{"id":"6a1dd804-0ab8-4ce6-8c8e-47591bb4d8d7","arxiv_id":"2508.19233","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":5,"one_line_summary":"Temperature dependence of Lanczos coefficients is governed by two decoupled Toda chains, yielding a 'Krylov bootstrap' consistency criterion and exponentially small Krylov complexity at low temperature.","lead":"This paper derives that the temperature dependence of Lanczos coefficients, the numbers encoding operator growth in quantum systems, is governed by exactly solvable Toda-chain equations. This yields new analytic control at low temperature, a consistency 'bootstrap' criterion, and a mechanism for the observed splitting of the coefficients.","discovery_kind":"first_principles","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Central 'full generality' claim fails when degenerate energy gaps are present: the two-Toda-chain dynamics requires T_perp=0 (Eq. 2.11), which excludes operators with diagonal parts and symmetric spectra; the paper's own Sec 2.3 admits the flow is not closed in bn alone.","rationale":"The reader's weakest assumption—the invariance of the Krylov space under −½{H,·}, i.e., T⊥n=0 in Eq. (2.11)—is exactly the condition on which the central integrability result depends. I agree that this is the most load-bearing point. The derivation of the Lax equations (2.23) and the two-Toda-chain decomposition (2.35) is careful and plausible under that assumption, and the worked examples plus the 5-site Ising numerical check provide independent support. However, the abstract's claim of 'full generality' is contradicted by the paper's own Sec 2.3 and Appendix B: when degenerate gaps are present (including the zero gap for any operator with a diagonal part), the flow is not closed in bn alone, and the extended matrices t or Y enter. This is not merely a technical caveat; it excludes many physical systems, including CFTs and any operator with nonzero thermal expectation value. The harmonic oscillator is a crisp counterexample: the closed equation predicts dTK/dβ=0, while the exact derivative is nonzero and is only reproduced by including the Y term. The factor-2 inconsistency in Sec 3.4 and the suppressed CFT bootstrap computation are secondary issues; they affect particular applications but not the core Toda-chain derivation. The reader's CONDITIONAL verdict is appropriate: the central argument is defensible under the stated condition, but the presentation overclaims and several supporting computations are omitted. My read does not change that verdict.","tokens_in":25306,"tokens_out":10752,"duration_ms":112017,"concrete_test":"Use the harmonic oscillator example of Appendix B: for A0=x, TK(β)=ω coth(βω/2) I. The closed Lax equation (2.23) with B=½(T+−T−) gives B=0 and hence dTK/dβ=0, whereas the exact derivative is −ω²/(2 sinh²(βω/2)) I. The difference is exactly the Y term of Eq. (2.119) (see Eq. B.3). If one instead computes the full extended T′ and verifies that Eq. (2.114) holds with Y, then the need for the extended Krylov space is demonstrated. This check settles that the central closed Toda-chain claim is conditional on the no-degenerate-gaps assumption, contradicting 'full generality'.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's headline claim—that β-dependence of Lanczos coefficients is governed by two independent Toda chains 'in full generality'—rests on the condition in Sec 2.1 (after Eq. 2.9) that A has no matrix elements connecting degenerate energy levels, so the components T⊥n in Eq. (2.11) vanish. Under this condition the Krylov space is invariant under −½{H,·} and the Lax equations (2.23) close. When the condition fails—which includes every operator with a non-zero diagonal part (zero gap is always degenerate) and any system with symmetries producing equal energy gaps (CFT, translation-invariant chains, harmonic oscillator)—the β-evolution of the physical Lanczos coefficients is not closed: Eqs. (2.116)-(2.119) require the extended matrix t or Y, which is not determined by bn alone. The abstract's 'full generality' is therefore unsupported; the two-Toda-chain structure holds only in the restricted no-degenerate-gaps scenario. The paper itself acknowledges this (Sec 2.3, Discussion), but the abstract and the phrasing of the central result overclaim. The harmonic oscillator (Appendix B) provides an explicit counterexample to the closed dynamics.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the temperature (β) dependence of Lanczos coefficients b_n(β) for thermal two-point functions defined with the Wightman inner product. The authors derive Lax equations for the tridiagonal matrix M representing the Liouvillian and for the matrix T representing the superoperator -1/2{H,·} in Krylov space, under the condition that this superoperator does not move operators outside the Krylov space. In that no-degenerate-gaps setting, they show that even and odd Lanczos coefficients decouple into two independent Toda chains related via initial conditions. They apply this structure to the large-β limit, where half the Lanczos coefficients vanish and the other half approach energy gaps, and to the moderate-β regime, where they explain staggering via spectral gaps and delta-function weight in the spectral function. They also introduce a 'Krylov bootstrap' consistency argument, claiming that the known CFT Lanczos coefficients force a degenerate spectrum, and propose a numerical method to compute C_β(t) from β=0 data.","tokens_in":25526,"tokens_out":5248,"duration_ms":57502,"significance":"If the main derivation stands, this is a valuable contribution: it connects the recursion-method/Lanczos-coefficient formalism to the Toda hierarchy, gives an integrable interpretation of β-dependence, and produces concrete analytic predictions for low temperatures, including the exponential suppression of Krylov complexity. The paper contains clean derivations in Sec. 2.1, exact solvable examples (Examples I–III), and a numerical demonstration of the flow method in Sec. 3.2. These are genuine strengths. However, the advertised 'full generality' is not achieved: the two-Toda-chain result holds only under a restrictive no-degenerate-gaps condition, and several of the application-level claims rest on asserted or heuristic steps. These issues are fixable but require substantive revision.","major_comments":[{"comment":"The abstract states the result 'in full generality,' but the derivation of the two independent Toda chains requires T⊥_n = 0 in Eq. (2.11), i.e. A has no matrix elements connecting degenerate energy levels. This excludes every operator with nonzero diagonal part and any system with degenerate nonzero gaps. When this condition fails, the flow of the physical Lanczos coefficients is not closed: Eqs. (2.116)–(2.118) involve t or Y, which are not determined by b_n alone. The harmonic oscillator example in Appendix B makes this explicit: T_K satisfies (B.2) only with an external Y. The paper should either restrict the central claim to the no-degenerate-gaps case or present the extended-space Lax dynamics as the general integrable statement.","section":"Abstract; Sec. 2.1, 2.3"},{"comment":"The Krylov bootstrap conclusion—that the explicit CFT Lanczos coefficients b_n^2=(n+1)(n+2Δ)π^2/β^2 force a degenerate spectrum—is asserted, not demonstrated. The sentence 'Evaluating the right hand side of (2.76), one finds it not to match the left hand side' is a load-bearing step. Please display the explicit mismatch (at least for the first non-trivial n), or provide a supplementary computation, so the claim is verifiable.","section":"Sec. 3.1, after Eq. (3.4)"},{"comment":"There is an internal inconsistency in the exponential decay. The text states b_0^2 ≈ e^{-βm/2}; substituting into Eq. (3.19) with b_1 finite gives ln K ≈ -βm/2 + O(1), not -βm. As written, Eq. (3.20) contradicts the abstract, the introduction, and Fig. 2. Please correct Eq. (3.20) to ln K ≈ -βm/2, or if the intended asymptotic is -βm, change the stated b_0^2 scaling accordingly.","section":"Sec. 3.4, Eq. (3.20)"},{"comment":"The large-β diagonalization argument is used to derive b_{2k+1}→0 and b_{2k}→|E_i-E_j|, which underlies all low-temperature predictions. The argument is heuristic: the boundedness of the partial sums in Eq. (3.7) is assumed rather than shown, the eigenvalue-ordering argument around Eq. (3.8) is a perturbation statement, and the claim that the doubly degenerate eigenvalues are ordered with doublets consecutive requires justification. Since these conclusions are central, please either supply a precise statement with the relevant convergence theorem from Refs. [15,16] or explicitly mark this part as heuristic.","section":"Sec. 3.3, Eqs. (3.7)–(3.9)"}],"minor_comments":[{"comment":"The notation T versus \\tilde{T} is confusing; the reader must track which matrix is tridiagonalized and which evolves in the original basis. Please define the relationship once and consistently.","section":"Sec. 2.1.1, Eqs. (2.33)–(2.35)"},{"comment":"The termination condition for finite Krylov space is stated parenthetically. Please spell out the index ranges and the convention for b_n=0 for n≥N, since this formula is used repeatedly.","section":"Eq. (2.25)"},{"comment":"Typo: there is a stray comma in 'b_n^2 = (n+1)(n+2Δ)π^2/β^2, .' Remove the comma.","section":"Eq. (3.2)"},{"comment":"The fitting constant is denoted '#'; use a conventional symbol such as c or A to avoid confusion.","section":"Fig. 2 caption"},{"comment":"Typo: 'Lnczos' should be 'Lanczos'.","section":"Appendix C"}],"recommendation":"major_revision","confidential_remarks":"The paper has a strong core and is likely acceptable after revision. The main concerns are the overclaim in the abstract regarding full generality and the missing explicit computation behind the Krylov bootstrap. The harmonic oscillator example in Appendix B is useful evidence that the closed two-Toda-chain form is not general; the authors should use it to calibrate the claims. The inconsistency in Eq. (3.20) should be fixed before publication."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"First, the genuinely useful result: the beta-dependence of the Lanczos coefficients is an isospectral deformation, and when the operator stays inside the Krylov space under -1/2{H,.}, the flow closes and decouples into two Toda chains. Sec 2.1 is careful linear algebra; the QR decomposition step and the commutator argument are standard and correct. The worked spin examples check out. The large-beta picture—half the coefficients vanishing, the other half going to energy gaps—is a nice application of Toda convergence theorems, and the numerical integration from beta=0 to beta=2 on a 5-site Ising chain is a reasonable proof-of-concept.\n\nThe soft spots are real but not fatal. The abstract says 'in full generality,' but the derivation requires that the Krylov space be invariant under -{H,.}, which fails for operators with diagonal parts and whenever there are degenerate gaps. The paper knows this: Sec 2.2 handles the diagonal-part case with an extra function mu, and Sec 2.3 admits the flow is not closed in b_n alone. The abstract and the phrasing around eq (2.23) overclaim. That needs fixing, not by changing the math but by stating the domain.\n\nTwo smaller issues. The CFT 'Krylov bootstrap' conclusion in Sec 3.1 rests on 'one finds it not to match'—the mismatch computation is not displayed. It may well be right, but it is a load-bearing assertion. And in Sec 3.4 there is a factor-of-two inconsistency: if b0^2 ~ e^{-beta m/2}, then K ~ b0^2 gives ln K ~ -beta m/2, not -beta m as printed in (3.20). The abstract claims K ~ e^{-beta m/2}, so (3.20) is the odd one out. The b0^2 estimate itself is asserted, not derived. These are addressable.\n\nSec 4 is honestly framed as a model, and the paper explicitly lists counterexamples (XY with h=1.1). I trust that section more because it does not oversell.\n\nWho should read this: anyone working on Krylov complexity or temperature-dependent Lanczos coefficients. The bootstrap idea is speculative but interesting. I would send it to a serious referee: the core result is worth vetting, and the referee can demand the missing computation and the exponent fix. I also think the 'full generality' claim needs to be walked back in the abstract before publication.","headline":"Two Toda chains for beta-dependence of Lanczos coefficients is real and worth engaging, but 'full generality' in the abstract oversells: the closed two-chain flow requires the no-degenerate-gaps condition, and there is a factor-2 slip in the K~e^{-beta m/2} claim.","tokens_in":26194,"tokens_out":3247,"would_cite":true,"duration_ms":32322,"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 Toda chains drive Lanczos coefficients as temperature changes","keywords":["Krylov complexity","Lanczos coefficients","Toda chain","integrable dynamics","thermal two-point function","isospectral deformation","Krylov bootstrap","staggering"],"falsifier":"Take a small system with no degenerate energy gaps and an off-diagonal initial operator (e.g. the four-state model of Sec 2.1.2), compute b_n(β) by direct Lanczos at β and β+δ, and compare the numerical derivative to the right-hand sides of (2.53)–(2.54); any discrepancy beyond integration error would falsify the closed Toda flow. Alternatively, run the same check on the harmonic oscillator: the paper itself shows the closed equations fail there because the extra term Y in (2.118) is nonzero, so the oscillator cleanly delimits the claim's domain.","tokens_in":25010,"feed_emoji":"🌡️","tokens_out":8575,"duration_ms":85673,"temperature":0.7,"pith_summary":"The paper shows that when the Lanczos recursion is built from a thermal inner product, changing the inverse temperature β is an isospectral deformation: the tridiagonal matrix representing the Liouvillian evolves by Lax equations whose even- and odd-index parts are two independent Toda chains. Because Toda flows diagonalize at late times, the very-low-temperature limit becomes tractable: odd Lanczos coefficients vanish, even ones tend to energy gaps, and time-averaged Krylov complexity falls as e^{-βm/2}. The same flow equations impose consistency conditions: applied to 2d CFT two-point functions, they force the spectrum to be degenerate (\"Krylov bootstrap\"), and they explain the empirically observed staggering of Lanczos coefficients as the combined effect of a spectral gap and a constant part of the correlator. A reader should care because this recasts temperature dependence—normally found by redoing the Lanczos algorithm at each β—as a solvable dynamical system connecting all temperatures, with analytic control in the low-temperature regime.","feed_headline":"Two Toda chains drive Lanczos coefficients as temperature changes","feed_subtitle":"At very low temperature, odd coefficients vanish, even ones become energy gaps, and Krylov complexity decays as e^{-βm/2}.","key_machinery":"The central objects are the tridiagonal Liouvillian matrix M (representing [H,·] in the Krylov basis) and the matrix T representing the superoperator -1/2{H,·}; their commutativity [M,T]=0, together with the QR decomposition of e^{T(β-β0)/2}, produces the Lax equations (2.23). The even and odd parts of T decouple into two Toda chains, related only through their conserved quantities at the initial condition. The eigenvalues of T are the energy sums -1/2(E_i+E_j), and the late-time diagonalization theorem for Toda flows is what yields the asymptotic Lanczos coefficients and the exponential decay of Krylov complexity.","core_discovery":"The central claim is that the β-dependence of Lanczos coefficients b_n(β) is governed by the Lax pair dT/dβ=[B,T], dM/dβ=[B,M] with B=(T_+−T_-)/2, where M is the Liouvillian in the Krylov basis and T represents -1/2{H,·}. After projecting on even and odd indices, this splits into two independent Toda chains whose only relation is the equality of conserved quantities at the initial condition. In the large-β limit βm≫1, the Toda flow forces T to diagonal form, so b_{2k+1}→0 and b_{2k}→|E_i−E_j|; the time-averaged Krylov complexity then decays as e^{-βm/2}. The paper also introduces the \"Krylov bootstrap\": consistency of the flow with the known CFT Lanczos coefficients b_n^2=(n+1)(n+2Δ)π^2/β^2","pith_inferences":["One could use the Krylov bootstrap in reverse: given measured or conjectured Lanczos coefficients at two temperatures, failure of the closed Lax equations would pinpoint the presence of degenerate energy gaps or of a diagonal component in the operator.","The late-time Toda diagonalization suggests a way to extract the energy-gap spectrum from the even Lanczos coefficients at low temperature; testing this on larger interacting systems would show whether the β→∞ limit is reached before finite-size effects dominate.","The β=0 numerical integration scheme, if it scales, would offer an alternative route to finite-temperature dynamics that does not require imaginary-time evolution or matrix-product-state thermalisation; the paper only demonstrates five sites, so scalability is the open question.","The paper's distinction between two staggering mechanisms (spectral gap m vs constant part κ) predicts that systems with the same spectral function shape but different κ will show different branch splittings; this is directly testable in exactly solvable models like the XY chain."],"forward_implications":["Thermal correlators at any β can be obtained by integrating the Lax equations from β=0 Lanczos data, bypassing a fresh Lanczos run at each temperature.","At very low temperature, the Krylov chain effectively truncates: odd coefficients vanish, even coefficients encode the energy gaps, so Lanczos data becomes a spectral probe of |E_i−E_j|.","Time-averaged Krylov complexity is exponentially small in βm/2 in the low-temperature limit, essentially independent of system size.","The Krylov bootstrap turns the flow equations into a consistency test on candidate spectral functions: any proposed b_n(β) that cannot satisfy the closed Lax equations within the Krylov space signals either degeneracies or a nonzero diagonal part of the operator."],"supporting_citations":[{"why":"Established the Toda-lattice form of Euclidean-time dependence of Lanczos coefficients, the template for the β-flow here.","marker":"[5]"},{"why":"Supplied the explicit CFT Lanczos coefficients used in the Krylov bootstrap argument.","marker":"[14]"},{"why":"Introduced Krylov complexity in quantum field theory and the spectral-gap/staggering observations, plus the Dyck-path formalism used in Sec 4.","marker":"[4]"},{"why":"Provided the exactly soluble Toda-type system used as Example II for explicit β-dependent Lanczos coefficients.","marker":"[13]"},{"why":"Gave the late-time diagonalization theorem for Toda flows underpinning the large-β asymptotic analysis.","marker":"[15]"},{"why":"Extended Toda-flow integrability and asymptotics to infinite-dimensional settings, supporting the β→∞ limit.","marker":"[16]"},{"why":"Supplied the orthogonal-polynomial theorem giving the even/odd branch asymptotics |b_e−b_o|=m for spectral gaps.","marker":"[17]"},{"why":"Gave the recursion relating Lanczos coefficients when a constant is added to the autocorrelation function, used for the κ-induced staggering analysis.","marker":"[18]"}],"fun_headline_variants":["Temperature flow of Lanczos coefficients splits into two Toda chains","At low T, odd Lanczos vanish, even become energy gaps","Krylov bootstrap ties integrable flow to degenerate spectra","Integrable dynamics govern Lanczos coefficient temperature dependence"],"cache_read_input_tokens":2688,"weakest_assumption_plain":"The derivation assumes that moving an operator toward lower temperature, by conjugating with e^{-βH/2}, never leaves the original Krylov space—true only when the operator has no matrix elements between equal-energy states and no on-diagonal part; if that fails, the equations for the Lanczos coefficients are not closed, and the paper's CFT conclusion relies on an asserted mismatch that is not shown.","fun_headline_variants_meta":{"raw":{"variants":["Temperature flow of Lanczos coefficients splits into two Toda chains","At low T, odd Lanczos vanish, even become energy gaps","Krylov bootstrap ties integrable flow to degenerate spectra","Integrable dynamics govern Lanczos coefficient temperature dependence"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000198,"raw_usage":{"total_tokens":1208,"prompt_tokens":754,"completion_tokens":454,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":498,"completion_tokens_details":{"reasoning_tokens":386}},"tokens_in":498,"tokens_out":454,"duration_ms":5087,"temperature":1.0,"reasoning_tokens":386,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-05T15:51:05.058049+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take a small system with no degenerate energy gaps and an off-diagonal initial operator (e.g. the four-state model of Sec 2.1.2), compute b_n(β) by direct Lanczos at β and β+δ, and compare the numerical derivative to the right-hand sides of (2.53)–(2.54); any discrepancy beyond integration error would falsify the closed Toda flow. Alternatively, run the same check on the harmonic oscillator: the paper itself shows the closed equations fail there because the extra term Y in (2.118) is nonzero, so the oscillator cleanly delimits the claim's domain.","supporting_citations":[{"cited_title":"Dymarsky and A","cited_arxiv_id":null,"evidence_quote":"Established the Toda-lattice form of Euclidean-time dependence of Lanczos coefficients, the template for the β-flow here."},{"cited_title":"Dymarsky and M","cited_arxiv_id":null,"evidence_quote":"Supplied the explicit CFT Lanczos coefficients used in the Krylov bootstrap argument."},{"cited_title":"Kac and P","cited_arxiv_id":null,"evidence_quote":"Provided the exactly soluble Toda-type system used as Example II for explicit β-dependent Lanczos coefficients."},{"cited_title":"Moser, Finitely many mass points on the line under the influence of an exponential potential – an integrable system , in Dynamical Systems, Theory and Applications , J","cited_arxiv_id":null,"evidence_quote":"Gave the late-time diagonalization theorem for Toda flows underpinning the large-β asymptotic analysis."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Extended Toda-flow integrability and asymptotics to infinite-dimensional settings, supporting the β→∞ limit."},{"cited_title":"Chihara, An Introduction to Orthogonal Polynomials , Dover (2011)","cited_arxiv_id":null,"evidence_quote":"Supplied the orthogonal-polynomial theorem giving the even/odd branch asymptotics |b_e−b_o|=m for spectral gaps."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Gave the recursion relating Lanczos coefficients when a constant is added to the autocorrelation function, used for the κ-induced staggering analysis."}],"review_version":1}