The computational modeling for liquid/metal interfaces, such as heterogeneous catalysis, corrosion, electrochemistry, lubrification, and biomedical applications, it is particularly challenging because of configurationally disordered nature of the aqueous molecules [1, 2]. Moreover, methods for calculating the free energies at the liquid/metal interfaces are not well-established. For instance, solvent molecules affect the energies of adsorbed catalytic intermediates species and participate in catalytic reactions. Influencing, thus, significantly catalytic performance [3]. In order to get such molecular insight into the liquid/metal interfaces, different computational methods such as Density Functional Theory (DFT), classical molecular dynamics (cMD), and/or
ab-initio molecular dynamics (AIMD) can be used. DFT is needed to describe bond breaking and forming, but the aqueous phase molecules (configurationally disordered) require multiple “snapshots†to account for the plausible effects. However, this can quickly become too computationally demanding to do in DFT. Thus, our group established a multiscale approach for calculating the free energies of heterogeneous catalytic system, specifically, by employing an explicit solvation method using a multiscale sampling (MSS) approach. This MSS approach combines DFT with cMD, which in general consists of three steps: 1) DFT is used to optimize the reaction intermediate over the catalyst under vacuum, 2) Water molecules are added and arranged into liquid structures using cMD while the reaction intermediate is held fixed, and 3) DFT is used to calculate the electronic energy for the whole system (intermediate, catalyst, and solvent). While this strategy has been shown to accurately capture the free energies of solvation of fixed intermediates on catalyst surfaces [3], it leaves questions unanswered. Specifically, since the intermediate is held fixed, this strategy ignores conformational contributions to the free energy caused by the thermal fluctuations in the internal geometry of the adsorbate. Our goal in this work is to learn how extensively such movements influence the free energies of catalytic species. To do this, we must develop a force field (FF) that can capture the chemical bond that is formed between the catalytic species and the catalyst surface. In this work, we derive a FF that can be used to model CH3OH adsorption at a Pt(111) catalyst surface. We named the FF GDL19, as it is inspired by the GAL17 FF [4], which is a DFT-based FF which describes the H2O-Pt binding interaction by means of gaussianic anisotropic and angular correction potentials. In GDL19, the CH3OH-Pt intermolecular interactions (bond, angles, and dihedrals) are described by gaussianic anisotropic potentials, while the internal CH3OH geometry, as well as the nonbonded interactions, are described by an existing FF, specifically, the optimized potentials for liquid simulations (OPLS-AA) FF [5]. The philosophy behind GDL19 is to keep it as “minimalistic†as possible in its functional form. We use this force field to calculate the free energy of solvation of CH3OH* and compare it with the analogous value when the adsorbate is held fixed, enabling interrogation of the extent to which internal conformational changes influence the free energies of solvation of catalytic intermediates. We further use the force field to calculate the rate constant for adsorption of CH3OH on a catalyst surface from liquid H2O and compare with existing models for calculating adsorption rate constants. The current models either assume that the CH3OH molecule is an ideal gas or that
CH3OH adsorbs physically to the catalyst surface; thus, the GDL19 FF will enable interrogation of the extent to which these assumptions influence the adsorption rate constant, as well.
To map the CH3OH-Pt interaction energy, we constructed a set of 400 frames for the CH3OH-Pt system, by dragging the CH3OH over the Pt surface. Partial optimizations of the CH3OH were performed using the VASP code [6], in these calculations, we held the OCH3OH-Pt distance fixed and allowed the CH3OH internal geometry to relax. We used these in the parameterization of GDL19. Specifically, we are using a simulated annealing routine implemented in Python3 to develop the GDL19 parameters. The FF is tested using cMD simulations perform in LAMMPS [7]. Moreover, ab-initio Molecular Dynamics (AIMD) simulations were carried out to qualitatively assess the trajectory produced in cMD.
References
1. Besson, M., Gallezot, P., and Pinel, C., Chemical Reviews, 114, 1827 (2014)
2. M. Saleheen and A. Heyden, ACS Catalysis, 8, 2188 (2018).
3. Zhang X., DeFever R., Sarupria S., and Getman R., J. Chem. Inf. Model., Article ASAP (2019)
4. Steinmann, S., Ferreira De Morais, R., Götz, A., Fleurat-Lessard, P., Iannuzzi, M., Sautet, P., and Michel, C., JCTC, 14, 3238 (2018).
5. Jorgensen, W., Maxwell, D., and Tirado-Rives, J., JACS, 118, 11225 (1996).
6. J. Hafner, J. Comput. Chem., 29, 2044 (2008).
7. S. Plimpton, J. Comput. Phys., 117, 1 (1995).
8. W. L. Jorgensen, J. Chandrasekhar, J. D. Madura, R. W. Impey, and M. L. Klein, J. Chem. Phys., 79, 926 (1983)
