{"id":"d10922f7-ab07-4109-a472-ea77bd16bee4","arxiv_id":"2608.10868","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":1,"one_line_summary":"Heavy-tailed hopping in a noninteracting Anderson model produces a parametrically long intermediate subdiffusive regime that always ends in ordinary diffusion.","lead":"This paper derives analytical transport formulas for a disordered lattice with heavy-tailed hopping, showing a long-lived subdiffusive regime that crosses over to normal diffusion. It offers a single-particle benchmark that could help scientists separate interaction-driven slow dynamics from pure disorder effects.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The μ→1 divergence in Eq. (16) is asserted in a limit that the paper's own validity condition excludes: the saddle-point approximation breaks down below a threshold μ*>1, so the central asymptotic claim lacks a controlled derivation.","rationale":"The reader identified the same weakest spot: the derivation is controlled only for μ > μ*, while the headline divergence is stated at μ → 1+, so the paper's own validity condition cuts off the asymptotic limit. I agree that this is the most load-bearing concern because the central quantitative prediction is precisely the 1/sqrt(μ−1) divergence of R_sub; if the saddle-point and delta-function approximations break down in that limit, Eq. (16) is not established, regardless of how plausible the crossover picture is. I do not fully agree that the paper should be rejected or that the concern is fatal: the authors explicitly disclose the threshold, the same mechanism could still produce a parametrically large window if μ* is sufficiently close to 1 for small P_D h0, and the one direct simulation at μ = 1.6 is consistent with the predicted preasymptotic enhancement. The missing piece is a controlled analysis of the double limit and a numerical check of the exact saddle-point equation near μ*. The reader's CONDITIONAL verdict already captures this: the claim is plausible but not fully verified. My stress-test therefore does not move the verdict, though it sharpens the required condition: the divergence must be shown to hold from the exact equation, not only from the approximate saddle-point solution. The paper deserves credit for clearly flagging the limitation rather than hiding it, but the central asymptotic statement is exactly where the control is lost.","tokens_in":13716,"tokens_out":6340,"duration_ms":63668,"concrete_test":"Estimate μ* for the parameters of Figs. 2 and 4 from the condition Γ0(μ) = W using Eq. (10), then numerically solve the full saddle-point equation (32) without the delta-function approximation (38) at μ just above μ* (e.g., μ = 1.05 and 1.02 with h0/W chosen so that μ* < μ) and extract R_sub from the resulting λ0(κ) and c0(κ). If the exact R_sub follows 1/sqrt(μ−1) over μ ∈ (μ*, 1.1), the concern is benign; if it saturates or the divergence softens, Eq. (16) is an artifact of extrapolating outside the valid regime.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim is Eq. (16): R_sub ~ r0/sqrt(μ−1) as μ→1+. But the derivation of the correlation function (8)–(9) rests on the saddle-point solution gκ(φ) ≈ (βΓκ/4 φ†φ)^(μ/2) in Eq. (33), obtained using the delta-function approximation (38), justified by φ†Lφ ∼ W/h0^2 >> 1/W, and on the condition Γ0 ≪ W. The paper itself states in 'Physical discussion and regime of validity' that a(μ) in Eq. (10) diverges as μ→1, so Γ0 ≪ W inevitably fails for μ below a threshold μ* > 1; 'there always remains a threshold μ* > 1 such that the theory is valid only for μ > μ*.' Thus the predicted divergence as μ→1+ is not reached within the controlled domain of the calculation. For any fixed P_D(E)h0, the approach to μ=1 exits the valid regime before Eq. (16) can be established. The μ=1.03 and μ=1.05 curves in Fig. 2 may lie below μ*, and the only direct simulation is at μ=1.6, far from the asymptotic claim. This does not disprove the existence of a long anomalous crossover window, but the specific 1/sqrt(μ−1) divergence is an extrapolation across an uncontrolled boundary unless the double limit (P_D h0 → 0 together with μ → 1+) is analyzed and shown not to alter the scaling.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript develops an analytical, supersymmetric field-theoretic treatment of transport in a noninteracting Anderson model with power-law (heavy-tailed) distributed hopping amplitudes. It derives a closed-form correlation function, Eq. (8), and from it the mean-square displacement and an effective dynamical exponent ξ(t). The central claim is that for 1<μ<2 the system exhibits a long intermediate subdiffusive regime before crossing over to asymptotic diffusion, and that as μ approaches the localization transition μ→1+, the spatial extent of the subdiffusive window diverges as R_sub ~ r0/√(μ−1), while the dimensionless crossover time stays finite. The analytical results are compared with a direct time-evolution simulation in three dimensions at μ=1.6.","tokens_in":13979,"tokens_out":8049,"duration_ms":71343,"significance":"If the central claim holds, the paper provides a single-particle, disorder-only mechanism for long-lived subdiffusive transport that qualitatively mimics Griffiths-type dynamics near the MBL transition, which would be a valuable null model for interaction-effect studies. The paper's strengths are its analytic closed-form expressions, the unified treatment of orthogonal and unitary symmetry classes, the explicit prediction for the frequency-dependent conductivity, and a direct numerical check. However, the headline μ→1+ divergence is obtained in a limit that the paper's own validity conditions exclude, and the numerical verification uses one parameter set with a fitted energy scale, so the claimed universality and the μ-dependence of the effect are not yet convincingly established.","major_comments":[{"comment":"The paper's central claim, Eq. (16) (R_sub ~ r0/√(μ−1) as μ→1+), is not controlled by the derivation. The section 'Physical discussion and regime of validity' states that a(μ) in Eq. (10) diverges as μ→1 and that 'there always remains a threshold μ* > 1 such that the theory is valid only for μ > μ*.' The saddle-point solution Eq. (33) and the delta-function approximation Eq. (38) are only justified in this valid regime, so the approach to μ=1 exits the controlled domain before the divergence can be established. To sustain the claim, the authors must analyze the double limit μ→1+ together with P_D(E)h0→0 and show that the scaling R_sub ~ 1/√(μ−1) is not altered; otherwise Eq. (16) is an extrapolation across an uncontrolled boundary.","section":"Physical discussion and regime of validity; Eq. (16)"},{"comment":"The numerical verification is too weak to support the μ-dependence of the theory. Only one value, μ=1.6, is simulated in 3D, and the energy scale Γ0 is not computed from Eq. (10) but is extracted by fitting the long-time tail of the same data to the predicted diffusive law, Eq. (12). The comparison with the full theory therefore does not independently test the predicted timescale or its μ-dependence. The figure also shows no error bars and only n=20–50 disorder realizations. Tests closer to μ=1, where the subdiffusive window is predicted to be large, or an independent determination of Γ0, are needed.","section":"Direct numerical simulations; Fig. 4"},{"comment":"Equation (13), the asymptotic form of the mean-square displacement, is load-bearing for the crossover time and length in Eqs. (14)–(16), but its derivation is not shown. The End Matter states that it follows from a higher-order small-κ expansion of Eq. (9), yet the expansion is not presented. Please include the expansion or an explicit reference to the derivation, together with the conditions under which the constant term is valid, because that constant term is what produces the preasymptotic enhancement over diffusion.","section":"End Matter, Derivation of main results; Eq. (13)"},{"comment":"The delta-function approximation Eq. (38) and the separation of g(ϕ) into real- and imaginary-part contributions in Eq. (43) are imported from Ref. [27] and Ref. [29], respectively. The justification of Eq. (38) relies on the estimate ϕ†Lϕ ~ W/h0^2 in Eq. (44) from Ref. [27]. Since these steps are essential for the saddle-point solution Eq. (33) and hence for the main result Eq. (8), the manuscript should either reproduce these estimates in the present spatial setting or state explicitly that the zero-dimensional results are assumed to carry over and under what conditions.","section":"End Matter, Saddle-point equation; Eq. (38)"}],"minor_comments":[{"comment":"The mean-square displacement is written as ⟨ψ(t)|Σ_n (r_n − r_n)^2|ψ(t)⟩; the second r_n should be the position of the initial state, e.g., r_0, otherwise the expression is identically zero.","section":"Eq. (18)"},{"comment":"Reference [14] has a malformed author list; it should be corrected to 'Yan V. Fyodorov, Alexander D. Mirlin, and Hans-Jürgen Sommers.'","section":"Reference [14]"},{"comment":"Fig. 4 reports no error bars; please add them or state that the scatter between realizations is smaller than the symbol size.","section":"Fig. 4 caption"},{"comment":"The notation Γκ should be defined as Γ_κ when first introduced, and the subscript should be used consistently instead of Γκ.","section":"Eq. (9) and throughout"},{"comment":"For β=2, the distribution of hopping amplitudes should state explicitly whether the magnitude |t_nm| and the phase θ_nm are independent random variables.","section":"Table following Eq. (3)"}],"recommendation":"major_revision","confidential_remarks":"The paper is within the journal's scope. My main technical concern is that the headline μ→1+ divergence in Eq. (16) is not supported by the derivation, since the paper's own validity conditions exclude μ below a threshold μ* > 1. This is fixable either by proving the double-limit scaling or by explicitly demoting the μ→1 claim to an extrapolation/conjecture. The numerical check is also too minimal to independently confirm the predicted μ-dependence. If the authors address these points, the intermediate-regime theory could be a worthwhile contribution; I therefore recommend major revision rather than rejection."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: the paper delivers a clean single-particle mechanism for Griffiths-like subdiffusion, built on the authors' earlier supersymmetric LRP toolkit. The genuinely new piece is the spatial transport calculation, Eqs. (8)–(16), and the picture it paints is plausible. But the headline divergence as μ→1 is claimed in a limit that the paper itself says is outside the controlled domain.\n\nWhat is good: the model is clearly specified, the three energy scales W, Γ0, and Δ are laid out properly, and the crossover from an anomalous window to asymptotic diffusion is derived rather than fitted. The formulas for the effective exponent and the subdiffusive length scale are new. The paper is also honest: it states explicitly that the theory breaks down below a threshold μ*, and it frames the MBL connection as a qualitative benchmark, not a proof of Griffiths physics. The direct simulation at μ=1.6 reproduces the predicted enhancement and crossover, which is a useful check.\n\nSoft spots, in proportion. The derivation of Eqs. (9) and (13) is not shown in the preprint; key integrals and the saddle-point solution are imported from Refs. [27–29], mostly by the same group. That makes independent verification laborious, but it is not a flaw by itself. The numerics are a single parameter set with a fitted Γ0 and no error bars; fine as a consistency check, weak as a test of the μ-dependence. The larger issue is the central asymptotic claim: R_sub ~ 1/√(μ−1) as μ→1+. The paper's own validity condition requires μ > μ* with μ* > 1, because the coefficient a(μ) in Eq. (10) diverges at μ=1. So the divergence is predicted on the other side of a boundary the derivation does not control. The stress-test note is right. This does not kill the paper: the qualitative subdiffusive window at moderate μ is supported, and it is plausible that a suitable double limit (P_D h0 → 0 together with μ → 1+) recovers the scaling, but that double limit is not analyzed here.\n\nWho it is for: people working on MBL subdiffusion and rare-region physics should know this single-particle null model exists. It deserves peer review; the flaw is addressable. I would send it out, with the request that the authors either prove the divergence in the appropriate double limit or scale back the claim to the controlled regime.","headline":"Solid single-particle null model for Griffiths-like subdiffusion; the headline divergence is predicted outside the paper's own validity regime.","tokens_in":14572,"tokens_out":3592,"would_cite":true,"duration_ms":32651,"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":"This paper establishes that a noninteracting Anderson model with heavy-tailed hopping amplitudes produces an intermediate subdiffusive regime whose spatial extent diverges as the localization transition is approached, while the crossover…","keywords":["heavy-tailed hopping","Anderson localization","anomalous diffusion","subdiffusive crossover","mean-square displacement","rare-region transport","many-body localization"],"falsifier":"Directly simulate the three-dimensional Hamiltonian for a sequence of exponents approaching $\\mu=1^+$ within the controlled parameter regime, and extract the distance at which the local exponent $\\xi(t)$ reaches a fixed value such as 0.9. The central claim requires that distance to grow as $(\\mu-1)^{-1/2}$ while the renormalized crossover time $\\Gamma_0 t_\\xi$ stays essentially constant; observing a saturation of the distance or a diverging crossover time would invalidate the claimed mechanism.","tokens_in":13455,"feed_emoji":"⏳","tokens_out":17264,"duration_ms":132441,"temperature":0.7,"pith_summary":"The paper aims to establish that a single noninteracting particle moving on a disordered lattice with power-law-tailed hopping amplitudes can diffuse sublinearly—mean-square displacement growing with an effective exponent below 1—for a long intermediate time before ordinary diffusion takes over. The heavy-tail exponent $\\mu$ controls the hopping statistics; as $\\mu$ approaches $1$ from above, the spatial extent of this subdiffusive window grows as $R_{\\rm sub}\\sim r_0/\\sqrt{\\mu-1}$, while the crossover time measured in units of $\\Gamma_0^{-1}$ stays finite. If true, this gives a minimal single-particle mechanism that reproduces the slow, rare-region-like transport often attributed to many-body effects near the many-body localization transition. The result matters because it provides an analytically controlled benchmark for separating disorder-driven dynamics from interaction-driven dynamics in numerical studies of disordered systems.","feed_headline":"Rare strong hops make a single particle mimic many-body subdiffusion","feed_subtitle":"A single-particle model shows a subdiffusive window that widens near localization, benchmarking many-body studies.","key_machinery":"The engine of the calculation is the saddle-point solution $g_\\kappa(\\varphi)\\simeq(\\beta\\Gamma_\\kappa \\varphi^\\dagger\\varphi/4)^{\\mu/2}$ for the supersymmetric effective action, which replaces the heavy-tailed hopping ensemble by an effective medium characterized by the self-consistent hybridization scale $\\Gamma_\\kappa$. This yields the correlation function $\\hat K_q(\\kappa)=c_0(\\kappa)/(1-\\tilde w(q)\\lambda_0(\\kappa))$, whose curvature in $q$ and dependence on $\\kappa$ determine the conductivity and, through the exact relation to $\\langle r^2(t)\\rangle$, the mean-square displacement. The heavy-tail exponent $\\mu$ enters through the stable distribution of hopping amplitudes and controls both the diffusion coefficient and the singular prefactor that produces the $(\\mu-1)^{-1/2}$ growth of the subdiffusive window.","core_discovery":"The paper derives an analytical expression for the disorder-averaged retarded–advanced correlation function and from it the time-dependent mean-square displacement. The central quantitative result is that for $1<\\mu<2$ the effective exponent $\\xi(t)=d\\log\\langle r^2(t)\\rangle/d\\log t$ lies below 1 over an intermediate window, with the asymptotic form $\\xi(t)\\approx \\Gamma_0 t/(\\Gamma_0 t+(2-\\mu)\\Gamma(2-2/\\mu)(\\mu-1))$, and only reaches the diffusive value $\\xi=1$ at long times. As $\\mu\\to 1^+$, the length scale $R_{\\rm sub}$ over which the local exponent stays below a fixed value diverges as $r_0/\\sqrt{\\mu-1}$, whereas the corresponding crossover time $\\Gamma_0 t_\\xi$ approaches the finite constant $2\\xi/(1-\\xi)$. The paper therefore predicts a parametrically extended spatial crossover, not a true asymptotic subdiffusive phase.","pith_inferences":["The same correlation function should produce a subdiffusive window in local observables such as the return probability or the participation entropy; the paper does not compute these, but the mechanism is directly carried by $\\hat K_q(\\kappa)$.","If the divergence is purely spatial and not temporal, then finite-time measurements at a fixed distance will misread the crossover as an asymptotic anomalous phase; checking system-size dependence of the effective exponent would expose the crossover.","Because the paper's own validity analysis leaves a threshold $\\mu^*>1$, the most informative tests are those that push $\\mu$ toward $1$ while scaling the model parameters to keep the saddle-point regime controlled—whether $R_{\\rm sub}$ truly diverges or saturates remains an open quantitative question.","A natural next step is to add weak interactions to this single-particle model and ask whether the heavy-tail subdiffusive window survives or is cut off; the present paper stops at the noninteracting level."],"forward_implications":["For every $1<\\mu<2$, transport is asymptotically diffusive, but the effective exponent $\\xi(t)$ remains below 1 over an intermediate window that can extend over distances much larger than the microscopic hopping length $r_0$.","As the localization transition is approached, the distance over which subdiffusion appears grows without bound as $r_0/\\sqrt{\\mu-1}$, while the crossover time in units of $\\Gamma_0^{-1}$ stays finite.","The frequency-dependent conductivity displays an apparent power-law $\\sigma(\\omega)\\sim\\omega^{\\alpha(\\omega)}$ with exponent below 1 at intermediate frequencies, before settling into the diffusive dc limit at low frequency.","A noninteracting disordered Hamiltonian can reproduce the slow-transport signatures often taken as evidence for many-body physics, so those signatures alone are not diagnostic of interactions.","The closed-form expressions provide a benchmark for numerical simulations of interacting disordered models, allowing disorder-driven subdiffusion to be identified and subtracted."],"supporting_citations":[{"why":"Supplies the supersymmetric field-theoretic method and the sparse random hopping model that the lattice calculation extends.","marker":"[14]"},{"why":"Provides the distribution of the real part of the self-energy and the estimate used to justify the delta-function approximation in the saddle-point integration.","marker":"[27]"},{"why":"Gives the unitary-symmetry solution of the zero-dimensional heavy-tailed random-matrix model used to evaluate the saddle-point equation.","marker":"[28]"},{"why":"Provides the stable-distribution-averaged integral and the semiclassical formula for the hybridization scale, fixing the time rescaling used in the numerical comparison.","marker":"[29]"},{"why":"Defines the heavy-tailed random-matrix ensemble and the hybridization scale that underlies the mapping of the lattice model.","marker":"[26]"},{"why":"Relates the retarded-advanced correlation function to the frequency-dependent conductivity, connecting the calculation to the mean-square displacement.","marker":"[30]"},{"why":"Establishes the statistical independence of the real and imaginary parts of the self-energy, justifying the separation of energy scales throughout the calculation.","marker":"[31]"}],"fun_headline_variants":["Rare hops produce transient subdiffusion in one particle","Single-particle model mimics many-body subdiffusion via heavy tails","Analytical theory: heavy-tailed hopping creates subdiffusive window","No interactions needed: rare hops slow a single particle","Subdiffusion without many-body effects in a simple Anderson model"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The derivation relies on a saddle-point approximation that is controlled only when the hopping range contains many sites, the dimensionless conductance is large, and the heavy-tail exponent stays above a threshold $\\mu^*>1$; as $\\mu\\to 1^+$ the regime of validity shrinks, so the predicted divergence of the subdiffusive window is asserted in a limit the calculation does not fully control.","fun_headline_variants_meta":{"raw":{"variants":["Rare hops produce transient subdiffusion in one particle","Single-particle model mimics many-body subdiffusion via heavy tails","Analytical theory: heavy-tailed hopping creates subdiffusive window","No interactions needed: rare hops slow a single particle","Subdiffusion without many-body effects in a simple Anderson model"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000191,"raw_usage":{"total_tokens":1327,"prompt_tokens":913,"completion_tokens":414,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":529,"completion_tokens_details":{"reasoning_tokens":331}},"tokens_in":529,"tokens_out":414,"duration_ms":4303,"temperature":1.0,"reasoning_tokens":331,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T15:23:59.569922+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Directly simulate the three-dimensional Hamiltonian for a sequence of exponents approaching $\\mu=1^+$ within the controlled parameter regime, and extract the distance at which the local exponent $\\xi(t)$ reaches a fixed value such as 0.9. The central claim requires that distance to grow as $(\\mu-1)^{-1/2}$ while the renormalized crossover time $\\Gamma_0 t_\\xi$ stays essentially constant; observing a saturation of the distance or a diverging crossover time would invalidate the claimed mechanism.","supporting_citations":[{"cited_title":"A novel field theoretical approach to the An- derson lo- calization : sparse random hopping model.Journal de Physique I, 2 (8):1571–1605, 1992","cited_arxiv_id":null,"evidence_quote":"Supplies the supersymmetric field-theoretic method and the sparse random hopping model that the lattice calculation extends."},{"cited_title":"Spectral properties of Levy Rosenzweig-Porter model via supersymmetric approach.SciPost Phys., 18:010, 2025","cited_arxiv_id":null,"evidence_quote":"Provides the distribution of the real part of the self-energy and the estimate used to justify the delta-function approximation in the saddle-point integration."},{"cited_title":"Density of states correlations in L´ evy Rosenzweig-Porter model via supersymmetry approach","cited_arxiv_id":null,"evidence_quote":"Gives the unitary-symmetry solution of the zero-dimensional heavy-tailed random-matrix model used to evaluate the saddle-point equation."},{"cited_title":"Local den- sity of states correlations in the L´ evy-Rosenzweig-Porter random matrix ensemble.SciPost Phys., 19:015, 2025","cited_arxiv_id":null,"evidence_quote":"Provides the stable-distribution-averaged integral and the semiclassical formula for the hybridization scale, fixing the time rescaling used in the numerical comparison."},{"cited_title":"Biroli and M","cited_arxiv_id":null,"evidence_quote":"Defines the heavy-tailed random-matrix ensemble and the hybridization scale that underlies the mapping of the lattice model."},{"cited_title":"Disordered electronic system as a model of interacting matrices.Physics Reports, 67(1):15–24, 1980","cited_arxiv_id":null,"evidence_quote":"Relates the retarded-advanced correlation function to the frequency-dependent conductivity, connecting the calculation to the mean-square displacement."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Establishes the statistical independence of the real and imaginary parts of the self-energy, justifying the separation of energy scales throughout the calculation."}],"review_version":1}