arXiv subjects
Publications and source records attributed to Ken-Ichi Ishikawa.
We present results for the large-$N$ limit of the chiral condensate computed from twisted reduced models. We followed a two-fold strategy, one constiting in extracting the condensate from the quark-mass dependence of the pion mass, the other consisting in extracting the condensate from the mode number of the Dirac operator.
We present results for the axial charge and root-mean-square (RMS) radii of the nucleon obtained from 2+1 flavor lattice QCD at the physical point with a large spatial extent of about 10 fm. Our calculations are performed with the PACS10 gauge configurations generated by the PACS Collaboration with the six stout-smeared $O(a)$ improved Wilson-clover quark action and Iwasaki gauge action at $β$ = 1.82 and 2.00 corresponding to lattice spacings of 0.085 fm and 0.063 fm respectively. We first evaluate the value of $g_A/g_V$ , which is not renormalized in the continuum limit and thus ends up with the renormalized axial charge. Moreover, we also calculate the nucleon elastic form factors and determine three kinds of isovector RMS radii such as electric, magnetic and axial ones at the two lattice spacings. We finally discuss the discretization uncertainties on renormalized axial charge and isovector RMS radii towards the continuum limit.
We present results for the scalar and tensor isovector-couplings ($g_S$ and $g_T$) of the nucleon measured at the physical point ($M_π=135$ MeV) with a single lattice spacing of $0.085\ \mathrm{fm}$ in 2+1 flavor QCD. Our calculations are carried out with two ensembles of gauge configurations generated by the PACS Collaboration with nonperturbatively ${\cal O}(a)$ improved Wilson quark action and Iwasaki gauge action on $(10.9\ {\rm fm})^4$ and $(5.5\ {\rm fm})^4$ lattices, where the finite-size effect on the nucleon mass was not shown at the level of the statistical precision less than 0.5%. We compute the nucleon three-point correlation functions in the vector, axial, scalar, and tensor channels. We confirm that our previous result of the nucleon axial coupling on the large spatial volume of $(10.9\ {\rm fm})^4$ has no finite-size effect at the level of the statistical precision of 1.9%. For the renormalization, we first renormalize $g_S$ and $g_T$ nonperturbatively using the RI/SMOM$_{(γ_μ)}$ scheme, a variant of Rome-Southampton RI/MOM scheme with reduced systematic errors, as the intermediate scheme. We evaluate our final results at the renormalization scale of 2 GeV in the $\overline{\rm MS}$ scheme through matching procedure between the RI/SMOM$_{(γ_μ)}$ and $\overline{\rm MS}$ schemes with the help of perturbation theory, and then obtain $g_S=0.927(71)_{\rm stat}(22)_{\rm syst}$ and $g_T=1.036(6)_{\rm stat}(20)_{\rm syst}$.
We compute the first four perturbative coefficients of the internal energy for the twisted reduced principal chiral model (TRPCM) using numerical stochastic perturbation theory (NSPT). This matrix model has the same large $N$ limit as the ordinary principal chiral model (PCM) at infinite volume. Indeed, we verify that the first three coefficients match the analytic result for the PCM coefficients at large $N$ with a precision of three to four significant digits. The fourth coefficient also matches our own NSPT calculation of the corresponding PCM coefficient at large $N$. The finite-$N$ corrections to all coefficients beyond the leading order are smaller for TRPCM than for PCM. We analyze the variance to determine the feasibility of extending the calculations to higher orders.
In this paper we study the large $N$ limit of four-dimensional Supersymmetric Yang-Mills on the lattice using twisted reduced models. We have generated configurations with dynamical massive gluinos and various lattice 't Hooft couplings, and verified that the Pfaffian remains positive. We have determined the lattice spacing in terms of various observables obtaining compatible results. Extrapolating to the massless gluino limit we obtain the lattice spacing dependence on the bare couplings for the supersymmetric theory. The observed dependence goes along the expected behaviour predicted by perturbation theory.
We present preliminary results for the axial, scalar and tensor charges of the nucleon measured in 2+1 flavor QCD with the physical light quarks ($m_π=135$ MeV). Our simulations are carried out with gauge configurations generated by the PACS Collaboration with the stout-smeared $O(a)$ improved Wilson fermions and Iwasaki gauge action at a single lattice spacing of $0.085\ (\mathrm{fm})$. There are two lattice ensembles of the PACS gauge configurations, which have physical lattice sizes over $(10\ \mathrm{fm})^4$ and $(5\ \mathrm{fm})^4$, respectively. We compute the nucleon three-point correlation functions in the axial, scalar, and tensor channels. For the renormalization, we use the Rome-Southampton method as the intermediate scheme in order to evaluate the renormalization constants for the scalar and tensor currents in fully nonperturbative manner. We then evaluate the renormalized values of the scalar and tensor charges ($g_S$ and $g_T$) in the $\overline{\rm MS}$ scheme at the renormalization scale of 2 GeV with a help of the continuum perturbation theory for the matching between two schemes. We compare our preliminary results of $g_S$ and $g_T$ with those of other collaboration results.
We present results for the nucleon structure functions and form factors obtained from 2+1 flavor lattice QCD with physical light quark masses ($m_π=135$ MeV) in a large spatial extent of about 10 fm. Our calculations are performed with the PACS10 gauge configurations generated by the PACS Collaboration with the six stout-smeared ${\mathscr{O}}(a)$ improved Wilson-clover quark action and Iwasaki gauge action at $β=1.82$ and $2.00$ corresponding to lattice spacings of $0.085$ fm and $0.064$ fm respectively. The lower moments of structure functions, $\langle x \rangle_{u-d}$ and $\langle x \rangle_{Δu - Δd}$ given by the twist-2 operators being properly renormalized, are evaluated in the $\overline{\rm MS}$ scheme at the renormalization scale of 2 GeV only at $β=1.82$, since the renormalization factors at $β=2.00$ have not yet determined nonperturbatively in the RI/MOM scheme. Instead, at two lattice spacings, we evaluate appropriate ratios of $g_{A}/g_{V}$ and $\langle x \rangle_{u-d}/\langle x \rangle_{Δu -Δd}$, which are not renormalized in the continuum limit. These quantities thus can be directly compared with the experimental data without the renormalization.
We report an implementation of a multigrid solver for the Clover fermion on supercomputer Fugaku, which uses A64FX CPU with Arm architecture. On Fugaku, a highly optimized implementation of BiCGStab solver with domain decomposed preconditioner for the Clover fermion, called QCD Wide SIMD library (QWS), is available. We use the preconditioner in QWS as a smoother of the multigrid solver. Our implementation shows reasonably good scaling and performance. The code is developed by using Bridge++ code framework and its extension.
A portable implementation of elaborated algorithm is important to use variety of architectures in HPC applications. In this work we implement and benchmark an algebraic multi-grid solver for Lattice QCD on three different architectures, Intel Xeon Phi, Fujitsu A64FX, and NVIDIA Tesla V100, in keeping high performance and portability of the code based on the object-oriented paradigm. Some parts of code are specific to an architecture employing appropriate data layout and tuned matrix-vector multiplication kernels, while the implementation of abstract solver algorithm is common to all architectures. Although the performance of the solver depends on tuning of the architecture-dependent part, we observe reasonable scaling behavior and better performance than the mixed precision BiCGSstab solvers.
To investigate the properties of the large $N$ limit of $\mathcal{N} = 1$ SUSY Yang-Mills theory, we have started a study for a reduced matrix model with an adjoint Majorana fermion. The gauge action is based on the Wilson action and the adjoint-fermion one is the Wilson-Dirac action on a reduced lattice with twisted gauge boundary condition. We employ the RHMC algorithm in which the absolute value of the Pfaffian is incorporated. The sign of the Pfaffian is included with the re-weighting method and separately measured as an observable. In this talk, we show the configuration generation status towards the large $N$ limit and the behavior of the lowest/lower eigenvalue(s) of the Wilson-Dirac adjoint fermion operator. We investigated the sign of the Pfaffian and the critical hopping parameters for the chiral limit. The sign of the Pfaffian is always positive on the configurations we have generated. The critical hopping parameters derived from the eigenvalues of the Dirac operator are consistent with those derived from the PCAC mass relation with non-singlet flavor adjoint fermions.
We present results on the world's first over 100 PFLOPS single precision lattice QCD quark solver on the japanese new supercomputer Fugaku. We achieve a factor 38 time speedup from the supercomputer K on the same problem size, $192^4$, with 102 PFLOPS, 10% floating-point operation efficiency against single precision floating-point operation peak. The evaluation region is the single precision BiCGStab for a Clover-Wilson Dirac matrix with Schwarz Alternating Procedure domain decomposition preconditioning using Jacobi iteration for the local domain matrix inversion.
We present a direct calculation for the first derivative of the isovector nucleon form factors with respect to the momentum transfer $q^2$ using the lower moments of the nucleon 3-point function in the coordinate space. Our numerical simulations are performed using the $N_f = 2 + 1$ nonperturbatively $O(a)$-improved Wilson quark action and Iwasaki gauge action near the physical point, corresponding to the pion mass $M_π=138$ MeV, on a (5.5 fm)$^4$ lattice at a single lattice spacing of $a = 0.085$ fm. In the momentum derivative approach, we can directly evaluate the mean square radii for the electric, magnetic, and axial-vector form factors, and also the magnetic moment without the $q^2$ extrapolation to the zero momentum point. These results are compared with the ones determined by the standard method, where the $q^2$ extrapolations of the corresponding form factors are carried out by fitting models. We find that the new results from the momentum derivative method are obtained with a larger statistical error than the standard method, but with a smaller systematic error associated with the data analysis. Within the total error range of the statistical and systematic errors combined, the two results are in good agreement. On the other hand, two variations of the momentum derivative of the induced pseudoscalar form factor at the zero momentum point show some discrepancy. It seems to be caused by a finite volume effect, since a similar trend is not observed on a large volume, but seen on a small volume in our pilot calculations at a heavier pion mass of $M_π= 510$ MeV. Furthermore, we discuss an equivalence between the momentum derivative method and the similar approach with the point splitting vector current.
A thick-restart Lanczos type algorithm is proposed for Hermitian $J$-symmetric matrices. Since Hermitian $J$-symmetric matrices possess doubly degenerate spectra or doubly multiple eigenvalues with a simple relation between the degenerate eigenvectors, we can improve the convergence of the Lanczos algorithm by restricting the search space of the Krylov subspace to that spanned by one of each pair of the degenerate eigenvector pairs. We show that the Lanczos iteration is compatible with the $J$-symmetry, so that the subspace can be split into two subspaces that are orthogonal to each other. The proposed algorithm searches for eigenvectors in one of the two subspaces without the multiplicity. The other eigenvectors paired to them can be easily reconstructed with the simple relation from the $J$-symmetry. We test our algorithm on randomly generated small dense matrices and a sparse large matrix originating from a quantum field theory.
We present the nucleon form factors and root-mean-square (RMS) radii measured on a (10.8 fm$)^4$ lattice at the physical point. We compute the form factors at small momentum transfer region in $q^2\le 0.102$ GeV$^2$ with the standard plateau method choosing four source-sink separation times $t_{\rm sep}$ from 0.84 to 1.35 fm to examine the possible excited state contamination. We obtain the electric and magnetic form factors and their RMS radii for not only the isovector channel but also the proton and neutron ones without the disconnected diagram. We also obtain the axial-vector coupling and the axial radius from the axial-vector form factor. We find that these three form factors do not show large $t_{\rm sep}$ dependence in our lattice setup. On the other hand, the induced pseudoscalar and pseudoscalar form factors show the clear effects of the excited state contamination, which affect the generalized Goldberger-Treiman relation.
We have evaluated perturbation coefficients of Wilson loops up to $O(g^8)$ for the four-dimensional twisted Eguchi-Kawai model using the numerical stochastic perturbation theory (NSPT) in arXiv:1902.09847. In this talk we present a progress report on the higher order calculation up to $O(g^{63})$, for which we apply a fast Fourier transformation (FFT) based convolution algorithm to the multiplication of polynomial matrices in the NSPT aiming for higher order calculation. We compare two implementations with the CPU-only version and the GPU version of the FFT based convolution algorithm, and find a factor 9 improvement on the computational speed of the NSPT algorithm with SU($N=225$) at $O(g^{31})$. The perturbation order dependence of the computational time, we investigate it up to $O(g^{63})$, shows a mild scaling behavior on the truncation order.
We present results on the axial, scalar and tensor isovector-couplings of the nucleon from 2+1 flavor lattice QCD with physical light quarks ($m_π$ = 135 MeV) in large spatial volume of (10.8 fm)$^3$. The calculations are carried out with the PACS10 gauge configurations generated by the PACS Collaboration with the stout-smeared $\mathcal{O}(a)$ improved Wilson fermions and Iwasaki gauge action at $β=1.82$ corresponding to the lattice spacing of 0.084 fm. For the renormalization, we use the RI/SMOM scheme, a variant of Rome-Southampton RI/MOM scheme with reduced systematic errors, as the intermediate scheme. We then evaluate our final results in the $\overline{\rm MS}$ scheme at a scale of 2 GeV, using the continuum perturbation theory for the matching scale of RI/SMOM and $\overline{\rm MS}$ schemes and running.
We present the results of an exploratory study of the numerical stochastic perturbation theory (NSPT) applied to the four dimensional twisted Eguchi-Kawai (TEK) model. We employ a Kramers type algorithm based on the Generalized Hybrid Molecular Dynamics (GHMD) algorithm. We have computed the perturbative expansion of square Wilson loops up to $O(g^8)$. The results of the first two coefficients (up to $O(g^4)$) have a high precision and match well with the exact values. The next two coefficients can be determined and even extrapolated to large $N$, where they should coincide with the corresponding coefficients for ordinary Yang-Mills theory on an infinite lattice. Our analysis shows the behaviour of the probability distribution for each coefficient tending to Gaussian for larger $N$. The results allow us to establish the requirements to extend this analysis to much higher order.
We port Domain-Decomposed-alpha-AMG solver to the K computer. The system has 8 cores and 16 GB memory per node, of which theoretical peak is 128 GFlops (82,944 nodes in total). Its feature, as many as 256 registers per core and as large as 0.5 byte/Flop ratio, requires a different tuning from other machines. In order to use more registers, we change some of the data structure and rewrite matrix-vector operations with intrinsics. The performance is improved by more than a factor two for twelve solves including the setup. The efficiency is still about 5% after the optimization, which is lower than a previously tuned mixed precision solver for the K computer, 22%. The throughput is, however, more than two times better for a physical point configuration.