate voltage converts linearly and knownly to electron density, with the zero-field integer fillings ν = 4N resetting the unbound charge density; the paper gives no independent capacitance or density calibration.","fun_headline_variants_meta":{"raw":{"variants":["Tunneling spectroscopy resolves Hofstadter's fractal subbands","Hofstadter's butterfly seen in twisted bilayer graphene flat bands","Self-similar Hofstadter spectrum directly imaged in moiré graphene","Moiré electrons display Hofstadter's fractal energy structure","Spectroscopy captures Hofstadter's fractal spectrum in twisted bilayer"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001157,"raw_usage":{"total_tokens":4831,"prompt_tokens":1023,"completion_tokens":3808,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":639,"completion_tokens_details":{"reasoning_tokens":3718}},"tokens_in":639,"tokens_out":3808,"duration_ms":23931,"temperature":1.0,"reasoning_tokens":3718,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T21:27:50.274352+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A Hall-effect or capacitance measurement on the same device that shows the Vg-to-ν conversion is nonlinear or offset would shift every (t, s) label and break the self-similarity collapse, even if the raw spectral peaks are genuine. Alternatively, counting Hofstadter subbands at a flux where 1/q is not one of the reported values, say q = 7, and finding the number of subbands differs from q would falsify the fractionalization claim.","supporting_citations":[{"cited_title":"& MacDonald, A","cited_arxiv_id":null,"evidence_quote":"Provides the continuum model of twisted bilayer graphene used to predict the flat bands and to compute the Hofstadter spectrum."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Predicted multiple flat bands and the topological Hofstadter butterfly near the second magic angle, motivating the experiment."},{"cited_title":"Quantised Hall effect in a two-dimensional periodic potential","cited_arxiv_id":null,"evidence_quote":"Gives the relation between Hall density and Chern number used to label gaps with (t, s)."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Defines the bound-charge topological invariant s from the electron count per unit cell in the zero-field limit."},{"cited_title":"Scheer, Jonah Herzog-Arbeitman, Kevin P","cited_arxiv_id":null,"evidence_quote":"Provides the continuum model LDOS calculation method used to simulate the measured dI/dV maps."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Shows how strain designs flat bands and is used to include heterostrain effects in the model."}],"review_version":1}