Assessing the persistence of chalcogen bonds in solution with neural network potentials

Abstract

Non-covalent bonding patterns are commonly harvested as a design principle in the field of catalysis, supramolecular chemistry and functional materials to name a few. Yet, their computational description generally neglects finite temperature and environment effects, which promote competing interactions and alter their static gas-phase properties. Recently, neural network potentials (NNPs) trained on Density Functional Theory (DFT) data have become increasingly popular to simulate molecular phenomena in condensed phase with an accuracy comparable to ab initio methods. To date, most applications have centered on solid-state materials or fairly simple molecules made of a limited number of elements. Herein, we focus on the persistence and strength of chalcogen bonds involving a benzotelluradiazole in condensed phase. While the tellurium-containing heteroaromatic molecules are known to exhibit pronounced interactions with anions and lone pairs of different atoms, the relevance of competing intermolecular interactions, notably with the solvent, is complicated to monitor experimentally but also challenging to model at an accurate electronic structure level. Here, we train direct and baselined NNPs to reproduce hybrid DFT energies and forces in order to identify what are the most prevalent non-covalent interactions occurring in a solute-Cl$^-$-THF mixture. The simulations in explicit solvent highlight competition with chalcogen bonds formed with the solvent and the short-range directionality of the interaction with direct consequences for the molecular properties in the solution. The comparison with other potentials (e.g., AMOEBA, direct NNP and continuum solvent model) also demonstrates that baselined NNPs offer a reliable picture of the non-covalent interaction interplay occurring in solution.

How Robust Is the Reversible Steric Shielding Strategy for Photoswitchable Organocatalysts?

“…While obtaining both highly accurate energetics and converged statistical sampling has typically hampered the appropriate description of the flexibility of organic molecules, efforts in our group are being made to achieve ab initio accuracy by correcting semiempirical potentials with machine learning models, including implicit 66 or explicit treatment of solvation. 65 , 95 …”

“…While obtaining both highly accurate energetics and converged statistical sampling has typically hampered the appropriate description of the flexibility of organic molecules, efforts in our group are being made to achieve ab initio accuracy by correcting semiempirical potentials with machine learning models, including implicit 66 or explicit treatment of solvation. 65,95 ■ CONCLUSIONS In this work, we have illustrated the importance of thoroughly exploring of the free energy landscape of flexible photoswitchable organocatalysts to understand their catalytic performance. Without any a priori knowledge of the system, the complex FES of three N-alkylated azobenzene-tethered piperidine photoswitches were successfully mapped, the energetic basins corresponding to the structural minima identified, and the experimentally observed reactivity trends rationalized according to the catalysts' conformational behavior.…”

“… Overall, the relative population of catalytically active or inactive states of PSa–c , corroborated by comparison to the experimentally observed reactivity trends, is not expected to significantly change. While obtaining both highly accurate energetics and converged statistical sampling has typically hampered the appropriate description of the flexibility of organic molecules, efforts in our group are being made to achieve ab initio accuracy by correcting semiempirical potentials with machine learning models, including implicit or explicit treatment of solvation. , …”

Learning Conductance: Gaussian Process Regression for Molecular Electronics

“…Recently, machine learning (ML) approaches have become an alternative to traditional quantum chemical calculations, which can bring down computational cost by an order of magnitude or more. − Examples involve the generation of force fields for MD simulations, − prediction of charge transfer integrals, , or even trying to directly solve the Schrödinger equation . While artificial neural networks are attractive in situations where large data sets are available for training [On the flipside, this implies that artifical neural networks typically also require more data points for training.…”

Learning Conductance: Gaussian Process Regression for Molecular Electronics

“…Recently, machine learning (ML) approaches have become an alternative to traditional quantum chemical calculations, which can bring down computational cost by an order of magnitude or more [60][61][62][63][64][65][66][67][68][69] . Examples involve the generation of force fields for MD simulations [70][71][72][73][74][75] , prediction of charge transfer integrals 76,77 or even trying to directly solve the Schrödinger equation 78 . While neural networks are attractive in situations, where large datasets are available for training, other methods such as Gaussian process regression (GPR) or kernel ridge regression (KRR) can cope with smaller data sets 61 and have been applied successfully to predict, e.g., interatomic potentials 79 .…”