Skip to main content
. 2021 Mar 11;121(16):10142–10186. doi: 10.1021/acs.chemrev.0c01111

Figure 1.

Figure 1

Accurate ab initio methods are computationally demanding and can only be used to study small systems in gas phase or regular periodic materials. Larger molecules in solution, such as proteins, are typically modeled by force fields, empirical functions that trade accuracy for computational efficiency. Machine learning methods are closing this gap and allow to study increasingly large chemical systems at ab initio accuracy with force field efficiency.