Accurate and Compatible Force Fields for Molecular Oxygen, Nitrogen, and Hydrogen to Simulate Gases, Electrolytes, and Heterogeneous Interfaces

Abstract

Gas molecules and interfaces with liquids and solids play a critical role in living organisms, sorption, catalysis, and the environment. Monitoring adsorption and heterogeneous interfaces remains difficult in experiments, and earlier models for molecular simulations lead to errors over 100% in fundamental molecular properties. We introduce conceptually new force field parameters for molecular oxygen, nitrogen, and hydrogen that reduce deviations to <5%. We employ a combination of a harmonic bond stretching potential and Lennard-Jones parameters with 12-6 and 9-6 options, leading to computed bond lengths, Raman peaks, liquid densities, vaporization enthalpies, and free energies of hydration in impressive agreement with experiments. Reliable free energies of hydration were obtained upon validation of density and vaporization energy without significant further parameter adjustments. We illustrate applications to O2 adsorption on Pt electrocatalysts and N2 adsorption in zeolites, showing <5% deviation in adsorption energies measured in experiments without additional fitting parameters. We discuss the chemical interpretation of all parameters and explain the reasons for discrepancies in earlier models. Compatibility with the Interface Force Field (IFF), CHARMM, AMBER, OPLS-AA, GROMOS, DREIDING, CVFF, PCFF, COMPASS, and QM/MM methods enables reliable simulations of gases and liquid/solid interfaces with biopolymers, minerals, and metals. The parametrization protocol can be applied to similar molecules.

Adsorption and Diffusion of Oxygen on Pure and Partially Oxidized Metal Surfaces in Ultrahigh Resolution

“…In this Letter, we derive missing understanding in atomic detail. We analyzed the dynamics of O 2 molecules on 11 clean and partially oxidized Al and Ni (100), (110), and (111) surfaces using over 150 000 molecular mechanics (MM) and molecular dynamics (MD) simulations with the INTERFACE force field (IFF). ,− IFF matches lattice parameters of metals within 0.1%, includes recent order-of-magnitude improved parameters for gases, and reproduces surface energies of metals as well as gas-metal binding energies with only about 5% error relative to experiments. ,, Therefore, molecular simulations with IFF fill a gap in current methods by covering functionality unavailable in other force fields and by DFT methods, as well as high computational efficiency (see Section S1.1 in the Supporting Information for details).…”

“…We chose model surfaces that include six clean low-index surfaces (111), (100), and (110) of Ni and Al, as well as five representative partially oxidized surfaces (Ni (100)-O-p2 × 2, Ni (100)-O-c2 × 2, Ni (110)-O-c2 × 1, Ni (111)-O-p2 × 2, and Al (111)-O-p1 × 1), which were derived from experimental X-ray and LEED data (Figure , see details and references in Section S1.2 in the Supporting Information). Geometry optimizations and molecular dynamics simulations employed IFF parameters for metal surfaces and O 2 , as well as specifically developed parameters for oxides and partially oxidized surfaces (Section S1.3, Tables S1 to S3, and Figure S1 in the Supporting Information). These additions reproduce oxygen–metal bond distances and oxide geometries within 1%, as well as the interlayer spacing and layer buckling of the partially oxidized surfaces in agreement with LEED data (Section S1.4 and Figure S1 in the Supporting Information).…”

“…We analyzed the dynamics of O 2 molecules on 11 clean and partially oxidized Al and Ni (100), (110), and (111) surfaces using over 150 000 molecular mechanics (MM) and molecular dynamics (MD) simulations with the INTERFACE force field (IFF). 3,52−54 IFF matches lattice parameters of metals within 0.1%, 55 includes recent order-of-magnitude improved parameters for gases, 53 and reproduces surface energies of metals as well as gas-metal binding energies with only about 5% error relative to experiments. 3,53,55 Therefore, molecular simulations with IFF fill a gap in current methods by covering functionality unavailable in other force fields and by DFT methods, as well as high computational efficiency (see Section S1.1 in the Supporting Information for details).…”

Patterning of Self-Assembled Monolayers of Amphiphilic Multisegment Ligands on Nanoparticles and Design Parameters for Protein Interactions

“…However, we note remaining inconsistencies in CGenFF even after modifications. For example, the 12–6 Lennard-Jones well depth of hydrogen atoms of 0.045 kcal/mol is much higher than validated values of 0.015 kcal/mol in IFF, and even larger than the well depth of 0.032 kcal/mol of carbon atoms in CGenFF, which are clearly more polarizable than hydrogen atoms (Table ).…”

Quantum-based modeling of protein-ligand interaction: The complex of RutA with uracil and molecular oxygen

“…[17] and those for the molecular oxygen taken from Ref. [18] , while water molecules were treated as TIP3P. Classical MD trajectories were simulated with the NAMD 3.0 software package.…”