{"id":"70f766d8-cd4b-4431-a00f-8097b826cd0d","arxiv_id":"1908.05941","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":1,"one_line_summary":"In relaxed twisted bilayer graphene at angles near 0.2 degrees, an interlayer bias reveals quasi-one-dimensional states confined to the network of stacking-boundary channels.","lead":"This paper extends a continuum model of twisted bilayer graphene to very small twist angles, including lattice relaxation. It shows that applying an electric bias between the layers exposes a network of one-dimensional conducting channels along the boundaries between different stacking regions.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The channel-state claim rests on a 75-harmonic continuum truncation that is not validated at 0.221°, where the moiré length is ~64 nm and the channel width is a small fraction of that period; without a convergence check, the bias-liberated states could be artifacts of the cutoff.","rationale":"The paper's strongest claim is a single-particle statement about the existence of bias-insensitive channel states at θ=0.221°. For that claim to hold, the model must faithfully represent the relaxation-induced channel network at that angle. The weakest point is exactly the one the reader flags: the model relies on a harmonic truncation (75 G-vectors, 271 basis states) whose accuracy at θ≈0.2° is inherited from [33] rather than demonstrated here. The central claim lives precisely in the regime where the fixed cutoff is most questionable, because the channel width is a small fraction of the 64 nm moiré period and the projected couplings are angle-dependent (Fig. 3b). Other possible objections, such as the subjectivity of channel-state identification in Fig. 5, the neglect of intervalley coupling, and the fixed interlayer distance, are acknowledged in the text or are secondary. The harmonic-convergence issue is more load-bearing because it concerns the numerical representation of the very object (the channel) that the paper claims to discover. If the 75-vector model is under-resolved, the bias-liberated states and their wave-function localization could change or disappear with more harmonics; if it is converged, the central claim is supported. Because the reader already made this concern the basis of a CONDITIONAL verdict, my pass does not change the verdict: it remains conditional pending the convergence check.","tokens_in":10972,"tokens_out":10070,"duration_ms":111514,"concrete_test":"Recompute Figs. 5 and 6 at θ=0.221° for V=0 and V=100 meV with systematically increased harmonic sets (e.g., 127, 217, and 331 G-vectors with correspondingly larger bases). Compare (i) the energies of the lowest bias-insensitive bands at Γ, K, and M; (ii) the density of states at the Fermi level for V=125 meV; and (iii) the fraction of wave-function weight within a corridor of width about 5 nm around the AB/BA domain walls at the M point. If these quantities change by more than about 10% when going from 75 to 217 harmonics, the truncated model is not converged and the channel-state claim is not established; if they are stable, the reader's conditional concern is resolved.","verdict_should_be":"UNCHANGED","load_bearing_attack":"At the marginal angle used for the central claim (θ=0.221°), the moiré period is L≈64 nm. The Methods section fixes the interlayer-coupling projection at 75 G-vectors and a 271-state basis, and the method is validated only by reference to earlier work [33], which did not reach this angle. No convergence check with respect to the harmonic cutoff is shown at 0.221°. This matters because the central claim concerns relaxation-induced AB/BA domain-wall channels whose width is a small fraction of L (visible in Fig. 3a), and the paper itself states that the projected couplings 'now do change with the twist-angle' (Sec. I, Fig. 3b/S5). The claim that under an interlayer bias a bias-insensitive set of quasi-1D channel states survives is therefore conditional on those channel-resolving coupling components being captured within the 75-vector cutoff. If the relevant Fourier components lie at larger G, the apparent channel states in Figs. 5 and 6 could change or disappear with more harmonics. The 1.3× rescaling of t(r12) to place the magic angle at 1.05° does not resolve this: it calibrates a low-harmonic parameter, not the high-harmonic content needed to represent narrow channels.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper generalizes the continuum model of twisted bilayer graphene to include lattice relaxation and a large number of interlayer harmonics, and applies it to marginal twist angles near 0.2 degrees. The authors compute relaxed atomic structures using the LCBOPI+KC potential, project the interlayer hopping onto a 75-G-vector basis, and study the effect of an interlayer bias on the low-energy spectrum. Their central claim is that at small angles an applied bias gaps out states localized in AB-stacked regions, liberating a set of bias-insensitive quasi-one-dimensional states confined to the relaxation-induced channels separating AB and BA domains. They compare their results with the Efimkin-MacDonald helical network model and the Zhang-MacDonald-Mele domain-wall model, and also provide wave-function plots showing channel localization and interlayer beats. The main conclusion is that the low-energy bands and density of states tend to a bias-independent limit, consistent with a situation in which only channel states remain.","tokens_in":11235,"tokens_out":4166,"duration_ms":40463,"significance":"If the central claim holds, the paper provides a concrete and testable prediction: at marginal twist angles around 0.2 degrees, an interlayer bias can isolate a robust network of quasi-one-dimensional channel states that are insensitive to the bias and confined to AB/BA domain walls. This is directly relevant to recent experiments on marginally twisted bilayers (Yoo et al., Xu et al., Rickhaus et al.) and goes beyond earlier model studies by including realistic relaxation and a more complete interlayer-coupling expansion. Strengths of the paper include the extensive numerical spectra, density-of-states plots, and wave-function visualizations, and the fact that the central bias-liberation prediction is not used to fit any constant, giving it independent predictive content. The method extends the authors' prior continuum-projection framework to smaller angles, and the robustness of the qualitative behavior is asserted across alternative relaxation and hopping models, though the supporting data for those alternatives is not shown.","major_comments":[{"comment":"","section":"Methods and Sec. I (Fig. 3b, Fig. S5)"},{"comment":"","section":"Sec. I, Eq. (1)"},{"comment":"","section":"Sec. I, Fig. 5 and Conclusions"}],"minor_comments":[{"comment":"","section":"Sec. I, after Eq. (1)"},{"comment":"","section":"Fig. 2a and Fig. S5"},{"comment":"","section":"Sec. I, first Results paragraph"},{"comment":"","section":"Sec. I, Fig. 4 discussion"}],"recommendation":"major_revision","confidential_remarks":"The paper is largely a direct application of the authors' own continuum-projection framework from Ref. [33], and the self-citation density is high; the incremental novelty lies in the application to marginal angles and bias-induced channel states, which is of topical interest. The main risk is the absence of a harmonic-truncation convergence check at the central angle of 0.221°, which is load-bearing for the paper's central claim. The manuscript would be publishable after this convergence test is provided and the channel-state classification is made quantitative. I also note that the robustness claim about alternative models is asserted without showing data, which should be addressed in revision."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The genuinely new thing here is that Walet and Guinea take their earlier continuum-projection framework, add lattice relaxation and a large number of harmonics, and push it down to 0.221°, where the moiré unit cell has ~268,000 atoms. They then show that an interlayer bias separates a set of quasi-1D channel states that are largely bias-insensitive. The wave-function analysis—chiral interlayer beats, concentration on AB/BA domain walls, comparison with the Efimkin-McDonald and Zhang-McDonald-Mele models—is real evidence, not just a hand-wave. The paper is also honest about what it does not include (intervalley coupling at larger angles, full many-body screening) and about the subjective element in coloring bands as channel or non-channel states. That gets credit.\n\nThe soft spots are real but not fatal. The largest is the missing convergence check on the 75-G-vector cutoff at the central angle. With a moiré period around 64 nm and channel widths that are a small fraction of that, the Fourier content needed to resolve narrow channels is exactly where a truncation error would bite. The paper validates the method at larger angles, but the regime where the central claim lives is new, and the projected couplings are angle-dependent—so a plot showing the spectrum stable as the harmonic number goes from, say, 75 to 150+ would close the biggest gap. Without it, the bias-liberated states could in principle be an artifact of the cutoff. The 1.3× rescaling of the interlayer hopping to put the magic angle at 1.05° is a free parameter, though the central prediction is not fitted to anything and so has independent content. The channel-state identification is partly by eye, and the lack of code or data makes reproduction harder. Those are minor-to-moderate, not enough to sink the paper.\n\nOverall, the central claim is plausible and the method is a genuine step forward for this subfield. The paper deserves a serious referee; the revision should ask for a harmonic-convergence test at small angles and, if possible, a more automated way to label channel states.","headline":"A useful extension of the continuum model to relaxed marginal-angle TBG, with credible bias-liberated channel states; the missing harmonic-convergence check at 0.221° is the main thing standing between this and a fully convincing claim.","tokens_in":11740,"tokens_out":1537,"would_cite":true,"duration_ms":18176,"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":"Applying an interlayer electric bias to relaxed twisted bilayer graphene at marginal angles frees a stable network of quasi-one-dimensional channel states, leaving a bias-independent low-energy spectrum.","keywords":["twisted bilayer graphene","marginal twist angle","lattice relaxation","one-dimensional channels","interlayer bias","continuum model","chiral edge states","moiré"],"falsifier":"A direct tight-binding or other numerically exact calculation for a relaxed 0.2-degree bilayer that shows the low-energy spectrum remaining strongly bias-dependent at high bias—or a scanning probe experiment at 0.2 degrees with bias that finds bias-sensitive states coexisting with the channel network—would contradict the claim that the low-energy limit is bias independent and channel-only.","tokens_in":10770,"feed_emoji":"⚡","tokens_out":4686,"duration_ms":41142,"temperature":0.7,"pith_summary":"This paper argues that in relaxed twisted bilayer graphene at marginal twist angles near 0.2 degrees, an applied interlayer electric bias separates the electronic spectrum into two parts: states tied to AB-stacked regions, which shift in energy with the bias, and a stable network of quasi-one-dimensional states confined to the relaxation-induced channels between AB and BA regions, which stay put. The authors show that as the bias grows, the low-energy bands and density of states converge to a bias-independent limit, which they read as the signature of a standalone channel network. They establish this using a generalized continuum model that includes lattice relaxation and a large number of interlayer harmonics, making small-angle calculations tractable without a costly tight-binding treatment of hundreds of thousands of atoms. If correct, the work identifies a concrete experimental route—applying a bias to marginally twisted bilayers—for isolating and studying one-dimensional chiral channel states.","feed_headline":"Bias reveals 1D channels in marginally twisted graphene","feed_subtitle":"At 0.2° twists, low-energy states become bias-independent—the signature of a liberated network of one-dimensional channels.","key_machinery":"The machinery is a generalized continuum model for twisted bilayer graphene that starts from a relaxed atomic structure (computed with the LCBOPI+KC potential), projects the interlayer hopping onto Bloch states, and includes 75 reciprocal-lattice harmonics rather than the usual three of the Bistritzer-MacDonald model. This makes small-angle calculations feasible with accuracy comparable to tight binding at larger angles, despite over 268,000 atoms in the moiré unit cell at 0.221 degrees. The paper uses an exponential Koster-Slater interlayer hopping, scaled so the first magic angle sits near 1.05 degrees, and applies an interlayer bias as a probe to distinguish bias-sensitive AB-region states from bias-insensitive channel states.","core_discovery":"The central discovery is that lattice relaxation at marginal twist angles creates a network of narrow channels separating AB and BA stacked domains, and that these channels support quasi-one-dimensional states that are insensitive to an interlayer bias. In the authors' continuum calculations for a relaxed bilayer at 0.221 degrees, the low-energy bands and density of states approach a bias-independent limit as the bias is increased from 0 to 200 meV, whereas unrelaxed layers show little spectral change and larger angles show only weak sensitivity. The bias liberates the channel states by gapping the AB-region states, leaving a stable network whose wave functions concentrate on the channels with chiral interlayer beats. The paper also finds that a single pair of nearly straight, linearly dispersing bands persists, and that crossings occur at K, K' and near M, differing from simpler one-channel models. The authors conclude that the occurrence of channel states is a robust consequence of relaxation in bilayer graphene, with details depending on the theoretical model and experimental environment.","pith_inferences":["The bias-independent limit implies that the channel network could serve as a platform for studying one-dimensional conductance quantization and interferometry, since the AB-region states can be gapped away without destroying the channels.","If the channel states survive at finite temperature and moderate disorder, the marginal-angle regime with bias might exhibit transport dominated by the channel network's topology, potentially mimicking features of electrical networks in other moiré systems.","The method's reliance on a relaxed structure means that the channel network's geometry—and thus the device behavior—could be tunable by strain or substrate encapsulation, a testable prediction beyond the free-standing planar case."],"forward_implications":["At angles below about 0.5 degrees, applying an interlayer bias of order 100 meV should drive the system into a regime where only channel states remain near the Fermi energy, enabling transport through a quasi-1D network.","The generalized continuum model with many harmonics extends quantitative band-structure calculations to twist angles as small as 0.1 degrees, where the moiré unit cell contains about a million atoms, at no significant increase in computational cost.","The channel states show a chiral structure with interlayer beats, so their detection in scanning tunneling microscopy would provide a direct signature of the relaxation-induced channels.","Only one pair of nearly straight, linearly dispersing bands survives, with crossings at K, K' and near M, suggesting that simplified one-channel models miss important scattering and energy-dependent velocity effects."],"supporting_citations":[{"why":"Supplies the continuum projection method and the relaxed LCBOPI+KC atomic structures that the paper generalizes to small angles.","marker":"[33]"},{"why":"The standard Bistritzer-MacDonald continuum model that this paper extends by adding more interlayer harmonics and relaxation.","marker":"[36]"},{"why":"The Efimkin-MacDonald single-channel helical network model that the paper compares against and finds quantitative differences from.","marker":"[30]"},{"why":"Earlier calculation of helical networks in twisted bilayer graphene under interlayer bias, without lattice relaxation, providing the baseline for the bias effect.","marker":"[29]"},{"why":"Lattice relaxation study that informs the picture of AB/BA domain growth and channel formation.","marker":"[32]"},{"why":"Experimental study of marginal-angle twisted bilayers down to 0.1 degrees, providing the experimental context for the angle range.","marker":"[3]"},{"why":"Experimental observation of giant oscillations in a triangular network of one-dimensional states in marginally twisted graphene, motivating the channel-network picture.","marker":"[4]"},{"why":"Provides the valley Chern numbers and boundary modes for chiral edge states at AB/BA interfaces, used in the comparison of channel-state structure.","marker":"[11]"}],"fun_headline_variants":["Relaxation liberates bias-stable channels in twisted bilayer graphene","Bias-independent 1D channels emerge in relaxed twisted bilayer graphene","Marginal twist graphene shows robust 1D channels from lattice relaxation","Relaxed twisted bilayer graphene: 1D channels ignore bias"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The calculations assume that the relaxed atomic structure from the LCBOPI+KC potential, together with the 75-vector harmonic truncation of interlayer coupling, correctly reproduces the low-energy electronic structure at angles near 0.2 degrees, without an independent check against a full tight-binding calculation at those angles.","fun_headline_variants_meta":{"raw":{"variants":["Relaxation liberates bias-stable channels in twisted bilayer graphene","Bias-independent 1D channels emerge in relaxed twisted bilayer graphene","Marginal twist graphene shows robust 1D channels from lattice relaxation","Relaxed twisted bilayer graphene: 1D channels ignore bias"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000198,"raw_usage":{"total_tokens":1314,"prompt_tokens":836,"completion_tokens":478,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":452,"completion_tokens_details":{"reasoning_tokens":403}},"tokens_in":452,"tokens_out":478,"duration_ms":4814,"temperature":1.0,"reasoning_tokens":403,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T12:59:40.204612+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A direct tight-binding or other numerically exact calculation for a relaxed 0.2-degree bilayer that shows the low-energy spectrum remaining strongly bias-dependent at high bias—or a scanning probe experiment at 0.2 degrees with bias that finds bias-sensitive states coexisting with the channel network—would contradict the claim that the low-energy limit is bias independent and channel-only.","supporting_citations":[{"cited_title":"Guinea and N","cited_arxiv_id":null,"evidence_quote":"Supplies the continuum projection method and the relaxed LCBOPI+KC atomic structures that the paper generalizes to small angles."},{"cited_title":"Bistritzer and A","cited_arxiv_id":null,"evidence_quote":"The standard Bistritzer-MacDonald continuum model that this paper extends by adding more interlayer harmonics and relaxation."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"The Efimkin-MacDonald single-channel helical network model that the paper compares against and finds quantitative differences from."},{"cited_title":"San-Jose and E","cited_arxiv_id":null,"evidence_quote":"Earlier calculation of helical networks in twisted bilayer graphene under interlayer bias, without lattice relaxation, providing the baseline for the bias effect."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Experimental study of marginal-angle twisted bilayers down to 0.1 degrees, providing the experimental context for the angle range."},{"cited_title":"Zhang, A","cited_arxiv_id":null,"evidence_quote":"Provides the valley Chern numbers and boundary modes for chiral edge states at AB/BA interfaces, used in the comparison of channel-state structure."}],"review_version":1}