{"id":"8f144499-39b4-4e44-bd67-3a5c20e2a193","arxiv_id":"2608.12553","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":13,"one_line_summary":"FIR resonances of MATBG flat bands are observed and attributed to c-electron transitions renormalized by c-f hybridization, with an optical selection rule from a hidden rotational symmetry.","lead":"This paper reports the first far-infrared spectroscopic measurements of flat-band electrons in magic-angle twisted bilayer graphene, using a millikelvin photocurrent platform. It interprets the resonances with the topological heavy-fermion model and derives optical selection rules, opening a new experimental window on moiré materials.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The v=+4 FIR resonances are not shown to be absorption features: the photocurrent/photovoltage detection transfer function is uncalibrated, so m1-m3 could be measurement artifacts.","rationale":"The reader's weakest assumption identifies the same load-bearing concern: the measured photovoltage or photocurrent has not been shown to faithfully represent optical absorption. I agree that this is the most fragile link in the central claim. If the detection transfer function contains spectral structure, then the observed m1-m3 modes, their B-dependence, and even the filling dependence could be instrumental rather than intrinsic MATBG resonances. The stated CONDITIONAL verdict is therefore appropriate, and my analysis does not move it; instead, it sharpens the condition by identifying a concrete calibration experiment that would settle the issue. The secondary concern about fitting the THF parameters to the same spectra is real but does not change the verdict: even if the theoretical interpretation is overparameterized, a genuine experimental observation of resonances would remain a significant result. The paper's internal controls (Vph vs Iph consistency, KBr vs Mylar comparison) reduce but do not eliminate the concern, because those controls share the same uncalibrated detection physics. A reference measurement with a non-moiré bolometer is the decisive check.","tokens_in":14385,"tokens_out":7935,"duration_ms":100478,"concrete_test":"Calibrate the detection transfer function H(ω) by measuring the normalized photovoltage/photocurrent spectrum of a bolometric reference with no moiré flat bands (e.g., a thin graphite resistor or a large-twist-angle graphene device) placed in the identical optical path and under identical bias, temperature, and magnetic-field conditions, then divide the MATBG v=+4 spectra by this reference spectrum. If the m1-m3 peaks survive the division with unchanged positions and B-dependence, they are sample-specific absorption features; if they are reduced or removed, they are artifacts of the photodetection chain rather than intrinsic flat-band resonances.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim assumes that the measured photovoltage/photocurrent interferograms, after division by the 1200 K blackbody spectrum, are proportional to the MATBG optical absorption. This is not established. In the 'Millikelvin FIR Spectroscopy' section, Vph or Iph is recorded at a fixed d.c. bias on a device whose resistance at v=+4 is large and strongly temperature-dependent; the photo-signal is therefore plausibly dominated by bolometric heating or photothermoelectric response, whose spectral transfer function H(ω) is never calibrated. Normalizing by the source spectrum (Extended Data Fig. 3) removes only the source envelope, not H(ω) or optical-path transmission. Each spectrum in Fig. 3 is additionally normalized to its maximum, so any frequency-dependent detection efficiency, e.g., from the hBN phonon, beam-splitter absorption, or substrate/contact interference, can masquerade as resonant modes. Because the same uncalibrated detection chain is used for all fillings, the apparent filling-dependence (Fig. 2e,g) is also consistent with a gate-dependent bolometric sensitivity rather than a change in intrinsic absorption. Unless a resonance-free reference or independent absorption measurement is provided, the assignment of m1-m3 to inter-Landau-level transitions is not uniquely supported.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript reports far-infrared (FIR) photocurrent and photovoltage spectroscopy of magic-angle twisted bilayer graphene (MATBG) at millikelvin temperatures, using a custom-built platform. At filling v=+4, the authors observe magnetic-field-dependent resonances (m1–m3) that they attribute to inter-Landau-level transitions of itinerant c-electrons renormalized by hybridization with localized f-electrons, based on calculations within the topological heavy-fermion (THF) model. They further extract an optical selection rule Δm=±1 that they trace to an angular momentum offset originating from the Γ3 irreducible representation. At charge neutrality, they report low-energy resonances below the on-site Coulomb energy, which they leave largely unexplained. The paper includes transport characterization of three devices and detailed THF simulations of the optical conductivity.","tokens_in":14782,"tokens_out":3358,"duration_ms":35105,"significance":"If the central assignment is correct, this work would constitute the first observation of characteristic FIR resonances of flat-band electrons in MATBG, providing direct spectroscopic access to the c-f hybridization and to the symmetry that organizes the low-energy states. The experimental platform itself, enabling FIR spectroscopy at millikelvin temperatures and in magnetic fields, is a valuable technical advance, and the use of three independent devices with consistent transport and photo-response data is a notable strength. The THF-based calculations are detailed and include parameter-sensitivity studies. However, the current evidence for the central claim is weakened by two load-bearing issues: the uncalibrated photodetection transfer function, and the fact that the THF parameters are fitted to the same spectra that are then presented as 'theoretical predictions.' These issues currently prevent the manuscript from fully establishing the uniqueness of the resonance assignment.","major_comments":[{"comment":"The measured photocurrent/photovoltage is not demonstrated to be proportional to the intrinsic optical absorption of the MATBG flat bands. The spectra in Fig. 3 are normalized to their maximum value at each magnetic field, and the normalization by the blackbody source spectrum (Extended Data Fig. 3) removes only the source envelope, not the frequency-dependent detection efficiency of the bolometric/photothermoelectric response, the beam path, or the contacts. Because the same uncalibrated detection chain is used for all fillings and the device resistance at v=+4 is large and strongly temperature-dependent, the apparent resonances could in principle arise from gate-dependent bolometric sensitivity or from spectral features in the optical path. To support the assignment of m1–m3 to inter-Landau-level transitions, the authors should provide a reference measurement with a known flat spectral response (e.g., a bolometer in the same optical configuration, or a material with known FIR absorption) or otherwise calibrate the detection transfer function.","section":"Millikelvin FIR Spectroscopy (main text) and Fig. 3"},{"comment":"The gray curves labeled 'theoretical prediction' in Fig. 3c are not independent predictions: the THF parameters in Extended Data Table 1 are obtained by best fitting to the same experimental spectra, as stated in the main text ('systematic extraction of interacting parameters by best fitting to experimental spectra'). The agreement therefore reflects a fit, and the statement 'remarkable agreement with the experimental data' overstates the evidential value. To make the comparison meaningful, the authors should either perform an out-of-sample test (e.g., using the D2 parameters to predict the D1 spectra, or predicting the v=-4 spectra or CNP response) or explicitly present the curves as fits and quantify the parameter uncertainties and degeneracies.","section":"Fig. 3c and Extended Data Table 1"},{"comment":"The optical selection rule Δm=±1 and the angular momentum offset from the Γ3 irrep are derived within the same fitted THF model. Because the model's parameters are adjusted to reproduce the observed peak positions, the selection-rule assignment is not independently corroborated; for example, the claim that the m1 mode would be dark without the offset depends on the fitted band structure and matrix elements. An independent symmetry-based argument, or a calculation using parameters constrained by other measurements (e.g., transport or QTM data), would strengthen this central conclusion.","section":"Fig. 3h and 'Optical Selection Rules'"}],"minor_comments":[{"comment":"The phrase 'an record electron temperature' should read 'a record electron temperature'.","section":"Abstract and Introduction"},{"comment":"The term 'theoretical prediction' in the figure caption is misleading given that the parameters are fitted to the experiment; consider replacing with 'best-fit calculation'.","section":"Fig. 3c"},{"comment":"The table lists parameter values such as n* = -4658.0 and n'* = 1671.0 without explicit units; please state the units (e.g., meV for energies, or the appropriate THF-model units) in the table caption.","section":"Extended Data Table 1"},{"comment":"The CNP resonances are presented without a model, and the text explicitly states 'we currently do not understand its exact origin.' This is acceptable for a reported observation, but the statement 'new many-body modes' in the abstract is somewhat stronger than the evidence presented; consider softening it.","section":"Fig. 4 and CNP discussion"}],"recommendation":"major_revision","confidential_remarks":"The experimental platform and the multi-device dataset are impressive, and the THF modeling is thorough. However, the two central pillars—uncalibrated detection and parameter fitting circularity—are exactly what a critical referee will probe. If the authors can supply a calibration reference or an out-of-sample prediction, the paper could become a strong addition to the field. In its current form, the central spectroscopic assignment is not uniquely supported, so I recommend major revision."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The one thing to know: this is the first far-infrared spectroscopy of MATBG flat bands, and it opens a genuinely new experimental window (millikelvin FTIR photocurrent/photovoltage). The core observation—filling-dependent FIR resonances that develop below the insulating gap, with field-dependent modes at v=+4 and distinct low-energy structure at CNP—is new and, on the evidence, real. Three devices with different gates give consistent spectra, and the transport characterization is solid. The theory-data comparison is also impressive at first glance: the THF model reproduces the m1–m3 slopes, the D1/D2 difference in m1 curvature, and the relative brightness of the modes. The \"c-electron antenna\" picture and the Δm=±1 selection rule with the Γ3 angular-momentum offset are a substantive conceptual step, and the explanation of why the f-sector is optically dark is convincing.\n\nNow the soft spots. The stress-test note about the detection chain is legitimate and is not refuted in the paper. The measured quantity is a photovoltage or photocurrent at finite DC bias, which can contain bolometric and photothermoelectric contributions. The spectral transfer function of that chain is never calibrated. Normalizing by the 1200 K blackbody curve removes the source envelope, but not the wavelength-dependent detection efficiency, and each spectrum is then normalized to its maximum. So any frequency-dependent artifact—hBN phonon, beam-splitter features, contact or substrate interference—could masquerade as a resonance. That does not prove the resonances are artifacts, but it means the assignment of m1–m3 to inter-Landau-level transitions is not uniquely supported by the data. I'd want at least one independent check: a known resonance, a different detection geometry, or a direct absorption measurement on a larger device.\n\nThe second issue is the fitting. Extended Data Table 1 lists thirteen THF parameters extracted by fitting the same experimental spectra that the “prediction” curves in Fig. 3c are then compared against. That is a genuine circularity, and it weakens the claim that the theory predicts the observed modes. The sensitivity analysis in Extended Data Fig. 7 helps, but does not eliminate the concern.\n\nThe CNP section is also oversold in the abstract. The body is appropriately cautious—the authors say they do not understand the exact origin of the 40 meV mode and leave detailed modeling for later. The abstract's wording about \"emergence of new many-body modes\" goes beyond what the body demonstrates.","headline":"First real FIR spectra of MATBG flat bands and a genuinely new experimental capability, but the mode assignment leans on an uncalibrated photodetection chain and on THF parameters fitted to the very spectra they are asked to predict.","tokens_in":15380,"tokens_out":1801,"would_cite":true,"duration_ms":19815,"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":"Magic-angle graphene's flat bands emit bright far-infrared resonances explained by the topological heavy-fermion model.","keywords":["magic-angle twisted bilayer graphene","far-infrared spectroscopy","topological heavy-fermion model","Landau level transitions","optical selection rules","flat-band fermions","millikelvin photocurrent spectroscopy","moiré materials"],"falsifier":"Measure the m1-m3 modes under circularly polarized far-infrared light at v=+4: the claimed Δm = ±1 selection rule predicts that left- and right-circular polarizations should excite different sets of inter-Landau-level transitions, so a null polarization dependence would refute the assignment. Alternatively, a device with a twist angle measurably further from the magic angle should show the predicted W3/U1-driven changes in the m1 slope and the m3 intercept, testing the c-f interaction picture without relying on the same fitted parameters.","tokens_in":14195,"feed_emoji":"🔬","tokens_out":6582,"duration_ms":58621,"temperature":0.7,"pith_summary":"The paper reports the first observation of characteristic far-infrared resonances of flat-band electrons in magic-angle twisted bilayer graphene, using a photocurrent and photovoltage Fourier-transform spectrometer that operates at millikelvin temperatures. It argues that the resonances are not simply transitions of the high-density heavy electrons; itinerant topological c-electrons act as an optical antenna, and the resonance energies are renormalized by hybridization with localized f-electrons. At full filling v=+4, the magnetic-field-dependent modes m1-m3 are identified as bright inter-Landau-level transitions obeying the selection rule Δm = ±1, which follows from an angular momentum offset tied to the Γ3 irreducible representation. At charge neutrality, pronounced low-energy resonances appear below the on-site Coulomb energy, which the authors take as evidence of new many-body modes. If correct, these results establish resonant FIR spectroscopy as a direct probe of the interacting flat bands and their organizing symmetry.","feed_headline":"First far-infrared resonances seen in magic-angle graphene","feed_subtitle":"Millikelvin spectra reveal bright Landau-level modes and a selection rule tied to the Γ3 symmetry.","key_machinery":"The load-bearing object is the topological heavy-fermion (THF) model, which decomposes the MATBG flat bands into itinerant topological Dirac fermions (c-electrons) and localized heavy fermions (f-electrons) at AA stacking sites, coupled by a momentum-dependent hybridization. The paper computes the optical conductivity of interacting Landau levels in this model and identifies the modes m1-m3 as bright c-to-c inter-Landau-level transitions across the hybridization gap, with frequencies set by f-electron interactions such as the Hubbard U1 and the c-f repulsions W1 and W3. In the rotationally invariant limit, an emergent SO(2) symmetry labels Landau levels by angular momentum m, and the Γ3 irreducible representation of the ΓM-point states shifts the conduction-band index by one, yielding the selection rule Δm = ±1. This machinery connects the spectra to microscopic interaction parameters and explains why the light c-sector dominates the optical response while the heavy f-sector renormalizes the resonance positions.","core_discovery":"The central claim is that the observed far-infrared resonances in MATBG are genuine optical excitations of its interacting flat bands, and that the topological heavy-fermion model describes them quantitatively. At v=+4, the modes m1-m3 are bright inter-Landau-level transitions of the itinerant c-electrons, with energies renormalized by hybridization with localized f-electrons, and they obey the optical selection rule Δm = ±1. This selection rule is forced by an angular momentum offset at the ΓM point that originates from the Γ3 irreducible representation; without this offset, the negatively dispersing m1 mode would be dark, so the joint appearance of m1-m3 validates the hidden rotational symmetry. At charge neutrality, resonances centered around 12-24 meV lie below the on-site Coulomb energy U1 ~ 45 meV, indicating that they arise from many-body or composite modes rather than simple single-particle flat-band transitions, and an unidentified mode near 40 meV develops at high fields.","pith_inferences":["A natural extension is to apply the same c-antenna picture to other moiré flat-band systems such as twisted bilayer MoTe2: if the light-sector dominance is generic, their FIR spectra should also show sharp c-like transitions with energies renormalized by heavy-sector interactions, a prediction testable with the same platform.","The Γ3 angular-momentum offset predicts a circular-dichroic signature: circularly polarized FIR should excite different inter-Landau-level transitions for left and right polarization, offering a direct way to map the hidden symmetry without relying on parameter fitting.","The low-energy charge-neutrality resonances may be collective, flat-band exciton-like modes; a discriminating test would be to follow them under in-plane magnetic field or as a function of temperature, since single-particle c-to-c transitions and collective modes typically respond differently to these knobs."],"forward_implications":["The v=+4 spectrum provides a direct measure of the interacting band structure: the negative magnetic-field slope of m1 bounds the interacting flat-band width from below, and the zero-field intercept of m3 measures the gap to the remote bands at the ΓM point.","Because c-electrons dominate the optical coupling, the brightness pattern of the resonances (m3 strongest, m1 and m2 weaker) becomes a fingerprint of light-heavy hybridization, allowing interaction parameters such as W1, W3, and U1 to be extracted from spectroscopy.","The selection rule Δm = ±1 with a Γ3-induced angular momentum offset is testable: without the offset, m1 would be dark, so the observation of m1-m3 together validates the hidden SO(2) organizing symmetry of the THF description.","The platform extends resonant FIR spectroscopy into the millikelvin regime, making correlated insulator states and possibly superconductivity of small moiré devices accessible to optical study.","At charge neutrality, the sub-U1 resonances indicate many-body modes distinct from single-particle flat-band transitions, and the unidentified high-field mode near 40 meV marks a concrete open problem for future theory."],"supporting_citations":[{"why":"Supplies the millikelvin FIR platform on which all reported photocurrent and photovoltage spectra were measured.","marker":"[7]"},{"why":"Introduces the topological heavy-fermion model of MATBG that the paper uses for all optical conductivity calculations.","marker":"[22]"},{"why":"Extends the THF model to magnetic fields and provides the canonical quantization used to compute Landau-level optical transitions.","marker":"[25]"},{"why":"Provides analytical approximations for THF model parameters, including the quadratic twist-angle scaling of W3 used to interpret device differences.","marker":"[23]"},{"why":"Supplies the efficient representation of atomistic strain and relaxation used in the simulations.","marker":"[24]"},{"why":"Reports the Rashba-like Landau-level crossings in transport that the paper compares with the computed THF Landau-level sequence.","marker":"[70]"},{"why":"Provides quantum twisting microscope evidence for the c and f sectors that motivates the light-sector versus heavy-sector interpretation.","marker":"[64]"},{"why":"States Kohn's theorem, the theoretical foil that makes flat-band optical resonances sensitive to interactions and motivates the measurement.","marker":"[26]"}],"fun_headline_variants":["Far-infrared light exposes flat-band fermions in MATBG","Hidden symmetry revealed in magic-angle graphene's optical response","Millikelvin spectroscopy cracks flat-band selection rule","Heavy fermions steer far-infrared light in magic-angle graphene"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The inference that the measured photovoltage or photocurrent interferograms, after normalization by the blackbody spectrum, faithfully represent the optical absorption of the MATBG flat bands rests on the premise that no spurious wavelength-dependent bolometric, thermoelectric, or gate-dependent detection artifact shapes the spectra; this premise enters where the conductivity change due to light absorption is monitored via Iph or Vph in the Millikelvin FIR Spectroscopy section.","fun_headline_variants_meta":{"raw":{"variants":["Far-infrared light exposes flat-band fermions in MATBG","Hidden symmetry revealed in magic-angle graphene's optical response","Millikelvin spectroscopy cracks flat-band selection rule","Heavy fermions steer far-infrared light in magic-angle graphene"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.0003,"raw_usage":{"total_tokens":1745,"prompt_tokens":972,"completion_tokens":773,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":588,"completion_tokens_details":{"reasoning_tokens":708}},"tokens_in":588,"tokens_out":773,"duration_ms":6679,"temperature":1.0,"reasoning_tokens":708,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-16T00:04:55.153805+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure the m1-m3 modes under circularly polarized far-infrared light at v=+4: the claimed Δm = ±1 selection rule predicts that left- and right-circular polarizations should excite different sets of inter-Landau-level transitions, so a null polarization dependence would refute the assignment. Alternatively, a device with a twist angle measurably further from the magic angle should show the predicted W3/U1-driven changes in the m1 slope and the m3 intercept, testing the c-f interaction picture without relying on the same fitted parameters.","supporting_citations":[],"review_version":1}