New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Problems with ADC(2) in Psi4? #1033

Open
ssh2 opened this Issue Jun 6, 2018 · 2 comments

Comments

Projects
None yet
2 participants
@ssh2
Contributor

ssh2 commented Jun 6, 2018

Hi everyone,

I was testing the ADC(2) code in Psi4 on small molecules and noticed that the results I obtained were significantly different from Q-Chem. The SCF and MP2 energies match very well.

For N2 (r = 1.2 A) with cc-pVDZ basis, here's a summary of the energies for the 8 lowest-lying singlet states from Q-Chem 4.4.1 compared to Psi4 1.1:

Q-Chem 4.4.1:
MP2 ground-state energy: -109.2579498247
Excitation energy: 8.385211 eV
Excitation energy: 8.385211 eV
Excitation energy: 8.832813 eV
Excitation energy: 9.441957 eV
Excitation energy: 9.441957 eV
Excitation energy: 13.868446 eV
Excitation energy: 13.868446 eV
Excitation energy: 14.791430 eV

Psi4 1.1:
MP2 ground-state energy: -109.25794982519506
-> 1 A state : 0.3113289 (a.u.), 8.4716885 (eV)
-> 2 A state : 0.3113289 (a.u.), 8.4716885 (eV)
-> 3 A state : 0.3343892 (a.u.), 9.0991916 (eV)
-> 4 A state : 0.3565394 (a.u.), 9.7019279 (eV)
-> 5 A state : 0.3565394 (a.u.), 9.7019279 (eV)
-> 6 A state : 0.5069575 (a.u.), 13.7950140 (eV)
-> 7 A state : 0.5069575 (a.u.), 13.7950140 (eV)
-> 8 A state : 0.5127312 (a.u.), 13.9521243 (eV)

The Q-Chem energies agree with my standalone code. For N2 with STO-3G basis I simply get zero excitation energies with Psi4:

==> ADC(2) Computation <==                       

-> 1 A state : 0.0000000 (a.u.), 0.0000000 (eV)
Non-iterative: 0.3864160 (a.u.), 10.5149136 (eV)
┆ ┆ Occ Vir Coefficient
---------------------------------------------

Converged in   2 iteration.                      
Squared norm of the S component:  1.0000000      
The S vector is rotated up to 90.000 (deg.)      
#WARNING: Strongly rotated from the CIS state!   

-> 2 A state : 0.0000000 (a.u.), 0.0000000 (eV)
Non-iterative: 0.4145495 (a.u.), 11.2804653 (eV)
┆ ┆ Occ Vir Coefficient
---------------------------------------------

Converged in   2 iteration.                      
Squared norm of the S component:  1.0000000      
The S vector is rotated up to 90.000 (deg.)      
#WARNING: Strongly rotated from the CIS state!   

Any idea about what's going on?

My input files for cc-pvdz:

Psi4:
molecule {
0 1
N 0.000000 0.000000 0.000000
N 0.000000 0.000000 1.200000
symmetry c1
}

set {
basis cc-pvdz
guess sad
reference rhf
scf_type pk
roots_per_irrep [8]
}

set_num_threads(12)
energy('adc')

Q-Chem:
$molecule
0 1
N 0.0 0.0 0.0
N 0.0 0.0 1.2
$end

$rem
jobtype sp
method adc(2)
basis cc-pvdz
mem_total 4000
mem_static 100
cc_symmetry false
ee_singlets 8
$end

@CDSherrill

This comment has been minimized.

Show comment
Hide comment
@CDSherrill

CDSherrill Jun 7, 2018

Member
Member

CDSherrill commented Jun 7, 2018

@ssh2

This comment has been minimized.

Show comment
Hide comment
@ssh2

ssh2 Jun 7, 2018

Contributor

It looks like one of the corrections could be the "partial renormalization scheme", but according to the manual it is turned off by default...

I tried using partial renormalization (set pr true) and it is indeed giving different ADC(2) energies:
-> 1 A state : 0.3074681 (a.u.), 8.3666314 (eV)
-> 2 A state : 0.3074681 (a.u.), 8.3666314 (eV)
-> 3 A state : 0.3282148 (a.u.), 8.9311786 (eV)
-> 4 A state : 0.3502345 (a.u.), 9.5303635 (eV)
-> 5 A state : 0.3502345 (a.u.), 9.5303635 (eV)
-> 6 A state : 0.5040061 (a.u.), 13.7147013 (eV)
-> 7 A state : 0.5040061 (a.u.), 13.7147013 (eV)
-> 8 A state : 0.5127427 (a.u.), 13.9524352 (eV)

I wonder if the original problem is related to convergence issues when starting with CIS guess vectors. For all states in N2 the ADC(2) code in Psi4 gives the following warning:
"#WARNING: Strongly rotated from the CIS state!"

Contributor

ssh2 commented Jun 7, 2018

It looks like one of the corrections could be the "partial renormalization scheme", but according to the manual it is turned off by default...

I tried using partial renormalization (set pr true) and it is indeed giving different ADC(2) energies:
-> 1 A state : 0.3074681 (a.u.), 8.3666314 (eV)
-> 2 A state : 0.3074681 (a.u.), 8.3666314 (eV)
-> 3 A state : 0.3282148 (a.u.), 8.9311786 (eV)
-> 4 A state : 0.3502345 (a.u.), 9.5303635 (eV)
-> 5 A state : 0.3502345 (a.u.), 9.5303635 (eV)
-> 6 A state : 0.5040061 (a.u.), 13.7147013 (eV)
-> 7 A state : 0.5040061 (a.u.), 13.7147013 (eV)
-> 8 A state : 0.5127427 (a.u.), 13.9524352 (eV)

I wonder if the original problem is related to convergence issues when starting with CIS guess vectors. For all states in N2 the ADC(2) code in Psi4 gives the following warning:
"#WARNING: Strongly rotated from the CIS state!"

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment