MOSP: A User-interface Package for Simulating Metal Nanoparticle Structure and Reactivity under Operando Conditions

Technical Summary of MOSP: A User-interface Package for Simulating Metal Nanoparticle Structure and Reactivity under Operando Conditions

Problem Statement
Metal nanoparticles (NPs) are central to heterogeneous catalysis, where their catalytic reactivity is intrinsically linked to their atomic structure and the assembly of active sites. While traditional kinetic simulations, such as mean-field microkinetic modeling (MKM) and kinetic Monte Carlo (KMC), have successfully predicted catalytic behavior based on fixed model structures, recent in situ experimental observations reveal that NP structures dynamically reconstruct in response to changing reaction conditions. Traditional approaches that rely on static structures fail to account for these reaction-condition-induced shape changes, creating a gap in understanding the dynamic structure-property relationships essential for rational catalyst design.

Methodology
To address this limitation, the authors present MOSP (Multi-scale Operando Simulation Package), an open-source, Python-based graphical user interface (GUI) that integrates two distinct computational models:

1. Multiscale Structure Reconstruction (MSR) Model:

   • Based on Wulff theory, this module predicts the equilibrium shape of metal NPs under specific reaction atmospheres.
   • It calculates the interface tension ($\gamma_{hkl}^{int}$) of each crystal facet by modifying the vacuum surface tension with terms accounting for gas coverage, adsorption energies, and lateral interactions between adsorbates.
   • Adsorbate coverages are determined by solving the Fowler-Guggenheim adsorption isotherm, incorporating partial pressures, temperatures, and entropy changes.
   • The model quantitatively predicts equilibrium structures for particles larger than 3 nm, considering finite-size effects.

2. Kinetic Monte Carlo (KMC) Model:

   • This module simulates reaction dynamics on the specific NP structures generated by the MSR module (or imported existing structures).
   • The system state is defined by a surface configuration of discrete sites, each characterized by element, position, and occupation value.
   • The simulation employs the Direct Method (DM) of the Gillespie algorithm to solve the master equation for time evolution.
   • Events include adsorption, desorption, diffusion, and surface reactions. Rate constants are derived from collision theory (for adsorption/desorption) and transition state theory (for reactions/diffusion), utilizing Brønsted-Evans-Polanyi (BEP) relations and generalized coordination numbers (GCN) to map site-specific energy profiles.

Key Contributions

• Integrated Workflow: MOSP provides a unified framework to couple thermodynamic structure prediction (MSR) with kinetic reaction simulation (KMC), enabling the exploration of dynamic structure-activity relationships.
• User Accessibility: The package features an intuitive GUI built with wxpython, allowing users to input complex parameters (e.g., lateral interaction matrices, scaling relations, reaction networks) without writing code. It includes interactive visualization panels for NP structures (using pyOpenGL) and simulation results (using pandas and matplotlib).
• Modularity: While designed for coupled operation, the MSR and KMC modules can function independently. MSR can export equilibrium structures, and KMC can load pre-existing structures in XYZ format.
• Open Source: The source code is publicly available, facilitating community adoption and extension.

Results and Demonstrations
The paper validates MOSP through two illustrative examples:

1. CO Oxidation over Pt Nanoparticles:

   • The MSR module successfully reconstructed Pt NPs (radius 2.5 nm) under varying pressures (0–2000 Pa) at 800 K.
   • Results showed a pressure-dependent structural evolution: as pressure increased, the fraction of (111) facets decreased while (110) facets increased, shifting the NP shape from a truncated octahedron to a quasi-rhombic dodecahedron.
   • Subsequent KMC simulations (10 million steps) on the 500 Pa structure revealed that edges and corners were the primary contributors to activity, with calculated coverages of CO (0.01) and O (0.11) and a turnover frequency (TOF) of $8.33 \times 10^5$ site$^{-1}$s$^{-1}$.
   • Performance benchmarks indicated that MSR execution time scales with the number of atoms ($N_{atoms}^{2.4}$), while KMC scales nearly linearly ($N_{atoms}^{0.93}$).

2. Water-Gas Shift (WGS) Reaction over Cu Nanoparticles:

   • A KMC simulation was performed on a 2190-atom truncated octahedron Cu NP at 650 K and 4000 Pa.
   • The model included five adsorbates and fourteen events following the carboxyl mechanism.
   • Results showed CO as the dominant adsorbate (coverage 0.67).
   • Crucially, the simulation highlighted kinetic coupling at the atomic level: low-coordination sites were identified as favorable for CO$_2$ formation, while high-coordination sites favored H$_2$ formation, despite similar overall TOFs for both products.

Significance
The paper positions MOSP as an efficient tool for researchers to predict the structure and catalytic properties of metal NPs under operando conditions. By bridging the gap between static structural assumptions and dynamic experimental realities, MOSP aids in the screening of optimal reaction conditions and the rational design of catalysts. The authors emphasize that the package is designed to run on standard laptops, making advanced operando simulation accessible without reliance on high-performance computing clusters for initial screening.