Partial Charges Optimized by Genetic Algorithms for Deep Eutectic Solvent Simulations

Abstract

Deep eutectic solvents (DESs) are a class of solvents often composed of ammonium-based chloride salts and a neutral hydrogen bond donor (HBD) at specific ratios. These cost-effective and environmentally friendly solvents have seen significant growth in multiple fields, including organic synthesis, and in materials and extractions because of their desirable properties. In the present work, a new software called genetic algorithm machine learning (GAML) was developed that utilizes a genetic algorithm (GA) approach to facilitate the development of optimized potentials for liquid simulation (OPLS)-based force field (FF) parameters for eight unique DESs based on three ammonium-based salts and five HBDs at multiple salt:HBD ratios. As an initial test of GAML, partial charges were created for 86 conventional solvents based on neutral organic molecules that yielded excellent overall mean absolute deviations (MADs) of 0.021 g/cm 3 , 0.63 kcal/mol, and 0.20 kcal/mol compared to experimental densities, heats of vaporization (ΔH vap ), and free energies of hydration (ΔG hyd ), respectively. FFs for DESs constructed from ethylammonium, N,N-diethylethanolammonium, and N-ethyl-N,Ndimethylethanolammonium chloride salts were then parameterized using GAML with exceptional agreement achieved at multiple temperatures for experimental densities, surface tensions, and viscosities with MADs of 0.024 g/cm 3 , 4.2 mN/m, and 5.3 cP, respectively.

Comparison and Validation of Force Fields for Deep Eutectic Solvents in Combination with Water and Alcohol Dehydrogenase

“…Very recently the authors of the OPLS-DES force field have parametrized further DES components, whereby a different parameterization strategy was used. 62 The TIP3P force field 63 was used to represent water in all MD simulations that are performed within this work. In case of simulations using CGenFF/CHARMM36m, the CHARMM-TIP3P variant 64,65 has been used, which is a slight modification of the original TIP3P model.…”

“…The forces from the LJ interactions were thereby smoothly switched to zero between 0.9 and 1.1 nm. Very recently the authors of the OPLS-DES force field have parametrized further DES components, whereby a different parameterization strategy was used …”

Simulation of deep eutectic solvents: Progress to promises

“…110 For example, Zhong, Velez, and Acevedo reported OPLS-AA parameters for DESs constructed from ethylammonium, N,N-diethylethanolammonium, and N-ethyl-N,N-dimethylethanolammonium chloride salts by developing a genetic algorithm approach that automates the creation of partial charges fit to experimental physicochemical properties, for example, surface tension and viscosity. 111,112 Another avenue for improving agreement with experiment is the tuning of nonbonded parameters to treat polarization implicitly. For example, when Ferreira et al tested various FF combinations for CCEtg using an unscaled charge scheme (±1e), the self-diffusion coefficients of Ch + and ethylene glycol were found to be underestimated by a factor of 8 compared to experiment; whereas, they observed large accuracy gains when scaling the atomic partial charges by 0.8.…”

“…Machine learning methods may offer new opportunities to improve the accuracy of FFs, as recently shown for partial charges 110 . For example, Zhong, Velez, and Acevedo reported OPLS‐AA parameters for DESs constructed from ethylammonium, N , N ‐diethylethanolammonium, and N ‐ethyl‐ N , N ‐dimethylethanolammonium chloride salts by developing a genetic algorithm approach that automates the creation of partial charges fit to experimental physicochemical properties, for example, surface tension and viscosity 111,112 …”

Comparison between Ab Initio Molecular Dynamics and OPLS-Based Force Fields for Ionic Liquid Solvent Organization

“…The high computational cost of AIMD often limits simulations to short timescales of ∼100 ps and a reduced simulation box size composed of 32 ion pairs that prohibits description of the medium–long structural order present in ILs. , This speaks to the present need of highly accurate atomic-level IL force fields (FFs). − Our group published a nonpolarizable OPLS-based IL FF in 2009 (i.e., OPLS-2009IL), which was reevaluated in 2017 using a new charge scaling of ±0.8 e to mimic polarization and charge transfer effects . The scaled parameters yielded excellent agreement with experimental densities, heats of vaporization (Δ H vap ), viscosities, diffusion coefficients, heat capacities, surface tensions, and other relevant solvent data.…”