aroundpeking / copt

Basis set optimization for correlation calculation

cOpt: basis set optimization for correlation calculation

This is an optimization module for basis set, based on ABACUS and LibRPA. And starting point is recommended by SIAB.

Installation

git clone https://github.com/AroundPeking/cOpt.git
cd cOpt
conda create -n cOpt
conda activate cOpt
pip install .

Usage

You should prepare opt.json to set optimization parameters.

A folder abacus_inputs containing all inputs for ABACUS calculation in iterations, e.g. single atom RPA calculation. Besides, ORBITAL_RESULTS.txt is needed to be the starting point. Here SPIN*_CHG.cube is recommended to speed up SCF calculation in every iteration.

A folder to store all outputs, i.e. work_dir.

Then run

python cOpt/cOpt/driver.py

you can get file iter.$SLURM_JOB_ID.out, folder best_orb_$SLURM_JOB_ID and opting_$SLURM_JOB_ID.

PARAMETERS in opt.json

dft: str, support "rpa_pbe" or "hf" now.

fix and mod: list, e.g. if you want to optimize DZP while fixing SZ, set fix=[1, 1, 0] and mod=[2, 2, 1]. Don't recommend to optimize SZ by cRPA due to distorted DFT.

init_chg: bool, true is recommended to speed up SCF calculation in every iteration. If true, density SPIN*_CHG.cube should be put in abacus_inputs.

maxiter: int, maximum number of iterations.

opt_method and method: str. If opt_method="local opt", "Powell" (recommended), "CG", "BFGS", ... are optional. If opt_method="global opt", method=basinhopping.

freq_disp: int, freqency of displaying change of object in iter.$SLURM_JOB_ID.out.

work_dir: str, outputs directory.

abacus_inputs: str, ABACUS inputs directory.

abacus and librpa: str, ABACUS and LibRPA executable program.