In this work we explore the potential of a new data-driven approach to the design of exchange-correlation (XC) functionals. The approach, inspired by "surrogate" functions from optimization and convolutional filters in computer vision, utilizes convolutions of the electron density in order to form a feature space to represent local electronic environments. These features are orbital free, and provide a systematic route to including information at various length scales. Dimensional reduction and regression models are utilized to quantitatively connect these orbital-free descriptors to the exchange-correlation energies and potentials extracted from hybrid functionals. In this talk the approach will be introduced and demonstrated for the construction of a GGA-B3LYP surrogate functional that is tested for small hydrocarbon molecules. The use of uncertainty quantification and advanced visualization tools will provide insight into the advantages and challenges of surrogate functionals as a new route for designing exchange-correlation functionals and embedding higher level physics into density functional theory.

Development and analysis of "surrogate" hybrid functionals using neural networks, uncertainty quantification, and visualization

255th National Meeting (2018)

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.