** Reference to paper:

https://doi.org/10.1021/acs.jpclett.1c03395

** DOI:

10.1021/acs.jpclett.1c03395

** Title:

Accurate Simulations of the Reaction of H2 on a Curved Pt Crystal Through Machine Learning

** Authors:

Gerrits, Nick

** Contact e-mail:

n.gerrits@lic.leidenuniv.nl

** Abstract:

Theoretical studies on molecule-metal surface reactions have so far been limited to small surface unit cells due to computational costs.
Here, for the first time molecular dynamics simulations on very large surface unit cells at the level of DFT are performed, allowing a direct comparison to experiments performed on a curved crystal.
Specifically, the reaction of D2 on a curved Pt crystal is investigated with a neural network potential (NNP).
The developed NNP is also accurate for considerably larger surface unit cells than those that have been included in the training data, allowing dynamical simulations on very large surface unit cells that otherwise would have been intractable.
Important and complex aspects of the reaction mechanism are discovered such as diffusion and a shadow effect of the step.
Furthermore, conclusions from simulations on smaller surface unit cells are not always transferable to larger surface unit cells, limiting the applicability of theoretical studies on smaller surface unit cells to heterogeneous catalysts with small defect densities.

** Description per file:
CRPResults.tar.gz - Contains files and scripts to obtain sticking probabilities with particular molecular beam parameters from MD results obtained with the CRP PESs.

** Folder ASECalculator:
Contains files (except RuNNer itself, could also be interchanged with n2p2) to use ASE and compute minimum barrier heights (or anything else one could do with ASE).

** Folder Figure01:
plotsticking.py - Generates the figure; requires one to compute the energies of the entire DFT dataset with the HDNNP to generate the first subpanel.
reactionprobability.pdf

** Folder Figure02:
plotsticking_crosssection_v2.py - Generates the figure
sticking_crosssection.pdf

** Folder Figure03:
Several files to generate the figure, requires direct access to MD analyzed by check-H2-NN.py to generate the site density plots.
Blender v2.90.1 is used to generate the top and side views, and GIMP is used to include those views into the site density plots.
Note that Analysis_site_density.py is also used indirectly to generate figures 4 and 5.

** Folder Figure04:
plotstepratio.py - Generates the figure
stepratio.pdf

** Folder Figure05:
plotshadowratio.py - Generates the figure
shadowratio.pdf

** Folder FigureS01:
plotsticking_trapping.py - Generates the figure
sticking_trapping.pdf

** Folder FigureS02:
Analysis_x_disp.py - Obtains the data from the analysis of the MD trajectories, requires check-H2-NN.py
plotx_disp.py - Generates the figure
Xdisplacement.pdf

** Folder FigureS03:
plotsticking_facet.py - Generates the figure
sticking_facet.pdf

** Folder FigureS04:
crosssection_fit.pdf
plotcrosssection_fit.py - Generates the figure

** Folder FigureS05:
Analysis_site_density.py - Generates the figure directly from MD data, requires check-H2-NN.py
site_density_SI.pdf

** Folder DFT:

**** Subfolder HDNNP:
Inputs for the dataset to train and test the HDNNP with.

LGIsubmitter_977.py - Example of how the calculations are submitted.
slab_H2-poscar.py - Generates the POSCARs. Unfortunately, when uploading the files to the pubs a mistake was made and the most recent version was overwritten by accident. Therefore, this is a slightly older version, but it remains unclear what the differences are supposed to be. Nevertheless, it shows how in general the structures are generated. Please see the SI for more details.
slab_H2-poscar_surfacemotion.py - If one wants to include surface atom motion in the training data.

VASP inputs:
INCAR
KPOINTS
POTCAR - Not supplied here due to licensing, used PAW_PBE Pt 04Feb2005 and PAW_PBE H 15Jun2001
vdw_kernel.bindat - Not supplied here due to licensing

Number of k-points for each surface:
111 - 6x6x1
211 - 6x6x1
533 - 6x6x1
322 - 4x6x1
755 - 4x6x1
433 - 4x6x1
977 - 4x6x1

Technically, for 433 and 977 it would have been better to use 3x6x1 to be really consistent, but it seems that it did not yield any issues.

**** Subfolder Slab:
Resulting surfaces with interlayer relaxations can be found in MolecularDynamics/standard_fs
INCAR - VASP input
KPOINTS - VASP input
slab-poscar.py - Generates the POSCAR

** Folder MolecularDynamics:
AIMD - Contains all the files (except the vdW kernel and POTCAR for licensing reasons) to perform AIMD on Pt(211).
facet_angle.xlsx - Contains the incidence angle for the stepped surfaces (search for 'angle between the two vectors') as simulated for experimental conditions.
standard_fs/ - Contains all the files to perform MD simulations
submit-dynamics.py - Submits the calculations. Look at this script to have a better understanding of how the calculations are set up. The codes are your documentation.

** Folder RuNNer:
Contains all the files required by RuNNer (or by LAMMPS with the correct USER module or n2p2) to use the HDNNP.