{"id":"a6426b1e-0d8c-4cdc-b8b4-72c1428f8da6","arxiv_id":"2506.19770","paper_version":1,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":1,"one_line_summary":"At fixed lattice spacing and volume with unphysical quark masses, direct QCD+QED simulation and the RM123 perturbative expansion give consistent U-spin window contributions to a_mu^HVP, with the direct simulation about three times more precise.","lead":"This lattice study compares two ways of including electromagnetic corrections in calculations of the muon magnetic anomaly: simulating QCD plus QED directly, or expanding around isospin-symmetric QCD using the RM123 method. Both approaches agree on the U-spin window contribution, but the direct simulation has roughly three times smaller uncertainty at fixed statistics.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"No significant objection identified: the precision comparison is well-supported at fixed spacing, volume, and sample count; the RM123 error is shown to be gauge-noise dominated and the fixed-bare-parameter check controls scheme dependence.","rationale":"I read the paper as making a carefully scoped claim: for the U-spin intermediate-window contribution at one lattice spacing and volume, with 2000 gauge configurations, direct QCD+QED sampling yields a 2.5-3x smaller uncertainty than RM123 including full sea-quark effects. The evidence for this claim is strong: the RM123 error budget is decomposed and the dominant sea-sea variance is shown to saturate with stochastic sources, indicating the error is genuine gauge noise; the fixed-bare-parameter comparison (table 10) provides an independent check that both methods describe the same lattice theory, with all hadronic observables agreeing within errors; and the final LCP-matched results are consistent. The reader's identified weakest assumption (same renormalized theory, leading-order truncation, unimproved currents) is real but not load-bearing for the precision claim: even if the central values had a small O(a) or O(e^4) bias, the comparison of statistical uncertainties at fixed samples is unaffected, and the fixed-bare comparison controls the scheme dependence. The self-referential LCP targets in eq. (6.1) are acknowledged and do not propagate to a_mu. I found no internal inconsistency or unsupported assertion that would change the verdict. The one minor gap is the lack of a published variance-saturation check for the sea-valence diagrams, which is easily testable and does not invalidate the conclusion as presented.","tokens_in":32833,"tokens_out":16793,"duration_ms":177387,"concrete_test":"On the A400a00 ensemble, recompute the sea-valence contribution to delta a_mu (second row of diagrams in table 1) with N_eta=320 pseudofermion sources, doubling the 160 used in the paper, and compare the variance to the N_eta=160 result. If the sea-valence variance does not decrease, the RM123 total uncertainty of 21 is confirmed as gauge-noise dominated and the stated advantage of the non-perturbative method at fixed sample count is fair. If the variance decreases by more than about 10%, part of the claimed 2.5-3x advantage is attributable to stochastic noise that could be removed by more pseudofermion sources, and the comparison should be re-evaluated at the gauge-noise limit.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim is a precision comparison at fixed lattice spacing, volume, and number of gauge configurations. The RM123 uncertainty of 21 (vs 7 non-perturbatively) is attributed to sea-quark gauge noise, and figure 6 demonstrates saturation of the sea-sea variance with pseudofermion sources. The independent fixed-bare-parameter comparison (table 10) shows agreement between both methods for phi_i, t0, and a_mu, which supports the claim that both evaluate the same renormalized theory without relying on the self-referential LCP targets of eq. (6.1). The modified renormalization condition in Sec. 6 does not affect the target observable a_mu, and the fixed-bare comparison is the more fundamental consistency test. The use of the non-compact photon action for the RM123 diagrams versus the compact action in the non-perturbative simulation is a discretization difference of O(a^2) in the QED sector, which is negligible at the quoted precision (roughly 2e-5 relative on the ~1% QED correction). The leading-order truncation at O(e^2, Delta m_f) is consistent with the expansion parameters (e^2 ~ 0.0073 and Delta m ~ 0.005 in lattice units), and the agreement of central values within 0.4 sigma supports its adequacy. The only residual uncertainty I see is that stochastic saturation is demonstrated explicitly only for the sea-sea diagrams, while the sea-valence diagrams also contribute to the RM123 error; the paper states N_eta=160 reaches gauge noise for all estimators, but no saturation plot is shown for sea-valence. This is a minor, testable gap, not a demonstrated flaw.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper compares two ways of including electromagnetic effects in lattice hadronic vacuum polarization: a fully non-perturbative QCD+QED simulation with C* boundary conditions and the perturbative RM123 expansion around an isospin-symmetric QCD theory. The observable is the U-spin window contribution to the muon anomaly, a_mu^{U,w}, computed on two matched ensembles at a single lattice spacing and volume with N_f=1+2+1 dynamical quarks at unphysical masses. The comparison is done both at fixed bare parameters and at a fixed line of constant physics, with sea-quark effects included fully in both approaches. The final results are a_mu^{U,w} x 10^11 = 1094(21) for RM123 and 1085(7) for the non-perturbative simulation, so the non-perturbative result is about three times more precise at fixed sample count. The paper also reports consistency between the two methods for the scale-setting quantity t0 and the hadronic observables phi_i. The central claim is that, at fixed lattice spacing, volume, and Monte Carlo sample count, the direct QCD+QED simulation is advantageous over the RM123 expansion, with the RM123 uncertainty dominated by sea-quark gauge noise.","tokens_in":33166,"tokens_out":6812,"duration_ms":74905,"significance":"If the result holds, it provides useful guidance for future QCD+QED lattice calculations of the muon anomaly, where per-mille precision on HVP is needed. The paper is careful and transparent: it uses two current discretizations, two matching prescriptions, Gamma-method error estimation, AIC model averaging for fit-range systematics, and an explicit variance-saturation test for the stochastic sea-sea estimators. The fixed-bare-parameter comparison in Table 10 is a valuable independent consistency check that mitigates the self-referential nature of the modified LCP conditions in Eq. (6.1). The main limitations — a single lattice spacing and volume, unphysical quark masses, and the absence of a total-cost comparison — are stated clearly in the text. The paper is a solid groundwork study rather than a final physical prediction, and that scope is respected throughout.","major_comments":[],"minor_comments":[{"comment":"The text states that N_eta=160 pseudofermion sources reach gauge noise for all estimators, but Fig. 6 shows saturation only for the sea-sea contribution; please state explicitly whether a similar check was performed for the sea-valence diagrams, or explain why their stochastic error is subdominant.","section":"Sec. 5.2 / Fig. 6"},{"comment":"The modified LCP targets are set to the central values measured on A380a07, so the fixed-LCP comparison of the phi_i is partly tautological; I recommend highlighting more prominently that the fixed-bare-parameter comparison in Table 10 is the independent consistency test, and that the target observable a_mu is not affected by this choice.","section":"Sec. 6.1 / Eq. (6.1)"},{"comment":"The label 'isoQCD+RM123|eq' is undefined; please spell out 'electro-quenched' in the caption or define the abbreviation in the text.","section":"Table 10"},{"comment":"The column 'alpha' lists 0 for A400a00 and 0.007299 for A380a07, while the text says the latter is 'close to the physical value of alpha'; please clarify in the caption that this is the bare coupling, not the renormalized alpha_R.","section":"Table 2"},{"comment":"The statement that the expansion of Eq. (3.11) contains only odd powers of e would benefit from a brief explanation or a reference, since it is used to justify the absence of an e^2 term in the SW expansion.","section":"Sec. 4.1, after Eq. (4.12)"}],"recommendation":"accept","confidential_remarks":"This is a careful technical comparison from an established collaboration. The headline claim is properly scoped to fixed sample count, and the authors are transparent that total computational cost is not quantified. The only point I would keep in mind for the editor is that the abstract's word 'advantageous' could be overgeneralized by readers to mean advantage at fixed cost; the body of the paper is explicit about this, so no revision is needed."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"First direct comparison of the two approaches at matched volume and spacing with full sea-quark effects under C* boundary conditions. The central claim—that with fixed statistics the non-perturbative simulation beats the RM123 expansion in precision for the U-spin window—holds up. 1094(21) vs 1085(7) is consistent and the uncertainty ratio is about 2.5–3.\n\nWhat earns credit: the analysis is careful. Two current discretizations, two matching prescriptions, Gamma-method errors, AIC model averaging, and an explicit variance-saturation test showing the sea-sea contribution reaches gauge noise. The fixed-bare-parameter comparison in table 10 is the more fundamental consistency test, and it agrees for t0, phi_i, and a_mu; the modified LCP targets in eq. (6.1) are partially self-referential for phi_i, but that does not touch a_mu, and the authors are upfront about it. The paper also states its limitations cleanly: one lattice spacing, one volume, unphysical masses.\n\nSoft spots, in proportion: the conclusion is explicitly conditional on fixed statistics and does not include the full cost comparison—generation cost is factor ~2.5 for QCD+QED, but the RM123 measurement overhead is not fully quantified, and the paper says so. The leading-order truncation at O(e^2, delta m_f) is consistent with the expansion parameters and the 0.4-sigma agreement supports it, but QED-dependent O(a) effects from using unimproved currents and isoQCD c_sw are not controlled; the authors acknowledge this and it does not invalidate the fixed-spacing comparison. The variance saturation plot is shown only for sea-sea diagrams; the text says N_eta=160 reaches gauge noise for all estimators, but no saturation plot for sea-valence. That is a minor, testable gap, not a demonstrated flaw.\n\nWho this is for: lattice practitioners working on QCD+QED and the muon HVP; also useful for anyone weighing RM123-style expansions against direct simulations. It deserves a serious referee. I would send it to peer review and expect the referees to focus on the fixed-LCP matching choice and the missing sea-valence saturation plot, both addressable in revision.","headline":"First matched comparison of full QCD+QED vs RM123 with sea effects; the precision claim holds at fixed spacing, though the paper's own caveats should be kept.","tokens_in":33765,"tokens_out":1763,"would_cite":true,"duration_ms":16983,"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":"Direct QCD+QED simulation gives about three times better precision than the RM123 expansion for the U-spin window contribution to the muon anomaly, at fixed statistics.","keywords":["QCD+QED lattice simulation","C* boundary conditions","RM123 method","muon anomalous magnetic moment","hadronic vacuum polarization","isospin-breaking effects","U-spin window","sea-quark diagrams"],"falsifier":"Compute the same U-spin window with the RM123 method including the next-order terms ($O(e^4)$ and $O(\\Delta m_f^2)$) or on a second, finer lattice spacing with matched physics; if the RM123 prediction moves by more than its quoted uncertainty relative to the non-perturbative result, the truncation and $O(a)$ assumptions behind the comparison fail.","tokens_in":32633,"feed_emoji":"⚛️","tokens_out":10871,"duration_ms":99323,"temperature":0.7,"pith_summary":"This paper asks which way of putting electromagnetism into lattice QCD gives better predictions for the hadronic vacuum polarization (HVP) piece of the muon anomalous magnetic moment: simulate QCD+QED together from the start, or start from isospin-symmetric QCD and add electromagnetic and quark-mass corrections perturbatively, as in the RM123 method. Using the U-spin window contribution $a_\\mu^{U,w}$ — a windowed slice of the HVP at Euclidean separations 0.4–1 fm — as the test observable, at one lattice spacing, one volume, and the same number of Monte Carlo samples, the two methods agree within errors, but the direct simulation is more precise: $a_\\mu^{U,w}\\times10^{11}=1085(7)$ versus $1094(21)$. The RM123 error is dominated by statistical noise in the sea-quark isospin-breaking diagrams, which more stochastic estimators cannot reduce once gauge noise is reached. The paper concludes that, sample for sample, simulating the full QCD+QED distribution is the advantageous route, with configuration-generation cost and tuning effort left aside.","feed_headline":"Simulating QCD+QED directly beats RM123 expansion ~3x in precision","feed_subtitle":"U-spin window: 1085(7) vs 1094(21) x10^-11, a 2.5-3x uncertainty reduction at fixed statistics.","key_machinery":"The load-bearing setup is the C* (charge-conjugation-periodic) boundary-condition formulation of finite-volume QCD+QED, which permits a local, gauge-invariant treatment of the photon and allows the same action to be either sampled directly or expanded perturbatively around the isospin-symmetric point. The probe observable is the U-spin window contribution $a_\\mu^{U,w}$, built from the non-singlet current $V_\\mu = \\frac{1}{2}(\\bar{s}\\gamma_\\mu s - \\bar{d}\\gamma_\\mu d)$ and the intermediate Euclidean-time window $t\\in[0.4,1]\\,\\mathrm{fm}$, which is quark-line connected because the $d$ and $s$ quarks remain degenerate. On the RM123 side the argument is carried by the systematic expansion of the Dirac operator, the pfaffian (sea-quark effects), and the point-split current to order $O(e^2,\\Delta m_f)$, organized into valence-valence, sea-valence and sea-sea diagrams; the comparison is made both at fixed bare parameters and after propagating the uncertainty of matching the hadronic observables $\\phi_0,\\dots,\\phi_3$ that define the line of constant physics.","core_discovery":"The paper's central claim is a direct comparison: at a fixed lattice spacing and volume, and with a fixed number of gauge configurations and quark sources, the non-perturbative QCD+QED computation of $a_\\mu^{U,w}$ reaches a relative uncertainty of about 0.6–0.7%, while the RM123 expansion including all sea-quark effects reaches only about 1.9%. The two predictions are consistent, $1085(7)\\times10^{-11}$ and $1094(21)\\times10^{-11}$ after matching to a common line of constant physics, and the electro-quenched version of RM123, which neglects sea-quark effects, has essentially the same uncertainty as the direct simulation. This identifies the sea-quark diagrams as the entire source of RM123's extra error. The paper therefore claims it is advantageous to simulate the full QCD+QED distribution given a fixed number of samples, while noting that configuration-generation cost and tuning effort are not part of the comparison.","pith_inferences":["If the precision gap persists at physical quark masses and finer lattice spacings, direct QCD+QED simulation should become the default production method for the hadronic vacuum polarization, with RM123 serving as a cross-check rather than the primary estimator.","Because the U-spin current is protected from disconnected diagrams by a residual SU(2) symmetry, this comparison probably understates how hard the full electromagnetic current is for RM123, where singlet and disconnected contributions do not cancel.","The modified renormalization condition sets the $\\phi_i$ targets to the central values measured on the QCD+QED ensemble, so the fixed-LCP comparison is partially self-referential for those tuning observables; an independently determined line of constant physics could change the quoted non-perturbative tuning uncertainty, though not the direct $a_\\mu^{U,w}$ comparison at fixed bare parameters.","A natural next test is to rerun the same comparison at a smaller lattice spacing; if the RM123 correction shifts relative to the non-perturbative result, the neglected $O(a e^2)$ and higher-order terms are not small at $a\\approx0.054$ fm."],"forward_implications":["At fixed statistics, the non-perturbative QCD+QED result for the U-spin window has a 0.6–0.7% uncertainty, about 2.5–3 times smaller than the RM123 result's 1.9%.","The entire extra RM123 uncertainty comes from sea-quark (sea-valence and sea-sea) diagrams; the electro-quenched approximation is as precise as the direct simulation.","The sea-sea variance is gauge-noise dominated for more than about 100 pseudofermion sources, so reducing RM123's error forces more gauge configurations, and the paper expects this cost to grow with volume.","The two methods agree within errors, so RM123 remains a valid cross-check of direct QCD+QED at this lattice spacing and at a pion mass of about 400 MeV.","Matching to a fixed line of constant physics increases RM123's uncertainty from 1.6% to 1.9% but leaves the non-perturbative result essentially unchanged."],"supporting_citations":[{"why":"Defines the C* boundary-condition formulation of finite-volume QCD+QED used by both implementations.","marker":"[34]"},{"why":"Introduces the RM123 perturbative expansion of lattice observables in the electromagnetic coupling and quark-mass shifts.","marker":"[38]"},{"why":"Extends the RM123 method to leading isospin-breaking effects; the diagrammatic expansion used here follows it.","marker":"[39]"},{"why":"Supplies the two ensembles (isoQCD and QCD+QED), the renormalization scheme, and the earlier cost and precision estimates this comparison builds on.","marker":"[40]"},{"why":"Provides the gradient-flow scale value $\\sqrt{8t_0^*}=0.415$ fm used to convert lattice units to physical units.","marker":"[43]"},{"why":"Defines the intermediate-window weight function with $t_1=0.4$ fm, $t_2=1$ fm, and $\\Delta=0.15$ fm.","marker":"[16]"},{"why":"Provides the kernel $\\tilde K(t;m_\\mu)$ used in the window integral.","marker":"[50]"},{"why":"Analyzes the error scaling of sea-quark isospin-breaking effects, supporting the claim that RM123's sea-sea variance grows with volume.","marker":"[57]"}],"fun_headline_variants":["Direct QCD+QED simulation triples precision over RM123 method","Full QCD+QED yields 3x tighter muon g-2 window than RM123","QCD+QED direct: 1085(7) vs RM123's 1094(21) in U-spin window","At fixed samples, QCD+QED beats RM123 by ~3x in uncertainty","U-spin window: direct QCD+QED wins over RM123 with 3x edge"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the RM123 expansion, cut off at first order in $e^2$ and the quark-mass shifts, reproduces the same renormalized QCD+QED theory as the direct simulation at $a\\approx0.054$ fm, with the clover coefficient left at its isospin-symmetric value and the currents unimproved; if electromagnetic lattice artefacts or neglected higher-order terms are not small, the agreement between the two methods would be misleading.","fun_headline_variants_meta":{"raw":{"variants":["Direct QCD+QED simulation triples precision over RM123 method","Full QCD+QED yields 3x tighter muon g-2 window than RM123","QCD+QED direct: 1085(7) vs RM123's 1094(21) in U-spin window","At fixed samples, QCD+QED beats RM123 by ~3x in uncertainty","U-spin window: direct QCD+QED wins over RM123 with 3x edge"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00057,"raw_usage":{"total_tokens":2721,"prompt_tokens":994,"completion_tokens":1727,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":610,"completion_tokens_details":{"reasoning_tokens":1601}},"tokens_in":610,"tokens_out":1727,"duration_ms":11792,"temperature":1.0,"reasoning_tokens":1601,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T18:25:22.118738+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute the same U-spin window with the RM123 method including the next-order terms ($O(e^4)$ and $O(\\Delta m_f^2)$) or on a second, finer lattice spacing with matched physics; if the RM123 prediction moves by more than its quoted uncertainty relative to the non-perturbative result, the truncation and $O(a)$ assumptions behind the comparison fail.","supporting_citations":[],"review_version":2}