In this section we will describe how to do LCAO calculations using ABACUS. Again the crystal Si in the diamond structure will be taken as an example.
In the work directory, copy the INPUT, KPT and STRU file, the pseudopotential file(Si.pz-vbc.UPF), and in addition the numerical atomic orbital file(Si_lda_8.0au_50Ry_2s2p1d) from the examples/02a_Si_diamond_lcao_scf/ directory.
Note the keyword
basis_type lcao
in the INPUT file.
The information printed on the screen is different from that obtained using the plane-wave basis:
*********************************************************
* *
* WELCOME TO ABACUS *
* *
* 'Atomic-orbital Based Ab-initio *
* Computation at UStc' *
* *
* Website: http:https://abacus.ustc.edu.cn/ *
* *
*********************************************************
Fri Jul 2 11:23:16 2021
MAKE THE DIR : OUT.ABACUS/
DONE(0.0615518 SEC) : SETUP UNITCELL
DONE(0.0656908 SEC) : INIT K-POINTS
---------------------------------------------------------
Self-consistent calculations for electrons
---------------------------------------------------------
SPIN KPOINTS PROCESSORS NBASE
1 64 1 26
---------------------------------------------------------
Use Systematically Improvable Atomic bases
---------------------------------------------------------
ELEMENT ORBITALS NBASE NATOM XC
Si 2s2p1d-8au 13 2 PZ-LDA
---------------------------------------------------------
Initial plane wave basis and FFT box
---------------------------------------------------------
SET NONLOCAL PSEUDOPOTENTIAL PROJECTORS
DONE(1.73298 SEC) : INIT PLANEWAVE
UNIFORM GRID DIM : 36 * 36 * 36
UNIFORM GRID DIM(BIG): 18 * 18 * 18
DONE(1.7346 SEC) : INIT CHARGE
DONE(1.73558 SEC) : INIT POTENTIAL
START POTENTIAL : atomic
---------------------------------------------------------
SELF-CONSISTENT :
---------------------------------------------------------
ITER ETOT(eV) EDIFF(eV) DRHO2 TIME(s)
GE1 -2.138667e+02 0.000000e+00 1.652e-01 4.690e+00
GE2 -2.139153e+02 -4.859216e-02 3.480e-02 4.970e+00
GE3 -2.139161e+02 -8.407097e-04 4.131e-03 4.760e+00
GE4 -2.139161e+02 -1.051285e-05 4.859e-05 4.630e+00
GE5 -2.139161e+02 2.951636e-08 6.059e-06 5.120e+00
GE6 -2.139161e+02 -3.891962e-09 1.608e-07 5.210e+00
GE7 -2.139161e+02 1.061725e-10 7.775e-10 4.930e+00
|CLASS_NAME---------|NAME---------------|TIME(Sec)-----|CALLS----|AVG------|PER%-------
A Run_lcao lcao_line 25.505 1 26 1e+02 %
B ORB_control set_orb_tables 1.6457 1 1.6 6.4 %
C LCAO_Orbitals Read_Orbitals 1.0609 1 1.1 4.2 %
C ORB_gen_tables gen_tables 0.58477 1 0.58 2.3 %
D ORB_table_phi init_Table 0.33613 1 0.34 1.3 %
B LOOP_ions opt_ions 23.744 1 24 93 %
C LOOP_elec solve_elec_stru 23.744 1 24 93 %
D LOOP_elec set_matrix_grid 0.13358 1 0.13 0.52 %
D Grid_Technique init 0.10495 1 0.1 0.41 %
D LOOP_elec before_solver 5.2475 1 5.2 21 %
E LCAO_Hamilt set_lcao_matrices 5.2453 1 5.2 21 %
G LCAO_gen_fixedH build_Nonlocal_mu 5.1953 1 5.2 20 %
X ORB_gen_tables snap_psibeta 3.6577 1703858 2.1e-06 14 %
D LOOP_elec solver 18.363 1 18 72 %
D ELEC_scf scf 18.363 1 18 72 %
E ELEC_cbands_k cal_bands 13.126 7 1.9 51 %
K Gint_k vlocal 8.6586 7 1.2 34 %
F LCAO_Hamilt calculate_Hk 3.927 448 0.0088 15 %
G LCAO_nnr folding_fixedH 3.7241 448 0.0083 15 %
G Diago_LCAO_Matrix genelpa 2.2571 224 0.01 8.8 %
G Diago_LCAO_Matrix genelpa1 0.1584 448 0.00035 0.62 %
G Diago_LCAO_Matrix genelpa2 0.29915 448 0.00067 1.2 %
E Local_Orbital_Cha sum_bands 5.1631 7 0.74 20 %
F LCAO_Charge cal_dk_k 0.20373 7 0.029 0.8 %
F Gint_k charge 4.9517 7 0.71 19 %
----------------------------------------------------------------------------------------
START Time : Fri Jul 2 11:23:16 2021
FINISH Time : Fri Jul 2 11:23:42 2021
TOTAL Time : 26
SEE INFORMATION IN : OUT.ABACUS/
The string GEn in the first column means that the genelpa eigenvalue solver is used, and this is the n-th self-consistent KS iteration. In contrast, the output information from the PW calculation has the string CGn in its first column, indicating the conjugate gradients (CG) method is used to solve the Kohn-Sham equation.
In many cases, only one Γ point (i.e., k=0) is needed in the calculations. In these cases, one can set in the INPUT file:
- gamma_only If set to 1, only Γ point is used in the calculation. The gamma_only algorithm is expected to be faster than the standard algorithm.