{"id":"35b48ca0-3035-43a9-8bb1-7aa2e72d7a78","arxiv_id":"2411.13529","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"A GW-based tight-binding model combined with Chen's derivative rule reproduces scanning tunneling spectra of Bi2Se3, Bi2Te2Se, and Bi2Te3 and reveals bulk and surface orbital contributions.","lead":"This paper calculates and measures how electrons tunnel between a scanning probe tip and the surfaces of three bismuth-chalcogenide topological insulators, using quantum many-body corrected band structures. The method lets researchers identify which atoms and orbitals contribute to each feature in the tunneling spectrum, and shows that plain density functional theory is too crude for these materials.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The claimed 'excellent agreement' and 'accurate reproduction' rely on a non-self-consistent surface barrier plus per-spectrum Fermi-level shifts and normalization; the GW-vs-DFT ranking is probably robust, but the absolute quantitative claim is not yet established.","rationale":"The reader's weakest-assumption analysis identifies the non-self-consistent surface model and the per-spectrum Fermi-level shift and normalization as the main vulnerability. My read agrees: the load-bearing condition for the abstract claim is that the calculated dI/dV spectra respond to the GW corrections in the same way as the real surface, and the step-potential/Dirichlet model is where that condition is least secure. The paper itself flags this in observation (xii) and in the discussion of discrepancies in Figs. 3, 5, and 8. The GW-vs-DFT inference in Section V uses relative peak separations, which are much less affected by a global shift or vertical normalization, so that specific conclusion is reasonably supported. What is not supported with current evidence is the stronger wording 'accurately reproduced' and 'excellent agreement' in the abstract, because the agreement includes fitted shifts, normalization, and acknowledged deviations without quantified uncertainty. A targeted numerical test that replaces the vacuum step with a self-consistent surface potential would directly probe whether the surface approximation biases the comparison; if the GW-vs-DFT ranking survives that test, the central claim would be materially strengthened. Since the reader already assigned CONDITIONAL, my stress-test does not move the verdict; it sharpens the condition under which the paper would be acceptable.","tokens_in":28368,"tokens_out":7442,"duration_ms":89638,"concrete_test":"Recompute the Bi2Se3 dI/dV curves of Fig. 8 with the GW and DFT Hamiltonians, but replace the abrupt vacuum step of Eq. (8) with a self-consistently determined surface potential, for example from a DFT slab calculation projected onto the same TB basis, keeping all other parameters fixed. Compare the first three peak positions and the gap ΔE_exp between the negative and first positive peak. If the GW-vs-DFT ranking survives and the peak shifts are within about 100 meV, the surface approximation is not the limiting factor and the central claim stands; if the ranking changes, or if the GW third-peak discrepancy (2.25 eV vs 2.5 eV) is removed or worsened, the quantitative agreement is not robust and the 'accurately reproduced' claim would need to be weakened.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The weakest load-bearing assumption is in Section III B: the half-space Hamiltonian is a hard Dirichlet truncation of the bulk TB Hamiltonian, Eq. (5), and the vacuum is an abrupt laterally averaged potential step, Eq. (8), with literature work functions and an ad hoc boundary placed 1.3 Å outside the outermost atomic cores. No surface relaxation, surface dipole, band bending, or boundary-localized orbital shifts are included; observation (xii) explicitly concedes this and defends it a posteriori from the agreement. The quantitative comparison is then made after shifting each experimental spectrum by a per-spectrum ΔEF and normalizing the theoretical and experimental dI/dV through the integrated current, as described in Section IV A. These are unconstrained degrees of freedom per spectrum. Because the barrier height and boundary position control the energy-dependent decay constant in Eq. (11), errors in the surface potential can be partially absorbed by ΔEF and the overall normalization. The strongest inference, that GW rather than DFT is required, is supported by the relative peak-separation argument in Section V, which is less sensitive to these degrees of freedom: ΔE_K^GW < ΔE_exp < ΔE_M^GW while both DFT gaps are smaller. However, the headline claims of 'excellent agreement' and 'accurate reproduction' are not fully quantitative: no error bars are given, one arbitrary shift is applied per spectrum, and acknowledged discrepancies at E ≈ 0.5 eV in Fig. 3, at E > 0.2 eV in Fig. 5, and in the third peak position in Fig. 8 are attributed to unspecified surface effects. The same surface approximation is applied to GW and DFT, so the relative DFT-vs-GW comparison is probably sound; the absolute statement that GW 'accurately reproduces' the experimental spectra is the part that remains conditional.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents a numerical scheme for computing STM tunneling current and differential conductance dI/dV for the topological insulators Bi2Se3, Bi2Te2Se, and Bi2Te3, using Wannier-interpolated tight-binding Hamiltonians derived from DFT+GW calculations. The method combines half-space band structures obtained by Dirichlet truncation of the bulk TB Hamiltonian, a simple step-like vacuum barrier, and Chen's derivative rule for s-wave metal tips and s/p-wave CO-terminated tips. The authors compare calculated and experimentally measured dI/dV spectra over bias ranges from near the band gap to roughly 2 eV, identify signatures of topological boundary modes, bulk-band hybridizations, Van Hove singularities, and orbital character, and argue that GW-corrected Hamiltonians reproduce the data substantially better than bare DFT Hamiltonians.","tokens_in":28666,"tokens_out":5017,"duration_ms":53138,"significance":"If fully established, this would be a valuable quantitative bridge between GW quasiparticle band structures and atomic-scale STM/STS measurements, with the practical advantage that the tight-binding scheme permits dense Brillouin-zone sampling that is computationally infeasible in direct first-principles surface calculations. The paper has several genuine strengths: the GW-TB Hamiltonians are independent first-principles inputs from prior work, the k-grid integration is far denser than in previous first-principles comparisons, the CO-tip treatment includes orbital mixing and tip tilt through the probe-particle model, and the GW-versus-DFT discrimination in Section V is based on a relative peak-separation criterion that is less sensitive to the fitted shifts. The inclusion of runnable-script cues in figure captions also signals a welcome degree of reproducibility. The main weakness is that the headline claims of 'excellent agreement' and 'accurate reproduction' are stronger than the evidence, given acknowledged discrepancies and several per-spectrum adjustable parameters.","major_comments":[{"comment":"The half-space description is a hard Dirichlet truncation of the bulk GW-TB Hamiltonian combined with an abrupt, laterally averaged vacuum barrier with literature work functions and an ad hoc boundary position z_b = 1.3 Å beyond the outermost atomic cores. This neglects surface relaxation, band bending, and boundary-localized orbital shifts, as the authors concede in observation (xii). Because Eq. (11) makes the vacuum decay constant kappa(n,k_parallel,Delta) directly dependent on Phi and z_b, any error in the surface potential can be partially absorbed by the per-spectrum Fermi-level shift and the overall normalization. Since the absolute 'excellent agreement' claim rests on the shape of dI/dV over eV-scale ranges, I request a sensitivity check: for example, vary z_b and Phi within reasonable bounds and show that the calculated peak positions and relative intensities move by less than the claimed agreement. Without such a check, the quantitative claim is not fully established; the robust statement is the relative-gap comparison of Section V.","section":"Section III B, Eq. (8), observation (xii)"},{"comment":"The abstract and observation (iv) claim 'excellent agreement' and 'accurately reproduced', but the manuscript itself reports discrepancies at load-bearing points. In Fig. 3(a) the calculated dI/dV is significantly lower than experiment for E = 0.5 ± 0.1 eV. In Fig. 5(a) there are deviations above E ≈ 0.2 eV that are not reproduced, with a surface 2DEG suggested as a possible cause. In Fig. 8(a) the third GW peak lies at 2.25 eV versus 2.5 eV in experiment, a 0.25 eV shift. These are acknowledged in the text and are not incidental. No error bars or uncertainty estimates are given for the spectra or for the extracted peak positions. The headline claim should be downgraded to a qualitative or relative statement, or the agreement should be quantified with a stated error metric per energy range.","section":"Section IV A, Section IV B, Section V; Figs. 3, 5, 8"},{"comment":"The quantitative comparison is not parameter-free. Each experimental spectrum is shifted by a fitted per-spectrum Delta_EF, and the theoretical and experimental dI/dV are normalized through the integrated current traces as described in Section IV A; the Brillouin-zone broadening or smoothing and the boundary position z_b are also adjustable. These degrees of freedom can compensate for systematic errors in the barrier model and in the absolute energy alignment. The GW-versus-DFT inference in Eq. (21) is less sensitive to these choices because it uses relative peak separations, but the paper should state this distinction explicitly and report the sensitivity of the spectra to each fitted parameter, for example by tabulating Delta_EF values across spectra and locations on the same sample.","section":"Section IV A, Fig. 3(b)"}],"minor_comments":[{"comment":"The text describing Fig. 8(a) says the DFT calculation is the green line, while the figure caption says the DFT-Hamiltonian spectrum is red; these color assignments need to be reconciled.","section":"Section V, Fig. 8"},{"comment":"Several figure captions contain 'Run the script ...' commands (for example, 'Run the script Bi2Te3_WF_Eval_25Feb2023.m to get this figure'); such reproducibility instructions belong in a code-availability section or supplement rather than in the printed figure captions.","section":"Figure captions, Figs. 4, 5, 6"},{"comment":"In the paragraph discussing deviations above E = 0.2 eV, the text refers to 'Figure IV B (a)', which should be 'Figure 5(a)'.","section":"Section IV B"},{"comment":"The text uses 'Chens derivative rule' without the possessive apostrophe; it should read 'Chen's derivative rule'.","section":"Section III C"},{"comment":"The sentence 'as discussed in Section IV C and Section 8' contains a broken cross-reference: there is no Section 8, and the intended reference is likely Section IV C, where the corresponding peaks are discussed.","section":"Section VI A"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is within the journal's scope and the methodological core is sound. The main risk is overclaiming in the abstract and overview: the absolute 'excellent agreement' claim is not yet supported by a controlled sensitivity analysis, whereas the relative GW-versus-DFT conclusion is well argued. I recommend requesting a sensitivity analysis and a softened or qualified version of the headline claim before acceptance."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: this is a solid methods paper whose headline overclaims. The scheme — GW-derived Wannier tight-binding Hamiltonians plus Chen's derivative rule for metal and CO tips — is a genuine step forward for interpreting dI/dV on the Bi2Se3 family, and the large-bias data on Bi2Te3 is genuinely new. The reinterpretation of the gap-region linear dI/dV as partly due to bulk valence bands rather than a clean lower Dirac cone is also a nice correction to prior work. What the paper does well: the pipeline from first-principles GW to a half-space TB Hamiltonian to real-space tunneling matrix elements is described clearly, and the orbital-resolved maps plus CO-tip tilting via the probe-particle model show real craftsmanship. The DFT-vs-GW comparison is the strongest part: the peak-separation inequality (Eq. 21) is far less sensitive to the fitted Fermi-level shifts and normalization than the absolute spectra, and it does support the claim that GW, not bare DFT, is needed for this family. Now the soft spots. The phrase 'excellent agreement' is doing more work than the evidence supports. Each experimental spectrum is shifted by an adjustable ΔEF and normalized; the surface model is a hard Dirichlet truncation with a laterally averaged potential step and an ad hoc boundary position; and the text itself concedes discrepancies at ~0.5 eV in Fig. 3, above 0.2 eV in Fig. 5, and in the third peak of Fig. 8. These are not fatal — the authors are honest about the surface approximation (observation xii) and about defect issues on Bi2Te2Se — but they mean the 'accurate reproduction' claim should be downgraded to 'good qualitative and semi-quantitative agreement, with the main features and the GW-vs-DFT ordering robust.' The lack of public code or data is a minor irritation; the Hamiltonians come from published Ref. [20] and the method is described well enough to reproduce in principle. My overall take: the contribution is real and publishable, but the conclusion needs recalibration. The physics result — many-body corrections are required to explain these STS spectra — is probably right. The surface model is the main uncertainty, and a referee should push for a sensitivity check on the truncation and boundary position. Recommendation: send it to peer review. It deserves serious refereeing, and with revisions toning down the absolute claim and quantifying the discrepancies, it should be accepted. I would cite the methods part in work on STS modeling. For a reading group, it's worth a look if you work on topological insulators or scanning probe theory, but it's not a must-read.","headline":"Solid methods paper with a real GW-vs-DFT result buried under an overclaimed 'excellent agreement' — worth refereeing, needs toning down.","tokens_in":733,"tokens_out":798,"would_cite":true,"duration_ms":30911,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["71.15.Mb","71.15.Qe","73.20.-r","68.37.Ef"],"model":"deepseek-v4-flash","headline":"A GW-based tight-binding scheme quantitatively reproduces measured STM tunneling spectra of three topological insulators, showing that bare DFT cannot describe this material family.","keywords":["topological insulator","scanning tunneling spectroscopy","differential conductance","GW approximation","Wannier functions","Chen's derivative rule","Bi2Se3","Bi2Te3"],"falsifier":"Measure dI/dV across a sample region where the local work function or band bending is known to vary, for example near a step edge or on a sample intentionally doped to move the Fermi level far from the Dirac point. The fixed vacuum-barrier model uses a single work function and a single global Fermi-level shift; if the conduction-band onset shifts relative to the bulk-gap features by more than the fit tolerance as the local band bending changes, the laterally averaged step-potential approximation is falsified. Alternatively, a self-consistent surface calculation that moves the Dirac point or conduction band edge relative to the bulk valence band by more than the fit tolerance would break the scheme's central agreement.","tokens_in":1892,"feed_emoji":"🔬","tokens_out":3046,"duration_ms":101602,"temperature":0.7,"pith_summary":"This paper establishes that scanning tunneling spectroscopy on the topological insulators Bi2Se3, Bi2Te2Se, and Bi2Te3 can be predicted quantitatively from first-principles many-body calculations. The authors build half-space tight-binding Hamiltonians from GW-corrected, Wannier-interpolated band structures and use Chen's derivative rule to compute the tunneling current and differential conductance for both metal and CO-terminated tips. The calculated spectra closely track 4.4 K STM/AFM experiments across the Dirac point, bulk band edges, and Van Hove singularities. Direct comparison shows that plain density functional theory shifts the spectral peaks and underestimates the trivial band gaps, whereas the GW Hamiltonians reproduce the measured positions and amplitudes. If correct, the scheme turns STS into a quantitative probe of many-body quasiparticle band structures.","feed_headline":"GW model reproduces STS spectra of Bi2Se3 family","feed_subtitle":"A tight-binding scheme with many-body corrections matches experimental tunneling spectra; plain DFT fails.","key_machinery":"The key machinery is a four-step pipeline: maximally localized Wannier functions and Wannier-interpolated tight-binding Hamiltonians built from first-principles DFT+GW calculations; half-space Hamiltonians obtained by Dirichlet truncation of the bulk hopping matrices between quintuple layers; propagation of the resulting Bloch states through a laterally averaged vacuum potential with a work-function step; and Chen's derivative rule, which turns the tunneling matrix element into differential operators acting on the sample wavefunction at the tip apex. For CO-terminated tips the orbital mix includes s and p orbitals, with the CO tilt included through a mechanistic probe-particle model. The pipeline converts a quasiparticle band structure into the experimental dI/dV signal with no free parameters other than a global Fermi-level shift and an overall normalization.","core_discovery":"The central claim is that the measured dI/dV spectra of the Bi2Se3 family are quantitatively reproduced by a tight-binding Hamiltonian whose parameters are taken from GW many-body calculations, evaluated with Chen's derivative rule for the tip-sample tunneling matrix element. The paper shows that a DFT-based tight-binding model systematically underestimates the trivial band gaps at the Brillouin-zone boundary and places the spectral peaks at the wrong energies, while the GW-based model places them at the measured positions and captures the linear conductance increase inside the bulk gap, the hybridization of the topological surface state with bulk bands, and the suppression of tunneling matrix elements near the conduction-band edge. A specific finding is that the linear increase of dI/dV below the Dirac point in Bi2Se3 comes from tunneling into bulk valence bands hybridized with the surface state rather than from an isolated lower Dirac cone, modifying the interpretation of previous step-edge and gap-opening studies.","pith_inferences":["Editorial inference: if the scheme holds, the same pipeline could be applied to other layered or weakly bonded materials where GW corrections are large, turning STS into a routine test of many-body band structures.","Editorial inference: the hard-Dirichlet boundary condition implies that surface-specific reconstructions or band bending at the top quintuple layer would show up as systematic deviations; comparing spectra on different surface terminations or under an externally tuned two-dimensional electron gas would test this.","Editorial inference: the energy-dependent tip-orbital mix was held constant, so making it energy-dependent could improve the quantitative agreement of the dI/dV maps and would probe how much of the residual deviation comes from the tip model.","Editorial inference: adding defects to the tight-binding model, as suggested in the paper, could predict impurity resonances at the atomic scale and directly connect the observed dI/dV maps to specific native point defects."],"forward_implications":["STS spectra computed from GW Hamiltonians can be used to read off the energy position of the Dirac point and the bulk band edges in all three compounds, even where the Dirac point is buried in the valence bands.","The lower linear segment in the Bi2Se3 spectrum is attributed to bulk valence bands hybridized with the surface state, not an isolated lower Dirac cone, which changes how to interpret future gap-opening experiments.","Calculated dI/dV maps distinguish the orbital character of Bloch bands at the atomic scale, making the band inversion at the Gamma point visible without ARPES.","The method works over wide bias ranges (up to plus or minus 2 V on Bi2Te3), where Van Hove singularities from flat bands at the Brillouin-zone boundary produce characteristic peaks that can be used to quantify trivial band gaps.","Because the tight-binding model is computationally cheap, the Brillouin-zone integration can be done on dense 80-by-80 or 90-by-90 grids, eliminating the numerical artifacts that limited earlier first-principles comparisons."],"supporting_citations":[{"why":"Supplies the GW-corrected tight-binding Hamiltonians of all three compounds that are the starting point of the calculation.","marker":"[20]"},{"why":"Provides Bardeen's tunneling formula that defines the current as a sum over matrix elements.","marker":"[22]"},{"why":"Provides Chen's derivative rule that approximates the tunneling matrix element by differential operators on the sample wavefunction.","marker":"[23]"},{"why":"Provides the Wannier-interpolation formalism used to construct real-space tight-binding Hamiltonians from first-principles band structures.","marker":"[21]"},{"why":"Supplies the minimal four-band model of the surface Dirac cone that the paper extends and compares against for the surface-state band structure.","marker":"[24]"},{"why":"Earlier first-principles-based dI/dV comparison on Bi2Te3 limited by a coarse k-grid, which the tight-binding approach supersedes.","marker":"[14]"},{"why":"Previous experimental attribution of the linear dI/dV rise below the Dirac point to the lower Dirac cone, which the present calculation modifies.","marker":"[17]"},{"why":"Describes the structural AFM identification of surface defects and tip preparation used in the experiments.","marker":"[15]"}],"fun_headline_variants":["GW tight-binding reproduces Bi2Se3-family STS spectra","DFT fails; GW model matches Bi2Se3 STS spectra","Linear STS rise in Bi2Se3 stems from hybridized bulk bands","GW model reveals bulk bands shape Bi2Se3 STS"],"cache_read_input_tokens":31232,"weakest_assumption_plain":"The load-bearing assumption is that a hard truncation of the bulk GW tight-binding Hamiltonian plus a laterally averaged vacuum barrier with a literature value of the work function is an accurate model of the actual surface, so that the only free adjustments—an overall bias shift and a normalization—are enough to match every measured spectrum.","fun_headline_variants_meta":{"raw":{"variants":["GW tight-binding reproduces Bi2Se3-family STS spectra","DFT fails; GW model matches Bi2Se3 STS spectra","Linear STS rise in Bi2Se3 stems from hybridized bulk bands","GW model reveals bulk bands shape Bi2Se3 STS"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001414,"raw_usage":{"total_tokens":5699,"prompt_tokens":922,"completion_tokens":4777,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":538,"completion_tokens_details":{"reasoning_tokens":4704}},"tokens_in":538,"tokens_out":4777,"duration_ms":37264,"temperature":1.0,"reasoning_tokens":4704,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T16:18:16.065663+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure dI/dV across a sample region where the local work function or band bending is known to vary, for example near a step edge or on a sample intentionally doped to move the Fermi level far from the Dirac point. The fixed vacuum-barrier model uses a single work function and a single global Fermi-level shift; if the conduction-band onset shifts relative to the bulk-gap features by more than the fit tolerance as the local band bending changes, the laterally averaged step-potential approximation is falsified. Alternatively, a self-consistent surface calculation that moves the Dirac point or conduction band edge relative to the bulk valence band by more than the fit tolerance would break the scheme's central agreement.","supporting_citations":[{"cited_title":"Aguilera, C","cited_arxiv_id":null,"evidence_quote":"Supplies the GW-corrected tight-binding Hamiltonians of all three compounds that are the starting point of the calculation."},{"cited_title":"Bardeen, Tunnelling from a Many-Particle Point of View, Phys","cited_arxiv_id":null,"evidence_quote":"Provides Bardeen's tunneling formula that defines the current as a sum over matrix elements."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides Chen's derivative rule that approximates the tunneling matrix element by differential operators on the sample wavefunction."},{"cited_title":"Zhang, C","cited_arxiv_id":null,"evidence_quote":"Supplies the minimal four-band model of the surface Dirac cone that the paper extends and compares against for the surface-state band structure."},{"cited_title":"Netsou, D","cited_arxiv_id":null,"evidence_quote":"Earlier first-principles-based dI/dV comparison on Bi2Te3 limited by a coarse k-grid, which the tight-binding approach supersedes."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Previous experimental attribution of the linear dI/dV rise below the Dirac point to the lower Dirac cone, which the present calculation modifies."},{"cited_title":"Liebig, C","cited_arxiv_id":null,"evidence_quote":"Describes the structural AFM identification of surface defects and tip preparation used in the experiments."}],"review_version":1}