Accelerate Materials Development Research Articles

Moving toward autonomous manufacturing by accelerating hydrodynamic fabrication of microstructures using deep neural networks

Manufacturing of microstructures using a microfluidic device is a largely empirical effort due to the multi-physical nature of the fabrication process. As such, in moving toward autonomous manufacturing, models are desired that will predict microstructure attributes (e.g., size, porosity, and stiffness) based on known inputs, such as sheath and core fluid flow rates. Potentially more useful is the prospect of inputting desired microfiber features into a design model to extract appropriate manufacturing parameters. In this study, we demonstrate that deep neural networks (DNNs) trained with sparse datasets augmented by synthetic data can produce accurate predictive and design models to accelerate materials development. For our predictive model with known sheath and core flow rates and bath solution percentage, calculated solid microfiber dimensions are shown to be greater than 95% accurate, with porosity and Young's modulus exhibiting greater than 90% accuracy for a majority of conditions. Likewise, the design model is able to recover sheath and core flow rates with 95% accuracy when provided values for microfiber dimensions, porosity, and Young's modulus. As a result, DNN-based modeling of the microfiber fabrication process demonstrates high potential for reducing time to manufacture of microstructures with desired characteristics.

μ2mech: A software package combining microstructure modeling and mechanical property prediction

We have developed a graphical user interface (GUI) based package μ2mech to perform phase-field simulation for predicting microstructure evolution. The package can take inputs from ab initio calculations and CALPHAD (Calculation of Phase Diagrams) tools for quantitative microstructure prediction. The package also provides a seamless connection to transfer output from the mesoscale phase field method to the microscale finite element analysis for mechanical property prediction. Such a multiscale simulation package can facilitate microstructure-property correlation, one of the cornerstones in accelerated materials development within the integrated computational materials engineering (ICME) framework.

(Invited) accelerating Polymer Discovery for Multifunctional Polymer Thin Films: High-Throughput Characterization

Identifying new formulations of polymers that have multiple highly tuned properties is a grand challenge of materials development. This is especially important in polymer thin film devices where several distinct transport properties must be controlled. For instance, solid-state battery electrolyte materials must be conductive to ions, but remain electronically insulating. In order to achieve the materials advances needed to transform the use of such polymer thin films, new paradigms of accelerated materials development are needed.In this talk, we discuss our recent efforts to accelerate the development of advanced polymer thin film materials for energy applications. A crucial feature of accelerating the discovery of high performing formulations and processing conditions is a rapid means to assess the properties of polymer films with spatial resolution commensurate with nanoscopic defects and throughput that can accommodate macroscopic devices. We initially focus on the challenge of verifying that films are electrically insulating. Maintaining electrical insulation across large films is a particularly insidious challenge as small “pin-hole” defects can result in an entire device failing. We explore an optical method to detect and characterize conductive defects based upon the electrochemiluminescence of luminol. In particular, when a voltage is applied to a conductive surface that is protected by a nominally insulating film, regions where the film is not present or is electrically conductive will result in the local generation of light. A major virtue of this approach is that the minimum detectable feature is not directly determined by the wavelength of light or the optical magnification in use, but rather the intensity of light generated per unit area on the surface. To explore this approach, we conduct an extensive series of optimization experiments to maximize the intensity of light generated by electrochemiluminescence by adjusting the timing, voltage, and solution properties. The result of this optimization is that lines narrower than 100 nm are readily detectable using this method. Finally, we present efforts to use this approach to screen centimeter-square areas to find and characterize nanoscale defects.Looking beyond the important task of characterizing films at high throughput, we conclude the talk with a discussion of how this approach fits in a broader materials discovery pipeline. In particular, recent years have seen a number of examples of autonomous experimentation systems that combine automation to perform experiments and machine learning to select experiments to vastly increased the speed at which knowledge is gained. The reported high-throughput analytical process is anticipated to form an important link in this process by providing a path to rapidly and effectively screening films for electrical properties.

Combinatorial screening via high-throughput preparation: Thermoelectric performance optimization for n-type Bi–Te–Se film with high average zT > 1

Thermoelectric materials have drawn extensive interest due to the direct conversion between electricity and heat, however, it is usually a time-consuming process for applying traditional “sequential” methods to grow materials and investigate their properties, especially for thermoelectric films that typically require fine microstructure control. High-throughput experimental approaches can effectively accelerate materials development, but the methods for high-throughput screening of the microstructures require further study. In this work, a combinatorial high-throughput optimization solution of material properties is proposed for the parallel screening and optimizing of composition and microstructure, which involves two distinctive types of high-throughput fabrication approaches for thin films, along with a new portable multiple discrete masks based high-throughput preparation platform. Thus, Bi2Te3–xSex thin film library with 196 throughputs for locating the optimized composition is obtained in one growth cycle. In addition, another thin film library composed of 31 materials with traceable process parameters is built to further investigate the relationship between microstructure, process, and thermoelectric performance. Through high-throughput screening, the Bi2Te2.9Se0.1 film with (00l) orientation is prepared with a peak zT value of 1.303 at 353 K along with a high average zT value of 1.047 in the interval from 313 to 523 K. This method can be also extended to the discovery of other functional thin films with a rapid combinatorial screening of the composition and structure to accelerate material optimization.

Serial rotation electron diffraction for rapid phase analysis to accelerate materials development.

Serial rotation electron diffraction for rapid phase analysis to accelerate materials development.

A Materials Acceleration Platform for Organic Laser Discovery.

Conventional materials discovery is a laborious and time-consuming process that can take decades from initial conception of the material to commercialization. Recent developments in materials acceleration platforms promise to accelerate materials discovery using automation of experiments coupled with machine learning. However, most of the automation efforts in chemistry focus on synthesis and compound identification, with integrated target property characterization receiving less attention. In this work, an automated platform is introduced for the discovery of molecules as gain mediums for organic semiconductor lasers, a problem that has been challenging for conventional approaches. This platform encompasses automated lego-like synthesis, product identification, and optical characterization that can be executed in a fully integrated end-to-end fashion. Using this workflow to screen organic laser candidates, discovered eight potential candidates for organic lasers is discovered. The lasing threshold of four molecules in thin-film devices and find two molecules with state-of-the-art performance is tested. These promising results show the potential of automated synthesis and screening for accelerated materials development.

Topological feature engineering for machine learning based halide perovskite materials design

Accelerated materials development with machine learning (ML) assisted screening and high throughput experimentation for new photovoltaic materials holds the key to addressing our grand energy challenges. Data-driven ML is envisaged as a decisive enabler for new perovskite materials discovery. However, its full potential can be severely curtailed by poorly represented molecular descriptors (or fingerprints). Optimal descriptors are essential for establishing effective mathematical representations of quantitative structure-property relationships. Here we reveal that our persistent functions (PFs) based learning models offer significant accuracy advantages over traditional descriptor based models in organic-inorganic halide perovskite (OIHP) materials design and have similar performance as deep learning models. Our multiscale simplicial complex approach not only provides a more precise representation for OIHP structures and underlying interactions, but also has better transferability to ML models. Our results demonstrate that advanced geometrical and topological invariants are highly efficient feature engineering approaches that can markedly improve the performance of learning models for molecular data analysis. Further, new structure-property relationships can be established between our invariants and bandgaps. We anticipate that our molecular representations and featurization models will transcend the limitations of conventional approaches and lead to breakthroughs in perovskite materials design and discovery.

Experimental-numerical analysis of microstructure-property linkages for additively manufactured materials

The innovation of new or improved products fabricated from additive manufacturing processes with desired properties depends on a multitude of trials as stated by the National Science and Technology Council (2011). Therefore, a systematic approach is essential to accelerate materials development. This can be realised by developing systematic materials knowledge in the form of process-structure-property relationships. In this envisioned framework, the present work aims to derive the structure-property linkages of additively manufactured Ti-6Al-4V alloy. The main focus is to investigate the influence of potential defects, in the form of pores, inherited from the fabrication process on the fatigue properties. The complicated polycrystalline microstructure, including porosity at a microscale, is obtained by processing light microscopy and x-ray computed tomography measurements. A detailed statistical analysis is performed to obtain a low-dimensional representation of the structure. Based on these statistical measures, a suitable reconstruction algorithm is developed to create pore distributions that are incorporated into synthetic statistical volume elements (SVEs) generated from DREAM.3D by Groeber and Jackson (2014). Using these SVEs, microscale crystal plasticity simulations in DAMASK, see Roters et al. (2019), are performed to obtain the material properties such as yield strength and fatigue indicator parameters (FIPs). A detailed numerical analysis is carried out to study the influence of pore statistics such as size distribution or porosity fraction. Data analysis is carried out to rank-order the SVEs based on FIPs. Furthermore, a comparison with Murakami’s empirical square root area concept is made.

(Invited) Hydrogen Materials Advanced Research Consortium R&D Activities

The Hydrogen Materials—Advanced Research Consortium (HyMARC) is a National Lab-led consortium working to advance and develop hydrogen storage materials that meet DOE targets for energy density, kinetics and cost for a wide range of applications. The goal of HyMARC is to enable a doubling of the hydrogen storage density compared with state-of-the-art compressed storage systems. DOE’s Hydrogen and Fuel Cell Technologies Office launched HyMARC in 2015, bringing together world-class National Lab capabilities in materials modelling, characterization and synthesis; and fostering collaboration with industry and academia to overcome R&D challenges in materials-based hydrogen storage. The consortium is led by teams at Sandia National Laboratories and the National Renewable Energy Laboratory and includes groups at Lawrence Livermore, Lawrence Berkeley, and Pacific Northwest National Laboratories. HyMARC is working to develop foundational understanding of phenomena governing thermodynamics and kinetics of hydrogen release and uptake in all classes of hydrogen storage materials, including metal hydrides, adsorbents, and hydrogen carriers. It seeks to accelerate materials development by addressing critical R&D gaps in hydrogen storage—leveraging recent advances in predictive multiscale modeling, machine learning, high-resolution in situ characterization, and novel materials synthesis techniques to realize their goal to enable twice the energy density for hydrogen storage in the coming years. This presentation will describe HyMARC’s ongoing R&D activities and past accomplishments which have pushed the bar forward for materials-based hydrogen storage.

Machine Learning-Driven Biomaterials Evolution.

Biomaterials is an exciting and dynamic field, which uses a collection of diverse materials to achieve desired biological responses. While there is constant evolution and innovation in materials with time, biomaterials research has been hampered by the relatively long development period required. In recent years, driven by the need to accelerate materials development, the applications of machine learning in materials science has progressed in leaps and bounds. The combination of machine learning with high-throughput theoretical predictions and high-throughput experiments (HTE) has shifted the traditional Edisonian (trial and error) paradigm to a data-driven paradigm. In this review, each type of biomaterial and their key properties and use cases are systematically discussed, followed by how machine learning can be applied in the development and design process. The discussions are classified according to various types of materials used including polymers, metals, ceramics, and nanomaterials, and implants using additive manufacturing. Last, the current gaps and potential of machine learning to further aid biomaterials discovery and application are also discussed.

Perspectives on multiscale modelling and experiments to accelerate materials development for fusion

Prediction of material performance in fusion reactor environments relies on computational modelling, and will continue to do so until the first generation of fusion power plants come on line and allow long-term behaviour to be observed. In the meantime, the modelling is supported by experiments that attempt to replicate some aspects of the eventual operational conditions. In 2019, a group of leading experts met under the umbrella of the IEA to discuss the current position and ongoing challenges in modelling of fusion materials and how advanced experimental characterisation is aiding model improvement. This review draws from the discussions held during that workshop. Topics covering modelling of irradiation-induced defect production and fundamental properties, gas behaviour, clustering and segregation, defect evolution and interactions are discussed, as well as new and novel multiscale simulation approaches, and the latest efforts to link modelling to experiments through advanced observation and characterisation techniques.

On the use of non-destructive, gigahertz ultrasonics to rapidly screen irradiated steels for swelling resistance

Transient grating spectroscopy (TGS), a non-contact ultrasonic materials analysis technique, is proposed to rapidly and indirectly assess relative void swelling resistance of multiple structural materials. Statistically significant changes in the frequency of probed surface acoustic waves (SAWs) suggest that newly developed steels containing nanosized precipitates show higher resistance to void swelling when compared to their simpler, commercial analogues. The higher reduction in SAW frequency seen in the simpler steels, proportional to porosity, indicates more void formation which is directly validated by TEM examinations. This example illustrates the minimum set of targeted TGS studies required to quickly and inexpensively rank materials by relative void swelling resistance, and hence, accelerate materials development and characterization.

Microfluidic Synthesis of Semiconductor Materials: Toward Accelerated Materials Development in Flow

AbstractControlled synthesis of semiconductor nano/microparticles has attracted substantial attention for use in numerous applications from photovoltaics to photocatalysis and bioimaging due to the breadth of available physicochemical and optoelectronic properties. Microfluidic material synthesis strategies have recently been demonstrated as an effective technique for rapid development, controlled synthesis, and continuous manufacturing of solution‐processed semiconductor nano/microparticles, due to enhanced parametric control enabling precise tuning of material properties, size, and morphologies. In this review, the basics of microfluidic material synthesis approaches complemented with recent advances in the flow fabrication of metal oxide, chalcogenide, and perovskite semiconductor particles are discussed. Furthermore, advancements in artificial intelligence (AI)‐driven materials–space exploration and accelerated formulation optimization using modular microfluidic reactors are outlined. Finally, future directions for the fabrication of semiconducting materials in flow and the implementation of AI with automated microfluidic reactors for accelerated material discovery and development are presented.

Machine-learning-driven discovery of polymers molecular structures with high thermal conductivity

The ability to efficiently design new and advanced polymers with functional thermal properties is hampered by the high-cost and time-consuming experiments. Machine learning is an effective approach that can accelerate materials development by combining material science and big data techniques. Here, machine learning methods were used to predict the thermal conductivity of various single-chain polymers, and the relationship between molecular structures of polymer repeating units and thermal conductivity was also been investigated. The predict model starts from a benchmark dataset generated by large-scale molecular dynamics computations. In predict models, the polymers were ‘fingerprinted’ as simple, easily attainable numerical representations, which helps to develop an on-demand property prediction model. Further, potential quantitative relationship between molecular structures of polymer and thermal conductivity property was analyzed, and hypothetical polymers with ideal thermal conductivity were identified. The methods are shown to be general, and can hence guide the screening and systematic identification of high thermal conductivity.

Role of uncertainty estimation in accelerating materials development via active learning

An active learning strategy using sampling based on uncertainties shows the promise of accelerating the development of new materials. We study the efficiencies of the active learning iteration loop with different uncertainty estimators to find the “best” material in four different experimental datasets. We use a bootstrap approach aggregating with support vector regression as the base learner to obtain uncertainties associated with model predictions. If the bootstrap replicate number B is small, the variance estimated by the empirical standard error estimator is found to be close to the true variance, whereas the jackknife based estimators give an upward or downward biased estimation of variance. As B increases, the bias of the jackknife based estimators decreases and the variance estimated finally converges to the true one. Therefore, the empirical standard error estimator needs the least number of iteration loops to find the best material in the datasets, especially when the bootstrap replicate number B is small. Our work demonstrates that an appropriate Bootstrap replicate B is conducive to minimizing calculation costs during the materials property optimization by active learning.

Machine learning: Accelerating materials development for energy storage and conversion

AbstractWith the development of modern society, the requirement for energy has become increasingly important on a global scale. Therefore, the exploration of novel materials for renewable energy technologies is urgently needed. Traditional methods are difficult to meet the requirements for materials science due to long experimental period and high cost. Nowadays, machine learning (ML) is rising as a new research paradigm to revolutionize materials discovery. In this review, we briefly introduce the basic procedure of ML and common algorithms in materials science, and particularly focus on latest progress in applying ML to property prediction and materials development for energy‐related fields, including catalysis, batteries, solar cells, and gas capture. Moreover, contributions of ML to experiments are involved as well. We highly expect that this review could lead the way forward in the future development of ML in materials science.image

Economically Sustainable Growth of Perovskite Photovoltaics Manufacturing

Summary The significant capital expense of photovoltaics manufacturing has made it difficult for new cell and module technologies to enter the market. We present two technoeconomic models that analyze the sustainable growth of perovskite manufacturing for an R2R single-junction technology and a perovskite-silicon tandem module, focusing on the impacts of economies of scale and average selling price on profitability. We establish a cost range of $3.30/W to $0.53/W for flexible modules manufactured in factory sizes ranging from 0.3 MW/year to 1 GW/year. In addition, we model the cost to manufacture a tandem module consisting of a single-junction perovskite cell stacked in 4-terminal configuration onto a silicon cell and show how an existing manufacturer can grow at a faster rate by co-investing in tandems. Our analyses highlight potential routes to market for perovskite photovoltaics and the possibility to sustainably grow a photovoltaics manufacturing company even in markets with higher labor rates.

Predicting material properties by integrating high-throughput experiments, high-throughput ab-initio calculations, and machine learning

ABSTRACT High-throughput experiments (HTEs) have been powerful tools to obtain many materials data. However, HTEs often require expensive equipment. Although high-throughput ab-initio calculation (HTC) has the potential to make materials big data easier to collect, HTC does not represent the actual materials data obtained by HTEs in many cases. Here we propose using a combination of simple HTEs, HTC, and machine learning to predict material properties. We demonstrate that our method enables accurate and rapid prediction of the Kerr rotation mapping of an FexCoyNi1-x-y composition spread alloy. Our method has the potential to quickly predict the properties of many materials without a difficult and expensive HTE and thereby accelerate materials development.

Inverse design of porous materials using artificial neural networks.

Generating optimal nanomaterials using artificial neural networks can potentially lead to a notable revolution in future materials design. Although progress has been made in creating small and simple molecules, complex materials such as crystalline porous materials have yet to be generated using any of the neural networks. Here, we have implemented a generative adversarial network that uses a training set of 31,713 known zeolites to produce 121 crystalline porous materials. Our neural network takes in inputs in the form of energy and material dimensions, and we show that zeolites with a user-desired range of 4 kJ/mol methane heat of adsorption can be reliably produced using our neural network. The fine-tuning of user-desired capability can potentially accelerate materials development as it demonstrates a successful case of inverse design of porous materials.

Using artificial intelligence to accelerate materials development

Using artificial intelligence to accelerate materials development