{"id":"20046eef-1736-4628-89e7-e91ded738232","arxiv_id":"2506.02716","paper_version":2,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"By combining Fibonacci and Bronze Mean hopping sequences with a staggered Aubry-Andre-Harper potential, the authors find a window in which previously localized eigenstates become extended again.","lead":"This numerical study models a one-dimensional wire where hopping strengths follow a Fibonacci or Bronze Mean pattern and on-site energies follow the Aubry-Andre-Harper form. It reports that as modulation strength rises, states localize, then a subset extends again, a reentrant localization effect.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The thermodynamic-limit claim is under-supported: Fig. 4's N→∞ curve rests on an unspecified extrapolation and only three system sizes for the Bronze Mean case, so the reported finite NPR could be a slowly decaying finite-size tail.","rationale":"The reader's conditional verdict already flags robustness; I agree with the need for conditions. Their identified weakest assumption is the special choice b equal to the hopping symbol ratio. I view that as a model-design choice rather than a fatal flaw: the paper explicitly scans other irrational b and finds no RL, and the claim is about correlated hopping, for which this b is the natural definition of correlation. The more load-bearing issue is the unsupported N→∞ extrapolation in Fig. 4, because the headline claim explicitly asserts finite NPR in the thermodynamic limit. The text does not state the fitting model, error estimates, or whether the extrapolation is stable; with only three BM system sizes a sublinear decay could masquerade as saturation. This is testable with a scaling collapse and larger sizes. Since the reader's verdict is already CONDITIONAL and our concern reinforces that conditionality, no verdict change is needed.","tokens_in":12272,"tokens_out":7801,"duration_ms":95522,"concrete_test":"Recompute the RL-window average ⟨NPR⟩ as a function of N for both sequences, using exactly the stated eigenstate windows, and perform a documented extrapolation: plot ⟨NPR⟩ vs 1/N and vs 1/N^ω, fit a + b N^{-ω} with bootstrap errors, and require the intercept a to be positive at 3σ and stable when the next generation (Fibonacci N≈28658, BM N≈55800) is added. In addition, track the number of eigenstates in the window with IPR_n ≤ 2/N; this count should grow linearly with N. If the intercept is consistent with zero or the count saturates, the claimed thermodynamic RL phase is a finite-size artifact.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The strongest claim is that ⟨NPR⟩ stays finite as N→∞ inside the RL windows (Fibonacci λ≈1.55–2.1; BM λ≈0.7–1.3). The only direct evidence is Fig. 4 and the sentence 'we further extrapolate ⟨NPR⟩ in the thermodynamic limit.' No extrapolation form, no error bars, and no goodness-of-fit are given. For the BM sequence only N=1550, 5117, 16898 are shown; with three points a function like c N^{-α} (α<1) or c/log N can be mistaken for saturation. The averaging window itself is chosen after the RL regions are identified ('central 44%' for Fibonacci, 'central 18%' for BM), so ⟨NPR⟩ is conditioned on the phenomenon to be demonstrated. If the finite-N signal comes from a finite fraction of critical/near-extended states whose fraction decays with N, the intercept would be zero and the reentrant phase would not exist in the thermodynamic limit. The b-matching issue raised by the reader is real but less decisive: the authors explicitly test generic b and report monotonic localization, so the special b is a defensible model choice; the missing scaling analysis is the load-bearing weakness.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies a one-dimensional tight-binding chain with a staggered Aubry-André-Harper (AAH) on-site potential and off-diagonal hopping amplitudes arranged in Fibonacci or Bronze Mean (BM) quasiperiodic sequences. The authors compute eigenstate fractal dimensions, inverse participation ratio (IPR), normalized participation ratio (NPR), and a mixed-phase indicator η to argue that, as the quasiperiodic modulation strength λ increases, the system exhibits a reentrant localization (RL) transition: after an initial localization, a subset of states becomes delocalized again within a finite λ window, before final localization sets in. They report RL windows for Fibonacci (λ ≈ 1.55–2.1) and BM (λ ≈ 0.7–1.3) at t_A=1, t_B=2.5, and claim that finite-size scaling of ⟨NPR⟩ supports persistence of the RL phase in the thermodynamic limit. They also map η in the t_B/t_A–λ plane and find extended mixed-phase regions, multiple RL transitions for BM, and sensitivity of the effect to the incommensurability parameter b, which they set to the asymptotic ratio of the hopping sequence. The paper concludes that correlated quasiperiodic hopping provides a new minimal route to RL.","tokens_in":12554,"tokens_out":4085,"duration_ms":40755,"significance":"If substantiated, the paper extends the phenomenology of reentrant localization to models where both diagonal and off-diagonal modulations are quasiperiodic, without explicit dimerization or long-range hopping. The multi-pronged numerical approach (fractal dimension, IPR/NPR, phase maps) is a strength, and the explicit statement that generic b values destroy RL is an honest test of the mechanism. The model is simple and potentially realizable in photonic or cold-atom settings. However, the central thermodynamic-limit claim currently rests on an under-specified extrapolation and a post hoc selection of averaging windows, which limits the force of the conclusions.","major_comments":[{"comment":"The thermodynamic limit extrapolation that underlies the central claim (⟨NPR⟩ remains finite in the RL windows) is not described. The text says 'we further extrapolate ⟨NPR⟩ in the thermodynamic limit' but gives no functional form, number of points used in the fit, error bars, or goodness-of-fit. For the BM sequence only three system sizes (N=1550, 5117, 16898) are shown, which is insufficient to distinguish a genuinely finite asymptotic value from a slow decay such as c N^{-α} or c/log N. Please specify the extrapolation procedure, show the fitted curves, and test alternative decay forms with a quantitative comparison. Without this, the assertion that the RL phase survives the N→∞ limit is not supported.","section":"Section III, Fig. 4"},{"comment":"The averaging windows for ⟨IPR⟩ and ⟨NPR⟩ (28%–72% of states for Fibonacci, 41%–59% for BM) are chosen after identifying the RL regions in the fractal-dimension plots, as explicitly stated in the text. This post hoc selection conditions the averaged quantities on the phenomenon to be demonstrated. To make the RL claim robust, the authors should show that the non-monotonic behavior of ⟨NPR⟩ persists across a range of reasonable energy windows, or derive the window from an independent criterion (e.g., mobility-edge positions). They should also verify that the fraction of states in the chosen window does not shrink with system size, since a shrinking fraction would undermine the thermodynamic-limit interpretation.","section":"Section III, Fig. 3 and Eq. (4)"}],"minor_comments":[{"comment":"The N→∞ curves in Fig. 4 are drawn as smooth lines without markers or error bars; please indicate how these were obtained and include confidence intervals if possible. For the BM case, the three system sizes span a relatively narrow range, and adding intermediate or larger sizes would strengthen the scaling analysis.","section":"Section III, Fig. 4"},{"comment":"Equation (5) defines the fractal dimension via a limit N→∞, while the paper computes D_n at finite N. The text states this method 'does not require extrapolation', but a finite-size estimate of D_n is not the same as the limit. Please clarify that D_n shown in Fig. 2 is a finite-size estimate, or provide a scaling analysis for D_n.","section":"Section II, Eq. (5)"},{"comment":"The paragraph justifying b = (1+√5)/2 and b = (3+√13)/2 is helpful, but the statement that RL also occurs for the inverses and twice these values is qualitative. A quantitative statement about the range of b values that support RL (e.g., a plot or a bound) would clarify whether the effect is a fine-tuned resonance or a broader phenomenon.","section":"Section III, paragraph on b dependence"},{"comment":"The claim that 'a universal feature across all known RL studies' is the presence of a staggered quasiperiodic on-site potential is a strong generalization based on limited cases; please soften or provide a more systematic survey.","section":"Section III, closing paragraph"},{"comment":"The paper states that other substitutional sequences (Thue-Morse, Copper mean) were tested but results are omitted because no RL was observed. Including a brief summary of these null results (or at least the parameter range explored) would improve reproducibility and transparency.","section":"General"}],"recommendation":"major_revision","confidential_remarks":"The paper addresses a timely topic in quasiperiodic systems and the basic observation of non-monotonic ⟨NPR⟩ is interesting. The main weakness is not the physics but the evidence for the thermodynamic-limit claim: the extrapolation is undocumented and the averaging windows are selected post hoc. These issues are fixable with additional numerical analysis, so a major revision is appropriate rather than rejection. The b-tuning issue is handled honestly but the 'purely' phrasing in the abstract and introduction should be tempered."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: this is an honest, thorough numerical study of a genuinely new family of reentrant localization (RL) models. The combination—Fibonacci and Bronze Mean hopping sequences in a staggered AAH potential, with the incommensurability parameter b locked to the asymptotic t_A/t_B ratio—is not in the cited literature, and the paper gives a fairly complete numerical characterization: fractal dimension maps, IPR/NPR, eta phase diagrams, and wavefunction snapshots. I believe the central observation, that these sequences produce an extended window between two localized phases, is real; the evidence is internally consistent and the paper is careful about what it claims.\n\nThe strongest soft spot is the thermodynamic-limit claim. The finite-size scaling in Fig. 4 has no extrapolation form, no error bars, and only three system sizes for the Bronze Mean chain. A function like c N^{-α} with α<1 can masquerade as a finite intercept. This matters because the persistence of the RL phase in the thermodynamic limit is a headline claim. The fix is easy: show the extrapolation explicitly (e.g., <NPR> vs 1/N or log N with a stated fit), add more sizes, and report the uncertainty.\n\nA second concern, smaller than the first: the averaging windows for <IPR> and <NPR> are chosen after the RL regions are identified (central 44% for Fibonacci, 18% for BM). The authors are transparent about this, and the D_n plots show the same reentrant streaks, so it is not pure cherry-picking. Still, a sensitivity check (vary the window boundaries, or show the full-spectrum measures) would blunt the criticism that the result is conditioned on the phenomenon.\n\nThe b-matching is less of a problem. They explicitly test generic irrational b and report monotonic localization, and they show RL for inverses and multiples of the sequence ratio. So the correlated choice is a defensible model condition, not a hidden tuning knob. I would not hang a rejection on it.\n\nAlso minor: no code/data, and the eta maps are at fixed N ≈ 1600 with no scaling. These are addressable and not disqualifying.\n\nWho should read this: anyone working on quasiperiodic localization or mobility edges. It is a modest but useful addition to the RL literature, distinct from the Tabanelli et al. off-diagonal IAAF study. It deserves serious peer review; a referee should push on the scaling analysis and the averaging-window sensitivity.","headline":"A genuinely new family of reentrant localization models, supported by consistent numerics, but the thermodynamic-limit claim rests on an under-documented extrapolation that a referee should push on.","tokens_in":13057,"tokens_out":2660,"would_cite":true,"duration_ms":26866,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["71.23.-k","71.30.+h","72.15.Rn"],"model":"deepseek-v4-flash","headline":"This paper claims that in a one-dimensional chain with a staggered Aubry-André-Harper potential, correlated quasiperiodic hopping following Fibonacci or Bronze Mean sequences makes some localized eigenstates become extended again in a…","keywords":["reentrant localization","quasiperiodic chain","Aubry-André-Harper model","Fibonacci sequence","Bronze Mean sequence","correlated hopping","mobility edge","participation ratio"],"falsifier":"A decisive check is to repeat the Fibonacci-hopping calculation with a generic irrational b, such as $b=1/\\sqrt{2}$, keeping all other parameters fixed: the paper's mechanism predicts monotonic localization with no reentrant $\\langle\\mathrm{NPR}\\rangle$ bump, so any surviving RL window at generic $b$ would falsify the correlation picture. A complementary experimental test would be a photonic waveguide array with Fibonacci-ordered couplings and staggered AAH site energies tuned to the golden-mean frequency, where the claim predicts re-extended transport near $\\lambda\\approx1.8$ between localized regimes.","tokens_in":12076,"feed_emoji":"🔁","tokens_out":11049,"duration_ms":103821,"temperature":0.7,"pith_summary":"This paper claims that reentrant localization—the reappearance of extended states after a system has already localized—can be driven purely by correlated quasiperiodic hopping, without dimerization or long-range couplings. The setting is a one-dimensional tight-binding chain whose on-site energies are a staggered Aubry-André-Harper potential and whose nearest-neighbor hoppings follow either a Fibonacci or a Bronze Mean substitution sequence. As the potential strength $\\lambda$ grows, the eigenstates first localize, then a subset re-extends inside a finite window (for Fibonacci, roughly $1.55\\lesssim\\lambda\\lesssim2.1$; for Bronze Mean, roughly $0.7\\lesssim\\lambda\\lesssim1.3$), and then localize again at larger $\\lambda$. The authors argue the reentrant window survives the thermodynamic limit because $\\langle\\mathrm{NPR}\\rangle$ extrapolates to a finite value, and they attribute the effect to a deliberate correlation: the on-site incommensurability parameter $b$ is set to the asymptotic symbol ratio of the hopping sequence. If the claim is right, it identifies a new minimal route to reentrant localization in quasiperiodic systems with correlated disorder in both diagonal and off-diagonal terms.","feed_headline":"Correlated hopping reopens extended states in a localized chain","feed_subtitle":"A finite fraction of states re-extends at λ≈1.55–2.1 (Fibonacci) and λ≈0.7–1.3 (Bronze Mean), beyond the localization threshold.","key_machinery":"The load-bearing object is the joint modulation: the staggered (alternating-sign) Aubry-André-Harper site potential $\\epsilon_i=\\lambda\\cos(2\\pi b i)$ with the incommensurability parameter $b$ fixed to the asymptotic ratio of the number of $t_A$ bonds to $t_B$ bonds in the hopping sequence—$(1+\\sqrt{5})/2$ for the Fibonacci substitution $A\\to AB$, $B\\to A$, and $(3+\\sqrt{13})/2$ for the Bronze Mean substitution $A\\to AAAB$, $B\\to A$. This choice aligns the two quasiperiodic modulations and is what generates the correlated disorder. The argument is carried by eigenstate diagnostics: the fractal dimension $D_n=-\\lim_{N\\to\\infty}\\log(\\mathrm{IPR}_n)/\\log N$, the averaged inverse and normalized participation ratios $\\langle\\mathrm{IPR}\\rangle$ and $\\langle\\mathrm{NPR}\\rangle$, and the mixed-phase indicator $\\eta=\\log_{10}[\\langle\\mathrm{IPR}\\rangle\\langle\\mathrm{NPR}\\rangle]$, whose finite values in intermediate windows mark the reentrant phase.","core_discovery":"On the paper's own terms, the discovery is a two-step localization-delocalization-localization transition in a one-dimensional tight-binding model with a staggered Aubry-André-Harper on-site potential and hopping that alternates according to Fibonacci or Bronze Mean substitution rules. As $\\lambda$ increases from zero, the eigenstates go from extended to localized, then a subset re-extends in a finite window, and finally localizes again; the reentrant window persists as $N\\to\\infty$, with the extrapolated $\\langle\\mathrm{NPR}\\rangle$ staying finite inside the window and vanishing outside it. The mechanism is the correlation between diagonal and off-diagonal quasiperiodicities: choosing $b=(1+\\sqrt{5})/2$ for Fibonacci hopping and $b=(3+\\sqrt{13})/2$ for Bronze Mean hopping aligns the on-site potential with the asymptotic ratio of the hopping sequence, allowing constructive interference that re-extends a fraction of states at intermediate $\\lambda$. The paper also maps the mixed-phase indicator $\\eta=\\log_{10}[\\langle\\mathrm{IPR}\\rangle\\langle\\mathrm{NPR}\\rangle]$ over the $t_B/t_A$–$\\lambda$ plane, showing that RL occupies a finite hopping-ratio band, that the Bronze Mean case has extra mixed-phase islands and multiple RL transitions, and that generic irrational $b$ destroys the effect.","pith_inferences":["Inference: the paper's $\\eta$-maps imply the RL window should be tunable through $t_B/t_A$; shifting the hopping ratio should move the two bounding mobility edges in a predictable way, which offers a direct experimental knob that the paper does not discuss.","Inference: the matching condition suggests a broader search strategy—any substitution sequence whose inflation ratio is a quadratic irrational could be paired with an AAH potential at that frequency; only Fibonacci, Bronze Mean, and a few others were tested here.","Inference: because the model is strictly non-interacting and single-particle, the fate of the reentrant window under weak interactions or periodic driving is an open question that the paper does not address."],"forward_implications":["The reentrant extended phase is not a finite-size artefact: extrapolated $\\langle\\mathrm{NPR}\\rangle$ stays finite as $N\\to\\infty$ inside the RL windows.","The two-step localization-delocalization-localization transition is bounded by two distinct single-particle mobility edges, so extended, localized, and critical states coexist in the spectrum at the same $\\lambda$.","Matching the on-site frequency to the hopping sequence's inflation ratio is necessary: generic irrational $b$ removes RL, while inverses and doubles of the matching ratios also produce it.","In the $t_B/t_A$–$\\lambda$ plane, RL occupies a finite hopping-ratio band (about 2.3–3.0 for $t_B/t_A$), and the Bronze Mean sequence additionally shows an isolated mixed-phase island and multiple RL transitions near $t_B/t_A\\approx2.8$.","Correlated quasiperiodic hopping alone, without dimerization or long-range hopping, suffices to produce RL when combined with the staggered AAH potential."],"supporting_citations":[{"why":"Supplies the Aubry-André-Harper on-site potential and the standard incommensurate localization transition it induces.","marker":"[12, 13]"},{"why":"First demonstrated reentrant localization in a quasiperiodic potential; defines the phenomenon this paper extends to correlated hopping.","marker":"[26]"},{"why":"Showed RL from the interplay of hopping dimerization and quasiperiodic modulation; the baseline scenario this paper replaces with correlated quasiperiodic hopping.","marker":"[29]"},{"why":"Experimental observation of RL and criticality in the interpolating Aubry-André-Fibonacci model; used to justify the phase choice and experimental accessibility.","marker":"[27]"},{"why":"Source of the Fibonacci, Bronze Mean, and other substitution sequences and of the asymptotic symbol-ratio method used to set b.","marker":"[41]"},{"why":"Define IPR, NPR, and the mixed-phase indicator eta, the diagnostics all claims are built on.","marker":"[42, 43]"},{"why":"Off-diagonal interpolating Aubry-André-Fibonacci model with multiple RL transitions; the closest prior hopping-modulated RL that this work generalizes.","marker":"[38]"}],"fun_headline_variants":["Reentrant localization from Fibonacci and Bronze Mean hopping","Correlated hopping reopens states in a localized quasiperiodic chain","Fibonacci and Bronze Mean hopping trigger reentrant delocalization","Quasiperiodic hopping sequences restore extended states at high disorder","Reentrant localization: hopping correlations reopen extended states"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The reentrant effect depends entirely on setting the on-site potential's frequency equal to the hopping sequence's asymptotic symbol ratio; if that matching is an ad hoc tuning rather than a legitimate physical condition, localization becomes monotonic and the central claim loses its force.","fun_headline_variants_meta":{"raw":{"variants":["Reentrant localization from Fibonacci and Bronze Mean hopping","Correlated hopping reopens states in a localized quasiperiodic chain","Fibonacci and Bronze Mean hopping trigger reentrant delocalization","Quasiperiodic hopping sequences restore extended states at high disorder","Reentrant localization: hopping correlations reopen extended states"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000372,"raw_usage":{"total_tokens":2056,"prompt_tokens":1079,"completion_tokens":977,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":695,"completion_tokens_details":{"reasoning_tokens":894}},"tokens_in":695,"tokens_out":977,"duration_ms":8638,"temperature":1.0,"reasoning_tokens":894,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-07T11:17:03.296608+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A decisive check is to repeat the Fibonacci-hopping calculation with a generic irrational b, such as $b=1/\\sqrt{2}$, keeping all other parameters fixed: the paper's mechanism predicts monotonic localization with no reentrant $\\langle\\mathrm{NPR}\\rangle$ bump, so any surviving RL window at generic $b$ would falsify the correlation picture. A complementary experimental test would be a photonic waveguide array with Fibonacci-ordered couplings and staggered AAH site energies tuned to the golden-mean frequency, where the claim predicts re-extended transport near $\\lambda\\approx1.8$ between localized regimes.","supporting_citations":[{"cited_title":"Kohmoto and Y","cited_arxiv_id":null,"evidence_quote":"First demonstrated reentrant localization in a quasiperiodic potential; defines the phenomenon this paper extends to correlated hopping."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Showed RL from the interplay of hopping dimerization and quasiperiodic modulation; the baseline scenario this paper replaces with correlated quasiperiodic hopping."},{"cited_title":"Kohmoto, B","cited_arxiv_id":null,"evidence_quote":"Experimental observation of RL and criticality in the interpolating Aubry-André-Fibonacci model; used to justify the phase choice and experimental accessibility."},{"cited_title":"Padhan, M","cited_arxiv_id":null,"evidence_quote":"Source of the Fibonacci, Bronze Mean, and other substitution sequences and of the asymptotic symbol-ratio method used to set b."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Off-diagonal interpolating Aubry-André-Fibonacci model with multiple RL transitions; the closest prior hopping-modulated RL that this work generalizes."}],"review_version":1}