Bulletin of the American Physical Society
APS March Meeting 2021
Volume 66, Number 1
Monday–Friday, March 15–19, 2021; Virtual; Time Zone: Central Daylight Time, USA
Session V60: Machine Learning for Correlative and Analytical MeasurementsFocus Live
|
Hide Abstracts |
Sponsoring Units: GIMS Chair: Stephen Jesse, University of Tennessee |
Thursday, March 18, 2021 3:00PM - 3:12PM Live |
V60.00001: Sequential Bayesian experimental design for everyday measurements Robert McMichael, Sergey Dushenko, Sean M Blakley This talk gives an overview of software that increases speed and precision of routine measurements. In experiments where one would traditionally fit data to extract parameters, the ‘optbayesexpt’ package recommends measurement settings “on the fly” based on analysis of accumulating data. The algorithm uses optimal Bayesian experimental design to predict settings with the best chance of reducing parameter uncertainty. In simulations and in tests, we demonstrate order-of-magnitude speedup in measurements of Lorentzian peaks and significant speedup of exponential decay measurements relative to measure-then-fit strategies. The package, written in Python, includes a server script that communicates with instrument control software in any popular instrument control language. Demonstrations include magnetic resonance spectra of NV centers in diamond, and simulations include calibration of π-pulses for spin control. See the manual at https://pages.nist.gov/optbayesexpt/ and software at https://github.com/usnistgov/optbayesexpt. |
Thursday, March 18, 2021 3:12PM - 3:24PM Live |
V60.00002: Machine Learning Correlates Charge Density Wave with the Local Gap in Cuprate Superconductors Kaylie Hausknecht, Tatiana Webb, Michael Boyer, Yi Yin, Takeshi Kondo, Tsunehiro Takeuchi, Hiroshi Ikuta, Eric Hudson, Jenny E. Hoffman With the advent of atomic resolution imaging techniques comes the challenge of disentangling the intrinsic electronic properties of materials from their stochastic atomic-scale disorder. In the past decade, machine learning (ML) image analysis techniques have rapidly evolved, while their applications in physics are just emerging. Here, we use ML to test local correlation hypotheses between spatially resolved measurements of disordered materials to overcome the limitations of standard Fourier analysis techniques. By training on a simulated density wave (DW) dataset, we develop a convolutional neural network (CNN) to uncover the doping-dependence of the DW in the cuprate superconductor (Pb,Bi)2(Sr,La)2CuO6+δ (Bi-2201) imaged via scanning tunneling microscopy. In Bi-based cuprates, the electronic inhomogeneity, caused by local variations in doping, limits the precision with which the DW wavevector can be measured. Our ML algorithm overcomes this limitation and allows clear differentiation between commensurate and incommensurate DW instabilities with physically distinct mechanisms. More broadly, our work lays the foundation for a ML approach to quantify intrinsic periodic order and correlations in datasets where these trends are masked by disorder. |
Thursday, March 18, 2021 3:24PM - 3:36PM Live |
V60.00003: Using Machine Learning for noise reduction in X-ray Photon Correlation Spectroscopy data to quantify time series dynamics Tatiana Konstantinova, Lutz Wiegart, Anthony DeGennaro, Andi Barbour Computational methods of noise reduction in data allow for reliable extraction of useful signals from a limited amount of experimental data. This opens the door for optimal use of experimental resources and obtaining information from intrinsically limited, e.g. destructive or out-of-equilibrium, measurements. Here, I present the application of the convolutional deep learning models for the reduction of noise in intensity-intensity correlation functions from X-ray Photon Correlation Spectroscopy (XPCS) experiments. This approach creates a filter tailored to specific types of noise encountered in XPCS experiments and results in up to 20-fold reduction of required experimental data. |
Thursday, March 18, 2021 3:36PM - 3:48PM Live |
V60.00004: Neural network temperature predictions based on the optical properties of quantum dots John Colton, Charles Lewis, James W Erikson, Carrie E McClure, Jordan Bryan, Marissa Iraca, Derek Sanchez, Greg Nordin, Troy Munro The vital role of temperature in many biological processes creates a need to develop improved microscopic temperature sensors. The photoluminescence (PL) of quantum dots (QDs) has previously been used as a temperature probe in microfluidic devices, but the accuracy of the reconstructed temperature was limited to about 1 K over a temperature range of tens of degrees. In this work we present a machine learning neural network algorithm which uses a combination of normalized spectral and time-resolved PL data of QD emission in a microfluidic device to predict temperature. The neural network was trained with PL collected under known temperatures, then tested with holdout PL data not involved in the training process. The accuracy of the temperature predictions was 0.1 K. We present results for CdTe QDs, as well as recent extensions of this work into other types of QDs. While an ongoing issue has been long term accuracy of the temperature predictions as the properties of the QDs can vary over time, this method demonstrates a potential approach to accurately sense temperature in microfluidic (and possibly nanofluidic) devices via optical measurements. |
Thursday, March 18, 2021 3:48PM - 4:00PM Live |
V60.00005: Reverse modelling for Lorentz transmission electron microscopy William Perry, Min He, Ying Zhang, Xiaoguang Zhang Lorentz Transmission Electron Microscopy (LTEM) is a versatile measurement technique that can probe the magnetic configuration of thin films under a wide variety of experimental conditions. One limitation of this measurement technique is that it is only sensitive to projection of the sample’s magnetic moment onto a two dimensional plane that is perpendicular to an electron beam which is passed through the sample. We have developed an algorithm which linearizes the relationship between the experimental sample and measured phase change. The resulting underdetermined system can be augmented by solving to two or more different measurements simultaneously and can be solved using iterative optimizers. We apply this to measurements of NdFeB to obtain a magnetic configuration which would not be discoverable by treating measurements individually. |
Thursday, March 18, 2021 4:00PM - 4:36PM Live |
V60.00006: Navigating atomic-scale disorder with correlative tunneling microscopy and defect manipulation Invited Speaker: Petro Maksymovych Disorder is a powerful approach to elicit and control quantum properties in the bulk, as evidenced from record Tc, quantum phase transitions and exotic quasiparticles, predicted or evidenced in disordered superconducting and topological materials. Here we will introduce several new approaches to reveal, understand and introduce disorder in unconventional superconductors on the nanoscale, by augmenting scanning tunneling microscopy with machine learning, force microscopy and new defect manipulation methods. We will demonstrate reduced-dimensionality manifolds as a natural choice for sparse representation of heterogeneous disorder effects, with the potential to mirror the success of integral transforms for quantitative analysis of periodic structures. The techniques of metric analysis enables optimization of machine learning techniques to both amplify the disorder signatures in the hyperspectral data-sets, detect artefacts and reveal otherwise hidden properties of spectral weight transfer. Finding correlations within datasets can reveal the very mechanism of tunneling and imaging modes, such as statistically identifying the regime of Josephson tunneling in STM with a superconducting tip. Finally, correlative imaging of atomic-scale forces and tunneling current reveals unexpected properties of superconductor surfaces, including subsurface defects and surface phases, and enables injection of new kinds of defects with sub-nm resolution. The combination of these techniques paves way to comprehensive analysis of disorder across measurements and materials, and ultimately developing the “axis of disorder” as an approach to control superconducting and topological properties with near atomic-scale accuracy. |
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