iOpenShell » Q-Chem specific questions » EOM-IP for Benzene

EOM-IP for Benzene

Page: 1

Author Post
Registered: Jun 2016
Posts: 2
Dear All,

I am trying to run an eom-ip calculation on benzene in a cc-pvdz basis but I am experiencing some very strange behavior with the Davidson iterations in the eom section. Depending on the exact parameters used, it sometimes fails to converge (producing nans in the output) while in other cases it converges to the wrong values predicting an IP of 5 ev for Benzene. I have also tried with the new GPLMR method but the results are the same. A similar behavior is observed for an aug-cc-pVDZ basis. As I am rather new to this package I could be missing something so any advice would be highly appreciated. I have tried playing the convergence thresholds etc. but this did not seem to change the problematic behavior. My input is provided below.

Thank you in advance.


0 1
C 0.000000 1.392000 0.000000
C 1.205507 0.696000 0.000000
C 1.205507 -0.696000 0.000000
C 0.000000 -1.392000 0.000000
C -1.205507 -0.696000 0.000000
C -1.205507 0.696000 0.000000
H 0.000000 2.473000 0.000000
H 2.141681 1.236500 0.000000
H 2.141681 -1.236500 0.000000
H 0.000000 -2.473000 0.000000
H -2.141681 -1.236500 0.000000
H -2.141681 1.236500 0.000000

max_scf_cycles 200
scf_guess core
method eom-ccsd
basis cc-pVDZ
ip_states [4,4,4,4,4,4,4,4]
cc_trans_prop true
scf_convergence 8
thresh 14
n_frozen_core fc
mem_static 2000
cc_memory 20000
cc_state_to_opt [6,1]
eom_davidson_threshold 00109
eom_davidson_max_iter 100
Registered: May 2016
Posts: 4
Hi Shachar,

My bet it that it is because you are requesting 4 transitions for each irrep. Some of these transitions will be very high in energy. Try requesting only the irrep you are interested in, or if not sure then request one of each irrep to start.
Registered: Jun 2016
Posts: 2
Hi @Sgozem,

Thanks for the reply. I am not sure this is the problem since the wrong results are already produced in the ionization energies calculation (right hand eom) long before the transition moments calculation begins. Clearly, it is some instability in the Davidson diagonalization, the calculation almost converges and then jumps of to some strange energy values and converges to them. I tried with another program (CFOUR) and there the Davidson procedure works so I guess it has to do with some fine tuning of the algorithm.

Registered: Sep 2007
Posts: 175
Samer is right, you've got into convergence problems because you are requesting too many states in such a small system (and also seek tight convergence). The algorithm might work or might get tripped. The same can happen with Cfour, you just got lucky. Ask for a reasonable number of states per irrep.

Page: 1

iOpenShell » Q-Chem specific questions » EOM-IP for Benzene