The main aim of this analysis is to study the impact of the usage of the pseudopotentials (psp) on the evalutation of the statical polarizability for a wide class of molecules given by the dataset of the paper of Head-Gordon (HG). In order to obtain this information we compare the results produced by different psp (keeping fixed the xc potential) with the benchmark value provided by an all electron computation performed with the same xc. Results of HG are obtained exactly in this way and so represents a natural benchmark for our tests (the only possible subtle point is given by the fact that HG uses a localized basis set, but this fact should not introduce a bias in the evaluation of the statical polarizability if the completeness of the basis is sufficient). As ulterior benchmark we can also use the results of the Norvergian group (when available), since they perform all electron computations in a wavelet basis.
The analysis is structured in a set of notebooks built in a sort of top-down approach. The (hierarchycally ordered) list of notebooks is:
-
HG Dataset/Benchmark data : perform a parsing of the results of HG and build a dictionary saved in hg_data.yaml. The dictionary contains the reference results for the CCSD(T) method and for the DFT computations with xc = lda,pbe,pbe0. Also, for each molecule the type of the study (nsp or sp) is reported.
-
Calculations/Construction of the dataset : This notebook reads the hg_data.yaml file and build two dictionaries, one for nsp and one for the sp dataset extraced by the HG data. For each molecule, the dictionary contains the reference results and the 'study' key. The allowed values of the study type are the tuples: (lda_pt,hgh_k),(lda_pw,hgh_k),(pbe,hgh_k),(pbe0,hgh_k). Moreover, for a subset of molecules also two type of nlcc psp are provided (based on the pbe xc), so for these cases there are two further studies : (pbe,nlcc_aw) and (pbe,nlcc_ss).
-
Calculations/Single study nsp calculator : Define (and test) the complete procedure to compute the statical polarizability. The procedure contains various convergence tests, in order to identify :
- the converged value of rmult for the gs computation,
- the converged value of the field intensity,
- the converged value of rmult for the evaluation of the statical polarizability tensor.
-
Calculations/Dataset calculator : Read the dataset dictionaries built by 'Construction of the dataset' and compute the statical polarizability for all the molecules and study type. Save the results in the files (sp)nsp__results.yaml. This files are the starting point for subsequent data analysis. The statical polarizability is computed according to the methods (sp)nsp_workflow defined in the Routines.Py. The procedures used in these methods are described in the notebooks 'Single study sp/nsp calculator'.
-
Calculations/Data analysis : Load the dictionary with the results written in the (sp)nsp_results.yaml file and perform data analysis and post processing. Actually the data analysis consists in:
- check if the all the studies have produced converged results.
- display the converged value of the field intensity and of rmult
- compute the RMSRE and the MRE, as defined in the paper of HG and build one table that collects the effects of the choice of the various classes of psp on the statical polarizability.