Discovery Of New Materials Research Articles

Machine learning algorithms for optimization of magnetocaloric effect in all-d-metal Heusler alloys

This study examines the application of machine learning algorithms, specifically the Random Forest regression model, to optimize the magnetocaloric effect in all-d-metal Heusler alloys. The model was trained using descriptors related to the mean properties of individual atoms, the properties of simple compounds in their ground state, and measures of chemical disorder. It demonstrated high accuracy in predicting structural properties, while exhibiting moderate accuracy in predicting magnetic properties. To identify optimal alloy compositions, a genetic algorithm was used to find those with the greatest differences in magnetization during martensitic transitions. Using this combined approach, the Ni–Co–Mn–Ti alloy system was thoroughly explored, resulting in the discovery of an alloy with a maximum magnetization difference. These results are consistent with previous research based on density functional theory and highlight the effectiveness of integrating machine learning with genetic algorithms for the discovery of new materials with outstanding magnetocaloric properties. The study emphasizes the need for further refinement of models capable of accurately predicting complex magnetic interactions, which is essential for fully leveraging the potential of all-d-metal Heusler alloys in practical applications.

Autonomous Scanning Tunneling Microscopy Imaging via Deep Learning.

Scanning tunneling microscopy (STM) is a powerful technique that provides the ability to manipulate and characterize individual atoms and molecules with atomic-level precision. However, the processes of scanning samples, operating the probe, and analyzing data are typically labor-intensive and subjective. Deep learning (DL) techniques have shown immense potential in automating complex tasks and solving high-dimensional problems. In this study, we developed an autonomous STM framework powered by DL to enable autonomous operations of the STM without human interventions. Our framework employs a convolutional neural network (CNN) for real-time evaluation of STM image quality, a U-net model for identifying bare surfaces, and a deep Q-learning network (DQN) agent for autonomous probe conditioning. Additionally, we integrated an object recognition model for the automated recognition of different adsorbates. This autonomous framework enables the acquisition of space-averaging information using STM techniques without compromising the high-resolution molecular imaging. We achieved measuring an area of approximately 1.9 μm2 within 48 h of continuous measurement and automatedly generated the statistics on the molecular species present within the mesoscopic area. We demonstrate the high robustness of the framework by conducting measurements at the liquid nitrogen temperature (∼78 K). We envision that the integration of DL techniques and high-resolution microscopy will not only extend the functionality and capability of scanning probe microscopes but also accelerate the understanding and discovery of new materials.

FOS: A fully integrated open-source program for Fast Optical Spectrum calculations of nanoparticle media

FOS, which means light in Greek, is an open-source program for Fast Optical Spectrum calculations of nanoparticle media. This program takes the material properties and a description of the system as input, and outputs the spectral response including the reflectance, absorptance, and transmittance. Previous open-source codes often include only one portion of what is needed to calculate the spectral response of a nanoparticulate medium, such as Mie theory or a Monte Carlo method. FOS is designed to provide a convenient fully integrated format to remove the barrier as well as providing a significantly accelerated implementation with compiled Python code, parallel processing, and pre-trained machine learning predictions. This program can accelerate optimization and high throughput design of optical properties of nanoparticle or nanocomposite media, such as radiative cooling paint and solar heating liquids, allowing for the discovery of new materials and designs. FOS also enables convenient modeling of lunar dust coatings, combustion particulates, and many other particulate systems. In this paper we discuss the methodology used in FOS, features of the program, and provide four case studies.

Stable crystal structure prediction using machine learning-based formation energy and empirical potential function

Crystal structure prediction aims to predict stable and easily experimentally synthesized materials, which accelerates the discovery of new materials. It is worth noting that the stability of materials is the basis for ensuring high performance and reliable application of materials. Among which, the thermodynamic and molecular dynamics stability is especially important. Therefore, this paper proposes a method to predict stable crystal structures using formation energy and Lennard-Jones potential as evaluation indicators. Specifically, we use graph neural network models to predict the formation energy of crystals, and employ empirical formulas to calculate the Lennard-Jones potential. Then, we apply Bayesian optimization algorithms to search for crystal structures with low formation energy and Lennard-Jones potential approaching zero, in order to ensure the thermodynamic stability and dynamics stability of materials. In addition, considering the impact of the bonding situation between atoms in the crystal on the structural stability, this article uses contact map to analyze the atomic bonding situation of each crystal to screen out more stable materials. Finally, the experimental results show that the method we proposed can not only reduce the time for crystal structure prediction, but also ensure the stability of crystal materials.

Machine learning assisted study of phase and properties in cobalt-free AlCrxCuFeNi2 high-entropy alloys

The application of machine learning (ML) accelerates the discovery of new materials, the prediction and optimization of properties, and the analysis of the correlation between structure and properties. The design, characterization, and mechanical properties of the novel AlCrxCuFeNi2 cobalt-free high-entropy alloys (HEAs) were investigated in this study. Through the experimental validation combined with the predictive K-nearest neighbor model (KNN) and the extreme gradient boosting model (XGBoost), the influence of Cr content on phase evolution and mechanical behavior were analyzed comprehensively. The results indicate that the optimized KNN and XGBoost models can achieve the 10-fold average cross validation (CV) accuracies of 0.761 and 0.836, respectively. Eight components of AlCrxCuFeNi2 HEAs are selected for experimental verification. The experimental results shows that the microstructures of the eight alloys are composed of FCC + BCC phases, which is agreement with the prediction. With the increase of Cr content, the dendrite morphology becomes finer and the lattice constants increase. The addition of Cr results in the hardness values from 332.4 HV to 447.7 HV and the deviation from the predicted ones is between 5.4 HV and 37.3 HV. It is worth noting that wear resistance is consistent with the change in hardness values, the specific wear rate decreases from 3.152 × 10−4 mm3/Nm to 6.357 × 10−5 mm3/Nm, and when x = 2.0, the wear resistance is the best. This research underscores the efficacy of ML in expediting the discovery and optimization of HEAs, offering valuable insights for promoting materials design and engineering applications.

Accelerating Computational Materials Discovery with Machine Learning and Cloud High-Performance Computing: from Large-Scale Screening to Experimental Validation.

High-throughput computational materials discovery has promised significant acceleration of the design and discovery of new materials for many years. Despite a surge in interest and activity, the constraints imposed by large-scale computational resources present a significant bottleneck. Furthermore, examples of very large-scale computational discovery carried out through experimental validation remain scarce, especially for materials with product applicability. Here, we demonstrate how this vision became reality by combining state-of-the-art machine learning (ML) models and traditional physics-based models on cloud high-performance computing (HPC) resources to quickly navigate through more than 32 million candidates and predict around half a million potentially stable materials. By focusing on solid-state electrolytes for battery applications, our discovery pipeline further identified 18 promising candidates with new compositions and rediscovered a decade's worth of collective knowledge in the field as a byproduct. We then synthesized and experimentally characterized the structures and conductivities of our top candidates, the NaxLi3-xYCl6 (0≤ x≤ 3) series, demonstrating the potential of these compounds to serve as solid electrolytes. Additional candidate materials that are currently under experimental investigation could offer more examples of the computational discovery of new phases of Li- and Na-conducting solid electrolytes. The showcased screening of millions of materials candidates highlights the transformative potential of advanced ML and HPC methodologies, propelling materials discovery into a new era of efficiency and innovation.

High‐Throughput Screening of Low‐Bandgap Organic Semiconductors for Photovoltaic Applications: In the Search of Correlations

Low‐bandgap nonfullerene acceptors (NFAs) offer a unique potential for photovoltaic (PV) applications, such as transparent PV and agrivoltaics. Evaluating each new PV system to achieve the optimum thickness, microstructure, and device performance is, however, a complex multiparametric challenge with large time and resource requirements. Herein, the PV potential of low‐bandgap donor and NFA materials by combining high‐throughput screening and statistical methods is evaluated. The use of thickness gradients (20–600 nm) facilitates the fabrication of more than 2000 doctor‐bladed devices from 24 different low‐bandgap blend combinations. The corresponding power conversion efficiencies varies significantly, from 0.06% to 10.45% across materials and thicknesses. The self‐consistency of the large dataset allows to perform a parameter sensitivity study as well as parameter correlation analysis. These reveal that the choice of materials and energy alignment‐related features (i.e., electron affinity offset, ionization energy offset, bandgap, and energy loss) has the largest influence on final device performance, while processing conditions appear less important for the final efficiencies. Our study demonstrates that high‐throughput experimentation is a perfect match for correlation analyses in order to gain a statistically meaningful understanding of these systems, potentially accelerating the discovery of new materials.

NSGAN: a non-dominant sorting optimisation-based generative adversarial design framework for alloy discovery

The design and discovery of new materials is fundamental to advancing scientific and technological innovation. The recent emergence of the materials genome concept holds great promise in revolutionising materials science by enabling the systematic utilisation of data for efficient prediction and optimisation of ‘superior’ materials. However, the materials genome approach can be stymied by the vast complexity of design spaces, which often demand substantial computational resources and sophisticated data processing capabilities. To address these challenges, this work introduces a generative design framework called the non-dominant sorting optimisation-based generative adversarial networks (NSGAN). Capitalising on the synergies of genetic algorithms (GA) and generative adversarial networks (GANs), NSGAN provides a robust and efficient approach for tackling high-dimensional multi-objective optimisation design problems. To validate the efficacy of the proposed framework, we applied the model to a comprehensive dataset of aluminium alloys. Additionally, an online tool was created as a supplementary resource, offering a brief introduction to this innovative method for the wider scientific community. This study explores the potential of a predictive and data-driven approach in material design, indicating a promising pathway for widespread applications in the field of materials science.

A physics-and-data co-driven material design strategy for multicomponent diboride ceramics

The vast chemical composition space of multicomponent ceramics provides opportunity for regulating their properties and performances. Other than trial-and-error and screening, a physics-and-data co-driven material design strategy was reported to efficiently explore chemical composition space through establishing prediction model between physical properties and chemical composition via physics-guided data mining method. Firstly, causal and correlative relationships were discriminated according to physicochemical knowledge. Secondly, ensemble features such as chemical formula and chemical bond were digitized by constructing descriptors for crystalline solids with atomic features including atomic weight and radius contained in the Periodic Table of elements. Finally, quantitative causal relations between material structure/properties and ensemble descriptors were mined out of small dataset using a least square regression method, on which physical prediction model established therefore for forward predicting structure/properties and inverse designing chemical composition. Choosing diboride ceramics as a case, those causal analytic relations of composition-structure and composition-property were successfully derived and illustrated to predict the value of structure and properties accompanying with quantitative relation between chemical pressure and chemical bonding ensemble descriptors. This method bridges the connection between atomic features and macroscopic properties via ensemble descriptors, which significantly accelerates the discovery of new materials.

Generalization of graph-based active learning relaxation strategies across materials

Although density functional theory (DFT) has aided in accelerating the discovery of new materials, such calculations are computationally expensive, especially for high-throughput efforts. This has prompted an explosion in exploration of machine learning (ML) assisted techniques to improve the computational efficiency of DFT. In this study, we present a comprehensive investigation of the broader application of Finetuna, an active learning framework to accelerate structural relaxation in DFT with prior information from Open Catalyst Project pretrained graph neural networks. We explore the challenges associated with out-of-domain systems: alcohol ( C>2 ) on metal surfaces as larger adsorbates, metal oxides with spin polarization, and three-dimensional (3D) structures like zeolites and metal organic frameworks. By pre-training ML models on large datasets and fine-tuning the model along the simulation, we demonstrate the framework’s ability to conduct relaxations with fewer DFT calculations. Depending on the similarity of the test systems to the training systems, a more conservative querying strategy is applied. Our best-performing Finetuna strategy reduces the number of DFT single-point calculations by 80% for alcohols and 3D structures, and 42% for oxide systems.

Semantics-Enabled Data Federation: Bringing Materials Scientists Closer to FAIR Data

The development and discovery of new materials can be significantly enhanced through the adoption of FAIR (Findable, Accessible, Interoperable, and Reusable) data principles and the establishment of a robust data infrastructure in support of materials informatics. A FAIR data infrastructure and associated best practices empower materials scientists to access and make the most of a wealth of information on materials properties, structures, and behaviors, allowing them to collaborate effectively, and enable data-driven approaches to material discovery. To make data findable, accessible, interoperable, and reusable to materials scientists, we developed and are in the process of expanding a materials data infrastructure to capture, store, and link data to enable a variety of analytics and visualizations. Our infrastructure follows three key architectural design philosophies: (i) capture data across a federated storage layer to minimize the storage footprint and maximize the query performance for each data type, (ii) use a knowledge graph-based data fusion layer to provide a single logical interface above the federated data repositories, and (iii) provide an ensemble of FAIR data access and reuse services atop the knowledge graph to make it easy for materials scientists and other domain experts to explore, use, and derive value from the data. This paper details our architectural approach, open-source technologies used to build the capabilities and services, and describes two applications through which we have successfully demonstrated its use. In the first use case, we created a system to enable additive manufacturing data storage and process parameter optimization with a range of user-friendly visualizations. In the second use case, we created a system for exploring data from cathodic arc deposition experiments to develop a new steam turbine coating material, fusing a combination of materials data with physics-based equations to enable advanced reasoning over the combined knowledge using a natural language chatbot-like user interface.

Challenges in High-Throughput Inorganic Materials Prediction and Autonomous Synthesis

Materials discovery lays the foundation for many technological advancements. The prediction and discovery of new materials are not simple tasks. Here, we outline some basic principles of solid-state chemistry, which might help to advance both, and discuss pitfalls and challenges in materials discovery. Using the recent work of Szymanski [Nature 624, 86 (2023)], which reported the autonomous discovery of 43 novel materials, as an example, we discuss problems that can arise in unsupervised materials discovery and hope that by addressing these, autonomous materials discovery can be brought closer to reality. We discuss all 43 synthetic products and point out four common shortfalls in the analysis. These errors unfortunately lead to the conclusion that no new materials have been discovered in that work. We conclude that there are two important points of improvement that require future work from the community, as follows. (i) Automated Rietveld analysis of powder x-ray diffraction data is not yet reliable. Future improvement of such, and the development of a reliable artificial-intelligence-based tool for Rietveld fitting, would be very helpful, not only for autonomous materials discovery but also for the community in general. (ii) We find that disorder in materials is often neglected in predictions. The predicted compounds investigated herein have all their elemental components located on distinct crystallographic positions but in reality, elements can share crystallographic sites, resulting in higher-symmetry space groups and—very often—known alloys or solid solutions. This error might be related to the difficulty of modeling disorder in a computationally economical way and needs to be addressed both by computational and experimental material scientists. We find that two thirds of the claimed successful materials in Szymanski are likely to be known compositionally disordered versions of the predicted ordered compounds. We highlight important issues in materials discovery, computational chemistry, and autonomous interpretation of x-ray diffraction. We discuss concepts of materials discovery from an experimentalist point of view, which we hope will be helpful for the community to further advance this important new aspect of our field. Published by the American Physical Society 2024

Machine learning based hybrid ensemble models for prediction of organic dyes photophysical properties: Absorption wavelengths, emission wavelengths, and quantum yields

Fluorescent organic dyes are extensively used in the design and discovery of new materials, photovoltaic cells, light sensors, imaging applications, medicinal chemistry, drug design, energy harvesting technologies, dye and pigment industries, and pharmaceutical industries, among other things. However, designing and synthesizing new fluorescent organic dyes with desirable properties for specific applications requires knowledge of the chemical and physical properties of previously studied molecules. It is a difficult task for experimentalists to identify the photophysical properties of the required chemical molecule at negligible time and financial cost. For this purpose, machine learning-based models are a highly demanding technique for estimating photophysical properties and may be an alternative approach to density functional theory. In this study, we used 15 single models and proposed three different hybrid models to assess a dataset of 3066 organic materials for predicting photophysical properties. The performance of these models was evaluated using three evaluation parameters: mean absolute error, root mean squared error, and the coefficient of determination (R2) on the test-size data. All the proposed hybrid models achieved the highest accuracy (R2) of 97.28%, 95.19%, and 74.01% for predicting the absorption wavelengths, emission wavelengths, and quantum yields, respectively. These resultant outcomes of the proposed hybrid models are ∼1.9%, ∼2.7%, and ∼2.4% higher than the recently reported best models’ values in the same dataset for absorption wavelengths, emission wavelengths, and quantum yields, respectively. This research promotes the quick and accurate production of new fluorescent organic dyes with desirable photophysical properties for specific applications.

Organic chemical Nano sensors: synthesis, properties, and applications.

Nanosensors work on the "Nano" scale. "Nano" is a unit of measurement around 10- 9 m. A nanosensor is a device capable of carrying data and information about the behavior and characteristics of particles at the nanoscale level to the macroscopic level. Nanosensors can be used to detect chemical or mechanical information such as the presence of chemical species and nanoparticles or monitor physical parameters such as temperature on the nanoscale. Nanosensors are emerging as promising tools for applications in agriculture. They offer an enormous upgrade in selectivity, speed, and sensitivity compared to traditional chemical and biological methods. Nanosensors can be used for the determination of microbe and contaminants. With the advancement of science in the world and the advent of electronic equipment and the great changes that have taken place in recent decades, the need to build more accurate, smaller and more capable sensors was felt. Today, high-sensitivity sensors are used that are sensitive to small amounts of gas, heat, or radiation. Increasing the sensitivity, efficiency and accuracy of these sensors requires the discovery of new materials and tools. Nano sensors are nanometer-sized sensors that, due to their small size and nanometer size, have such high accuracy and responsiveness that they react even to the presence of several atoms of a gas. Nano sensors are inherently smaller and more sensitive than other sensors.

Influence of redox engineering on the trade-off relationship between thermopower and electrical conductivity in lanthanum titanium based transition metal oxides

Discovery of new materials plays a critical role in developing advanced high-temperature thermoelectric (TE) applications. Transition metal oxides (TMOs) are one of the attractive candidates for high-temperature TE applications due...

Machine learning assisted crystal structure prediction made simple

Crystal structure prediction (CSP) plays a crucial role in condensed matter physics and materials science, with its importance evident not only in theoretical research but also in the discovery of new materials and the advancement of novel technologies. However, due to the diversity and complexity of crystal structures, trial-and-error experimental synthesis is time-consuming, labor-intensive, and insufficient to meet the increasing demand for new materials. In recent years, machine learning (ML) methods have significantly boosted CSP. In this review, we present a comprehensive review of the ML models applied in CSP. We first introduce the general steps for CSP and highlight the bottlenecks in conventional CSP methods. We further discuss the representation of crystal structures and illustrate how ML-assisted CSP works. In particular, we review the applications of graph neural networks (GNNs) and ML force fields in CSP, which have been demonstrated to significantly speed up structure search and optimization. In addition, we provide an overview of advanced generative models in CSP, including variational autoencoders (VAEs), generative adversarial networks (GANs), and diffusion models. Finally, we discuss the remaining challenges in ML-assisted CSP.

Single-atom alloy structure and unique bonding properties of Au104Ag40(PET)60 nanoclusters.

The detailed characterization of AuAg alloy nanoclusters is essential to guide the discovery of species ideal for applications in various fields including catalysis and biomedicine. This work presents structural analysis of the Au104Ag40(PET)60 species through X-ray absorption spectroscopy (XAS). First, XAS fitting is utilized to model the distribution of Au and Ag atoms within the structure. Our proposed model assigns Ag atoms to the vertex sites of the second shell of the metal core, as well as the outermost staple sites. This distribution reveals Au104Ag40(PET)60 to be a Ag single-atom alloy. The proposed model shows outstanding agreement with the coordination number values derived from XAS. XAS near-edge analysis is employed to investigate the alloy bonding interactions between Au and Ag. Substantial d-electron transfer from Au to Ag is observed in this sample, beyond the magnitude of previously studied AuAg NCs. This work enhances the understanding of the structure-property relationship of AuAg alloy NCs, offering insights which can be applied to other large NCs and even NPs. These insights will in turn aid the discovery of new materials for use in various applications.

A Neural Thermometer for Predicting Phase Transitions of Unknown Systems

A novel convolutional neural network predicts phase transition temperatures from spin configurations without prior information about order parameters, paving the way for the discovery of new materials in condensed matter physics.

Stokes shift prediction of fluorescent organic dyes using machine learning based hybrid cascade models

Fluorescent organic dyes are widely used in various fields, including science and technology, research and development, medicine, and drug delivery. Multitudinous attempts have been made by experimentalists to develop such fluorescent organic dyes with the desired Stokes shift property at negligible cost and time. For quickly and accurately predicting the Stokes shift property of fluorescent organic dye, we proposed eight hybrid models based on the combination of nine single machine-learning models. To fulfill the objective, we considered a dataset of 3066 fluorescent organic materials and evaluated the performance of each model using three evaluation parameters: mean absolute error (MAE), root mean squared error (RMSE), and the coefficient of determination (R2). The hybrid cascade model of Extreme Gradient Boosting Regression and Light Gradient Boosting Machine Regression (XGBR + LGBMR) performed best for Stokes shift prediction, with MAE of 13.83 nm, RMSE of 19.95 nm, and R2 of 86.18 %. The prediction performance of all the undertaken models was validated by the experimental data of four xanthene dyes (Rh-19, Rh–B, Rh-6G, and Rh-110). In this regard, the XGBR + DTR (Extreme Gradient Boosting Regression + Decision Tree Regression) model was the best performer, with errors ranging from 5 to 13 nm for four dyes. The resultant errors are much smaller than the recently reported synthesized material with an error of 30 nm. The proposed models allow for rapid and cost-effective screening of a wide range of fluorescent organic dyes, which assists the researchers in gaining prior knowledge of materials and accelerates the discovery of new materials.

On Some Glyphs of Khitan Small Script

In 2017, Further Research on Khitan Small Script was published, which revised and summarized the phonetic value of 300 glyphs. However, with the discovery of new materials and an increasing number of researchers, new progress has been made in the reconstruction of Khitan small script. This paper aims to introduce the latest research results on the reconstruction of 8 glyphs in Khitan small script.