Feature: initialize planewave basis wavefunction by numerical atomic orbitals

[kirk0830]

Introduction and performance

A simple test would be illustrative, I choose the conventionally considered hard-to-converge system examples/smearing/lcao_fe, make following changes:

Tested result:

Usage

• specify psi_initializer 1 in INPUT file, will enable the newly-implemented psi initializer, supports random, atomic, atomic+random, nao, nao+random method. For *+random method, 5% random wavefunction is mixed, in accordance with the old version
• use init_wfc nao the newly added, where nao stands for Numerical Atomic Orbital

Unittests' Note:

The unittest system has been well-estabilished for quite a while, but it is hard for wavefunction initialization function to make really unit test. Therefore tests are designed in a top-down way: use the value generated from old implemented one as reference value.

Feature in future

Stochastic DFT scheme is not supported currently, its related implementation will depend on actual demands. Therefore currently if INPUT has two lines like:

psi_initializer 1
esolver_type sdft

, will switch psi_initializer to 0 automatically.

[Qianruipku]

If results of KG do not change too much, you can update them because they are sensitive.

[kirk0830]

If results of KG do not change too much, you can update them because they are sensitive.

Thanks Qianrui, I find failed tests are all SDFT related. Maybe there are variables not initialized, I will examine and find out.