Machine-learning-assisted insight into spin ice Dy2Ti2O7

Machine learning for analysis of experimental scattering and spectroscopy data in materials chemistry

The rapid growth of materials chemistry data, driven by advancements in large-scale radiation facilities as well as laboratory instruments, has outpaced conventional data analysis and modelling methods, which can require enormous manual effort. To address this bottleneck, we investigate the application of supervised and unsupervised machine learning (ML) techniques for scattering and spectroscopy data analysis in materials chemistry research. Our perspective focuses on ML applications in powder diffraction (PD), pair distribution function (PDF), small-angle scattering (SAS), inelastic neutron scattering (INS), and X-ray absorption spectroscopy (XAS) data, but the lessons that we learn are generally applicable across materials chemistry. We review the ability of ML to identify physical and structural models and extract information efficiently and accurately from experimental data. Furthermore, we discuss the challenges associated with supervised ML and highlight how unsupervised ML can mitigate these limitations, thus enhancing experimental materials chemistry data analysis. Our perspective emphasises the transformative potential of ML in materials chemistry characterisation and identifies promising directions for future applications. The perspective aims to guide newcomers to ML-based experimental data analysis.

Capturing dynamical correlations using implicit neural representations

Understanding the nature and origin of collective excitations in materials is of fundamental importance for unraveling the underlying physics of a many-body system. Excitation spectra are usually obtained by measuring the dynamical structure factor, S(Q, ω), using inelastic neutron or x-ray scattering techniques and are analyzed by comparing the experimental results against calculated predictions. We introduce a data-driven analysis tool which leverages 'neural implicit representations' that are specifically tailored for handling spectrographic measurements and are able to efficiently obtain unknown parameters from experimental data via automatic differentiation. In this work, we employ linear spin wave theory simulations to train a machine learning platform, enabling precise exchange parameter extraction from inelastic neutron scattering data on the square-lattice spin-1 antiferromagnet La2NiO4, showcasing a viable pathway towards automatic refinement of advanced models for ordered magnetic systems.

Classification of magnetic order from electronic structure by using machine learning

Identifying the magnetic state of materials is of great interest in a wide range of applications, but direct identification is not always straightforward due to limitations in neutron scattering experiments. In this work, we present a machine-learning approach using decision-tree algorithms to identify magnetism from the spin-integrated excitation spectrum, such as the density of states. The dataset was generated by Hartree-Fock mean-field calculations of candidate antiferromagnetic orders on a Wannier Hamiltonian, extracted from first-principle calculations targeting BaOsO[Formula: see text]. Our machine learning model was trained using various types of spectral data, including local density of states, momentum-resolved density of states at high-symmetry points, and the lowest excitation energies from the Fermi level. Although the density of states shows good performance for machine learning, the broadening method had a significant impact on the model's performance. We improved the model's performance by designing the excitation energy as a feature for machine learning, resulting in excellent classification of antiferromagnetic order, even for test samples generated by different methods from the training samples used for machine learning.

DeepStruc: towards structure solution from pair distribution function data using deep generative models

Structure solution of nanostructured materials that have limited long-range order remains a bottleneck in materials development. We present a deep learning algorithm, DeepStruc, that can solve a simple monometallic nanoparticle structure directly from a Pair Distribution Function (PDF) obtained from total scattering data by using a conditional variational autoencoder. We first apply DeepStruc to PDFs from seven different structure types of monometallic nanoparticles, and show that structures can be solved from both simulated and experimental PDFs, including PDFs from nanoparticles that are not present in the training distribution. We also apply DeepStruc to a system of hcp, fcc and stacking faulted nanoparticles, where DeepStruc recognizes stacking faulted nanoparticles as an interpolation between hcp and fcc nanoparticles and is able to solve stacking faulted structures from PDFs. Our findings suggests that DeepStruc is a step towards a general approach for structure solution of nanomaterials.

Dynamical fractal and anomalous noise in a clean magnetic crystal

Fractals-objects with noninteger dimensions-occur in manifold settings and length scales in nature. In this work, we identify an emergent dynamical fractal in a disorder-free, stoichiometric, and three-dimensional magnetic crystal in thermodynamic equilibrium. The phenomenon is born from constraints on the dynamics of the magnetic monopole excitations in spin ice, which restrict them to move on the fractal. This observation explains the anomalous exponent found in magnetic noise experiments in the spin ice compound Dy₂Ti₂O₇, and it resolves a long-standing puzzle about its rapidly diverging relaxation time. The capacity of spin ice to exhibit such notable phenomena suggests that there will be further unexpected discoveries in the cooperative dynamics of even simple topological many-body systems.

Machine-Learning Spectral Indicators of Topology

Topological materials discovery has emerged as an important frontier in condensed matter physics. While theoretical classification frameworks have been used to identify thousands of candidate topological materials, experimental determination of materials' topology often poses significant technical challenges. X-ray absorption spectroscopy (XAS) is a widely used materials characterization technique sensitive to atoms' local symmetry and chemical bonding, which are intimately linked to band topology by the theory of topological quantum chemistry (TQC). Moreover, as a local structural probe, XAS is known to have high quantitative agreement between experiment and calculation, suggesting that insights from computational spectra can effectively inform experiments. In this work, computed X-ray absorption near-edge structure (XANES) spectra of more than 10 000 inorganic materials to train a neural network (NN) classifier that predicts topological class directly from XANES signatures, achieving F₁ scores of 89% and 93% for topological and trivial classes, respectively is leveraged. Given the simplicity of the XAS setup and its compatibility with multimodal sample environments, the proposed machine-learning-augmented XAS topological indicator has the potential to discover broader categories of topological materials, such as non-cleavable compounds and amorphous materials, and may further inform field-driven phenomena in situ, such as magnetic field-driven topological phase transitions.

Machine Learning Methods for Multiscale Physics and Urban Engineering Problems

We present an overview of four challenging research areas in multiscale physics and engineering as well as four data science topics that may be developed for addressing these challenges. We focus on multiscale spatiotemporal problems in light of the importance of understanding the accompanying scientific processes and engineering ideas, where "multiscale" refers to concurrent, non-trivial and coupled models over scales separated by orders of magnitude in either space, time, energy, momenta, or any other relevant parameter. Specifically, we consider problems where the data may be obtained at various resolutions; analyzing such data and constructing coupled models led to open research questions in various applications of data science. Numeric studies are reported for one of the data science techniques discussed here for illustration, namely, on approximate Bayesian computations.

Investigation of the monopole magneto-chemical potential in spin ices using capacitive torque magnetometry

The single-ion anisotropy and magnetic interactions in spin-ice systems give rise to unusual non-collinear spin textures, such as Pauling states and magnetic monopoles. The effective spin correlation strength (J_eff) determines the relative energies of the different spin-ice states. With this work, we display the capability of capacitive torque magnetometry in characterizing the magneto-chemical potential associated with monopole formation. We build a magnetic phase diagram of Ho₂Ti₂O₇, and show that the magneto-chemical potential depends on the spin sublattice (α or β), i.e., the Pauling state, involved in the transition. Monte Carlo simulations using the dipolar-spin-ice Hamiltonian support our findings of a sublattice-dependent magneto-chemical potential, but the model underestimates the J_eff for the β-sublattice. Additional simulations, including next-nearest neighbor interactions (J₂), show that long-range exchange terms in the Hamiltonian are needed to describe the measurements. This demonstrates that torque magnetometry provides a sensitive test for J_eff and the spin-spin interactions that contribute to it.

Magnetic-field induced phase transitions in spin-ice materials have been investigated with various experimental techniques. Here, the authors demonstrate the capability of capacitive torque magnetometry in probing magnetic interaction energies and establishing magnetic phase boundaries in Ho₂Ti₂O₇.

Optimization and inference of bin widths for histogramming inelastic neutron scattering spectra

A data-driven bin-width optimization for the histograms of measured data sets based on inhomogeneous Poisson processes was developed in a neurophysiology study [Shimazaki & Shinomoto (2007). Neural Comput. 19, 1503-1527], and a subsequent study [Muto, Sakamoto, Matsuura, Arima & Okada (2019). J. Phys. Soc. Jpn, 88, 044002] proposed its application to inelastic neutron scattering (INS) data. In the present study, the results of the method on experimental INS time-of-flight data collected under different measurement conditions from a copper single crystal are validated. The extrapolation of the statistics on a given data set to other data sets with different total counts precisely infers the optimal bin widths on the latter. The histograms with the optimized bin widths statistically verify two fine-spectral-feature examples in the energy and momentum transfer cross sections: (i) the existence of phonon band gaps; and (ii) the number of plural phonon branches located close to each other. This indicates that the applied method helps in the efficient and rigorous observation of spectral structures important in physics and materials science like novel forms of magnetic excitation and phonon states correlated to thermal conductivities.

Supervised and unsupervised machine learning of structural phases of polymers adsorbed to nanowires

We identify configurational phases and structural transitions in a polymer nanotube composite by means of machine learning. We employ various unsupervised dimensionality reduction methods, conventional neural networks, as well as the confusion method, an unsupervised neural-network-based approach. We find neural networks are able to reliably recognize all configurational phases that have been found previously in experiment and simulation. Furthermore, we locate the boundaries between configurational phases in a way that removes human intuition or bias. This could be done before only by relying on preconceived, ad hoc order parameters.

Anomalous magnetic noise in an imperfectly flat landscape in the topological magnet Dy2Ti2O7

Noise generated by motion of charge and spin provides a unique window into materials at the atomic scale. From temperature of resistors to electrons breaking into fractional quasiparticles, "listening" to the noise spectrum is a powerful way to decode underlying dynamics. Here, we use ultrasensitive superconducting quantum interference device (SQUIDs) to probe the puzzling noise in a frustrated magnet, the spin-ice compound Dy₂Ti₂O₇ (DTO), revealing cooperative and memory effects. DTO is a topological magnet in three dimensions-characterized by emergent magnetostatics and telltale fractionalized magnetic monopole quasiparticles-whose real-time dynamical properties have been an enigma from the very beginning. We show that DTO exhibits highly anomalous noise spectra, differing significantly from the expected Brownian noise of monopole random walks, in three qualitatively different regimes: equilibrium spin ice, a "frozen" regime extending to ultralow temperatures, and a high-temperature "anomalous" paramagnet. We present several distinct mechanisms that give rise to varied colored noise spectra. In addition, we identify the structure of the local spin-flip dynamics as a crucial ingredient for any modeling. Thus, the dynamics of spin ice reflects the interplay of local dynamics with emergent topological degrees of freedom and a frustration-generated imperfectly flat energy landscape, and as such, it points to intriguing cooperative and memory effects for a broad class of magnetic materials.

Machine learning for magnetic phase diagrams and inverse scattering problems

Machine learning promises to deliver powerful new approaches to neutron scattering from magnetic materials. Large scale simulations provide the means to realise this with approaches including spin-wave, Landau Lifshitz, and Monte Carlo methods. These approaches are shown to be effective at simulating magnetic structures and dynamics in a wide range of materials. Using large numbers of simulations the effectiveness of machine learning approaches are assessed. Principal component analysis and nonlinear autoencoders are considered with the latter found to provide a high degree of compression and to be highly suited to neutron scattering problems. Agglomerative heirarchical clustering in the latent space is shown to be effective at extracting phase diagrams of behavior and features in an automated way that aid understanding and interpretation. The autoencoders are also well suited to optimizing model parameters and were found to be highly advantageous over conventional fitting approaches including being tolerant of artifacts in untreated data. The potential of machine learning to automate complex data analysis tasks including the inversion of neutron scattering data into models and the processing of large volumes of multidimensional data is assessed. Directions for future developments are considered and machine learning argued to have high potential for impact on neutron science generally.

Machine learning for neutron reflectometry data analysis of two-layer thin films *

Neutron reflectometry (NR) is a powerful tool for probing thin films at length scales down to nanometers. We investigated the use of a neural network to predict a two-layer thin film structure to model a given measured reflectivity curve. Application of this neural network to predict a thin film structure revealed that it was accurate and could provide an excellent starting point for traditional fitting methods. Employing prediction-guided fitting has considerable potential for more rapidly producing a result compared to the labor-intensive but commonly-used approach of trial and error searches prior to refinement. A deeper look at the stability of the predictive power of the neural network against statistical fluctuations of measured reflectivity profiles showed that the predictions are stable. We conclude that the approach presented here can provide valuable assistance to users of NR and should be further extended for use in studies of more complex n-layer thin film systems. This result also opens up the possibility of developing adaptive measurement systems in the future.

Unsupervised machine learning of topological phase transitions from experimental data

Identifying phase transitions is one of the key challenges in quantum many-body physics. Recently, machine learning methods have been shown to be an alternative way of localising phase boundaries from noisy and imperfect data without the knowledge of the order parameter. Here, we apply different unsupervised machine learning techniques, including anomaly detection and influence functions, to experimental data from ultracold atoms. In this way, we obtain the topological phase diagram of the Haldane model in a completely unbiased fashion. We show that these methods can successfully be applied to experimental data at finite temperatures and to the data of Floquet systems when post-processing the data to a single micromotion phase. Our work provides a benchmark for the unsupervised detection of new exotic phases in complex many-body systems.

Principal component analysis of diffuse magnetic neutron scattering: a theoretical study

We present a theoretical study of the potential of principal component (PC) analysis to analyse magnetic diffuse neutron scattering data on quantum materials. To address this question, we simulate the scattering functionSqfor a model describing a cluster magnet with anisotropic spin-spin interactions under different conditions of applied field and temperature. We find high dimensionality reduction and that the algorithm can be trained with surprisingly small numbers of simulated observations. Subsequently, observations can be projected onto the reduced-dimensionality space defined by the learnt PCs. Constant-field temperature scans correspond to trajectories in this space which show characteristic bifurcations at the critical fields corresponding to ground-state phase boundaries. Such plots allow the ground-state phase diagram to be accurately determined from finite-temperature measurements.

Thermal expansion and phase transformation in the rare earth di-titanate (R2Ti2O7) system

Characterization of the thermal expansion in the rare earth di-titanates is important for their use in high-temperature structural and dielectric applications. Powder samples of the rare earth di-titanates R₂Ti₂O₇ (or R₂O₃·2TiO₂), where R = La, Pr, Nd, Sm, Gd, Dy, Er, Yb, Y, which crystallize in either the monoclinic or cubic phases, were synthesized for the first time by the solution-based steric entrapment method. The three-dimensional thermal expansions of these polycrystalline powder samples were measured by in situ synchrotron powder diffraction from 25°C to 1600°C in air, nearly 600°C higher than other in situ thermal expansion studies. The high temperatures in synchrotron experiments were achieved with a quadrupole lamp furnace. Neutron powder diffraction measured the monoclinic phases from 25°C to 1150°C. The La₂Ti₂O₇ member of the rare earth di-titanates undergoes a monoclinic to orthorhombic displacive transition on heating, as shown by synchrotron diffraction in air at 885°C (864°C-904°C) and neutron diffraction at 874°C (841°C-894°C).

Machine learning for neutron scattering at ORNL*

Machine learning (ML) offers exciting new opportunities to extract more information from scattering data. At neutron scattering user facilities, ML has the potential to help accelerate scientific productivity by empowering facility users with insight into their data which has traditionally been supplied by scattering experts. Such support can help in both speeding up common modeling problems for users, as well as help solve harder problems that are normally time consuming and difficult to address with standard methods. This article explores the recent ML work undertaken at Oak Ridge National Laboratory involving neutron scattering data. We cover materials structure modeling for diffuse scattering, powder diffraction, and small-angle scattering. We also discuss how ML can help to model the response of the instrument more precisely, as well as enable quick extraction of information from neutron data. The application of super-resolution techniques to small-angle scattering and peak extraction for diffraction will be discussed.

Interpretable, calibrated neural networks for analysis and understanding of inelastic neutron scattering data

Deep neural networks (NNs) provide flexible frameworks for learning data representations and functions relating data to other properties and are often claimed to achieve 'super-human' performance in inferring relationships between input data and desired property. In the context of inelastic neutron scattering experiments, however, as in many other scientific scenarios, a number of issues arise: (i) scarcity of labelled experimental data, (ii) lack of uncertainty quantification on results, and (iii) lack of interpretability of the deep NNs. In this work we examine approaches to all three issues. We use simulated data to train a deep NN to distinguish between two possible magnetic exchange models of a half-doped manganite. We apply the recently developed deterministic uncertainty quantification method to provide error estimates for the classification, demonstrating in the process how important realistic representations of instrument resolution in the training data are for reliable estimates on experimental data. Finally we use class activation maps to determine which regions of the spectra are most important for the final classification result reached by the network.

Scattering Signatures of Bond-Dependent Magnetic Interactions

Bond-dependent magnetic interactions can generate exotic phases such as Kitaev spin-liquid states. Experimentally determining the values of bond-dependent interactions is a challenging but crucial problem. Here, I show that each symmetry-allowed nearest-neighbor interaction on triangular and honeycomb lattices has a distinct signature in paramagnetic neutron-diffraction data, and that such data contain sufficient information to determine the spin Hamiltonian unambiguously via unconstrained fits. Moreover, I show that bond-dependent interactions can often be extracted from powder-averaged data. These results facilitate experimental determination of spin Hamiltonians for materials that do not show conventional magnetic ordering.

Machine learning and big scientific data

This paper reviews some of the challenges posed by the huge growth of experimental data generated by the new generation of large-scale experiments at UK national facilities at the Rutherford Appleton Laboratory (RAL) site at Harwell near Oxford. Such 'Big Scientific Data' comes from the Diamond Light Source and Electron Microscopy Facilities, the ISIS Neutron and Muon Facility and the UK's Central Laser Facility. Increasingly, scientists are now required to use advanced machine learning and other AI technologies both to automate parts of the data pipeline and to help find new scientific discoveries in the analysis of their data. For commercially important applications, such as object recognition, natural language processing and automatic translation, deep learning has made dramatic breakthroughs. Google's DeepMind has now used the deep learning technology to develop their AlphaFold tool to make predictions for protein folding. Remarkably, it has been able to achieve some spectacular results for this specific scientific problem. Can deep learning be similarly transformative for other scientific problems? After a brief review of some initial applications of machine learning at the RAL, we focus on challenges and opportunities for AI in advancing materials science. Finally, we discuss the importance of developing some realistic machine learning benchmarks using Big Scientific Data coming from several different scientific domains. We conclude with some initial examples of our 'scientific machine learning' benchmark suite and of the research challenges these benchmarks will enable. This article is part of a discussion meeting issue 'Numerical algorithms for high-performance computational science'.