Large changes in excitation energies with EOM-CCSD(ft)

Hello,

I am trying to run EOM-CCSD(fT) calculations on a system containing 11 heavy atoms, albeit with def2-svp basis sets.

With EOM-CCSD, I get a value of 1.08 eV. With ADC2, SCS-ADC2 & CC2, I get around 1.02-1.10 eV. Results from double hybrid TDDFT calculations are also around this range. These are all perfectly reasonable since the experimental data is around 0.99 eV. With larger def2-TZVP basis sets, results from this WFT and TDDFT methods change only minimally.

However, with EOM-CCSD(fT), the excitation energy jumps from 1.08 eV to 2.57 eV. The fT correction to the ground state is 0.055 a.u. more than for the first singlet excited state. With EOM-CC3 using another package, I get 1.06 eV, but that calculation took almost a month, requesting for 8 states in total.

Does anyone have insights into why this could be? It just seems rather odd that the fT corrections make EOM-CCSD “worse” by about 1.50 eV.

With the perturbative triples inclusion, is it possible to converge to a different energy state? With EOM-CCSD, the first and second excited states are 1.08 and 3.54 eV, respectively.

Thanks.

This is my input card:

$rem
MEM_STATIC = 10000
MEM_TOTAL = 120000
SYM_IGNORE TRUE METHOD EOM-CCSD(ft)
CCMAN2 false
BASIS def2-SVP
JOBTYPE SP
SCF_ALGORITHM DIIS
MAX_SCF_CYCLES 512
EE_SINGLETS 6
EE_TRIPLETS 2
CC_SYMMETRY FALSE
STATE_ANALYSIS false
CC_EOM_PROP false
CC_TRANS_PROP false
CC_CONVERGENCE 5
EOM_NGUESS_SINGLES 7
EOM_NGUESS_DOUBLES 1
EOM_DAVIDSON_CONVERGENCE 4
EOM_DAVIDSON_THRESHOLD 00102
EOM_DAVIDSON_MAX_ITER 32
EOM_DAVIDSON_MAXVECTORS 240
$end

~ ~ ~ ~

What is the system? If it is badly behaved, a questionable result is expected.

The (ft) correction should not be used for EOM-EE, only for EOM-SF and EOM-IP. This is explained in our papers: