Accurate Probabilities for Highly Activated Reaction of Polyatomic Molecules on Surfaces Using a High-Dimensional Neural Network Potential: CHD₃ + Cu(111)

Abstract

An accurate description of reactive scattering of molecules on metal surfaces often requires the modeling of energy transfer between the molecule and the surface phonons. Although ab initio molecular dynamics (AIMD) can describe this energy transfer, AIMD is at present untractable for reactions with reaction probabilities smaller than 1%. Here, we show that it is possible to use a neural network potential to describe a polyatomic molecule reacting on a mobile metal surface with considerably reduced computational effort compared to AIMD. The highly activated reaction of CHD3 on Cu(111) is used as a test case for this method. It is observed that the reaction probability is influenced considerably by dynamical effects such as the bobsled effect and surface recoil. A special dynamical effect for CHD3 + Cu(111) is that a higher vibrational efficacy is obtained for two quanta in the CH stretch mode than for a single quantum.

Dynamical Study of the Dissociative Chemisorption of CHD3 on Pd(111)

“…Another important dynamical aspect of the reaction of methane is the energy transfer from the molecule to the surface. 54 Figure 6 compares for scattered trajectories this energy transfer from CHD 3 to Cu(111), 54 Pt(111), 65 Ni(111), 12 and Pd(111), where the energy transfer E T is defined aswhere V and K are the potential free and kinetic energy of methane at the initial (i) and final (f) steps of the scattered trajectories, respectively. In general, it is observed that the lower the surface atom mass is, the higher the energy transfer is from methane to the surface.…”

“…Energy transfer from methane to Ni(111) (blue squares), Pd(111) (black circles), Pt(111) 65 (red triangles), and Cu(111) 54 (green diamonds) compared to the refined Baule model. The solid lines without symbols indicate results predicted by the refined Baule model, whereas the dashed and dotted lines with solid and open symbols indicate laser-off and ν 1 = 1 results, respectively.…”

“…However, to gather further support for this idea, better statistics are needed, which could perhaps be obtained on the basis of QCT dynamics on a PES also incorporating the effect of surface atom motion. 54…”

Designing new SRP density functionals including non-local vdW-DF2 correlation for H2 + Cu(111) and their transferability to H2 + Ag(111), Au(111) and Pt(111)

“…We still discuss the valuable experimental results on associative desorption since they do provide insight into the reaction dynamics. However, as we will discuss, it is fraught with difficulty to make a direct quantitative comparison between calculations on dissociative chemisorption and associative desorption experiments without improving our dynamical model by incorporating phonons and ehp excitations, which is challenging to do [148][149][150] . Recently, this has been done for H 2 + Cu(111) 75 using ab initio molecular dynamics with electronic friction (AIMDEF) calculations employing the PBE 115 DF.…”

DScribe: Library of descriptors for machine learning in materials science

“…In practice, multiple symmetry functions of each type are used in the descriptor, each with a different parametrization (ζ, λ, η, R s and κ), encoding different portions of the chemical envi-ronment. While no optimal parameter set exists, the choice of parameters is guided by the generally desired properties of a descriptor previously listed (unique, continuous and compact), and various alternatives for different applications can be found in the literature [51][52][53].…”