Bulletin of the American Physical Society
APS March Meeting 2020
Volume 65, Number 1
Monday–Friday, March 2–6, 2020; Denver, Colorado
Session B40: Building the bridge to exascale: applications and opportunities for materials, chemistry, and biology IIFocus
|
Hide Abstracts |
Sponsoring Units: DCOMP DCMP DAMOP DCP Chair: Anouar Benali, Argonne Natl Lab Room: 705 |
Monday, March 2, 2020 11:15AM - 11:51AM |
B40.00001: Accelerating Large-Scale GW Calculations on Hybrid GPU-CPU Systems Invited Speaker: Mauro Del Ben Large-scale GW calculations are state-of-the-art to accurately describe excited state phenomena in materials, which is critical for the design of novel new devices based on complex materials with applications in many fields. Application of the GW method to complex systems is often perceived as being limited due to high computational cost. Reduced time to solution can be achieved through novel methods, algorithms and optimal implementations on modern HPC systems. In particular accelerators such as GPU’s can speed-up by order of magnitudes conventional CPU-only implementations and additionally reduce the energy per flop consumption. However, porting a large scale HPC code to hybrid GPU-CPU systems and achieve best performance is non trivial and faces several challenges. This talk showcases the various techniques used to accelerate the Material Science code BerkeleyGW on hybrid architectures targeting to accelerate large scale simulations with thousands of atoms. These techniques include the efficient use of accelerated libraries, pinned host memory, asynchronous memory transfer, streams, batched operations, shared memory, and the overlapping of MPI communication and GPU computation. We achieve good strong- and weak-scaling on thousands of GPUs, and a 16x improvement is obtained on FLOPs/Watt efficiency compared to the CPU-only implementation. We show in this way that GW calculations on systems made of thousands of atoms can be performed with excellent time to solutions, of the order of minutes, even running on a moderate number of hybrid nodes. |
Monday, March 2, 2020 11:51AM - 12:03PM |
B40.00002: Accuracy limits of quantum Monte Carlo in the weak-interaction limit Roman Fanta, Matus Dubecky Fixed-node diffusion Monte Carlo (FNDMC) method with single-determinant (SD) trial wave functions gains popularity as a scalable benchmark approach suitable for large noncovalent systems. The talk will summarize recent improvements and accuracy limits of the best available bias-cancellation-based SD FNDMC computational protocol suitable for a large class of noncovalent systems. |
Monday, March 2, 2020 12:03PM - 12:15PM |
B40.00003: Simulating realistic features of chemical and materials systems with massively parallel many-body perturbation theory calculations Derek Vigil-Fowler, Jacob Clary, Charles Musgrave, Aaron M Holder With experiments obtaining increasingly fine resolution in time, energy, and position, there is a growing need for high-fidelity calculations that can match this rapidly improving resolution. Many-body perturbation theory is one approach that has had significant successes in the accurate simulation of the optoelectronic properties of many materials classes and also of surface chemistry. These methods have traditionally been applied to comparatively simple materials with relatively small numbers of atoms and atom types in a simulation cell, but increasing computing power allows for calculations on systems with hundreds and evens thousands of atoms in a simulation cell. In this talk we will discuss our work on simulating the electrochemical oxygen reduction reaction (ORR) on a system of FeN4clusters in graphene using the many-body random phase approximation (RPA) and including the effects of zero-point energy, vibrational entropy, solvation, and applied bias. We will also discuss calculations using the many-body GW-Bethe Salpeter equation (GW-BSE) approach to calculate the optoelectronic properties of complex materials including double perovskites. |
Monday, March 2, 2020 12:15PM - 12:27PM |
B40.00004: Large temperature sensitivity of optical emission via synergy of Raman and photoluminescence effects in BaTiO3-based phosphor ARNAB DE, Rajeev Ranjan The growing interest in luminescence-based optical thermometry in recent years is motivated by the prospect of designing phosphor materials capable of exhibiting large temperature sensitivity of the emission spectrum. The prevalent material design strategies predominantly rely on ascertaining temperature sensitivity using fluorescence intensity ratio (FIR) of the Stark lines in the photoluminescence (PL) emission spectrum of rare-earth(s) doped materials. This approach has yielded temperature sensitivity in the range 0-5% . Here we propose a new strategy for achieving high temperature sensitivity by considering a twin combination of Raman and PL spectrum of the phosphor material, and by introducing Raman-PL-intensity ratio (RPIR) as a parameter for measuring relative temperature sensitivity(Sr). The effectiveness of this concept is demonstrated on Eu/Er-doped BaTiO3 wherein a large Sr range (-3 % to +6 %) and large color tuning (red to green) was achieved by exploiting the mechanisms which resulted in a contrasting effect of temperature on Raman and the PL part of the total spectra. |
Monday, March 2, 2020 12:27PM - 12:39PM |
B40.00005: NWChemEx – Computational Chemistry for the Exascale Era Hubertus van Dam, Edoardo Apra, Raymond Bair, Jeffery S Boschen, Eric J. Bylaska, Wibe A De Jong, Thomas H Dunning, Niranjan Govind, Robert J Harrison, Kristopher Keipert, Karol Kowalski, Sriram Krishnamoorthy, Suraj Kumar, Erdal Mutlu, Bruce Palmer, Ajay Panyala, Bo Peng, Ryan M Richard, T P Straatsma, Edward F Valeev, Marat Valiev, David B Williams-Young, Chao Yang, Theresa L Windus NWChemEx is an ECP project for computational chemistry that builds on the success of NWChem. NWChem is an early co-design project started in 1992, to address distributed memory parallel computers. The code's modular design, the Global Arrays for distributed data handling, and code generation using the Tensor Contraction Engine, provided a platform for building scalable chemistry capabilities. Key capabilities of the code are MD, DFT methods, and coupled cluster. |
Monday, March 2, 2020 12:39PM - 12:51PM |
B40.00006: Accessible and collaborative web interface for HPC materials simulations Timur Bazhirov High Performance Computing (HPC) predictive capabilities are of key importance to chemical and materials sciences and engineering, contributing to the goal of the Materials Genome Initiative for enhancing the rate of breakthroughs in complex materials chemistry and materials design. However, recent advances in theory, algorithms, and HPC hardware for materials and chemical sciences, especially aimed at the exascale, are yet to be widely available to the majority of scientifically and technically capable communities in the United States [1]. We outline the concept for the creation of a web-enabled infrastructure for predictive theory and modeling [2] able to facilitate access to, coordination and sharing of information and data produced by scalable codes adapted for exascale computing. In addition, we explain how our web-based infrastructure can help promote universal standards for data inputs and outputs in the multitude of codes and methodologies and discuss the use cases. |
Monday, March 2, 2020 12:51PM - 1:03PM |
B40.00007: Large-scale many-body perturbation theory calculations on leadership class facilities Marco Govoni, He Ma, Francois Gygi, Giulia Galli Many-body perturbation theory (MBPT) has been shown to provide an accurate description of excited state properties for the simulation of spectroscopic signatures of materials and molecules. The advent of exascale computing offers the appealing opportunity to expand the applicability of MBPT methods to systems of unprecedented size and complexity, e.g. to thousands of electrons. We will discuss methodological advances implemented in the WEST code [http://west-code.org] for both GW and BSE calculations. We will present new functionalities enabled by the concurrent use of WEST and the Qbox code [http://qboxcode.org], with focus on interoperability paradigms. We will discuss the advantages of developing interoperable software, and we will present results for the calculation of spectroscopic properties of defective insulators and semiconductors hosting optically addressable spin-defects. |
Monday, March 2, 2020 1:03PM - 1:15PM |
B40.00008: Electronic density and atomic forces in solids by plane-wave auxiliary-field quantum Monte Carlo Siyuan Chen, Mario Motta, Fengjie Ma, Shiwei Zhang We present accurate electronic densities for several prototypical solids, including the ionic crystal NaCl, covalent-bond semiconductor Si, and metalic Cu. These results are obtained using ab initio auxiliary-field quantum Monte Carlo (AFQMC) method1 working in a plane-wave basis with multiple-projector, norm-conserving pseudopotentials. AFQMC has been shown to be an excellent many-body total energy method. Computation of observables and correlation functions other than the ground-state energy requires back-propagation2, which we have implemented in this work in the plane-wave basis AFQMC framework. Our (near-exact) results are used to benchmark several density functionals, which can be useful in constructing improved density functionals. Besides density, we have also implemented the calculation of atomic forces, which paves the way for geometry optimization in solids. |
Monday, March 2, 2020 1:15PM - 1:27PM |
B40.00009: A pseudo-BCS wavefunction from density matrix decomposition - application in auxilary-field quantum Monte Carlo Zhi-Yu Xiao, Hao Shi, Shiwei Zhang We present a method to construct BCS-like (pseudo-BCS) wave functions from the one-body density matrix. |
Monday, March 2, 2020 1:27PM - 1:39PM |
B40.00010: Electronic band gaps from Quantum Monte Carlo methods Yubo Yang, Vitaly Gorelov, CARLO PIERLEONI, David Ceperley, Markus Holzmann We develop a method for calculating the fundamental electronic gap of semiconductors and insulators using grand canonical Quantum Monte Carlo simulations. We discuss the origin of the bias introduced by supercell calculations of finite size and show how to correct the leading and sub-leading finite size errors either based on observables accessible in the finite-sized simulations or from DFT calculations. Our procedure is applied to solid molecular hydrogen and compared to experiment for carbon and silicon crystals. arXiv: 1910.07531 |
Monday, March 2, 2020 1:39PM - 1:51PM |
B40.00011: RuCl3 electronic structure by quantum Monte Carlo Abdulgani Annaberdiyev, Cody Melton, Raymond C Clay, Luke Shulenburger, Guangming Wang, Lubos Mitas The alpha phase of RuCl3 crystal consists of 2D stacked honeycomb layers coupled with van der Waals interactions. Both bulk and multi/single layers of bonded RuCl6 octahedra have been produced and exhibit several promising electronic, magnetic and structural properties. The ground state is a magnetic insulator with several possible magnetic states that are energetically very close. Furthermore, the material allows for a formation of putative spin liquid phases and other unconventional magnetic states due to Ru spin-orbit coupling and favorable orbital occupations. We study these systems using fixed-node quantum Monte Carlo (QMC) with averaged spin-orbit as well as with recently developed fixed-phase spin-orbit QMC that is based on two-component spinors formalism. In particular, we attempt to understand the role of electron correlations as they affect the ground and excited states. We carry out calculations for the low-lying magnetic states vs non-magnetic state and attempt to reveal the impact of explicit spin-orbit effects on magnetic and electronic properties. |
Monday, March 2, 2020 1:51PM - 2:03PM |
B40.00012: Heavy-atom systems in quantum Monte Carlo: pseudopotentials and beyond Guangming Wang, Abdulgani Annaberdiyev, Cody Melton, Michael Bennett, Luke Shulenburger, Lubos Mitas We study the electronic structure of selected systems with heavy atoms such as Ru, Ir, Pb, Bi and I, using quantum Monte Carlo (QMC). This is motivated by expanding real space QMC to systems with strong spin-orbit interactions and significant correlation effects. Such studies require accurate effective core potentials (ECPs), the ability to obtain accurate spinors and the eventual inclusion of multi-reference expansions of trial wave functions. We start by assessing the accuracy of ECPs and their impact on the most basic quantities such as the lowest energy excitations and binding in atomic and molecular systems. Moreover, we try to assess the errors caused by averaged vs. explicit spin-orbit interaction using the recently developed two-component spinor fixed-phase QMC method. We study also the corresponding biases that stem from the fixed-node vs fixed-phase approximations. Furthermore, we try to explore the cases where spin-orbit and correlation are of the same magnitude and can impact important quantities such as band gaps and magnetic states in periodic materials. |
Monday, March 2, 2020 2:03PM - 2:15PM |
B40.00013: Mitigating the Sign Problem Through Basis Rotations Ryan Levy, Bryan Clark Quantum Monte Carlo simulations of quantum many body systems are plagued by the fermion sign problem. The cost of overcoming the sign problem (and its overall existence) is basis dependent. In this talk, we show how to use sign-free quantum Monte Carlo simulations to minimize the effect of the sign problem by optimizing over the choice of basis on large two-dimensional systems. This can be done despite the fact that the sign problem makes it difficult to even compute the average sign of such a simulation. We numerically illustrate these techniques by decreasing the 'badness' of the sign problem through single-particle basis rotations on one and two dimensional Hubbard systems. |
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