Tutorial 2: Benchmarking of Input Settings

General

Every RSA simulation contains two important choices: The size of the unit cell as well as the number of rotations for every adsorbate. While the cell size can be adjusted based on a requested accuracy, the number of rotations should be motivated by the physics of the modeled system. Examples are small rotation steps for adsorbates that are considered to be freely rotating, a step size of 60° or 120° if an adsorbate is adapting to the symmetry of a Cu(111) surface, or even values of 180° or 360° (0°) if the adsorbate is assumend to be static. The value and number of valid rotations is therefore benchmarked to gain information on the sensitivity of the RSA simulations to this value.

Input File Setup

In this tutorial we will again use aniline as our adsorbate. Store the following coordinates in an xyz file:

14
xyz file for aniline
N   5.04118   8.64882   8.56544
C   6.26128   5.16607   9.00043
C   6.26490   6.56249   8.89207
C   5.04571   7.26376   8.83488
C   3.83123   6.55488   8.89628
C   3.84412   5.15850   9.00450
C   5.05496   4.45477   9.03852
H   7.21439   4.63227   9.04668
H   7.21136   7.11047   8.87828
H   2.88123   7.09678   8.88507
H   2.89468   4.61854   9.05367
H   5.05865   3.36626   9.11942
H   5.88382   9.14949   8.85915
H   4.19684   9.14457   8.86255

The structure of the main input file is identical to the previous tutorial. You must create different input files to benchmark the cell size and rotation values.

# Adsorption on Cu(111)
Molecule
    rotationmodus = angle
    rotationangle = 60.0
    fixpointtype = atoms
    fixpointatoms = 1
    structure = ...ADJUST-YOUR-PATH.../aniline.xyz
End

For a benchmark of the rotation values change the rotationangle settings from 360° to 5° or even 2° in case you are patient (recommended: 360°, 180°, 90°, 60°, 45°, 10°, 5°). Keep the cell size fixed during this benchmark. Start with the largest value to get a feeling of the growing compute time with decreasing values.

# Lattice of the Cu(111) surface
Lattice
    transx = 30
    transy = 18
    
    vectors
        2.51883   0.00000   0.00000
        0.00000   4.36274   0.00000
        0.00000   0.00000   1.00000
    end
End

For a benchmark of the cell size change the transx and transy settings from 50x30 to 10x6 (recommended: 50x30, 40x24, 30x18, 20x12, 10x6). Again, keep other settings fixed. Start with the smallest value to get a feeling of the compute time. All other settings are identical to the previous tutorial:

# On-top grid points of the Cu(111) surface
Grid
    points
        0.00000   0.00000   0.00000
        1.25942   2.18137   0.00000
    end
End

# General settings for adsorption of aniline  
Events
    steps = 1000

    eventlist
	    1 ads 1 1.0
    end
End

Running the Simulation

To start the simulations use the perform_multiple_rsa_runs function. Again, you should get a message that the RSA simulations are done after some time based an your chosen benchmark settings.

NRuns = 1000
inputfile_path = "...ADJUST-YOUR-PATH.../input.inp"
rsa_results, Nmolecules, molecules, Ngrids, grids, lattice, events, timings = perform_multiple_rsa_runs(NRuns, inputfile_path);

Evaluation

Use the plot_count_area_histograms function to get the mean and variance of your simulations with changing inputs.

histo, mean, variance, minvalue, minvalue_id, maxvalue, maxvalue_id, avgvalue, avgvalue_id = plot_count_area_histograms(NRuns, rsa_results, Nmolecules, molecules, lattice, plotonly = false, area = 0.6);
mean[4]
variance[4]