Bulletin of the American Physical Society
APS March Meeting 2017
Volume 62, Number 4
Monday–Friday, March 13–17, 2017; New Orleans, Louisiana
Session C1: Computational Discovery and Design of Novel Materials IIIFocus
|
Hide Abstracts |
Sponsoring Units: DMP DCOMP Room: 260 |
Monday, March 13, 2017 2:30PM - 2:42PM |
C1.00001: Construction of interatomic potentials for multicomponent systems with stratified neural networks Samad Hajinazar, Aleksey Kolmogorov Recent application of neural networks (NNs) to modeling interatomic interactions has shown the learning machines' encouragingly accurate performance for select elemental and multicomponent systems. In an effort to build extended libraries of NN-based models we have introduced a hierarchical training in which NNs for multicomponent systems are obtained by sequential fitting from the bottom up: first unaries, then binaries, and so on. Advantages of constructing NN sets with shared parameters include acceleration of the training process and intact description of the constituent systems. In the test case of the Cu-Pd-Ag ternary and its subsystems, NNs trained in the traditional and stratified fashions are found to have essentially identical accuracy for defect energies, phonon dispersions, formation energies, etc. The models' robustness is further illustrated via unconstrained evolutionary structure searches in which the NN is used for the local optimization of crystal unit cells. The use of NN instead of DFT in these simulations accelerates structure prediction by several orders of magnitude. The NN module is available in the MAISE package. [Preview Abstract] |
Monday, March 13, 2017 2:42PM - 2:54PM |
C1.00002: High-dimensional artificial neural network potentials for boron and its application to searching for new structures Woohyun Han, In-Ho Lee, Keejoo Chang The construction of accurate potential-energy surfaces (PES) with respect to lattice parameters and atomic coordinates is an important step for the atomistic simulations of structural phases. Recently, artificial neural networks (ANN) have been suggested to be a promising technique for constructing the PES of complex systems due to advantages in efficiency and accuracy, as compared to interatomic potentials and first-principles calculations. Elemental boron exhibits a variety of allotropes consisting of icosahedra as structural units, which are attributed to the electron deficiency, compared with carbon. Thus, it is a challenging task to generate accurate potentials for boron. In this work, we report high-dimensional ANN potentials for elemental boron, which are generated by the Behler-Parrinello approach. The weight parameters for the ANN potentials are optimized by using the machine learning technique, and training sets are obtained from first-principle calculations. The generated ANN potentials well reproduce the energy vs volume curves, phonon spectrum, and molecular dynamics simulations for several known boron allotropes. We combine the ANN potentials with the conformational space annealing algorithm for global optimization and discuss its applications. [Preview Abstract] |
Monday, March 13, 2017 2:54PM - 3:06PM |
C1.00003: Sampling Models for Machine Learned Atomistic Potentials - The Case of Water Amber Maharaj, Isaac Tamblyn Computational restrictions on system size and accessible time scales in atomistic simulations inhibit the investigation of mesoscopic materials. While \textit{ab initio} methods provide accurate energies and forces, they are unable to efficiently sample configuration space. Classical force fields and semi empirical methods give an efficient sampling of configuration space, but with weights and energies that differ from \textit{ab initio} methods. Here we consider the relationship between the optimal configurational sampling method and electronic structure approach for calculating total energies and forces (for the particular case of water) of a liquid. Using the Atomic Energy Network (aenet) approach, we apply machine learning to the generation of interaction potentials of water models such as TIP3P, TIP4P and TIP5P, and reference data from DFTB, and DFT. The use of machine learned potentials reduces the algorithmic complexity of simulations while achieving accuracy comparable to \textit{ab initio} methods. To verify the accuracy of the generated potentials, physically observable quantities are computed and compared to \textit{ab initio} and experimental data. [Preview Abstract] |
Monday, March 13, 2017 3:06PM - 3:42PM |
C1.00004: Many-Body Tensor Representation for Machine Learning of Materials Invited Speaker: Matthias Rupp Computational discovery and design of novel materials requires large numbers of accurate electronic structure calculations, whose high computational cost is a limiting factor. Machine learning can significantly reduce the number of necessary calculations by interpolating between a set of reference calculations. For this, a numerical representation of atomistic systems that supports interpolation is crucial. We present a many-body tensor representation that can encode both molecules and crystals, has proper mathematical structure, is invariant to translation, rotation, and nuclear permutations, unique, continuous, differentiable, fast to compute, and exhibits excellent empirical performance on benchmark datasets. [Preview Abstract] |
Monday, March 13, 2017 3:42PM - 3:54PM |
C1.00005: Simulating Strongly Anharmonic and Mechanically Unstable Crystals at Finite Temperature John C. Thomas, Anton Van der Ven For a wide range of materials, linear elasticity and quasiharmonic models of lattice dynamics are either inaccurate or unphysical. In particular, these methods predict many high-temperature crystal phases to be dynamically unstable, due to their non-convex free energies at low temperature. Unfortunately, the few theoretical methods that go beyond simple harmonic approximations are still inadequate for rigorous and predictive simulation of most strongly anharmonic materials. We describe a simulation framework that enables accurate first-principles prediction of finite-temperature properties of anharmonic and mechanically unstable crystal phases. This framework relies on basis functions, constructed in terms of lattice deformation and atomic displacements, that are invariant to rigid-body translation and rotation, as well as space-group operations of the ideal crystal. This basis set is used to specify order parameters and parameterize highly accurate crystal Hamiltonians, which can be employed within molecular dynamics or Monte Carlo simulation to predict free energies, structural phase transitions, and nonlinear elastic properties. We illustrate the relevance of this approach to thermoelectric semiconductors and metal hydrides. [Preview Abstract] |
Monday, March 13, 2017 3:54PM - 4:06PM |
C1.00006: Discovery of Novel Oxides Using Machine Learning and First-Principles Calculations Antoine Emery, Logan Ward, Chris Wolverton Oxide materials are used for a variety of technologically relevant applications such as solid oxide fuel cell, water splitting and transparent conductors. Up until now, mostly binary and simple ternary oxides have been carefully synthesized and characterized. As a result, there are opportunities to discover new, more complex and more efficient materials for numerous applications. As the number of possible compounds is prohibitively large to explore entirely experimentally or via first-principles calculations, we use machine learning to reduce the number of compositions to be calculated via more costly methods such as density functional theory (DFT). We show that this approach reduces significantly the time spent calculating unstable compounds, allowing the exploration of larger structures and wider chemical spaces. The machine learning-aided DFT approach presented in this work also showcases a reliable framework enabling the acceleration of materials discovery. [Preview Abstract] |
Monday, March 13, 2017 4:06PM - 4:18PM |
C1.00007: Large-scale high-throughput computer-aided discovery of advanced materials using cloud computing Timur Bazhirov, Mohammad Mohammadi, Kevin Ding, Sergey Barabash Recent advances in cloud computing made it possible to access large-scale computational resources completely on-demand in a rapid and efficient manner. When combined with high fidelity simulations, they serve as an alternative pathway to enable computational discovery and design of new materials through large-scale high-throughput screening. Here, we present a case study for a cloud platform implemented at Exabyte Inc. We perform calculations to screen lightweight ternary alloys for thermodynamic stability. Due to the lack of experimental data for most such systems, we rely on theoretical approaches based on first-principle pseudopotential density functional theory. We calculate the formation energies for a set of ternary compounds approximated by special quasirandom structures. During an example run we were able to scale to 10,656 CPUs within 7 minutes from the start, and obtain results for 296 compounds within 38 hours. The results indicate that the ultimate formation enthalpy of ternary systems can be negative for some of lightweight alloys, including Li and Mg compounds. We conclude that compared to traditional capital-intensive approach that requires in on-premises hardware resources, cloud computing is agile and cost-effective, yet scalable and delivers similar performance. [Preview Abstract] |
Monday, March 13, 2017 4:18PM - 4:30PM |
C1.00008: Finding patterns, correlations, and descriptors in materials-science data using subgroup discovery Mario Boley, Bryan R. Goldsmith, Jilles Vreeken, Luca M. Ghiringhelli, Matthias Scheffler Data analytics applied to materials-science data often focuses on the inference of a global prediction model for some physical or chemical property of interest for a given class of materials, such as activation barriers or binding energies. However, the underlying mechanism for some target property could differ for different materials within a large pool of materials-science data. Consequently, a global model fitted to the entire dataset may be difficult to interpret and may well hide or incorrectly describe the actuating physical mechanisms. In these situations, local models would be advantageous to global models. Subgroup discovery (SGD) is presented here as a data-mining approach to find interpretable local models of a target property in materials-science data. We first demonstrate that SGD can identify physically meaningful models that classify the crystal structures of 82 octet binary semiconductors as either rocksalt or zincblende. The SGD framework is subsequently applied to 24 400 configurations of neutral gas-phase gold clusters with 5 to 14 atoms to discern general patterns between geometrical and physicochemical properties. [Preview Abstract] |
Monday, March 13, 2017 4:30PM - 4:42PM |
C1.00009: Pareto fronts for multiobjective optimization design on materials data Abhijith Gopakumar, Prasanna Balachandran, James E. Gubernatis, Turab Lookman Optimizing multiple properties simultaneously is vital in materials design. Here we apply infor- mation driven, statistical optimization strategies blended with machine learning methods, to address multi-objective optimization tasks on materials data. These strategies aim to find the Pareto front consisting of non-dominated data points from a set of candidate compounds with known character- istics. The objective is to find the pareto front in as few additional measurements or calculations as possible. We show how exploration of the data space to find the front is achieved by using uncer- tainties in predictions from regression models. We test our proposed design strategies on multiple, independent data sets including those from computations as well as experiments. These include data sets for Max phases, piezoelectrics and multicomponent alloys. [Preview Abstract] |
Monday, March 13, 2017 4:42PM - 4:54PM |
C1.00010: Lead-free Halide Perovskites via Functionality-directed Materials Screening Lijun Zhang, Dongwen Yang, Jian Lv, Xingang Zhao, Ji-Hui Yang, Liping Yu, Su-Huai Wei, Alex Zunger Hybrid organic-inorganic halide perovskites with the prototype material of CH$_3$NH$_3$PbI$_3$ have recently attracted much interest as low-cost and high-performance photovoltaic absorbers but one would like to improve their stability and get rid of toxic Pb. We used photovoltaic-functionality-directed materials screening approach to rationally design via first-principles DFT calculations Pb-free halide perovskites. Screening criteria involve thermodynamic and crystallographic stability, as well as solar band gaps, light carrier effective masses, exciton binding, etc. We considered both single atomic substitutions in AMX$_3$ normal perovskites (altering chemical constituents of A, M and X individually) as well as double substitution of 2M into B+C in A$_2$BCX$_6$ double-perovskites. Chemical trends in phase stabilities and optoelectronic properties are discussed with some promising cases exhibiting solar cell efficiencies comparable to that of CH$_3$NH$_3$PbI$_3$. [Preview Abstract] |
Monday, March 13, 2017 4:54PM - 5:06PM |
C1.00011: Designing Semiconductor Heterostructures Using Digitally Accessible Electronic-Structure Data Ethan Shapera, Andre Schleife Semiconductor sandwich structures, so-called heterojunctions, are at the heart of modern applications with tremendous societal impact: Light-emitting diodes shape the future of lighting and solar cells are promising for renewable energy. However, their computer-based design is hampered by the high cost of electronic structure techniques used to select materials based on alignment of valence and conduction bands and to evaluate excited state properties. We describe, validate, and demonstrate an open source Python framework which rapidly screens existing online databases and user-provided data to find combinations of suitable, previously fabricated materials for optoelectronic applications. The branch point energy aligns valence and conduction bands of different materials, requiring only the bulk density functional theory band structure. We train machine learning algorithms to predict the dielectric constant, electron mobility, and hole mobility with material descriptors available in online databases. Using CdSe and InP as emitting layers for LEDs and CH$_3$NH$_3$PbI$_3$ and nanoparticle PbS as absorbers for solar cells, we demonstrate our broadly applicable, automated method. [Preview Abstract] |
Monday, March 13, 2017 5:06PM - 5:18PM |
C1.00012: A Data-Driven Approach for Characterization of Ternary Al(fcc)-Al$_{2}$Cu(tet)-Ag$_{2}$Al(hcp) Eutectic Alloys Irmak Sargin, Scott Beckman We present a new data-driven approach to characterize the microstructure of the ternary eutectic Al-Ag$_{2}$Al-Al$_{2}$Cu alloy. In this approach a range of microstructural descriptors are developed and used within the PCA and PLSR analysis methods. The similarity between the ideal microstructures and those experimentally obtained are quantified and the percentage similarity to ideal structures is used as a new means for cataloguing microstructures. This quantified comparison of the idealized structures to the experimental ones provides insight about the microstructural evolution. Such an approach can be applied to many different areas of materials science allowing the important relationships between microstructure and physical properties to be identified. This form of analysis allows determination of which deviations from the ideal attributes result in variation from the ideal physical properties. [Preview Abstract] |
Monday, March 13, 2017 5:18PM - 5:30PM |
C1.00013: Ab initio Studies of Filling-enforced Nodal Semimetals Ru Chen, Hoi Chun Po, Jeffrey B. Neaton, Ashvin Vishwanath We present a new search criterion for nodal semimetals based on electron filling and symmetries. With this new criterion, several material candidates are discovered using ab initio calculations and experimentally characterized material databases. We discuss specific material candidates, such as filling-enforced Dirac semimetals and Dirac nodal-line semimetals. [Preview Abstract] |
Follow Us |
Engage
Become an APS Member |
My APS
Renew Membership |
Information for |
About APSThe American Physical Society (APS) is a non-profit membership organization working to advance the knowledge of physics. |
© 2024 American Physical Society
| All rights reserved | Terms of Use
| Contact Us
Headquarters
1 Physics Ellipse, College Park, MD 20740-3844
(301) 209-3200
Editorial Office
100 Motor Pkwy, Suite 110, Hauppauge, NY 11788
(631) 591-4000
Office of Public Affairs
529 14th St NW, Suite 1050, Washington, D.C. 20045-2001
(202) 662-8700