Computational Materials Sciences (CMS) - Public Abstracts

MLAMD - Ames National Laboratory (PI Cai-Zhuang Wang)

Accelerating the discovery and design of advanced functional materials using AI/ML and exascale computing 

This center will develop exascale-capable computational codes and workflows that integrate materials theories and methods with AI/ML tools, materials databases, and the software stack developed through the Exascale Computing Program (ECP). The goal is to enable dramatic speed up in the prediction of new materials in new composition-structure-property spaces, and the identification of synthetic pathways for new materials. The center will also produce AI/ML integrated and validated open-source, public-access community codes, and associated databases on exascale computing facilities, enabling science-based predictive design and discovery of functional materials that otherwise would be impractical or impossible to investigate timely. This project will also significantly develop the workforce for materials research and development in exascale computing and AI/ML era.

High-performance functional materials are critical for advanced technology innovation and sustainable development. Recent advances in artificial intelligence/machine learning (AI/ML), data infrastructures, and exascale computing hardware/software, offer exciting opportunities to develop new transformative strategies to significantly speed up the pace of materials design and discovery. Specifically, we will focus on advancing an AI/ML-assisted materials discovery and design framework, and on computational prediction of synthesis pathways.

This center aligns well with the Materials Genome Initiative (MGI) goal and supports DOE’s missions on Exascale Computing and Artificial Intelligence for Science. We will discovery and design (with experimental validation) several novel ternary or quaternary compounds with unique and desirable magnetic or superconducting properties to demonstrate the power of AI/ML and exascale computing in significant reducing the time-to-solution for clean energy and advanced technology innovations.

MICCoM - Argonne National Laboratory (PI Giulia Galli)

Midwest Integrated Center for Computational Materials (MICCoM)

MICCoM develops and disseminates interoperable computational tools – open-source software, data, simulation methods, and validation procedures – that enable the community to simulate and predict the properties of functional materials. The focus is on two areas:  solid-state materials for quantum technology applications and materials on low power electronics.  The former is a class of materials with potential impact in designing new qubits, quantum sensors, and quantum communication’s devices. The latter class of materials is related to the field of microelectronics, with potential, long term impact on semiconductor manufacturing. 

The center’s scientific strategy is built under the premise that the functionality of materials depends critically on the integration of dissimilar, often defective components and on the interfaces that arise between them.  Hence the emphasis is placed on understanding and characterizing with atomic precision defects and interfaces, and on predicting finite-temperature spectroscopic and coherence properties, to understand the interaction of materials with light and external fields.  The description of dissimilar and defective components requires the development of first-principles electronic-structure methods, coupled to appropriate dynamical descriptions of matter and advanced sampling techniques, in order to predict multiple properties of complex systems and to capture the relevant length and time scales of importance to materials design. A key hallmark of MICCoM is the coupling of classical and quantum codes, together with their use on several high-performance computing (HPC) peta- and exascale architectures and, since recently, also on quantum computers. MICCoM has been engaged since its inception in verification, validation, and benchmarking activities for multiple properties of materials and is fostering close collaborations between theory and experiments.

Coordination and collaboration on HPC computational research for precision design of materials between universities and national laboratories will be continued. MICCoM’s ccommunity outreach activities, in addition to documentation of codes and use cases made available to the public on a regular basis, include the organization of schools and workshops.

COMSCOPE - Brookhaven National Laboratory (PI Gabriel Kotliar)

Comscope: Center for Computational Design of Functional Strongly Correlated Materials and Theoretical Spectroscopy

Strongly correlated materials hold a promise of revolutionary functionalities ranging from energy transmission to superior thermoelectric performance. However, to understand the properties and functionalities of strongly correlated materials is a difficult task. Standard analytical tools are not well suited for the study of these materials – these were instead designed to understand materials when interactions are weak, and the single electron picture is adequate to their description. While strongly correlated materials can be understood, it is always an arduous, time intensive task. This in turn makes material discovery difficult if not impossible. What is then needed are tools by which a user can rapidly and easily characterize strongly correlated materials and so reveal their possible functionalities.

Comscope, the BNL-led Computational Materials Science Center, has led the way in developing such tools. In the past funding cycle, it has developed three software packages by which a user can characterize the properties of a strongly correlated material. It has released FlapwMBPT, an all electron, LAPW, first-principles code, Comsuite, a package built on the principles of dynamical mean field theory, and EDRIXS, a package able to describe RIXS spectroscopies and so is of particular relevance to the BNL synchrotron, National Synchrotron Light Source II (NSLS II). These software packages are community software: they are open source and released under a GPL license.

We will extend these tools so that they are more user-friendly and are able to better account for strong correlations in a wider body of materials. Of particular interest is the thermopower in strongly correlated materials such as FeSb2, a material which has a record thermopower at low temperatures, which will require a multiscale effort. We are thus developing a quantum molecular dynamics approach that employs our methodologies for strongly correlated electron systems.

The software that we are developing will see extensive experimental validation. Our team thus includes scientists able to synthesize and characterize a wide variety of strongly correlated materials. It is an important part of our validation program to tie into the community at NSLS II to understand x-ray spectroscopies.

C2SEPEM - Lawrence Berkeley National Laboratory (PI Steven Louie)

Center for Computational Study of Excited-State Phenomena in Energy Materials

The mission of the Center for Computational Study of Excited-State Phenomena in Energy Materials (C2SEPEM) is to develop new ab initio theories, methods, algorithms, and open-source software to elucidate and predict excited-state phenomena.  Excited-state phenomena (such as electron transport and optical response) in materials typically give rise to their defining attributes and determine their usefulness.  Computational discovery and design of functional materials utilizing excited-state properties require accurate and predictive descriptions of quantum many-body interactions as well as the correlation and coherence of excitations, both in- and out-of-equilibrium. To achieve this goal, C2SEPEM aims to advance the frontiers of fundamental theories and methods, their efficient and scalable implementation utilizing exascale and post-exascale high-performance computing resources, and their experimental validations, through efforts of a diverse and multidisciplinary team consisting of physical scientists, applied mathematicians, and computational scientists.  Studies on quasiparticle excitations, optical spectra, correlated multi-particle excitations, couplings of phonons with electrons and excitons, nonlinear optical processes, field-driven time-dependent phenomena, and more are carried out for bulk and reduced-dimensional systems.  These phenomena, which are central to many fields within physics, chemistry, materials science, and engineering, are particularly important in processes of energy generation, transport, and storage.  Ab initio approaches based on field-theoretical methods, advanced numerical methods and algorithms, as well as high-performance computing, are developed with the help of machine learning and artificial intelligence techniques.  The planned activities develop widely applicable open-access community codes and associated databases for the discovery and design of functional materials with unique properties.  In addition to expanding the frontiers of knowledge, an end product is an integrated open-source software package, BerkeleyGW, with advanced capabilities to predict and understand excited-state phenomena from first principles for a broad range of material systems.

NPNEQ - Lawrence Livermore National Laboratory (PI Tadashi Ogitsu)

Nonperturbative Studies of Functional Materials Under Nonequilibrium Conditions

The Center for Nonperturbative Studies of Functional Materials Under Nonequilibrium Conditions (NPNEQ) is designed to facilitate advancement of ultrafast science via development and dissemination of novel quantum dynamics software tools. Such open-source software tools include real-time time-dependent density functional theory code, for quantum mechanical time evolution of spins, electrons and ions relevant to a wide range of applications, including switching, memory, and optoelectronic devices, as well as new materials processing methods. The developed community software tools support ab initio, non-perturbative studies far from equilibrium. These quantum dynamics simulation software tools are being validated through collaborative scientific research on the phenomena where quantum mechanical spin-electron-ion-photon coupling plays crucial roles.

Most of the traditional quantum dynamics simulation methods and software available today provide solutions for time-independent version of the quantum mechanical Schrödinger equation by mapping it to eigenvalue problems for electrons, solved for a given atomic structure (i.e. adiabatic approximation or Born-Oppenheimer approximation). This level of theory has been the foundation of the currently available technologies such as the semiconductor devices used in von Neumann computer architectures, however, development of truly time-dependent quantum theory and the computer simulation tools based on such a theory can play crucial roles for the advancement of the next-generation of opto-electronic devices (for example, simulating decoherence in quantum computers).

NPNEQ intends to move the needle by developing novel methodologies and providing software tools to study truly time-dependent quantum mechanical problems. The software is designed to fully leverage state of the art high-performance computers to address increased computational cost for higher level of approximation. The novel methodologies will be validated using the state-of-the-art ultrafast experimental capabilities being developed at SLAC/Stanford University.

CPSFM - Oak Ridge National Laboratory (PI Paul Kent)

Center for Predictive Simulation of Functional Materials

Meeting the challenge of computational discovery, fundamental understanding, and design of real functional materials with unique physical properties ultimately relies on the availability of sufficiently reliable and predictive computational methods.  Achieving the desired predictiveness for general materials remains highly challenging, particularly where strong electron correlations, spin-orbit interactions, and van der Waals interactions couple with the atomic structure of the materials. At the same time, these interactions can result in novel physical properties that are highly desired for energy applications. The mission to the Center for Predictive Simulation of Functional Materials is therefore to provide a leap forward in state-of-the-art methods that can make highly accurate and reliable quantum-mechanics-based predictions in these classes of materials.

To achieve this goal, the Center focuses on advances in Quantum Monte Carlo (QMC) techniques, their implementation in the open-source QMCPACK code, and the workflow software and supporting data required for broad utilization. Validation and initial scientific application of these new methods will be to materials and properties where existing theoretical and computational methods are not predictive, including quantum materials such as challenging layered magnets and layered materials that are proposed to exhibit novel quantum phases. QMCPACK has been expressly designed to run performance portably from laptops up to exascale machines such as Frontier at OLCF and Aurora at ALCF. The combination of improved QMC techniques and advanced computational implementation is expected to enable application to new condensed matter systems as well as enable higher-throughput uses for machine learning and artificial-intelligence based applications.

The team consists of condensed matter theory, materials, and high-performance computing staff, professors, students and postdoctoral research associates at Argonne, Sandia, and Oak Ridge National Laboratories, and North Carolina State and Brown Universities.

HeteroFAM – Pacific Northwest National Laboratory (PI Eric Bylaska)

Navigating the Design Space of Heterostructures: Advancing Functionality of Modeling for Two-Dimensional Materials and Transition Metal Oxides

The materials design space for groundbreaking advancements in technology and industry is vast and complex. This project aims to enhance and innovate the design of new functional materials, such as graphene and other two-dimensional materials with emergent electronic properties and metal oxides with specific magnetic structures, through the development of novel software tools. Understanding and harnessing emergent electronic states could create more efficient semiconductors and advanced sensors. Tailoring atomic-level interactions would give rise to novel magnetic properties for applications in advanced data storage, magnetic sensors, and spintronic devices.

Our goal is to create user-friendly interfaces, enabling the research community from various backgrounds and skill levels to use these advanced tools for material design and discovery, ensuring more contributions to scientific advancement. We are developing powerful materials modeling software that runs on some of the world’s fastest supercomputers. Our software uses artificial intelligence and advanced electronic structure methods to predict materials properties and behaviors. Through the development of web applications and large databases, our software will make it easier for scientists and engineers to find and design materials with the best properties for their needs.

COMMS - Pennsylvania State University (PI Long-Qing Chen)

COMMS: Center for COmputational Mesoscale Materials Science 

Objectives: The central goal of the Center is to develop mesoscale computational models, efficient numerical algorithms for exascale computation, and software validated by experiments for quantum and functional materials. The focus is on distilling and translating the essential physics of strong electron correlation, topological spin, charge, orbital and lattice textures, and dynamical phenomena on the ultrafast time scales into phase-field phenomenology towards predicting emergent mesoscale quantum orders and pattern formations from femtosecond-to-equilibrium time scales. Two specific objectives of the renewal proposal are to: (1) Significantly expand the scope and capability of mesoscale computational models based on the phase-field method for understanding and predicting the formation and dynamical evolution of mesoscale structures in systems exhibiting complex polar and magnetic textures, metal-insulator transitions, superconductivity, and quantum Hall effect; and (2) Develop new software modules and tools that will be deployed in the open-source environment under Q-POP(Quantum Phase-Field Open-Source Package) and parallelized to enable petascale and exascale computing for understanding mesoscale quantum phenomena towards accelerating materials insertion into next-generation neuromorphic computing chips, terahertz spintronic and magnonic devices, and superconducting qubits for quantum memories and transducers.

Research and methods: Mesoscale science of quantum materials is a new frontier for the applications of the phase-field method. Built upon the accomplishments of the current award period, the specific efforts of the renewal proposal are to: (1) Further extend our dynamic phase-field model (DPFM) of coupled structural and electronic carrier dynamics to photon dynamics through the study of the formation and responses of mesoscale structures to external mechanical and electromagnetic fields under the influence of the electronic system such as light-excited electronic carriers; (2) Construct a set of novel phase-field models of coupled phase transitions such as metal-insulator transitions, magnetic phase transitions, and superconducting phase transitions to study mesoscale electronic and structural pattern formation and evolution under the influence of lattice strain and chemical doping; (3) Develop and deploy the main modules of Q-POP, namely, Q-POP-IMT, Q-POP-FerroDyn, and Q-POP-SuperCon, as well as a number of powerful input/output (I/O), data-generation, and mesostructure characterization tools through annual workshops on phase-field method and software for quantum materials; and (4) Experimentally validate and iteratively refine the theory and computational tools using atomic-scale controlled materials synthesis in tandem with cutting-edge quantum characterization methods.

The project brings together an interdisciplinary team of experts in mesoscale phase-field method and model development, thermodynamics from electronic-structure calculations, thermodynamic theory of phase transitions, advanced numerical algorithms and peta-/exascale implementation, experimental characterization of mesostructures and their responses, phase-field modeling of ferroic and quantum devices, and analytical and numerical thermodynamic calculations of coupled electronic and structural phase transitions including their size effect. The project will involve extensive collaborations with experts outside the core team on dynamical mean field theory, crystal growth, and experimental characterization of mesoscale structures of quantum materials at several DOE Labs and academic institutions.

Outcome and Impacts: An outcome is an experimentally validated software package, Q-POP, parallelized to enable petascale and exascale computing for understanding and predicting the mesostructures of quantum and functional materials and their responses to external stimuli towards designing device architectures for harnessing these functionalities.

QMC-HAMM - University of Illinois, Urbana-Champaign (PI Lucas Wagner)

QMC-HAMM: High accuracy multiscale models using quantum Monte Carlo

This project is centered on the development of high-accuracy multiscale models using quantum Monte Carlo (QMC-HAMM), a set of community-serving software, models, and data that enable one to link highly accurate many-electron microscopic quantum simulations with multiscale modeling that can achieve large length and time scales. The main objective is to derive high accuracy coarse-grained models from the quantum mechanics calculations, while also assessing uncertainty in those models. Because the underlying data is computational, machine learning techniques can be applied to learn coarse-grained models for the materials under study from the computational data itself. Here we move beyond density functional theory and use QMC as the underlying data source, which improves the accuracy of the derived models. In contrast to density functional theory, QMC simulates the correlations between electrons explicitly, which results in more accurate results.

QMC-HAMM is focused on developing tools, and data to address the multiscale problem. There are three main classes of tools that are developed in this work: (1) automated software to manage complex quantum Monte Carlo workflows and data, (2) model derivation software that also evaluates the uncertainty in the software from the statistical data available from QMC and publishes the models in industry-standard formats, and (3) an intelligent data selector that directs the automated software in (1) to gather data that will most reduce the uncertainty of the models in (2). The software developed is modular, open source, and publicly distributed, and can be mixed and matched by other researchers. The data produced from (1) is published and indexed. While the tools are applicable to any material system, there are two main applications of interest. The first is hydrogen at high pressure, which is predicted to superconduct at high temperatures and is present in the Jovian planets and exoplanets. The second is 2D material systems, such as twisted bilayer graphene, which is a uniquely well-controlled system in which the twist angle controls the effective length scale and is an excellent test bed for the algorithms. For both systems, the models developed will describe all useful degrees of freedom at a coarse-grained level. This complete description will allow exploration of phase diagrams and assignment of clear origins from the observed physical behavior.

The work will impact several fields. At the smallest length scale, the workflow and software stack can be used to develop high accuracy models. Moving to a slightly larger length scale, the underlying data is used by researchers working on better quantum mechanical calculations; these data are state-of-the-art reference values. At the largest length scale, the models derived here are more reliable than others in the literature and include evaluated uncertainties, and so should be preferred when studying twisted bilayer graphene or hydrogen at high pressure. By publishing the models in a Python package that can interface with many other software packages, these high accuracy models are made accessible. Because the models have all relevant degrees of freedom, they can provide insight into the origin of the complex phase diagrams and emergent behavior in these systems.

EPW - University of Texas, Austin (PI Feliciano Giustino)

Toward Exascale Computing of Electron-Phonon Couplings for Finite-Temperature Materials Design

The overarching aim of this effort is to enable accurate, fast, scalable, and reproducible ab initio calculations of electron-phonon couplings for the design of advanced functional materials, and to make these tools widely accessible to the computational materials science community.

In semiconductors, metals, insulators, and superconductors, the vibrations of the crystal lattice can have a significant impact on their electronic properties. For example, under standard operating conditions, the electrical resistivity of semiconductors in microelectronic devices increases with temperature because electrons experience increased scattering from thermal vibrations of the atoms. This phenomenon leads to electronic devices heating up when performing intensive compute tasks, something that we are all very familiar with. At the microscopic scale, these processes can be understood in terms of electrons exchanging energy with the crystalline lattice in the form of vibration quanta called phonons. Beyond semiconductor devices, the coupling between electrons and phonons influences myriad other materials properties and functionalities, including the absorption of light in solar cells, the emission of light in LEDs, the dissipation of excess heat by thermal conductors, and even the entanglement of qubits in quantum computers. Therefore, the ability to compute electron-phonon couplings with predictive accuracy is key to design and develop a variety of advanced materials and devices for applications in energy conversion and storage, solid-state lighting, energy-efficient electronics, and quantum technologies.

Here, we will develop innovative ab initio methods, algorithms, and software to investigate an array of materials properties that are beyond the reach of existing computational methods; we will consolidate and expand the EPW code, a large open-source software project for calculating electron-phonon couplings and related materials properties; and we will progress toward exascale computing by harnessing the power of leadership-class DOE supercomputers. We will push the frontiers of high-performance computing of electron-phonon couplings for predictive materials design. We will develop advanced many-body approaches to investigate the formation, energetics, and dynamics of small and large polarons. We will develop Monte Carlo simulators and real-time solvers to predict the carrier transport properties of semiconductors under high-field and out-of-equilibrium conditions, as well as methods to investigate magneto-transport in topological materials. Furthermore, we will continue refactoring the EPW code to leverage many-core architectures and GPU acceleration on DOE supercomputers, and we will simplify the calculation workflows to enable more systematic investigations of electron-phonon couplings. We will strive to increase accessibility, transparency, and reproducibility of these calculations by making all new developments fully and promptly available to the scientific community, through a six-months release cycle under GPL open-source license. To support the growth of data-driven materials research, we will adopt portable, descriptive, and AI-ready data formats, and we will invest into training of users and developers of ab initio electronic structure software.

Overall, this project will accelerate the pace of discovery in materials research by making advanced ab initio many-body calculations of electronic, optical, and transport properties of materials more widely accessible, by leveraging the power of DOE exascale supercomputers, and by enabling predictive design of materials for energy, microelectronics, and quantum technologies.