welcome: please sign in
location: Diff for "scf"
Differences between revisions 66 and 67
⇤ ← Revision 66 as of 2017-11-26 06:25:17 →
Size: 14114
Editor: huangc
Comment:
← Revision 67 as of 2018-10-17 22:51:05 → ⇥
Size: 14156
Editor: leiyibo
Comment:
Deletions are marked like this. Additions are marked like this.
Line 228: Line 228:
The maximum Number of SCF iteration The maximum Number of SCF iteration. For MCSCF module, maxiter = 2 is proper.

SCF

Contents

  1. SCF
    1. General keywords
      1. RHF/UHF/ROHF
      2. RKS/UKS/ROKS
      3. Occupy
      4. Alpha
      5. Beta
      6. Charge
      7. Spin
      8. keyword
    2. DFT functional keywords
      1. DFT
      2. RS
      3. D3
    3. DFT grid keywords
      1. NPTRAD
      2. NPTANG
      3. Grid
      4. Gridtype
      5. Partitiontype
      6. Numinttype
      7. NoSymGrid
      8. DirectGrid
      9. NoDirectGrid
      10. NoGridSwitch
      11. ThreshRho
      12. ThreshBSS
      13. Coulpot
      14. Coulpotlmax
      15. Coulpottol
    4. SCF convergence
      1. MAXITER
      2. NODIIS
      3. MaxDiis
      4. THRENE
      5. THRDEN
      6. ThreshConverg
      7. THRDIIS
      8. DIISmode
      9. Vshift
      10. Damp
      11. Icheck
      12. IAUFBAU
    5. Diagonalization
      1. Blkiop
      2. iviop
    6. Print and output SCF orbital
      1. print
      2. iprtmo
      3. Noscforb
      4. Pyscforb
      5. Molden
    7. Expert keywords
      1. IfNoDeltaP
      2. IfDeltaP
      3. Optscreen
      4. Nok2Prim
      5. FixDif
      6. Jengin
      7. LinK
      8. Guess
      9. Cutlmotail
      10. CHECKLIN
      11. TOLLIN
      12. ifPair
      13. pinalpha , pinbeta
  2. Depend Files
  3. Examples
    1. How to perform a direct DFT calculation with B3LYP functional?
    2. How to read molecular orbital as initial guess orbital or restart SCF calculation?

HF/DFT.

General keywords

RHF/UHF/ROHF

  • Must input one of them if Hartree-Fock calculation is required. Require for restricted/unrestricted/restricted open shell Hartree-Fock calculations.

Example: 

$Scf
RHF
$end

RKS/UKS/ROKS

  • Must input one of them if Kohn-Sham calculation is required. Require for restricted/unrestricted/restricted open shell Kohn-Sham calculations.

Occupy

  • Used in RHF/RKS. Set double occupied number of each irreps. The following line is an integer array, $noccu(i),i=1,\cdots, nirreps$.

Alpha

  • Used in UHF/ROHF/UKS/ROKS. Set number of alpha orbitals in each irreps. The following line is an integer array, $nalpha(i),i=1,\cdots, nirreps$.

Beta

  • Used in UHF\/ROHF\/UKS\/ROKS. Set number of beta orbitals in each irreps. The following line is an integer array,$ nbeta(i),i=1,\cdots , nirreps$.

Charge

  • Charge of the state.

Spin

  • Spin of the state. The value is 2S+1.

keyword

DFT functional keywords

DFT

  • DFT functional used in Korn-Sham calculation. Commonly used functionals: SVWN5, BLYP, B3LYP, CAM-B3LYP, etc.

  • LSDA:
    • ex_fun=1 co_fun=1
    SVWN5:
    • ex_fun=1 co_fun=2
    PW91:
    • ex_fun=3 co_fun=3
    SAOP:
    • ex_fun=30 co_fun=30
    BLYP:
    • ex_fun=4 co_fun=8
    BHHLYP
    • ex_fun=21 co_fun=8
    B2PLYP
    • ex_fun=22 co_fun=8
    B3LYP
    • ex_fun=20 co_fun=0
    LC-BLYP
    • ex_fun=104 co_fun=8
    CAM-B3LYP
    • alpha=0.19d0 beta=0.46d0 ex_fun=120 co_fun=0
    B3PW91
    • ex_fun=27 co_fun=0
    PBE PBE0 VBLYP GBLYP SF5050 LC-BVWN5 LC-BLYP SAOP

RS

  • Alpha and beta value in CAM calculation. The following line are two float number. For example : 0.33 0.15

D3

  • Grimmers dispersion corrrection for DFT.

DFT grid keywords

NPTRAD

  • Number of radius grid points.

NPTANG

  • Number of angular grid points.

Grid

  • Set DFT grid. Support values are: Ultra Coarse, Coarse, Medium, Fine, Ultra Fine, SG1.

Gridtype

  • DFT grid type control, integer.
    • 0, Radial(new kind Chebeshev used by Becke) Angular(Lebedev).
    • 1, Radial(Chebeshev) Angular(Lebedev).
    • 2, Radial(Eular-Maclarin) Angular(Lebedev).
    • 3, Radial(ut_rad) Angular(Lebedev).
    Default: 0

Partitiontype

  • DFT grid partition type control, integer.
    • 0: Becke partition. 1: Stratmann-Scuseria-Frisch partition.
    Default: 0

Numinttype

  • Numerical integration code control, in form x*10+y.
    • y: 0, use default numerical integration code, else debug old numerical integration code.
    • x: print control parameter for default numerical integration code, only useful when y==0
    Default: 0

NoSymGrid

  • Do not use symmetry dependent grid. Only for debugging.

DirectGrid

  • Use DirectGrid. Basis set values on the grid points are calculated directly. Default: Direct SCF, use direct grid. None Direct SCF, do not use direct grid.

NoDirectGrid

  • Force to do not use direct grid.

NoGridSwitch

  • For direct SCF, DFT grid can be switched. At the beging of iteration, Ultra coase grid will be used. After energer change is little than a value, such as 1.d-4, the medium grid or user setted grid

    will be used. NoGridSwitch dissiable grid switch and use default grid directly.

ThreshRho

  • When use debug numerical integration (see NUMINTTYPE):

    • If the numerical integral $\rho_{\mu}^k<threshRho$, the basis $\chi_{\mu}$ will be neglected at grid batch k. The $\rho_{\mu}^k$ is defined as

      • $\rho_{\mu}^k=\sum_{i\in batch k} w_i*\chi_{\mu}(r_i)\chi_{\mu}(r_i)$

      Default value: $ThreshRho=\frac{thresh\_{ene}}{maxradgrid*maxanggrid*natom}$

    When use default numerical integration:

    • Neglect the basis $\chi_{\mu}(r)$, if $r\geq r_{\mu}$. The $r_{\mu}$ is defined as

      • $f(r_{\mu})=\int_{r_{\mu}}^{\infty} \chi_{\mu}^2 r^2dr = ThreshRho$

      If input values is $\eta$,

      • $ ThreshRho = \left \{ \begin{array}{ll} 10^{-10} & default \\ 10^{-\eta} & \eta \geq 1 \\ \eta & \eta < 1 \end{array} \right. $

ThreshBSS

  • Only useful when use default numerical integration.

    • Neglect the basis $\chi_{\mu}(r)$ at grid batch k, if $f_{\mu}\leq ThreshBSS$. The $f_{\mu}$ is defined as

      • $f_{\mu}=\sum_{r_g \in k} |\chi_{\mu}| \sqrt{w(r_g)}$

      If input values is $\eta$,

      • $ ThreshBSS = \left \{ \begin{array}{ll} min(10^{-8}, \sqrt{ThreshRho}/N) & default,\ N=\frac{NumberOfGridBatch\times 10}{NumberOfAtom} \\ 10^{-\eta} & \eta \geq 1 \\ \eta & \eta < 1 \end{array} \right. $

Coulpot

  • Integer number, control parameter for generation of Coulomb (Vc) and Nuclear attraction (Vn) matrix.
    • 0 get both Vc and Vn by analytical integration.
    • 1 get coulomb potential with multipolar expansion, and get Vc by numerical integration.
    • 2 get coulomb potential with multipolar expansion, and get both Vc and Vn by numerical integration.
    default: 0
    • NOTE, when do Hartree-Fock or Hybrid DFT with numerical integration of coulomb matrix, must use sketeleton matrix method to do 2e-integral, i.e., need keywork skeleton in module Compass.

Coulpotlmax

  • Max L value for coulomb potential multipolar expansion. Default value: 8

Coulpottol

  • Cutoff threshold parameter for coulomb potential multipolar expansion, more higher more accurate. Default value: 8.

SCF convergence

MAXITER

The maximum Number of SCF iteration. For MCSCF module, maxiter = 2 is proper.

NODIIS

Logical control parameter. Disable DIIS.

MaxDiis

  • Maxim number of Diis space. Default: 8

THRENE

  • Convergence threshhold for energy. Default: 1.d-8.

THRDEN

  • Convergence threshhold for density matrix. Default: 3.d-6.

ThreshConverg

  • Convergence threhhold. Two float value: DeltaE DeltaD 

$SCF
threshconverg                                                               
1.d-6 1.d-4
$END

THRDIIS

  • Threshold to turn on DIIS. Default: 0.15.

DIISmode

DIISmode: 
0: diisdim goes from 0 to maxdiis, then cycles to 0. And reset to 0 when diis fails. 
1: diisdim goes from 0 to maxdiis, keeps maxdiis. And throw the oldest vector (reduce diisdim) when diis fails.
Default: 0.

Vshift

  • Level shift value.

Damp

  • Damping value.

Icheck

  • Check Aufbau law.

IAUFBAU

Control parameter of electron occupation protocol in each SCF iteration. IAUFBAU = 1, electron occupation obeys Aufbau principle(default); IAUFBAU = 2, electrons complies with specific occupation pattern based on maximum occupation method(mom); IAUFBAU = 3, electrons complies with specific occupation pattern based on maximum occupation method(mom).Update MO coefficients and reorder occupied orbitals in each iteration. WARNING if IAUFBAU=2 or 3 without initial guess=read (this means initial guess is bad), the result is unpredictable.

Diagonalization

Blkiop

7 and 8 for iVI diagonalization otherwise specific pFLMO diagonalization: 1: SAI, 2: DDS, 3: DNR, 4: DGN, 5: FNR, 6: FGN 8: CHC rotation with Fock screen, full diagonalization 7: iVI diagonalization, specific by iviop

iviop

1：CHC rotation with Fock screen, automatic switch betwwen iVI and Dsyev. 2：iVI for GEP (generalized eigenvalue problem) diagonalization. 3：iVI for EP (eigenvalue problem) with Cholesky decomposition of S.

print

  • Print level.

iprtmo

  • Require to print MO coefficients. Values: 1 Only print orbital energy and occupation numbers. 2 Print all information.

Noscforb

  • Not output SCF orbital in .scforb file.

Pyscforb

  • Output SCF orbital into Pyscf format file.

Molden

  • Output SCF orbital into Molden format file.

Expert keywords

IfNoDeltaP

  • Dissable using DeltaP to update Fock matrix.

IfDeltaP

  • Delta P is used to update density matrix. In direct SCF calculation, delta P will be used in integral prescreening instead of P. Default: true.

Optscreen

  • For debugging. Set a strict threshold (thresh_rho=1.d-4) for integral prescreening directly.

Nok2Prim

  • Disable primative integral screenning via K2 integrals. Use (SS|SS) esitimating primative integral value and perform screening. Default: Direct SCF, use K2 primative screening.
    • None Direct SCF, use (SS|SS) integral.

FixDif

  • Fix factor for incremental fock update. If the factor is not fixed, use the formular

$fac=1-\frac{D^{n+1}-D^n}{D^{n+1}*D^{n+1}}$

$F^{n+1}=F^n+fac*\delta F$

  • if using fixed factor, fac=1.d0.

Jengin

  • Use Jengin method calculate J matrix. In debugging, not support now.

  • Use LinK calculate K matrix. In debugging, not support now.

Guess

  • Method to get initial guess orbital. The following line is a string.
    • Values: Atom, Hcore, Huckel, Read. If Read is used, the old orbital will be read. The old orbital saves in a file named "inporb" in BDF_TMPDIR . It is generated by previous SCF calcualtion with the name Task.scforb.

Cutlmotail

  • Methods to cut long Coulomb tails of Local molecular orital. Values: -1 Do not cut tail. 1 Project a LMO into fragment with largest Lowdin population.
    • 2 Similar with 1, but project a LMO into predefined group of fragments with largest Lowdin population. 3 Very stick cutoff. Project LMO to a fragment plus several atoms. The threshhold is 1.d-4.
    Comment: Method 1 is prefered if fragments are well defined. We can easy reduce compuations times in post
    • SCF calcualtion based on LMO because diffirent fragment interaction policy can be predefined, which will reduce ERIs need to be calculated.

CHECKLIN

Check if the basis sets is linear dependent. If diffuse basis set is used, SCF do not converge or ridiculous energy observed, it is better to check linear dependent of the basis set. 

$SCF
checklin
$END

TOLLIN

Tolerance of basis set linear dependent. Default value 1.d-7. 

$SCF
tollin
1.d-5
$END

ifPair

used to excite electrons (MOM)with following keywords:
hpalpha,hpbeta
then with number of partical-hole pairs N
then with 2N lines specificate partical-hole pairs. (0 is do nothing, indexes start from 1)
eg. the molecular is has 4 irreducible representation, we want to excite electrons from orb 5,6 to 8,9 in rep 1 and 3 to 4 in rep 3 (alpha) & 7 to 8 in rep 1(beta): 

ifpair
hpalpha
2
5 0 3 0
8 0 4 0
6 0 0 0
9 0 0 0
hpbeta
1
7 0 0 0
8 0 0 0

this should be combined with iaufbau=2 or 3.
WARNING: this function will not check whether partical orbital is filled or whether hole orbiltal is not filled.

pinalpha , pinbeta

specificate fix orbitals
first line specificates the number of fix orbitals
then with N lines specificate fix orbitals. (0 is do nothing, indexes start from 1)
(somewhat likes hpalpha/hpbeta input)
these keywords leads to SCF_solver from solve FC=SCE to $\tilde{F}U=UE,\tilde{F}=C^\dagger FC$

Depend Files

Filename

Description

Format

Examples

How to perform a direct DFT calculation with B3LYP functional?

$COMPASS 
Title
 Cocaine Molecule test run, CC-PVDZ
Basis
  CC-PVDZ
Geometry
  XYZ               # The molecule geometry will be read from file $BDFTASK.xyz 
End Geometry
Skeleton          # This keyword must be used.
$End

$xuanyuan
Direct              # Direct SCF.
Schwarz          # Schwarz prescreening.
$end

$scf
RKS
DFT functional
 B3LYP
Molden     # This keyword is used to output SCF orbital to molden format file.
$end

How to read molecular orbital as initial guess orbital or restart SCF calculation?

Suppose you have performed a calculation and generated aSCF orbital file in your work directory as test.scforb. Usually, this file atomically generated by SCF module. This file also can be used to restart SCF calculation via read it as initial guess orbital. 

$COMPASS 
Title
 Cocaine Molecule test run, CC-PVDZ
Basis
  CC-PVDZ
Geometry
  XYZ    # The molecule geometry will be read from file $BDFTASK.xyz. 
End Geometry
Skeleton          # This keyword must be used.
$End

$xuanyuan
Direct              # Direct SCF.
Schwarz          # Schwarz prescreening.
$end

# Copy orbital file test.scforb as inporb in BDF_TMPDIR
% cp $BDF_WORKDIR/test.scforb $BDF_TMPDIR/inporb

$scf
RKS
DFT functional
 B3LYP
Guess       # Read orbital as  initial guess orbital
 Read
Molden     # This keyword is used to output SCF orbital to molden format file.
$end

scf (last edited 2022-11-03 14:37:07 by bsuo)