{"id":"c9156d4e-33b5-4b7f-9cd2-04c804625733","arxiv_id":"hep-lat/0607020","paper_version":2,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"A lattice action with extra negative-mass Wilson fermions produces a measured gap in the spectral density of the hermitian Wilson-Dirac operator near zero, removing the near-zero-mode bottleneck for overlap fermions.","lead":"Adding two flavors of unphysical heavy Wilson fermions to the lattice gauge action suppresses the near-zero modes of the hermitian Wilson-Dirac operator, opening a gap in its spectral density. The authors measure this gap directly in quenched simulations and argue it makes dynamical overlap-fermion simulations practical.","discovery_kind":"unclear","skeptic_critique":{"model":"selfhosted/deepseek-v4.1-flash","headline":"The paper's own fit gives ρ(λ_W) ~ λ_W^2, which vanishes at zero but is not a gap; the determinant suppression cannot produce a hard gap, so the claimed 'gap' needed for HJL locality is likely a finite-volume/statistics effect.","rationale":"The paper's practical contribution is solid: the extra Wilson fermions demonstrably suppress near-zero modes (Figs. 2, 3), reduce the sign-function cost, and the topology-conservation check (Section V) is a clean direct measurement. The β-shift calculation and the twisted-mass cancellation are well done. My concern is narrower but load-bearing for the abstract's theoretical claim: the determinant mechanism yields ρ ~ λ^2, not a hard gap, and the paper's own fit says so. The reader's weakest_assumption (HMC sampling exactness) is a real secondary issue — the spikes in ΔH and ⟨e^{-ΔH}⟩ values in Table I are acknowledged — but even with perfect sampling the ρ ~ λ^2 behavior would remain, so it is not the most load-bearing point. The reader's rationale does note 'the null result in the lowest spectral-density bin rests on limited statistics without a quoted upper bound,' which is adjacent to my concern, hence 'partial' agreement. I recommend CONDITIONAL: the method is sound for its practical purpose, but the 'gap → HJL locality' claim should either be backed by a demonstration that a true gap survives the infinite-volume limit, or softened to 'strong suppression / ρ(0)=0,' which is what the data actually show.","tokens_in":16779,"tokens_out":8716,"duration_ms":165940,"concrete_test":"Re-measure ρ(λ_W) in the lowest bins with ~10× the statistics and on a larger volume (e.g., 24^3×48), and fit ρ = c λ^2 near zero. If c > 0 (as the current fit suggests), compute the predicted count of modes below 0.02 as c λ^2 V N_configs and compare with the observed zero; consistency means no gap. Equivalently, check whether the number of modes in a fixed window |λ_W| < 0.02 grows linearly with V. A true gap would keep this count at zero as V increases; ρ ~ λ^2 would make it grow.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The central theoretical claim is that the extra Wilson fermions produce a gap in ρ(λ_W) near zero, satisfying the necessary condition for the Hernandez-Jansen-Lüscher locality proof. But the mechanism is a determinant weight det[H_W^2/(H_W^2+μ^2)], which for a mode at λ gives λ^2/(λ^2+μ^2) ≈ λ^2/μ^2 for λ << μ. The resulting density is therefore ρ(λ) ∝ λ^2 near zero — vanishing at λ=0 but with no gap: modes exist arbitrarily close to zero with density ~ λ^2. The paper's own Section IV fit ('consistent with ~ λ^2_W') confirms this, and it is in tension with the abstract's 'we find a gap.' The 'no events in the lowest bin' is consistent with a small λ^2 coefficient and finite statistics: with V ≈ 1.3×10^5 and ~100 configs, the expected count below 0.02 is O(1), so absence is not evidence of a gap. HJL locality requires ρ=0 in a neighborhood, not merely ρ(0)=0. Thus the necessary condition is not demonstrated; the observed 'gap' is plausibly a finite-volume artifact that fills in as V grows.","agreement_with_reader":"partial"},"referee_report":{"model":"selfhosted/deepseek-v4.1-flash","summary":"The manuscript proposes a lattice action in which two flavors of unphysical Wilson fermions with a large negative mass m0 = -1.6/a are added to the gauge action, together with a pair of twisted-mass pseudo-fermions, so that the near-zero modes of the hermitian Wilson-Dirac operator H_W(m0) are suppressed by the determinant factor det[H_W^2/(H_W^2+mu^2)]. Quenched HMC simulations are performed on a 16^3 x 32 lattice for three gauge actions (standard Wilson plaquette, a modified plaquette action with 1/epsilon = 2/3, and the Iwasaki RG action) and three values of the twisted mass (mu = 0, 0.2, 0.4). The low-lying spectrum of H_W is measured with the Lanczos algorithm, and the spectral density rho(lambda_W) is found to be nonzero at lambda_W = 0 for mu = 0 but consistent with a lambda_W^2 behavior for mu != 0. The paper also checks conservation of the topological charge via the index of the overlap-Dirac operator, discusses the topology dependence of the quenched pion mass, and computes the one-loop beta shift due to the extra fields.","tokens_in":17041,"tokens_out":10250,"duration_ms":182870,"significance":"The proposal is simple, practical, and directly addresses a real bottleneck in dynamical overlap-fermion simulations: the cost of the matrix sign function and the discontinuity encountered when a low-lying eigenvalue of H_W crosses zero. The numerical demonstration across three gauge actions, the twisted-mass trick to reduce the beta shift, and the explicit topology-conservation check are all valuable. The paper is a methods paper supported by simulation evidence, and its central practical claim (that near-zero modes are strongly suppressed and the sign-function cost is reduced) is well supported by the data shown.","major_comments":[{"comment":"The abstract (sentence s2) states that the paper finds 'a gap near lambda_W = 0', and the Introduction and Conclusion tie this gap to the Hernandez-Jansen-Luescher necessary condition for exponential locality. However, Section IV states that the measured density is 'consistent with ~ lambda_W^2' for mu != 0. A determinant weight lambda^2/(lambda^2+mu^2) produces rho(lambda) proportional to lambda^2 near the origin, i.e. a vanishing density at lambda = 0 but not a hard gap in any neighborhood. The HJL locality proof requires rho = 0 on an interval, not merely rho(0) = 0. The manuscript should either soften 'gap' throughout to 'vanishing spectral density at lambda_W = 0' and state explicitly which condition HJL actually requires, or provide a quantitative demonstration (a fit or an upper bound) that the density is zero on a finite interval at the quoted statistics. As written, the abstract's 'gap' is stronger than the data support, and this is load-bearing for the locality claim.","section":"Abstract and Section IV","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The paper acknowledges large spikes in Delta H and states 'In the following analysis we assume that the gauge configurations are properly sampled.' The values of <exp(-Delta H)> in Table I are mostly consistent with 1, but the assumption is stated rather than demonstrated. Please quantify the fraction of trajectories with anomalously large Delta H, the fraction rejected by the Metropolis test, and whether the measured spectral density in Fig. 3 is stable when the delta_tau = 0.005 test run (mentioned in Section III) is used for the measurement. Since the spectral density and the topology-conservation check are the central evidence, a demonstration that the results are insensitive to the sampling issue would materially strengthen the paper.","section":"Section III and Table I","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The null result in the lowest bin for mu != 0 rests on 30-100 configurations with a bin size of 0.0192. The statement 'there is no events in the lowest bin' should be converted into a quantitative upper bound on rho(0), or a fitted coefficient of the lambda_W^2 behavior, so that the mu = 0.2 and mu = 0.4 cases can be compared quantitatively with each other and with the mu = 0 case. As presented, the histograms in Fig. 3 do not allow the reader to judge whether the suppression is statistically significant or whether the difference between mu = 0.2 and 0.4 is meaningful. A fit analogous to the rho(0) fits quoted for the mu = 0 runs would resolve this.","section":"Section IV and Fig. 3","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The abstract states that 'the simulation including dynamical overlap fermions becomes feasible', but the present paper is entirely quenched and the dynamical overlap simulations are described as 'currently carrying out'. The feasibility claim is therefore an expectation, not a result demonstrated here. Please soften the wording (e.g. 'is expected to become feasible') or cite the ongoing work explicitly, so that the abstract does not overstate what the quenched study establishes.","section":"Abstract (sentence s4) and Section VIII","authors_acknowledge":null,"authors_acknowledge_p":null}],"minor_comments":[{"comment":"There are several typographical errors that should be corrected: 'occurence' (Section III) should be 'occurrence', 'peformed' (Acknowledgments) should be 'performed', 'the the matrix sign function' (Introduction) has a doubled word, and 'corres ponds' (Section III) should be 'corresponds'. The spelling 'L¨ uscher' should be 'Lüscher'.","section":"Throughout","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The vertical axis scales differ among the three panels of Figs. 2 and 3, and the paper notes this in the text. A common scale, or an inset showing the ratio of the mu != 0 density to the mu = 0 density, would make the comparison across gauge actions much easier for the reader.","section":"Figs. 2 and 3","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The modified plaquette action with the denominator 1 - (1 - ReTr P/3)/epsilon is unusual. A brief sentence explaining why this particular rational form was chosen, rather than a hard cutoff at epsilon, would help the reader understand the construction and its relation to the constraint used in Ref. [5].","section":"Eq. (3)","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The text states that <exp(-Delta H)> is 'consistent with 1 within one standard deviation' for most cases, but several entries (e.g. 0.995(3) and 1.005(4)) are closer to 1.5-1.7 sigma. The wording should reflect the actual deviations, or the errors should be interpreted more carefully.","section":"Table I","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"Reference [32] is given only as an arXiv number (hep-lat/0607024) without a journal reference. Please check whether it has since been published and update the citation accordingly.","section":"Reference [32]","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The quenched topology dependence of am_PS (about 10% between |Q| = 0 and 6) is a useful observation, but the extrapolation to the unquenched case is asserted rather than shown. A sentence clarifying that this is a quenched artifact, and that the dynamical case is under study, would prevent the reader from over-interpreting Fig. 4.","section":"Section VI","authors_acknowledge":null,"authors_acknowledge_p":null}],"recommendation":"major_revision","confidential_remarks":"This is a useful methods paper, and the practical claim (suppression of near-zero modes and reduced sign-function cost) is well supported. The main hurdle is the gap-versus-lambda^2 wording, which is load-bearing for the locality claim in the abstract; it is fixable by rewording and by adding a quantitative bound or fit. The authors disclose that Vranas [20] was submitted after this work was essentially completed; the editor may wish to ensure the relationship between the two papers is clearly stated in the published version."},"author_rebuttal":null,"desk_editor":{"model":"selfhosted/deepseek-v4.1-flash","letter":"The thing to know up front: the extra negative-mass Wilson fermions do what the authors claim — they crush the near-zero mode density of H_W — and the paper measures this directly with Lanczos eigenvalues across three gauge actions. That measurement is the real contribution, and it is done properly. The twisted-mass pseudo-fermion trick to cancel the β shift is a genuinely useful practical addition (the one-loop shift drops from 0.79 to 0.05 at μ = 0.2), and the topology-conservation check in Sec. V is clean: they start from Q = 0 and Q = −2 ensembles, count zero modes of the chirally projected overlap operator, and the count holds. They are also upfront that the idea is old (refs [16,17,18], Lüscher private communication) and that Vranas had a concurrent paper. No games there. The statement that an exact zero mode cannot appear is correct — the determinant weight vanishes at λ = 0 — and that much is solid. Now the soft spot that matters for the framing. The abstract says 'we find a gap near λ_W = 0.' The mechanism is a determinant weight λ²/(λ² + μ²), which for λ ≪ μ goes as λ²/μ², giving ρ(λ) ∝ λ². That vanishes at the origin but is not a gap; modes exist arbitrarily close to zero. The paper's own Sec. IV says the curve is 'consistent with ~λ²_W,' which is the honest statement and contradicts the abstract's word 'gap.' The 'no events in the lowest bin' is exactly what a λ² suppression plus ~100 configurations predicts, not evidence of a hard gap. Since HJL locality needs ρ = 0 on a neighborhood, not just ρ(0) = 0, the necessary condition is not actually demonstrated. The body is more careful than the abstract; the abstract should be fixed. Second, smaller: the HMC sampling. Sec. III reports ΔH spikes and ⟨e^{−ΔH}⟩ not always 1 within errors, then says 'we assume the configurations are properly sampled.' They show that halving δτ fixes it, so this reads as a practical caveat rather than a hidden bias, but it is stated rather than verified for the production runs. Third, minor: 30–100 configs, no upper bound quoted on ρ(0) with the extra fermions. None of this kills the method. The cost-reduction argument and the topology conservation stand independently of whether you call the suppression a gap. Who should read it: anyone running or planning overlap or domain-wall simulations. It is a methods paper, not a physics result. Recommendation: send it out. The referee should press on the gap/λ² wording and on the sampling assumption, but the work earns the time.","headline":"A solid, honest methods paper whose abstract says 'gap' where its own fit says ρ ∝ λ²; the practical case for the action survives regardless.","tokens_in":17692,"tokens_out":3005,"would_cite":true,"duration_ms":62035,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"pith_extraction":{"msc":[],"pacs":["11.15.Ha","12.38.Gc","11.30.Rd"],"model":"selfhosted/deepseek-v4.1-flash","headline":"Adding two unphysical heavy quarks to the lattice action opens a gap at zero in the spectrum of the hermitian Wilson-Dirac operator, which is what the locality proof for overlap fermions needs and what makes their numerical sign functionche","keywords":["lattice QCD","overlap fermions","hermitian Wilson-Dirac operator","near-zero modes","spectral density","topological charge","hybrid Monte Carlo","chiral fermions"],"falsifier":"Repeat the spectral measurement with a molecular dynamics step small enough that $langle e^{-Delta H} rangle = 1$ within errors on every ensemble, and check whether the lowest eigenvalue bin stays empty and $rho(0)$ stays zero; a reappearance of eigenvalues below $0.02$, or a nonzero $rho(0)$ in that cleaner run, would undercut the claimed gap.","tokens_in":18143,"feed_emoji":"⚛️","tokens_out":3299,"duration_ms":73339,"temperature":0.7,"pith_summary":"The paper proposes a lattice gauge action that adds two flavors of unphysical Wilson fermions with a large negative mass $m_0$ of order the inverse lattice spacing, together with a pair of twisted-mass pseudo-fermions. Because those extra fields contribute a determinant factor $[H_W(m_0)^2/(H_W(m_0)^2+mu^2)]$, the exact zero mode of the hermitian Wilson-Dirac operator $H_W(m_0)$ cannot appear and the near-zero modes are strongly suppressed. The authors measure the spectral density $rho(lambda_W)$ on quenched configurations for three gauge actions and find a gap near $lambda_W = 0$ with the extra fermions switched on, where without them the density at zero is nonzero. The payoff is twofold: the gap supplies a necessary condition for the Hernandez-Jansen-Luescher proof of exponential locality of the overlap-Dirac operator, and the small number of low modes cuts the cost of evaluating the matrix sign function enough to make dynamical overlap-fermion simulations practical.","feed_headline":"Heavy ghost quarks open a gap in the lattice Dirac spectrum","feed_subtitle":"Two unphysical Wilson fermions with large negative mass push near-zero modes out of the spectrum, making overlap fermions cheaper.","key_machinery":"The load-bearing object is the ratio of determinants $det[H_W(m_0)^2/(H_W(m_0)^2 + mu^2)]$ generated by the extra fermion sector of equation (4): two flavors of Wilson fermion at large negative mass plus one pseudo-fermion field with a flavor-twisted mass term $i mu gamma_5 tau_3$. The numerator forces the partition function to vanish whenever an eigenvalue of $H_W(m_0)$ approaches zero, so the near-zero modes that plague the sign function are statistically excluded rather than merely projected out; the twisted mass $mu$ sets the width of the excluded window, and the pseudo-fermion is arranged so that its contribution largely cancels the heavy Wilson fermion's ultraviolet effects in the combination $[D_W + i mu gamma_5] D_W^{-1}$. The same mechanism is what freezes the spectral flow of $H_W$ through zero, and hence the topological charge.","core_discovery":"On a $16^3 times 32$ lattice at roughly $0.125$ fm, the collaboration generates quenched-plus-extra-fermion ensembles with the standard Wilson plaquette action, a modified plaquette action with $1/epsilon = 2/3$, and the Iwasaki (RG-improved) action, each at $mu = 0$, $0.2$ and $0.4$, with $m_0 = -1.6$ throughout. Without the extra fields the spectral density at zero is nonzero and grows with the rougher action, $rho(0) = 1.509(3) times 10^{-3}$ for the standard plaquette action, $0.494(1) times 10^{-3}$ with $1/epsilon = 2/3$, and $0.177(1) times 10^{-3}$ for the RG action. With $mu = 0.2$ or $0.4$ no eigenvalue below about $0.02$ appears at all and the density near the origin behaves as $lambda_W^2$, so the gap is present for all three gauge actions and is widest for the Iwasaki action. The paper also reports that the topological charge, defined as the index of the overlap-Dirac operator built on $H_W(m_0)$, does not change during the molecular dynamics evolution: configurations started in the trivial sector show no chiral zero mode, and configurations started at $Q = -2$ retain exactly two left-handed zero modes. Finally, a one-loop calculation gives a $\\beta$ shift of $0.786(4)$ at $m_0 = -1.6$ from the extra Wilson fermions alone, falling to $0.0502(4)$ once the twisted-mass pseudo-fermions at $mu = 0.2$ are included.","pith_inferences":["If the gap survives at larger volumes and finer lattice spacings with the same $m_0$, the extra-fermion action becomes a generic preconditioner for any simulation that needs a well-defined sign function, not only for overlap fermions; a volume scan of $rho(0)$ at fixed physical lattice spacing would test this directly.","The fixed-topology property is a double-edged tool: it makes $epsilon$-regime work clean, but the paper's own quenched data show a $10%$ shift in $am_{PS}$ between $|Q| = 0$ and $|Q| = 6$, so any unquenched claim of small topology dependence should be checked against the same quantity rather than assumed from the $O(1/V)$ argument.","Because the suppression acts through a determinant ratio, its strength depends on $mu$ and on the number of extra flavors; a systematic scan of $mu$ at fixed lattice spacing would separate the desired gap from the residual beta shift more cleanly than the two-value comparison shown here."],"forward_implications":["The overlap-Dirac operator built on these configurations satisfies the gap condition required by the Hernandez-Jansen-Luescher locality proof, so exponential locality is no longer blocked by a nonzero density of near-zero modes.","The numerical cost of the matrix sign function drops, since the projection of low-lying eigenmodes scales with the eigenvalue density; the paper reports that the added Wilson-fermion inversion is cheap by comparison with the overlap inversion itself.","The reflection/refraction procedure used in dynamical overlap simulations to track sign-function jumps at $lambda_W = 0$ becomes unnecessary, because the crossing never happens for a small enough molecular dynamics step.","Topological charge is conserved along HMC trajectories, which is an advantage for $epsilon$-regime simulations at fixed topology and a disadvantage for sampling the $theta = 0$ vacuum.","The twisted-mass pseudo-fermions keep the beta shift small, avoiding the strong-coupling regime where the Wilson plaquette action shows a remnant of the fundamental-adjoint phase transition."],"supporting_citations":[{"why":"Supplies the Hernandez-Jansen-Luescher proof of exponential locality whose necessary condition, a gap in the $H_W$ spectrum, the paper is trying to satisfy.","marker":"[15]"},{"why":"Gives the smoothness bound $||1 - P_{mu nu}|| < 1/20.49$ under which $H_W$ is guaranteed to have a gap, the benchmark the paper compares its measured plaquette deviations against.","marker":"[5]"},{"why":"Earlier proposal to add heavy Wilson fermions to the lattice action, the idea the paper says has been around for years without detailed numerical study.","marker":"[16]"},{"why":"Vranas's contemporaneous paper on the same mechanism, cited as the closest independent treatment.","marker":"[20]"},{"why":"Source of the modified plaquette action with the denominator in equation (3), one of the three gauge actions tested here.","marker":"[23]"},{"why":"Defines the Iwasaki gauge action with $c_1 = -0.331$, the third action tested and the one showing the strongest near-zero-mode suppression.","marker":"[26]"},{"why":"Provides the chirally projected operator method used to count exact zero modes and hence verify topological charge conservation.","marker":"[30]"},{"why":"Supplies the $O(1/V)$ estimate for how hadron masses in a fixed topological sector deviate from the $theta = 0$ values.","marker":"[31]"},{"why":"One-loop vacuum polarization constant for Wilson fermions, the reference value against which the new negative-mass calculation is compared.","marker":"[35]"}],"fun_headline_variants":["Negative-mass Wilson fermions suppress near-zero Dirac modes","Ghost Wilson fermions open a gap in the overlap Dirac spectrum","Spectral gap found for three gauge actions with extra Wilson fermions","Topological charge stays fixed as ghost fermions open a Dirac gap"],"cache_read_input_tokens":0,"weakest_assumption_plain":"The central measurement assumes the hybrid Monte Carlo sampling is exact even though the runs show occasional large spikes in $Delta H$ and values of $langle e^{-Delta H} rangle$ that are not always consistent with one within errors; the paper states explicitly that it assumes the configurations are properly sampled.","fun_headline_variants_meta":{"raw":{"variants":["Negative-mass Wilson fermions suppress near-zero Dirac modes","Ghost Wilson fermions open a gap in the overlap Dirac spectrum","Spectral gap found for three gauge actions with extra Wilson fermions","Topological charge stays fixed as ghost fermions open a Dirac gap"]},"model":"selfhosted/deepseek-v4.1-flash","effort":"low","attempts":1,"cost_usd":0.0,"raw_usage":{"total_tokens":1728,"prompt_tokens":1123,"completion_tokens":605,"prompt_tokens_details":null,"completion_tokens_details":{"reasoning_tokens":545}},"tokens_in":1123,"tokens_out":605,"duration_ms":12790,"image_pages":[],"temperature":1.0,"reasoning_budget":50,"reasoning_tokens":545,"transport_provider":"selfhosted","cache_read_input_tokens":0,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-11-11T06:58:52.623996+00:00","model_set":{"reader":"selfhosted/deepseek-v4.1-flash"},"falsifier":"Repeat the spectral measurement with a molecular dynamics step small enough that $langle e^{-Delta H} rangle = 1$ within errors on every ensemble, and check whether the lowest eigenvalue bin stays empty and $rho(0)$ stays zero; a reappearance of eigenvalues below $0.02$, or a nonzero $rho(0)$ in that cleaner run, would undercut the claimed gap.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the Hernandez-Jansen-Luescher proof of exponential locality whose necessary condition, a gap in the $H_W$ spectrum, the paper is trying to satisfy."},{"cited_title":"Bounds on the Wilson Dirac Operator","cited_arxiv_id":"hep-lat/9911004","evidence_quote":"Gives the smoothness bound $||1 - P_{mu nu}|| < 1/20.49$ under which $H_W$ is guaranteed to have a gap, the benchmark the paper compares its measured plaquette deviations against."},{"cited_title":"Domain wall fermions in vector theories","cited_arxiv_id":"hep-lat/0001006","evidence_quote":"Earlier proposal to add heavy Wilson fermions to the lattice action, the idea the paper says has been around for years without detailed numerical study."},{"cited_title":"Gap Domain Wall Fermions","cited_arxiv_id":"hep-lat/0606014","evidence_quote":"Vranas's contemporaneous paper on the same mechanism, cited as the closest independent treatment."},{"cited_title":"Topology conserving gauge action and the overlap-Dirac operator","cited_arxiv_id":"hep-lat/0510116","evidence_quote":"Source of the modified plaquette action with the denominator in equation (3), one of the three gauge actions tested here."},{"cited_title":"Iwasaki, UTHEP-118","cited_arxiv_id":null,"evidence_quote":"Defines the Iwasaki gauge action with $c_1 = -0.331$, the third action tested and the one showing the strongest near-zero-mode suppression."},{"cited_title":"Numerical techniques for lattice QCD in the $\\epsilon$--regime","cited_arxiv_id":"hep-lat/0212012","evidence_quote":"Provides the chirally projected operator method used to count exact zero modes and hence verify topological charge conservation."},{"cited_title":"QCD at Fixed Topology","cited_arxiv_id":"hep-lat/0302005","evidence_quote":"Supplies the $O(1/V)$ estimate for how hadron masses in a fixed topological sector deviate from the $theta = 0$ values."},{"cited_title":"Phys. Lett. B 100, 331","cited_arxiv_id":null,"evidence_quote":"One-loop vacuum polarization constant for Wilson fermions, the reference value against which the new negative-mass calculation is compared."}],"review_version":1}