Linker IT Software
Google
Web www.oraxcel.com
menubar-top-links menubar-top-rechts
Home Help Search Login
Welcome, Guest. Please Login.
SQL*XL: Database to Excel bridge litLIB: Excel power functions pack ExcelLock: Locking and securing your valuable Excel spreadsheets encOffice: Protect your Excel file easy and safe encOffice: Protect your Excel file easy and safe
Pages: 1
RASSCF (Read 15990 times)
Gerrit-Jan Linker
YaBB Administrator
*****




Posts: 75
RASSCF
17.05.08 at 13:00:24
 
RASSCF
 
RASSCF: Restricted Active Space SCF
 
Since the number of CSFs quickly increases in CASSCF with the number of active orbitals, along with the computational cost, it may be desirable to use a smaller set of CSFs. One way to make this selection is to restrict the number of electrons in certain subspaces, done in the restricted active space SCF method (RASSCF). One could, for instance, allow only single and double excitations from some strongly-occupied subset of active orbitals, or restrict the number of electrons to at most 2 in another subset of active orbitals.
 
In CASSCF the orbital space is divided into 5 subspaces:

  • external space
    The external space is similar to the virtual or secondary space in the CASSCF description. The external space consists of unoccupied orbitals that remain unoccupied in the CAS expansion.
  • RAS3
    Only a fixed number of added electrons is allowed. The excited electrons in the RAS3 space either come from the RAS1 or RAS2 space.
  • RAS2
    In the RAS2 space a full CI is formed.
  • RAS1
    Only a fixed number of holes is allowed. The number of electrons allowed to be excited from the RAS1 space to the RAS2 or the RAS3 space is limited.
  • inactive space
    The inactive space is similar to the inactive space in a CASSCF description. The inactive space contains doubly occupied orbitals that remain doubly occupied in the CAS expansion.

 
Polarisation CI:
A RASSCF calculation with at maximum one hole in the RAS1 and one electron in the RAS3.
 
See also:
CASSCF
http://www.oraxcel.com/cgi-bin/yabb2/YaBB.pl?num=1198499037
Quantum Mechanical methods
http://www.oraxcel.com/cgi-bin/yabb2/YaBB.pl?num=1202388089
 
Back to top
 
« Last Edit: 12.10.13 at 12:24:33 by Gerrit-Jan Linker »  

Gerrit-Jan Linker
Linker IT Software
Email WWW Gerrit-Jan Linker   IP Logged
Gerrit-Jan Linker
YaBB Administrator
*****




Posts: 75
Re: RASSCF
Reply #1 - 14.08.08 at 14:10:17
 
RASSCF in MOLCAS
 
Example input file:
 
1.   &RASSCF &END
2.  Title
3.  edo-ttf
4.  NACTEL
5.  2 0 0
6.  INAC
7.  66
8.  RAS2
9.  1
10. SPIN
11. 1
12. OUTPUT ALL
12. End of Input
 

  • Title in line 3 can be any text
  • NACTEL: 3 numbers:
    - number of active electrons
      All elections minus 2*(inactive + frozen orbitals)
    - max holes in RAS1
    - max electrons in RAS3
  • INAC: The number of inactive (doubly occupied) orbitals in each symmetry.
  • RAS2: The number of orbitals for each symmetry in the RAS2 subspace.
  • Keywords RAS1 and RAS3 can be used to specify the number of orbitals in these subspaces.
  • OUTPUT FEW FULL displays the most occupied orbitals and a few secondary but with full details. OUTPUT FEW COMPACT will list them with coefficients larger than 0.1. OUTPUT NOTHING will not print any orbitals. OUTPUT ALL will print all orbitals.
  • The keyword ITER can be used to limit the maximum number of iterations (as put on the next line of input)

 
Note that you can specficy to use 2 active electrons and set all but one orbital as inactive (or use another set of parameters to only get one CSF). This renders to a single configuration wavefunction and hence a HF calculation is done!
 
Note that if the RAS1 and RAS3 subspaces are left empty (this is the default) a CASSCF calculation is actually performed.
 
See:
MOLCAS RASSCF
Totorial with input examples and explanation
http://www.teokem.lu.se/molcas/documentation/manual/node66.html
 
http://info.iqm.unicamp.br/manuais/tutorials/node9.html
Back to top
 
« Last Edit: 21.09.11 at 15:29:37 by Gerrit-Jan Linker »  

Gerrit-Jan Linker
Linker IT Software
Email WWW Gerrit-Jan Linker   IP Logged
Gerrit-Jan Linker
YaBB Administrator
*****




Posts: 75
Re: RASSCF
Reply #2 - 29.08.08 at 21:15:07
 
RASSCF output
 
In the RASSCF output a printout is given of the wavefunction.
The occupation of active orbitals is given together with the spin coupling of the open shells. In this printout (see example below) the following notations are used:
 
2: fully occupied orbital
0: unoccupied orbital
u: singly occupied orbital. Electron has spin up
d: singly occupied orbital. Electron has spin down
 
Example output:
 
     conf/sym  111111     Coeff  Weight
            1  222220  -0.06035 0.00364
           10  2u222d   0.31147 0.09701
           11  2u22d2  -0.05097 0.00260
           12  2u2d22  -0.14250 0.02031
           13  u2222d   0.11921 0.01421
           15  u22d22  -0.05454 0.00297
           17  2ud222  -0.07971 0.00635
           19  202222  -0.80371 0.64595
           20  ud2222  -0.43503 0.18925
           21  022222  -0.11774 0.01386
 
There are a few important CSF's in this CI wavefunction. CSF nr 19 which contibutes 64.5%, SCF nr 20 which contributes 18.9% and CSF nr 10 which contributes 9.7%.
Back to top
 
« Last Edit: 29.08.08 at 21:15:40 by Gerrit-Jan Linker »  

Gerrit-Jan Linker
Linker IT Software
Email WWW Gerrit-Jan Linker   IP Logged
Gerrit-Jan Linker
YaBB Administrator
*****




Posts: 75
RASSCF: Specify CI roots
Reply #3 - 18.09.08 at 15:55:31
 
RASSCF: Specify CI roots
 
The CIROOT keyword can be used in the RASSCF input to specify the CI roots. This input makes it possible to perform orbital optimisations for the average energy of a number of states.
 
Syntax:
 
CIROOT
NROOTS LROOTS IALL
ROOT1 ROOT2 ROOT3 ...
WEIGHT1 WEIGHT2 WEIGHT3 ...
 
The default input is not to specify the CIROOT keyword at all. Only one CI root is calculated, the lowest.
 
NROOTS is the number of roots over which the average is to be taken.
LROOTS is the size of the CI matrix in the Davidson procedure. LROOTS must be as least as large as the index of the highest roow in row 3 of the input.
IALL is a flag to specify whether more input is given in row 3 to indicate over which roots the average is to be taken.
 
A simple average calculation over 3 roots:
CIROOT
3 3 1
 
In fact the IALL of 1 is optional here.
 
For an average over 5 roots namely roots 1,2,5, 6 and 7 the input is:
CIROOT
5 7
1 2 5 6 7
 
It is possible to give weights to the roots. In the next example root 1 is weighted 60% and the other roots 10%.  
 
CIROOT
3 7
1 2 5 6 7
6 1 1 1 1
 
To calculate the second root only (no average) you need to specify:
 
CIROOT
1 2
2
 
Orbitals can be optimised for one root but the energy of other roots can be calculated too (with those orbitals). In the following example I calculate 3 roots but optimise for the first only:
CIROOT
1 3
1
1
Back to top
 
« Last Edit: 10.12.10 at 09:38:25 by Gerrit-Jan Linker »  

Gerrit-Jan Linker
Linker IT Software
Email WWW Gerrit-Jan Linker   IP Logged
Gerrit-Jan Linker
YaBB Administrator
*****




Posts: 75
Divergence in the RASSCF energy
Reply #4 - 28.09.08 at 10:40:01
 
Divergence in the RASSCF energy
 
      Quote:

=================================
!!! Warning: Divergence in the RASSCF energy !!!
iteration  81
RASSCF energy   -41.9947585793
energy difference 4.3591558588
=================================

!!! The program was enforced to stop !!!

 
When divergence in the RASSCF energy occurs it is worth trying different input orbitals. If you started with guess orbitals you may want to do a calculation obtaining input vectors and use these as input vectors in other calculations.
 
Also using a level shifter may help to overcome this problem.
Back to top
 
« Last Edit: 17.09.15 at 12:29:43 by Gerrit-Jan Linker »  

Gerrit-Jan Linker
Linker IT Software
Email WWW Gerrit-Jan Linker   IP Logged
Gerrit-Jan Linker
YaBB Administrator
*****




Posts: 75
LEVShift in RASSCF
Reply #5 - 21.08.09 at 20:14:55
 
LEVShift in RASSCF
 
If the RASSCF job does not converge you can use a levelshift to keep the levels artificial apart. A levelshift of 0.5, 1.0 or 1.5 hartree should be enough.  
 
Command:
levshift
1.0
 
Example of a non converging job:
 
     Iter CI   SX   CI       RASSCF       Energy    max BLB   max BLB  max ROT   Level Ln srch  Step   QN    Time(min)
         iter iter root      energy       change    element    value    param    shift minimum  type update     CPU
      86   5   42    3   -46.24192005    0.71E-02*  59  67 1  0.12E-01* -0.10E-01*  0.46   0.07     LS    YES      5.72
      87   4  100    3   -46.24765088   -0.57E-02*  57 150 1  0.18E-01* -0.46E-01*  0.43   1.49     QN    YES      5.83
      88   6  100    3   -46.24169397    0.60E-02*  57 150 1  0.16E-01* -0.20E-01*  0.39   0.23     LS    YES      6.10
      89   5   89    3   -46.24617131   -0.45E-02*  69 477 1  0.96E-02*  0.60E-01*  0.41   1.89     QN    YES      5.90
      90   6  100    3   -46.24698146   -0.81E-03*  63 169 1  0.92E-02* -0.28E-01*  0.45   0.64     QN    YES      6.10
      91   6   43    3   -46.24864574   -0.17E-02*  56 100 1  0.17E-01*  0.23E-01*  0.39   0.69     QN    YES      5.86
      92   4   49    3   -46.24990980   -0.13E-02*  63 169 1 -0.46E-02*  0.26E-01*  0.41   1.13     QN    YES      5.60
      93   5   67    3   -46.25040417   -0.49E-03*  63 169 1 -0.50E-02*  0.25E-01*  0.40   1.80     SX    YES      5.80
      94   6  100    3   -46.24779107    0.26E-02*  59  67 1  0.13E-01*  0.14E-01*  0.46   0.17     SX    YES      6.10
      95   5   39    3   -46.25033411   -0.25E-02*  56 100 1  0.16E-01* -0.76E-01*  0.39   0.74     QN    YES      5.71
      96   5   66    3   -46.25030769    0.26E-04*  69 477 1  0.77E-02*  0.51E-01*  0.41   0.35     QN    YES      5.79
      97   9   41    3   -46.25244383   -0.21E-02*  69 477 1  0.83E-02*  0.15E+00*  0.42   2.50     QN    YES      6.28
      98   7  100    3   -46.24644073    0.60E-02*  59  67 1  0.13E-01*  0.26E-01*  0.47   0.21     SX    YES      6.24
      99   5   47    3   -46.25637790   -0.99E-02*  56 100 1  0.13E-01*  0.98E-01*  0.44   1.47     QN    YES      5.73
     100   5   54    3   -46.25484742    0.15E-02*  69 477 1  0.11E-01*  0.24E-01*  0.43   0.30     SX    YES      5.76
     101   6   34    3   -46.25910785   -0.43E-02*  59  67 1  0.55E-02* -0.34E-01*  0.46   0.77     QN    YES      5.84
     102   4   77    3   -46.25874818    0.36E-03*  57 150 1  0.10E-01* -0.15E-01*  0.44   0.31     SX    YES      5.71
Back to top
 
« Last Edit: 12.08.14 at 11:12:30 by Gerrit-Jan Linker »  

Gerrit-Jan Linker
Linker IT Software
Email WWW Gerrit-Jan Linker   IP Logged
Gerrit-Jan Linker
YaBB Administrator
*****




Posts: 75
RASSCF orbital files
Reply #6 - 26.08.09 at 07:59:28
 
RASSCF orbital files
 
RASSCF stores the orbitals in a set of files.
 
project.RasOrb : RASSCF orbitals. In case of a state average these are the orbitals for the average. The orbitals for the individual roots are named projet.RasOrb.n where n is the root number
 
project.SpdOrb : RASSCF spin orbitals.
 
You can use the RasOrb files as input vectors for a new calculation.
Back to top
 
 

Gerrit-Jan Linker
Linker IT Software
Email WWW Gerrit-Jan Linker   IP Logged
Gerrit-Jan Linker
YaBB Administrator
*****




Posts: 75
RASSCF: 1 active electron in degenerate HOMO
Reply #7 - 26.08.09 at 13:42:36
 
RASSCF: 1 active electron in degenerate HOMO
 
Be careful with 1 active electron in a degenerate HOMO. One of the degenerate highest orbitals will be favoured and this leads to a wrong set of orbitals. Better is to include all degenerate orbitals in the active space (RAS2) and to include all the electrons in them as active electrons. If you have an even number of electrons you can also specify zero active orbitals (RAS2=0) and zero active electrons. With one electron active a state average over the degenerate orbitals will be a solution.
Back to top
 
 

Gerrit-Jan Linker
Linker IT Software
Email WWW Gerrit-Jan Linker   IP Logged
Gerrit-Jan Linker
YaBB Administrator
*****




Posts: 75
RASSCF: parallel running
Reply #8 - 16.10.13 at 08:09:48
 
RASSCF: parallel running
 
A quick analysis of parallel processing with RASSCF was performed. A small RAS space was used (RAS1=1, RAS3=4). Timings in the attached plot are per RASSCF iteration. One to four 2.6 GHz cores were used on 12 core nodes.
 
Quick conclusion: run with 3 or 2 cores is most efficient.
Back to top
« Last Edit: 16.10.13 at 08:13:30 by Gerrit-Jan Linker »  

rasscf_molcas78.png

Gerrit-Jan Linker
Linker IT Software
Email WWW Gerrit-Jan Linker   IP Logged
Gerrit-Jan Linker
YaBB Administrator
*****




Posts: 75
RASSCF: geometry optimisation
Reply #9 - 22.12.14 at 22:37:59
 
RASSCF: geometry optimisation
 
The RLXROOT keyword can be used to specify which root to relax the geometry to.
Back to top
 
 

Gerrit-Jan Linker
Linker IT Software
Email WWW Gerrit-Jan Linker   IP Logged
Gerrit-Jan Linker
YaBB Administrator
*****




Posts: 75
RASSCF : compact display of orbitals
Reply #10 - 17.09.15 at 12:31:43
 
RASSCF : compact display of orbitals
 
For basic analysis of the orbitals a compact display is often more useful than the detailed display.  
To force the compact display, the following input option can be used in rasscf:
 
ORBAPPEAR
compact
Back to top
 
 

Gerrit-Jan Linker
Linker IT Software
Email WWW Gerrit-Jan Linker   IP Logged
Pages: 1