{"id":"2e7cfd61-1006-4a2a-9f74-d99aab3f9458","arxiv_id":"1908.03198","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"The paper derives analytic corrections that let fermion-boson vertices and Bethe-Salpeter equations be solved with a small frequency box, improving convergence from 1/ν_max to 1/ν_max^4.","lead":"This paper derives a mathematical shortcut for computing interaction vertices in strongly-correlated electron systems, so only a small set of frequencies has to be handled numerically. It could make a family of many-body methods faster and more accurate.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Claimed O(1/ν_max^4) accuracy hinges on an assumed 1/ν^3 decay of irreducible-vertex corrections beyond Eq. (3); this exponent is not derived and the numerics test only one model.","rationale":"We read the paper in good faith and found the derivation internally consistent within its stated ladder-like assumptions. We traced the central claim to the assumed asymptotic behavior Eq. (3) and the subleading decay 1/ν^3 of the irreducible vertex deviations; this is exactly the reader's weakest assumption. The paper itself uses 'expected' for this decay (Sec. IV, Conclusion), and the numerical check in Sec. V is a single-model test without error bars or a direct power-law analysis. We considered whether an internal algebraic gap (Eq. (24) to Eq. (25)) is more load-bearing, but the alternative derivation from Ref. [33] and the numerical consistency of the final formulas with CT-QMC susceptibilities reduce that risk. We also considered the restriction to vertices depending on a single momentum transfer q; since the paper explicitly targets ladder-like non-local extensions, this is a scope condition rather than a flaw. Therefore the conditional verdict is appropriate: the method is plausible and numerically supported, but the central accuracy scaling rests on an unproven asymptotic exponent. Our concrete test would settle whether that exponent is actually 3 or smaller.","tokens_in":11924,"tokens_out":17856,"duration_ms":174872,"concrete_test":"Compute the exact or numerically exact irreducible vertex Φ_{νν'ω} of the single-orbital Hubbard atom (or the DMFT impurity) at large Matsubara frequencies with |ν-ν'| fixed, and extract the exponent α of δΦ = Φ - (U_q + Φ̄) ∝ 1/ν^α. If α < 3, the out-of-box sums in Eqs. (8) and (25) acquire a residual error O(1/ν_max^{α+1}), so the claimed improvement would degrade accordingly. A direct way: evaluate the left- and right-hand sides of Eq. (25) for the Hubbard atom with box sizes Nf varying by a factor of 2–4, and test whether the deviation between the physical Φ and the box-corrected expression scales as 1/Nf^4 rather than 1/Nf^3.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's central improvement — reducing the finite-box error from O(1/ν_max) to O(1/ν_max^4) — depends on the assertion (Sec. IV, after Eq. (25), and Conclusion) that the difference between the true irreducible vertex and the asymptotic form Eq. (3) decays as 1/max(|ν|,|ν'|)^3. All correction functions X_q and Z_q (Eqs. (7), (8), (25)) are constructed from the U_q + Φ̄ tail; any component δΦ of Φ outside this tail with a slower decay is not accounted for and enters the out-of-box sums at lower order. The paper does not derive the 1/ν^3 exponent; it is stated only as 'expected.' The numerical verification in Sec. V is limited to the spin channel of a single 2D Hubbard model (U=10t, t'=0.15t, n=0.96, T=0.08t), fits the corrected vertices with a+b/ν^4+c/ν^5 without error bars, and does not directly measure the exponent of δΦ or demonstrate that a 1/ν^2 or 1/ν^3 contamination is absent. Since the method is proposed for a broad class of non-local DMFT extensions (ladder DΓA, dual fermion/boson, TRILEX, DMF2RG), the central accuracy claim is not yet established beyond this single test case.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript proposes an analytic treatment of fermionic frequency tails in Bethe-Salpeter equations and fermion-boson vertices for non-local extensions of DMFT. The key idea is to split all frequency summations into a small box, where vertices are known numerically, and the outside region, where the asymptotic form of the irreducible vertex, Eq. (3), is used. The outside contribution is expressed in closed form through the quantities X_q and Z_q, yielding explicit formulas for the full and reduced fermion-boson vertices (Eqs. (8), (14), (21)), the full and irreducible susceptibilities (Eqs. (11), (18)), and the two-particle irreducible vertex (Eq. (25)). The method is tested on spin-sector DMFT results for a two-dimensional Hubbard model, with reported agreement with CT-QMC local susceptibilities and a much weaker dependence on frequency-box size than the uncorrected calculation.","tokens_in":12118,"tokens_out":8254,"duration_ms":82445,"significance":"If the central claims hold, the work is a useful contribution: the correction formulas are explicit, parameter-free, and reduce the need for large frequency boxes and for separate numerical treatment of the large box in earlier schemes. The algebraic nature of the derivation and the concrete numerical demonstration of improved convergence with box size are strengths. However, the headline O(1/ν_max^4) error-scaling claim rests on an unproven decay assumption for the residual irreducible vertex, and the numerical support is limited to one channel and one model parameter set, which weakens the generality asserted in the abstract and conclusion.","major_comments":[{"comment":"The central claim that the method achieves O(1/ν_max^4) accuracy is not established. The argument assumes that the deviation of the true irreducible vertex from the asymptotic form (3) decays as 1/max(|ν|,|ν'|)^3; this exponent is not derived, and the text describes it only as 'expected.' Because X_q and Z_q are built entirely from the U_q + \\barΦ tail, any component of Φ outside that tail with a slower power law will enter the out-of-box sums at lower order. The numerical fits in Sec. V (Fig. 3) use an a + b/ν_max^4 + c/ν_max^5 form, which presupposes the claimed scaling and therefore cannot by itself confirm the exponent. To support the conclusion, the author should either derive (or cite a derivation of) the 1/ν^3 decay of the residual, or directly measure this residual as a function of frequency and demonstrate the absence of 1/ν^2 or 1/ν^3 contamination.","section":"Section IV, discussion following Eq. (25)"},{"comment":"The key relation (25) is obtained by 'algebraic transformations' that are not shown, and Eq. (21) similarly refers to 'algebraic manipulations' of Appendix C of Ref. [30]. Since Eq. (25) is the central new result and is used to convert the box-restricted vertex Φ^{box} into the physical Φ, the derivation should be reproduced in detail, for example in an appendix. Without this, the reader cannot verify the consistency of signs, the treatment of the out-of-box sums, or the order at which terms are neglected. The numerical checks in Sec. V are too limited to substitute for this derivation.","section":"Section IV, transition from Eq. (24) to Eq. (25)"},{"comment":"The numerical validation covers only the spin channel of a single two-dimensional Hubbard model with U=10t, t'=0.15t, n=0.96, T=0.08t (and one higher-temperature half-filled run). The paper advertises the method for a broad class of non-local DMFT extensions (ladder DΓA, dual fermion/boson, TRILEX, DMF2RG), but the asymptotic form (3) and the assumed 1/ν^3 decay of its correction are model- and approximation-dependent. In addition, the comparison of the local susceptibility from Eq. (16) with the CT-QMC result is partly a consistency check, because the tail functions X_q and Z_q themselves use local susceptibilities obtained from the same CT-QMC solver. At minimum, the authors should state these limitations explicitly in the conclusions and provide evidence in at least one further channel, parameter set, or non-local context.","section":"Section V"}],"minor_comments":[{"comment":"The axis label 'tX0, Z' in panel (b) is unclear; please spell out the plotted quantities (e.g., tX_0^{(1)}, tX_0^{(2)}, Z_{ν,0}-1) and ensure that all axis labels render correctly.","section":"Fig. 3"},{"comment":"The statement that the second term in X_q and the difference Z_q-1 are 'verified numerically' should be qualified: the verification is performed only for the spin channel of the Hubbard model, not for the full range of claimed applications.","section":"Sec. III, after Eq. (8)"},{"comment":"The extrapolations shown by dotted lines and the fits with a+b/ν_max^4+c/ν_max^5 are presented without error bars or fit parameters; reporting these numbers would make the claimed scaling easier to assess.","section":"Sec. V, Fig. 3"},{"comment":"The sentence 'we find the results of extrapolation consistent with those for vertices, obtained without account of finite frequency box effects' is not quantified; please provide the extrapolated vertex values and the corresponding fit parameters.","section":"Sec. V"}],"recommendation":"major_revision","confidential_remarks":"The paper is a single-author contribution and builds closely on Refs. [32,33]; the novelty relative to 'Method 2' of Ref. [33] should be delineated more clearly in the introduction and in the discussion of Eq. (25). This is not a reason to reject, but it would help the editor and the reader assess the incremental contribution. The main revision requested concerns the derivation of the central relation and the evidence for the 1/ν^4 scaling claim."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThis paper gives closed-form analytic corrections for finite-frequency-box truncation in non-local extensions of DMFT. What is actually new: instead of numerically treating vertex asymptotics in a large outer box, as in the box-splitting schemes of Kunes and Tagliavini et al., Katanin derives formulas involving X_q and Z_q so that only the small box needs numerical treatment. The core relation, Eq. (25), connecting the physical 2PI vertex to the box vertex via U_q − Zν Ũ_q Zν′, is the useful piece. The derivation is parameter-free, and the paper is honest that Eq. (25) can be obtained from Method 2 of Ref. [33]; the new contribution is making the correction analytic and cheap.\n\nWhat it does well: the algebra is internally consistent, and the numerical tests on the 2D Hubbard model show corrected vertices approaching CT-QMC susceptibilities to 0.01–0.3% and convergence with box size that is much flatter than the uncorrected 1/ν_max behavior. That is credible evidence the machinery works in the tested regime. The citation pattern is fine; prior work is credited, and self-citation is not a problem here.\n\nSoft spots: first, the central accuracy claim—O(1/ν_max^4) error—rests on the assumed 1/ν^3 decay of the irreducible vertex beyond the Eq. (3) asymptotic. That exponent is not derived; it is described as expected, and the numerics do not directly measure the decay of δΦ. This is a real limitation, but not a fatal one: the method remains useful even if the true decay is slower, it just weakens the advertised scaling. Second, the step from Eq. (24) to Eq. (25) is compressed as algebraic transformations. The author points to the alternative derivation, but a reader should not have to reconstruct Ref. [33] to verify the main result. Third, the numerical validation is one model, one parameter set, spin channel only, no error bars on the fits. For a method paper aimed at broad use, another channel or a second model would strengthen the claim. These are addressable, not disqualifying.\n\nWho it is for: anyone doing ladder DΓA, dual fermion/boson, TRILEX, or 2PI-fRG who is currently brute-forcing frequency boxes. I would send it to a serious referee, with the request to expand the derivation of Eq. (25) and clarify the status of the decay assumption. It is not a breakthrough, but it is a practical and honest improvement.","headline":"A genuinely useful, honestly-scoped methodological advance: analytic out-of-box corrections for non-local DMFT vertices, with the advertised O(1/ν^4) scaling resting on an unproven but plausible decay assumption.","tokens_in":12737,"tokens_out":1660,"would_cite":true,"duration_ms":16587,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper derives analytic formulas that remove most of the finite-frequency-box error in non-local dynamical mean-field theory, improving the truncation error from O(1/ν_max) to O(1/ν_max^4).","keywords":["fermion-boson vertex","Bethe-Salpeter equation","dynamical mean-field theory","non-local correlations","frequency box asymptotics","two-particle irreducible vertex","Hubbard model"],"falsifier":"Take a model in which the asymptotic form is known to fail, such as a non-ladder vertex with retarded non-local interactions, evaluate the full two-particle irreducible vertex on a very large frequency box, and compare the left and right sides of Eq. (25) as the box size grows. If the difference does not shrink like $1/\\nu_{\\max}^4$ — or if the correction pushes the vertex away from the exact large-box value — the assumed tail is the culprit.","tokens_in":11632,"feed_emoji":"📉","tokens_out":11062,"duration_ms":103161,"temperature":0.7,"pith_summary":"This paper claims that, in non-local extensions of dynamical mean-field theory, the error caused by truncating the Matsubara frequency grid to a finite box can be reduced from a slow $1/\\nu_{\\max}$ decay to a much faster $1/\\nu_{\\max}^4$ decay. The author derives closed formulas that analytically account for all contributions of frequencies outside the box, so only the numerically exact vertices inside a small box are needed. These formulas express the full fermion-boson vertices, susceptibilities, and the two-particle irreducible vertex in terms of two correction functions, $X_q$ and $Z_{\\nu q}$, built from the known high-frequency tail of the vertex. The central identity, Eq. (25), relates the physical two-particle irreducible vertex to the vertex obtained by inverting the Bethe-Salpeter equation inside the box. This makes finite-box artifacts nearly negligible and allows much smaller frequency boxes in practice.","feed_headline":"Correlated-electron vertex errors now fall as 1/ν^4","feed_subtitle":"Analytic tail corrections let small frequency boxes replace large ones in non-local dynamical mean-field theory.","key_machinery":"The analytic tail correction $X_q$ and the vertex correction factor $Z_{\\nu q}$ are the objects that carry the argument. $X_q$ is the sum over frequencies outside the box of the bare bubble plus one insertion of the asymptotic vertex tail; $Z_{\\nu q}$ measures how much the three-leg vertex is renormalized by the tail outside the box. Both are computed from the asymptotic form (3)-(4), which uses only local charge, spin, and particle-particle susceptibilities, so they are cheap to evaluate for arbitrarily large frequencies. Eq. (25) is the load-bearing identity: $\\Phi_{\\nu\\nu'q} = \\Phi^{\\mathrm{box}}_{\\nu\\nu'q} + U_q - Z_{\\nu q}\\tilde U_q Z_{\\nu'q}$, with $\\tilde U_q = U_q/(1-U_q X_q)$. This identity removes the $1/\\nu_{\\max}$ error and leaves the smaller $1/\\nu_{\\max}^4$ error.","core_discovery":"The central claim is that the out-of-box frequency contributions, which previously required numerical treatment of the asymptotic tail in a larger box, can be evaluated analytically to leading order. Starting from the asymptotic form (3) of the two-particle irreducible vertex, the paper defines $X_q$ and $Z_{\\nu q}$ and derives Eqs. (8), (11), (14), (18), and (25) for the full fermion-boson vertex, the non-local susceptibility, the reduced fermion-boson vertex, the irreducible susceptibility, and the two-particle irreducible vertex, respectively. The relation $\\Phi_{\\nu\\nu'q} = \\Phi^{\\mathrm{box}}_{\\nu\\nu'q} + U_q - Z_{\\nu q}\\tilde U_q Z_{\\nu'q}$ is the key identity: it converts a matrix inversion performed only inside the box into the physical vertex. In the companion numerical test on the two-dimensional Hubbard model, the corrected vertices change very little with box size, while the uncorrected ones extrapolate slowly; the corrected error is consistent with $1/\\nu_{\\max}^4$ scaling.","pith_inferences":["If the same asymptotic form holds in the particle-particle channel, an analogous set of box corrections could be derived for pairing susceptibilities, where the tail is built from the same local susceptibilities.","The quartic improvement is conditional on the decay of the deviation from Eq. (3); in models with long-range or retarded interactions the effective exponent would be lower, and the size of $X_q$ and $Z_{\\nu q}$ would reveal that.","The construction suggests a general recipe for any truncated Matsubara summation: add back the analytically known tail before inverting, rather than enlarging the box."],"forward_implications":["A single calculation with a small frequency box should suffice for ladder-type non-local DMFT calculations, instead of running several box sizes and extrapolating.","The method gives a direct relation between the local and non-local two-particle irreducible vertices, making it easier to feed local input into non-local Bethe-Salpeter equations.","The same formulas apply to both charge and spin channels, and they can be adapted to cases where the vertex depends on more than one momentum transfer.","The claimed accuracy means the main remaining error is set by the physical decay of the irreducible vertex toward its asymptotic form, not by the truncation."],"supporting_citations":[{"why":"It introduced the two-box splitting of the frequency grid, the approach that this paper makes analytic.","marker":"[32]"},{"why":"It supplied the asymptotic form of the two-particle irreducible vertex used in Eq. (3) and the method that Eq. (25) generalizes.","marker":"[33]"},{"why":"It provided the algebraic manipulations, used here to simplify the reduced fermion-boson vertex and to derive Eq. (21).","marker":"[30]"},{"why":"It defined the reduced fermion-boson vertex that Eqs. (13)-(14) are built around.","marker":"[34]"}],"fun_headline_variants":["Analytic tail corrections replace numerical vertex extrapolation","Small frequency boxes now exact with analytic tail terms","Vertex asymptotics handled analytically in non-local DMFT","Out-of-box vertex contributions solved analytically","Fermion-boson vertices fixed analytically beyond the box"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The whole scheme depends on the high-frequency asymptotic form (3) of the two-particle irreducible vertex, together with the assumption that deviations from this form decay at least as $1/\\max(|\\nu|,|\\nu'|)^3$; if that tail is wrong or decays too slowly, the size of the correction terms $X_q$ and $Z_{\\nu q}$ would be misestimated.","fun_headline_variants_meta":{"raw":{"variants":["Analytic tail corrections replace numerical vertex extrapolation","Small frequency boxes now exact with analytic tail terms","Vertex asymptotics handled analytically in non-local DMFT","Out-of-box vertex contributions solved analytically","Fermion-boson vertices fixed analytically beyond the box"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000422,"raw_usage":{"total_tokens":2190,"prompt_tokens":986,"completion_tokens":1204,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":602,"completion_tokens_details":{"reasoning_tokens":1131}},"tokens_in":602,"tokens_out":1204,"duration_ms":10140,"temperature":1.0,"reasoning_tokens":1131,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T14:21:38.834306+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take a model in which the asymptotic form is known to fail, such as a non-ladder vertex with retarded non-local interactions, evaluate the full two-particle irreducible vertex on a very large frequency box, and compare the left and right sides of Eq. (25) as the box size grows. If the difference does not shrink like $1/\\nu_{\\max}^4$ — or if the correction pushes the vertex away from the exact large-box value — the assumed tail is the culprit.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"It introduced the two-box splitting of the frequency grid, the approach that this paper makes analytic."},{"cited_title":"Tagliavini, S","cited_arxiv_id":null,"evidence_quote":"It supplied the asymptotic form of the two-particle irreducible vertex used in Eq. (3) and the method that Eq. (25) generalizes."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"It provided the algebraic manipulations, used here to simplify the reduced fermion-boson vertex and to derive Eq. (21)."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"It defined the reduced fermion-boson vertex that Eqs. (13)-(14) are built around."}],"review_version":1}