Machine Learning in Condensed Matter and Materials Physics

Overview

Explore machine learning applications in condensed matter and materials physics through this 56-minute talk from the Alan Turing Institute. Delve into how ML methods enhance simulations of matter's building blocks, from electronic to molecular structures, boosting computational techniques like density functional theory and molecular dynamics. Discover the potential for new physical insights and exotic material engineering. Learn about atomic-scale modeling, predictive modeling challenges, and ML approaches for properties beyond potentials. Examine symmetry-adapted ML for tensors, electron charge density predictions, and ML techniques incorporating molecular orbital theory. Investigate the combination of structural and functional properties in ML models, enabling first-principles accuracy at realistic time and size scales. Trace the historical development of related concepts and explore how machine learning can uncover quantum emergence from quantum matter data.

Syllabus

Intro
Atomic-scale modeling of real materials ⚫ Foundations for the predictive modeling of chemicals and materials Key challenge: accurate electronic properties + sampling of fluctuations/defec
Predicting properties beyond potentials • Symmetry-adapted ML for tensors: CCSD-quality molecular polarizabilities & d • Electron charge density for molecules (and condensed phases!) • Single-particle Hamiltonians: ML that knows molecular orbital theory!
Structural and functional properties, combined • Predicting any property accessible to quantum calculations • Realistic time and size scales, with first-principles accuracy and mapping of stru functional properties
TUNNELING DENSITY OF STATES IN 1962
X-ray diffraction in 1913
Projective Measurements in 1922
DETERMINED BY WEIGHTS AND BIAS
Hypothesis test
Learn the sorting criteria for emerger
Discoveries
Machine Learning Quantum Emergence From Quantum Matter Data