NRLMOL_INPUT.DAT: Controlling a calculation#

The NRLMOL_INPUT.DAT is an auxiliary input file that allows setting of control parameters for calculation. If it does not exist, then it will be created by the FLOSIC executable using default values.

The NRLMOL_INPUT.DAT with default values of parameters as of today (August 15, 2019) is given below.

# Put Y,N or number next to the equal sign to determine execution
# Don't forget the quotation marks for the letters
# All variables in this list end with V

&input_data
ATOMSPHV      = 'N'
BASISV        = 'DEFAULT' ! Specify basis for calculation(basis.txt)
CALCTYPEV     = 'LBFGS'
DFTD3V        = 'N' ! Set to Y to do include Grimmes DFT-D3 dispersion
DIAG1V        =  1  ! diagonalization to use on regular arrays (diagge.f90)
DIAG2V        =  1  ! diagonalization to use on packed arrays (diag_dspgv.f90)
DIAG3V        =  0  ! diagonalization to use on parallel (sdiagge_n.f90)
DMATV         = 'N' ! Create/use/mix density matrix
DOSOCCUV      = 'N' ! Controls wether to calculate density of states (only in DFA)
FIXMV         = 'N' ! Fix spin moment
JNTDOSV       = 'N' ! This calculates jonit density of states (only in DFA)
MAXSCFV       = 100 ! Maximum SCF iterations
MIXINGV       = 'P' ! (H)amiltonian (P)otential (D)ensity matrix mixing
NONSCFV       = 'N' ! Set to Y to do a non SCF calculation
NONSCFFORCESV = 'N' ! Set to Y to calculate forces in a non SCF calculation
NWFOUTV       = 10  ! Write WFOUT file for every N-th iteration
POPULATIONV   = 'N' ! Population analysis
RHOGRIDV      = 'N' ! Set to Y to execute RHOGRID
SCALEDLBFGSV  = 'Y' ! Set to Y to scaled LBFGS (only in SIC)
SCFTOLV       = 1.0D-6 ! SCF tolerance
SPNPOLV       = 'N' ! Run spin polarized calculation from CLUSTER
VERYFINEMESHV = 'N' ! Set to Y to use very fine mesh
SYMMETRYV     = 'N' ! Set to Y to detect symmetry
WFGRIDV       = 'N' ! set to Y to write orbitals in cube format (only in DFA)
WFFRMV        = 'N' ! set to Y to write Fermi orbitals in cube format (only in SIC)
&end

A large number of entries related to calculation of certain properties are given as yes (Y) or no (N). Note that certain calculations such as, for example, the calculation of joint density of states (JNTDOSV) is often useful only at the final converged geometry.

Most of the variables are already explained briefly in the NRLMOL_INPUT.DAT file. Below we add some comments on a few of them.

ATOMSPHV#

Set to Y to calculate charge and spin charge in each inequivalent atom integrated over a sphere of specified radius.

BASISV#

This variable specifies which basis set is employed. The 'DEFAULT' basis is the NRLMOL Basis, optimized for the PBE functional

Other bases are available in the basis.txt file, found in the basis subdirectory. This file lists all the basis sets available.

The following bases are available:

3-21G

6-31G-Blaudeau

Ahlrichs_TZV

Partridge_Uncontracted_1

3-21Gs

6-31Gs

Ahlrichs_VDZ

Partridge_Uncontracted_2

3-21GSP

6-31Gs-Blaudeau

Ahlrichs_VTZ

Partridge_Uncontracted_3

3-21Gs_Polarization

6-31Gs_Polarization

DZVP2

Partridge_Uncontracted_4

3-21ppG

6-31Gss

DZVP

Saddlej

3-21ppGs

6-31Gss_Polarization

GAMESS_PVTZ

STO-2G

6-311G

6-31pGs

GAMESS_VTZ

STO-3G

6-311Gss

6-31ppG

Huzinaga_polarization

STO-3Gs

6-311Gss_Polarization

6-31ppGs

IGLO-II

STO-3Gs_Polarization

6-311pGs

6-31ppGss

IGLO-III

STO-6G

6-311ppG2d2p

Ahlrichs_Coulomb_Fitting

McLeanChandler_VTZ

TZVP

6-311ppGss

Ahlrichs_Polarization

MIDI_Huzinaga

6-31G

Ahlrichs_pVDZ

MINI_Huzinaga

The user only needs to specify the string preceding the .basis extension in the input file. For example, if the user wants to use 6-31G basis then he should replace 'DEFAULT' with '6-31G'.

CALCTYPEV#

Available choices:

  • SCF-ONLY: This choice is used when the user is not interested in an atomic geometry optimization. Atomic forces are not computed with this setting.

  • CONJUGATE-GRADIENT: Atomic geometry optimization is done using the conjugate-gradient algorithm.

  • LBFGS : …WIP…

DIAG1V#

This variable allows different diagonalization algorithms to use for diagonalization of the Hamiltonian.

The options for this variable are:

  1. DSGVX (LAPACKL: computes selected eigenvalues, and optionally eigenvectors).

  2. DSGVD (Default if matrix size is below 100). Uses a divide and conquer algorithm to compute eigenvectors.

  3. DSGV (This is the slowest of the available)

DIAG2V#

This variable allows diagonalization of the Hamiltonian using packed storage format for memory savings. This is useful for large system sizes. Available options are: #. DSPGVX (LAPACK: computes selected eigenvalues, and optionally eigenvectors). #. DSPGVD (Default if matrix size is below 100). Uses a divide and conquer algorithm to compute eigenvectors. #. DSPGV (QR- factorization. This is the slowest of the available)

The option 1 (DSPGVD) is the fastest if all eigenvectors are required otherwise use the default 0.

Note

FLOSIC does not use SCALAPACK

POPULATIONV#

When it is set to Y, Mulliken and Lowdin Population analysis calculations are run. Note that this is available only for spin polarized.

RHOGRIDV#

When set to Y, it will generate a CUBE file for visulaization of total and spin density (spin up - spin down).