In this repository we provide exemplary code on how to construct and evaluate a state-to-distribution (STD) model for a reactive atom-diatom collision system.

• python 3.7
• TensorFlow 2.4
• SciKit-learn 0.20

We recommend to use Miniconda for the creation of a virtual environment.

Once in miniconda, you can create a virtual enviroment called StD from the .yml file with the following command

conda env create --file StD.yml

On the same file, there is a version of the required packages. Additionally, a .txt file is included, if this is used the necessary command for the creation of the environment is:

conda create --file StD.txt 

To activate the virtual environment use the command:

conda activate StD

You are ready to run the code.

To predict product state distributions for fixed nitial conditions from the test set (77 data sets). Go to the evaluation_InitialCondition folder.

Don't remove (external_plotting directory).

python3 evaluate.py 

The evaluate.py file predicts product state distributions for all initial conditions within the test set and compares them with reference data obtained from quasi-classical trajectory similations (QCT).

Edit the code evaluation.py in the folder evaluation_InitialCondition to specify whether accuracy measures should be calculated based on comparison of the NN predictions and QCT data solely at the grid points where the NN places its predictions (flag "NN") or at all points where QCT data is available (flag "QCT") based on linear interpolation. Then run the code to obtain a file containing the desired accuracy measures, as well as a PDF with the corresponding plots. The evaluations are compared with available QCT data located in QCT_Data/Initial_Condition_Data.

To predict product state distributions from thermal reactant state distributions go to the evaluation_Temperature folder.

Edit the code evaluation.py in the folder evaluation_Temperature, to specify which of the four studied cases

• Ttrans=Trot=Tvib (indices_set1.txt)
• Ttrans != Tvib =Trot (indices_set2.txt)
• Ttrans=Tvib != Trot (indices_set3.txt)
• Ttrans != Tvib != Trot (indices_set4.txt)

you want to analyse.

Then run the code with the following command to obtain a file containing the desired accuracy measures, as well as a PDF with the corresponding plots for three example temperatures.

Don't remove (external_plotting directory).

python3 evaluate.py

The evaluations are compared with the available QCT data in QCT_Data/Temp_Data.

The complete list of temperatures and can be read from the file tinput.dat in data_preprocessing/TEMP/tinput.dat .

Julian Arnold, Debasish Koner, Juan Carlos San Vicente, Narendra Singh, Raymond J. Bemish, and Markus Meuwly,

!*Complete name of paper or do you want to cite the repository? Also, add an email or responsable*