Pith. sign in

REVIEW 3 major objections 5 minor 32 references

Smeared $R$-ratio in isospin symmetric QCD with Low Mode Averaging

T0 review · 3 major / 5 minor · reviewed 2026-08-09 · deepseek-v4-flash

Pith's one-line read Combining Low Mode Averaging with the HLT spectral-density reconstruction reduces the statistical error of the light vector-vector correlator by a factor of about 3.6 and resolves the rho resonance in the smeared R-ratio at a Gaussian…

desk verdict The LMA gain is solid and useful; the sigma=250 MeV smeared R-ratio is interesting but the HLT systematic control at that width is not yet demonstrated. read the letter →

arxiv 2502.03187 v1 pith:5KQUCJAD submitted 2025-02-05 hep-lat

classification hep-lat
keywords latticeQCDsmearedR-ratioLowModeAveragingspectraldensityreconstructionvector-vectorcorrelatorrhoresonancemuong-2isospinsymmetric
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper is a methods demonstration in lattice QCD: it combines Low Mode Averaging (LMA) with the HLT spectral-density reconstruction to compute the light-quark connected contribution to the smeared R-ratio in isospin-symmetric QCD. The authors report that LMA reduces the statistical error of the vector-vector correlator by a factor of about 3.6 while increasing the cost by only about 2.7, giving a net speedup of about 5 to reach a fixed accuracy. With that gain they are able to reduce the Gaussian smearing width to about 250 MeV, roughly half the width of previous lattice extractions. At a center-of-mass energy near 770 MeV the relative error on the smeared R-ratio falls from about 11% to about 3%, which makes the rho resonance visible directly in the lattice data. The results are preliminary and blinded, but they establish a practical route to finer energy resolution in first-principles determinations of the R-ratio.

What carries the argument

The machinery has two parts. Low Mode Averaging (LMA) deflates the Hermitian Dirac operator $Q_W = \gamma_5 D_W$: its $N_v$ lowest eigenpairs are computed exactly and used to build an all-to-all infrared propagator, while the ultraviolet part is still estimated stochastically. Because the stochastic sources are time- and spin-diluted, the deflated and non-deflated two-point correlators differ only in their infrared part, so the exact infrared contribution can be added back without touching the ultraviolet noise. The second part is the HLT spectral-density reconstruction, which approximates the target smearing kernel by a short exponential sum $\sum_\tau g_\tau e^{-a\omega\tau}$, fixing the coefficients $g_\tau$ by minimizing a weighted $L^2$ distance to the desired Gaussian kernel subject to a covariance penalty. The correlator $C(t) = (12\pi^2)^{-1}\int d\omega\, e^{-\omega t}\,\omega^2 R(\omega)$ then supplies the smeared R-ratio. The gain in signal-to-noise is quantified by comparing deflated and non-deflated correlators with adjusted numbers of configurations and stochastic sources.

What would settle it

Run the HLT analysis at $\sigma = 250$ MeV with several truncation values $\tau_{\rm max}$ and with 400 versus 500 eigenvectors: if the reconstructed R-ratio at $E \sim 770$ MeV moves by more than the quoted roughly 3% statistical error when $\tau_{\rm max}$ is increased, or if the two lattice regularizations no longer agree after unblinding, then the systematic control claim for the narrower width would be disproven.

Watch

Extended reading notes

Core claim

The central claim is that deflating the low eigenmodes of the Dirac operator does not merely reduce noise in the correlator; it changes which physics can be extracted from the same lattice ensembles. In the mixed-action twisted-mass setup, the deflated two-point function is obtained from the stochastic correlator by removing the stochastically estimated infrared part and adding back the exact all-to-all infrared part. With around 400 eigenvectors and about 1000 time-diluted stochastic sources per configuration, the signal-to-noise ratio on the light vector-vector correlator improves by a factor of 3.63(15) and 3.57(16) for the two current regularizations, at a cost increase of 2.7(2), i.e. a net speedup of about 5. Feeding these correlators into the HLT spectral-density reconstruction yields the connected $u/d$ contribution to the Gaussian-smeared R-ratio $R^\ell_\sigma(E)$ with $\sigma = 250$ MeV, where the relative error at $E \sim 770$ MeV drops from roughly 11% to 3%, exposing the rho resonance. The same pipeline is applied at three lattice spacings (about 0.057, 0.068 and 0.080 fm) and the results from the two regularizations are mutually compatible, so the authors regard the rho signal as a genuine physics outcome rather than a discretization artifact.

Load-bearing premise

The load-bearing premise is that the HLT reconstruction stays systematically under control at the narrower Gaussian width of 250 MeV, because the method had previously been validated only for widths of 440 to 630 MeV and this paper does not yet give an explicit error analysis for truncation or excited-state contamination at the new width.

Editorial extensions

If this is right

  • At $E \sim 770$ MeV and $\sigma = 250$ MeV, the relative error on the smeared R-ratio drops from about 11% to about 3%, making the rho resonance visible in the lattice data.
  • The signal-to-noise gain is 3.63(15) for one current regularization and 3.57(16) for the other; with a computational cost increase of about 2.7(2), the net speedup to reach a given accuracy is about 5(1).
  • The method is demonstrated at three lattice spacings (about 0.057, 0.068 and 0.080 fm) with volumes up to about 5.5 fm, so the smeared R-ratio can be computed with controlled discretization effects.
  • Results from the two lattice regularizations are compatible within the reached precision, allowing low-energy lattice artifacts specific to each regularization to be identified.
  • Combining LMA with HLT makes Gaussian widths of 250 MeV feasible, roughly half the previously accessible width, enabling a more direct comparison with phenomenological smeared R-ratio data.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Going beyond the paper: if the HLT truncation systematics are confirmed at 250 MeV, the same LMA plus HLT combination could plausibly reach widths near 150-200 MeV, where the omega and phi resonances, and possibly excited vector states, would come into view rather than just the rho.
  • The gain factor is expected to grow on larger physical volumes because the number of low Dirac modes scales with the spacetime volume, so on boxes larger than the 5.1-5.5 fm used here the statistical advantage of LMA should become even more pronounced.
  • Because LMA removes only the exactly-deflated infrared noise, its benefit should transfer to other bilinear channels, such as axial or scalar correlators, using the same eigenvector set and at no extra inversion cost; testing this on the same ensembles would be a direct extension.
  • A practical consequence for the muon g-2 program is that cutting the statistical error on the connected light-quark contribution by about 3.6 times at fixed cost would allow the disconnected and strange or charm contributions to be computed at comparable precision, which often dominate the error budget.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 5 minor

Summary. The paper reports a proof-of-concept application of Low Mode Averaging (LMA) to the light-quark connected vector-vector correlator in the ETMC mixed-action setup, with the goal of enabling Hansen-Lupo-Tantalo (HLT) spectral reconstruction of the smeared R-ratio at a Gaussian width of sigma = 250 MeV. The LMA estimator is defined in Eq. (13), and the authors show that it reduces the statistical error of the correlator by a factor of 3.63(15) for TM and 3.57(16) for OS regularizations, with a net computational speedup of about 5. Using these improved correlators, they present preliminary, blinded results for the light connected contribution to the smeared R-ratio at sigma = 250 MeV, reporting that the relative error at E around 770 MeV drops from about 11% to 3%, which they interpret as enough to resolve the rho resonance.

Significance. If the HLT systematic errors at sigma = 250 MeV are under control, this paper would be a significant methodological advance: LMA makes spectral reconstruction at smaller Gaussian widths feasible, and the quantitative error-reduction benchmark against data from Ref. [11] is a concrete and reproducible result. The LMA part is well supported: Eq. (13) is a clean deflation identity, and the measured noise reductions and cost model give a net speedup of about 5. The HLT extension to sigma = 250 MeV, however, is the load-bearing novelty, and it is not yet backed by a systematic error analysis. The agreement between TM and OS is a useful cross-check but does not constrain the shared HLT bias. The paper is a preliminary proceedings contribution, and the central physics claim therefore remains conditional on the missing systematic validation.

major comments (3)
  1. [Sec. 3, Eqs. (15)-(17), Fig. 5] The central claim that sigma = 250 MeV is sufficient to resolve the rho resonance is not yet supported by a systematic error analysis for the HLT reconstruction at this width. The approximation space in Eq. (15) is a truncated exponential basis, and the Tikhonov regularization in Eq. (17) introduces a bias; both become more severe as the Gaussian kernel narrows. Ref. [1] validated HLT for widths of 440-630 MeV, while the present paper uses sigma = 250 MeV without providing a tau_max scan, a lambda-stability study, or a synthetic-data test at this width. Consequently, the reported reduction from 11% to 3% at E ~ 770 MeV is purely statistical. The TM/OS agreement in Fig. 5 (right) is a useful cross-check, but it does not constrain the shared HLT systematic. I ask for an explicit HLT stability analysis or a synthetic-data validation at sigma = 250 MeV before the rho-resolution claim is made.
  2. [Sec. 3, Fig. 5] No systematic error budget is provided for the smeared R-ratio: there is no continuum extrapolation (only two lattice spacings, B64 and C80, are shown), no estimate of finite-volume, O(a), or isospin-breaking effects, and the results are blinded. As a result, the error bars in Fig. 5 do not represent the total uncertainty relevant for comparing with the rho peak, and the TM/OS agreement cannot be taken as evidence of continuum behavior. A proceedings paper may report preliminary status, but the phrase 'enough to appreciate the rho resonance' should be qualified as referring to statistical precision only, pending the systematic analysis.
  3. [Sec. 3, Fig. 5] The plotted quantity is the connected light-quark (u/d) contribution to the smeared R-ratio, not the full R-ratio defined in Eq. (2), which also receives disconnected and heavier-quark contributions. The text should consistently say 'light-quark connected contribution' when discussing the rho signal, and it should state how much of the rho peak is expected to arise from this partial contribution. Without this qualification, the claim that the full R-ratio has been resolved at sigma = 250 MeV is stronger than what the data show.
minor comments (5)
  1. [Eq. (14), Sec. 2.2] The formula for the Gain contains a stray symbol 'vt' before the two ratios; this is likely a LaTeX artifact and should be removed.
  2. [Title and Abstract] The name 'Hansen-Lupo-Tantatlo' is misspelled; it should be 'Hansen-Lupo-Tantalo', matching the reference list and Sec. 2.3.
  3. [Sec. 2.2, Table 1] For the B64 ensemble the optimal value Neig = 400 is supported by Fig. 2, but Table 1 lists Neig = 530 for C80 and D96 without showing the corresponding optimization scans; the choice for the finer ensembles should be justified or described as a scaling assumption.
  4. [Sec. 2.3, Eq. (16)] The notation for the weight function is inconsistent: the text says 'weight-functions w_n > 0', while Eq. (16) uses w_n(omega); the relation between the subscript n and the functional A_n should be stated explicitly.
  5. [References] Ref. [8] is incomplete ('JHEP 04 (2004) .') and Ref. [15] is a preprint 'In preparation'; these entries should be completed where possible.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the LMA gain is measured and the HLT output is defined by a known kernel fit, with the sigma=250 MeV systematic an open uncertainty rather than a circular reduction.

full rationale

The paper's central results are not circular. The LMA method is presented through the exact relation Eq. (13), C_defl = C_stoch - C_IR_stoch + C_IR_exact, which is an identity for the deflated estimator, not a fit. The quoted gain in signal-to-noise, 3.63(15) and 3.57(16), is computed from measured correlator errors and the stated numbers of configurations and sources; it is not adjusted to produce any particular final value. The HLT reconstruction is likewise not circular: the coefficients g_tau are determined by minimizing A_n[g] = ∫ dω w_n(ω) |K(ω;g) - 12π² G_σ(E-ω)/ω²|², in which the target is the known Gaussian smearing kernel, not the measured R-ratio. The subsequent R_σ(E) is then a linear combination of the lattice correlators. The comparison with the no-LMA result and the TM/OS cross-check in Fig. 5 are independent checks, and the stated blinding further reduces the possibility of tuning toward a desired phenomenological shape. Refs. [1], [2], and [11] include overlapping authors, but they supply the HLT algorithm and the earlier no-LMA data rather than an unverified uniqueness theorem or a fitted parameter disguised as a prediction. The main weakness, that HLT at σ=250 MeV has not yet been validated with a tau_max scan or a systematic error analysis at that width, is a correctness and robustness limitation, not a circularity: the estimator is still defined by the kernel objective and could be tested against synthetic data or other ensembles, as the paper itself implies by labeling the results preliminary and blinded. No load-bearing derivation reduces by construction to its inputs.

Assumptions & free parameters 4 free parameters · 4 assumptions · 0 invented entities

No new particles, forces, or dimensions are introduced. The free parameters are computational choices that affect cost and statistical precision, not the physics content of the result. The axioms are standard tools in lattice QCD, with the volume-scaling assumption being the least certain but not biasing the central result.

free parameters (4)
  • Number of low modes Neig = 400 (B64), 530 (C80, D96)
    Chosen by scanning Neig=200,300,400,500 on B64 for best gain-to-cost tradeoff; similar values are used on the other ensembles.
  • Number of stochastic sources Neta = 1024 (B64), 960 (C80, D96)
    Stochastic noise saturates at roughly 1000 time-wall sources per configuration; values are chosen for stable noise behavior.
  • IR eigenmode threshold lambda_tilde_thrs = approximately 8 mu_sea
    Used to define the low-mode subspace; the exact value is a practical convenience and affects computational cost, not the physics result.
  • HLT Tikhonov parameter lambda = varied in stability analysis
    Regularization parameter in the HLT functional; the paper reports that it is varied and optimized in a stability analysis, but the final numerical choice is not given.
assumptions (4)
  • standard math Spectral representation C(t) = (1/12pi^2) * integral domega exp(-omega t) omega^2 R(omega)
    Eq. (4) connects the Euclidean vector-vector correlator to the R-ratio. This is a standard dispersion relation in lattice QCD.
  • domain assumption Eigenvalue density near zero scales as Banks-Casher: (Delta N / Delta lambda_tilde)(0) proportional to Lambda_QCD^3 L^3 T
    Invoked in Section 2.2 to scale the required number of low modes to other lattice spacings and volumes. If the scaling is inaccurate, the chosen Neig values may be suboptimal, though the method remains unbiased.
  • standard math Spin-diluted stochastic sources allow the UV contribution to be identical between deflated and non-deflated propagators
    Used in Section 2.1, Eq. (10)-(13), to justify computing the IR part exactly and the UV part stochastically. This is a mathematical property of the source dilution.
  • domain assumption Identical stochastic sources can be generated on different node configurations using site-specific random seeds
    Practical requirement for splitting the stochastic and IR computation runs, stated in Section 2.1 with reference to QUDA-like seeding. This is not proven in the paper.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Smeared $R$-ratio in isospin symmetric QCD with Low Mode Averaging." pith.science (2026). https://pith.science/paper/5KQUCJAD

@misc{pith2026250203187,
  author       = {Pith},
  title        = {Pith review of: Smeared $R$-ratio in isospin symmetric QCD with Low Mode Averaging},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/5KQUCJAD}},
  note         = {Machine review of arXiv:2502.03187}
}
abstract

Low Mode Average (LMA) is a technique to improve the quality of the signal-to-noise ratio in the long time separation of Euclidean correlation functions. We report on its beneficial impact in computing the vector-vector light connected two-point correlation functions and derived physical quantities in the mixed action lattice setup adopted by ETM collaboration. We focus on preliminary results of the computation within isospin symmetric QCD (isoQCD) of the $R$-ratio smeared with Gaussian kernels of widths down to $\sigma\sim250$ MeV, which is enough to appreciate the $\rho$ resonance around 770 MeV, using the Hansen-Lupo-Tantatlo (HLT) spectral-density reconstruction method.

Figures

Figures reproduced from arXiv: 2502.03187 by the authors.

Figure 1
Figure 1. The first-principle lattice results of 2023 [1] compared to the smeared experimental 𝑅-ratio from KNT19 compilation [3]. = {TM, OS} corresponding to the so-called Twisted-Mass (TM) and Osterwalder-Seiler (OS) regularizations, see Ref. [6] for details. This correlator is our primary observable directly computed via lattice simulations and it is connected to the 𝑅-ratio by the well known formula 𝐶(𝑡) = 1 12𝜋 2 ∫ ∞ 𝐸th… view at source ↗
Figure 2
Figure 2. The gain in the signal-to-noise ratio defined in Eq. (14), computed for the deflated two-point correlators on a 643 × 128 lattice with 𝐿 = 5.1 fm, for TM lattice regularization. Results are shown for 𝑁eig = 200, 300, 400, 500, 𝑁𝑈 = 56 and 𝑁𝜂 = 256. The optimal choice is 𝑁eig = 400, achieving a gain statistically compatible with that for 500 eigenvectors while requiring fewer computational resources. Ensemble name (s… view at source ↗
Figure 3
Figure 3. Left: Comparison of the two-point e.m. current correlator in isoQCD, on a 643 × 128 lattice with 𝐿 = 5.1 fm, by using the old setup, with no LMA and (𝑁𝑈, 𝑁𝜂) = (780, 1000) and the new setup, with LMA and a reduced statistics (𝑁𝑈, 𝑁𝜂) = (158, 1000) - the full statistics data are blinded, further details will be given in Ref. [15]. Right: the gain in the signal-to-noise ratio defined in Eq. (14). We reach a reduction … view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: The frequency histogram (Δ𝑁/Δ𝜆˜) (𝜆˜) of eigenvalues 𝜆˜ of (𝑄 2 𝑊 + 𝜇 2 ) 1/2 for B64, C80 and D96 ensembles (see Tab. 1 for details on the gauge ensembles). where 𝜏 is an integer variable and 𝑎 is the lattice spacing. The distance between the target kernel and its rep…
Figure 5
Figure 5. Figure 5: Preliminary blinded values of the connected contribution of the 𝑢/𝑑 quark mass to the smeared 𝑅𝜎 (𝐸) with central energies up to 1.6 GeV and a fixed resolution of 𝜎 = 250 MeV. Left: Results for the B64 ensemble, obtained through a spectral reconstruction analysis using…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

32 extracted references · 11 canonical work pages

  1. [1]

    Alexandrou et al., Probing the Energy-Smeared R Ratio Using Lattice QCD, Phys

    Extended Twisted Mass Collaboration (ETMC) collaboration, C. Alexandrou et al., Probing the Energy-Smeared R Ratio Using Lattice QCD, Phys. Rev. Lett.130 (2023) 241901 [2212.08467]

  2. [15]

    Alexandrou et al., Light quark contributions to the muon anomalous magnetic moment in lattice QCD with twisted-mass fermions,In preparation(2025)

    Extended Twisted Mass Collaboration (ETMC) collaboration, C. Alexandrou et al., Light quark contributions to the muon anomalous magnetic moment in lattice QCD with twisted-mass fermions,In preparation(2025)

  3. [11]

    Alexandrou, S

    Extended Twisted Mass Collaboration collaboration, C. Alexandrou, S. Bacchio, P.Dimopoulos,J.Finkenrath,R.Frezzotti,G.Gagliardietal., Latticecalculationoftheshort and intermediate time-distance hadronic vacuum polarization contributions to the muon magnetic moment using twisted-mass fermions,Phys. Rev. D107 (2023) 074506

  4. [2]

    Hansen, A

    M. Hansen, A. Lupo and N. Tantalo,Extraction of spectral densities from lattice correlators, Phys. Rev. D99(2019) 094508 [1903.06476]

  5. [3]

    Keshavarzi, D

    A. Keshavarzi, D. Nomura and T. Teubner,𝑔− 2 of charged leptons,𝛼(𝑀2 𝑍) , and the hyperfine splitting of muonium, Phys. Rev. D101 (2020) 014029 [1911.00367]

  6. [4]

    Frezzotti and G

    R. Frezzotti and G. C. Rossi,Chirally improving Wilson fermions. 1. O(a) improvement, JHEP 08(2004) 007 [hep-lat/0306014]. 9 Smeared𝑅-ratio in isospin symmetric QCD with Low Mode Averaging Francesca Margari

  7. [5]

    Frezzotti and G

    R. Frezzotti and G. C. Rossi,Chirally improving Wilson fermions. II. Four-quark operators, JHEP 10 (2004) 070 [hep-lat/0407002]

  8. [6]

    Alexandrou et al., Strange and charm quark contributions to the muon anomalous magnetic moment in lattice QCD with twisted-mass fermions, 2411.08852

    Extended Twisted Mass Collaboration (ETMC) collaboration, C. Alexandrou et al., Strange and charm quark contributions to the muon anomalous magnetic moment in lattice QCD with twisted-mass fermions, 2411.08852

Show all 32 references
  1. [7]

    H. Neff, N. Eicker, T. Lippert, J. W. Negele and K. Schilling,Low fermionic eigenmode dominance in qcd on the lattice, Phys. Rev. D64 (2001) 114509

  2. [8]

    Giusti et al.,Low-energy couplings of QCD from current correlators near the chiral limit, JHEP 04 (2004)

    L. Giusti et al.,Low-energy couplings of QCD from current correlators near the chiral limit, JHEP 04 (2004)

  3. [9]

    T. A. DeGrand and S. Schaefer,Improving meson two point functions in lattice QCD, Comput. Phys. Commun.159(2004) 185 [hep-lat/0401011]

  4. [10]

    Borsanyi et al.,Leading hadronic contribution to the muon magnetic moment from lattice QCD, Nature 593 (2021) 51 [2002.12347]

    S. Borsanyi et al.,Leading hadronic contribution to the muon magnetic moment from lattice QCD, Nature 593 (2021) 51 [2002.12347]

  5. [12]

    Banks and A

    T. Banks and A. Casher,Chiral symmetry breaking in confining theories, Nuclear Physics B 169 (1980) 103

  6. [13]

    Leutwyler and A

    H. Leutwyler and A. V. Smilga,Spectrum of Dirac operator and role of winding number in QCD, Phys. Rev. D46(1992) 5607

  7. [14]

    Luscher,Local coherence and deflation of the low quark modes in lattice QCD,JHEP 07 (2007) 081 [0706.2298]

    M. Luscher,Local coherence and deflation of the low quark modes in lattice QCD,JHEP 07 (2007) 081 [0706.2298]

  8. [16]

    Alexandrou et al.,Status of the ETMC ensemble generation effort, PoS LATTICE2024 (2025) 429

    C. Alexandrou et al.,Status of the ETMC ensemble generation effort, PoS LATTICE2024 (2025) 429

  9. [17]

    Jansen and C

    K. Jansen and C. Urbach,tmLQCD: A Program suite to simulate Wilson Twisted mass Lattice QCD,Comput. Phys. Commun.180 (2009) 2717 [0905.3331]

  10. [18]

    A.Abdel-Rehim,F.Burger,A.Deuzeman,K.Jansen,B.Kostrzewa,L.Scorzatoetal., Recent developments in the tmLQCD software suite, PoS LATTICE2013(2014) 414 [1311.5495]

  11. [19]

    Deuzeman, K

    A. Deuzeman, K. Jansen, B. Kostrzewa and C. Urbach,Experiences with OpenMP in tmLQCD,PoS LATTICE2013(2014) 416 [1311.4521]. 10 Smeared𝑅-ratio in isospin symmetric QCD with Low Mode Averaging Francesca Margari

  12. [20]

    Kostrzewa, S

    ETM collaboration, B. Kostrzewa, S. Bacchio, J. Finkenrath, M. Garofalo, F. Pittler, S. Romiti et al.,Twisted mass ensemble generation on GPU machines,PoS LATTICE2022 (2023) 340 [2212.06635]

  13. [21]

    Deuzeman, S

    ETM collaboration, A. Deuzeman, S. Reker and C. Urbach,Lemon: an MPI parallel I/O library for data encapsulation using LIME, Comput. Phys. Commun.183 (2012) 1321 [1106.4177]

  14. [22]

    Frommer, K

    A. Frommer, K. Kahl, S. Krieg, B. Leder and M. Rottmann,Adaptive Aggregation-Based Domain Decomposition Multigrid for the Lattice Wilson–Dirac Operator, SIAM J. Sci. Comput. 36(2014) A1581 [1303.1377]

  15. [23]

    Alexandrou, S

    C. Alexandrou, S. Bacchio, J. Finkenrath, A. Frommer, K. Kahl and M. Rottmann,Adaptive Aggregation-based Domain Decomposition Multigrid for Twisted Mass Fermions,Phys. Rev. D 94(2016) 114509 [1610.02370]

  16. [24]

    Bacchio, C

    S. Bacchio, C. Alexandrou and J. Finkerath,Multigrid accelerated simulations for Twisted Mass fermions,EPJ Web Conf.175 (2018) 02002 [1710.06198]

  17. [25]

    Alexandrou, S

    C. Alexandrou, S. Bacchio and J. Finkenrath,Multigrid approach in shifted linear systems for the non-degenerated twisted mass operator, Comput. Phys. Commun.236 (2019) 51 [1805.09584]

  18. [26]

    B. Joó, D. D. Kalamkar, T. Kurth, K. Vaidyanathan and A. Walden,Optimizing Wilson-Dirac Operator and Linear Solvers for Intel® KNL, inHigh Performance Computing(M. Taufer, B. Mohr and J. M. Kunkel, eds.), (Cham), pp. 415–427, Springer International Publishing, 2016

  19. [27]

    Schröck, S

    M. Schröck, S. Simula and A. Strelchenko,Accelerating Twisted Mass LQCD with QPhiX, PoS LATTICE2015(2016) 030 [1510.08879]

  20. [28]

    M. A. Clark, R. Babich, K. Barros, R. C. Brower and C. Rebbi,Solving Lattice QCD systems of equations using mixed precision solvers on GPUs, Comput. Phys. Commun.181 (2010) 1517 [0911.3191]

  21. [29]

    Babich, M

    R. Babich, M. A. Clark, B. Joo, G. Shi, R. C. Brower and S. Gottlieb,Scaling Lattice QCD beyond 100 GPUs, inSC11 International Conference for High Performance Computing, Networking, Storage and Analysis Seattle, Washington, November 12-18, 2011, 2011, 1109.2935, DOI

  22. [30]

    M. A. Clark, B. Joó, A. Strelchenko, M. Cheng, A. Gambhir and R. C. Brower,Accelerating Lattice QCD Multigrid on GPUs Using Fine-Grained Parallelization, inSC ’16: Proceedings of the International Conference for High Performance Computing, Networking, Storage and Analysis, pp....

  23. [31]

    11 Smeared𝑅-ratio in isospin symmetric QCD with Low Mode Averaging Francesca Margari

    Jülich Supercomputing Centre,JUWELS: Modular Tier-0/1 Supercomputer at the Jülich Supercomputing Centre,Journal of large-scale research facilities5 (2019) . 11 Smeared𝑅-ratio in isospin symmetric QCD with Low Mode Averaging Francesca Margari

  24. [32]

    Jülich Supercomputing Centre,JUWELS Cluster and Booster: Exascale Pathfinder with Modular Supercomputing Architecture at Juelich Supercomputing Centre, Journal of large-scale research facilities7 (2021) . 12

Pith tools

Reviewed August 9, 2026 · model on record in the stance chip above.