SearcharxivSearch

arXiv subjects

Yusuke Namekawa

Publications and source records attributed to Yusuke Namekawa.

At least 19 recordsLinked to original sources

Path optimization method for the sign problem: Insights from random matrix models

The path optimization method is applied to the Stephanov model and the chiral random matrix model, both of which share several properties with QCD, to mitigate the sign problem caused by the fermion determinant. The Stephanov model serves as a prototypical model of finite-density QCD, while the chiral random matrix model represents an ideal system featuring the Silver Blaze phenomenon. We show that the path optimization successfully improves the average phase factor in the Stephanov model at high chemical potential, reproducing the analytical results with reduced statistical errors. However, it fails to improve the average phase factor in the Stephanov model at low chemical potential, as well as in the chiral random matrix model. This tendency in the phase factor behavior seems to be closely related to the global sign problem.

hep-lat

Analyzing the two-dimensional doped Hubbard model with the Worldvolume HMC method

We apply the Worldvolume Hybrid Monte Carlo (WV-HMC) method [arXiv:2012.08468] to the two-dimensional Hubbard model, which is known to suffer from a severe sign problem when the system is doped (away from half filling). We show that the method predicts physical observables with controlled statistical errors on an $8 \times 8$ lattice at temperature $T/t = 1/6.4 \approx 0.156$ and interaction strength $U/t = 8.0$ ($t$ is the hopping amplitude), for which the standard determinant quantum Monte Carlo fails.

hep-lat

Mixed precision solvers with half-precision floating point numbers for Lattice QCD on A64FX processor

We investigate the use of half-precision floating-point numbers (FP16) in mixed-precision linear solvers for lattice QCD simulations. Since the emergence of GPUs for general-purpose, mixed-precision algorithms that combine single-precision (FP32) with double-precision (FP64) arithmetics have become widely used in this field and others. While FP32-based methods are now well established, we examine the practicality of using FP16. In this work, we introduce rescaling steps in both the outer iterative refinement step and the inner BiCGStab solver to avoid numerical instability. In our experiments with a simple Wilson kernel, the solver shows improved stability, and the additional iteration count compared to the FP64 version remains within 20\%, indicating that the FP16 version is practical for use. We believe that the proposed rescaling methods can also benefit other mixed precision preconditioners in avoiding underflows.

hep-lat

Enhancing the ergodicity of Worldvolume HMC via embedding generalized thimble HMC

The Worldvolume Hybrid Monte Carlo (WV-HMC) method [arXiv:2012.08468] is an efficient and versatile algorithm that mitigates the sign problem while resolving the ergodicity issues inherent in Lefschetz-thimble approaches. We focus on cases where the maximum flow time can be kept small, such as when applying WV-HMC to the doped Hubbard model utilizing a redundant, nonphysical parameter. An optimal choice of this parameter significantly reduces the sign problem on the original integration surface. This allows for small flow times, thereby enabling the simulation of larger system sizes at a modest computational cost. However, when the worldvolume reduces to a thin layer, phase-space exploration becomes inefficient, and ergodicity problems may reemerge. To address this limitation in WV-HMC, we propose embedding generalized thimble HMC (GT-HMC) into the WV-HMC framework. GT-HMC performs updates on a single deformed surface at a fixed flow time. Despite its inherent ergodicity issues at the zeros of the Boltzmann weight, GT-HMC efficiently explores the allowed region and typically permits larger molecular dynamics step sizes than WV-HMC. Consequently, it is highly effective in regions where ergodicity issues are less severe. We prove that GT-HMC can be consistently embedded within WV-HMC and confirm that the standalone and combined algorithms agree within statistical errors for the two-dimensional doped Hubbard model on an $8 \times 8$ lattice. This combined algorithm enables simulations on larger spacetime lattices. We demonstrate the feasibility of this approach by extrapolating the number and energy densities to the zero Trotter step limit at fixed temperature $T/t = 1/6.4\simeq 0.156$ and repulsive interaction $U/t = 8.0$. Even with modest sample sizes, we achieve controlled statistical errors across the entire range of the chemical potential.

cond-mat.str-el

Applying the Worldvolume Hybrid Monte Carlo method to the Hubbard model away from half filling

The Worldvolume Hybrid Monte Carlo (WV-HMC) method [arXiv:2012.08468] is an efficient algorithm for addressing the numerical sign problem at moderate computational cost. It mitigates the sign problem while avoiding the ergodicity issues inherent in approaches based on Lefschetz thimbles. In this study, we apply WV-HMC to the two-dimensional Hubbard model doped away from half filling, which is known to suffer from a severe sign problem. We compute the number density and the energy density on lattices of size $6 \times 6$ and $8 \times 8$ at temperature $T/t = 1/6.4 \simeq 0.156$ and interaction strength $U/t = 8.0$, using Trotter number $N_t = 20$ (Trotter step $\epsilon = 0.32$). Our results demonstrate that WV-HMC remains effective even in parameter regimes where standard (non-thimble) determinant quantum Monte Carlo methods fail. In this work, fermion matrix inversions are performed using direct solvers, leading to a computational cost of $O(N^3)$, where $N$ denotes the number of degrees of freedom and is proportional to the spacetime lattice volume. An alternative algorithm employing pseudofermions and iterative solvers, which reduces the cost to $O(N^2)$ at the expense of careful parameter tuning, will be discussed in a separate publication.

cond-mat.str-el

On the validity of the complex Langevin method near the deconfining phase transition in QCD at finite density

In our previous paper [JHEP 10 (2020) 144], we found that the complex Langevin (CL) method works for QCD at finite density on the $16^3 \times 32$ lattice in the low-temperature high-density regime within the range $\mu / T = 1.6 - 9.6$ with $\mu$ and $T$ being the quark chemical potential and the temperature, which enabled us to see a clear trend towards the formation of the Fermi sphere. Here we investigate the validity of the CL method on the $24^3 \times 12$ lattice in the deconfined phase near the deconfinement phase transition. As before, we use four-flavor staggered fermions and judge the validity using the criterion based on the probability distribution of the drift term. The spatial extent is $L = (1.3 - 2.7 {\rm ~fm} )> \Lambda_{\rm LQCD}^{-1} \sim 1{\rm ~fm}$, in contrast to our previous study with $L < \Lambda_{\rm LQCD}^{-1}$. We find that the CL method works in a broad region up to $\mu / T = 4.8$, while it starts to fail as we approach the phase boundary due to the singular drift problem, which can be understood qualitatively by extending the Banks-Casher relation to the case at finite density.

hep-lat

Path optimization method for the sign problem caused by fermion determinant

The path optimization method with machine learning is applied to the one-dimensional massive lattice Thirring model, which has the sign problem caused by the fermion determinant. This study aims to investigate how the path optimization method works for the sign problem. We show that the path optimization method successfully reduces statistical errors and reproduces the analytic results. We also examine an approximation of the Jacobian calculation in the learning process and show that it gives consistent results with those without an approximation.

hep-lat

Update of kaon semileptonic form factor using $N_f=2+1$ PACS10 configurations

We calculate the form factors for the kaon semileptonic decay process using the PACS10 configurations, whose physical volume is more than (10 fm)$^4$ very close to the physical point. The configurations were generated with the Iwasaki gauge action and $N_f=2+1$ stout-smeared nonperturbatively $O(a)$-improved Wilson quark action at the three lattice spacings, 0.085, 0.063, and 0.041 fm. We present updated results for the form factors, and discuss their continuum extrapolations, momentum transfer interpolation, and short chiral extrapolation to tune the simulated pion and kaon masses to the physical ones. From the results with various analyses, the systematic error of the form factor at the zero momentum transfer is estimated. The value of $|V_{us}|$ is determined using our result, and is compared with those using the previous calculations and also those determined through the kaon leptonic decay process.

hep-lat

$|V_{us}|$ from kaon semileptonic form factor in $N_f = 2+1$ QCD at the physical point on (10 fm)$^4$

We present a preliminary result of the kaon semileptonic form factor calculated at the smallest lattice spacing in the PACS10 configurations, whose physical volumes are more than (10 fm)$^4$ at the physical point. The configurations were generated using the Iwasaki gauge action and $N_f=2+1$ stout-smeared nonperturbatively $O(a)$ improved Wilson quark action at the three lattice spacings, 0.085, 0.063, and 0.041 fm. The value of $|V_{us}|$ in the continuum limit is estimated from our results including the preliminary one. We compare our result of $|V_{us}|$ with the previous results and those through the kaon leptonic decay.

hep-lat

Application of the path optimization method to a discrete spin system

The path optimization method, which is proposed to control the sign problem in quantum field theories with continuous degrees of freedom by machine learning, is applied to a spin model with discrete degrees of freedom. The path optimization method is applied by replacing the spins with dynamical variables via the Hubbard-Stratonovich transformation, and the sum with the integral. The one-dimensional (Lenz-)Ising model with a complex coupling constant is used as a laboratory for the sign problem in the spin model. The average phase factor is enhanced by the path optimization method, indicating that the method can weaken the sign problem. Our result reproduces the analytic values with controlled statistical errors.

hep-lat

Bridge++ 2.0: Benchmark results on supercomputer Fugaku

Bridge++ is a general-purpose code set for lattice QCD simulations aiming at a readable, extensible, and portable code while keeping practical high performance. The new version 2.0 employs machine-dependent optimization, enabling flexible data layout in float/double precision, while it was fixed layout and only with the double precision in previous versions. We report the performance on supercomputer Fugaku with Arm A64FX-SVE architecture by Fujitsu.

hep-lat

Color superconductivity on the lattice -- analytic predictions from QCD in a small box

We investigate color superconductivity on the lattice using the gap equation for the Cooper pair condensate. The weak coupling analysis is justified by choosing the physical size of the lattice to be smaller than the QCD scale, while keeping the aspect ratio of the lattice small enough to suppress thermal excitations. In the vicinity of the critical coupling constant that separates the superconducting phase and the normal phase, the gap equation can be linearized, and by solving the corresponding eigenvalue problem, we obtain the critical point and the Cooper pair condensate without assuming its explicit form. The momentum components of the condensate suggest spatially isotropic s-wave superconductivity with Cooper pairs formed by quarks near the Fermi surface. The chiral symmetry in the massless limit is spontaneously broken by the Cooper pair condensate, which turns out to be dominated by the scalar and the pseudo-scalar components. Our results provide useful predictions, in particular, for future lattice simulations based on methods to overcome the sign problem such as the complex Langevin method.

hep-lat

Improving efficiency of the path optimization method for a gauge theory

We investigate efficiency of a gauge-covariant neural network and an approximation of the Jacobian in optimizing the complexified integration path toward evading the sign problem in lattice field theories. For the construction of the complexified integration path, we employ the path optimization method. The $2$-dimensional $\text{U}(1)$ gauge theory with the complex gauge coupling constant is used as a laboratory to evaluate the efficiency. It is found that the gauge-covariant neural network, which is composed of the Stout-like smearing, can enhance the average phase factor, as the gauge-invariant input does. For the approximation of the Jacobian, we test the most drastic case in which we perfectly drop the Jacobian during the learning process. It reduces the numerical cost of the Jacobian calculation from ${\cal O}(N^3)$ to ${\cal O}(1)$, where $N$ means the number of degrees of freedom of the theory. The path optimization using this Jacobian approximation still enhances the average phase factor at expense of a slight increase of the statistical error.

hep-lat

Momentum transfer dependence of kaon semileptonic form factor on (10 fm)$^4$ at the physical point

We calculate the kaon semileptonic form factors using the two sets of the PACS10 configuration, whose physical volumes are more than (10 fm)$^4$ at the physical point. The lattice spacings are 0.063 and 0.085 fm. The configurations were generated using the Iwasaki gauge action and $N_f=2+1$ stout-smeared nonperturbatively $O(a)$-improved Wilson quark action. From the momentum transfer dependence of the form factors, we evaluate the slope and curvature for the form factors at the zero momentum transfer. Furthermore, we calculate the phase space factor, which is used to obtain $|V_{us}|$ through the kaon semileptonic decay. These results are compared with previous lattice results and experimental values.

hep-lat

$K_{\ell 3}$ form factors at the physical point: Toward the continuum limit

We present updated results for the form factors of the kaon semileptonic $(K_{\ell 3})$ decay process calculated with $N_f = 2 + 1$ nonperturbatively $O(a)$-improved Wilson quark action and Iwasaki gauge action at the physical point on large volumes of more than (10 fm)$^4$. In addition to our previous calculation at the lattice spacing $a = 0.085$ fm, we perform a calculation at the second lattice spacing of $0.063$ fm. Using the results for the form factors extracted from 3-point functions with the local and also conserved vector currents at the two lattice spacings, continuum extrapolation and interpolation of the momentum transfer are carried out simultaneously to obtain the value of the form factor $f_+(0)$ at the zero momentum transfer in the continuum limit. After investigation of stability of $f_+(0)$ against several fit forms and different data, we obtain $f_+(0) = 0.9615(10)(^{+47}_{\ -3})(5)$, where the first, second, and third errors are statistical, systematic errors from choice of the fit forms and isospin breaking effect, respectively. Combining our value of $f_+(0)$ and experimental input of the $K_{\ell 3}$ decay, one of the Cabibbo-Kobayashi-Maskawa matrix elements $|V_{us}|$ is determined as $|V_{us}| = 0.2252(^{\ +5}_{-12})$, whose error contains the experimental one as well as that in the lattice calculation. This value is reasonably consistent with the ones determined from recent lattice QCD results of $f_+(0)$ and also the one determined through the kaon leptonic decay process. We observe some tension between our value and $|V_{us}|$ evaluated from the unitarity of the CKM matrix with $|V_{ud}|$, while it depends on the size of the error of $|V_{ud}|$. It is also found that $|V_{us}|$ determined with our phase space integrals through six $K_{\ell 3}$ decay processes is consistent with the above one using $f_+(0)$.

hep-lat

Numerical sign problem and the tempered Lefschetz thimble method

The numerical sign problem is a major obstacle to the quantitative understanding of many important physical systems with first-principles calculations. Typical examples for such systems include finite-density QCD, strongly-correlated electron systems and frustrated spin systems, as well as the real-time dynamics of quantum systems. In this talk, we argue that the "tempered Lefschetz thimble method" (TLTM) [M. Fukuma and N. Umeda, arXiv:1703.00861] and its extension, the "worldvolume tempered Lefschetz thimble method" (WV-TLTM) [M. Fukuma and N. Matsumoto, arXiv:2012.08468], may be a reliable and versatile solution to the sign problem. We demonstrate the effectiveness of the algorithm by exemplifying a successful application of WV-TLTM to the Stephanov model, which is an important toy model of finite-density QCD. We also discuss the computational scaling of WV-TLTM.

hep-lat

Gauge invariant input to neural network for path optimization method

We investigate the efficiency of a gauge invariant input to a neural network for the path optimization method. While the path optimization with a completely gauge-fixed link-variable input has successfully tamed the sign problem in a simple gauge theory, the optimization does not work well when the gauge degrees of freedom remain. We propose to employ a gauge invariant input, such as plaquette, to overcome this problem. The efficiency of the gauge invariant input to the neural network is evaluated for the 2-dimensional $U(1)$ gauge theory with a complex coupling. The average phase factor is significantly enhanced by the path optimization with the plaquette input, indicating good control of the sign problem. It opens a possibility that the path optimization is available to complicated gauge theories, including Quantum Chromodynamics, in a realistic setup.

hep-lat

Color superconductivity in a small box: a complex Langevin study

It is expected that the color superconductivity (CSC) phase appears in QCD at low temperature and high density. On the basis of the lattice perturbation theory, a possible parameter region in which the CSC occurs has been predicted. In this work, we perform complex Langevin simulation on an $8^3\times 128$ lattice using four-flavor staggered fermions. We find, in particular, that the quark number has plateaux with respect to the chemical potential similar to our previous study, indicating the formation of the Fermi sphere. A diquark-antidiquark operator, which is an order parameter of color superconductivity, is formulated on the lattice using the U(1) noise. Our result for this operator is found to fluctuate violently when the Fermi surface coincides with the energy levels of quarks. We also discuss partial restoration of the chiral symmetry at high density.

hep-lat