Methane dissociation on Ni(111): A fifteen-dimensional potential energy surface using neural network method

Area of Science:

• Computational Chemistry
• Surface Science
• Chemical Physics

Background:

• Accurate potential energy surfaces (PES) are crucial for understanding gas-surface interactions.
• Methane (CH4) adsorption and reaction on metal surfaces like Nickel (Ni) are fundamental in catalysis and chemical processes.
• Previous PES models may lack the accuracy needed for detailed dynamical simulations.

Purpose of the Study:

• To develop a highly accurate, 15-dimensional PES for CH4 interacting with a rigid Ni(111) surface.
• To validate the PES accuracy through various tests, including vibrational spectra and transition state analysis.
• To investigate the sticking probability of CH4 on the Ni(111) surface.

Main Methods:

• Utilized neural network (NN) fitting to a large database of ab initio density functional theory (DFT) energy points (approx. 194,208 points).
• Employed a specific NN architecture (25-60-60-1) to achieve high accuracy.
• Applied sudden approximations, following Jackson's group methodology, to calculate sticking probabilities.

Main Results:

• Achieved a highly accurate PES with low root mean square errors (10.11 meV for entrance, 17.00 meV for interaction/product regions).
• The developed PES accurately reproduces DFT results, especially for critical transition state geometries.
• Calculated sticking probabilities (S0) for ground state CH4 at experimental surface temperatures, noting an underestimation.

Conclusions:

• The NN-based PES provides a reliable representation of the CH4-Ni(111) interaction.
• The model's accuracy facilitates further studies in surface reaction dynamics and catalysis.
• Further investigation is needed to explain the observed underestimation in sticking probability.