Search

Data-driven synthesis planning has seen remarkable successes in recent years by virtue of modern approaches of artificial intelligence that efficiently exploit vast databases with experimental data on chemical reactions. However, this success story is intimately connected to the availability of existing experimental data. It may well occur in retrosynthetic and synthesis design tasks that predictions in individual steps of a reaction cascade are affected by large uncertainties. In such cases, it will, in general, not be easily possible to provide missing data from autonomously conducted experiments on demand. However, first-principles calculations can, in principle, provide missing data to enhance the confidence of an individual prediction or for model retraining. Here, we demonstrate the feasibility of such an ansatz and examine resource requirements for conducting autonomous first-principles calculations on demand., Competing Interests: There are no conflicts to declare., (This journal is © The Royal Society of Chemistry.)

Density functional theory (DFT) is used to calculate the relative free energies of deprotonation of the isomers of iron-group hydride complexes MHXL 4 where M = Fe, Ru, Os, L 4 = (CO) 4 or (PMe 2 CH 2 CH 2 PMe 2 ) 2 for a wide range of anionic ligands X. The free energies of the most stable isomers are used to calculate relative pK a values where K a refers to the acid dissociation constant for the equilibrium MHXL 4 → [MXL 4 ] - + H + . These are used to test the proposal that the pK a for a given metal complex in THF can be simply calculated by adding the contributions to the total pK a value from each ligand L; these are called ligand acidity constants (LAC) A L used in the LAC equation [R. H. Morris, J. Am. Chem. Soc., 2014, 136, 1948-1959]. The A L are calculated using A L = 0.2 as a reference for the hydride ligand. The A L of certain less polarizable X ligands are found to be fairly constant (±1 to ±2 units) and consistent with the proposed LAC method for a range of the complexes considered: 2 for Me - , 1 for OH - , 1 for NH 2 - , 0 for B(OCH 2 CH 2 O - ) - , -1 for OMe - , -2 for SH - and -2 for SMe - . Other ligands have more variable A L values (±3 to ±5 units) because of high polarizability or other reasons: 1 for OtBu - , -1 for F - , -2 for BMe 2 - , -2 for NMe 2 - , -3 for Cl - , -3 for PMe 2 - , -3 for Br - , -5 for I - , -6 for CN - and -12 for SiCl 3 - . Iodide stabilizes the anion [MIL 4 ] - more than does fluoride in [MFL 4 ] - making iodide the more acidifying ligand despite its lower electronegativity. DFT is also used to validate the charge correction in the LAC equation.

Recent reports on the necessity of using externally orthogonal orbitals in subsystem density-functional theory (SDFT) [Annu. Rep. Comput. Chem., 8, 2012, 53; J. Phys. Chem. A, 118, 2014, 9182] are re-investigated. We show that in the basis-set limit, supermolecular Kohn-Sham-DFT (KS-DFT) densities can exactly be represented as a sum of subsystem densities, even if the subsystem orbitals are not externally orthogonal. This is illustrated using both an analytical example and in basis-set free numerical calculations for an atomic test case. We further show that even with finite basis sets, SDFT calculations using accurate reconstructed potentials can closely approach the supermolecular KS-DFT density, and that the deviations between SDFT and KS-DFT decrease as the basis-set limit is approached. Our results demonstrate that formally, there is no need to enforce external orthogonality in SDFT, even though this might be a useful strategy when developing projection-based DFT embedding schemes.