{"id":"b8fe14e8-1d9f-4bed-8b4d-59f9320287c0","arxiv_id":"2607.25411","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":7,"one_line_summary":"In twisted graphene, moderate disorder can lengthen the electron mean free path in flat bands, quasicrystalline stacks show a first-reported sub-ballistic transport exponent, and corrugations cap spin lifetimes at nanoseconds.","lead":"This doctoral thesis uses million-atom simulations to show that moderate disorder can lengthen the mean free path of electrons in magic-angle twisted bilayer graphene, that graphene quasicrystals spread electrons sub-ballistically, and that atomic ripples limit spin lifetimes in suspended graphene to nanoseconds. It matters because it offers a unified, experimentally testable picture of how geometry and disorder — not only interactions — control transport in twisted graphene,","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Mean-free-path extraction at the flat band depends on clean-system v_F and a 600-fs plateau that is least reliable where ℓ is largest; the W = 3γ0/4 → 3γ0/2 comparison may be an artifact.","rationale":"The reader identified the same load-bearing weakness: the mean-free-path extraction assumes a converged diffusive plateau and a single clean-system Fermi velocity. My analysis sharpens this by showing that exactly the anomalous regime — flat-band E = 0 at weaker disorder — is where both assumptions are most fragile. The paper is otherwise transparent about its limitations (inaccessible diffusive regime for weakest disorders, KPM broadening, no error bars), which is why this remains a conditional concern rather than a demonstrated falsification. The proposed check directly targets the two coupled assumptions and would settle whether the reported ℓ enhancement is physical. Since the reader's verdict is already CONDITIONAL and this concern does not move it to ACCEPT or REJECT, the appropriate recommendation is UNCHANGED.","tokens_in":55447,"tokens_out":6267,"duration_ms":74135,"concrete_test":"Repeat the Kubo/MSD calculation at E = 0 for W = 3γ0/4 and W = 3γ0/2 with the time horizon extended to at least 6000 fs (matching the clean-system long-time runs) and with several disorder realizations. Independently extract a disorder-resolved Fermi velocity, e.g. v_F(W,E) = lim_{t→0} sqrt(D(t)/t), and recompute ℓ(W,E) = 2D_plateau/v_F(W,E). If the inequality ℓ(W = 3γ0/2) > ℓ(W = 3γ0/4) survives both the plateau-convergence check and the v_F renormalization, the disorder-induced delocalization claim is supported; if it reverses or becomes statistically insignificant, the headline claim is an extraction artifact.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim — that Anderson disorder increases the mean free path at charge neutrality in MATBLG — rests on converting the simulated diffusion coefficient D(E,t) into ℓ via Eqs. (3.16)–(3.17): D = ½ v_F² τ_p and ℓ = v_F τ_p, so ℓ = 2D/v_F. The paper extracts v_F from the clean-system ballistic slope (Sec. 5.2.3, Fig. 5.3 inset) and does not report a disordered-system v_F at the same energy. This is load-bearing because Anderson broadening changes which states contribute at E = 0: the effective ⟨v²⟩ at the flat band can rise substantially even while the true scattering time τ_p falls. An increase in D(0) is then not evidence of a longer τ_p or longer ℓ. A second, compounding issue is the plateau itself: the diffusive plateau is read at t = 600 fs, but the clean system required 6000 fs to reach its asymptotic ballistic regime, and the diffusive-onset time τ_p = ℓ/v_F grows with ℓ. Thus the largest claimed ℓ values are exactly those least likely to have converged by 600 fs, so the W = 3γ0/4 versus W = 3γ0/2 comparison may be an artifact of comparing a partially-converged D with a fully-converged one. The paper's own caveat (Sec. 5.2.5) that the diffusive regime is computationally inaccessible for the weakest disorders strengthens, rather than removes, this concern.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript (a PhD thesis deposited on arXiv) uses large-scale real-space Kubo/Kernel Polynomial Method simulations to study transport in superperiodic, quasicrystalline, and disordered graphene stacks. Its central claim, developed in Sec. 5.2, is that in magic-angle twisted bilayer graphene a finite window of Anderson disorder (W ≈ 3γ0/4 to 3γ0/2) broadens the flat bands and increases the diffusion coefficient and mean free path at charge neutrality — a disorder-induced delocalization that reverses at W = 2γ0. Supporting results include a sub-ballistic exponent α ≈ 0.84 for dodecagonal graphene quasicrystal approximants (Sec. 5.3.3), fragility of quasicrystalline states to disorder and to an added third layer (Secs. 5.3.4, 5.4), and nanosecond spin lifetimes in corrugated monolayer graphene from curvature-induced Rashba fields (Ch. 7). The manuscript is transparent about its limitations, including an explicit statement that the magnetism chapter (Ch. 6) is a work in progress and that the diffusive regime is not accessed at the weakest disorders.","tokens_in":55685,"tokens_out":6690,"duration_ms":71719,"significance":"If the central claim survives scrutiny, it is a conceptually important counterexample to the usual disorder-localization scaling: moderate disorder can delocalize flat-band states by broadening the flat bands and reducing the effective scattering rate. The connection to the quantum metric via the SWM sum rule is a valuable geometric perspective, and the quasicrystal exponent and spin-lifetime results address open questions. The manuscript's strengths are its large-scale atomistic KPM calculations, explicit structural relaxation, a code-availability appendix, and unusually honest caveats about unaccessed diffusive regimes and unreproduced literature results. However, the main claim's mean-free-path extraction pipeline contains a load-bearing fragility: the reported ℓ values use the clean-system Fermi velocity and a 600 fs plateau assumption, and the largest ℓ values are Fermi-golden-rule extrapolations anchored to a single converged point. Fixing this requires new analysis rather than editing.","major_comments":[{"comment":"The disorder-induced delocalization claim is supported by ℓ(E,W) in Fig. 5.5, obtained from the diffusion coefficient through ℓ = 2D/v_F using the clean-system Fermi velocity v_F(E) (inset of Fig. 5.3). Because Anderson disorder broadens the flat bands and changes the spectral composition at E=0, the effective ⟨v²⟩ at the flat band is not guaranteed to equal the clean-system v_F². A rise in D(0) from W=3γ0/4 to W=3γ0/2 can therefore be produced by a rise in ⟨v²⟩ even if the true scattering time τ_p is unchanged or reduced. Please provide a disorder-resolved v_F(E,W) — for example from the short-time ballistic slope of D(t) or from a direct evaluation of the velocity operator spectral function — and show that τ_p = 2D/v_F² also increases. In addition, the 600 fs window is not evidently a converged diffusive plateau for the W=3γ0/4 case: the clean system required 6000 fs to reach its asymp","section":"Sec. 5.3.3 / Fig. 5.6 and Fig. 5.14"},{"comment":"The first estimate of the sub-ballistic exponent α ≈ 0.84 for dodecagonal graphene quasicrystals is fitted from D(t) in the interval t < 13 fs in a 29.8° periodic approximant. The manuscript itself states that this window is before the electronic spreading reaches the approximant's superperiodicity; the shadowed region in Fig. 5.14 marks this limitation. A power-law fit in such a short, pre-periodicity window can reflect a transient inherited from the initial condition or from the periodic approximant rather than the asymptotic quasicrystalline exponent. Please provide a stability analysis: fits over several time windows, comparison with the 31° approximant and/or the Koshino-Moon continuum model, system-size dependence, and an estimate of the contamination from the initial state. Without this, the 'first estimate' is a promising but unsecured transient.","section":"Chapter 6"},{"comment":"Chapter 6 is presented as a thesis chapter on magnetism in twisted bilayer graphene, but the text explicitly states (p. x) that it 'represents a work in progress' and that it was 'not possible with the available computational resources to replicate previous results in the literature with our formalism.' The numerical results shown in Figs. 6.1–6.3 are convergence failures or magnetizations that remain at the tolerance threshold. Including an unreplicated chapter in the manuscript undermines the 'unified picture' promised in the conclusions. The authors should either remove this chapter or reframe it as an explicit, self-contained negative result with a technical analysis of why convergence is not achieved (for example, KPM broadening versus the relevant gap scale, initial conditions, or mean-field instability). This is a self-acknowledged missing result, not a presentation issue.","section":null}],"minor_comments":[{"comment":"The sentence 'The increase in QM at lower disorder is expected for weakly disordered cases, since the cleaner the system, the longer the corresponding mean free path and localization length' appears to conflict with the delocalization interpretation and should be clarified. As written, it is difficult to see why a longer mean free path in a cleaner system implies an increased quantum metric at lower disorder.","section":"Sec. 5.2.5 / Fig. 5.5 (bottom)"},{"comment":"The simulation length is stated as L = 2084; since the text describes periodic boundary conditions and powers of two are standard in such calculations, please confirm whether this is a typo for L = 2048.","section":"Sec. 4.1"},{"comment":"The caption says 'Choice of parameters for our twisted bilayer graphene simulator,' but the section models a gate-defined Bernal bilayer with an artificial superlattice, not a twisted bilayer. Please correct the caption to avoid confusion.","section":"Table 4.1"},{"comment":"Notation is inconsistent for the diffusion coefficient: D(t) is sometimes the time-dependent quantity ½ dΔX²/dt and sometimes the asymptotic diffusion constant. Please define explicitly whether the plotted values are D(t = 600 fs) or the saturated plateau value, and use separate symbols for the two.","section":"Multiple sections"},{"comment":"The code availability appendix would be more useful with a URL or repository identifier and a list of key dependencies/versions, rather than a general statement.","section":"App. C"}],"recommendation":"major_revision","confidential_remarks":"The stress-test note's concern is valid: the central claim is supported by a mean-free-path extraction that assumes a clean-system Fermi velocity and a 600 fs diffusive plateau, while the manuscript itself admits the diffusive regime is inaccessible at the lowest disorders. That said, the direct D(t) comparison at W=3γ0/4 versus W=3γ0/2 is an honest numerical result, and the authors are unusually explicit about limitations. This is not a reject; it is a request for new analysis. If the authors can supply a disorder-resolved Fermi velocity and demonstrate that the plateau is converged (or provide longer-time data), I would support acceptance. I would also ask the editor to consider whether the self-described work-in-progress Chapter 6 should remain in the version of record."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Read the thesis. The headline claim in MATBLG is a genuine numerical observation, but the mean-free-path numbers are softer than the abstract implies. The ℓ extraction uses the clean-system Fermi velocity and a 600-fs plateau; the largest weak-disorder values (ℓ≈76, 306 nm) are Fermi-golden-rule extrapolations from a single converged point, not simulated plateaus. The stress-test note is fair on this. Still, the qualitative effect is supported by the directly computed LDoS delocalization at W=3γ0/4→3γ0/2 and the recovery at W=2γ0. So the effect is likely real; the numbers are order-of-magnitude.\n\nWhat's genuinely new: the Ch. 4 transport fingerprints separating periodic, superperiodic, and quasiperiodic modulations—especially the tunable α∈[0.05,0.2] for the six-wave quasicrystalline potential—and the gate-defined quantum-dot array modeling in Sec. 4.3. The trilayer QC+MATBLG stability analysis in Sec. 5.4 is a useful addition. The spin-transport chapter (Ch. 7) is published work and the most polished. The thesis is candid about failure: Ch. 6 admits non-replication of prior mean-field magnetism, and Sec. 5.2.5 explicitly says the diffusive regime is inaccessible at weakest disorders.\n\nSoft spots, in order of severity. First, the missing disordered-system v_F. Near the flat band, disorder broadening changes the spectral composition at E=0; D=½⟨v²⟩τ_p, so an increase in D could partly reflect a higher effective velocity rather than a longer scattering time. Reporting a control with a disordered v_F would settle this. Second, the sub-ballistic exponent α≈0.84 is fitted from t<13 fs in a periodic approximant, with no error bars. That's preliminary. Third, the KPM broadening for the quantum metric (66 meV) is far larger than the few-meV flat-band width, so the geometry-transport link is suggestive. Fourth, the thesis relies on a single Slater-Koster parameter set; standard, but a sensitivity check would help.\n\nWho's it for: graduate students and researchers in moiré transport wanting a method-oriented compilation. I'd send it to a serious referee. The computational basis is substantial, the caveats are honest, and the qualitative delocalization claim deserves scrutiny. The extraction protocol needs a supplementary section with a disordered v_F and plateau-convergence tests. If those come out clean, the result stands.","headline":"The MATBLG disorder-induced delocalization is likely real but the ℓ numbers rest on clean-system v_F and a 600-fs plateau; the thesis is honest, useful, and worth refereeing.","tokens_in":56388,"tokens_out":5674,"would_cite":true,"duration_ms":57644,"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":"Moderate disorder can increase the mean free path of flat-band electrons in magic-angle twisted bilayer graphene.","keywords":["twisted bilayer graphene","magic angle","flat bands","random on-site disorder","mean free path","quantum metric","graphene quasicrystal","spin lifetime"],"falsifier":"Extend the disordered flat-band simulations to longer evolution times (beyond 600 fs) and larger systems at W ≈ γ0/4 and γ0/2; if the diffusion coefficient does not plateau or the effective Fermi velocity is renormalized by disorder, the reported 76 nm and 306 nm mean free paths would be extraction artifacts. On the experimental side, a transport measurement that tunes disorder in a magic-angle device and observes a non-monotonic flat-band mobility would confirm the mechanism.","tokens_in":55112,"feed_emoji":"🌀","tokens_out":7148,"duration_ms":73257,"temperature":0.7,"pith_summary":"This thesis uses large-scale real-space quantum-transport simulations to argue that geometry itself—twist angles, moiré superlattices, quasicrystalline order, and atomic corrugation—acts as an effective potential shaping electronic and spin transport in graphene stacks. Its central claim is that in magic-angle twisted bilayer graphene, within a bounded window of random on-site disorder where the flat bands remain visible, increasing disorder broadens those bands and reduces the scattering rate, so the mean free path grows instead of shrinking; only above that window do the flat bands dissolve and normal localization scaling resume. The same calculations give a first estimate of the sub-ballistic diffusion exponent (about 0.84) for dodecagonal graphene quasicrystals and show that this quasicrystalline order and its self-similar localization patterns are fragile against disorder and an added third layer. A separate study of suspended corrugated graphene shows that curvature-induced fluctuating spin-orbit fields can bring spin lifetimes down to nanoseconds even while charge transport stays nearly ballistic.","feed_headline":"Disorder can lengthen electron paths in twisted graphene","feed_subtitle":"Simulations show a finite disorder window where flat-band electrons travel farther as disorder grows.","key_machinery":"The carrying mechanism is the time-dependent diffusion coefficient D(E,t) = ½ d⟨ΔX²⟩/dt, computed by a Chebyshev-polynomial expansion of the real-space tight-binding Hamiltonian on million-atom samples. In a disordered system D(t) is expected to plateau; the plateau value yields the mean free path through the semiclassical relation ℓ = v_F τ_p with τ_p = 2D/v_F². The paper's main step is to identify that at the magic-angle flat band this plateau rises as the random disorder W is increased from 3γ0/4 to 3γ0/2, and to explain it as disorder-induced band broadening that reduces the scattering rate. The quantum metric is obtained from an optical-conductivity sum rule, G_xx ∝ ∫ dω Re σ_xx(ω)/ω, w","core_discovery":"This thesis establishes, through large-scale real-space quantum-transport simulations of atomistically relaxed structures, that in magic-angle twisted bilayer graphene the flat bands are not only localization-prone but also respond counterintuitively to disorder. For random on-site disorder strengths in a finite window—roughly W ≈ 3γ0/4 to 3γ0/2, with flat-band features washed out by W ≈ 2γ0—the mean free path at charge neutrality increases with disorder rather than decreasing. The proposed mechanism is that disorder broadens the very narrow flat bands, lowering the density of available final states and hence the scattering rate, while delocalizing the real-space wave functions away from the","pith_inferences":["Inference: The disorder window studied is comparable to charge inhomogeneity from common substrates, so sample-to-sample variations in correlated-phase transport may be partly a disorder-window effect rather than intrinsic physics.","Inference: The same band-broadening mechanism might be sought in other flat-band moiré systems, such as transition-metal dichalcogenide heterobilayers, where an analogous disorder window could produce disorder-enhanced mobility testable in gated devices.","Inference: The quantum-metric–mean-free-path correlation suggests that disorder could act as a dial for superfluid weight in moiré superconductors; optical-conductivity measurements at controlled disorder would provide a direct test.","Inference: Because the quasicrystal sub-ballistic exponent is extracted from a very short time window, extending the calculation to larger approximants or to true 30-degree samples with absorbing boundaries would test whether the power law persists or is a short-time transient."],"forward_implications":["If the disorder-induced delocalization is real, the flat-band mean free path in magic-angle twisted bilayer graphene is non-monotonic: it rises with disorder up to about W ≈ 3γ0/2, then falls, implying a noise-tolerance window for flat-band transport.","The accompanying rise in the quantum metric extracted from optical conductivity makes the same single-particle mechanism measurable through the optical sum, not just through dc transport.","The graphene quasicrystal's sub-ballistic exponent α ≈ 0.84 predicts anomalous, non-Drude optical and temperature scaling, useful as an experimental fingerprint of quasicrystalline order.","Quasicrystalline resonances are fragile: they are destroyed by moderate disorder and by proximity to a third layer, so observing them requires very clean, isolated 30-degree interfaces.","Curvature-induced Rashba fields set a nanosecond-scale upper bound on spin lifetimes in suspended graphene, even where charge mean free paths remain long."],"fun_headline_variants":["Disorder boosts electron travel in twisted graphene","Twisted graphene: more disorder, longer paths","Counterintuitive: disorder lengthens electron path","Flat bands: disorder enhances mean free path","In twisted graphene, messiness helps electrons go far"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The mean free paths are extracted by assuming that the diffusion coefficient reaches a true plateau within the 600 fs simulation window and that the clean-system Fermi velocity remains the correct conversion factor; for the weakest disorders that plateau is not actually reached, so the largest reported mean free paths rest on a perturbation-theory extrapolation rather than a directly observed diffusive regime.","fun_headline_variants_meta":{"raw":{"variants":["Disorder boosts electron travel in twisted graphene","Twisted graphene: more disorder, longer paths","Counterintuitive: disorder lengthens electron path","Flat bands: disorder enhances mean free path","In twisted graphene, messiness helps electrons go far"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000109,"raw_usage":{"total_tokens":900,"prompt_tokens":774,"completion_tokens":126,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":518,"completion_tokens_details":{"reasoning_tokens":56}},"tokens_in":518,"tokens_out":126,"duration_ms":2624,"temperature":1.0,"reasoning_tokens":56,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-01T02:32:14.033254+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Extend the disordered flat-band simulations to longer evolution times (beyond 600 fs) and larger systems at W ≈ γ0/4 and γ0/2; if the diffusion coefficient does not plateau or the effective Fermi velocity is renormalized by disorder, the reported 76 nm and 306 nm mean free paths would be extraction artifacts. On the experimental side, a transport measurement that tunes disorder in a magic-angle device and observes a non-monotonic flat-band mobility would confirm the mechanism.","supporting_citations":[],"review_version":1}