Search

The presence of grain boundaries (GBs) has a great impact on the coefficient of thermal expansion (CTE) of polycrystals. However, direct measurement of local expansion of GBs remains challenging for conventional methods due to the lack of spatial resolution. In this work, we utilized the valence electron energy loss spectroscopy (EELS) in a scanning transmission electron microscope (STEM) to directly measure the CTE of Σ5 and 45°GBs of SrTiO3 at a temperature range between 373 and 973 K. A CTE that was about 3 times larger was observed in Σ5 GB along the direction normal to GB plane, while only a 1.4 time enhancement was found in the 45° GB. Our result provides direct evidence that GBs contribute to the enhancement of CTE in polycrystals. Also, this work has revealed how thermodynamic properties are varied in different GB structures and demonstrated the potential of EELS for probing local thermal properties with nanometer-scale resolution.

In this study, we propose a physically informed transfer learning approach for materials informatics (MI) using a quantum deep descriptor (QDD) obtained from the quantum deep field (QDF). The QDF is a machine learning model based on density functional theory (DFT) and can be trained with a large database of molecular properties. The pre-trained QDF model can provide an effective molecular descriptor that encodes the fundamental quantum-chemical characteristics (i.e., the wave function or orbital, electron density, and energies of a molecule) learned from the large database; we refer to this descriptor as a QDD. We show that a QDD pre-trained with certain properties of small molecules can predict different properties (e.g., the band gap and dielectric constant) of polymers compared with some existing descriptors. We believe that our DFT-based, physically informed transfer learning approach will not only be useful for practical applications in MI but will also provide quantum-chemical insights into materials in the future. All codes used in this study are available at https://github.com/masashitsubaki.

Graphite intercalation compounds (GICs) are formed by inserting various atoms and molecules between the layers of graphite. Among these GICs, five structures have been reported wherein a single atom M was intercalated into graphite with chemical formulae MC6 and MC8 and different stacking orders. The formation energies of the GICs with these crystal structures and various intercalants (M) were investigated using first-principles calculation. Consequently, the formation energies of all the GICs reported to be synthesized were negative, indicating a good agreement with the experimental results. Our study also provides insights into the relative stability required for synthesizing GICs. Furthermore, we determined the negative formation energies of GICs for M =Sc, Y, Fr, and Ra, which are yet to be reported.

Real-space observations in the middle stage of spinodal decomposition have not been achieved because of the small sizes and unclear interfaces of the phases. Here, we performed in situ real-space experiments with heating to quantitatively reveal the local dynamics and structures of a calcium aluminosilicate glass in the middle stage of spinodal decomposition by scanning transmission electron microscopy. This glass separate into CaO–Al2O3–SiO2 phase and SiO2 phases. The observations revealed that the CaO–Al2O3–SiO2 phases with low Ca concentrations behaved as if they were in the middle stage, whereas the CaO–Al2O3–SiO2 phases with high Ca concentrations coarsened as if they were in the final stage. This coexistence of stages suggested that the behaviors of the phases depend not only on time, but also on their local relative concentrations. Differing from the well-known final stage of spinodal decomposition, the phases with relatively high Ca concentrations in the middle stage showed peculiar directional migration guided by the phases with relatively low Ca concentrations. These findings have not been observed before and thus have the potential to provide a new way to control the phase-separated structure.

Excited states are different quantum states from their ground states, and spectroscopy methods that can assess excited states are widely used in materials characterization. Understanding the spectra reflecting excited states is thus of great importance for materials science. However, understanding such spectra remains difficult because excited states have usually different atomic or electronic configurations from their corresponding ground states. If excited states could be predicted from ground states, the knowledge of the excited states would be improved. Here, we used an artificial neural network to predict the excited states of the core-electron absorption spectra from their ground states. Consequently, our model correctly learned and predicted the excited states from their ground states, providing several thousand times computational efficiency. Furthermore, it showed excellent transferability to other materials. Also, we found two physical insights about excited states: core-hole effects of amorphous silicon oxides are stronger than those of crystalline silicon oxides, and the excited-ground states relationships of some metal oxides are similar to those of the silicon oxides, which could not be obtained by conventional spectral simulation nor found until using machine leaning.

Understanding atomic arrangement in network glass is crucial to develop new glassy materials for future applications in medicine, optics, and electronics. While nuclear magnetic resonance (NMR) and...

Artificial neural networks are applied to quantify the properties of organic molecules by introducing a new descriptor, a core‐loss spectrum, which is typically observed experimentally using electron or X‐ray spectroscopy. Using the calculated C K‐edge core‐loss spectra of organic molecules as the descriptor, the neural network models quantitatively predict both intensive and extensive properties, such as the gap between highest occupied molecular orbital (HOMO) and lowest unoccupied molecular orbital (LUMO) (HOMO–LUMO gap) and internal energy. The prediction accuracy estimated by the mean absolute errors for the HOMO–LUMO gap and internal energy is 0.205 and 97.3 eV, respectively, which are comparable with those of previously reported chemical descriptors. This study indicates that the neural network approach using the core‐loss spectra as the descriptor has the potential to deconvolute the abundant information available in core‐loss spectra for both prediction and experimental characterization of many physical properties. The study shows the practical potential of machine‐learning‐based material property measurements taking advantage of experimental core‐loss spectra, which can be measured with high sensitivity, high spatial resolution, and high temporal resolution.

The presence of grain boundaries (GBs) has a great impact on the coefficient of thermal expansion (CTE) of polycrystals. However, direct measurement of local expansion of GBs remains challenging for conventional methods due to the lack of spatial resolution. In this work, we utilized the valence electron energy loss spectroscopy (EELS) in a scanning transmission electron microscope (STEM) to directly measure the CTE of Σ5 and 45°GBs of SrTiO

One of the important issues of studying grain boundaries (GBs) which has recently attracted increasing interests is to investigate the phase behavior of GBs that one GB with determined disorientation and plane orientation (known as macroscopic parameters) can exist as distinct phases and perform phase transition. While such an issue has been investigated in fcc and bcc metals, GB phases in other elemental materials have not been reported. This work by applying molecular dynamics (MD) simulation explored totally around 7000 meta-stable GB phases of the ∑9(22‾1‾) symmetric tilt GB of silicon, germanium and diamond carbon as diamond-structured elemental materials. Meta-stable phases commonly exist in different elements were discovered and some of them were successfully verified to be reasonable by first-principle simulation. The verified meta-stable GBs were subsequently proved to have different capability to transform to the ground-stable GB at elevated temperature under MD simulation and to perform different pre-melting behaviors. We discovered a bi-directional structural reconstruction mechanism of the unit structure belonging to one of the verified meta-stable phases, by which the unit structures can transform to identical unit structures of the ground-stable GB which can present ‘opposite orientation’. Through computing the kinetic barriers by nudged-elastic-band and annealing simulation using MD, the integral behavior of the unit structures’ reconstruction is found to be a first-order like phase transition. Our work extended the research on GB phases from metals to diamond-structured materials and discovered a new GB phase transition mechanism which has not been reported before.

Aberration-corrected scanning transmission electron microscopy (STEM) is widely used for atomic-level imaging of materials but severely requires damage-free and thin samples (lamellae). So far, the preparation of the high-quality lamella from a bulk largely depends on manual processes by a skilled operator. This limits the throughput and repeatability of aberration-corrected STEM experiments. Here, inspired by the recent successes of “robot scientists”, we demonstrate robotic fabrication of high-quality lamellae by focused-ion-beam (FIB) with automation software. First, we show that the robotic FIB can prepare lamellae with a high success rate, where the FIB system automatically controls rough-milling, lift-out, and final-thinning processes. Then, we systematically optimized the FIB parameters of the final-thinning process for single crystal Si. The optimized Si lamellae were evaluated by aberration-corrected STEM, showing atomic-level images with 55 pm resolution and quantitative repeatability of the spatial resolution and lamella thickness. We also demonstrate robotic fabrication of high-quality lamellae of SrTiO3 and sapphire, suggesting that the robotic FIB system may be applicable for a wide range of materials. The throughput of the robotic fabrication was typically an hour per lamella. Our robotic FIB will pave the way for the operator-free, high-throughput, and repeatable fabrication of the high-quality lamellae for aberration-corrected STEM.

Graphite is used as an anode material in conventional lithium-ion batteries owing to its ability to form stable Li-intercalated graphite intercalation compounds (Li GICs). Its application to sodium-ion batteries has long been of great interest, but the instability of Na GICs hampers its implementation. First-principles calculations were performed to gain physical insight into the intercalation process in alkali-metal (AM) GICs, where $\mathrm{AM}=\mathrm{Li}$, Na. In this study, the structure, stability, and diffusion properties of AM GICs with various in-plane AM concentrations were systematically investigated using a van der Waals density functional simulation, and the differences between Li and Na GICs were discussed. Li GICs were found to be quite stable over a wide range of in-plane Li concentrations, with a change in the favorable stacking sequence of graphite. In terms of diffusion, the migration energy for Li in graphite increases as the graphite stacking transition occurs, suggesting that hindering the stacking transition could realize fast and uniform Li diffusion. In contrast, Na GICs are less stable than Li GICs because of following two reasons: (1) interaction between Na and carbon is less stable than that between Li and carbon, and (2) a larger amount of deformation in the interlayer distance is necessary. The Na GICs tend to be stabilized by increasing the number of Na-carbon interactions. Namely, fasted Na diffusion is expected in the Na-rich phase. Our systematic simulations of the formation energy and migration energy of Na GICs with different structures and in-plane AM concentrations suggested that the expansion of graphite layers prior to Na intercalation could achieve graphite anodes for Na-ion batteries.

Structure determination of glass remains an important issue in glass science. The electron microscope with its high spatial resolution and versatile functions has been widely applied to observe phase separation and structural heterogeneity in glass. While elemental analysis such as energy dispersive spectroscopy (EDS) and electron energy loss spectroscopy (EELS) may provide local compositional information with nanometer-scale resolution, structural information in a glass network cannot be directly obtained. Here, a novel way to probe local coordination is employed using electron energy loss fine structure (ELNES) in the scanning transmission electron microscope (STEM). The method is demonstrated in a phase-separated aluminosilicate glass with multiple Al-coordinated species. With the support of ab initio calculation, two exciton-like peaks in the Al L2,3-edge at around 77 and 80 eV are attributed to 4-fold and 5,6-fold Al excitations, respectively. Mapping of the relative intensity ratio for two peaks in a phase-separated microstructure reveals a heterogeneous distribution of highly coordinated Al species in real space. The finding is in agreement with previous MD simulation that 5- and 6-fold Al species are favored to form in the Al-rich phase. This work has demonstrated that complex network structure within the phase-separated region can now be studied via STEM-EELS.

Deep neural networks (DNNs) have been used to successfully predict molecular properties calculated based on the Kohn--Sham density functional theory (KS-DFT). Although this prediction is fast and accurate, we believe that a DNN model for KS-DFT must not only predict the properties but also provide the electron density of a molecule. This letter presents the quantum deep field (QDF), which provides the electron density with an unsupervised but end-to-end physics-informed modeling by learning the atomization energy on a large-scale dataset. QDF performed well at atomization energy prediction, generated valid electron density, and demonstrated extrapolation. Our QDF implementation is available at https://github.com/masashitsubaki/QuantumDeepField_molecule.

Characterization is indispensable for developing functional materials and molecules. In particular, spectroscopy provides atomic configuration, chemical bonding, and vibrational information, which are crucial for understanding the mechanism underlying the functions of a material and molecule. Despite its importance, the interpretation of spectra using “human-driven” methods, such as manual comparison of experimental spectra with reference/simulated spectra, is becoming difficult owing to the increase in experimental data. To overcome the limitations of “human-driven” methods, new data-driven approaches based on machine learning were developed. In this chapter, we review our machine learning method for spectral analysis. Hierarchical clustering, a decision tree, and a feedforward neural network were combined to investigate the core loss spectroscopy, namely electron energy loss near edge structures (ELNES) spectrum, which is identical to the X-ray absorption near edge structure (XANES) spectrum. Hierarchical clustering and the decision tree are used to interpret and predict ELNES/XANES, while the feedforward neural network is used to obtain hidden information about the material structure and properties from the spectra. Further, we construct a prediction model that is robust against noise by data augmentation. Finally, we apply our method to noisy spectra and predict six properties accurately. In summary, the proposed approaches can pave the way for fast and accurate spectrum interpretation/prediction as well as the local measurement of material functions.

We measured the local composition and thickness of SiO2-based glass material from diffraction. By using four dimensional scanning transmission electron microscopy (4D-STEM), we obtained diffraction at each scanning point. Comparing the obtained diffraction with simulated diffraction patterns, we try to measure the local composition and thickness. Although this method requires some constraints, this method measured local composition and thickness with 1/10 or less electron dose of EELS., Comment: 15 pages, 10 figures, 6 supporting figures

Spectroscopy is indispensable for determining atomic configurations, chemical bondings, and vibrational behaviours, which are crucial information for materials development. Despite their importance, the interpretation of spectra using “human-driven” methods, such as the manual comparison of experimental spectra with reference/simulated spectra, is difficult due to the explosive increase in the number of experimental spectra to be observed. To overcome the limitations of the “human-driven” approach, we develop a new “data-driven” approach based on machine learning techniques by combining the layer clustering and decision tree methods. The proposed method is applied to the 46 oxygen-K edges of the ELNES/XANES spectra of oxide compounds. With this method, the spectra can be interpreted in accordance with the material information. Furthermore, we demonstrate that our method can predict spectral features from the material information. Our approach has the potential to provide information about a material that cannot be determined manually as well as predict a plausible spectrum from the geometric information alone.

Silicate glasses are indispensable for optical and photonics applications, and their properties are affected by phase-separated structures. Understanding the phase separation behavior inside the glasses is thus crucial for controlling their optical properties. Here, we attempt to identify the phase-separated structure inside silicate glass by high-angular annular dark field-scanning transmission electron microscopy (HAADF-STEM) combined with a multi-slice image simulation. In addition to the phase-separated structure, we also demonstrate that the identifications of the type and stage of the phase-separation are possible by the HAADF observation in combination with a phase separation simulation.

Data-centric approaches have become increasingly popular in materials science, also known as informational materials science. Nanostructures often play essential roles in materials properties. Nanoinformatics is an important subset of informational materials science and a powerful tool for characterization and design of nanostructures. It allows discovery of meaningful and useful information and patterns from experimental and theoretical data and databases. This article reviews progress in nanoinformatics and informational materials science. Data-centric approaches for materials property description, construction of interatomic potentials, discovery of new inorganic compounds, efficient characterization of ionic transport and interfacial structures, hyperspectral image data analysis, and design of catalytic nanoparticles are outlined.

The discovery of molecules with specific properties is crucial to developing effective materials and useful drugs. Recently, to accelerate such discoveries with machine learning, deep neural networks (DNNs) have been applied to quantum chemistry calculations based on the density functional theory (DFT). While various DNNs for quantum chemistry have been proposed, these networks require various chemical descriptors as inputs and a large number of learning parameters to model atomic interactions. In this paper, we propose a new DNN-based molecular property prediction that (i) does not depend on descriptors, (ii) is more compact, and (iii) involves additional neural networks to model the interactions between all the atoms in a molecular structure. In the consideration of the molecular structure, we also model the potentials between all the atoms; this allows the neural networks to simultaneously learn the atomic interactions and potentials. We emphasize that these atomic "pair" interactions and potentials are characterized using the global molecular structure, a function of the depth of the neural networks; this leads to the implicit or indirect consideration of atomic "many-body" interactions and potentials within the DNNs. In the evaluation of our model with the benchmark QM9 data set, we achieved fast and accurate prediction performances for various quantum chemical properties. In addition, we analyzed the effects of learning the interactions and potentials on each property. Furthermore, we demonstrated an extrapolation evaluation, i.e., we trained a model with small molecules and tested it with large molecules. We believe that insights into the extrapolation evaluation will be useful for developing more practical applications in DNN-based molecular property predictions.

Grain boundary segregation of dopants plays a crucial role in materials properties. To investigate the dopant segregation behavior at the grain boundary, an enormous number of combinations have to be considered in the segregation of multiple dopants at the complex grain boundary structures. Here, two data mining techniques, the random-forests regression and the genetic algorithm, were applied to determine stable segregation sites at grain boundaries efficiently. Using the random-forests method, a predictive model was constructed from 2% of the segregation configurations and it has been shown that this model could determine the stable segregation configurations. Furthermore, the genetic algorithm also successfully determined the most stable segregation configuration with great efficiency. We demonstrate that these approaches are quite effective to investigate the dopant segregation behaviors at grain boundaries.

1 Recently, powerful methods for determining grain boundary structures with the aid of machine learning techniques have been proposed. However, the application of these methods to oxide materials has not been reported. Herein, we describe a Bayesian optimization method (Kriging) for effective and accurate determination of grain boundary structures of complex materials, namely metal oxides, including MgO, TiO2, and CeO2. The efficiency of this method is ~500 times higher than that of conventional all candidate calculations. We reveal that the grain boundary energy surface of metal oxides is very similar to that of metallic materials, enabling the use of the Kriging method to determine grain boundary structures.

It has been reported that dislocations with 1/3 $$ \left\langle {\bar{1}101} \right\rangle $$ edge component of the Burgers vector are formed in {1 $$ \bar{1} $$ 04}/ $$ \left\langle {11\bar{2}0} \right\rangle $$ low-angle grain boundaries of alumina (α-Al2O3). These dislocations dissociate into two partial dislocations with a stacking fault on the (0001) plane (Tochigi et al. in J Mater Sci 46:4428–4433, 2011). However, the dissociation reaction of these dislocations has not been determined so far. In this study, the structures of the dissociated dislocations and the (0001) stacking fault were investigated by transmission electron microscopy and theoretical calculations. It was revealed that the dissociated dislocations were generated from the 1/3 $$ \left\langle {\bar{1}101} \right\rangle $$ perfect edge dislocation by the reaction of 1/3 $$ \left\langle {\bar{1}101} \right\rangle $$ → 1/18 $$ \left\langle {\bar{4}223} \right\rangle $$ + 1/18 $$ \left\langle {\bar{2}4\bar{2}3} \right\rangle $$ . Furthermore, electron energy loss spectroscopy analysis was performed to examine the atomic/electronic structure of the (0001) stacking fault. In the observed spectra, a chemical shift and intensity decrease were found at the oxygen K-edge. Theoretical spectrum analysis using first-principles calculations revealed that the characteristic features of the spectra are originated from the local atomic configurations of the (0001) stacking fault.

Given the strong dependence of material structure and properties on the length and strength of constituent bonds and the fact that surface adsorption and chemical reactions are initiated by the formation of bonds between two systems, bonding parameters are of key importance for material design and industrial processes. In this study, a machine learning (ML)-based model is used to accurately predict bonding properties from information pertaining to isolated systems before bonding. This model employs the density of states (DOS) before bond formation as the ML descriptor and accurately predicts binding energy, bond distance, covalent electron amount, and Fermi energy even when only 20% of the whole dataset is used for training. The results show that the DOS of isolated systems before bonding is a powerful descriptor for the accurate prediction of bonding and adsorption properties., 10 pages, 3figures

Understanding structures and spatial distributions of molecules in liquid phases is crucial for the control of liquid properties and to develop efficient liquid-phase processes. Here, real-space mapping of molecular distributions in a liquid was performed. Specifically, the ionic liquid 1-Ethyl-3-methylimidazolium bis(trifluoromethanesulfonyl)imide (C2mimTFSI) was imaged using atomic-resolution scanning transmission electron microscopy. Simulations revealed network-like bright regions in the images that were attributed to the TFSI- anion, with minimal contributions from the C2mim+ cation. Simple visualization of the TFSI- distribution in the liquid sample was achieved by binarizing the experimental image.

The electron energy loss near edge structures (ELNES) appearing in an electron energy loss spectrum obtained through transmission electron microscopy (TEM) have the potential to unravel atomic and electronic structures with sub-nano meter resolution. For this reason, TEM-ELNES has become one of the most powerful analytical methods in materials research. On the other hand, theoretical calculations are indispensable in interpreting the ELNES spectrum. Here, the basics and applications of one-particle, two-particle and multi-particle ELNES calculations are reviewed. A key point for the ELNES calculation is the proper introduction of the core-hole effect. Some applications of one-particle ELNES calculations to huge systems of more than 1000 atoms, and complex systems, such as liquids, are reported. In the two-particle calculations, the importance of the correct treatment of the excitonic interaction is demonstrated in calculating the low-energy ELNES, for example at the Li-K edge. In addition, an unusually strong excitonic interactions in the O-K edge of perovskite oxides is identified. The multi-particle calculations are necessary to reproduce the multiplet structures appearing at the transition metal L2,3-edges and rare-earth M4,5-edges. Applications to dilute magnetic semiconductors and Li-ion battery materials are presented. Furthermore, beyond the 'conventional' ELNES calculations, theoretical calculations of electron/X-ray magnetic circular dichroism (MCD) and the vibrational information in ELNES, are reported.