Why no input file template for structure optimization？

[Pin Chen]

Dear developers,

I am new to abipy and try to use this tool to do high-throughput computation. I have downloaded a lot cif files from ICSD to prepare the input files for abinit. In the beginning, I need to optimize all the structures . However, I do not find how to create input files with optimization by abinp.py , and only the following calculations are considered:

    gs                  Build Abinit input for ground-state calculation.
    ebands              Build Abinit input for band structure calculations.
    phonons             Build Abinit input for phonon calculations.
    g0w0                Generate input files for G0W0 calculations.
    anaph               Build Anaddb input file for the computation of phonon
                        bands DOS.
    wannier90           Build wannier90 template input file from Abinit
                        input/output file. possibly with electron bands
    lobster             Build and print Lobster input file from directory with
                        files file (for pseudos) and GSR.nc output file
                        (multidatasets are not allowed). Supports also
                        directory with vasprun.xml and POTCAR file.

So, my question is how can i add calculation method into abinp.py (for example, generating input files for structure optimization)? Can you give me some guides?

Regards.

Pin.

[Jalen Cates]

I believe what you need is in the functions of the gs calculations. Refer to the Jupyter Notebook of flows for examples of high-throughput flows. "Adding a function" can be done within a python environment with custom written scripts saved in a local directory. If you set the python interpreter correctly and import the modules, then it should work fine that way.

Cheers,
Jalen

[Pin Chen]

Another issue is somewhat irrelevant to the topic, but it is very important for our subsequent calculations. I kindly ask for your opinion. The abinit software is very powerful in the development of potential functions. The related literature [https://science.sciencemag.org/content/351/6280/aad3000.full] shows that the accuracy of the abnit JTH02 method is very close to that of the vasp GW method, but the latter calculation (O(N^5)) is very time consuming. I am not very familiar with JTH02 Method, is the calculation time of this method similar to GW?

By the way, the main Material databases, such as Material project, aflow, and jarvis, are all calculated by using vasp. If Abinit's JTH02 method has advantages in terms of calculation accuracy and speed, it will be a competitive DFT calculation engine in the method of large-scale high throughput .