We present the Materials Learning Algorithms (MALA) package, a scalable machine learning framework designed to accelerate density functional theory (DFT) calculations suitable for large-scale atomistic simulations. Using local descriptors of the atomic environment, MALA models efficiently predict key electronic observables, including local density of states, electronic density, density of states, and total energy. The package integrates data sampling, model training and scalable inference into a unified library, while ensuring compatibility with standard DFT and molecular dynamics codes. We demonstrate MALA's capabilities with examples including boron clusters, aluminum across its solid-liquid phase boundary, and predicting the electronic structure of a stacking fault in a large beryllium slab. Scaling analyses reveal MALA's computational efficiency and identify bottlenecks for future optimization. With its ability to model electronic structures at scales far beyond standard DFT, MALA is well suited for modeling complex material systems, making it a versatile tool for advanced materials research.
The properties of electrons in matter are of fundamental importance. They give rise to virtually all material properties and determine the physics at play in objects ranging from semiconductor devices to the interior of giant gas planets. Modeling and simulation of such diverse applications rely primarily on density functional theory (DFT), which has become the principal method for predicting the electronic structure of matter. While DFT calculations have proven to be very useful, their computational scaling limits them to small systems. We have developed a machine learning framework for predicting the electronic structure on any length scale. It shows up to three orders of magnitude speedup on systems where DFT is tractable and, more importantly, enables predictions on scales where DFT calculations are infeasible. Our work demonstrates how machine learning circumvents a long-standing computational bottleneck and advances materials science to frontiers intractable with any current solutions.
Abstract: Due to a beneficial balance of computational cost and accuracy, real-time time-dependent density-functional theory has emerged as a promising first-principles framework to describe electron real-time dynamics. Here we discuss recent implementations around this approach, in particular in the context of complex, extended systems. Results include an analysis of the computational cost associated with numerical propagation and when using absorbing boundary conditions. We extensively explore the shortcomings for describing electron–electron scattering in real time and compare to many-body perturbation theory. Modern improvements of the description of exchange and correlation are reviewed. In this work, we specifically focus on the Qb@ll code, which we have mainly used for these types of simulations over the last years, and we conclude by pointing to further progress needed going forward. Graphical abstract: [Figure not available: see fulltext.].
The long-standing problem of predicting the electronic structure of matter on ultra-large scales (beyond 100,000 atoms) is solved with machine learning.
The focus of this project is to accelerate and transform the workflow of multiscale materials modeling by developing an integrated toolchain seamlessly combining DFT, SNAP, LAMMPS, (shown in Figure 1-1) and a machine-learning (ML) model that will more efficiently extract information from a smaller set of first-principles calculations. Our ML model enables us to accelerate first-principles data generation by interpolating existing high fidelity data, and extend the simulation scale by extrapolating high fidelity data (102 atoms) to the mesoscale (104 atoms). It encodes the underlying physics of atomic interactions on the microscopic scale by adapting a variety of ML techniques such as deep neural networks (DNNs), and graph neural networks (GNNs). We developed a new surrogate model for density functional theory using deep neural networks. The developed ML surrogate is demonstrated in a workflow to generate accurate band energies, total energies, and density of the 298K and 933K Aluminum systems. Furthermore, the models can be used to predict the quantities of interest for systems with more number of atoms than the training data set. We have demonstrated that the ML model can be used to compute the quantities of interest for systems with 100,000 Al atoms. When compared with 2000 Al system the new surrogate model is as accurate as DFT, but three orders of magnitude faster. We also explored optimal experimental design techniques to choose the training data and novel Graph Neural Networks to train on smaller data sets. These are promising methods that need to be explored in the future.
We present a numerical modeling workflow based on machine learning (ML) which reproduces the total energies produced by Kohn-Sham density functional theory (DFT) at finite electronic temperature to within chemical accuracy at negligible computational cost. Based on deep neural networks, our workflow yields the local density of states (LDOS) for a given atomic configuration. From the LDOS, spatially-resolved, energy-resolved, and integrated quantities can be calculated, including the DFT total free energy, which serves as the Born-Oppenheimer potential energy surface for the atoms. We demonstrate the efficacy of this approach for both solid and liquid metals and compare results between independent and unified machine-learning models for solid and liquid aluminum. Our machine-learning density functional theory framework opens up the path towards multiscale materials modeling for matter under ambient and extreme conditions at a computational scale and cost that is unattainable with current algorithms.
Through a combination of single crystal growth, experiments involving in situ deposition of surface adatoms, and complimentary modeling, we examine the electronic transport properties of lithium-decorated ZrTe5 thin films. We observe that the surface states in ZrTe5 are robust against Li adsorption. Both the surface electron density and the associated Berry phase are remarkably robust to adsorption of Li atoms. Fitting to the Hall conductivity data reveals that there exist two types of bulk carriers: those for which the carrier density is insensitive to Li adsorption, and those whose density decreases during initial Li depositions and then saturates with further Li adsorption. We propose this dependence is due to the gating effect of a Li-adsorption-generated dipole layer at the ZrTe5 surface.
We present a numerical modeling workflow based on machine learning (ML) which reproduces the the total energies produced by Kohn-Sham density functional theory (DFT) at finite electronic temperature to within chemical accuracy at negligible computational cost. Based on deep neural networks, our workflow yields the local density of states (LDOS) for a given atomic configuration. From the LDOS, spatially-resolved, energy-resolved, and integrated quantities can be calculated, including the DFT total free energy, which serves as the Born-Oppenheimer potential energy surface for the atoms. We demonstrate the efficacy of this approach for both solid and liquid metals and compare results between independent and unified machine-learning models for solid and liquid aluminum. Our machine-learning density functional theory framework opens up the path towards multiscale materials modeling for matter under ambient and extreme conditions at a computational scale and cost that is unattainable with current algorithms.
The self-interstitial atom (SIA) is one of two fundamental point defects in bulk Si. Isolated Si SIAs are extremely difficult to observe experimentally. Even at very low temperatures, they anneal before typical experiments can be performed. Given the challenges associated with experimental characterization, accurate theoretical calculations provide valuable information necessary to elucidate the properties of these defects. Previous studies have applied Kohn-Sham density functional theory (DFT) to the Si SIA, using either the local density approximation or the generalized gradient approximation to the exchange-correlation (XC) energy. The consensus of these studies indicates that a Si SIA may exist in five charge states ranging from -2 to +2 with the defect structure being dependent on the charge state. This study aims to re-examine the existence of these charge states in light of recently derived "approximate bounds"on the defect levels obtained from finite-size supercell calculations and new DFT calculations using both semi-local and hybrid XC approximations. We conclude that only the neutral and +2 charge states are directly supported by DFT as localized charge states of the Si SIA. Within the current accuracy of DFT, our results indicate that the +1 charge state likely consists of an electron in a conduction-band-like state that is coulombically bound to a +2 SIA. Furthermore, the -1 and -2 charge states likely consist of a neutral SIA with one and two additional electrons in the conduction band, respectively.
Recent years have seen an explosion in research efforts discovering and understanding novel electronic and optical properties of topological quantum materials (TQMs). In this LDRD, a synergistic effort of materials growth, characterization, electrical-magneto-optical measurements, combined with density functional theory and modeling has been established to address the unique properties of TQMs. Particularly, we have carried out extensive studies in search for Majorana fermions (MFs) in TQMs for topological quantum computation. Moreover, we have focused on three important science questions. 1) How can we controllably tune the properties of TQMs to make them suitable for quantum information applications? 2) What materials parameters are most important for successfully observing MFs in TQMs? 3) Can the physical properties of TQMs be tailored by topological band engineering? Results obtained in this LDRD not only deepen our current knowledge in fundamental quantum physics but also hold great promise for advanced electronic/photonic applications in information technologies.
Tin-germanium alloys are increasingly of interest as optoelectronic and thermoelectric materials as well as materials for Li/Na ion battery electrodes. However, the lattice incompatibility of bulk Sn and Ge makes creating such alloys challenging. By exploiting the unique strain tolerance of nanosized crystals, we have developed a facile synthetic method for homogeneous SnxGe1-x alloy nanocrystals with composition varying from essentially pure Ge to 95% Sn while still maintaining the cubic structure.
Detailed understanding of solid–solid interface structure–function relationships is critical for the improvement and wide deployment of all-solid-state batteries. The interfaces between lithium phosphorous oxynitride (LiPON) solid electrolyte material and lithium metal anode, and between LiPON and LixCoO2 cathode, have been reported to generate solid–electrolyte interphase (SEI)-like products and/or disordered regions. Using electronic structure calculations and crystalline LiPON models, we predict that LiPON models with purely P−N−P backbones are kinetically inert towards lithium at room temperature. In contrast, transfer of oxygen atoms from low-energy LixCoO2(104) surfaces to LiPON is much faster under ambient conditions. The mechanisms of the primary reaction steps, LiPON structural motifs that readily reacts with lithium metal, experimental results on amorphous LiPON to partially corroborate these predictions, and possible mitigation strategies to reduce degradations are discussed. LiPON interfaces are found to be useful case studies for highlighting the importance of kinetics-controlled processes during battery assembly at moderate processing temperatures.
The formulation of carrier transport through heterojunctions by tunneling and thermionic emission is derived from first principles. The treatment of tunneling is discussed at three levels of approximation: numerical solution of the one-band envelope equation for an arbitrarily specified potential profile; the WKB approximation for an arbitrary potential; and, an analytic formulation assuming constant internal field. The effects of spatially varying carrier chemical potentials over tunneling distances are included. Illustrative computational results are presented. The described approach is used in exploratory physics models of irradiated heterojunction bipolar transistors within Sandia's QASPR program.
We show scaling results for materials of interest in Sandia Radiation-Effects and High-Energy-Density-Physics Mission Areas. Each timing is from a self-consistent calculation for bulk material. Two timings are given: (1) walltime for the construction of the CR exchange operator (Exchange-Operator) and (2) walltime for everything else (non-Exchange-Operator).
