Towards accurate prediction of catalytic activity in IrO₂ nanoclusters via first principles-based variable charge force field

Abstract

IrO 2 is one of the most efficient electrocatalysts for the oxygen evolution reaction (OER), and also has other applications such as in pH sensors. Atomistic modeling of IrO 2 is critical for understanding the structure, chemistry, and nanoscale dynamics of IrO 2 in these applications. Such modeling has remained elusive due to the lack of an empirical force field (EFF) for IrO 2 . We introduce a first-principles-based EFF that couples the Morse (MS) potential with a variable charge equilibration method, QEq. The EFF parameters are optimized using a genetic algorithm (GA) on a density functional theory (DFT)-based training set. The resultant Morse plus QEq EFF, "MS-Q" in short, successfully reproduces the lattice constants, elastic constants, binding energies, and internal coordinates of various polymorphs of IrO 2 from DFT calculations. More importantly, it accurately captures key metrics for evaluating structural and chemical properties of catalysts such as surface energetics, equilibrium shape, electrostatic charges, oxygen vacancy formation energies, relative stability of low index rutile IrO 2 surfaces, and pressureinduced phase transformations. The MS-Q EFF is used to predict the oxygen binding energy (E ad ), a wellknown descriptor for OER activity, on various sites of a nanocatalyst. We find E ad to be more favorable at low coordination sites, i.e. edges and corners, compared to planar facets; E ad is also correlated with charge transfer between the adsorbed O and nanocrystal, highlighting the importance of variable charge electrostatics in modeling catalysis on metal oxide surfaces. Our variable charge force field offers encouraging prospects for carrying out large-scale reactive simulations to evaluate catalytic performance of IrO 2 surfaces and nanostructures. 4 interactions. 47-50 For example, a Morse+QEq (MS-Q) force field for titanium oxide, developed by Swamy and Gale, accurately predicted the bulk moduli and binding energies of various titanium oxide polymorphs, when compared to quantum mechanical calculations, but could not predict the relative surface stability in rutile TiO 2 . 48, 51 There are several possibilities for this shortcoming: the lack of surface energies in the training set, 51 the lack of parametrizing of QEq parameters, and/or the lack of adequate global parameter space search. We hypothesize that it is possible to successfully use MS-Q to describe the IrO 2 system by addressing these missing elements.Our approach for developing a MS-Q force-field incorporates several key elements: the inclusion of diverse local environments in an elaborate DFT training set that includes surfaces, the parameterization of QEq parameters, and the use of an evolutionary algorithm to parameterize MS-Q. The use of different local environments including broken bonds in the training set enhances the probability that chemical reactions can be modeled with the resultant force field. The QEq parameters, namely, electronegativity and chemical hardness, may depend on chemical environments and therefore should ...

Periodic DFT Study of Rutile IrO2: Surface Reactivity and Catechol Adsorption

“…Wulff structure is in perfect agreement with previous works 22,24 . This is explained by the fact that PBEsol overstabilizes the bulk compared to the surfaces rendering then a higher value for the surface energy.…”

“…A combination of surface science and Density Functional Theory (DFT) has been used to support the stability of the oxygen rich (100) iridia termination 23 . Sen et al 24 have used DFT to parametrize a force field for IrO 2 nanoparticles. Novell-Leruth et al 22 have studied the mixture of different rutile oxides and give a brief description of the structural and energetic properties for IrO 2 surfaces.…”

Unraveling the Planar-Globular Transition in Gold Nanoclusters through Evolutionary Search

“…However, finding the global minimum structure for a given composition requires global minimization techniques such as simulated annealing23, genetic algorithm (GA)24, basin21 and minima hopping25, and particle swarm26 methods, coupled with a reliable local optimizer. These stochastic global optimization methods have increasingly been utilized in materials discovery and development272829303132.…”

Machine Learning Force Field Parameters from Ab Initio Data

“…Here, to overcome the challenge encountered with force field parameterization using ML techniques, we have used the genetic algorithm (GA), which is an evolutionary algorithm that mimics the process of natural selection. 37 Previous successful applications of the genetic algorithm in similar contexts include the determination of the parameters for the ReaxFF reactive force field, 2–3, 5 and various force fields including Morse+QEq charge transfer ionic potential (CTIP), 6 and a new hybrid bond-order potential (HyBOP) for materials system. 2–3, 5–6, 38–39 …”

“…The vdW parameters in the force field were optimized using the interaction energy from 1,250 methanol clusters by employing the developed ML/GA framework. 6–7, 38 …”