arXiv subjects

Yuki Nagai

Publications and source records attributed to Yuki Nagai.

At least 19 recordsLinked to original sources

arXiv2026

Anomalous vortex shape in a frustrated superconductor hosting chiral multicomponent order parameters

Multicomponent chiral superconductors can host spatially separated superconducting components within a single Abrikosov vortex, leading to unconventional, anisotropic vortex shapes forbidden in conventional single-component superconductors. Here, using spectroscopic scanning tunneling microscopy at 0.3 K, we investigate the mixed state of the spinel superconductor LiTi2O4 (Tc = 13 K). In the low-field regime, where vortex-vortex interactions are minimized and the intrinsic single-vortex shape is preserved, reasonably isolated triangular Abrikosov vortices are directly visualized with high statistical significance. Remarkably, the vortex orientation is locked to crystallographic domains rather than the magnetic-field direction, revealing a hidden chirality selectivity embedded in the zero-field electronic state.

cond-mat.supr-con↗

arXiv2026

Numerical Hints for Dyon Condensation at $θ=2π$ via Wilson-'t Hooft Loops in $SU(2)$ Yang-Mills Theory

Yang-Mills theories at $θ$ and $θ+2π$ are unitarily equivalent, but their $2π$ periodicity has a nontrivial realization. Recent developments in generalized symmetries rigorously prove that confinement vacua at $θ=0$ and $2π$ should belong to different symmetry-protected topological (SPT) states with the $1$-form center symmetry. For its examination, we measure the Wilson-'t Hooft loop operators at $θ=2π$ for the $SU(2)$ Wilson lattice gauge action and discuss their long-distance behaviors. This requires us to identify the gauge topological charge in the presence of defects, and we employ the $1$-form covariant DBW2 gradient flow to smear lattice gauge fields. We find a clear perimeter-law signal for the dyonic Wilson-'t Hooft loop at $θ=2π$, providing numerical evidence in the pure $SU(2)$ Yang-Mills theory for the theoretically expected dyon condensation at $θ=2π$.

hep-lat↗

arXiv2026

AccelNet: Exact backward-compatible acceleration of polynomial angular descriptors through Cartesian moment factorization

We present AccelNet, an exact, backward-compatible method for accelerating existing trained aenet and n2p2 neural-network potentials without retraining. For angular terms with separable one-neighbor weights and a finite polynomial dependence on $\cos θ$, the method exploits their hidden finite-rank structure to replace explicit neighbor-pair loops by one-neighbor Cartesian moments. AccelNet reads models trained with either package and reproduces their descriptors, energies, and analytic forces to floating-point roundoff. We verified this equivalence for H$_2$O and TiO$_2$ models and tested the resulting potentials in LAMMPS molecular-dynamics simulations. The implementation, model-conversion tools, and LAMMPS interfaces are released as open-source software.

cond-mat.mtrl-sci↗

arXiv2026

Target-Distribution-Guided Cross-Functional Fine-Tuning of Machine-Learning Interatomic Potentials

Cross-functional fine-tuning of machine-learning interatomic potentials (MLIPs) is often treated as a relabeling problem, where configurations generated at one density-functional level are relabeled using a higher-fidelity target functional. However, the resulting training data may be drawn from the wrong equilibrium distribution, because the statistical weights of configurations change across exchange--correlation functionals. Here we address this distribution mismatch using a target-distribution-guided workflow based on self-learning hybrid Monte Carlo (SLHMC), in which trial configurations are proposed by a machine-learning potential and accepted or rejected using target-functional density-functional-theory energies. Using rutile TiO$_2$ as a test system, we fine-tune the MACE-MP-0 foundation potential toward PBE, r$^2$SCAN, and HSE06 target functionals. The resulting adapted potentials reproduce target-anchored nearest-neighbor Ti--O distributions, radial distribution functions, and the NPT cell metrics examined here more accurately than the foundation-model and off-target relabeling controls considered in this work. In particular, HSE06-guided fine-tuning improves structural and thermodynamic properties that are difficult to access with direct hybrid-functional molecular dynamics because of the computational cost of exact exchange. These results indicate that target-distribution coverage is an essential component of cross-functional MLIP transfer, and that accurate target-level labels alone may be insufficient when the configurational distribution is mismatched.

cond-mat.mtrl-sci↗

arXiv2026

Parameter Optimization of Domain-Wall Fermion using Machine Learning

We study a parameter optimization of domain-wall fermions to improve chiral symmetry based on machine learning. Domain-wall fermions involve coefficients along the fifth dimension, which can be treated as trainable parameters to reduce the chiral symmetry violation caused by the finite extent of the fifth dimension. As the loss function, we use the residual mass estimated stochastically on a single gauge configuration. Numerical tests on a $L^3\times T\times L_5=4^3\times8\times8$ lattice demonstrate the feasibility of this framework.

hep-lat↗

arXiv2026

Lattice Gauge Theory via LLVM-Level Automatic Differentiation

We enable the automatic construction of Hybrid Monte Carlo (HMC) forces in lattice gauge theory by performing reverse-mode automatic differentiation at the level of optimized LLVM intermediate representation, making the approach applicable to any language that lowers lattice action code to LLVM. In practice, this means that once the action evaluation routine is implemented, the corresponding HMC force can be generated automatically from the same code path, without deriving or maintaining a separate force routine. The method preserves conventional imperative, in-place implementations and enables a single-source workflow in which forces are generated directly from the action code while inheriting compiler optimizations. We perform end-to-end reverse-mode differentiation of both gauge and Wilson fermion actions. For the Wilson fermion case, we find that the force generated by automatic differentiation achieves performance comparable to a conventional hand-written fermion force implementation. The same differentiation pipeline targets both CPU and GPU backends, providing a practical route to performance-portable force construction for compositional lattice actions.

hep-lat↗

arXiv2025

Puzzling Isotonic Odd-Even Staggering of Charge Radii in Deformed Rare Earth Nuclei

The nuclear charge radius is a fundamental observable that encodes key aspects of nuclear structure, deformation, and pairing. Isotonic (constant neutron number) systematics in the deformed rare-earth region have long suggested that odd-$Z$ nuclei are more compact than their even-$Z$ neighbors - except for Lu, whose recommended radius appeared anomalously large relative to Yb and Hf. We report a high-precision determination of the natural-abundance-averaged Lu-Yb charge-radius difference using extreme-ultraviolet spectroscopy of highly charged Na-like and Mg-like ions, supported by high-accuracy relativistic atomic-structure calculations - a recently introduced method with the unique ability to measure inter-element charge radius differences. Combined with muonic-atom and optical isotope-shift data, our result resolves the longstanding Lu inversion anomaly and reestablishes a pronounced odd-even staggering along the $N=94$ isotonic chain. The magnitude of this staggering is unexpectedly large, far exceeding that observed in semi-magic nuclei and in deformed isotopic sequences. State-of-the-art nuclear density functional theory calculations, including quantified uncertainties, fail to reproduce this enhancement, possibly indicating missing structural effects in current models. Our work demonstrates the power of highly charged ions for precise, element-crossing charge-radius measurements and provides stringent new constraints for future theoretical and experimental studies of nuclear-size systematics.

physics.atom-ph↗

arXiv2025

Extreme Ultraviolet Spectroscopy of Highly Charged Lu and Yb Ions for Nuclear Charge Radius Determination

We report a high-precision determination of the natural-abundance-averaged nuclear charge-radius difference between Yb and Lu using extreme ultraviolet (EUV) spectroscopy of highly charged ions (HCIs). By measuring the $D_1$ transition energies in Na- and Mg-like charge states of Lu and Yb confined in the Tokyo electron-beam ion trap, we extract meV-level energy shifts that are directly sensitive to nuclear-size effects. Transition-energy differences obtained from these spectra are compared with state-of-the-art relativistic many-body perturbation theory, including a new treatment of Mg-like ions. We develop a generalized framework to propagate uncertainties arising from nuclear deformation and surface diffuseness and evaluate corresponding nuclear-sensitivity coefficients. Combining Na- and Mg-like results yields mutually consistent radius differences, demonstrating the robustness of both the experimental calibration and the theoretical predictions. To determine absolute isotopic radii, we perform a generalized least-squares optimization incorporating our HCI constraints together with optical-isotope-shift data and muonic-atom results. This analysis establishes that the $^{175}$Lu charge radius is smaller than that of $^{174}$Yb, restoring the expected odd-even staggering across the $N=94$ isotonic chain. Our recommended value, $R(^{175}\text{Lu}) = 5.291(11)$ fm, reduces the uncertainty of the Lu radius by a factor of three compared with the previous electron-scattering result and resolves a long-standing anomaly in rare-earth nuclear systematics. This work demonstrates that EUV spectroscopy of HCIs provides a powerful and broadly applicable method for precision nuclear-structure studies in heavy, deformed nuclei. The techniques developed here enable future investigations of isotonic and isoelectronic sequences, including radioactive nuclides and higher-$Z$ systems.

physics.atom-ph↗

arXiv2025

JuliaQCD: Portable lattice QCD package in Julia language

We develop a new lattice gauge theory code set JuliaQCD using the Julia language. Julia is well-suited for integrating machine learning techniques and enables rapid prototyping and execution of algorithms for four dimensional QCD and other non-Abelian gauge theories. The code leverages LLVM for high-performance execution and supports MPI for parallel computations. Julia's multiple dispatch provides a flexible and intuitive framework for development. The code implements existing algorithms such as Hybrid Monte Carlo (HMC), many color and flavor, supports lattice fermions, smearing techniques, and full QCD simulations. It is designed to run efficiently across various platforms, from laptops to supercomputers, allowing for seamless scalability. The code set is currently available on GitHub https://github.com/JuliaQCD.

hep-lat↗

arXiv2025

CASK: A Gauge Covariant Transformer for Lattice Gauge Theory

We propose a Transformer neural network architecture specifically designed for lattice QCD, focusing on preserving the fundamental symmetries required in lattice gauge theory. The proposed architecture is gauge covariant/equivariant, ensuring it respects gauge symmetry on the lattice, and is also equivariant under spacetime symmetries such as rotations and translations on the lattice. A key feature of our approach lies in the attention matrix, which forms the core of the Transformer architecture. To preserve symmetries, we define the attention matrix using a Frobenius inner product between link variables and extended staples. This construction ensures that the attention matrix remains invariant under gauge transformations, thereby making the entire Transformer architecture covariant. We evaluated the performance of the gauge covariant Transformer in the context of self-learning HMC. Numerical experiments show that the proposed architecture achieves higher performance compared to the gauge covariant neural networks, demonstrating its potential to improve lattice QCD calculations.

hep-lat↗

arXiv2024

Imaging Josephson Vortices on Curved Junctions

Understanding the nature of vortices in type-II superconductors is crucial for comprehending exotic superconductors and advancing the application of superconducting materials in future electronic devices. This study uses spectroscopic scanning tunneling microscopy to visualize Josephson vortices along crystalline domain boundaries in the superconducting spinel oxide LiTi2O4 (LTO). Our experimental results reveal that the local curvature of the Josephson junction dictates the positioning of Josephson vortices. Self-consistent solutions of the Bogoliubov-de Gennes and gap equations theoretically corroborate this observation. In addition to enhancing our understanding of the physics of Josephson vortex formation, this study offers potential guidelines for developing vortex-based superconducting devices.

cond-mat.supr-con↗

arXiv2024

Self-learning path integral hybrid Monte Carlo with mixed ab initio and machine learning potentials for modeling nuclear quantum effects in water

The introduction of machine learned potentials (MLPs) has greatly expanded the space available for studying Nuclear Quantum Effects computationally with ab initio path integral (PI) accuracy, with the MLPs' promise of an accuracy comparable to that of ab initio at a fraction of the cost. One of the challenges in development of MLPs is the need for a large and diverse training set calculated by ab initio methods. This data set should ideally cover the entire phase space, while not searching this space using ab initio methods, as this would be counterproductive and generally intractable with respect to computational time.In this paper, we present the self-learning PI hybrid Monte Carlo Method using a mixed ab initio and ML potential (SL-PIHMC-MIX), where the mixed potential allows for the study of larger systems and the extension of the original SL-HMC method [Nagai et al., Phys. Rev. B 102, 041124 (2020)] to PI methods and larger systems. While the MLPs generated by this method can be directly applied to run long-time ML-PIMD simulations, we demonstrate that using PIHMC-MIX with the trained MLPs allows for an exact reproduction of the structure obtained from ab initio PIMD. Specifically, we find that the PIHMC-MIX simulations require only 5,000 evaluations of the 32-bead structure, compared to the 100,000 evaluations needed for the ab initio PIMD result.

physics.chem-ph↗

arXiv2024

Gauge covariant neural network for quarks and gluons

We propose gauge-covariant neural networks along with a specialized training algorithm for lattice QCD, designed to handle realistic quarks and gluons in four-dimensional space-time. We show that the smearing procedure can be interpreted as an extended version of residual neural networks with fixed parameters. To demonstrate the applicability of our neural networks, we develop a self-learning hybrid Monte Carlo algorithm in the context of two-color QCD, yielding outcomes consistent with those from the conventional Hybrid Monte Carlo approach.

hep-lat↗

arXiv2024

Kolmogorov--Arnold networks in molecular dynamics

We explore the integration of Kolmogorov Networks (KANs) into molecular dynamics (MD) simulations to improve interatomic potentials. We propose that widely used potentials, such as the Lennard-Jones (LJ) potential, the embedded atom model (EAM), and artificial neural network (ANN) potentials, can be interpreted within the KAN framework. Specifically, we demonstrate that the descriptors for ANN potentials, typically constructed using polynomials, can be redefined using KAN's non-linear functions. By employing linear or cubic spline interpolations for these KAN functions, we show that the computational cost of evaluating ANN potentials and their derivatives is reduced.

cond-mat.mtrl-sci↗

arXiv2024

Self-learning Monte Carlo with equivariant Transformer

Machine learning and deep learning have revolutionized computational physics, particularly the simulation of complex systems. Equivariance is essential for simulating physical systems because it imposes a strong inductive bias on the probability distribution described by a machine learning model. However, imposing symmetry on the model can sometimes lead to poor acceptance rates in self-learning Monte Carlo (SLMC). Here, we introduce a symmetry equivariant attention mechanism for SLMC, which can be systematically improved. We evaluate our architecture on a spin-fermion model (\textit{i.e.}, double exchange model) on a two-dimensional lattice. Our results show that the proposed method overcomes the poor acceptance rates of linear models and exhibits a similar scaling law to large language models, with model quality monotonically increasing with the number of layers. Our work paves the way for the development of more accurate and efficient Monte Carlo algorithms with machine learning for simulating complex physical systems.

cond-mat.str-el↗

arXiv2023

Equivariant Transformer is all you need

Machine learning, deep learning, has been accelerating computational physics, which has been used to simulate systems on a lattice. Equivariance is essential to simulate a physical system because it imposes a strong induction bias for the probability distribution described by a machine learning model. This reduces the risk of erroneous extrapolation that deviates from data symmetries and physical laws. However, imposing symmetry on the model sometimes occur a poor acceptance rate in self-learning Monte-Carlo (SLMC). On the other hand, Attention used in Transformers like GPT realizes a large model capacity. We introduce symmetry equivariant attention to SLMC. To evaluate our architecture, we apply it to our proposed new architecture on a spin-fermion model on a two-dimensional lattice. We find that it overcomes poor acceptance rates for linear models and observe the scaling law of the acceptance rate as in the large language models with Transformers.

hep-lat↗

arXiv2023

Atomic diffusion due to hyperatomic fluctuation for quasicrystals

A quasicrystal is an ordered but non-periodic structure understood as a projection from a higher dimensional periodic structure. Some physical properties of quasicrystals are different from those of conventional solids. An anomalous increase in heat capacity at high temperatures has been discussed for over two decades as a manifestation of a hidden high dimensionality of quasicrystals. A plausible candidate for this origin has been phason, which has excitation modes originating from additional degrees of freedom in the higher-dimensional lattice. However, most theoretical studies on phasons have used toy models. A theoretical study of the heat capacity of realistic quasicrystals or their approximants has yet to be conducted because of the huge computational complexity. To bridge this gap between experiment and theory, we show experiments and molecular simulations on the same material, an Al--Pd--Ru quasicrystal, and its approximants. We show that at high temperatures, aluminum atoms diffuse with discontinuous-like jumps, and the diffusion paths of the aluminum can be understood in terms of jumps corresponding to hyperatomic fluctuations in six-dimensional space. It is concluded that the anomaly in the heat capacity of quasicrystals arises from extra degrees of freedom due to hyperatomic fluctuations that play a role in diffusive Nambu--Goldstone modes.

cond-mat.mtrl-sci↗

arXiv2022

sparse-ir: optimal compression and sparse sampling of many-body propagators

We introduce sparse-ir, a collection of libraries to efficiently handle imaginary-time propagators, a central object in finite-temperature quantum many-body calculations. We leverage two concepts: firstly, the intermediate representation (IR), an optimal compression of the propagator with robust a-priori error estimates, and secondly, sparse sampling, near-optimal grids in imaginary time and imaginary frequency from which the propagator can be reconstructed and on which diagrammatic equations can be solved. IR and sparse sampling are packaged into stand-alone, easy-to-use Python, Julia and Fortran libraries, which can readily be included into existing software. We also include an extensive set of sample codes showcasing the library for typical many-body and ab initio methods.

physics.comp-ph↗