Bulletin of the American Physical Society
APS March Meeting 2020
Volume 65, Number 1
Monday–Friday, March 2–6, 2020; Denver, Colorado
Session W39: Machine Learning for Quantum Matter VIFocus
|
Hide Abstracts |
Sponsoring Units: DCOMP GDS DMP Chair: Giacomo Torlai, Simons Foundation Room: 703 |
Friday, March 6, 2020 8:00AM - 8:36AM |
W39.00001: Differentiable programming tensor networks and quantum circuits Invited Speaker: JinGuo Liu Computation is playing an increasingly important role in the studies of complex quantum systems. Efficient and exact gradients from automatic differentiation (AD) changes the way we program. This talk covers a brief survey of the state of the art differential programming frameworks, and their applications to condensed matter physics and quantum computing. These application range from optimizing infinite tensor network states to simulating variational quantum algorithms. Lastly, I will introduce reversible computing as the host of next generation differential programming framework, which may unleash the full power of AD for differentiable scientific computing. |
Friday, March 6, 2020 8:36AM - 8:48AM |
W39.00002: Machine learning effective models from a Boltzmann perspective Jonas Rigo, Andrew Mitchell We investigate the derivation of effective models for quantum impurity type problems using machine learning methods. Parameters of the effective model are optimized with respect to the parent Hamiltonian by using a classical probability distribution extracted from a diagrammatic expansion of the partition function. The classical probability distributions are naturally in the form of an energy-based model within this framework, making clear the connection to Boltzmann machines. In this case, the energy-based model is the effective model and has a physical meaning. The information geometry inspired derivation of the cost function predicts that the best fitting effective model has the same thermal expectation value of effective interactions as in the parent model. However, we show that this does not necessarily yield an effective model with the same low-energy physics as the parent model due to information monotonicity along RG flow [1]. |
Friday, March 6, 2020 8:48AM - 9:00AM |
W39.00003: Automatic design of Hamiltonians Kiryl Pakrouski We formulate an optimization problem of Hamiltonian design. Given a variational ansatz for a Hamiltonian we construct a loss function to be minimised as a weighted sum of relevant Hamiltonian properties specifying thereby the search query. Using fractional quantum Hall effect as a test system we illustrate how the framework can be used to determine a generating Hamiltonian of a finite-size model wavefunction (Moore-Read Pfaffian and Read-Rezayi states) or to find optimal conditions for an experiment. We also discuss how the search for approximate generating Hamiltonians may be used to find simpler and more realistic models implementing the given exotic phase of matter by experimentally accessible interaction terms. Based on arXiv:1907.05898. |
Friday, March 6, 2020 9:00AM - 9:12AM |
W39.00004: Direct and Reverse Structure-Electronic Property Relationship Prediction with Deep Learning and Bayesian Optimization Artem Pimachev, Sanghamitra Neogi Ab-initio computational approaches fail to predict electronic properties of systems with increasing size and complexity. It is highly desirable to formulate methods that can translate the information from ab-initio techniques across length scales, and predict electronic properties of technological applications at relevant length scales. We use a deep learning model to find the complex relationship between the geometrical attributes of heterostructures (HS) and the electronic structure properties predicted by ab-initio calculations. We test the usefulness of the model by predicting electronic transport properties of unknown Si/Ge HS based on the relationship, and comparing with experimental data. We consider the local atomic environment features and the global HS features as the descriptors of the model. The advantage of deep learning models becomes evident since the configuration space is vast due to the variability of HS fabrication. We propose a reverse approach based on Bayesian optimization to predict the structure from measured system’s properties of interest. This method is generically applicable to predict electronic properties of dynamically varying heterostructures and the discovery of new structures. |
Friday, March 6, 2020 9:12AM - 9:24AM |
W39.00005: Machine Learning of Single-Atom Defects in 2D Transition Metal Dichalcogenides with Sub-Picometer Precision Abid Khan, Bryan Clark, Chia-Hao Lee, Di Luo, Chuqiao Shi, Sangmin Kang, Wenjuan Zhu, Pinshane Huang Deep learning techniques based on fully convolutional networks (FCNs) have revolutionized image recognition in fields ranging from medical diagnosis to facial recognition. The ability of FCNs to identify objects in images opens new opportunities for accessing the underlying information in atomic-resolution images obtained using scanning transmission electron microscopy (STEM). The properties of two-dimensional transition metal dichalcogenides (2D TMDCs) are strongly influenced by atomic defects such as vacancies and substitutional dopants, but high-precision characterization of single-atom defects remains challenging because 2D materials are irradiation sensitive, produce low scattering signals, and require low-voltage imaging modes. While identifying defects by hand is possible, it severely limits our ability to process large quantities of atoms and obtain large-scale statistics. By employing deep learning techniques, we quickly identify and classify various defect species, including metal substitutions, chalcogen vacancies, and chalcogen substitutions. This approach lets us observe changes in atomic separations induced by these defects with sub-picometer precision. |
Friday, March 6, 2020 9:24AM - 9:36AM |
W39.00006: Dictionary Learning in Fourier Transform Scanning Tunneling Spectroscopy Jedrzej Wieteska, Yenson Lau, Tetsuo Hanaguri, John Wright, Ilya Eremin, Abhay Pasupathy Modern high-resolution microscopes, such as the scanning tunneling microscope, are commonly used to study specimens that have dense and aperiodic spatial structure. Extracting meaningful information from images obtained from such microscopes remains a formidable challenge. Fourier analysis is commonly used to analyze the underlying structure of fundamental motifs present in an image. However, the Fourier transform fundamentally suffers from severe phase noise when applied to aperiodic images. We have developed a new algorithm based on nonconvex optimization, applicable to any microscopy modality, that directly uncovers the fundamental motifs present in a real-space image. Apart from being quantitatively superior to traditional Fourier analysis, this novel algorithm also uncovers phase sensitive information about the underlying motif structure. We apply this algorithm to scanning tunneling microscopy images of an S-doped iron selenide superconductor to recover phase-sensitive quasiparticle interference in this material as function of sulfur doping. Implications of our results on the evolution of the superconducting gap structure across the putative nematic quantum critical point will be discussed. |
Friday, March 6, 2020 9:36AM - 9:48AM |
W39.00007: Machine Learning Tool for Crystal Structure Predictions Valentin Stanev, Haotong Liang, Aaron Kusne, Ichiro Takeuchi Structure is the most basic and important property of crystalline solids; it determines directly or indirectly most other materials characteristics. However, predicting the crystal structure of solids remains a formidable and not fully solved problem; standard theoretical tools for the task are computationally expensive and not always reliable. In this talk I will present an alternative approach that utilizes machine learning for crystal structure predictions. We developed a tool that can predict the Bravais lattice, space group and lattice parameters of a material based only on its chemical composition. It consists of a series of neural network models with predictors based on aggregate properties of the elements constituting the compound. The tool was trained and validated on more than 100,000 entries from the Inorganic Crystal Structure Database (ICSD). It demonstrates good predictive power and significantly outperforms alternative strategies. This machine learning tool is easy to use, and can be utilized both as an independent prediction engine or as a data-informed method to generate candidate structures for further exploration. |
Friday, March 6, 2020 9:48AM - 10:00AM |
W39.00008: Transferable and interpretable machine learning model for four-dimensional scanning transmission electron microscopy data Michael Matty, Michael Cao, Zhen Chen, Li Li, David Muller The challenge brought to scientific discovery by the data revolution may be overcome by data scientific approaches. Here we focus on 4D scanning transmission electron microscopy (STEM) data. With advances in detector technology, STEM records the full scattering distribution at each scan position in real space, producing a 4D phase-space distribution. An efficient approach is needed to turn these data into a real space image with subatomic resolution. Existing approaches are limited: annular dark field (ADF) imaging by low dose efficiency and resolution, and ptychography to a few atomic layers and by high computational cost. Here, we develop an efficient, interpretable machine learning model to map the entire STEM dataset to real space images. Our model has higher contrast than ADF, still distinguishes atomic species, and transfers well between samples of different lattice symmetry. We benchmark against conventional approaches using quantitative metrics for resolution and contrast. |
Friday, March 6, 2020 10:00AM - 10:12AM |
W39.00009: Tight-binding deep learning approach to band structures calculations Florian Sapper, Vittorio Peano, Florian Marquardt Band structures are ubiquitous in physics as they describe natural as well as engineered materials. Even for systems for which the numerical calculation of the band structure for a single configuration is in itself not proibitevely expensive, efficient numerical methods are highly valuable as they allow the systematic investigation of large sets of configurations. This is often required because configuration spaces are typically huge. |
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