Digital Twin for Chemical Sciences

Figure 1-4 are created in illustrator.

The procedure for generating data in Figure 3 can be found in the demo notebook in Supplementary Code.

The procedure for generating data of Figure 4 has been uploaded in fig4_figshare.zip file. Specifically, this involves the following three steps, as specified in fig4_figshare.zip. We first used the Data Generation.ipynb python notebook in the 1_inputs folder to generate the .csv files for one pressure condition, two pressure conditions, and six pressure conditions. They are included in the subfolders 1run, 2pressures, and 6pressures under the 1_inputs folder. These .csv files represent the subfigures in Figure 5a), b), and c) respectively. Next, we used the Basin Hopping.ipynb and Gaussian Processing.ipynb python notebooks under the 2_bh_and_gp folder to run Basin Hopping and Gaussian Process inverse solutions, respectively. The inputs for these notebooks are the .csv files provided in 1_input. The outputs of these notebooks are another set of .csv files saved in the basin_hopping and gaussian_processing subfolders. For streamlined experiences, we do not recommend overwriting these .csv files. We report all results that are stopped manually at 1000 iterations. Lastly, we use the Figure 4.ipynb notebook in the 3_outputs folder to plot the subfigures in d), e), f). Observing that Basin Hopping performs better than Gaussian Process, we plot the degeneracy result with 0.1 error cutoff to obtain the subfigures in g), h), i). The subfigures d-f are included in the subfigures subfolder. Additionally, the spectra .png files corresponding to subfigures a), b), and c) are included under the extras folder. We have also included a README file under each of the folders: 1_inputs, 2_bh_and_gp, 3_outputs, and extras, to help guide the process. Please change the path in the jupyter notebooks accordingly in order to run the code.