Bulletin of the American Physical Society
APS March Meeting 2018
Volume 63, Number 1
Monday–Friday, March 5–9, 2018; Los Angeles, California
Session B34: Petascale Science and Beyond: Applications and Opportunities for Materials, Chemical, and Bio Physics IIFocus
|
Hide Abstracts |
Sponsoring Units: DCOMP DBIO DCP DCMP Chair: Jack Wells, Oak Ridge National Lab Room: LACC 409A |
Monday, March 5, 2018 11:15AM - 11:51AM |
B34.00001: Data-driven Molecular Engineering of Solar-Powered Windows using Materials Database Auto-Generation Tools with Large-Scale Data-Mining Invited Speaker: Jacqueline Cole Large-scale data-mining workflows are increasingly able to predict successfully new chemicals that possess a targeted functionality. The success of such materials discovery approaches is nonetheless contingent upon having the right data source to mine, adequate supercomputing facilities and workflows to enable this mining, and algorithms that suitably encode structure-function relationships as data-mining workflows which progressively short list data toward the prediction of a lead material for experimental validation. |
Monday, March 5, 2018 11:51AM - 12:03PM |
B34.00002: ALCF Data Science and Machine Learning Programs: From Petascale to Exascale Nichols Romero, Elise Jennings, Álvaro Vázquez-Mayagoitia, Venkatram Vishwanath, Timothy Williams In addition to traditional high-performance computing simulations, ALCF is growing two additional computational science pillars: Data Science and Machine Learning. We anticipate two calls for proposals (CFPs) in 2018. One CFP will be for the ALCF Data Science Program (ADSP) which targets “big data” science problems for current leadership computing resources. Another CFP will be for the ALCF Early Science Program (ESP) targeting the Argonne 2021 exascale supercomputer. The goal of both programs is to explore and improve a variety of computational methods that will enable data-driven discoveries across all scientific disciplines. The projects will focus on data science techniques covering a wide area of discovery including but not limited to uncertainty quantification, statistics, machine learning, deep learning, databases, pattern recognition, image processing, graph analytics, data mining, real-time data analysis, and complex and interactive workflows. The ESP projects will ultimately have pre-production access to 2021 exascale system to run proposed calculations. The ADSP and ESP CFPs will be announced on the ALCF website, http://www.alcf.anl.gov. |
Monday, March 5, 2018 12:03PM - 12:15PM |
B34.00003: Towards Real-Time Analysis of Morphologies using Scattering Alexander Hexemer, Ronald Pandolfi, Dinesh Kumar, Harinarayan Krishnan, James Sethian The advent of high brightness sources, fast detectors and the increasing need of time-resolved experiments in small angle scattering has created an unprecedented data deluge and the needs for combining X-ray science with computer science. Over the last few years we have worked closely with our computational research and supercomputer division to enable the use of new math, fast networks, supercomputers and new algorithms at beamlines. The goal of such a superfacility would be immediate feedback for scientist during experiments. To streamline the user experience, we have developed a user friendly and plug-in based toolkit XI-CAM that connects different algorithms and allows execution on fast clusters and supercomputers with a push of a button. |
Monday, March 5, 2018 12:15PM - 12:27PM |
B34.00004: Atom-based polychromatic diffraction pattern predictions for ultra-fast time-resolved experiemnts with large scale GPU-accelerated parallel simulation code Juncheng E, Yiyang Zhang, Sen Chen, Sheng-Nian Luo To design and interpret single-shot ultra-fast time-resolved synchrotron- and XFEL-based diffraction experiments, it's desired to simulate the diffraction patterns directly from atomic configurations. GPU accelerated parallel simulation for direct, kinematics-based, simulations of x-ray/electron diffraction of large-scale atomic systems with mono-/polychromatic beams and arbitrary plane detector geometry is implemented here. With this code, nanocrystalline system size can be up to several billion atoms. In the system, nanocrystals can be of arbitrary crystal structures, grain shapes and sizes. Both X-ray and electron 2D diffraction patterns can be simulated with arbitrary beam energy in mono- or poly-chromatic beam cases. Intensities in reciprocal space can be directly exploded to help design experiment setups. The application to predict polychromatic diffraction pattern accounting for the polychromatic nature of x-ray implemented in ultra-fast time-resolved synchrotron- and XFEL-based experiments will presented. |
Monday, March 5, 2018 12:27PM - 12:39PM |
B34.00005: New Insights into the Geometric State for Two-Fluid Porous Medium Systems James McClure, Ryan Armstrong, Mark Berrill, Steffen Schlüter, Steffen Berg, Cass Miller, William Gray Mechanistic models for two-fluid flow in porous medium systems are closed with an empirical relation usually referred to as a capillary pressure-fluid saturation relation.The ad hoc and hysteretic nature of this closure relation has been the focus of attention for the last two decades. We show that capillary pressure can be represented as a function of fluid saturation, specific interfacial area, and the Euler characteristic. The posited form of the state equation is investigated by using data generated from synchrotron-based micro-CT and simulations performed on the Titan supercomputer to explore the possible geometric states for a representative set of media. The results conclusively demonstrate that hysteresis appearing in standard closure relations can be removed and that the resultant state equation describes not only equilibrium states but also dynamic states. This work provides an important foundational component for a new generation of high fidelity multiphase porous medium models. |
Monday, March 5, 2018 12:39PM - 12:51PM |
B34.00006: Increased productivity with real time atomistic simulations using Atomify Anders Hafreager, Svenn-Arne Dragly, Anders Malthe-Sørenssen The typical workflow when running atomistic simulations includes working with several programs. A text editor is needed to create and modify input scripts, the terminal to run the simulation, and programs like VMD or Ovito to visualize the system over time. If physical quantities are computed, the data is often plotted with MATLAB or Python, where additional scripts must be used. This is a tedious process, especially for teaching purposes and for people who are new in the field. I have developed Atomify; a high performance real time visualizer for atomistic simulations. It can simulate and render more than one million atoms with excellent frame rate on modern hardware. Atomify supports OpenMP acceleration, GPU acceleration, real time plotting of physical quantities, and an easy-to-use code editor in one single application. The software currently uses LAMMPS as physics engine, but it can easily be extended to support other codes like Gromacs, NAMD or OpenMM. Atomify is open-source software (GPL) written in C++ using the Qt framework. |
Monday, March 5, 2018 12:51PM - 1:03PM |
B34.00007: Automated Reaction Discovery from Combined Application of Transition State Search Algorithms Colin Grambow, Adeel Jamal, Yi-Pei Li, Judit Zádor, Yury Suleimanov, William Green With petascale computers, new chemistry can be discovered more rapidly in silico than in the laboratory. In this work, we investigate the unimolecular decomposition of γ-ketohydroperoxide using computational schemes that automatically search for elementary reaction steps, i.e., a combination of the Berny optimization algorithm with the freezing string method, the single- and double-ended growing string methods, the heuristic KinBot algorithm, and the single-component artificial force induced reaction method (SC-AFIR). These efforts lead to a discovery of 75 elementary unimolecular reactions of γ-ketohydroperoxide, 69 of which were previously unknown. All of the methods we adopted found the lowest energy transition state corresponding to the first step of the Korcek mechanism. However, each algorithm except for SC-AFIR discovered several reactions not detected by any of the other methods. The present work demonstrates both the strengths and weaknesses of existing schemes for automated reaction discovery and highlights the advantage of the combined application of several computational approaches. However, for the reliable discovery of all important reactions of any given reactants, further method development and assessment is required. |
Monday, March 5, 2018 1:03PM - 1:15PM |
B34.00008: Predictive Automated Combustion Chemistry: Massively Parallel High-Accuracy Thermochemistry Murat Keceli, Sarah Elliott, Yi-Pei Li, Matt Johnson, Carlo Cavallotti, Justin Wozniak, Yuri Georgievskii, Ahren Jasper, William Green, Stephen Klippenstein High fidelity mechanisms are crucial for improving the efficiency of combustion devices. We are developing a fuel chemistry code that automates reaction mechanism generation (RMG) with accurate thermochemical kinetics. Accurate computation of thermochemical parameters requires careful treatment of electronic and nuclear degrees of freedom. We couple accurate determinations of the lowest energy torsional conformer with extrapolation of electronic energies to the complete-basis-set, complete-correlation limit, and detailed examination of anharmonic vibrational effects in predicting the partition functions. To do so, we implemented a Python code, QTC, which automates all aspects of the workflow providing a unified interface for quantum chemistry packages and other codes developed by our team for torsional scans and partition function calculations. Parallelization through Swift scripts allows for very high strong scaling efficiency on supercomputers. Our approach is illustrated by generating a list of important species for butane combustion with RMG followed by large-scale automated benchmark thermochemistry calculations with QTC. |
Monday, March 5, 2018 1:15PM - 1:27PM |
B34.00009: High-Throughput Phonon Calculations With the Real-Space Multigrid Method (RMG) Jiayong Zhang, Wenchang Lu, Emil Briggs, Yongqiang Cheng, Anibal Ramirez-Cuesta, Jerry Bernholc We have combined our real-space multigrid code (RMG, www.rmgdft.org) with Phonopy to enable phonon calculations on large systems, up to ~1000 atoms, within the framework of Density Functional Theory. RMG is a real-space multigrid method based code for electronic-structure calculations, which is highly parallel with excellent scalability to thousands of nodes and GPUs. Phonopy is an open-source python-based package for pre/post-processing for phonon calculations with interfaces to many DFT codes. Employing the Finite Displacement Method (FDM), we build supercells with small perturbations from atoms’ equilibrium positions and perform self-consistent calculations to obtain interatomic forces in real space, which are transformed to reciprocal space to form dynamical matrices at arbitrary q values. Several techniques that reduce the computational cost or numerical errors and improve accuracy of the final results will also be discussed. We have used RMG for a variety of systems, four of which will be discussed in detail here: silicon, ZrH2, carbazole and ZIF-8. RMG’s results are also compared with inelastic neutron scattering data, measured at the VISION spectrometer at the SNS in ORNL, and other DFT codes for validation purposes. |
Monday, March 5, 2018 1:27PM - 1:39PM |
B34.00010: Theory of Band Gaps in Amorphous Nano-porous SiC Tyler Summers, Blair Tuttle, Andrew O'hara, Sokrates Pantelides Nano-porous SiC is an important insulator used as a back-end-of-the-line dielectric in scaled integrated circuits. In the present study, nano-porous SiC atomic models are created from cubic SiC supercells. First, a void of diameter of ~ 1 nm is hydrogen passivated. Then, amorphous models ae constructed with an atom type switching procedure, which was applied until the average atomic composition and bond densities converge to experimental values. Density functional theory calculations are used to explore the electronic and physical properties of various nano-porous SiC models. Specifically, we examine the effect of average atom composition and bonding on the bandgap of nano-porous SiC. |
Monday, March 5, 2018 1:39PM - 1:51PM |
B34.00011: Subgroup Discovery for Finding Interpretable Local Patterns in Data from Materials-Science Bryan Goldsmith, Mario Boley, Christopher Sutton, Jilles Vreeken, Matthias Scheffler, Luca Ghiringhelli We demonstrate that subgroup discovery (SGD) can help find physically meaningful descriptors from materials-science data obtained by first-principles calculations. In contrast to global modelling algorithms, SGD finds descriptions of subpopulations in which, locally, the target property takes on an interesting distribution. First, the SGD algorithm is formulated for materials applications. Next, SGD is applied to gas-phase gold clusters (having 5 to 14 atoms) to discern patterns between their geometrical and physicochemical properties. Additionally, SGD is shown to identify subgroups that classify 79 of the 82 octet binary materials as either rock salt or zincblende from only information of its chemical composition. SGD is also used to find descriptors that predict both the formation and bandgap energies of transparent conducting oxides. Lastly, an efficient optimal solver using branch-and-bound is developed for dispersion-corrected objective functions to help find improved subgroups. |
Monday, March 5, 2018 1:51PM - 2:03PM |
B34.00012: Variational Monte Carlo Study of the Two-Dimensional Hubbard Model Using Slater-Jastrow Wave Functions Jeong-Pil Song, Leonard Sprague, Chia-Chen Chang, Brenda Rubenstein We present numerical results for the 2D Hubbard and Extended Hubbard models on a variety of lattices using Variational Monte Carlo with a generalized Slater-Jastrow ansatz. We employ stochastic optimization to determine the thousands of Jastrow parameters that parameterize the ground state. In order to fully characterize the ground state physics, we calculate correlation functions for larger lattice sizes than are tractable using projector QMC algorithms and use those results to perform finite-size scaling analyses. Our results support previous findings that the half-filled one-band Hubbard model undergoes a transition to an antiferromagnetic phase at an infinitesimally small value of the critical on-site repulsion. We also discuss the application of our method to ab initio Hamiltonians. |
Monday, March 5, 2018 2:03PM - 2:15PM |
B34.00013: Improving efficiency of the continuous-time quantum Monte Carlo Mancheon Han, Hyoung Joon Choi The dynamical mean field theory (DMFT) is a method to investigate strongly correlated materials. This method considers correlations effects in the solid through the Anderson impurity model (AIM). Because there is no analytic solution to AIM, we have to adopt numerical methods called impurity solver. Among various impurity solvers, we implemented continuous-time quantum Monte Carlo approach, which is numerically exact and allow us to deal with any two particle interaction hamiltonian. We tried to improve our impurity solver to calculate the self energy more efficiently and tested its accuracy by comparing it with several model studies. Moreover, we constructed DFT (density functional theory) + DMFT program by combining the developed solver with the SIESTA program. Then we calculated electronic structures of some materials and compared them with experimental data. |
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