I would need to run some EOM-DEA-CCSD calculations with RI and CD approximations for computing the 2PA of some molecules, but qchem does not recognize my keywords (I have Qchem 6.0). here an example of my input:
$molecule
0 1
O 0.000000 0.000000 -0.069903
H 0.000000 0.757532 0.518435
H 0.000000 -0.757532 0.518435
$end
$rem
JOBTYPE sp
METHOD eom-ccsd
BASIS aug-cc-pvdz
AUX_BASIS rimp2-aug-cc-pvdz
DEA_SINGLETS [1,0,0,0]
N_FROZEN_CORE 1
CHOLESKY_TOL 3
CC_EOM_2PA 1
input_bohr false
cc_ref_prop true
cc_fullresponse 1
cc_trans_prop 1
cc_eom_prop true
scf_convergence 11
MAX_SCF_CYCLES 500
cc_convergence 9
CC_MAX_ITER 500
eom_davidson_convergence 7
eom_davidson_threshold 00105
thresh 14
mem_total 60000
cc_memory 55000
mem_static 4000
$end
In the output, I have only the GS energy computed at CCSD level and then this warning:
NIY
Solving for EOMDEA-CCSD A1 transitions to singlet states
Warning! Not enough singles to form guess subspace. Doubles guess vectors will be generated.
In case of poor EOM convergence, reduce requested DIP_STATES.
Then the calculation finish without any error few seconds later. Could you please let me know if everything is fine in my procedure or if I made some mystakes?
Hi Carmelo,
There are multiple facets to your query.
(1) RI/CD EOM-CC 2PA calculations are only available for EOM-EE-CCSD flavor, so remove the AUX_BASIS and CHOLESKY_TOL $rem variables from your input. You can still run 2PA calculations with the canonical EOM-DEA-CCSD method with Q-Chem 6.0.
(2) For canonical EOM-DEA 2PA calculations, only DEA to DEA 2PA calculations make sense as 2PA is an electron-conserving process. So, you need to request at least two DEA states, which you have not done in your input. So, you can set “DEA_SINGLETS = [2,0,0,0]”, for example, in your input. In addition, you need to set “CC_TRANS_PROP = 2” if you want transition properties between pairs of EOM-DEA target states or use CC_STATE_TO_OPT $rem variable.
(3) EOM-CC 2PA calculations in Q-Chem follow the expectation-value approach. So, you cannot use “CC_FULLRESPONSE = 1” with these calculations.
(4) You rarely need MAX_SCF_CYCLES and CC_MAX_ITER of 500. Typically, defaults are fine. In some cases, the number can be more than a 100. Also, is there a reason why you have set tight thresholds for EOM Davidson? The default values of EOM_DAVIDSON_CONVERGENCE and EOM_DAVIDSON_THRESHOLD should suffice for most systems.
Kaushik, addition even a single electron to a water molecule gives an unbound state that can go bananas in the diffuse basis set. It is even worse if two electrons are added. The resulting state is the state where two electrons sit on diffuse orbitals and go as far as possible from the molecule on diffuse orbitals. The overall electronic states sit in a continuum, which is often poorly behaved.
The issues above make me wonder whether indeed the 2PA from the dianion is needed. If it is needed, a completely different physics of electronic resonances must be accounted for with a different suite of methods.
Well, of course, EOM-DEA-CCSD is not meant to be used for such systems or target states (e.g., dianionic water). For certain target states (with doubly excited character) in systems such as polyenes (dicationic reference), EOM-DEA-CCSD might give better results for energies and first-order properties than other EOM-CC flavors. But 2PA is different (sum over states) and the currently implemented “2p+3p1h” EOM-DEA-CCSD method might still not give satisfactory results for such benchmark systems or transitions as the target configuration space for such an EOM-DEA-CCSD method is still limiting. The current EOM-DEA-CCSD 2PA implementation should therefore be used judiciously and drive the development of higher-level EOM-CC flavors. Thanks for pointing that out.
