A5018494557- Machine Learning in Materials Science
- Thermal properties of materials
- Advanced Thermoelectric Materials and Devices
- X-ray Diffraction in Crystallography
- High-pressure geophysics and materials
- Electronic and Structural Properties of Oxides
- Force Microscopy Techniques and Applications
- Thermal Radiation and Cooling Technologies
- Electron and X-Ray Spectroscopy Techniques
- Advanced Chemical Physics Studies
- Manufacturing Process and Optimization
- Additive Manufacturing and 3D Printing Technologies
- Thermal Expansion and Ionic Conductivity
- Ultrasonics and Acoustic Wave Propagation
- Geomagnetism and Paleomagnetism Studies
- Geophysical and Geoelectrical Methods
- Fuel Cells and Related Materials
- Carbon Nanotubes in Composites
- 3D Surveying and Cultural Heritage
- Solar Thermal and Photovoltaic Systems
- Adsorption and Cooling Systems
- Robot Manipulation and Learning
- Quantum, superfluid, helium dynamics
- Ga2O3 and related materials
- Spectroscopy and Chemometric Analyses
Sandia National Laboratories
2022-2024
Massachusetts Institute of Technology
2020-2022
Georgia Institute of Technology
2017-2018
We report the fabrication and measurement of thermophotovoltaic (TPV) cells with efficiencies >40%, which is a record high TPV efficiency first experimental demonstration high-bandgap tandem cells. was determined by simultaneous electric power output heat dissipation from device via calorimetry. The are two-junction devices comprising high-quality III-V materials band gaps between 1.0 1.4 eV that optimized for emitter temperatures 1900-2400{\deg}C. exploit concept band-edge spectral...
β-Ga2O3 is a wide-bandgap semiconductor of significant technological importance for electronics, but its low thermal conductivity an impeding factor applications. In this work, interatomic potential developed based on deep neural network model to predict the and phonon transport properties. Our trained by ab initio energy surface atomic forces, which reproduces dispersion in good agreement with first-principles calculations. We are able use molecular dynamics (MD) simulations anisotropic...
Chemical and physical properties of complex materials emerge from the collective motions constituent atoms.These are in turn determined by a variety interatomic interactions mediated local redistribution valence electrons about fixed core nuclear charges.Scientific engineering advances science, chemistry, many related fields benefit our ability to directly sample equilibrium kinetic probability distributions large collections atoms molecules.Classical molecular dynamics (MD) is widely used...
Abstract Lack of rigorous reproducibility and validation are significant hurdles for scientific development across many fields. Materials science, in particular, encompasses a variety experimental theoretical approaches that require careful benchmarking. Leaderboard efforts have been developed previously to mitigate these issues. However, comprehensive comparison benchmarking on an integrated platform with multiple data modalities perfect defect materials is still lacking. This work...
We present experimental measurements of the thermal boundary conductance (TBC) from 78–500 K across isolated heteroepitaxially grown ZnO films on GaN substrates. This data provides an assessment underlying assumptions driving phonon gas-based models, such as diffuse mismatch model (DMM), and atomistic Green's function (AGF) formalisms used to predict TBC. Our measurements, when compared previous data, suggest that TBC can be influenced by long wavelength, zone center modes in a material one...
Abstract Molecular dynamics simulations have been extensively used to study phonons and gain insight, but direct comparisons experimental data are often difficult, due a lack of accurate empirical interatomic potentials for different systems. As result, this issue has become major barrier realizing the promise associated with advanced atomistic-level modeling techniques. Here, we present general method specifically optimizing from ab initio inputs phonon transport properties, thereby...
Abstract Atomic vibrations influence a variety of phenomena in solids and molecules, ranging from thermal transport to chemical reactions. These can be decomposed into normal modes, often known as phonons, which are collective motions atoms vibrating at certain frequencies; this provides rigorous basis for understanding atomic motion its effects on material phenomena, since phonons detected excited experimentally. Unfortunately, traditional theories such the phonon gas model do not allow...
Abstract Machine learning (ML) enables the development of interatomic potentials with accuracy first principles methods while retaining speed and parallel efficiency empirical potentials. While ML traditionally use atom-centered descriptors as inputs, different models such linear regression neural networks map to atomic energies forces. This begs question: what is improvement in due model complexity irrespective descriptors? We curate three datasets investigate this question terms ab initio...
Current understanding of phonons is based on the phonon gas model (PGM), which best rationalized for crystalline materials. However, most phonons/modes in disordered materials have a different character and thus may contribute to heat conduction fundamentally way than described by PGM. For modes crystals, sinusoidal character, one can separate into two primary categories, namely acoustic optical modes. materials, such designations no longer rigorously apply. Nonetheless, phase quotient (PQ)...
A shadow molecular dynamics scheme for flexible charge models is presented where the Born-Oppenheimer potential derived from a coarse-grained approximation of range-separated density functional theory. The interatomic potential, including atomic electronegativities and charge-independent short-range part force terms, modeled by linear cluster expansion (ACE), which provides computationally efficient alternative to many machine learning methods. based on extended Lagrangian (XL) (BOMD) [Eur....
Molecular dynamics (MD) is a powerful technique that can be used to study thermal vibrations/phonons and properly account for their role in different phenomena are important mechanical engineering, chemistry, physics materials science. However, despite the widespread usage of MD various phenomena, direct comparisons between experiments simulations often associated with low fidelity, due inaccuracy interatomic potentials (IAPs) employed. This issue has become main barrier utilizing studying...
Lack of rigorous reproducibility and validation are major hurdles for scientific development across many fields. Materials science in particular encompasses a variety experimental theoretical approaches that require careful benchmarking. Leaderboard efforts have been developed previously to mitigate these issues. However, comprehensive comparison benchmarking on an integrated platform with multiple data modalities both perfect defect materials is still lacking. This work introduces...
The low thermal conductivity of polymers limits their use in numerous applications, where heat transfer is important.The two primary approaches to overcome this limitation, are mix other materials with high conductivity, or mechanically stretch the increase intrinsic conductivity.Progress along both these pathways has been stifled by issues associated interface resistance and manufacturing scalability respectively.Here, we report a novel polymer composite architecture that enabled employing...
Predictive modeling of the phonon/thermal transport properties materials is vital to rational design for a diverse spectrum engineering applications. Classical Molecular Dynamics (MD) simulations serve as tool simulate time evolution atomic level system dynamics and enable calculation thermal wide range materials, from perfect periodic crystals systems with strong structural compositional disorder, well their interfaces. Although MD does not intrinsically rely on plane wave-like phonon...
We address the problem of generating a phonon optimized interatomic potential (POP) for aluminum. The POP methodology, which has already been shown to work semiconductors such as silicon and germanium, uses an evolutionary strategy based on genetic algorithm (GA) optimize free parameters in empirical (EIP). For aluminum, we used Vashishta functional form. training data set was generated ab initio, consisting forces, energy vs. volume, stresses, harmonic cubic force constants obtained from...
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 atomic environment, MALA models efficiently predict key electronic observables, including states, density, and total energy. The package integrates data sampling, model training inference into unified library, while ensuring compatibility with standard...
Dynamic compression of iron to Earth-core conditions is one the few ways gather important elastic and transport properties needed uncover key mechanisms surrounding geodynamo effect. Herein, a machine-learned ab initio derived molecular-spin dynamics (MSD) methodology with explicit treatment for longitudinal spin-fluctuations utilized probe dynamic phase-diagram iron. This framework uniquely enables an accurate resolution phase-transition kinetics properties, as highlighted by compressional...
Material extrusion 3D printing has enabled an elegant fabrication pathway for a vast material library. Nonetheless, each requires optimization of parameters generally determined through significant trial-and-error testing. To eliminate arduous, iteration-based approaches, many researchers have used machine learning (ML) algorithms which provide opportunities automated process optimization. In this work, we demonstrate the use ML-driven approach real-time print-parameter in-situ monitoring...