Confused about ROOTS_PER_IRREP


I am trying to do an excited state energy calculation with CC3. I want to get the energy corresponding to the lowest B1 excitation for a molecule in C2v symmetry. However, the description for ROOTS_PER_IRREP says “Irreps denote the final state symmetry, not the symmetry of the transition”. What does that mean? Should the array not be [0,0,1,0]?


If the molecule has an A1 ground state in C2v then the lowest B1 transition would also correspond to the lowest B1 state.

In general though The ROOTS_PER_IRREP options should be set using the excited state symmetry, if you know the symmetry of the transition you can work out the final state symmetry using the product table for the group and the ground state symmetry.

Two examples from the C2v PG:

  1. CH2-triplet first B1 transition:
    • Ground state: B1
    • Excited state = B1 x B1 = A1
    • ROOTS_PER_IRREP = [ 1, 0, 0, 0]
  2. H20-singlet first B1 transition:
    • Ground state: A1
    • Excited State = B1 x A1 = B1
    • ROOTS_PER_IRREP = [0, 0, 1, 0]

What does that mean? Should the array not be [0,0,1,0]?

If the ground state is A1, then yes that would be the correct way to do the calculation you want. However if the ground state is not totally symmetric and you want the first B1 transition and not the first B1 state you will have to work out the symmetry of the final state to set the ROOTS_PER_IRREP option.

I was trying to do a EOM-CC3 energy computation with pseudopotentials and basis set corresponding to it (comparable with the aug-cc-pVDZ). I get an excitation energy of a B1 state of 4.9 eV and when I do an all electron calculation, I get an energy of 9.1
eV for the same. A calculation with a frozen core but with aug-cc-pVDZ also yields a value of 9.1 eV for this calculation. Why is that?



I couldn’t say from the information given however I agree that those results do not seem consistent. If you would be willing to share the three output (pseudopotential, all electron, and frozen core) files I may be able to provide some additional help.

Here are the input and output files.

(Attachment cn3-cc3_exc.out is missing)

(Attachment cn3-cc3_exc.inp is missing)

(Attachment cn3-cc3_exc-alle.out is missing)

(Attachment cn3-cc3_exc-alle.inp is missing)

cn3-cc3_exc_b1_bfd_pseudo_out.txt (32.7 KB)

cn3-cc3_exc_b1_bfd_pseudo_inp.txt (688 Bytes)

cn3-cc3_exc_b1_fc_out.txt (30.9 KB)

cn3-cc3_exc_b1_fc_inp.txt (721 Bytes)

Thanks for sharing your output files. I see a few things that may help.

  1. It looks like the ECP basis is accounting for the same number of electrons as the frozen core calculation, however it is changing the HF solution enough that the orbital occupations are changing.

  2. The ECP calculation may not be adding the contribution from the core potentials correctly as the SCF, MP2, and CC3 energy is significantly different between the two.

I believe our ECP functionality is somewhat new and it is possible that there is a bug in either the implementation of ECPs or that ECP contributions are not correctly propagating to the correlated part of the calculation.

Ok. But the frozen core calculation with a normal basis also does not give a reasonable number. It’s 2 eV off (higher) than an expected value. Why is that?



What is the source of the expected value?

The reference value is from a full CI calculation.

What do RIA, ria, RIjAB in the description of the largest components of the wavefunction mean? And what is libdpd, cscf?



RIA and Ria are the elements of the alpha- and beta-spin components, respectively, of the single-excitation contributions to the excitation vector. RIjAb are the alpha-beta-spin components of the double-excitation contributions.