Bulletin of the American Physical Society
APS March Meeting 2020
Volume 65, Number 1
Monday–Friday, March 2–6, 2020; Denver, Colorado
Session F20: Data Science I: Big Data & MLFocus
|
Hide Abstracts |
Sponsoring Units: GDS Chair: William Ratcliff, National Institute of Standards and Technology Room: 301 |
Tuesday, March 3, 2020 8:00AM - 8:36AM |
F20.00001: Information extraction, analysis and feedback for directing matter by design Invited Speaker: Bobby Sumpter Recent technical advances in the area of nanoscale imaging, spectroscopy and scattering/diffraction have provided tremendous capabilities for investigating materials structural, dynamical and functional characteristics. At the same time, advances in computational algorithms, including deep learning approaches, and computer capacities that are orders of magnitude larger and faster, have enabled extreme-scale simulations and deep data analytics of materials properties and processes starting with nothing but the identity of the atomic species and the basic principles of quantum and statistical mechanics and thermodynamics. This powerful confluence of capabilities/advances and the information bound in large volumes of high-quality data, offers new opportunities for advancing materials and chemical sciences. In this talk I will discuss how we are probing in-situ, chemical reactions and materials transformations, including hierarchical assembly, as a modality for direct feedback to an experiment in order to precisely impart directed energy (electrons, ions, photons, thermal) that manipulates a material at the nanoscale. This approach is enabled via the dual capability of high-resolution imaging and focused energy in-situ, with data rates, quality and volumes that allow for a deep learning framework to efficiently identify structures and dynamics across broad length and time scales. We have found that the approach can provide efficient mapping of solid-state reactions and transformations and overall allows a major step toward directing matter by design. |
Tuesday, March 3, 2020 8:36AM - 8:48AM |
F20.00002: High throughput search for plasmonic semiconductors using DFT databases Ethan Shapera, Andre Schleife The field of plasmonics aims to manipulate light via choice of materials and nanoscale structuring. Finding materials which exhibit low-loss responses to applied optical fields while remaining feasible for widespread use is an outstanding challenge. Online databases have compiled structural and electronic data for of tens of thousands of materials, but lack the expensive optical response calculations needed for selection of plasmonic materials. Understanding the optical response of semiconductors within DFT is complicated due to the DFT bandgap error and the influence of carrier doping density. We describe, validate, and demonstrate an approach which rapidly screens existing online databases to identify high quality factor plasmonic semiconductors. Using DFT bandstructures, we tabulate material bandgaps, optical transition energies, and effective masses. From correlations between bandstructure features and quality factor computed for a restricted set of semiconductors we predict CaMg2N2 , InGaO3, and LiInO2 as candidate high quality factor plasmonic semiconductors. Quality factors are verified with DFT to describe optical absorption and the Drude model for intraband transitions. |
Tuesday, March 3, 2020 8:48AM - 9:00AM |
F20.00003: Combined High-Throughput and Machine Learning Approach for Prediction of Lattice Thermal Conductivity Rinkle Juneja, George Yumnam, Swanti Satsangi, Abhishek Singh Search for materials via explicit evaluation of thermal conductivity (κl) either experimentally or computationally is very challenging. We carried out high-throughput screening on a dataset containing total of 2691 binary, ternary, and quaternary compounds. The κl values of 120 dynamically stable and nonmetallic compounds are calculated. Among these, 11 ultrahigh and 15 ultralow κl materials are identified. For the machine learning prediction models, the descriptor set is usually tuned via conventional algorithms such as least absolute selection and shrinkage operator (LASSO). However, we generated an extensive property map from high-throughput calculations to design a minimal set of descriptors directly related to the physics of κl. These simple descriptors are maximum phonon frequency, integrated Grüneisen parameter up to 3 THz, average atomic mass, and volume of the unit cell. Using these descriptors, a Gaussian process regression-based machine learning (ML) model is developed. The model predicts room temperature log-scaled κl with a very small root mean square error. The superior performance of the ML model can ensure a reliable and accelerated search for a multitude of low and high κl materials. |
Tuesday, March 3, 2020 9:00AM - 9:12AM |
F20.00004: Machine Learning-Assisted Design and Discovery of Next Generation 2D Materials Victor Venturi, Holden Low Parks, Zeeshan Ahmad, Venkat Viswanathan Atomically thin two-dimensional materials have attracted interest in the fields of electrochemistry, catalysis, and photonics due to the ease with which their properties may be tuned. First-principles calculations have proven an essential tool in the quest for new 2D materials with tailored properties. However, an exhaustive exploration of the parameter space is infeasible even in the monolayer case. In this work, we use a novel machine learning technique – Crystal Graph Convolutional Neural Networks (CGCNN) [1] – to train accurate models that can predict monolayer 2D material properties more efficiently than density functional theory simulations. Previously, CGCNN architectures have been demonstrated to successfully predict properties of solid electrolytes [2]. Here, we leverage their power to find design principles for 2D materials in light-absorbing and water-splitting applications. |
Tuesday, March 3, 2020 9:12AM - 9:24AM |
F20.00005: Revealing the Spectrum of Unknown Layered Materials with Super-Human Predictive Abilities Gowoon Cheon, Ekin Dogus Cubuk, Evan Antoniuk, Joshua Goldberger, Evan J. Reed We use semi-supervised learning to discover over 1000 new two-dimensional layered materials that have yet to be discovered or synthesized. We accomplish this by combining physics with machine learning on experimentally obtained data and verify a subset of candidates using density functional theory. Our model accelerates the discovery of layered materials by 13 times compared to random trial-and-error approaches. Even compared to expert scientists working in the field of two-dimensional materials, it is five times better than practitioners in the field at identifying layered materials and is comparable or better than professional solid-state chemists. We also find that our model is orders of magnitude faster than any human. |
Tuesday, March 3, 2020 9:24AM - 9:36AM |
F20.00006: Probing the microscopic origin of magnetism in two-dimensional materials using machine learning Trevor David Rhone, Efthimios Kaxiras Magnetic ordering in two-dimensions is at the forefront of research since the discovery of magnetism in monolayer CrI3 in 2017. We study two-dimensional (2D) materials with intrinsic magnetic order and explore the microscopic origins of magnetism in these novel materials. The Mermin-Wagner theorem asserts that magnetic ordering cannot occur in 2D without the presence of magnetocrystalline anisotropy (MCA), which arises due to spin-orbit coupling. 2D materials with magnetic order provide a platform with which to study magnetism in low dimensions – the positions of all the atoms are known in theoretical studies and some experimental studies. This contrasts studies of magnetism in thin films. We use data analytics to study the magnetic and thermodynamic properties of 2D materials. Crystal structures based on monolayer CrI3, are studied using density functional theory (DFT) calculations and machine learning. Magnetic properties, such as MCA and the magnetic moment are determined. We show that machine learning, combined with DFT, provides a means to learn patterns in 2D magnetic materials data, thereby providing insights into the microscopic origins of magnetic ordering in 2D. This approach to materials research also facilitates the rapid discovery of 2D magnets. |
Tuesday, March 3, 2020 9:36AM - 9:48AM |
F20.00007: Topological data analysis for magnetic domain structure characterization Masato Kotsugi Analysis of the magnetic domain structure is an important process in the development of advanced magnetic materials. A number of successful studies have been carried out by high-resolution magnetic imaging or LLG simulations. However, the treatment of the information of the whole image, especially the shape of the magnetic domain structure, has not been sufficiently discussed. |
Tuesday, March 3, 2020 9:48AM - 10:00AM |
F20.00008: Extracting Interpretable Physical Parameters from Spatiotemporal Systems using Unsupervised Learning Peter Lu, Samuel Kim, Marin Soljacic Real-world data from spatiotemporal systems is often difficult to analyze and interpret due to complex dynamics as well as uncontrolled experimental variables. We demonstrate an unsupervised machine learning technique for extracting interpretable physical parameters from noisy spatiotemporal data and for building a transferable predictive model of the system. This is accomplished without prior knowledge of the underlying dynamics or the governing partial differential equation (PDE). Numerical experiments using simulated data governed by PDEs show that our method accurately identifies and extracts relevant parameters that characterize independent variations in the system dynamics. Our method for discovering interpretable latent parameters in spatiotemporal systems will allow us to better understand real-world phenomena by analyzing datasets with varying dynamical behaviors that are difficult to disentangle. |
Tuesday, March 3, 2020 10:00AM - 10:12AM |
F20.00009: Prediction of Seismic Wave Arrivals Using a Convolutional Neural Network Jorge Garcia, Lauren Waszek Seismology uses energy from earthquakes to image the deep interior of Earth. Large amounts of seismic data are required in order to obtain detailed observations of the its internal structure; typical datasets comprise over 100,000 seismic records. With the exception of some basic processing methods, compilation of the data is performed by hand using simple visualization software. The most significant and time-consuming task is the identification and picking of seismic phases. Previous attempts at automating this procedure involve algorithms that generally underperform compared to a human expert. However, even among human-compiled datasets, consistency of phase arrival across and within datasets is a problem. The variation in decisions results in disagreement between obtained images, and subsequent interpretation of Earth’s structure and processes. We employ a Convolutional Neural Network (CNN) to predict the arrival time of the mantle shear-wave phases in a seismogram in an effort to accelerate and make consistent the task of data processing. We appraoch strategies to ensure correct prediction of arrival time and polarity of the seismic phase, as well as applying the model for identification of precursor signals of the same phase. |
Tuesday, March 3, 2020 10:12AM - 10:24AM |
F20.00010: Using Reinforcement Learning to Optimize Crystal Structure Determination William Ratcliff, Kate Meuse, Jessica Opsahl-Ong, Joeseph Rath, Paul Kienzle, Telon Yan, Ryan Cho, Abigail Wilson The first step to understanding the microscopic origins of the properties of a material is to determine the crystal structure. This can be accomplished with neutron diffraction. However, there are a small number of neutron sources in the world and thus it is critical to perform measurements as optimally as possible. We use reinforcement learning to address this problem. We compare several approaches within this framework including epsilon-greedy, Q-learning, and actor-critic. We find that in toy models, it is possible to measure a significantly smaller fraction of measurements than would commonly be performed to determine structural properties with the same accuracy. |
Tuesday, March 3, 2020 10:24AM - 10:36AM |
F20.00011: Active Learning for Quantum Experimental Controling Yadong Wu, Hui Zhai Experimental control problems are popular task for experimental physicists in quantum exper- iments. Generally human tune the experimental parameters step by step, hand by hand to fine the suitable experimental parameters to set up the experimental system. While this way of tuning parameters is not efficient and may miss the ’global minimum’. Here we apply one semi-supervised machine learning method, active learning, to this parameters tuning task to find the suitable param- eters automatically. We will show the advantage of this machine learning method by two simulated examples. First we set the Efimovian expansion as a benchmark. Putting the unitary fermi gases into a harmonic trap with time dependent trapping frequency ω(t), active learning told us the most efficient way to release the trap is ω(t) ∝ 1/t which is assistant with the theory. Then we apply this active learning to evaporative cooling issue. Comparing to the simulated result, this active learning can give us a better cooling trajectory which can reach to a lower temperature during less time. |
Tuesday, March 3, 2020 10:36AM - 10:48AM |
F20.00012: International Radiological Information Exchange (IRIX) Standards for Emergency Radiation Monitoring Data Reporting Sanjoy Mukhopadhyay The Incident and Emergency Centre (IEC) of the International Atomic Energy Agency (IAEA) has developed a web application, the International Radiation Monitoring Information System (IRMIS). IAEA Member States (MS) can share and visualize large quantities of radiation monitoring data (viz. gamma dose rate, isotope specific ground depositions and air concentrations) using IRMIS. The radiation monitoring data may be uploaded into IRMIS in the International Radiological Information Exchange (IRIX) format. IRIX is a technical standard developed by the IAEA, in cooperation with the MSs and the European Commission (EC). IRIX enables the development of interoperable systems and solutions for exchanging emergency information and data between organisations at both national and international level during a nuclear or radiological emergency. IRIX is an open format based on the Extensible Markup Language (XML), which makes it both machine and human readable. The system interface specification (or web-service specification) enables organizations to interconnect their emergency information systems to automate information exchange. The XML provides a software- and hardware-independent way of storing, transporting, and sharing data. The article will discuss applications of IRIX in IRMIS. |
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