Understanding the complex behavior of electrons has led to discoveries that have changed society, such as the data processing revolution made possible by the invention of the transistor.
Technological advances have allowed us to study the behavior of electrons much more closely than in the past, potentially allowing scientific breakthroughs that are changing worldwide like the personal computer. However, the data that these tools generate is too complex for humans to interpret.
A Cornell-led team has developed a methodology that allows machine learning to analyze the data generated by scanning tunneling microscopy (STM). Images of electronic motions in material surfaces with different energies provide information that can not be obtained by any other method.
"Some of these images were taken on materials that have been considered important and mysterious for two decades," said Professor Eun-Ah Kim of physics. "You wonder what secrets are hidden in these pictures, we want to unlock those secrets."
Kim is senior author of "Machine Learning in Electronic Quantum Imaging Experiments" published in Nature June 1
Co The authors include JC Séamus Davis, James Gilbert's White Distinguished Professor of Physical Sciences from Cornell, an innovator in STM-driven discovery studies in experimental quantum physics.
On the subatomic scale, a given sample contains trillions of trillion electrons that interact with each other and with the surrounding infrastructure. The behavior of electrons is determined in part by the tension between their two competing tendencies: to move, combined with kinetic energy; and to stay far apart, with repulsive interaction energy.
In this study, Kim and co-workers wanted to find out which of these tendencies are more important in a high-temperature superconducting material.
Electrons tunnel through a vacuum between the conductive tip of the microscope and the surface of the sample under investigation, providing detailed information about the behavior of the electrons.
"The problem is, when you record and record such data, you get image-like data, but it's not a natural picture, like an apple or a pear," Kim said. The data generated by the instrument is more of a pattern and about 10,000 times more complicated than a traditional waveform. "We do not have a good tool to study these types of records."
To interpret this data, the researchers simulated an ideal environment and added factors that would lead to changes in electron behavior. They then trained an artificial neural network – a kind of artificial intelligence that can learn a specific task using methods that are inspired by how the brain works – to identify the circumstances associated with different theories. When the researchers entered the experimental data into the neural network, they determined which of the theories most closely resembled the actual data.
According to Kim, this method confirmed the hypothesis that the repulsive interaction energy had a stronger influence on the behavior of the electrons.
Understanding how many electrons interact on different materials and under different conditions is likely to lead to more discoveries, including the development of new materials.
"The materials that led to the initial revolution of the transistors were We now have the opportunity to design much more complex materials," said Kim. "If these powerful tools can reveal important aspects that lead to a desired property, we would like to be able to make a material with this property."
Discrete energy levels without restriction – a new quantum trick
Yi Zhang et al. Machine Learning in Experiments to Depict Electronic Quantum Matter, Nature (2019). DOI: 10.1038 / s41586-019-1319-8
Machine Learning Uncovers Secrets of Quantum Physics (2019, June 20)
retrieved on June 20, 2019
This document is subject to copyright. Apart from any fair dealings for the purposes of private study or research, no
Part may be reproduced without written permission. The content is provided for informational purposes only.