Computational chemistry has two major goals: (i) making reliable first-principle predictions of the properties of matter including single molecules, molecular complexes, crystals, liquids and amorphous materials; (ii) providing insight, both qualitatitively via future-proof definitions of concepts, and quantitatively via fragment-based numerical explanations (up to atomic resolution). For the latter goal we propose Quantum Chemical Topology, which gathers QTAIM, ELF, IQA and several more research areas under the central idea of using the language of dynamical systems (attractor, critical point, basin, gradient path, separatrix,…).
For the first goal we propose to extend system size and time scale and develop FFLUX, a machine-learning force field based on quantum topological atoms. This choice guarantees that QCT’s chemical insight is preserved in the production phase of FFLUX. As such, a numerical narrative accompanies the predictions made by FFLUX. Moreover, goal (i) and (ii) will then be reached in a fully consistent way. An in-house method called Relative Energy Gradient (REG) helps this process as it is able, with minimal assumptions, to point out which fragments are responsible for the bevaiour of the total system.
Other than REG and FFLUX the group worked on Quantum Topological Molecular Similarity (QTMS), a Quantum Isostere Database (QID), and the Ab Initio Bond length (AIBL) pKa predictor. AIBL has the capacity to predict, after calibration, the pKa of a functional group of a molecule in aqueous solution, only from a relevant bond length in the gas phase. Spectacularly, in one case study, this method made a very accurate prediction before the pKa was measured.