{"id":"27fc74d8-270e-4208-b3f7-cb42c5bc1278","arxiv_id":"1909.02793","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":1,"one_line_summary":"A parquet-like scheme for the Hedin three-leg vertex self-consistently computes vertex corrections without four-point vertices or Bethe-Salpeter inversions, demonstrated on impurity models.","lead":"This paper derives a new set of self-consistent equations for the Hedin three-leg vertex, avoiding the expensive four-point vertex storage and Bethe-Salpeter inversion of the standard parquet method. If confirmed, this could make unbiased treatments of competing magnetic, charge, and pairing fluctuations feasible for correlated lattice models.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The AIM numerical test freezes Σ and π, so the full self-consistent cycle claimed in the abstract is not demonstrated.","rationale":"The reader's verdict is already CONDITIONAL, and the reader's weakest_assumption concerns the locality approximation for φ^firr in lattice extensions. My stress-test identifies a different, more immediate gap: the AIM numerical demonstration does not actually run the full self-consistent cycle, because the self-energies are held fixed. Since the central claim explicitly includes the self-energy and polarization as self-consistently determined quantities, this missing numerical test is load-bearing. The derivation itself appears coherent and is supported by the exact atomic-limit benchmark where the full cycle is run. Thus the concern does not overturn the conditional acceptance; it sharpens the condition: acceptance should explicitly require a demonstration of the full cycle on an AIM or small lattice. The reader's focus on lattice applicability and the locality of φ^firr is related but distinct; hence partial agreement.","tokens_in":19487,"tokens_out":3501,"duration_ms":37693,"concrete_test":"Implement the complete cycle of Fig. 5 for the AIM at U=4, β=5 without freezing Σ and π: start from the non-interacting guess (or an annealed guess), update w and g via (22a), (22b), (24), then λ via (15) and (18), then π and Σ via (20), (21), (23), using e.g. Broyden mixing; compare the converged λ, Σ, π with the exact DMFT/impurity-solver results. If the full map fails to converge to the exact fixed point, the abstract's claim that the scheme converges for the AIM is unsupported.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim is that the SBE decomposition yields a closed, self-consistent scheme that reconstructs the Hedin vertex, the polarization, and the self-energy. The derivation in Sec. III is algebraic and appears internally consistent. The load-bearing evidence gap is in Sec. IV B: for the Anderson impurity model the authors feed in the exact fully irreducible vertex together with the exact self-energies Σ and π, \"which we do not update in the cycle,\" and then iterate only the Hedin-vertex equations (15) and (18), using the exact vertex plus random perturbation as the initial guess. This demonstrates that the exact λ is an attractive fixed point of the vertex sub-cycle, but it does not demonstrate convergence of the full cycle involving the Dyson updates (22a), (22b), (24), the polarization updates (20), (21), and the Hedin equation (23). Feedback among these updates could destabilize the scheme, especially since the paper itself notes that the non-interacting initial guess does not converge to the exact result for the AIM parameters. As a consequence, the strongest empirical support for \"self-consistent reconstruction\" is limited to the atomic limit, while the more relevant AIM case tests only a subsystem of the proposed equations.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper derives a set of self-consistent equations for the Hedin three-leg vertex, the polarization, and the electronic self-energy, based on the single-boson-exchange (SBE) decomposition of the four-point vertex introduced in Ref. [49]. The closed cycle consists of the parquet-like vertex equations (15) and (18), the polarization relations (20) and (21), the Dyson-type updates for the screened interaction (22a), (22b), and the fermionic propagator (24), and the Hedin equation (23). The authors argue that this avoids the storage of four-point vertices and the inversion of Bethe-Salpeter equations, which are the bottlenecks of the traditional parquet formalism. They validate the scheme in the atomic limit, where a full self-consistent run from a non-interacting guess converges at U/T=2 and the exact strong-coupling vertex is reproduced at U/T=20 using an annealing initial guess, and for an Anderson impurity model, where the exact Hedin vertex is shown to be an attractive fixed point of the vertex equations when the self-energies are held fixed. The outlook discusses the SBE approximation (phi^firr=0) and an SBE-inspired DGammaA approximation with a local fully irreducible vertex.","tokens_in":19656,"tokens_out":10424,"duration_ms":108775,"significance":"If the advertised claims hold, the method is a numerically attractive alternative to parquet: it treats charge, spin, and particle-particle fluctuations on equal footing, preserves crossing symmetry and the consistency of the potential energy with the two-particle level (Appendices D and E), and has a computational cost scaling that is lower than that of standard parquet by one power of the momentum-frequency mesh. The algebraic derivation from the SBE decomposition is explicit and the paper provides useful exact identities and symmetry discussions. The strongest hardware-supported evidence is the full-cycle convergence of the atomic limit at weak coupling and the fixed-point stability of the exact vertex in the AIM. However, the numerical demonstration of the full self-consistent cycle is incomplete for the AIM, and several claims in the abstract and conclusions go beyond what is actually computed.","major_comments":[{"comment":"The AIM validation does not exercise the full calculation cycle claimed in the abstract. The text states that the exact self-energies Sigma and pi are provided as input \"which we do not update in the cycle,\" and only Eqs. (15) and (18) are iterated for the Hedin vertex. This demonstrates that the exact lambda is an attractive fixed point of the vertex sub-cycle, but it leaves untested the feedback among the Dyson updates (22a), (22b), (24), the polarization updates (20), (21), and the Hedin equation (23). Since the abstract and the conclusions describe the scheme as fully self-consistent and claim convergence of \"the calculation scheme starting from a fully irreducible vertex\" for the AIM, the numerical support is incomplete. The authors should either implement the full cycle for the AIM (for example, starting from a warm guess) or rephrase the claims to indicate that only the Hedin-vertex sub-cycle is tested.","section":"Sec. IV B (p. 8)"},{"comment":"Convergence from a generic initial guess is demonstrated only in the weak-coupling atomic limit. At U/T=20 in the atomic limit, the calculation is initialized by annealing from the exact solution for similar parameters, and the authors state that for U/T > 2 the non-interacting guess is \"quite inconvenient\" and does not converge. In the AIM, the successful runs start from a random perturbation of the exact vertex (Eq. (25)), and the text reports that the non-interacting limit is a \"poor initial guess\" from which the parquet does not converge to the exact result. Consequently, what is demonstrated is local stability of an exact fixed point, not convergence of the proposed scheme from a physically motivated cold start. The paper should either provide at least one successful cold-start convergence test in the non-trivial parameter regime or explicitly restrict the convergence claim to local stability.","section":"Sec. IV A and Sec. IV B (pp. 6-8)"},{"comment":"The SBE approximation (26), which is the only fully ab initio option with phi^firr=0 and thus the only option that completely eliminates the four-point input, is stated to converge in first tests of the atomic limit \"only for small enough values of the ratio U/T (results not shown).\" This unshown result undercuts the claim that this approximation is \"expected to yield a reasonable description of the lattice Hubbard model in the weak-coupling regime.\" The authors should either include the atomic-limit convergence tests for the SBE approximation or clearly label the statement as preliminary and unverified.","section":"Sec. V A (p. 9)"}],"minor_comments":[{"comment":"The text states that the calculations \"can be converged on a single core within a few minutes,\" but no timing data, stopping criterion, or iteration counts are reported; please specify the convergence criterion used for Figs. 6-9.","section":"Sec. IV A (p. 6)"},{"comment":"The report that the parquet does not converge from the non-interacting limit for the AIM is interesting but incomplete: the authors should describe what the iteration actually does in that case, namely whether it converges to a wrong fixed point, oscillates, or diverges.","section":"Sec. IV B (p. 8)"},{"comment":"The notation λ̄ and λ is used for right- and left-sided vertices, and the text says they are equal under time reversal; it would be helpful to state explicitly that all equations in the main text use this equality, to avoid confusion when Eq. (D1) is applied.","section":"Appendix D, Eq. (D1)"}],"recommendation":"major_revision","confidential_remarks":"The algebraic content and the conceptual proposal are sound and the paper is a natural continuation of Ref. [49]. The main revision needed is to align the numerical validation claims with what was actually computed: the AIM test freezes Sigma and pi, so the abstract's claim of demonstrating convergence of the full calculation scheme for the AIM is not supported. The authors should be asked to either run the full AIM cycle or explicitly reframe the claim as a test of the vertex sub-cycle. I would not reject on novelty grounds; the three-leg parquet construction is a useful contribution, and the scaling advantage over the standard parquet is plausible."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nHere is my read of Krien and Valli's parquet-like Hedin vertex paper.\n\nThe genuinely new thing is the closed set of three-leg equations. The SBE decomposition was already in their earlier PRB, but showing you can close the self-consistency cycle at the level of the Hedin vertex—avoiding four-point storage and Bethe-Salpeter inversion—is a real step forward. The derivation in Sec. III is explicit, the equations are written out, and the appendices give the crossing symmetry and exact relations. The complexity argument is plausible and matters.\n\nThe atomic-limit results are the strongest evidence. Starting from the non-interacting guess at U/T=2, the full cycle converges to the exact Hedin vertex; at U/T=20, with an annealed initial guess, it reproduces the exact vertex. That is a genuine demonstration that the closed set of equations works when the fully irreducible vertex is exact. The authors are careful about the decoupling ambiguity and about the Ward identity caveats.\n\nThe soft spot is exactly what the stress-test note flags. The AIM test in Sec. IV B feeds in the exact fully irreducible vertex plus the exact self-energies, and iterates only the Hedin-vertex equations. That shows the vertex sub-cycle has an attractive fixed point, but it does not show the full self-consistent cycle—with Dyson updates for g and w, polarization updates, and the Hedin equation—converges. The abstract's phrase \"convergence of the calculation scheme\" for the AIM overstates what is shown. The full cycle is only demonstrated in the atomic limit.\n\nAlso, the non-interacting guess does not converge for the AIM at these parameters; the authors say this, and they lean on random perturbations around the exact vertex. For a proof of concept, that is acceptable, but it limits how strongly one can claim the scheme is robust.\n\nThe lattice extension is honestly presented as an outlook, with the SBE-DGA approximation and its assumptions laid out. The question whether nonlocal correlations can be captured by U-reducible diagrams alone is open; the authors do not dodge it.\n\nOverall: the formal contribution is solid and worth a serious referee. The main demand on the authors should be to either soften the AIM claim or demonstrate a full-cycle convergence for an impurity model. I would cite this for the equations, and I'd take it to reading group.\n\nRecommendation: send to peer review.","headline":"The closed three-leg SBE parquet scheme is a real step forward, but the AIM convergence claim only holds for the vertex sub-cycle, not the full self-consistent cycle.","tokens_in":20198,"tokens_out":2535,"would_cite":true,"duration_ms":25861,"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 shows that the single-boson-exchange decomposition yields a closed set of parquet-like equations for the Hedin three-leg vertex, the polarization, and the self-energy, avoiding four-point vertex storage and Bethe-Salpeter…","keywords":["single-boson exchange","Hedin three-leg vertex","parquet equations","Anderson impurity model","Maki-Thompson diagrams","vertex divergences","dynamical vertex approximation","self-consistent many-body theory"],"falsifier":"Compute the spin susceptibility of the two-dimensional Hubbard model at half-filling with the SBE-DΓA input $\\phi^{\\mathrm{firr}}$ taken from the corresponding impurity model, and compare it with a controlled Monte Carlo simulation of the same lattice near the Néel temperature; if the generated nonlocal single-boson-exchange diagrams do not reproduce the growth of the antiferromagnetic susceptibility, the locality assumption fails. A cheaper consistency test is already reported in the paper: the approximation $\\phi^{\\mathrm{firr}} = 0$ converges only for small $U/T$ in the atomic limit, so mapping where it loses the exact solution delimits the method's reach.","tokens_in":19232,"feed_emoji":"⚛️","tokens_out":11688,"duration_ms":103310,"temperature":0.7,"pith_summary":"The paper's goal is to make the unbiased treatment of competing electronic fluctuations computationally affordable. It shows that the single-boson-exchange (SBE) decomposition of the vertex function can be rearranged into a closed set of parquet-like equations for the Hedin three-leg vertex, the polarization, and the self-energy, so that the full calculation never stores a four-point vertex or inverts a Bethe-Salpeter equation. The scheme sums Maki-Thompson diagrams self-consistently, sweeping in the mutual screening of charge, spin, and pairing fluctuations. A sympathetic reader would care because the result is a path to vertex corrections on real lattices where the full parquet equations are currently prohibitive. Convergence to the exact solution is demonstrated on the atomic limit and the Anderson impurity model.","feed_headline":"Three-leg parquet equations bypass costly four-point vertex","feed_subtitle":"Closed self-consistent scheme for vertex, polarization, and self-energy converges on impurity models","key_machinery":"The central object is the SBE decomposition, $f^{\\alpha} = \\phi^{\\mathrm{firr},\\alpha} + \\nabla^{\\mathrm{ph}} + \\nabla^{\\bar{\\mathrm{ph}}} + \\nabla^{\\mathrm{pp}} - 2U^{\\alpha}$, which rewrites the four-point vertex using only fully irreducible diagrams plus products of a Hedin three-leg vertex $\\lambda$ and a screened interaction $w$. Inserting this decomposition into the channel relations (12a)–(12c) and eliminating the channel-irreducible vertices $\\phi$ via the three-leg definitions (14) and (17) yields the parquet-like integral equations (15) and (18). These, together with the polarization updates (20)–(21), the Dyson equations (22a)–(22b) and (24), and the Hedin equation (23), form the self-consistent cycle in Fig. 5 whose core operation is the summation of Maki-Thompson diagrams.","core_discovery":"The central claim is that the SBE decomposition makes the parquet problem three-legged: a fully $U$-irreducible vertex $\\phi^{\\mathrm{firr}}$ fixes, through the coupled integral equations (15) and (18), the Hedin vertices $\\lambda^{\\mathrm{ch}}$, $\\lambda^{\\mathrm{sp}}$, and $\\lambda^{\\mathrm{s}}$; these in turn fix the polarization $\\pi$ and the screened interaction $w$ via Dyson equations, and the fermionic self-energy $\\Sigma$ via the Hedin equation. Iterating this cycle renormalizes propagators and vertex on equal footing, reconstructing the full vertex function as a sum of single-boson-exchange diagrams. On the atomic limit and the Anderson impurity model the iteration converges to the exact Hedin vertex, and for the impurity model the exact vertex is an attractive fixed point of the flow.","pith_inferences":["If the locality of $\\phi^{\\mathrm{firr}}$ holds in practice, the SBE-DΓA should reproduce the momentum-dependent vertex corrections needed for optical conductivity, a quantity that ladder approximations leave unchanged; this is a testable prediction for the two-dimensional Hubbard model.","The method's freedom from two-particle-self-energy divergences makes it a natural seed for functional renormalization group flows, where divergent intermediate vertices are a known obstacle.","Feeding $\\phi^{\\mathrm{firr}}$ from a small cluster rather than a single impurity into the lattice parquet equations would directly quantify how much nonlocal physics must come from the input vertex as opposed to the generated single-boson-exchange diagrams."],"forward_implications":["The two bottlenecks of the parquet formalism — storing four-point vertices and inverting Bethe-Salpeter equations — are removed; only three-leg objects are stored and iterated.","The computational cost per linear update scales as $(N_\\nu N_k)^2(N_\\omega N_q)$ instead of $(N_\\nu N_k)^3 N_\\omega N_q$, placing lattice calculations with a fine momentum grid within reach.","Because the SBE building blocks are connected to physical response functions, the scheme is free of the vertex divergences that complicate standard parquet solutions.","The crossing symmetry of the vertex can be enforced through a symmetry of the singlet Hedin vertex, ensuring thermodynamic consistency of the potential energy.","Using a local fully irreducible vertex from an auxiliary impurity model (SBE-DΓA) yields a concrete, approximation-only input for lattice Hubbard model calculations."],"supporting_citations":[{"why":"Supplies the single-boson-exchange decomposition of the four-point vertex on which the three-leg parquet equations are built.","marker":"[49]"},{"why":"Defines the Hedin three-leg vertex and the Hedin equation used to close the self-consistent cycle.","marker":"[46]"},{"why":"Introduces the dynamical vertex approximation whose local-input strategy motivates the SBE-DΓA lattice extension.","marker":"[21]"},{"why":"Provides the standard parquet solver whose cost scaling is the baseline the scheme improves upon.","marker":"[37]"},{"why":"Documents the vertex divergences that the SBE building blocks are designed to avoid.","marker":"[56]"},{"why":"Establishes the irreducible three-leg vertex and its connection to the two-particle self-energy used in Eqs. (6) and (16).","marker":"[63]"},{"why":"Defines the dynamical mean-field theory and the Anderson impurity model used as the numerical test bed.","marker":"[20]"},{"why":"Provides exact vertex functions of the atomic limit used to benchmark the parquet equations.","marker":"[57]"}],"fun_headline_variants":["Three-leg parquet scheme skips four-point vertex","Hedin vertex solved without four-point functions","Three-leg vertex scheme avoids Bethe-Salpeter inversion","Self-consistent three-leg vertex without four-point functions","Three-leg equations replace four-point vertex and Bethe-Salpeter"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The calculation depends on feeding in an accurate fully irreducible interaction vertex; if that input is approximated as purely local on a lattice, the method assumes the nonlocal physics is entirely rebuilt by the single-boson-exchange diagrams the equations generate — a step the paper itself flags as open in Sec. V.","fun_headline_variants_meta":{"raw":{"variants":["Three-leg parquet scheme skips four-point vertex","Hedin vertex solved without four-point functions","Three-leg vertex scheme avoids Bethe-Salpeter inversion","Self-consistent three-leg vertex without four-point functions","Three-leg equations replace four-point vertex and Bethe-Salpeter"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000772,"raw_usage":{"total_tokens":3404,"prompt_tokens":914,"completion_tokens":2490,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":530,"completion_tokens_details":{"reasoning_tokens":2412}},"tokens_in":530,"tokens_out":2490,"duration_ms":16958,"temperature":1.0,"reasoning_tokens":2412,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T04:38:52.434784+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute the spin susceptibility of the two-dimensional Hubbard model at half-filling with the SBE-DΓA input $\\phi^{\\mathrm{firr}}$ taken from the corresponding impurity model, and compare it with a controlled Monte Carlo simulation of the same lattice near the Néel temperature; if the generated nonlocal single-boson-exchange diagrams do not reproduce the growth of the antiferromagnetic susceptibility, the locality assumption fails. A cheaper consistency test is already reported in the paper: the approximation $\\phi^{\\mathrm{firr}} = 0$ converges only for small $U/T$ in the atomic limit, so mapping where it loses the exact solution delimits the method's reach.","supporting_citations":[{"cited_title":"Ayral and O","cited_arxiv_id":null,"evidence_quote":"Supplies the single-boson-exchange decomposition of the four-point vertex on which the three-leg parquet equations are built."},{"cited_title":"Pudleiner, P","cited_arxiv_id":null,"evidence_quote":"Defines the Hedin three-leg vertex and the Hedin equation used to close the self-consistent cycle."},{"cited_title":"Bergeron, V","cited_arxiv_id":null,"evidence_quote":"Introduces the dynamical vertex approximation whose local-input strategy motivates the SBE-DΓA lattice extension."},{"cited_title":"De Dominicis and P","cited_arxiv_id":null,"evidence_quote":"Provides the standard parquet solver whose cost scaling is the baseline the scheme improves upon."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Establishes the irreducible three-leg vertex and its connection to the two-particle self-energy used in Eqs. (6) and (16)."},{"cited_title":"Vilk and A.-M.S","cited_arxiv_id":null,"evidence_quote":"Defines the dynamical mean-field theory and the Anderson impurity model used as the numerical test bed."},{"cited_title":"Consistent partial bosonization of the extended Hubbard model","cited_arxiv_id":"1908.00536","evidence_quote":"Provides exact vertex functions of the atomic limit used to benchmark the parquet equations."}],"review_version":1}