Features

• Exact reproduction of occupied eigenstates (wavefunction and eigenenergy) of high-quality DFT planewave calculations (norm conserving or ultrasoft or PAW)
• [QO method] Efficient construction of highly localized atomic-orbital-like basis by rotation of occupied R(k) and direct construction of optimized combination subspace C(k) from the space spanned by R(k) and atomic-orbital Bloch subspace A(k).
• Generate localized quasi-atomic orbitals (QOs) in both real space and Fourier space representations
• Fully reproduce occupied band structure via tight binding calculation from QOs
• Support spin-unpolarized and spin-polarized calculations
• Efficiently and accurately generate 3D Fermi surface
• Output density of states(DOS) and projected density of states(PDOS)
• Mulliken charge and bond order analysis for solids/surface/molecule
• Support Vanderbilt ultrasoft pseudopotential package and convert it to NetCDF file format
• Support point group symmetry for Bloch wave functions and reconstruct Bloch states in full Brillion zone
• Interface between QOs package and various DFT packages (Dacapo and VASP)
• Various utilities such as "dacapo2cfg/cfg2dacapo" converters between Ju Li's cfg file and Campos/Dacapo NetCDF data file
• Various utilities such as "vasp2cfg/cfg2vasp" converters between Ju Li's cfg file and VASP data file
• Options to improve localization property by rescaling pseudo-atomic orbitals
• Output Tight binding hopping Hamiltonian and overlapping matrix based on QOs
• Output XCrySDen XSF file for 3D atomistic structures and QOs
• Output XCrySDen BXSF file for 3D Fermi surface

Manual

Download

Interface

Utilities

FAQ for the original QUAMBO method

1. Why is the condition number very bad(large) sometimes?
   
   This question is related to several possible reasons:
   
   (a) insufficient number of bands in your DFT calculation
   Normally beside N_occ occupied bands(ie. N_occ=10), you need to include N_unocc unoccupied bands to grab unoccupied antibonding states. To be safe, you can choose N_unocc = N_occ(ie. N_unocc=10), therefore N_total_bands = 2*N_occ(ie. N_total_bands=20). In DACAPO case, you can use sim.eband = ElectronicBands(20); In VASP case, you can set NBANDS = 20 in INCAR file.
   
   (b) insufficient K-point sampling in your DFT calculation
   Just like atomic or molecular orbitals, QUAMBOs are localized around ions in real-space and it may has a range of several Angstroms. That means when you calculate periodic systems(ie. crystal or surface), you need to include enough large periodic space in the Born von-Karman(BvK) boundary condition, otherwise QUAMBO would overlap with itself in other image cells. And BvK is normally realized by K-point sampling in DFT codes. You need to increase K-point sampling by setting sim.kpt = NetCDF.Entry( name="KpointSetup", value=[9,9,9]) in DACAPO case and setting
   Monkhorst-Pack
   9 9 9
   0. 0. 0.
   in KPOINTS file in VASP case.
   NOTE that only Monkhorst-Pack K-point sampling is supported by the current QUAMBO code.
   
   (c) incorrect or overcomplete pseudoatomic orbital basis-set in QUAMBO construction
   During QUAMBO construction, we do not need too many atomic orbitals as basis-set since some angular momentum channels will have very small portions in the bonding and antibonding orbitals. Therefore, overcomplete basis-set may give troubles in the matrix diagonalization. You may remove these pseudoatomic orbitals in QUAMBO construction by changing the "QuambInput.m" file:
   Input.ConvertPsp.AddOrDeleteOrbital = -1; % -1: delete orbital, 0: do nothing, 1 for add orbital
   Input.ConvertPsp.DeleteOrbitalType = [1]; % pseudopotential index
   Input.ConvertPsp.DeleteOrbitalIndex = [2]; % pseudoatomic orbital index for the above pseudopotential index

2. For chemical bonding analysis, what kind of basis-set shall I use?
   
   Actually we found that if we include additional angular momentum channels in QUAMBO construction, Mulliken charge will be different. For example, in the case of Pt(111) surface with Oxygen atom absorbed, we should use atomic orbitals sd for Pt and sp for O. If we use dsp for Pt and sp for O, Mulliken charge is different from the previous case. The reason is that the additional p orbitals share the same character as neighbors' orbitals. We need to use the minimal basis-set instead of using overcomplete basis-set.
   

3. Can I do K-point sampling in DFT calculation with symmetry?
   
   Yes, you can run your DFT calculation with symmetry included. In our QUAMBO code, We transform Bloch wavefunctions for each irreducible K-point to those Bloch wavefunctions for the corresponding reducible K-point in the first Brillouin zone. That means you can save a lot of time in DFT calculation by using the crystall or surface symmetry.
   

4. Can I do band structure calculation by my own K-point path during the post-processing of QUAMBO?
   
   Yes, you can. You only need to specify three entries in "QuamboInput.m" file.
   Input.BandStructure.CrystalStructure = ['DEF']; % use 'DEF' tag to define your own k-path
   Input.BandStructure.DefinedKPathName = ['GHNPGN']; % K-path label for each K-point [G for Gamma]
   Input.BandStructure.DefinedKPath = ...
   [ 0 0 0; ...
   -1/2 1/2 1/2; ...
   0 0 1/2; ...
   1/4 1/4 1/4; ...
   0 0 0 ; ...
   0 0 1/2;
   ]; % the corresponding K-path
   see the example at "example/DACAPO/Fe_bcc/QuamboInput.m" or "example/VASP/Fe_bcc/QuamboInput.m". We also include some pre-defined K-point paths for FCC, BCC and HCP. For example, you can use them by setting:
   Input.BandStructure.CrystalStructure = ['FCC'];
   

Bug report

Reference

To-Do

• With generated tight-binding Hamiltonian H and overlapping S matrices, calculate electrical conductance for molecular electronics by both Green's function method and Non-equilibrium Green's function method (top-priority)
• Calculate polarization via Berry phase for both molecular and extended systems
• Implement interfaces for Abinit and PWSCF
• Implement F90 version of QO package

Link