X2Cmmf question

Peterson, Kirk

unread,

Feb 20, 2021, 6:06:52 PM2/20/21

to dirac...@googlegroups.com

Dear Dirac experts,

is there any reason why an X2C-mmf CCSD calculation is not appropriate for an atom (even though the name is a bit inconsistent) ? I'm attempting this for Th atom but I note that it takes the DHF energy from the DFCOEF file which in this case is the average-of-configuration DHF energy, which is quite different from the lowest state from GOSCIP (which is used in a 4c-CCSD calculation). The CCSD then gives me an error like:

* REACMO: Coefficients read from file DFCOEF - Total energy: -26465.5416400058093

* Heading : X2Cmod - molecular-mean field 2c MOs Sat Feb 20 14:39:38 2021

SCF energy : -26465.541640005809313

II,JJ,AA,BB= 9 2 176 67

eps_II= -0.20841911691406145 eps_JJ= -0.20841911691406145 eps_AA= -0.21691291716792749 eps_BB= -0.21691291716792749

eII+eJJ-eAA-eBB= 1.6987600507732070E-002 ATVAL= 1.7033063894329577E-002 DHOLU= 9.9999999999999995E-007

Slave node 3 : --- SEVERE ERROR, PROGRAM WILL BE ABORTED ---

Date and time (Linux) : Sat Feb 20 14:41:06 2021

Zero or negative HOMO-LUMO gap

II,JJ,AA,BB= 9 2 176 67

eps_II= -0.20841911691406145 eps_JJ= -0.20841911691406145 eps_AA= -0.21691291716792749 eps_BB= -0.21691291716792749

eII+eJJ-eAA-eBB= 1.6987600507732070E-002 ATVAL= 1.7033063894329577E-002 DHOLU= 9.9999999999999995E-007

The input file is identical to my 4c calculation except that I just added .X2Cmmf under the Hamiltonian section. The calculation fails in the same way for .SPINFREE and for the DC (G) Hamiltonians.

My (SF) inp file is given below.

regards, -Kirk

**DIRAC

.TITLE

Th atom

.WAVE FUNCTION

.ANALYZE

**ANALYZE

.MULPOP

*MULPOP

.VECPOP

1..oo

.LABEL

SHELL

**HAMILTONIAN

.X2Cmmf

.SPINFREE

**INTEGRALS

*READIN

.UNCONTRACT

**WAVE FUNCTION

.SCF

.RESOLVE

.RELCCSD

*SCF

.CLOSED SHELL

44 44

.OPEN SHELL

1

2/10,0

.ERGCNV

1.d-8

.LSHIFT

5.0

.OLEVEL

1.0

.MAXITR

100

**MOLTRA

.ACTIVE

energy -2.5 20.0 0.5

**RELCC

.ENERGY

.PRINT

1

.NEL_F1

4 1 1 0 1 0 0 0 2 1 1 0 1 0 0 0

*CCENER

.MAXIT

250

.NTOL

7

*END OF

Johann Pototschnig

unread,

Feb 21, 2021, 7:56:18 PM2/21/21

to dirac-users

Hi Kirk,

The molecular mean field contributions are added at the SCF level.

Since the Fock matrix is not recomputed and the orbital energies are used to represent the Fock matrix,

it can happen if you select a specific occupation that virtual orbital energies are lower in energy than occupied ones.

The code checks for this and the computation is stopped if it happens.

This happens in open-shell molecules as well as atoms.

- X2C should be possible, you might need to activate the recomputation of the Fock matrix.

- It could be possible to select an occupation of the orbitals, that the HOMO-LUMO gap is positive,

but even in this case I would expect difficult convergence.

- The check for negative HOMO-LUMO gap could be deactivated, but I don't expect convergence in coupled cluster.

I computed a few open shell systems where the HOMO-LUMO gap was small but still positive and observed very slow convergence.

- A recomputation of the X2Cmmf Fock matrix should work, but is not implemented.

best wishes,

Johann

Peterson, Kirk

unread,

Feb 22, 2021, 11:06:43 AM2/22/21

to dirac...@googlegroups.com

Hi Johann,

thanks for the details below. This certainly makes sense. Too bad the recomputation of the X2Cmmf Fock matrix is not supported (yet) but it's good to know why it fails.

best wishes, -Kirk

--
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/f04c1a37-61c9-4cf2-b02e-eba30365b34en%40googlegroups.com.

Reply all

Reply to author

Forward