{"id":"7c716833-2964-4d52-b556-ca59f1209ec4","arxiv_id":"1908.06643","paper_version":3,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"The two-particle localization length ratio ξ2/ξ1 in a 1D disordered chain is a nonlinear, sublinear scaling function F(uξ1) once the single-particle localization length exceeds roughly 100 sites.","lead":"Two particles hopping through a random wire stay localized even when they interact, but how far they spread grows much more slowly with weak disorder than earlier theories predicted. The authors simulate record-large systems to extract this growth and find that the two-particle extension follows a single universal curve with a crossover at a characteristic scale.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Finite-size extrapolation at large u may be validated only at u=0, potentially bending the scaling function F(uξ1).","rationale":"I agree with the reader that the finite-size extrapolation is the weak point. The paper's own methodology section (Sec. IVA) establishes N*(W) and the 1/N scaling only for u=0, and the Appendix states that the constant fit is used systematically without providing a separate validation for large u. The claimed record values (F≈15 at ξ1≈400, u=3) lie at exactly the point where ξ2 is large and the unvalidated extrapolation is most consequential. There is also an internal tension: the text argues that the asymptotic scaling function is nonlinear and sublinear, but a missing upward correction to ξ2 at large uξ1 would tend to restore a faster (closer to linear) growth, directly weakening the novelty claim. The paper does provide several pieces of independent support: it benchmarks two methods (M1 and M2) against each other at u=0, compares with two earlier computations (Refs. 15 and 20), and tests single-parameter scaling via participation numbers. Those are real checks, but they do not independently establish the large-u extrapolation. No issue of internal inconsistency or fraud arises; the concern is about the unvalidated range of a standard extrapolation procedure. Thus the reader's CONDITIONAL verdict is appropriate, and I would not change it.","tokens_in":9063,"tokens_out":1519,"duration_ms":13269,"concrete_test":"Re-extract ξ2 for the record points (W=0.5, u=3, E=0) using the two-point extrapolation and diagnostic in Appendix A but with N up to 40000 (or with the number of fitting points increased to include only N≥12000), and compare the inferred ξ2(N→∞) with the reported values. If the extrapolated ξ2 increases by more than about 10% relative to the reported value for W=0.5, u=3, then the observed sublinear growth of F(x) is partly a finite-size artifact and the central claim needs revision.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The central scaling claim, ξ2/ξ1 = F(uξ1) rising to record F≈15 at ξ1=400 and u=3, rests on the finite-size extrapolation in Sec. IVA Step 2 and the Appendix. The extrapolation form ξ2(N)=ξ2+a/N and the constant-fit protocol were benchmarked only at u=0, where N*(W) is defined, and the extrapolated values (e.g., ξ2/ξ1≈0.56 for W=0.5) are close to the noninteracting value of 0.5. For u=3, however, ξ2 is about 30 times larger at ξ1≈400 (ξ2/ξ1≈16). The fractional correction a/N is then far larger, and there is no evidence that the asymptotic 1/N tail has been reached for these large ξ2 values. The data in Fig. 5(a) show that the largest F values come from W=0.5 and u=3 with no error bars and no demonstration of convergence with N; if the true ξ2 is larger than reported, the apparent sublinear slowdown of F(x) at large uξ1 could be, at least in part, a finite-size artifact.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the localization length ξ2 of two interacting particles in a one-dimensional Anderson chain using a projected Green's function method combined with finite-size scaling, for system sizes up to N=20000 and disorder strengths down to W=0.5. It reports that for weak disorder (ξ1>100) the ratio ξ2/ξ1 becomes a nonlinear function F(uξ1) of the single variable uξ1, with sublinear growth for large arguments, and claims record values ξ2/ξ1≈16 (F≈15 at ξ1=400, u=3). The paper also analyzes Fock-space connectivity to argue that momentum conservation becomes relevant in this regime and derives an effective model with Weff≈Δ1/ξ1^2 and teff≈u/ξ1.","tokens_in":9319,"tokens_out":7688,"duration_ms":77223,"significance":"If established, the result would supersede earlier claims of a constant scaling exponent for the two-interacting-particle problem and would provide a quantitative, falsifiable scaling prediction. The paper's strengths are its record system sizes, the systematic treatment of energy and interaction strength, and a parameter-free effective-model argument that motivates the scaling variable uξ1 rather than fitting it. However, the central data set rests on a finite-size extrapolation that is validated only at u=0, and the paper does not yet demonstrate convergence of the record ξ2 values or a clean collapse of the purported scaling function.","major_comments":[{"comment":"The extrapolation protocol is benchmarked only at u=0, where N*(W) is defined. For the record point W=0.5, u=3, the fitted interval (0.05N to 0.95N) spans at most 0.9N=18000 sites, while the claimed ξ2≈16ξ1≈6400, so the interval covers only about 2.8ξ2. A linear fit of ln|G| over less than three localization lengths cannot establish the asymptotic exponential tail, so the extracted ξ2(N) is a lower bound. Show ξ2(N) versus 1/N for W=0.5 and several u values, including u=3, over a range of N that enters the asymptotic tail, or determine an N*(W,u) from the convergence of the slope, and report the resulting uncertainty. Without this, the reported sublinear saturation of F(x) at large uξ1 could be an artifact of underestimating ξ2.","section":"Sec. IV A, Step 2; Fig. 5(a)"},{"comment":"The main text states that the extrapolation uses ξ2(N)=ξ2+a/N, while the appendix reports that the constant fit f(x)=c was used systematically. These are not the same procedure: the constant fit sets a=0 and returns an average over the available N, not the 1/N→0 limit. In the appendix example the two fits differ by 5% (272.52 vs 267.59), and for the much larger ξ2 at u=3 the difference can be larger. Specify which fit produced the data in Fig. 5(a) and give both extrapolated values for the record points; if the constant fit was used, the quoted ξ2(N→∞) is biased downward whenever ξ2(N) has a positive 1/N tail.","section":"Sec. IV A, Step 2; Appendix A"},{"comment":"The central claim that ξ2/ξ1=F(uξ1) is a single-parameter scaling function is not demonstrated by a collapse test. Fig. 5(a) plots curves for different W without showing that data points with the same uξ1 but different W agree within error, and the solid black line connects the maxima rather than representing F. Please include a collapse plot or a table of pairs (W,u) with comparable uξ1 and their measured ξ2/ξ1, with error bars, to distinguish a true scaling function from a family of curves.","section":"Sec. V; Fig. 5"},{"comment":"The u=0 benchmark yields ξ2/ξ1≈0.56 instead of the known 0.5, a 10% systematic discrepancy acknowledged in the text. Because the same fitting and extrapolation protocol is used for all u, this discrepancy sets a systematic floor for every F value. State how a 0.56→0.5 correction would propagate to the reported record F≈15 and to the fitted sublinear behavior; if the correction is not constant in uξ1, it could change the shape of F.","section":"Sec. IV A; Fig. 2"}],"minor_comments":[{"comment":"The notation 'where ... denotes the disorder average' is incomplete; define the averaging explicitly, for example as ξ2^{-1} = -lim_{|n-m|→∞} ⟨ln|⟨n,n|Ĝ|m,m⟩|⟩/|n-m|, and use the same notation consistently in Step 1.","section":"Sec. IV A, Eq. (6)"},{"comment":"Several error bars are not visible; please use larger markers or provide the numerical values of the extrapolated ξ2 and its error for each W, otherwise the claimed insensitivity of the extrapolation to the fitting method cannot be checked.","section":"Fig. 3"},{"comment":"The legend entry '2max' is undefined; state explicitly that it denotes the maximal ξ2 for the optimal u and W shown by the solid line, or relabel it in a self-explanatory way.","section":"Fig. 5(a)"},{"comment":"The statement 'for weak disorder W<1 the asymptotic connectivity regime of Fock states is observed' uses 'asymptotic' in a different sense from the later 'asymptotic scaling for ξ1>100'; define the two uses to avoid ambiguity.","section":"Sec. III"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is within the journal's scope and the central question is interesting, but the key quantitative claim depends on a finite-size extrapolation that is not yet demonstrated to be converged at the largest interaction strengths. I would ask the editor to require the revised manuscript to present a convergence analysis for the record points and a genuine collapse test for F(uξ1). No data repository or code is provided, which limits reproducibility."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"This paper is the first to push the TIP localization length calculation into the weak-disorder regime with serious system sizes, and the central observation—that ξ2/ξ1 scales as a sublinear function F(uξ1) once uξ1 > 50—is genuinely new and contradicts earlier linear or convex extrapolations. If it holds, it settles the long-standing exponent debate in favor of a crossover picture rather than a constant α. The Fock-space connectivity analysis is also a real step forward: it shows how momentum conservation selects about ξ1 resonant states and yields the effective disorder/hopping estimates that motivate the scaling variable. I credit that as a substantive mechanism.\n\nThe method is standard and the benchmarking is mostly honest. At u=0, they get ξ2/ξ1 ≈ 0.56 against the exact 0.5; that 10% error is not pretty, but it is far better than the 30% they quote from Ref. 20, and they discuss it. The use of two fitting protocols (M1/M2) and the convergence checks in Fig. 3 give me reasonable confidence that the N→∞ extrapolation is not wildly off at u=0.\n\nThe soft spot is the one the stress-test note identifies: the finite-size threshold N*(W) is calibrated at u=0 and then applied to all u. At u=3 the reported ξ2 is roughly 30 times larger than at u=0, so the same N* may not guarantee that the 1/N tail has been reached. The key scaling plot, Fig. 5(a), has no error bars, no released data or code, and the largest F values are exactly the ones where this worry is strongest. So the sublinear slowdown for large uξ1 could be partly a finite-size artifact. That said, the trend is consistent across several W values and the comparison against Ref. 20's overestimates is plausible, so I would not call the central claim dead. I would call it conditional: the shape of F(x) for x > 50 needs one more paper with either a u-dependent N* validation or direct convergence tests at large u.\n\nWho is this for? Anyone working on two-particle localization, many-body localization precursors, or Fock-space connectivity. It deserves a serious referee, not a desk reject. The referee should ask for error bars, a convergence criterion that does not rely on the u=0 benchmark, and ideally the data behind Fig. 5.","headline":"Solid numerical claim of a sublinear scaling crossover in TIP localization, but the finite-size extrapolation at large u needs tighter validation before I'd bet on the exact shape of F.","tokens_in":9884,"tokens_out":2473,"would_cite":true,"duration_ms":23623,"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":"Two interacting particles in a disordered one-dimensional chain localize up to 15 times farther than a single particle, and the gain is governed by a nonlinear scaling function that grows sublinearly in the weak-disorder limit.","keywords":["two interacting particles","Anderson localization","localization length","disorder","Fock space connectivity","interaction-induced delocalization","scaling function","Green's function method"],"falsifier":"A decisive test is to repeat the extraction at the record point W=0.5, u=3 using system sizes beyond the u=0 threshold, for example N=40000 with more disorder realizations, and to extrapolate ξ2(N) with a threshold determined at u=3 independently. If the extrapolated ratio ξ2/ξ1 then exceeds the reported value of about 16, or if the curve of F(uξ1) bends upward rather than saturating, the claim of a sublinear asymptotic scaling function is falsified. Equally, an independent transfer-matrix or exact-diagonalization computation at N>6000 that yields ξ2/ξ1≫16 at the same parameters would settle the question.","tokens_in":8853,"feed_emoji":"🔬","tokens_out":8407,"duration_ms":78747,"temperature":0.7,"pith_summary":"This paper studies two interacting particles moving on a one-dimensional disordered chain and asks how far the pair spreads compared with a single particle. The authors push numerical simulations to system sizes up to N=20000 and disorder strengths down to W=0.5, entering a weak-disorder regime where the single-particle localization length ξ1 exceeds 100. They find that the enhancement ratio ξ2/ξ1 is described by a scaling function F(uξ1) that grows only nonlinearly, with a sublinear tail at large argument, reaching a record value F≈15 for ξ1=400 and interaction u=3. If correct, this replaces earlier power-law predictions for the two-particle localization length with a crossover picture in which momentum conservation in the single-particle eigenstates sharply limits how many Fock states participate. The result matters because it determines how much interaction can extend localization in disordered wires, the basic building block of many-body localization discussions.","feed_headline":"Two particles localize 15 times farther than one","feed_subtitle":"Weak-disorder simulations show the two-particle gain grows slower than earlier power laws predicted.","key_machinery":"The central object is the projected two-particle Green's function $\\tilde G=\\tilde G_0/(1-u\\tilde G_0)$ restricted to doubly occupied sites, whose exponential decay defines ξ2 via $1/\\xi_2=-\\lim_{|n-m|\\to\\infty}\\ln|\\langle n,n|\\tilde G|m,m\\rangle|/|n-m|$. The computation is made feasible by a tridiagonal-structure reduction that turns the O($N^{4}$) evaluation of $\\langle n,n|G_0|m,m\\rangle$ into O($N^{3}$) sums over single-particle eigenfunctions. Finite-size scaling $\\xi_2(N)=\\xi_2+a/N$ is benchmarked against the known u=0 case, giving thresholds N*(W) up to about 6000. The key mechanism is the Fock-space coupling ratio $R=|uI_{\\mu,\\nu}^{\\mu',\\nu'}/(E_{\\mu'}+E_{\\nu'}-E_\\mu-E_\\nu)|$: for W<1, energy conservation plus emerging momentum conservation leaves only resonant pairs with ν=-μ, yielding effective scales $t_{\\rm eff}\\approx u/\\xi_1$ and $W_{\\rm eff}\\approx\\Delta_1/\\xi_1^2$, and hence the scaling variable uξ1.","core_discovery":"The claim is that, in the asymptotic weak-disorder regime (ξ1>100), the ratio of the two-particle to single-particle localization length obeys ξ2/ξ1=F(uξ1), with F(x) linear for small x but growing sublinearly for x>50. The authors report record values F=15 at ξ1=400, u=3, and they attribute the slowdown to the recovery of translational invariance and momentum conservation in the Fock-space matrix elements when W<1: resonantly coupled Fock states form fragile groups of only about ξ1 states, not $ξ1^{2}$, so the effective hopping and effective disorder scale differently than earlier estimates assumed. They also identify the largest enhancement at energy E=0 and interaction strength u≈3t, and they confirm single-parameter scaling by showing that the participation number PN is of the same order as ξ2.","pith_inferences":["The reported N*(W) is calibrated from u=0 data; at u≈3 the extracted ξ2 is about 30 times larger, so the same system-size threshold may be too small to resolve the exponential tail. A direct check would be to re-extract ξ2 at large u with a u-dependent N*, and the sublinear part of F(x) would be invalid if the values rise substantially.","One testable extension is to treat the resonant Fock group as an effective one-dimensional chain with hopping teff≈u/ξ1 and disorder Weff≈Δ1/ξ1^2, and compare its exact localization length with the paper's F(uξ1); any discrepancy would isolate corrections beyond the effective-chain picture.","The energy anomaly at E≈1 and u≈t, where ξ2 is enhanced over the band center, suggests that F(uξ1) may need an additional energy argument; computing F at fixed nonzero E would test whether the collapse to a single variable survives away from E=0.","If the sublinear F(x) continues to grow without saturation, then in the limit W→0 even the two-particle system is still localized, just with a slowly increasing length; observing F(x) at uξ1 beyond 3000 would discriminate between slow growth and a logarithmic saturation."],"forward_implications":["At fixed disorder, the largest two-particle localization length occurs at spectrum center E=0 and interaction strength u≈3t, the order of the single-particle bandwidth; the enhancement decreases for larger u as doubly occupied states leave the two-particle continuum.","Earlier power-law scalings such as ξ2∝ξ1^2 or ξ2∝ξ1^{1.6} do not describe the asymptotic regime; extrapolating the old linear scaling overestimates ξ2/ξ1 by at least a factor of six at uξ1=1500.","Finite-size corrections materially reduce earlier estimates: the Green's function result of Ref. 20, which reported ξ2≈9ξ1 at ξ1≈400, is corrected down to about ξ2≈6ξ1 once the same extrapolation is applied.","Single-parameter scaling holds: the participation number PN stays within 1 to 1.5 times ξ2 across all studied disorder strengths and interactions, so ξ2 controls both the exponential decay and the wavefunction extent.","The mechanism predicts that breaking the particle-hole symmetry of the clean chain, for example by next-nearest-neighbour hopping, shrinks the resonant Fock-state groups and further reduces ξ2/ξ1."],"supporting_citations":[{"why":"introduced the projected Green's function method for ξ2 and supplied the small-u scaling data this paper compares against","marker":"[15]"},{"why":"showed how the tridiagonal structure reduces the Green's function computation to O(N^3), allowing system sizes up to N=20000","marker":"[16]"},{"why":"established the connectivity analysis identifying restoration of momentum conservation and the resonance condition ν=-μ for W<1","marker":"[13]"},{"why":"provides the finite-size extrapolation form ξ2(N)=ξ2+a/N used to extract the thermodynamic limit","marker":"[21]"},{"why":"previous Green's function computation at ξ1≈400 reporting ξ2≈9ξ1, whose finite-size corrections this paper re-evaluates to about 6ξ1","marker":"[20]"},{"why":"predicted a nonuniversal interaction-dependent exponent and the suppression of enhancement for u above the bandwidth, used to set umax","marker":"[12]"},{"why":"the ξ2∝ξ1^2 power-law prediction that the paper's asymptotic regime deviates from","marker":"[10]"}],"fun_headline_variants":["Two-particle localization hits 15x gain at weak disorder","Disorder pairs: 15x localization boost, sublinear growth","Record 15x two-particle localization length ratio","Weak disorder reveals 15x two-particle localization gain"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing assumption is that the finite-size extrapolation formula ξ2(N)=ξ2+a/N and the system-size threshold N*(W) calibrated on the non-interacting u=0 benchmark remain valid for every interaction strength, including the large-u cases where the reported localization length is tens of times larger; if that extrapolation breaks down, the observed sublinear growth of the scaling function would be an artifact.","fun_headline_variants_meta":{"raw":{"variants":["Two-particle localization hits 15x gain at weak disorder","Disorder pairs: 15x localization boost, sublinear growth","Record 15x two-particle localization length ratio","Weak disorder reveals 15x two-particle localization gain"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000482,"raw_usage":{"total_tokens":2364,"prompt_tokens":912,"completion_tokens":1452,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":528,"completion_tokens_details":{"reasoning_tokens":1386}},"tokens_in":528,"tokens_out":1452,"duration_ms":10927,"temperature":1.0,"reasoning_tokens":1386,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T12:39:21.082552+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A decisive test is to repeat the extraction at the record point W=0.5, u=3 using system sizes beyond the u=0 threshold, for example N=40000 with more disorder realizations, and to extrapolate ξ2(N) with a threshold determined at u=3 independently. If the extrapolated ratio ξ2/ξ1 then exceeds the reported value of about 16, or if the curve of F(uξ1) bends upward rather than saturating, the claim of a sublinear asymptotic scaling function is falsified. Equally, an independent transfer-matrix or exact-diagonalization computation at N>6000 that yields ξ2/ξ1≫16 at the same parameters would settle the question.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"introduced the projected Green's function method for ξ2 and supplied the small-u scaling data this paper compares against"},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"showed how the tridiagonal structure reduces the Green's function computation to O(N^3), allowing system sizes up to N=20000"},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"established the connectivity analysis identifying restoration of momentum conservation and the resonance condition ν=-μ for W<1"},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"provides the finite-size extrapolation form ξ2(N)=ξ2+a/N used to extract the thermodynamic limit"},{"cited_title":"\\ Frahm ,\\ title title Eigenfunction structure and scaling of two interacting particles in the one-dimensional anderson model , \\ 10.1140/epjb/e2016-70114-7 journal journal Eur","cited_arxiv_id":null,"evidence_quote":"previous Green's function computation at ξ1≈400 reporting ξ2≈9ξ1, whose finite-size corrections this paper re-evaluates to about 6ξ1"},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"predicted a nonuniversal interaction-dependent exponent and the suppression of enhancement for u above the bandwidth, used to set umax"}],"review_version":1}