arXiv subjects

Noritsugu Sakuma

Publications and source records attributed to Noritsugu Sakuma.

8 recordsLinked to original sources

arXiv2026

Modelling magnetic material properties with uncertainty-aware neural networks

Machine learning is increasingly applied to accelerate the discovery of novel materials by exploring large compositional and structural design spaces. Yet, the scarcity of high-quality data and the frequent need for out-of-distribution prediction introduce substantial uncertainty, making the assessment of model reliability essential. In this work, we investigate uncertainty quantification as a means to evaluate model confidence in the context of permanent magnet research. In a first study, we benchmark classical and modern machine learning models for predicting intrinsic magnetic properties, focusing on the quality of their uncertainty estimates. We apply Gaussian negative log-likelihood loss and dropout-based Bayesian approximation as practical strategies for estimating predictive uncertainty. In a second study, we transfer these architectural features for uncertainty estimation to a more complex task: predicting coercivity from microstructural information using a graph neural network. Together, these studies demonstrate that uncertainty quantification not only enhances the trustworthiness of predictions but is also transferable across different modeling tasks.

cond-mat.mtrl-sci↗

arXiv2025

Impact of Lattice Distortions on Magnetocrystalline Anisotropy and Magnetization in (Nd$_{1-x}$Pr$_x$)$_2$Fe$_{14}$B Alloys

Nd$_{2}$Fe$_{14}$B -- a widely used permanent magnet -- has magnetocrystalline anisotropy constants that differ between the bulk and interface regions. This study explores the effects of lattice distortion on the magnetocrystalline anisotropy ($K_{\rm u}$) and magnetization of (Nd$_{1-x}$Pr$_x$)$_2$Fe$_{14}$B. Nd$_2$Fe$_{14}$B alloys were fabricated; scanning transmission electron microscopy revealed a compressive strain of up to 25% near grain boundaries. Using the full-potential Korringa--Kohn--Rostoker method, we calculated the strain dependence of $K_{\rm u}$, showing that although $K_{\rm u}$ is 4.2 MJ/m$^3$ under strain-free conditions at 0 K, it becomes negative in regions with 25% compressive strain. Additionally, Pr$_{2}$Fe$_{14}$B exhibits a larger $K_{\rm u}$ than Pr$_{2}$Fe$_{14}$B under undistorted conditions, whereas Pr-rich alloys exhibit a more pronounced reduction in $K_{\rm u}$ under strain. These findings highlight the critical influence of lattice distortions on magnetic properties. The calculated strain-dependent magnetic anisotropy parameters provide valuable inputs for future micromagnetic simulations, aiding the design of advanced magnetic materials.

cond-mat.mtrl-sci↗

arXiv2022

Description of collective magnetization processes with machine learning models

This work introduces a latent space method to calculate the demagnetization reversal process of multigrain permanent magnets. The algorithm consists of two deep learning models based on neural networks. The embedded Stoner-Wohlfarth method is used as a reduced order model for computing samples to train and test the machine learning approach. The work is a proof of concept of the method since the used microstructures are simple and the only varying parameters are the magnetic anisotropy axes of the hard magnetic grains. A convolutional autoencoder is used for nonlinear dimensionality reduction, in order to reduce the required amount of training samples for predicting the hysteresis loops. We enriched the loss function with physical information about the underlying problem which increases the accuracy of the machine learning approach. A deep learning regressor is operating in the latent space of the autoencoder. The predictor takes a series of previously encoded magnetization states and outputs the next magnetization state along the hysteresis curve. To predict the complete demagnetization loop, we apply a recursive learning scheme.

physics.comp-ph↗

arXiv2022

Exploring the hysteresis properties of nanocrystalline permanent magnets using deep learning

We demonstrate the use of model order reduction and neural networks for estimating the hysteresis properties of nanocrystalline permanent magnets from microstructure. With a data-driven approach, we learn the demagnetization curve from data-sets created by grain growth and micromagnetic simulations. We show that the granular structure of a magnet can be encoded within a low-dimensional latent space. Latent codes are constructed using a variational autoencoder. The mapping of structure code to hysteresis properties is a multi-target regression problem. We apply deep neural network and use parameter sharing, in order to predict anchor points along the demagnetization curves from the magnet's structure code. The method is applied to study the magnetic properties of nanocrystalline permanent magnets. We show how new grain structures can be generated by interpolation between two points in the latent space and how the magnetic properties of the resulting magnets can be predicted.

cond-mat.mtrl-sci↗

arXiv2021

Conditional physics informed neural networks

We introduce conditional PINNs (physics informed neural networks) for estimating the solution of classes of eigenvalue problems. The concept of PINNs is expanded to learn not only the solution of one particular differential equation but the solutions to a class of problems. We demonstrate this idea by estimating the coercive field of permanent magnets which depends on the width and strength of local defects. When the neural network incorporates the physics of magnetization reversal, training can be achieved in an unsupervised way. There is no need to generate labeled training data. The presented test cases have been rigorously studied in the past. Thus, a detailed and easy comparison with analytical solutions is made. We show that a single deep neural network can learn the solution of partial differential equations for an entire class of problems.

cs.LG↗

arXiv2020

Data Assimilation Method for Experimental and First-Principles Data: Finite-Temperature Magnetization of (Nd,Pr,La,Ce)$_{2}$(Fe,Co,Ni)$_{14}$B

We propose a data-assimilation method for evaluating the finite-temperature magnetization of a permanent magnet over a high-dimensional composition space. Based on a general framework for constructing a predictor from two data sets including missing values, a practical scheme for magnetic materials is formulated in which a small number of experimental data in limited composition space are integrated with a larger number of first-principles calculation data. We apply the scheme to (Nd$_{1-α-β-γ}$Pr$_α$La$_β$Ce$_γ$)$_{2}$(Fe$_{1-δ-ζ}$Co$_δ$Ni$_ζ$)$_{14}$B. The magnetization in the whole $(α, β, γ, δ, ζ)$ space at arbitrary temperature is obtained. It is shown that the Co doping does not enhance the magnetization at low temperatures, whereas the magnetization increases with increasing $δ$ above 320 K.

cond-mat.mtrl-sci↗

arXiv2019

Optimal uni-axial ferromagnetism in (La,Ce)$_2$Fe$_{14}$B for permanent magnets

Prospects for light-rare-earth-based permanent magnet compound R$_{2}$Fe$_{14}$B (R=La$_{1-x}$Ce$_{x}$ with $0 \le x\le 1$) are inspected from first principles referring to the latest experimental data. Ce-rich 2:14:1 compounds come with good structure stability, reasonably good combination of magnetization and magnetic anisotropy, while a drawback lies in the low Curie temperature that is only 120~K above the room temperature at the Ce$_2$Fe$_{14}$B limit. Best compromise is inspected on the basis of ab initio data for (La$_{1-x}$Ce$_{x}$)$_2$Fe$_{14}$B referring to the magnetic properties of the champion magnet compound Nd$_{2}$Fe$_{14}$B and prerequisite conditions imposed by practical utility.

cond-mat.mtrl-sci↗

arXiv2016

Grain-size dependent demagnetizing factors in permanent magnets

The coercive field of permanent magnets decreases with increasing grain size. The grain size dependence of coercivity is explained by a size dependent demagnetizing factor. In Dy free Nd$_2$Fe$_{14}$B magnets the size dependent demagnetizing factor ranges from 0.2 for a grain size of 55 nm to 1.22 for a grain size of 8300 nm. The comparison of experimental data with micromagnetic simulations suggests that the grain size dependence of the coercive field in hard magnets is due to the non-uniform magnetostatic field in polyhedral grains.

cond-mat.mtrl-sci↗