Q-Chem fatal error occurred in module setman/rpa_dia.C, line 59: Unable to diagonalize A-B

Questions
#1

Hi, I am know that there are similar topics were opened earlier but those answers did not help…Here is my input:

$molecule
read spp.txt
$end

$XC_FUNCTIONAL
C PBE 1.00 # Standart correlation
X wPBE 0.09 # beta=-alpha+1/eps
X PBE 0.71 # 1-1/eps
X HF 0.20 # alpha=Short-range HF
$end

$rem
jobtype sp # sinle point calculation
exchange general # XC_FUNCTIONAL
correlation none
basis 6-31G
omega 170 # tuned switching parameter
omega2 170 # tuned switching parameter
lrc_dft true # include longe-rage corr.
src_dft 2 # include short-rage corr.
HF_SR 200 # 1000xHF
HF_LR 290 # 1000x(1/eps)
scf_convergence 8
rpa 2 # Random Phase approx.
cis_n_roots 25 # How many roots you are asking
cis_singlets true # Only single excitations
cis_triplets false
max_cis_cycles 5000
pop_mulliken -1 # Print charges on atoms
max_scf_cycles 5000
mem_static 4000
mem_total 48000
print_general_basis true
print_orbitals 10
molden_format true
sts_fcd true # Print electronic coupling for dimer
sts_donor 1-164 # Donor atoms indexing for dimers
sts_acceptor 165-331 # Acceptor atoms indexing for dimers
sts_becke_partition 1
sts_multi_nroots 25 # Same as cis roots
sts_multi_type 1
solvent_method pcm # Polarizable continuum model (PCM)
SYM_IGNORE TRUE
CIS_DYNAMIC_MEM TRUE
CIS_RELAXED_DENSITY TRUE
USE_NEW_FUNCTIONAL TRUE
$end

$pcm
NonEquilibrium
Theory iefpcm
StateSpecific Perturb
$end

$solvent
Dielectric 3.5
$end

What I got is the following

Iter Rts Conv Rts Left Ttl Dev Max Dev

Q-Chem fatal error occurred in module setman/rpa_dia.C, line 59:

Unable to diagonalize A-B

Please submit a crash report at Q-Chem Crash Reporter

When I use rpa as 0 it stuck here

Iter Rts Conv Rts Left Ttl Dev Max Dev

no error messages…

I have 413 atoms in moleculer complex…

What is the issue do you think?

Thanks

#2

I have to add a note: When I try TDDFT with 331 atoms it went perfect—it is done.

Above erros messages in first post was from molecule which has 413 atoms which has the same core molecules as 331 but I added some site pigments as an environment which are only optimized with UFF and added to core 331 so total 413 atoms.

Do you think this can cause a diagonalize problem in the matrix?

Aksu

#3

It’s not about system size (this can happen in small molecules as well), it’s about the matrix A-B failing to be positive definite, which is incompatible with the diagonalization algorithm. This can be a sign of a triplet instability or I’ve also seen it happen at geometries that pass stability analysis but are near to a region of an instability on the potential surface. The theoretical background is discussed here,
https://doi.org/10.1016/B978-0-323-91738-4.00005-1
There’s nothing you can do except to invoke the Tamm-Dancoff approximation (set RPA=0), which decouples the excitation energy problem from the stability problem.

#4

Dear Prof. Herbert, Thanks for the reply.
However, as I wrote above I tried RPA 0 but it stopped at the stage of ‘‘Iter Rts Conv Rts Left Ttl Dev Max Dev’’ stage and there was no error…

Aksu

#5

That sounds like a memory problem. Can you:

  • Post a bit more of the output in the TDA case. (In particular, there might be a statement about memory usage just prior to the start of TDDFT iterations, depending on which version you are using.)

  • let me know what Q-Chem version it is

  • Try setting MEM_TOTAL to the physical memory available on your hardware.

#6

Yes it is indeed, I did not change the total memory but I cleaned up my disk.
It went successfull now with rpa set to 0.

(Usually it does not use the all memory that qchem asks.)

Thanks

Aksu

#7

The maximum memory is a conservative estimate based on the largest possible number of subspace vectors, which is in turn dictated by CIS_N_ROOTS and the maximum number of CIS iterations. You can get a better estimate by looking at the memory per subspace vector and then counting subspace vectors iteration by iteration. This paper,
https://doi.org/10.1021/acs.jctc.3c00673
discusses counting subspace vectors.

#8

Thanks for the educational discuss.
I note those paper you have pointed out.

Aksu

#9

In case, if other users of Qchem encounter still the diagonalize problem I am happy to say that I have solved my own issue. As I mentioned above, I have a core 331 atoms at the core which were optimized with Qchem and then I added environmental molecules but they were optimized UFF in Avogadro…That was the problem. I have no idea how UFF effect on the molecule geometry, this time I have optimized them only H atoms in Qchem then I added them to core molecule…TDDFT with RPA 2 works.
I insisted to use RPA 2 since RPA 0 (TDA) does not give satisfiying results for our system, espacially for the bright and dark statess.