{"id":"ffc29c96-5cf2-421d-9312-32474a06e9ed","arxiv_id":"2507.19265","paper_version":1,"verdict":"ACCEPT","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"low","formal_verification":"none","parameter_count":8,"one_line_summary":"Tuning the ratio of two couplings in a cubic-lattice clock model removes the leading and shrinks the subleading corrections to scaling, yielding eta = 0.03816(2) and 1/nu = 1.48872(5).","lead":"By simulating a clock model on a cubic lattice with an extra next-to-next-to-nearest-neighbor coupling, this paper obtains the most precise Monte Carlo estimates to date of two critical exponents of the three-dimensional XY universality class. The method also shows that the two main corrections to scaling can be made small, which matters because this universality class describes the superfluid transition in helium.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"FSS fits assume all corrections above ω_NR=2.02548 have exponent ≥3.4; the window (2.025,3.4) is untested by the ǫ4=3.6 check, so a term like L^{-(ω+ω_NR)}≈L^{-2.8} could bias η and y_t beyond the quoted errors despite good χ².","rationale":"The paper is careful, transparent about the ξ2nd/L bug, uses multiple Ansätze and Lmin stability checks, and the final exponents agree with the conformal bootstrap. The one place where the small quoted errors could be threatened is the assumed sparseness of the correction spectrum below exponent 3.4; this is exactly the reader's weakest assumption. I do not regard this as a demonstrated flaw, so the ACCEPT verdict stands; a targeted refit with an extra L^{-2.8} term would settle the residual risk without requiring new simulations. The absence of released code and raw data is a reproducibility limitation, not a correctness argument, and the paper is explicit about it. No ad hominem is intended; the critique targets the truncation assumption, not the author's conduct.","tokens_in":26980,"tokens_out":14768,"duration_ms":154621,"concrete_test":"Re-run the joint fit (29) and the final η/yt fits (40),(43) with an additional free-amplitude correction term L^{-(ω+ω_NR)}≈L^{-2.814}, and also with a free fourth exponent set to 2.4 and 3.0, for Lmin=8–20. If the central estimates of K3*, η, y_t move by no more than half the quoted error at Lmin≥13, the completeness assumption is confirmed; if they move further, the quoted error bars are understated.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central numerical claims (η=0.03816(2), y_t=1.48872(5)) rest on FSS Ansätze (29), (38), (40), (42), (43) that truncate the correction spectrum after ω=0.789, 2−η=1.9618, and ω_NR=2.02548. The paper's additional jmax=4 fits (AA, AB) add ǫ4=3.6, explicitly assuming all unmodeled operators have correction exponents ≳3.4 (Sec. V.E). This leaves the window (2.025,3.4) untested. In particular, the term L^{-(ω+ω_NR)}≈L^{-2.814}, generated by the product of the leading and lattice-anisotropy scaling fields, is explicitly ignored in Sec. V.C. Its amplitude is not automatically zero at the improved point: the anisotropy amplitude c is only reduced by about a factor of 9 at r*_K, not eliminated, so if b is small but nonzero in the joint fit at K3=0.04–0.05, the product b·c need not be negligible at the 1e-5 level of these high-statistics data. A small-amplitude omitted term at exponent ≈2.4–3.0 can shift K3*, η, and y_t while leaving χ²/DOF≈1; the paper itself states (Sec. V.D) that an acceptable p-value says little about such systematic errors. The paper's Lmin stability and multiple observable combinations mitigate this risk, but they do not eliminate the specific blind spot below 3.4.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the (q+1)-state clock model with nearest-neighbor and third-nearest-neighbor couplings on the simple cubic lattice, with the aim of removing leading and subleading corrections to scaling in the three-dimensional XY universality class. Using very high statistics Monte Carlo simulations, the author tunes the ratio rK=K3/K1 to eliminate the leading correction and finds that at a nearby value the anisotropy correction is reduced by about a factor of 9. Finite-size scaling of dimensionless observables (Za/Zp, U4, U6, ξ2nd/L) and of the magnetic susceptibility yields η=0.03816(2) and yt=1/ν=1.48872(5), consistent with conformal bootstrap results and improving on the author's previous Monte Carlo estimates.","tokens_in":27392,"tokens_out":10338,"duration_ms":100803,"significance":"If the quoted error bars are robust, this is the most accurate Monte Carlo determination of the 3D XY critical exponents to date, with precision comparable to the conformal bootstrap. The work also extends the improved-model program by showing that in a two-parameter family one can eliminate the leading correction while substantially reducing the subleading correction, even though both cannot be made to vanish simultaneously. The numerical effort (about 60 core-years at the optimal point) and the cross-checks among multiple observables, four FSS Ansätze, and Lmin stability are exemplary. The paper is also candid about residual systematic uncertainties, explicitly warning in Sec. V.D that an acceptable χ² says little about corrections not included in the Ansatz.","major_comments":[{"comment":"The FSS model space leaves the correction-exponent window between about 2.025 and 3.6 untested. The jmax=4 fits with ǫ4=3.6 (Sec. V.C) do not include terms such as L^{-(ω+ω_NR)} ≈ L^{-2.814}, which are generated by products of the leading and lattice-anisotropy scaling fields and are explicitly ignored in Sec. V.C because the amplitude b_i is asserted to be small. No quantitative bound on this amplitude is provided. Since the final error bars in Eqs. (41) and (44) are estimated from the spread over fits that all omit this term, a non-negligible amplitude in the window (2.025, 3.6) would bias η and y_t beyond the quoted errors. I request a fit that includes a correction at exponent ω+ω_NR, or an equivalent direct estimate of its amplitude from the data, together with a discussion of the resulting shift in the central values.","section":"V.C and V.E; Eqs. (29), (40), (43)"},{"comment":"The statistical errors of the Taylor coefficients used to extrapolate observables from K1,s to K1,c are ignored in the fits (Sec. V.C), and the improved-observable exponents p are fixed at values obtained from Eq. (A4) without propagating their errors (Sec. V.D and Appendix A). Given that the data have statistical errors at the 10^{-5} level, these ignored uncertainties could be comparable to the quoted error bars. Please demonstrate that the final estimates in Eqs. (41) and (44) and their error bars are stable when these uncertainties are propagated, or provide a quantitative statement of why the effect is negligible.","section":"V.C, V.D, Appendix A"}],"minor_comments":[{"comment":"The term \"unisotropy\" should be replaced by \"anisotropy\" (abstract, Sec. IV, Fig. 1, and elsewhere).","section":"Throughout"},{"comment":"The sentence \"In the case of AA and BB, ǫ4=3.6 is used\" appears to contain a typo; it should read \"AA and AB.\"","section":"V.C"},{"comment":"The list of values for Yq uses ellipses in a way that is difficult to follow; consider presenting the values in a table or in a more explicit sequence.","section":"III.A, Eq. (17)"},{"comment":"The notation for the subleading anisotropy exponent, ω'_NR, is defined implicitly in Eq. (27); please define it explicitly at first use.","section":"IV.C, Eq. (27) and surrounding text"},{"comment":"In the bullet list describing the corrections for the various slopes, the term \"L^{-yt-ω}\" would be clearer as \"L^{-(yt+ω)}\".","section":"V.E"},{"comment":"The statement \"For a discussion of χ/D.O.F. and the corresponding p-value see appendix A of Ref. [50]\" sends the reader to a different paper; a brief sentence defining the acceptance criterion here would improve the presentation.","section":"V.C"}],"recommendation":"major_revision","confidential_remarks":"The paper is a solo-author work with extensive self-citation, which is appropriate given the author's prior development of the improved-model program. The central numerical claims are likely correct, but the untested correction-exponent window is a specific concern about the quoted error bars that should be addressed before publication. The paper is well within the scope of the journal and the presentation is generally clear."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"What you should know upfront: this is the most precise Monte Carlo determination of the 3D XY critical exponents to date, and the two-improvement-point story is real. The stress-test worry about the untested correction window (2.025, 3.4) is legitimate in principle, but I don't think it bites at the quoted precision.\n\nWhat's new: the 3N q-state clock model at D→∞, the demonstration that the leading-improvement point r*_K = 0.1119(5) and the isotropy-restoring point r_iso = 0.140(8) are distinct and nearby, and the reduction of lattice anisotropy by about an order of magnitude at r*_K. The exponents η = 0.03816(2) and y_t = 1.48872(5) are consistent with the conformal bootstrap (η = 0.038176(44), y_t = 1.48864(22)) and with the author's earlier work, and they are two to four times more precise than any previous Monte Carlo result. The credibility comes from high statistics (about 60 core-years at the main parameter point), four FSS Ansätze, several independent observables, explicit checks of the ω_NR sensitivity, and a consistency check at a completely different parameter point. That is a lot of independent scaffolding.\n\nThe paper is also unusually transparent. The ξ2nd/L bug is disclosed and handled by dropping that observable from the joint fits. The Sec. V.D warning — that an acceptable χ²/DOF says little about unmodeled corrections — is demonstrated with the K3 = 0.05 fit, which gives η = 0.03863(5) with acceptable p-values and misses the final estimate by about nine times its quoted error. The final error bars are set by the spread across Ansätze, which is the right way to handle systematics.\n\nSoft spots, in proportion. The stress-test note is right that the jmax=4 fits (ǫ4 = 3.6) do not test the window (2.025, 3.4), and that L^{−(ω+ω_NR)} ≈ L^{−2.81} sits in that window. But at the improved point the amplitude of that term is the product b·c, and b is small there by construction. More importantly, the L_min stability plots over L = 8 to 24 at these statistics would show visible drift if an unmodeled correction with exponent near 2.8 and sizeable amplitude were present. I read those plots as the empirical answer to the concern. The actual weaknesses are minor: no code or raw data (for a paper this precise, I would want the data), Taylor coefficient errors ignored in the K1 extrapolation, and the p exponents fitted from a separate dataset with the errors of the ratio quantities neglected. None of these should move the central conclusions.\n\nWho it's for: specialists in 3D O(N) critical phenomena, precision lattice methods, and bootstrap comparisons, plus anyone using ν or η for the helium λ-transition. It deserves a serious referee. I would send it out, ask for data availability and an explicit test of an L^{−(ω+ω_NR)} term, and expect minor revision. It should be published.","headline":"The most precise Monte Carlo exponents for the 3D XY class to date, with a clean two-improvement-point story; the untested correction window (2.0-3.4) is a legitimate caveat but unlikely to move the last digit.","tokens_in":27963,"tokens_out":14173,"would_cite":true,"duration_ms":127742,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["05.50.+q","64.60.Fr","75.10.Hk"],"model":"deepseek-v4-flash","headline":"By tuning the ratio of two couplings in a clock model on the cubic lattice, the leading correction to scaling in the three-dimensional XY universality class can be eliminated, yielding η=0.03816(2) and ν=0.671718(23).","keywords":["three-dimensional XY universality class","clock model","finite-size scaling","corrections to scaling","improved lattice models","critical exponents","Monte Carlo simulation","lattice anisotropy"],"falsifier":"Compute the same dimensionless quantities at a point where the leading and subleading amplitudes are large, for example at $K_3=0$ or at $r_K\\simeq0.15$, and refit the data including an additional correction with exponent near $3.0$ or $3.4$; if the estimates of $\\eta$ or $y_t$ shift by more than the quoted uncertainty, the assumed correction spectrum is incomplete. Alternatively, extend the $K_3=0.0415$ dataset to $L=300$–$400$ and check whether fits with $L_{\\min}\\ge 24$ keep $\\eta$ inside $0.03816(2)$ and $y_t$ inside $1.48872(5)$.","tokens_in":26758,"feed_emoji":"🎯","tokens_out":16882,"duration_ms":148064,"temperature":0.7,"pith_summary":"This paper asks whether the 'improved model' strategy, tuning a Hamiltonian parameter so that the leading correction to scaling vanishes, can be extended to also control the subleading correction in the three-dimensional XY universality class. The author simulates a $q$-state clock model on the simple cubic lattice with an extra next-to-next-to-nearest-neighbor coupling $K_3$ and shows that at $r^*_K=(K_3/K_1)^*=0.1119(5)$ the amplitude of the leading correction vanishes, while the spatial anisotropy that causes subleading corrections is reduced by about a factor of nine. A nearby point $r^{\\rm iso}_K=0.140(8)$ makes the anisotropy vanish instead, so the two special points nearly coincide. High-statistics Monte Carlo data at the improved point give $\\eta=0.03816(2)$ and $y_t=1/\\nu=1.48872(5)$, i.e., $\\nu=0.671718(23)$, consistent with conformal-bootstrap results and several times more precise than the author's previous lattice estimates. The result gives a benchmark for the XY universality class and demonstrates how two correction amplitudes can be controlled in one lattice model.","feed_headline":"Tuned clock model pins down 3D XY exponents","feed_subtitle":"Setting K3/K1=0.1119 removes the leading correction; Monte Carlo yields η=0.03816(2) and ν=0.671718(23).","key_machinery":"The carrying object is the pair of amplitudes $b(D,r_K)$ and $c(D,r_K)$ of the leading irrelevant operator (correction exponent $\\omega\\approx0.789$) and of the lattice-anisotropy operator (correction exponent $\\omega_{NR}=2.02548$) in the finite-size expansion of dimensionless quantities such as the Binder cumulant $U_4$ and the partition-function ratio $Z_a/Z_p$. The argument moves along the critical line of the two-parameter model, using $K_3$ and the dynamic-dilution parameter $D$ as handles, until $b(D,r_K)=0$; independently, the anisotropy amplitude $c(D,r_K)=0$ is located in the high-temperature phase from the direction dependence of the exponential correlation length, where $r_i-1\\simeq a_i\\xi^{-\\omega_{NR}}$ for the $(1,1,0)$ and $(1,1,1)$ directions isolates the subleading operator. Improved observables such as $\\chi_{\\rm imp}=\\chi U_4^p$ cancel the leading correction and stabilize the exponent estimates.","core_discovery":"The paper's central claim is that a two-parameter family of $(q+1)$-state clock models on the simple cubic lattice contains a point on the critical line where the leading correction to scaling has zero amplitude, and that the same family contains a nearby point where the leading spatial-anisotropy correction vanishes. For the pure clock model reached in the limit $D\\to\\infty$, the ratio $r_K=K_3/K_1$ can be tuned to $r^*_K=0.1119(5)$ to eliminate the leading correction; at the nearby isotropy point $r^{\\rm iso}_K=0.140(8)$ the leading contribution to the spatial anisotropy vanishes, and at $r^*_K$ the residual anisotropy is smaller by a factor of about nine than at $K_3=0$. A finite-size scaling analysis of high-statistics data at $K_3=0.0415$, keeping the three correction terms with exponents $\\omega=0.789$, $2-\\eta$, and $\\omega_{NR}=2.02548$, yields $\\eta=0.03816(2)$, $y_t=1.48872(5)$ ($\\nu=0.671718(23)$), along with fixed-point values $(Z_a/Z_p)^*=0.320380(8)$ and $U_4^*=1.242934(10)$, and the leading-correction zero at $D^*=1.063(6)$ for $K_3=0$. The paper states explicitly that, unlike the analogous Ising case, no point where both correction amplitudes vanish exactly was found in this model; the improved point only makes the subleading anisotropy small.","pith_inferences":["A consequence left implicit in the paper is that the near-coincidence of $r^*_K$ and $r^{\\rm iso}_K$ may be a general feature of $O(N)$ universality classes on the cubic lattice, since the leading and anisotropy operators have comparable dimensions; if so, a single tuned ratio could nearly eliminate both corrections for other $N$ as well.","If the same two-coupling construction is applied to $O(N)$ models with $N=3,4$ or to the diluted Ising model, where the leading correction exponent is small, the dominant systematic error could shift from the leading to the next operator, improving precision beyond what a single tuned parameter allows.","A testable extension is to run the same tuned ratio on a body-centered or face-centered cubic lattice; a change of lattice geometry would alter the anisotropy amplitude while leaving the leading-correction amplitude essentially unchanged, separating the two mechanisms in a controlled way."],"forward_implications":["Future lattice simulations of the 3D XY universality class can work at $K_3=0.0415$ (with $q\\ge 24$) and reach exponent errors of a few $10^{-5}$ in $\\eta$ and $y_t$ with linear lattice sizes up to $L=200$.","Because the leading-correction and isotropy points are close, simulating at $r^*_K$ suppresses the anisotropy-induced systematic error by roughly a factor of nine, equivalent to reducing the required linear lattice size by almost a factor of three and the volume by about a factor of 27.","The fixed-point values $(Z_a/Z_p)^*=0.320380(8)$ and $U_4^*=1.242934(10)$ give model-independent numbers that other methods and experiments in the same universality class can be tested against.","The new $\\nu=0.671718(23)$ sharpens the known tension with the microgravity $^4$He measurement $\\nu=0.6709(1)$, making a reanalysis or a new measurement of the $\\lambda$-transition desirable."],"supporting_citations":[{"why":"Supplies the baseline q=8, K3=0 data and the previous exponent estimates that this work extends and improves.","marker":"[13]"},{"why":"Introduces the two-coupling lattice model and the isotropy-restoration strategy adapted here to the XY universality class.","marker":"[8]"},{"why":"Provides the conformal-bootstrap values of eta, y_t and omega used as benchmarks and as inputs for the correction spectrum.","marker":"[14]"},{"why":"Supplies the subleading correction exponent omega_NR=2.02548(41) that enters the FSS Ansatze.","marker":"[37]"},{"why":"Reviews the spectrum of O(N) CFT correction exponents and justifies the assumption that only a few operators dominate.","marker":"[26]"},{"why":"Introduces the improved-observable construction chi_imp = chi U4^p and gives earlier high-precision estimates used for comparison.","marker":"[21]"},{"why":"Defines the model class and the finite-size scaling framework for dimensionless quantities and slopes.","marker":"[25]"},{"why":"Provides the single-cluster algorithm used for the high-temperature and critical simulations.","marker":"[7]"},{"why":"Provides the wall-cluster algorithm used to compute the partition-function ratio Z_a/Z_p.","marker":"[44]"}],"fun_headline_variants":["Clock model tuning suppresses scaling corrections for XY exponents","Precise 3D XY exponents from clock model with suppressed corrections","Improved clock model: leading and subleading corrections small","Two-parameter clock model removes leading correction, sharpens XY exponents"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The analysis assumes that finite-size corrections are completely described by the three known sources with exponents $\\omega\\approx0.789$, $2-\\eta\\approx1.962$, and $\\omega_{NR}=2.02548$, and that all further corrections are negligible for lattice sizes $L=8$ to $13$ and above; if an unmodelled correction with exponent below about $3.4$ contributes non-negligibly, the quoted exponents and errors could be biased despite good fit quality.","fun_headline_variants_meta":{"raw":{"variants":["Clock model tuning suppresses scaling corrections for XY exponents","Precise 3D XY exponents from clock model with suppressed corrections","Improved clock model: leading and subleading corrections small","Two-parameter clock model removes leading correction, sharpens XY exponents"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001318,"raw_usage":{"total_tokens":5455,"prompt_tokens":1121,"completion_tokens":4334,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":737,"completion_tokens_details":{"reasoning_tokens":4265}},"tokens_in":737,"tokens_out":4334,"duration_ms":32558,"temperature":1.0,"reasoning_tokens":4265,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T17:57:13.767671+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute the same dimensionless quantities at a point where the leading and subleading amplitudes are large, for example at $K_3=0$ or at $r_K\\simeq0.15$, and refit the data including an additional correction with exponent near $3.0$ or $3.4$; if the estimates of $\\eta$ or $y_t$ shift by more than the quoted uncertainty, the assumed correction spectrum is incomplete. Alternatively, extend the $K_3=0.0415$ dataset to $L=300$–$400$ and check whether fits with $L_{\\min}\\ge 24$ keep $\\eta$ inside $0.03816(2)$ and $y_t$ inside $1.48872(5)$.","supporting_citations":[{"cited_title":"A Monte Carlo study of the three-dimensional XY universality class:Universal amplitude ratios","cited_arxiv_id":"0810.2716","evidence_quote":"Provides the wall-cluster algorithm used to compute the partition-function ratio Z_a/Z_p."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Introduces the two-coupling lattice model and the isotropy-restoration strategy adapted here to the XY universality class."},{"cited_title":"For a more comprehensive comparison with the literature see Sec","cited_arxiv_id":null,"evidence_quote":"Provides the conformal-bootstrap values of eta, y_t and omega used as benchmarks and as inputs for the correction spectrum."},{"cited_title":"For example for l = 8 one gets ωl=8 ≈ 6.03","cited_arxiv_id":null,"evidence_quote":"Supplies the subleading correction exponent omega_NR=2.02548(41) that enters the FSS Ansatze."},{"cited_title":"Teitel, Finite-size scaling study of the three-dimensional classi cal XY 38 model, Phys","cited_arxiv_id":null,"evidence_quote":"Reviews the spectrum of O(N) CFT correction exponents and justifies the assumption that only a few operators dominate."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Introduces the improved-observable construction chi_imp = chi U4^p and gives earlier high-precision estimates used for comparison."}],"review_version":1}