A Data-driven Approach to Correlated Quantum Many-Body Problems
Abstract
This work aims to exploit developments in machine learning to optimally fit wave functions for many-body problems to data sets of particle configurations, circumventing the usual exponential complexity of the wave function, and limitations of traditional approaches. This defines a wave function free from explicit adjustable parameters to optimize. Instead, the wave function at every point is defined by a statistically optimal interpolation of all data points, which leaves the resultant wave function manifestly size extensive, leading to well-defined properties and energetics in the thermodynamic limit. The flexibility of this description then stems from the amount and quality of the data that it accesses, towards a rigorously exact limit. As well as establishing this approach for both lattice models and first principles molecular and materials modelling, we will also consider its extensions for finite temperature, response, and its applications to outstanding problems in experimental reconstruction of a quantum state.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Aug 28, 2018
- Source ID
- FA95501810515
Entities
People
- George H Booth
Organizations
- Air Force Office of Scientific Research
- United States Air Force