GASCI hamiltonian non-hermicity

[Jakub Visnak]

Dear Dirac Experts,

I attempted GAS-CI calculation for non-symmetric geometry of UO2(H2O)5(2+) (*), with SORECP and def-TZVPP basis. The output (attached) ends with

"(GASCIP_RCISTD) Anti-hermiticity of reduced Hamiltonian (real,imag) 3.821D+00 3.036D+00

 FATAL ERROR, this is greater than threshold 1.000D-09"

Aside to the attached CAS(6,7), I have tried CAS(6,9) and CAS(12,12)  (thank you very much Kirk, the fall was due to the memory requirements, with  --mw=12000 --aw=15000 and 125GB in PBS script it works.) 

Except it subsequently falls on abov mentioned anti-hermicity after the first CI iteration, unfortunately (as in CAS(6,7) and CAS(6,9) cases).

I have tried to disable the two-electron integral screening as I supposed that the anti-hermicity must be some numerical artifact (?), but it didn't work. What should I do, please?

For D5h geometry (represented as C2v **) GAS-CI calculation worked for me (with different orbital spaces, more general than CAS, however)

Best regards,

Jakub Visnak

(*)  a cut from spin-free lowest-lying excited state geometry optimization started from a CMD snaphost geometry, presence of point-charges in optimization part gives C1 point group geometry

** from a ground state spin-free ECP/B3LYP geometry optimization in Vacuo

[Nike Dattani]

Dear Jakub,

Interesting. Have you had success with a smaller basis set?

With best wishes,

Nike

--
You received this message because you are subscribed to the Google Groups "dirac-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to dirac-users+unsubscribe@googlegroups.com.
Visit this group at https://groups.google.com/group/dirac-users.
For more options, visit https://groups.google.com/d/optout.

[Visscher, L.]

Dear Jakun,

This looks like a bug that one of the CI developers should have a look at. I'll post this also on our developer list and see whether somebody has time to take a closer look. I'm insufficiently familiar with the GASCIP module to say offhand what is causing the problem and whether there is a work-around for the current release.

best reagrds,

Luuk

--
You received this message because you are subscribed to the Google Groups "dirac-users" group.

To unsubscribe from this group and stop receiving emails from it, send an email to dirac-users...@googlegroups.com.

Visit this group at https://groups.google.com/group/dirac-users.
For more options, visit https://groups.google.com/d/optout.

<HF_UO2w5dz.out>

[jvi...@gmail.com]

Dear  Dirac Experts, 

now I have faced the same error message for C1* geometry of NUHFI molecule

 computationals: double-zeta basis set,  4c-CASCI(4,4)   (with GASCIP, as LUCIAREL is not implemented yet for C1*)

The error is " Master node : --- SEVERE ERROR, PROGRAM WILL BE ABORTED ---

 Date and time (Linux) : Sun Apr 25 13:59:33 2021

 Anti-hermiticity of red. Ham. > threshold

 Traceback from this point (gascip_rcistd_+0x3b81): 7 frames.

  6   function name:            gascip_rcistd_+0x3b81

  5   function name:                     rcist_+0x2ad

  4   function name:                   roptst_+0x1ef2

  3   function name:                 krci_calc_+0xade

  2   function name:                     dirac_+0xec4

  1   function name:                     MAIN__+0x7e8

  0   function name:                        main+0x2e"

what should be the work-around, please?

Do you think that FCIDUMP/MO_integrals might be affected by the same bug for the C1* geometries? 

Because the integrals we obtained does not lead to reasonable DMRG energies (lower than reasonable by several orders of magnitude).

Best regards,

Jakub

P.S.: Dear Luuk, did you obtain any information from the Developers on this topic, please?

P.S.II:  This molecule has been already investigated in an article  [1], including 4c-CCSD. 

[1] Wormit M., Olejniczak M., Deppenmeier A-L., Borschevsky A., Saue T., Schwerdtfeger: Strong enhancement of parity violation effects in chiral uranium compounds, Phys. Chem. Chem. Phys., 2014, 16, 17043, DOI:10.1039/c4cp01904k

[jvi...@gmail.com]

For some reasons, attaching files does not work for me here, so I will cut out inputs for SCF run:

inp.inp:

"**DIRAC

.TITLE

 NUHFI molecule, SCF try, DZ basis, 81 electronic pairs total, 13 valence pairs (including two pi)

.WAVE FUNCTION

.ANALYZE

**INTEGRALS

#*TWOINT

#.SCREEN

#-1.0D-14

*READIN

#.UNCONTRACT

**HAMILTONIAN

#.DOSSSS

**WAVE FUNCTION

.SCF

.RESOLVE

.KR CI

#.MP2 NO

*SCF

.ERGCNV

 5.D-11

.MAXITR

 800

.LSHIFT

 0.8

.2NDOPT

*RESOLVE

*KRMCSCF

.INACTIVE

75

.GAS SHELLS

1

12 12 / 12

.CI PROGRAM

LUCIAREL

.MAX MACRO

 200

*KRCICALC

.CI PROGRAM

GASCIP

#.FCIDUMP

.CIROOTS

 1 4

.MAX CI

 1000

.INACTIVE

 75

.GAS SHELLS

 1

 12 12 / 12

.MXCIVE

 20

#.OMEGAQ

.ANALYZ

#*MP2 NO

#.ACTIVE

#6..150

#.MULPOP

#active

**MOLECULE

**GENERAL

.ACMOUT

.PCMOUT

**ANALYZE

.PRIVEC

.MULPOP

#.PROJECTION

*PRIVEC

.VECPRI

all

.PRICMP

 1 1

*MULPOP

.VECPOP

all

#*PROJECTION

#.VECPRJ

#all

##.VECREF

##2

##AFTHXX

##1

##all

##AFCOXX

##2

##all

##.OWNBAS

#.ATOMS

#AFTHCX

#all

#AFCXXX

#all

#AFOXXX

#all

##.PROTHR

##-1.0D-4

#.POLREF

*END OF INPUT

"

mol.mol:

"DIRAC

NUHFI, C1 geometry (the second zero), ca. equilibrium geometry

there are 5 kinds of atoms (N,U,H,F,I), 0 total charge (the first zero), Angstroem, "A" should be on the exact 20th position

C   5  0 0         A

        7.    1

N     0.5255468   0.2608732    2.0653705

LARGE BASIS cc-pVDZ

       92.    1

U    -0.0303969  -0.0391832    0.4766824

LARGE BASIS dyall.v2z

        1.    1

H     1.5796088  -0.1407848   -0.6453847

LARGE BASIS cc-pVDZ

        9.    1

F    -0.9339928  -1.1591900   -0.9135786

LARGE BASIS cc-pVDZ

       53.    1

I    -0.7699855   2.4435462   -0.8140155

LARGE BASIS dyall.v2z

FINISH

"

directive (Dirac 17):

pam-dirac --scratch=$SCRATCHDIR/Dscr --noarch --gb=47 --ag=48 --inp=inp.inp --mol=mol.mol --outcmo --get="DFCOEF DFPCMO DFACMO"

and separate CI(4,4) run done after it (starting from the converged DHF orbitals):

inp.inp:

"**DIRAC

.TITLE

 NUHFI molecule, SCF try, DZ basis, 81 electronic pairs total, 13 valence pairs (including two pi)

.WAVE FUNCTION

.ANALYZE

**INTEGRALS

#*TWOINT

#.SCREEN

#-1.0D-14

*READIN

#.UNCONTRACT

**HAMILTONIAN

#.DOSSSS

**WAVE FUNCTION

#.SCF

#.RESOLVE

.KR CI

#*SCF

#.ERGCNV

# 5.D-11

#.MAXITR

# 800

#.LSHIFT

# 0.8

#.2NDOPT

#*RESOLVE

*KRCICALC

.CI PROGRAM

GASCIP

#.FCIDUMP

.CIROOTS

 1 4

.MAX CI

 1000

.INACTIVE

 79

.GAS SHELLS

 1

 4 4 / 4

.MXCIVE

 20

#.OMEGAQ

.ANALYZ

#*MP2 NO

#.ACTIVE

#6..150

#.MULPOP

#active

**MOLECULE

**GENERAL

.ACMOUT

.PCMOUT

**ANALYZE

.PRIVEC

.MULPOP

#.PROJECTION

*PRIVEC

.VECPRI

all

.PRICMP

 1 1

*MULPOP

.VECPOP

all

#*PROJECTION

#.VECPRJ

#all

##.VECREF

##2

##AFTHXX

##1

##all

##AFCOXX

##2

##all

##.OWNBAS

#.ATOMS

#AFTHCX

#all

#AFCXXX

#all

#AFOXXX

#all

##.PROTHR

##-1.0D-4

#.POLREF

*END OF INPUT

"

mol.mol:

the same

directive:

pam-dirac --scratch=$SCRATCHDIR/Dscr --noarch --gb=47 --ag=48 --inp=inp.inp --put="DFCOEF DFPCMO DFACMO" --incmo --mol=mol.mol --outcmo --get="DFCOEF DFPCMO DFACMO"

[Knecht Stefan]

Dear Jakub,

thanks for the inputs. I am currently looking into this issue but even with the DZ basis sets the calculations take some time. 

For the ground state I am getting a reference KRCI energy for the CAS(4,4) case of 

-35329.57894973 Hartree 

Does that make sense to you?

with best regards

Stefan

--
You received this message because you are subscribed to the Google Groups "dirac-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to dirac-users...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/dirac-users/5476b605-f790-4447-91c3-d093cecbad9bn%40googlegroups.com.

--

---------------------------------------------------------------------------
PD Dr. Stefan Knecht
---------------------------------------------------------------------------
GSI Helmholtzzentrum für Schwerionenforschung GmbH
Abteilung SHE Chemie
Planckstr. 1
64291 Darmstadt
Germany

room: SB3 3.163
phone: +49 6159 71 2588
web: http://stefanknecht.xyz
email: s.kn...@gsi.de
---------------------------------------------------------------------------
GSI Helmholtzzentrum für Schwerionenforschung GmbH
Planckstraße 1, 64291 Darmstadt, Germany, www.gsi.de

Commercial Register / Handelsregister: Amtsgericht Darmstadt, HRB 1528
Managing Directors / Geschäftsführung:
Professor Dr. Paolo Giubellino, Dr. Ulrich Breuer, Jörg Blaurock
Chairman of the GSI Supervisory Board / Vorsitzender des GSI-Aufsichtsrats:
Ministerialdirigent Dr. Volkmar Dietz
---------------------------------------------------------------------------
Johannes-Gutenberg Universität Mainz
Departement Chemie

Duesbergweg 10-14
55128 Mainz

Germany

---------------------------------------------------------------------------

[Knecht Stefan]

Dear Jakub,

could you please try the attached FCIDUMP file? My DMRG reference energy is in this case

-35329.5789201008 Hartree

The SCF energy is

-35329.578892973397 Hartree

Note that the SCF/FCI energy is different from the one of my previous mail since I had to run the SCF on my DMRG development branch which still uses an older CODATA set 

for the physical constants (speed of light, …) than the current master/release branch of Dirac. 

with best regards

Stefan

p.s.: as to why the GASCIP sub-space Hamiltonian gets non-Hermitian, I can’t say yet.

To view this discussion on the web visit https://groups.google.com/d/msgid/dirac-users/85EB5C65-F19D-4548-ADA7-8811E4FDD003%40ethz.ch.