Material Database Research Articles

ET-AL: Entropy-targeted active learning for bias mitigation in materials data

Growing materials data and data-driven informatics drastically promote the discovery and design of materials. While there are significant advancements in data-driven models, the quality of data resources is less studied despite its huge impact on model performance. In this work, we focus on data bias arising from uneven coverage of materials families in existing knowledge. Observing different diversities among crystal systems in common materials databases, we propose an information entropy-based metric for measuring this bias. To mitigate the bias, we develop an entropy-targeted active learning (ET-AL) framework, which guides the acquisition of new data to improve the diversity of underrepresented crystal systems. We demonstrate the capability of ET-AL for bias mitigation and the resulting improvement in downstream machine learning models. This approach is broadly applicable to data-driven materials discovery, including autonomous data acquisition and dataset trimming to reduce bias, as well as data-driven informatics in other scientific domains.

A method to computationally screen for tunable properties of crystalline alloys

Conventionally, high-throughput computational materials searches start from an input set of bulk compounds extracted from material databases, but, in contrast, many real functional materials are heavily engineered mixtures of compounds rather than single bulk compounds. We present a framework and open-source code to automatically construct and analyze possible alloys and solid solutions from a set of existing experimental or calculated ordered compounds, without requiring additional metadata except crystal structure. As a demonstration, we apply this framework to all compounds in the Materials Project to create a new, publicly available database of 600,000 unique "alloy pair" entries that can be used to search for materials with tunable properties. We exemplify this approach by searching for transparent conductors and reveal candidates that might have been excluded in a traditional screening. This work lays a foundation from which materials databases can go beyond stoichiometric compounds and approach a more realistic description of compositionally tunable materials.

A critical examination of robustness and generalizability of machine learning prediction of materials properties

Recent advances in machine learning (ML) have led to substantial performance improvement in material database benchmarks, but an excellent benchmark score may not imply good generalization performance. Here we show that ML models trained on Materials Project 2018 can have severely degraded performance on new compounds in Materials Project 2021 due to the distribution shift. We discuss how to foresee the issue with a few simple tools. Firstly, the uniform manifold approximation and projection (UMAP) can be used to investigate the relation between the training and test data within the feature space. Secondly, the disagreement between multiple ML models on the test data can illuminate out-of-distribution samples. We demonstrate that the UMAP-guided and query by committee acquisition strategies can greatly improve prediction accuracy by adding only 1% of the test data. We believe this work provides valuable insights for building databases and models that enable better robustness and generalizability.

Two-dimensional ferroelectrics from high throughput computational screening

We report a high throughput computational search for two-dimensional ferroelectric materials. The starting point is 252 pyroelectric materials from the computational 2D materials database (C2DB) and from these we identify 63 ferroelectrics. In particular we find 49 materials with in-plane polarization, 8 materials with out-of-plane polarization and 6 materials with coupled in-plane and out-of-plane polarization. Most of the known 2D ferroelectrics are recovered by the screening and the far majority of the predicted ferroelectrics are known as bulk van der Waals bonded compounds, which makes them accessible by direct exfoliation. For roughly 25% of the materials we find a metastable state in the non-polar structure, which may imply a first order transition to the polar phase. Finally, we list the magnetic pyroelectrics extracted from the C2DB and focus on the case of VAgP2Se6, which exhibits a three-state switchable polarization vector that is strongly coupled to the magnetic excitation spectrum.

Numerical simulation and experimental study of vacuum arc remelting (VAR) process for large-size GH4742 superalloy

Numerical modeling and industrial experiment were used to study the entire vacuum arc remelting (VAR) process of 660 mm GH4742 superalloy. In order to calculate the electromagnetic field, flow field, temperature field, and melt pool shape change throughout the entire VAR process, a 2D axisymmetric model with coupled multi-physical fields matching industry trials was constructed. The microstructure and elemental distribution conditions in the large-size GH4742 ingot were analyzed in combination with the industrial ingot dissection experiment, and the most likely locations of freckle defects were predicted. According to the research, the melt pool has two clockwise flow fields that are affected by buoyancy and Lorentz forces. The melt pool depth deepens gradually as melting proceeds, with a maximum melt pool depth of 0.32 m and a paste zone width of 0.16 m. Although the ingot's elemental distribution is largely consistent, there is microsegregation between the dendrites. Where the freckle is most likely to appear, at (0.322, 0.075), the Rayleigh has a great value of 1.1. For melt pool form and dendritic spacing throughout the entire melting process, the results of the industrial ingot dissection experiment and the simulation results show good agreement, proving the accuracy of the thermal history and solidification microstructure calculation for the entire melting process. Furthermore, the prospect of combining numerical simulation with a special material database for accurate defect prediction is proposed.

Visible-light-responsive Photocatalyst Based on Nitrogen-doped Bulk Oxide YTaO4-x Ny for Z-scheme Overall Water Splitting.

Nitrogen doping into oxide has been proved to be an effective strategy to extend visible-light utilization of oxides with layered or channeled structure, but it is shortage for the bulk oxides free of layered or channeled structure. Here, we report a novel nitrogen-doped bulk oxide (denoted as YTaO4-x Ny ) with good visible light response. As benefited from the strong hybridization of N 2p and O 2p electronic state according to the density functional theory (DFT) calculations, the band gap (3.6 eV) of YTaO4 precursor was sharply narrowed to 2.4 eV on YTaO4-x Ny due to the obvious uplift of the valence band maximum (VBM). After decorated with the platinum cocatalyst, the Pt/YTaO4-x Ny was used as the H2 -evolving photocatalyst for assembly of Z-scheme overall water splitting system by coupling with the PtOx /WO3 as the oxygen evolution photocatalyst and using the I- /IO3- as the shuttle ions. This work enriches the material database of nitrogen-doped oxide with wide visible light absorption and shows its promising application in solar-to-chemical conversion.

Physics guided deep learning for generative design of crystal materials with symmetry constraints

Discovering new materials is a challenging task in materials science crucial to the progress of human society. Conventional approaches based on experiments and simulations are labor-intensive or costly with success heavily depending on experts’ heuristic knowledge. Here, we propose a deep learning based Physics Guided Crystal Generative Model (PGCGM) for efficient crystal material design with high structural diversity and symmetry. Our model increases the generation validity by more than 700% compared to FTCP, one of the latest structure generators and by more than 45% compared to our previous CubicGAN model. Density Functional Theory (DFT) calculations are used to validate the generated structures with 1869 materials out of 2000 are successfully optimized and deposited into the Carolina Materials Database www.carolinamatdb.org, of which 39.6% have negative formation energy and 5.3% have energy-above-hull less than 0.25 eV/atom, indicating their thermodynamic stability and potential synthesizability.

Optical properties study of high-throughput horizontal composition gradient FeSe1-xTex thin film based on ellipsometry

The high-throughput superconductivity research model can effectively accelerate the material synthesis and physical characterization process. It can rapidly form a database of materials and reveal the hidden key physical laws, which provides significant assistance to the research of superconducting materials, especially the exploration of superconductivity mechanism. In this work, FeSe1-xTex thin film with spread composition was grown by combined laser molecular beam epitaxy, the third generation combined thin film preparation technique. The film thickness and surface morphology were obtained by atomic force microscopy, which showed that the sample surface was relatively smooth. In addition, we measured the distribution of critical elements and c-axis lattice constants within the different microzones along the direction of composition variation on the surface of the film, by X-ray photoelectron spectroscopy and X-ray diffraction, respectively, which verified that the film was prepared in accordance with a continuous composition gradient. Based on this, nine microzones of the film were tested using the ellipsometer equipped with a focused light spot. The dispersion model of the FeSe1-xTex layer was established using the Drude-Lorentz oscillator, and the optical constants of the film with horizontal composition gradient were obtained by inversion and fitting of the data within each microzone. The results indicate that the FeSe1-xTex thin film shows an obvious composition-dependent optical property. Our work reports the optical constants of FeSe1-xTex composition-spread thin film by using a nondestructive and accurate method, spectroscopic ellipsometry, which enriches the database of this material. This work may provide a reference for establishing the correlation between the optical and superconducting properties of FeSe-based superconductors.

Exploring and machine learning structural instabilities in 2D materials

We address the problem of predicting the zero-temperature dynamical stability (DS) of a periodic crystal without computing its full phonon band structure. Here we report the evidence that DS can be inferred with good reliability from the phonon frequencies at the center and boundary of the Brillouin zone (BZ). This analysis represents a validation of the DS test employed by the Computational 2D Materials Database (C2DB). For 137 dynamically unstable 2D crystals, we displace the atoms along an unstable mode and relax the structure. This procedure yields a dynamically stable crystal in 49 cases. The elementary properties of these new structures are characterized using the C2DB workflow, and it is found that their properties can differ significantly from those of the original unstable crystals, e.g., band gaps are opened by 0.3 eV on average. All the crystal structures and properties are available in the C2DB. Finally, we train a classification model on the DS data for 3295 2D materials in the C2DB using a representation encoding the electronic structure of the crystal. We obtain an excellent receiver operating characteristic (ROC) curve with an area under the curve (AUC) of 0.90, showing that the classification model can drastically reduce computational efforts in high-throughput studies.

Prediction of Material Properties of Inorganic Compounds Using Self-Attention Network

Due to the increase in material databases in recent years, there has been a lot of research regarding deep learning models which use large sizes of datasets and are aimed at the prediction of the material properties of inorganic compounds. Particularly, prediction models with Self-Attention structures, such as Roost and CrabNet, have garnered attention because of two reasons: (1) input variables are confined to the chemical composition of each formula and (2) Self-Attention enables models to learn individual element representations based on their chemical environment. However, the existing Self- Attention model yields low prediction accuracy when predicting structure-dependent material properties, such as the magnetic moment, for lack of structural information of compounds as input. In this research, based on the existing Self- Attention model, we set both elemental and structural information, especially the space group number and lattice constant, as input information and successfully construct a prediction model that is more versatile than existing methods. Furthermore, we visualized lists of promising materials by adopting Bayesian optimization. As a result, we have developed a system to propose desired materials for materials researchers.

Cost Estimation Model Based on Building Information Modeling and Virtual Reality for Customizing Presold Homes

Presold homes refer to houses that are sold before construction has begun or while the construction is still underway. To facilitate the selection and purchase of houses by clients, construction companies usually use sample houses, models, or videos to show the appearance of new houses being constructed. Clients have numerous opportunities to participate in the spatial planning and spatial allocation of the houses because these houses have not yet been constructed, and clients can direct the style and materials used for interior partition walls and furnishings. However, when discussing housing styles with their clients, construction companies often depend on their imagination to create three-dimensional images depicting the interior layout of houses. Construction companies can neither immediately present new styles when clients request changes to partition walls or styles nor estimate the costs of the changes. Consequently, clients cannot immediately decide on such changes, which then leads to poor communication between the two parties. Accordingly, this study designed a system that integrates building information modeling (BIM) and virtual reality (VR). Through this system, clients can change and immediately view house styles in immersive VR and decide on whether the changes meet their demands. Furthermore, the designed system is linked to a unit price database for materials. When clients make alterations in the system, it automatically calculates the costs of the changes and then helps the client determine whether the changes are within their budgets. The results of this study indicate that the designed system can accurately and rapidly achieve cost estimation. Furthermore, the proposed system ensures consistency between BIM and virtual environment information, thereby promoting the application of BIM in the stages of a building life cycle subsequent to building design.

A mid-infrared study of synthetic glass and crystal mixtures analog to the geochemical terranes on mercury

The MERTIS (MErcury Radiometer and Thermal Infrared Spectrometer) onboard of the BepiColombo ESA/JAXA mission to Mercury will map the surface of Mercury in the wavelength range of 7–14 μm and for the interpretation of these spectra a database of analog materials is needed. We analyzed bulk grain size fractions of a series of analog materials relevant to the distinct terranes of Mercury in diffuse reflectance in the mid-infrared (2.5 μm to 18 μm). Mineral mixtures cover a wide range of modal amounts of forsterite, enstatite, diopside and plagioclase, the resulting spectra can be divided into three distinct groups: (1) is dominated by a single glass feature, (2) by forsterite bands, and (3) by pyroxene bands. Despite often high contents, plagioclase features, are usually ‘overprinted’ by forsterite and pyroxene bands.Spectral parameter CF, an easy obtainable proxy for chemistry (SiO2) and polymerization (SCFM) places the hermean mixtures mostly in the intermediate and basaltic range. The correlation of parameters easily obtainable in remote sensing, Mg/Si ratio, and CF, allows differing materials from high-energy evaporation processes in impacts from such formed in igneous processes.Preliminary comparison with a spectrum covering most of the hermean surface shows some similarity with band positions of the Inter Crater Plain and Heavily Cratered Terrains (IcP-HCT) and High-Mg Northern Volcanic Plains (High-Mg NVP) mixtures, but none of our spectra is able to reproduce the remote sensing data entirely.

Two methods on one stone: Integrating visual and analytical techniques to clarify lithic raw material utilization in the Middle and Later Stone Age at Umhlatuzana rockshelter (South Africa)

We develop a study protocol to efficiently and accurately identify the raw material categories constituting the lithic assemblages at Umhlatuzana rockshelter, South Africa. We combine visual and analytical methods to establish a raw material database and to provide a more accurate insight into raw material selection during the Pleistocene Middle and Later Stone Age occupational sequence of Umhlatuzana. The protocol combines petrological properties (as studied on micromorphological samples), elemental composition of the specimens (as measured with p-XRF) and visual characterization by lithic analysts. We tested the protocol by applying it to a sample of piece-plotted lithics from four spits across the stratigraphic sequence. We document the intensive use of sandstone accounting for 25% of the tested sample. We also report a larger importance of hornfels and lower proportion of quartz than was reported in previous analyses (Kaplan 1990). The combination of micromorphological and p-XRF analysis of the Umhlatuzana assemblages demonstrates that with only visual inspection, the variability of raw materials used may be misinterpreted. With accurate raw material datasets, we are better equipped to answer techno-economic questions of the southern African Stone Age.

Introduction of Materials Genome Technology and Its Applications in the Field of Biomedical Materials

Traditional research and development (R&D) on biomedical materials depends heavily on the trial and error process, thereby leading to huge economic and time burden. Most recently, materials genome technology (MGT) has been recognized as an effective approach to addressing this problem. In this paper, the basic concepts involved in the MGT are introduced, and the applications of MGT in the R&D of metallic, inorganic non-metallic, polymeric, and composite biomedical materials are summarized; in view of the existing limitations of MGT for R&D of biomedical materials, potential strategies are proposed on the establishment and management of material databases, the upgrading of high-throughput experimental technology, the construction of data mining prediction platforms, and the training of relevant materials talents. In the end, future trend of MGT for R&D of biomedical materials is proposed.

Two-Dimensional Semiconductors with High Intrinsic Carrier Mobility at Room Temperature.

Two-dimensional semiconductors have demonstrated great potential for next-generation electronics and optoelectronics, however, the current 2D semiconductors suffer from intrinsically low carrier mobility at room temperature, which significantly limits their applications. Here we discover a variety of new 2D semiconductors with mobility 1 order of magnitude higher than the current ones and even higher than bulk silicon. The discovery was made by developing effective descriptors for computational screening of the 2D materials database, followed by high-throughput accurate calculation of the mobility using a state-of-the-art first-principles method that includes quadrupole scattering. The exceptional mobilities are explained by several basic physical features; particularly, we find a new feature: carrier-lattice distance, which is easy to calculate and correlates well with mobility. Our Letter opens up new materials for high performance device performance and/or exotic physics, and improves the understanding of the carrier transport mechanism.

A BIM–LCA Approach for the Whole Design Process of Green Buildings in the Chinese Context

The integrated description of the building geometry and the element attributes of the building information model (BIM) can reduce the effort needed to acquire data for life cycle assessment (LCA) and life cycle costing (LCC) at each design stage while supporting their potential for analyzing life cycle performances and feeding back to the design process. To support this, several methods and tools have been proposed that aim to obtain the life cycle performances of buildings following the level of model fidelity with the life cycle inventory (LCI) database at different scales. However, inconsistencies in decision-making caused by regional differences in LCA/LCC data sources, benchmarks, and building standards cannot be ignored. In this study, a scalable LCA/LCC method integrated with the BIM platform is proposed for the whole green building design process in the Chinese context, and it is implemented with a developed tool based on Revit. A national-/regional-specified database of building elements and materials is established. Referring to China’s carbon-neutral target and relevant standards for green buildings, the baseline values are deduced, and a reference building is defined accordingly to facilitate the evaluation and improvement of the design scheme. According to the Assessment Standard for Green Building (GB50378-2019) and the survey of architectural design practices in China, the key parameters at different design stages are defined. The method and tool are demonstrated using the case study of a school building, analyzing its life cycle carbon emissions and life cycle costs throughout the design process. The results show that the proposed method can facilitate the improvement of the scheme at different design stages and that it can cope with different data accuracies and different LODs in the building information model in the Chinese green building design process. Lastly, the uncertainties raised by the data quality and time-associated factors are discussed.

A framework driven by physics-guided machine learning for process-structure-property causal analytics in additive manufacturing

Data analytics with Machine Learning (ML) using physics knowledge and big data offers high potential to continuously transform raw data to newfound knowledge of Process-Structure-Property (PSP) causal relationships. In Additive Manufacturing (AM), however, realizing the potential is still limited largely due to the lack of a systematic way to learn the PSP relationships for various AM processes. To address the limitation, this paper proposes a novel framework driven by physics-guided ML, which consists of three tiers: (1) knowledge of predictive PSP models and physics, (2) PSP features of interest, and (3) raw AM data. The framework defines a PSP-learning process with two sub-processes. The first uses a knowledge-graph-guided top-down approach to generate the requirements for predictive analytics and data acquisition. The second uses a data-driven bottom-up approach to construct and model new PSP knowledge. Together, these processes connect the proposed framework to decision-making and control activities and physical and virtual AM systems, respectively. The paper includes a case study based on Laser Powder Bed Fusion processes including AM Metrology Testbed at the National Institute of Standards and Technology (NIST). The case study introduces predictive ML models and PSP knowledge extracted from the models. We also demonstrate the framework using an ML-Integrated Knowledge Extraction module called MIKE in NIST’s collaborative AM Material Database. The framework newly enables a systematic physics-guided data-driven approach for PSP in AM that can couple physics knowledge with the versatility of data-driven ML models. Using the approach, the framework continuously updates the models (1) to improve the understanding of dynamically generated AM data and (2) to link sub-models into coupled PSP models. Based on the improved understanding, the framework also facilitates decision-making and control activities for AM at multiple scales.

Prediction of lattice thermal conductivity with two-stage interpretable machine learning

Thermoelectric and thermal materials are essential in achieving carbon neutrality. However, the high cost of lattice thermal conductivity calculations and the limited applicability of classical physical models have led to the inefficient development of thermoelectric materials. In this study, we proposed a two-stage machine learning framework with physical interpretability incorporating domain knowledge to calculate high/low thermal conductivity rapidly. Specifically, crystal graph convolutional neural network (CGCNN) is constructed to predict the fundamental physical parameters related to lattice thermal conductivity. Based on the above physical parameters, an interpretable machine learning model–sure independence screening and sparsifying operator (SISSO), is trained to predict the lattice thermal conductivity. We have predicted the lattice thermal conductivity of all available materials in the open quantum materials database (OQMD) (https://www.oqmd.org/). The proposed approach guides the next step of searching for materials with ultra-high or ultra-low lattice thermal conductivity and promotes the development of new thermal insulation materials and thermoelectric materials.

High-throughput computation of Raman spectra from first principles

Raman spectroscopy is a widely-used non-destructive material characterization method, which provides information about the vibrational modes of the material and therefore of its atomic structure and chemical composition. Interpretation of the spectra requires comparison to known references and to this end, experimental databases of spectra have been collected. Reference Raman spectra could also be simulated using atomistic first-principles methods but these are computationally demanding and thus the existing databases of computational Raman spectra are fairly small. In this work, we developed an optimized workflow to calculate the Raman spectra efficiently and taking full advantage of the phonon properties found in existing material databases. The workflow was benchmarked and validated by comparison to experiments and previous computational methods for select technologically relevant material systems. Using the workflow, we performed high-throughput calculations for a large set of materials (5099) belonging to many different material classes, and collected the results to a database. Finally, the contents of database are analyzed and the calculated spectra are shown to agree well with the experimental ones.

An improved composition design method for high-performance copper alloys based on various machine learning models

The preparation of high-performance copper alloys generally considers alloying approaches to solve the conflicting problems of high strength and high electrical conductivity. The traditional “trial and error” research model is complicated and time-consuming. With the continuous accumulation of material databases and the advent of the “big data” era, machine learning has rapidly become a powerful tool for material design and development. In this paper, a total of 407 copper alloy data were collected. In the multi-objective prediction problem, the many-to-many prediction using back propagation neural network alone is improved to a many-to-one prediction. This improvement is based on back propagation neural network, tree model and support vector machine model. Through comparative analysis, an improved composition to property model was developed to predict the tensile strength and electrical conductivity of copper alloys, and the overall coefficient of determination reached 0.98; an improved property to composition model was developed to predict the composition of copper alloys, and the overall coefficient of determination reached 0.78. By combining these two models and the particle swarm optimization algorithm, an improved machine learning design system (MLDS) model was developed to achieve the composition prediction of copper alloy. The overall coefficient of determination reached 0.87, the prediction effect was better than the original MLDS model and with stronger stability. This method is of guiding significance for the alloy composition design of high-performance copper alloys. In addition, it also has certain reference value for the alloy composition design of other alloys.