{"id":"99bef202-96f7-475b-919d-6a3fa8102313","arxiv_id":"1908.05270","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"RF admittance measurements on Bi2Se3 capacitors yield Dirac fermion velocities of about 5 x 10^5 m/s at both the passivated surface and the hBN interface, and show the technique tolerates a trivial surface accumulation layer.","lead":"This paper measures the electrical capacitance of metal-insulator-topological insulator devices to probe the electronic states on the surface of the topological insulator Bi2Se3. The authors extract the Dirac fermion velocity and show the method still works when parasitic electron states are present on the other surface.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The Berglund energy calibration uses the total capacitance including the ungated-surface parasitic branch, so the reported Dirac velocities may be biased unless that branch is subtracted or shown to be voltage-independent.","rationale":"The reader's weakest assumption is the correct one: the paper's central quantitative output, v_D, is extracted from a c_q vs Δε_f plot, and Δε_f is obtained from a Berglund integral that treats the total measured capacitance as the gated surface branch. The paper itself demonstrates that a second branch contributes to the capacitance minimum, with magnitude large enough to affect the integral. Because the ungated-surface charge is coupled through the bulk, its contribution is not guaranteed to be constant in gate voltage, so the energy-axis error is not just a rigid offset; it can change the fitted slope. This is exactly a correctness risk internal to the analysis, not a disagreement with consensus: ARPES values are consistent, but the reported agreement cannot validate the calibration if the same data are used to set both axes. The proposed numerical test settles the question. We do not see a more load-bearing concern; other reservations (effective mass, no error bars) are secondary. Since the reader already flagged this assumption and assigned CONDITIONAL, our read supports that verdict without moving it.","tokens_in":10021,"tokens_out":16054,"duration_ms":152268,"concrete_test":"Re-analyze the raw capacitance data with the two-branch model C(V) = c_g c_q^g(V)/(c_g+c_q^g(V)) + (1/c_gBulk + 1/c_q^u(V))^{-1}, where c_q^g = e²(ε_f - ε_D)/(2πℏ²v²) and c_q^u includes a Dirac term at an unknown offset plus a 2DEG of mass 0.14m0 (for CAP2). Fit v², the gated Dirac-point voltage, the ungated Fermi offset, and the bulk coupling simultaneously. If the best-fit v² differs by more than ~10% from the values extracted via the single-surface Berglund integral, the reported Dirac velocities are biased by the parasitic ungated branch.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The load-bearing assumption is that the Berglund integral (paragraph after Fig. 3) converts the measured C(V) into the Fermi energy of the gated surface alone. That requires C(V) to be the gated-surface series branch, c_g c_q^gated/(c_g+c_q^gated). The paper's own c_min analysis (Section on capacitance minimum) shows this is not the case: the ungated surface is coupled through the depleted bulk, adding a parallel branch c_min = (1/c_gBulk + 1/c_q^ungated)^{-1}, with c_q^ungated ≈ 40 fF/μm² for CAP1 and >87 fF/μm² for CAP2 (if the 2DEG has bulk mass). Unless this branch is subtracted, the integral becomes Δε_f^meas = Δε_f^true - ∫ (c_min(V)/c_g) dV, a voltage-dependent rescaling of the energy axis. The paper neither subtracts this branch nor checks how the fitted Dirac velocity changes when the ungated surface is included in the calibration; the fitted v_D = 5.2 and 4.9×10^5 m/s therefore inherits an unquantified systematic error. The acknowledgment that the Dirac/2DEG decomposition of c_q^ungated is not independent confirms that the single-surface analysis is not self-consistent.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper reports RF admittance measurements of metal-hBN-Bi2Se3 capacitors at 10 K up to 10 GHz for two device geometries: CAP1, a top-gated hBN-passivated device, and CAP2, a bottom-gated device with an air-exposed top surface. From a distributed RC model the authors extract channel resistance and total capacitance, derive a quantum capacitance c_q, convert gate voltage to Fermi energy via the Berglund integral, and fit the Dirac linear relation c_q = e^2 Δε_f / (2πℏ²v²) to obtain Dirac velocities of 5.2×10^5 m/s for CAP1 and 4.9×10^5 m/s for CAP2. The capacitance minimum is interpreted as capacitive coupling to the ungated surface through the depleted bulk, and for CAP2 the authors infer an electron accumulation layer on the exposed top surface. The central claim is that RF compressibility can probe the gated Dirac surface even when the opposite surface hosts a parasitic density of states.","tokens_in":10257,"tokens_out":13660,"duration_ms":132479,"significance":"If the analysis is correct, the paper provides a useful technical demonstration: RF quantum capacitance can extract Dirac velocities at both a passivated Bi2Se3 surface and a buried hBN/Bi2Se3 interface, with values consistent with prior ARPES work. The two-device comparison is a genuine strength, as is the use of an external ARPES benchmark to anchor the velocity fit and the explicit physical model for the residual capacitance minimum. The paper also makes a practical, falsifiable claim about hBN passivation preventing surface accumulation layers. However, the extraction chain is not self-consistent: the same capacitance minimum that the authors attribute to the ungated surface is left inside the total capacitance used for both the c_q inversion and the Berglund energy calibration. Because the correction is likely large, the reported velocities carry an unquantified systematic error even though the qualitative claim is plausible.","major_comments":[{"comment":"The c_q extraction and the Berglund integral are applied to the total measured capacitance C(V), yet the paper's own analysis attributes c_min to a separate branch from the ungated surface, c_min = (1/c_gBulk + 1/c_q^ungated)^{-1}. If this branch is present, the inversion c = (1/c_g + 1/c_q)^{-1} is not the correct relation for the full circuit, and the Berglund integrand should contain the gated-surface series branch only. The correction is not small: from Table 1, c_min/c_g is roughly 10-15 for both devices, so C(V) can exceed c_g; taken literally, the formula for c_q would even become negative in the region where C exceeds c_g. Please specify the full equivalent circuit used for C(V), subtract or model the ungated-surface branch before extracting c_q and Δε_f, and report how v_D changes under a plausible range of this correction.","section":"After Eq. (3) and Fig. 3(c); the c_min analysis paragraph"},{"comment":"The central comparison is that the Dirac velocities, 5.2×10^5 m/s (CAP1) and 4.9×10^5 m/s (CAP2), are the same 'within experimental uncertainty,' but no uncertainties are quoted for v_D, no fit range is given, and the number of fitted points is not stated. This makes the equality claim unverifiable. Please report fit uncertainties, the Δε_f window used, and whether the fits allowed an intercept; since Eq. (1) predicts a zero intercept while the data show a finite c_min, a fit constrained through the origin would bias v_D, whereas an unconstrained intercept would introduce a second parameter that must be stated.","section":"Fig. 4(a) and Table 1"},{"comment":"The inference that the exposed top surface of CAP2 hosts a trivial electron accumulation layer is model-dependent: it assumes a 2DEG effective mass m = 0.14m0 and, as the manuscript notes, c_q^ungated = c_q2DEG + c_qTSS cannot be independently determined. The inequality c_min > 50 fF/μm² is not by itself a measurement of a 2DEG; other combinations of Dirac and massive states, or a different effective mass, could satisfy it. The conclusion should be presented as a plausible model consistent with earlier literature rather than as a unique determination.","section":"Section on the capacitance minimum (CAP2)"}],"minor_comments":[{"comment":"There are several typographical errors in the fabrication paragraphs, for example 'A seen in Fig. 1(d)' should be 'As seen', and 'in the addition to optimizing' should be 'in addition to optimizing'.","section":"Device fabrication, Fig. 1(d)"},{"comment":"The sentence containing 'through an insulating bulk with we can compute c_min using:' appears to be missing a word or value; please complete the sentence.","section":"CAP2 c_min paragraph"},{"comment":"Equation (3) as typeset contains an awkward denominator; please verify that the printed expression matches the standard distributed RC admittance formula.","section":"Eq. (3)"},{"comment":"Since CAP1 was reported in Ref. [15], the manuscript should explicitly indicate which data are new in this work (mainly CAP2 and the comparison) and avoid the impression that both devices are first reported here.","section":"CAP1 provenance and Fig. 3"},{"comment":"No error bars are shown in Fig. 4(a); adding representative error bars would strengthen the claim that the finite c_min is not an experimental artifact.","section":"Fig. 4(a)"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is within scope for a mesoscale physics journal, and the RF compressibility technique is of genuine interest. The main risk is the internal inconsistency between the Berglund calibration and the c_min model, which is load-bearing for the reported velocities. This appears fixable by reanalysis rather than fatal. I would also like the editor to ensure the authors distinguish new data from previously published CAP1 results."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Bottom line: this is a worthwhile incremental paper. The genuinely new part is the bottom-gated device (CAP2), where the quantum capacitance of the hBN/Bi2Se3 interface is measured, and the demonstration that a trivial accumulation layer on the opposite, ungated surface does not hide the ambipolar Dirac signature of the gated surface. The RF admittance work is careful, and the two-device comparison between hBN-passivated and air-exposed surfaces is a useful data point for the community.\n\nThe paper deserves a serious referee. The main soft spot is exactly what your stress-test note flags: the Berglund integral uses the total measured capacitance, which includes the ungated-surface branch (c_min) in parallel with the gated-surface quantum capacitance. That branch is sizeable (30 fF/um^2 for CAP1, 50-85 for CAP2) compared with c_g (2-3 fF/um^2), so near the Dirac point the energy axis is compressed unless the branch is subtracted or included in a two-surface model. The paper does neither, and quotes v_D with no uncertainty, so the claim that the two velocities are 'identical within experimental uncertainty' is unverifiable. My guess is the bias is not enormous--the extracted 5.2 and 4.9 x 10^5 m/s sit near ARPES values--but a 10-20% systematic error in v_D is plausible. That matters for a paper whose quantitative headline is the velocity.\n\nThe c_min interpretation itself is reasonable: the numbers require an accumulation layer for CAP2, and the authors are candid that they cannot decompose Dirac and 2DEG contributions. The assumption that the surface 2DEG mass equals the bulk value is standard but not tested; minor. The 'always' conclusion rests on two devices, but as a proof-of-principle it is fine.\n\nWho should read it: people using RF capacitance on topological insulators, and groups working on hBN/Bi2Se3 interfaces. It does not transform the field, but it does extend the tool to a regime with parasitic density of states, which is practically useful.\n\nRecommendation for review: send it out. Ask the authors to propagate uncertainties and to check how v_D changes if c_min is subtracted from c_q before the Berglund integral, or if a two-surface model is used for the energy calibration. Those are addressable in a revision.","headline":"Solid incremental RF-compressibility paper; the Berglund energy calibration needs a correction for the ungated-surface branch before the Dirac velocities can be taken at face value.","tokens_in":10875,"tokens_out":15031,"would_cite":false,"duration_ms":144618,"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":"Radio-frequency admittance of metal–hBN–Bi2Se3 capacitors resolves the quantum capacitance of a gated Dirac surface even when the opposite surface carries a trivial electron accumulation layer.","keywords":["topological insulator","quantum capacitance","RF admittance","Bi2Se3","Dirac surface states","compressibility","hexagonal boron nitride","ambipolar effect"],"falsifier":"Measure a CAP2-like device with a second gate that independently tunes the ungated top surface; if the extracted Dirac velocity changes as that surface's chemical potential is moved, the single-surface assumption behind the Berglund calibration is violated, whereas a constant velocity would confirm that the parasitic background is fixed.","tokens_in":9781,"feed_emoji":"📡","tokens_out":7734,"duration_ms":76910,"temperature":0.7,"pith_summary":"Radio-frequency admittance measurements on metal–hBN–Bi2Se3 capacitors can resolve the quantum capacitance of one gated Dirac surface even when the opposite, ungated surface hosts a trivial electron accumulation layer. The paper reports a capacitance minimum as the gate voltage is swept, the ambipolar signature of tuning through a Dirac point, in both a passivated top surface and a bottom hBN interface, and extracts Dirac velocities of $5.2\\times10^5$ m/s and $4.9\\times10^5$ m/s, in line with photoemission values. If this is right, RF compressibility becomes a practical way to study topological surface states in devices that also contain parasitic conducting channels, without needing a magnetic field.","feed_headline":"RF compressibility extracts Dirac velocity in Bi2Se3 capacitors","feed_subtitle":"Two capacitor designs yield speeds near 5×10^5 m/s, matching photoemission, even with a trivial surface electron layer.","key_machinery":"The central object is the quantum capacitance $c_q$ of a surface Dirac cone, obtained from the RF admittance $Y(\\omega)$ of a two-terminal capacitor through a one-dimensional distributed RC model, with the Fermi-energy axis set by the Berglund integral $\\Delta\\varepsilon_f = \\int_0^{V_g} dV\\,(1 - C(V)/c_g)$. The load-bearing identity is $c_q = e^2\\Delta\\varepsilon_f/(2\\pi\\hbar^2 v^2)$, whose slope gives the Dirac velocity, and the capacitance minimum is modeled as the series combination of the depleted-bulk geometric capacitance and the ungated surface's quantum capacitance, $c_{\\text{min}} = (1/c_{g,\\text{bulk}} + 1/c_q^{\\text{ungated}})^{-1}$. These pieces convert a raw admittance sweep into a statement about the surface band structure.","core_discovery":"The central claim is that the RF admittance of a metal–insulator–topological-insulator capacitor measures the compressibility of the Dirac surface state, and that this measurement survives the presence of a trivial accumulation layer on the ungated surface. In both device geometries, the quantum capacitance $c_q$ displays a pronounced minimum when the gated surface is tuned to the Dirac point, even though the ungated surface remains metallic. Fitting $c_q = e^2\\Delta\\varepsilon_f/(2\\pi\\hbar^2 v^2)$ gives Dirac velocities $v = 5.2\\times10^5$ m/s for the hBN-passivated top surface and $v = 4.9\\times10^5$ m/s for the hBN–Bi2Se3 interface, consistent with earlier photoemission results. The residual capacitance at the minimum is attributed to capacitive coupling through the depleted bulk of Bi2Se3 to the ungated surface, whose chemical potential sits about 170 meV above its Dirac point in the passivated device, and is high enough in the air-exposed device that a parabolic two-dimensional electron gas must contribute.","pith_inferences":["If the bulk-coupling explanation is correct, the ungated-surface offset should scale with Bi2Se3 thickness: a thinner flake raises $c_{g,\\text{bulk}}$ and should push $c_{\\text{min}}$ upward, so measuring $c_{\\text{min}}$ on flakes of 6, 8, and 12 quintuple layers would test the series-capacitance model directly.","The same RF geometry could be applied to Fermi-arc surfaces of Weyl semimetals, where a gated surface hosts the arc while the ungated surface contributes a trivial Fermi surface; a quantitative separation would need the two-dimensional electron gas contribution modeled independently rather than lumped into the total ungated-surface quantum capacitance.","One testable consequence of the identical measured velocities is that hBN encapsulation leaves the Dirac cone speed unchanged; an angle-resolved photoemission study on hBN-capped Bi2Se3 would confirm or contradict this directly."],"forward_implications":["In both CAP1 and CAP2, a gate voltage can tune the gated surface through its Dirac point even when the opposite surface hosts a trivial accumulation layer, so RF compressibility is a viable local probe of a single topological surface.","The hBN-passivated top surface and the hBN/Bi2Se3 interface yield the same Dirac velocity within uncertainty, so the interface does not strongly renormalize the surface Dirac cone.","The residual capacitance minimum is explained by bulk-mediated coupling to the ungated surface; for the air-exposed top surface this requires a populated parabolic two-dimensional electron gas, indicating that passivation prevents the accumulation layer.","Because a capacitance minimum and a linear $c_q(\\Delta\\varepsilon_f)$ curve are sufficient to extract the Dirac velocity, the same measurement can be applied to other topological or Dirac systems with parasitic bands as long as one surface can be gated through its Dirac point."],"supporting_citations":[{"why":"Supplies the distributed RC model and de-embedding routine used to extract capacitance and channel resistance from the RF admittance.","marker":"[1]"},{"why":"Supplies the Berglund integral that converts the capacitance–voltage trace into the Fermi-energy axis.","marker":"[6]"},{"why":"Establishes the RF quantum capacitance technique on Bi2Se3 capacitors and the CAP1 architecture that this work extends.","marker":"[15]"},{"why":"Provides the ARPES Dirac velocity and band structure against which the measured Dirac velocities are checked.","marker":"[23]"},{"why":"Provides comparative photoemission connecting the bulk Fermi surface and Dirac surface state, used to place the Dirac point relative to the conduction band.","marker":"[24]"},{"why":"Gives graphene quantum capacitance values whose much smaller minimum helps rule out a calibration artifact origin for the residual capacitance.","marker":"[25]"},{"why":"Supplies the Bi2Se3 bulk dielectric constant $\\kappa \\approx 100$ used in the bulk-coupling estimate of the capacitance minimum.","marker":"[26]"},{"why":"Supplies the effective mass $m = 0.14 m_0$ used to estimate the two-dimensional electron gas quantum capacitance of the accumulation layer.","marker":"[30]"}],"fun_headline_variants":["RF capacitance probes Dirac speed despite trivial surface states","Dirac velocity from RF compressibility in Bi2Se3 capacitors","Ambipolar quantum capacitance reveals Dirac fermions in Bi2Se3","Topological surface states measured via RF compressibility","RF admittance extracts Dirac velocity with parasitic electrons present"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing assumption is that the ungated surface acts as a constant parasitic capacitance during the gate sweep; if its chemical potential moves with the gate voltage through the bulk coupling the paper invokes to explain $c_{\\text{min}}$, then the Berglund-derived Fermi-energy axis is miscalibrated and the fitted Dirac velocity is biased.","fun_headline_variants_meta":{"raw":{"variants":["RF capacitance probes Dirac speed despite trivial surface states","Dirac velocity from RF compressibility in Bi2Se3 capacitors","Ambipolar quantum capacitance reveals Dirac fermions in Bi2Se3","Topological surface states measured via RF compressibility","RF admittance extracts Dirac velocity with parasitic electrons present"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000188,"raw_usage":{"total_tokens":1316,"prompt_tokens":916,"completion_tokens":400,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":532,"completion_tokens_details":{"reasoning_tokens":320}},"tokens_in":532,"tokens_out":400,"duration_ms":4321,"temperature":1.0,"reasoning_tokens":320,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T13:19:18.563037+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure a CAP2-like device with a second gate that independently tunes the ungated top surface; if the extracted Dirac velocity changes as that surface's chemical potential is moved, the single-surface assumption behind the Berglund calibration is violated, whereas a constant velocity would confirm that the parasitic background is fixed.","supporting_citations":[{"cited_title":"Transport scattering time probed through rf admittance of a graphene capacitor","cited_arxiv_id":null,"evidence_quote":"Supplies the distributed RC model and de-embedding routine used to extract capacitance and channel resistance from the RF admittance."},{"cited_title":"Surface states at steam-grown silicon-silicon dioxide interfaces","cited_arxiv_id":null,"evidence_quote":"Supplies the Berglund integral that converts the capacitance–voltage trace into the Fermi-energy axis."},{"cited_title":"rf Quantum Capacitance of the Topological Insulator Bi2Se3 in the Bulk Depleted Regime for Field-Effect Transistors","cited_arxiv_id":null,"evidence_quote":"Establishes the RF quantum capacitance technique on Bi2Se3 capacitors and the CAP1 architecture that this work extends."},{"cited_title":"Observation of a large-gap topological-insulator class with a single Dirac cone on the surface","cited_arxiv_id":null,"evidence_quote":"Provides the ARPES Dirac velocity and band structure against which the measured Dirac velocities are checked."},{"cited_title":"Bulk Fermi surface coexistence with Dirac surface state in Bi2Se3 : A comparison of photoemission","cited_arxiv_id":null,"evidence_quote":"Provides comparative photoemission connecting the bulk Fermi surface and Dirac surface state, used to place the Dirac point relative to the conduction band."},{"cited_title":"Interaction phenomena in graphene seen through quantum capacitance","cited_arxiv_id":null,"evidence_quote":"Gives graphene quantum capacitance values whose much smaller minimum helps rule out a calibration artifact origin for the residual capacitance."},{"cited_title":"Surface conduction of topological Dirac electrons in bulk insulating Bi2Se3","cited_arxiv_id":null,"evidence_quote":"Supplies the Bi2Se3 bulk dielectric constant $\\kappa \\approx 100$ used in the bulk-coupling estimate of the capacitance minimum."},{"cited_title":"Magneto-Optics of Massive Dirac Fermions in Bulk Bi2Se3","cited_arxiv_id":null,"evidence_quote":"Supplies the effective mass $m = 0.14 m_0$ used to estimate the two-dimensional electron gas quantum capacitance of the accumulation layer."}],"review_version":1}