Molcas Forum

Support and discussions for Molcas and OpenMolcas users and developers

You are not logged in.

Announcement

Welcome to the Molcas forum. You can choose an avatar and change the default style by going to "Profile" → "Personality" or "Display".

#1 2021-06-29 15:34:41

shosuro
Member
Registered: 2021-06-29
Posts: 1

DMRGSCF and DMRG energies differ / Convergence problems

Good day everyone!

We were forced to migrate on a new cluster system and since that day we have severe problems using DMRG and QCMacquis. We already solved some compiler problems and problems with OpenBlas and we managed to link mkl libraries, so we were very happy that the &RASSCF module did not quit with a rc=-6 error. The latest version is based on the following commands:

  • module load gcc/8.3.1 boost/1.73.0 hdf5/1.8.17 gsl/2.6  git/2.31.0 cmake/3.17.1

  • source /shared/apps/intel/2020.1/mkl/bin/mklvars.sh intel64

  • git clone https://gitlab.com/Molcas/OpenMolcas.git

  • cmake -DCMAKE_Fortran_COMPILER=gfortran -DCMAKE_C_COMPILER=gcc -DCMAKE_CXX_COMPILER=g++ -DDMRG=ON -DOPENMP=ON -DLINALG="MKL" -DHDF5=ON /path/to/openmolcas/source

  • At the following location «Build files have been written to...» the make command executed once again

  • an executable version of pymolcas is made

However, now we are facing the problem that apparently the DMRGSCF and DMRG energies differ and no convergence can be achieved while this was possible for the very same system on the previous cluster.

&GATEWAY
Title
Test
RICD
AMFI
COORD
structure5.xyz
BASIS
Atom1.ANO-RCC-VDZ, Atom2.ANO-RCC-VDZ, Atom3.ANO-RCC-MB, Atom4.ANO-RCC-MB
Group
NoSym

&SEWARD
RELA
R02O

>>> COPY structure5.RasOrb INPORB

&RASSCF
LUMORB
TYPEINDEX
NActEl=18
Ras2=16
Charge=+4
Spin=1
CIRoot=1 1 1
DMRG
RGInput
  max_bond_dimension=500
  nsweeps=5
EndRG

In the .log file we find in the wave function control section:

      Iter num  DMRG max tr DMRG  SX      DMRGSCF       Energy    max ROT   max BLB     max BLB  Level Ln srch  Step   QN     CPU Time
         sweeps/root weight/root iter     energy        change     param    element      value   shift minimum  type update   hh:mm:ss
      Total energies have been shifted. Add                -2000.00 au
        1  5   1    0.21E-01   1   18  -912.42366593    0.00E+00 -0.16E-01   69 113 1  0.31E-01*  0.36   0.00    SX    NO      0:15:28
        2  5   1    0.20E-01   1   17  -912.43026670   -0.66E-02*-0.14E-01   63 111 1  0.97E-02*  0.37   0.00    SX    NO      0:08:12
        3  5   1    0.20E-01   1   17  -912.42389868    0.64E-02*-0.92E-02   63 111 1  0.61E-02*  0.36   0.00    SX    NO      0:08:12
      ========================================================================================================================
 ###############################################################################
 ###############################################################################
 ###                                                                         ###
 ###                                                                         ###
 ###    ERROR: DMRGSCF and DMRG energies differ.                             ###
 ###                                                                         ###
 ###                                                                         ###
 ###############################################################################
 ###############################################################################
      iteration                     3
      DMRGSCF energy            -2912.4238986826
      DMRG energy               -2912.4323941599
      relative difference           0.0000029170
 About this difference:
 1) If possible, consider a larger M value
 2) Severe convergence problems. Maybe active
    space is unsuitable for this system?
      ========================================================================================================================
        4  5   1    0.20E-01   1   16  -912.43264114   -0.87E-02*-0.65E-02   63 111 1  0.34E-02*  0.36   0.00    SX    NO      0:08:07
        5  5   1    0.20E-01   1   16  -912.43285061   -0.21E-03*-0.13E-01   63 111 1  0.23E-02*  0.36   1.08    QN   YES      0:07:47
        6  5   1    0.20E-01   1   15  -912.43311885   -0.27E-03* 0.33E-02   67 117 1 -0.10E-02*  0.36   0.98    QN   YES      0:07:29
        7  5   1    0.20E-01   1   14  -912.43215538    0.96E-03* 0.23E-02   26 132 1 -0.81E-03*  0.36   0.00    LS   YES      0:07:28
        8  5   1    0.20E-01   1   15  -912.42974460    0.24E-02* 0.16E-02   68 126 1 -0.11E-02*  0.36  -0.00    LS   YES      0:07:29
      ========================================================================================================================
 ###############################################################################
 ###############################################################################
 ###                                                                         ###
 ###                                                                         ###
 ###    ERROR: DMRGSCF and DMRG energies differ.                             ###
 ###                                                                         ###
 ###                                                                         ###
 ###############################################################################
 ###############################################################################
      iteration                     8
      DMRGSCF energy            -2912.4297445952
      DMRG energy               -2912.4330686705
      relative difference           0.0000011413
 About this difference:
 1) If possible, consider a larger M value
 2) Severe convergence problems. Maybe active
    space is unsuitable for this system?

For another system we do not find that the DMRGSCF and DMRG energies differ, but we still cannot achieve convergence even for a (2,2) CAS.

&GATEWAY
Title = Monomer
AMFI
RICD
COORD = Monomer.xyz
BASIS = ANO-RCC-MB
GROUP = NOSYM

&SEWARD
RELA
R02O

&RASSCF
LUMORB
TYPEINDEX
DMRG
 RGINPUT
  max_bond_dimension=1000
  nsweeps=5
 ENDRG
SPIN = 1
RAS2 = 2 0 0 0
NACTEL = 2 0 0
INACTIVE = 169 0 0 0
CIROOT = 1 1 1
Stop Module: gateway at Mon Jun 28 14:41:50 2021 /rc=_RC_ALL_IS_WELL_
...
Stop Module: seward at Mon Jun 28 14:46:01 2021 /rc=_RC_ALL_IS_WELL_
...
      180  3   1    0.00E+00   1   24  -657.00034769   -0.18E-02* 0.17E-01  149 181 1  0.30E-01*  0.04   0.59    SX   YES      0:00:20
      181  2   1    0.00E+00   1   18  -656.99856725    0.18E-02* 0.17E-01  149 181 1 -0.45E-01*  0.03   0.41    SX   YES      0:00:20
      182  3   1    0.00E+00   1   24  -657.00034767   -0.18E-02* 0.17E-01  149 181 1  0.30E-01*  0.04   0.59    SX   YES      0:00:20
      183  2   1    0.00E+00   1   18  -656.99856694    0.18E-02* 0.17E-01  149 181 1 -0.45E-01*  0.03   0.41    SX   YES      0:00:20
      184  2   1    0.00E+00   1   24  -657.00034767   -0.18E-02* 0.17E-01  149 181 1  0.30E-01*  0.04   0.59    SX   YES      0:00:20
      185  3   1    0.00E+00   1   18  -656.99856663    0.18E-02* 0.17E-01  149 181 1 -0.45E-01*  0.03   0.41    SX   YES      0:00:20
      186  3   1    0.00E+00   1   24  -657.00034767   -0.18E-02* 0.17E-01  149 181 1  0.30E-01*  0.04   0.59    SX   YES      0:00:20
      187  2   1    0.00E+00   1   18  -656.99856630    0.18E-02* 0.17E-01  149 181 1 -0.45E-01*  0.03   0.41    SX   YES      0:00:20
      188  2   1    0.00E+00   1   24  -657.00034764   -0.18E-02* 0.17E-01  149 181 1  0.30E-01*  0.04   0.59    SX   YES      0:00:20
      189  2   1    0.00E+00   1   18  -656.99856599    0.18E-02* 0.17E-01  149 181 1 -0.45E-01*  0.03   0.41    SX   YES      0:00:20
      190  2   1    0.00E+00   1   24  -657.00034765   -0.18E-02* 0.17E-01  149 181 1  0.30E-01*  0.04   0.59    SX   YES      0:00:20
      191  2   1    0.00E+00   1   18  -656.99856569    0.18E-02* 0.17E-01  149 181 1 -0.45E-01*  0.03   0.41    SX   YES      0:00:20
      192  2   1    0.00E+00   1   24  -657.00034764   -0.18E-02* 0.17E-01  149 181 1  0.30E-01*  0.04   0.59    SX   YES      0:00:20
      193  2   1    0.00E+00   1   18  -656.99856536    0.18E-02* 0.17E-01  149 181 1 -0.45E-01*  0.03   0.41    SX   YES      0:00:20
      194  2   1    0.00E+00   1   24  -657.00034759   -0.18E-02* 0.17E-01  149 181 1  0.30E-01*  0.04   0.59    SX   YES      0:00:20
      195  2   1    0.00E+00   1   18  -656.99856505    0.18E-02* 0.17E-01  149 181 1 -0.45E-01*  0.03   0.41    SX   YES      0:00:20
      196  2   1    0.00E+00   1   24  -657.00034756   -0.18E-02* 0.17E-01  149 181 1  0.30E-01*  0.04   0.59    SX   YES      0:00:20
      197  2   1    0.00E+00   1   18  -656.99856473    0.18E-02* 0.17E-01  149 181 1 -0.45E-01*  0.03   0.41    SX   YES      0:00:20
      198  2   1    0.00E+00   1   24  -657.00034757   -0.18E-02* 0.17E-01  149 181 1  0.30E-01*  0.04   0.59    SX   YES      0:00:20
      199  2   1    0.00E+00   1   18  -656.99856444    0.18E-02* 0.17E-01  149 181 1 -0.45E-01*  0.03   0.41    SX   YES      0:00:20
      200  3   1    0.00E+00   1   24  -657.00034755   -0.18E-02* 0.17E-01  149 181 1  0.30E-01*  0.04   0.59    SX   YES      0:00:20
      No convergence after200 iterations
      201  2   1    0.00E+00   1   24  -656.99856422    0.18E-02* 0.17E-01  149 181 1 -0.45E-01*  0.04   0.59    SX   YES      0:00:20

--- Stop Module: rasscf at Mon Jun 28 15:56:11 2021 /rc=_RC_NOT_CONVERGED_ ---
.#######################.
.# Convergence problem #.
.#######################.

We would be entirely grateful for any ideas where the problems are coming from and in the best case how to solve them.

Offline

Board footer

Powered by FluxBB

Last refresh: Today 19:27:58