Re: Why openMX need wannier ( No.1 ) 
 Date: 2023/06/11 18:03
 Name: Kylin <kylinspecies@gmail.com>
 Although both PAO and Wannier orbitals can represent the same Hilbert space, Wannierization is favored for representing the band structure near the Fermi surface using a Hamiltonian with reduced size. Additionally, Wannier orbitals have a rigorous mathematical foundation, providing them with physical meaning and valuable visualization for further analysis.
Furthermore, a similar dimension reduction (Hamiltonian extraction) can be achieved using Localized Natural Orbitals (LNOs), where the atomic basis on each atom is contracted to represent any (partial) band structure with a minimal basis. However, we cannot provide a mathematical foundation as solid as the one for Wannierization.

Re: Why openMX need wannier ( No.2 ) 
 Date: 2023/09/09 22:10
 Name: Alexandre Cavalheiro <alexandre.dias@unb.br>
 Dear OpenMX developers, I'm Alexandre C. Dias, main developer of WanTiBEXOS code (https://doi.org/10.1016/j.cpc.2022.108636), our published version works with wannier90 Hamiltonians, however for larger systems, the wannierization procedure are problematic, in order to solve this, we are building a new version of our code that would works directly with DFT localized basis set packages Hamiltonians, we are trying to build an interface with Siesta, FHIAIMS and OpenMX, that are codes that externally write their DFT Hamiltonians and Overlap matrices. Our problem with OpenMX is that the format that you output these Hamiltonian and Overlap matrices are very difficult to understand and rewrite in a format readable for WanTiBEXOS package, so I'd like to ask for a help from the developers to rewrite this output Hamiltonian and Overlap matrices in a format similar to Wannier90 with with an extra collumn that contains the overlap elements, bellow I show an example of a Siesta Hamiltonian and Overlap, readable by WanTiBEXOS:
SOC 0.000000 !scissors operator 5.380821 ! Fermi level 3.165546 0.000000 0.000000 !lattice vectors 1.582773 2.741443 0.000000 0.000000 0.000000 16.000000 82 81 #rcell x rcell y rcell z i j ReH ImH S 0.000000 0.0 0.000000 1 1 10.055253 0.000000 0.999995 0.000000 0.0 0.000000 1 2 9.545580 0.000000 0.977032 0.000000 0.0 0.000000 1 3 0.000016 0.000000 0.000000 0.000000 0.0 0.000000 1 4 0.000000 0.000000 0.000000 0.000000 0.0 0.000000 1 5 0.166753 0.000000 0.000000 0.000000 0.0 0.000000 1 6 0.000000 0.000000 0.000000 0.000000 0.0 0.000000 1 7 0.000106 0.000000 0.000000 0.000000 0.0 0.000000 1 8 0.000010 0.000000 0.000000 0.000000 0.0 0.000000 1 9 0.000000 0.000000 0.000000
If you could do it, or build some script that converts the actual format, we can use OpenMX Hamiltonian to solve BSE and obtain excitonic effects in the optical properties, with a much lower computational cost and for larger systems, things that Yambo, Berkley GW and other codes can't do.

