{"id":"bd477ae5-0ad5-4219-8a77-44c6f3cf534c","arxiv_id":"2608.02916","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Finite twisted bilayer graphene flakes show GOE chaotic spectra at any tested rotation angle, and the Thouless energy scaling indicates the chaos comes from border scattering rather than bulk mechanisms.","lead":"This preprint reports numerical evidence that finite twisted bilayer graphene flakes show quantum chaos at all tested rotation angles, commensurate or not. It argues that the chaos comes from scattering at the flake borders, based on how the Thouless energy scales with system size.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Uncontrolled random interlayer sliding may be the actual source of both the GOE statistics and the E_Th~1/sqrt(N) scaling, so the border-induced scattering claim is underdetermined.","rationale":"The reader's weakest_assumption identifies the random sliding as a possible bulk scattering source that could produce both the GOE statistics and the ~1/sqrt(N) scaling. This is precisely the load-bearing point in my reading. The paper's central claim is that finite TBG flakes are chaotic and that the physics is dominated by border-induced scattering. For that claim to hold, the sliding must not itself be the cause of chaos or of the observed E_Th scaling. Figure 3 directly demonstrates that sliding can turn an integrable system chaotic, which raises a concrete alternative explanation. The E_Th scaling in Figure 5 is based on a small number of systems, a limited N range, and no ensemble averaging, so it cannot distinguish border-induced from sliding-induced mean-free-path scaling. My proposed test—varying the sliding vector and using a no-sliding symmetry-breaking control—would settle the question. If the results are robust to these changes, the claim is well supported; if not, the conclusion needs to be reframed. The rest of the paper, including the tight-binding model and the K-correlator method, is reasonable and the numerical data are consistent with the reported behavior, so a conditional verdict remains appropriate. The reader's verdict of CONDITIONAL is therefore unchanged by this stress-test.","tokens_in":889,"tokens_out":830,"duration_ms":84172,"concrete_test":"For a fixed large angle (e.g., θcom=21.8°), compute the adjacent-gap ratio r and the Thouless energy E_Th for at least five independent random sliding vectors at each system size, and for a no-sliding control using a symmetry-breaking flake shape (e.g., a stadium-like cut). If ⟨r⟩ remains near 0.536 and the E_Th exponent versus log N remains near -0.5 both across sliding realizations and in the no-sliding control, the border attribution is supported. If the GOE statistics or the -0.5 exponent vanish, or vary by more than the quoted error bars (±0.10 and ±0.12) across sliding choices, the sliding is the dominant source and the conclusion must be revised.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Section II states that a random sliding between layers is introduced in all cases unless otherwise stated and that no averaging over slidings is performed. Figure 3 shows that a circular θ=0 flake without sliding is integrable (Poisson-like r), while in Figure 2(a) the same geometry with sliding yields GOE statistics. The sliding alone can therefore generate the spectral signature used to infer chaos. The Thouless energy analysis in Figure 5, the only quantitative evidence for the border mechanism, also includes this sliding. If the sliding acts as a bulk perturbation whose effective scattering mean free path grows with the system size—for example, a weak disorder whose mean free path is not smaller than the sample—then E_Th~1/L~1/sqrt(N) would follow exactly as observed, with no physical role for the boundaries. The paper does not test this alternative: there are no multiple-sliding ensembles, no no-sliding large-angle control with symmetry-resolved statistics, and the scaling range spans only about a factor of 4.5 in N. The central conclusion that borders dominate scattering is therefore conditional on the unverified assumption that the sliding is spectrally irrelevant beyond breaking discrete symmetries.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper studies quantum chaos and diffusion in finite twisted bilayer graphene (TBG) flakes using a tight-binding model with open boundary conditions. The authors compute adjacent-gap-ratio statistics for circular and commensurate-supercell flakes at commensurate and incommensurate angles, always including a random interlayer sliding, and report GOE-like statistics across angles. They then extract the Thouless energy from the density-density correlator K(E) and study its scaling with system size for a θ≈21° incommensurate circular flake and a θ_com=21.8° commensurate supercell, finding E_Th ~ N^{-0.6}. They interpret this as E_Th ~ 1/L, hence a mean free path l~L, and conclude that border-induced scattering is the dominant single-particle chaotic and diffusive mechanism in the studied finite TBG systems.","tokens_in":12567,"tokens_out":7720,"duration_ms":70040,"significance":"If correct, this work identifies borders as a strong, angle-independent single-particle scattering mechanism in finite TBG flakes, which is directly relevant to transport measurements in mesoscopic twisted graphene. The paper's strengths are its systematic application of established spectral tools (GOE adjacent gap ratios and the K-correlator Thouless energy) to large tight-binding systems, including both commensurate and incommensurate geometries, and its explicit comparison of different flake shapes. However, the central mechanistic conclusion is conditional on an untested assumption about the random interlayer sliding: the sliding is introduced in all calculations and is not averaged over, and a control shown in the paper demonstrates that sliding alone can turn an integrable spectrum into a GOE-like one. The scaling analysis that supports the border-scattering claim is based on a short range of system sizes and a single sliding realization. The paper is therefore suggestive rather than conclusive, and the proposed mechanism needs additional controls before the central claim is established.","major_comments":[{"comment":"The load-bearing assumption that the random sliding only breaks spurious symmetries without altering the scattering mechanism is not tested. In Section III the authors state that 'a random sliding between layers is introduced for all cases unless otherwise stated' and that they are 'without averaging over different slidings'. Figure 3 shows that a θ=0 circular flake without sliding is integrable, while Figure 2(a) shows the same geometry with sliding yields GOE-like statistics. This directly demonstrates that the sliding alone can generate the spectral signature used to infer chaos. Moreover, the no-sliding θ=0 control is not symmetry-resolved; for a highly symmetric circular flake, the full-spectrum adjacent-gap-ratio can be biased by degeneracies and by independent angular-momentum sectors, so the comparison conflates symmetry breaking with the generation of chaos. To support the paper's interpretation, the authors should: (i) average r and K over several independent slidings, (ii) compute symmetry-resolved level statistics for no-sliding flakes, and (iii) include a large-angle no-sliding control with explicit symmetry resolution.","section":"Section III, Figs. 2 and 3"},{"comment":"The central scaling result is fitted from five system sizes per geometry over only a factor of about 4.5 in N, and it uses a single sliding realization. The two criteria for E_Th provide a crude systematic uncertainty, but they do not include realization-to-realization fluctuations arising from the random sliding. With slopes −0.59±0.10 and −0.64±0.12, the data are consistent with the claimed E_Th~1/√N scaling, but the confidence intervals do not exclude a nearby exponent, and the error bars are not statistical errors over disorder. The fitted exponent alone therefore cannot distinguish a boundary-induced mechanism from a weak bulk disorder whose effective mean free path grows with L. The statement in Section IV B that the chaotic behavior is 'caused by borders' is underdetermined by the presented data. The authors should provide ensemble averages over slidings, report the distribution of fitted slopes, and ideally compare against a bulk-disorder model with a controlled mean free path.","section":"Section IV B, Fig. 5"},{"comment":"The theoretical discriminant between intrinsic and boundary-induced chaos is presented as E_Th~1/L² versus E_Th~1/L, but the random sliding itself constitutes a bulk perturbation. A weak bulk disorder with mean free path l comparable to L would also produce E_Th~1/L~1/√N, with no role for boundaries. The paper does not provide any independent estimate of the scattering length associated with the sliding, and it does not vary the sliding amplitude or its spatial correlation length. Because Figure 3 shows that the sliding can by itself change the level statistics from Poisson-like to GOE-like, the alternative interpretation that the sliding acts as an effective bulk scattering source is not merely a formal possibility; it is a concrete competing explanation that the current data do not exclude.","section":"Section IV, Eq. (7)"}],"minor_comments":[{"comment":"There are several typographical and grammatical errors: the abstract contains 'as a experimentally relevant single-particle scattering mechanism' (should be 'an experimentally relevant'), and the Introduction contains 'TBG cavities where also found' (should be 'were also found') and 'a a dominant trigonal-warping mechanism'.","section":"Abstract and Introduction"},{"comment":"References [65] and [69] appear to be the same article (Pino, Kravtsov, Altshuler, and Ioffe, Physical Review B 96, 214205 (2017)), and References [19] and [70] also appear to be the same ACS Nano article. Please consolidate the duplicates.","section":"References"},{"comment":"The definition of K(E) uses δ(E−E_α+E_β), which is confusing because E is later plotted on a negative logarithmic scale; the text should state explicitly that E is the energy difference between eigenstates and define the normalization used for the GOE plateau value I_2/3.","section":"Section IV A, Eq. (8)"},{"comment":"The caption refers to a 'blue horizontal line is the average value of I^2/3', but the symbol I_2 is not defined in the text; please define I_2 and explain how its average is computed over the eigenstates.","section":"Section IV A, Fig. 4"},{"comment":"Section III states 'we therefore do not find pseudo-integrable statistics as in Ref. [45]', but the Conclusions suggest that the small dimensionless conductance of the commensurate flake 'may be related to the non-integrable (but neither chaotic) character' of the rhombus billiard. This apparent tension should be reconciled either by softening one of the statements or by adding a quantitative comparison with the pseudo-integrable expectation.","section":"Section III vs. Section V"},{"comment":"The text reports the number of system sizes only through the log(N) axis; it would improve transparency to state the actual number of data points per geometry and the corresponding linear sizes L, especially because the scaling conclusion rests on only a short range of N.","section":"Section IV B, Fig. 5"}],"recommendation":"major_revision","confidential_remarks":"The manuscript's central claim is potentially publishable, but the uncontrolled random-sliding protocol is its main vulnerability. I would ask the editor to require the authors to add a sliding-ensemble average, symmetry-resolved no-sliding controls, and a direct test of whether the E_Th~1/√N scaling persists across multiple sliding realizations. It would also strengthen the paper to include a data availability statement, since the scaling fit relies on a small number of points and a single realization."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The thing to know: this paper argues that finite twisted bilayer graphene flakes are quantum chaotic at essentially every twist angle, and that at large angles the chaos comes from border-induced scattering, diagnosed via E_Th scaling as 1/sqrt(N). The spectroscopic evidence for GOE level statistics is honestly presented and convincing for the geometries tested. The Thouless-energy diagnostic is a genuinely useful addition to the TBG-chaos toolkit, and the paper is clear about what it did and what the fits are. The self-citation for the K-correlator method is legitimate; that method is established and not rigged.\n\nThe problem is the random interlayer sliding. The paper says \"a random sliding between layers is introduced for all cases unless otherwise stated\" and, in the same breath, says the Hamiltonian \"does not contain any random quantity\" because they do not average over slidings. That is inconsistent. Figure 3 shows that sliding alone converts an integrable θ=0 flake into a GOE-looking one. So sliding is not spectrally neutral; it can generate the signal used to infer chaos. The central claim that borders dominate therefore needs a control the paper does not provide: multiple sliding realizations, or a no-sliding large-angle run with symmetry-resolved statistics. A single random shift is a global perturbation, not uncorrelated disorder, so the stress-test scenario of a bulk mean free path scaling with L is not obviously what is happening; but the paper does not rule it out.\n\nThe E_Th scaling itself is also thinner than the abstract implies. Five system sizes per geometry, spanning about a factor of 4.5 in N, give slopes -0.59 and -0.64 with ±0.1 errors. That excludes -1 and is consistent with -0.5, but it does not discriminate between boundary scattering and any mechanism with l ~ L. The different E_Th scales and dimensionless conductances for commensurate versus incommensurate flakes are interesting but also based on single realizations and the same limited fits.\n\nVerdict: the paper deserves a serious referee, but as it stands the central mechanistic conclusion is conditional. The fix is straightforward: average over slidings, show a no-sliding large-angle control, and report E_Th for more sizes. If those controls hold, this becomes a solid contribution. I would not cite the border-scattering claim in its current form, but I would probably bring it to a group meeting to argue about the confound.\n\nRecommendation: send it to peer review, but with the expectation of major revision focused on the sliding controls.","headline":"Finite TBG flakes show solid GOE chaos, but the border-scattering conclusion rests on an uncontrolled random sliding that the paper never averages over.","tokens_in":13079,"tokens_out":2998,"would_cite":false,"duration_ms":31504,"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":"Finite flakes of twisted bilayer graphene are quantum chaotic at both commensurate and incommensurate angles, and their single-particle diffusion at intermediate to large twist angles is dominated by scattering off the flake borders…","keywords":["twisted bilayer graphene","quantum chaos","Thouless energy","level statistics","random matrix theory","edge scattering","diffusion","moiré pattern"],"falsifier":"Repeat the level-statistics and Thouless-energy analysis over many independent random interlayer slidings for the same flake sizes and angles: if the ensemble-averaged gap ratio moves toward the Poisson value or the $E_{\\mathrm{Th}}$ slope moves from about $-0.5$ toward $-1$, the border-induced chaos claim would be refuted. A second check is to compare smooth-edge flakes with rough-edge flakes at fixed angle and size; reduced level repulsion in the smooth case would confirm the border mechanism.","tokens_in":12130,"feed_emoji":"🌀","tokens_out":11279,"duration_ms":85672,"temperature":0.7,"pith_summary":"The paper argues that finite flakes of twisted bilayer graphene are generically quantum chaotic at both commensurate and incommensurate angles, and that the dominant source of chaos and diffusion at intermediate to large twist angles is scattering off the flake borders, not the moiré pattern. It supports this by showing that the adjacent gap ratios of the energy spectrum match the Gaussian orthogonal ensemble (GOE) prediction, and that the Thouless energy (the energy at which eigenstate correlations become GOE-like) scales as $1/\\sqrt{N}$, which corresponds to a mean free path of the order of the system size. If the claim holds, single-particle transport in twisted bilayer graphene flakes is diffusive because of edge scattering, with direct consequences for nanoconstriction and conductance measurements.","feed_headline":"Borders make twisted graphene flakes chaotic","feed_subtitle":"Level statistics match random-matrix chaos; the Thouless energy scales with size, pointing to edge-driven diffusion.","key_machinery":"The load-bearing object is a real-space tight-binding Hamiltonian with distance-dependent $\\pi$ and $\\sigma$ hopping terms, diagonalized for open-boundary flakes. To detect level repulsion, the paper uses the adjacent gap ratio $r$, averaged over states between 3 and 5 eV, and compares its average and full distribution to the GOE predictions. To measure diffusion, it computes the density-density correlator $K(E)$ between eigenstates and defines the Thouless energy $E_{\\mathrm{Th}}$ as the scale at which $K$ reaches the GOE value $I_2/3$; the slope of $\\log E_{\\mathrm{Th}}$ versus $\\log N$ then discriminates border-induced chaos ($\\sim 1/\\sqrt{N}$) from intrinsic bulk scattering ($\\sim 1/N$). A random interlayer sliding is introduced in the main runs to break accidental symmetries and make the lattice more realistic.","core_discovery":"At energies around 4 eV, the level statistics of circular flakes and commensurate-supercell flakes with up to about 31,000 atoms show level repulsion consistent with the Gaussian orthogonal ensemble, independent of whether the twist angle is commensurate or incommensurate; only the $\\theta_{\\mathrm{com}}=1.05^\\circ$ case deviates slightly. The density-density eigenstate correlator reaches the GOE plateau at a Thouless energy that falls as $E_{\\mathrm{Th}}\\sim N^{-0.59}$ for incommensurate flakes and $N^{-0.64}$ for commensurate flakes, both compatible with $E_{\\mathrm{Th}}\\sim 1/\\sqrt{N}$. Since $E_{\\mathrm{Th}}$ is inversely proportional to the diffusion time and $N$ is proportional to the area, that scaling implies a mean free path of the order of the linear system size, so the scattering that produces chaos and diffusion comes from the borders rather than from an intrinsic bulk mechanism.","pith_inferences":["Because the random interlayer sliding is applied without averaging over realizations, an ensemble average over slidings would be the cleanest check that the GOE statistics and the $\\sim 1/\\sqrt{N}$ slope are not artifacts of a single disorder configuration.","If borders are the cause, flakes with atomically smooth edges at fixed angle and size should show weakened level repulsion and a drift of $\\langle r\\rangle$ toward the Poisson value.","The same density-density correlator method could map border-induced versus intrinsic diffusion in other moiré materials, such as twisted transition-metal dichalcogenides, where the interlayer coupling is different.","At magic-angle sizes, the large unit cell should produce a regime where the $E_{\\mathrm{Th}}$ versus $N$ slope is steeper; locating that crossover would separate moiré-cell chaos from edge chaos experimentally."],"forward_implications":["If $E_{\\mathrm{Th}}\\sim 1/\\sqrt{N}$, then transport through twisted bilayer graphene nanoconstrictions is limited by edge scattering; ballistic behavior of the kind seen in monolayer graphene should be much harder to achieve.","The dimensionless conductance $g=E_{\\mathrm{Th}}/\\Delta$ sits near 1 for commensurate rhombic flakes and above 1 for incommensurate circular flakes, so commensurability can change the metallic character even when the angle and size are similar.","At small twist angles, where the moiré unit cell approaches the system size, chaos can also arise from mixing of different band indices inside the large unit cell, so border scattering is not the only route to chaos.","For systems larger than the mean free path, the paper expects a crossover from $E_{\\mathrm{Th}}\\sim 1/L$ to $E_{\\mathrm{Th}}\\sim 1/L^2$, which would expose intrinsic scattering lengths at other angles."],"supporting_citations":[{"why":"Supplies the distance-dependent tight-binding Hamiltonian used for all spectral and diffusion calculations.","marker":"[47]"},{"why":"Earlier study of chaos in TBG under twisted boundary conditions that this work contrasts with its open-boundary treatment.","marker":"[32]"},{"why":"Report of chaos in TBG cavities with trigonal warping and growing edge effects that motivates the border-scattering question.","marker":"[33]"},{"why":"Introduces the adjacent gap ratio used to measure level repulsion.","marker":"[53]"},{"why":"Provides the GOE average $\\langle r\\rangle\\approx0.536$ against which the numerical spectra are compared.","marker":"[54]"},{"why":"Defines the Thouless energy and its relation to the diffusion time, grounding the $E_{\\mathrm{Th}}$ interpretation.","marker":"[62]"},{"why":"Fixes the scaling connection between Thouless energy, dimensionless conductance, and localization.","marker":"[63]"},{"why":"Supplies the density-density correlator method used to extract the Thouless energy from eigenstates.","marker":"[65]"}],"fun_headline_variants":["Edge scattering drives chaos in twisted graphene flakes","Graphene flake chaos traced to boundary scattering","Border-induced chaos in twisted graphene","Random-matrix chaos emerges from flake edges","Edges, not bulk, drive graphene flake chaos"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The random interlayer sliding used to break symmetry and make the lattice realistic is assumed not to be a bulk scattering source that by itself produces the GOE statistics and the $1/\\sqrt{N}$ Thouless-energy scaling; the paper does not average over sliding configurations to rule this out.","fun_headline_variants_meta":{"raw":{"variants":["Edge scattering drives chaos in twisted graphene flakes","Graphene flake chaos traced to boundary scattering","Border-induced chaos in twisted graphene","Random-matrix chaos emerges from flake edges","Edges, not bulk, drive graphene flake chaos"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000646,"raw_usage":{"total_tokens":2923,"prompt_tokens":852,"completion_tokens":2071,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":468,"completion_tokens_details":{"reasoning_tokens":2001}},"tokens_in":468,"tokens_out":2071,"duration_ms":13521,"temperature":1.0,"reasoning_tokens":2001,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T14:55:09.303279+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Repeat the level-statistics and Thouless-energy analysis over many independent random interlayer slidings for the same flake sizes and angles: if the ensemble-averaged gap ratio moves toward the Poisson value or the $E_{\\mathrm{Th}}$ slope moves from about $-0.5$ toward $-1$, the border-induced chaos claim would be refuted. A second check is to compare smooth-edge flakes with rough-edge flakes at fixed angle and size; reduced level repulsion in the smooth case would confirm the border mechanism.","supporting_citations":[{"cited_title":"Fang and E","cited_arxiv_id":null,"evidence_quote":"Supplies the distance-dependent tight-binding Hamiltonian used for all spectral and diffusion calculations."},{"cited_title":"Koshino, New Journal of Physics17, 015014 (2015)","cited_arxiv_id":null,"evidence_quote":"Earlier study of chaos in TBG under twisted boundary conditions that this work contrasts with its open-boundary treatment."},{"cited_title":"Gonçalves, H","cited_arxiv_id":null,"evidence_quote":"Report of chaos in TBG cavities with trigonal warping and growing edge effects that motivates the border-scattering question."},{"cited_title":"Oganesyan and D","cited_arxiv_id":null,"evidence_quote":"Provides the GOE average $\\langle r\\rangle\\approx0.536$ against which the numerical spectra are compared."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Defines the Thouless energy and its relation to the diffusion time, grounding the $E_{\\mathrm{Th}}$ interpretation."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Fixes the scaling connection between Thouless energy, dimensionless conductance, and localization."}],"review_version":2}