Machine learning corrections for DFTB

Due its computational efficiency, DFTB has positioned itself as a reliable quantum mechanical method for condensed phase applications, especially where an extensive sampling of the configurational space is important to the reactive process of interest such as chemical reactions in biological systems. However as a "low-method" it can neglect certain energy contributions, or show inaccurate transition states, as seen in thiol-disulfide exchange, a system of interest in project P4. To learn and correct these DFTB miscalculations is the motivation of our project. To achieve this, a Behler-Parrinello-type Neural Network with the Δ-machine learning implementation learns the energy value differences between the ab initio quantum chemical potential and DFTB for a given molecular structure.