Molcas Forum

Praseetha@97

Member

Registered: 2025-09-22

Posts: 2

OpenMolcas CASPT2 vertical excitations for furan vs. Thiel benchmark

Hi all,

I’m trying to reproduce the CASPT2 vertical excitation energies of furan from the Thiel benchmark (J. Chem. Phys. 128, 134110 (2008),    https://doi.org/10.1063/1.2889385. My values are systematically off by ~0.05–0.14 eV. I would appreciate help spotting what I’m doing differently (IPEA, MS , SA recipe, geometry, etc.).

What I’m comparing?

Target: Thiel (2008) Table XVIII vertical singlet excitations for furan.
My runs: OpenMolcas (version: please see below), SA-CASSCF → MS-CASPT2, C2v symmetry.

Exact input

&GATEWAY
coord=furan.xyz(from the paper)
Group = XY Y
basis=def-tzvp
&SEWARD
&SCF
Charge = 0
Spin= 1 
&RASSCF
Title= The CASSCF energy of furan
nActEl= 6 0 0
Inactive= 9 0 0 6
Ras2= 0 3 2 0 
Symmetry= 1; Spin= 1 
CIRoot= 4 4 1
LumOrb
&CASPT2
multistate= 4 1 2 3 4 
MaxIter= 80
Title = CASPT2 on gs

Questions
1. Is there a specificnstate-averaging recipe per irrep for furan in the benchmark that I should mirror exactly (number of A1/B1/B2 roots and weights)?

My small table (calc vs Thiel)

State   Irrep   Calc (eV)   Thiel (eV)   Δ (Calc–Ref)
1^1B2   B2      6.5715      6.43         +0.14
2^1A1   A1      6.5025      6.52         -0.02
3^1A1   A1      8.1663      8.22         -0.05

Offsets are small but persistent across states/irreps.
If you spot any mismatch with the original setup (IPEA, SA scheme per irrep, geometry source, or anything else), I’d be grateful for pointers. Thanks!

Ignacio

Administrator

From: Uppsala

Registered: 2015-11-03

Posts: 1,165

Re: OpenMolcas CASPT2 vertical excitations for furan vs. Thiel benchmark

According to the supporting information, the SA is over 4 states for A1 and 3 states for B2. Also, according to the main paper, the A1 excitation energies are computed with respect to the MS-CASPT2 ground state, but for other irreps it's against the SS-CASPT2 ground state. With that, I get:

State   Irrep   Calc (eV)   Thiel (eV)   Δ (Calc–Ref)
1^1B2   B2      6.3935      6.43         -0.04
2^1A1   A1      6.5025      6.52         -0.02
3^1A1   A1      8.1663      8.22         -0.05

which is slightly better. It can be further "improved" if you add `Shift = 0.1` to the CASPT2:

6.4471
6.5072
8.2507

I don't know if they used such a shift or something similar in their original calculations, it could also be some other bug fix or modification in all these years

Praseetha@97

Member

Registered: 2025-09-22

Posts: 2

Re: OpenMolcas CASPT2 vertical excitations for furan vs. Thiel benchmark

Thank you for your reply.
I’m getting consistent results for the A1 irrep, but my answers for the B1 irrep don’t match yours.

&GATEWAY
coord=/home/pxp479/thiel2008XYZs/furan.xyz
Group = XY Y
basis=def-tzvp
&SEWARD
&SCF
Charge = 0
Spin= 1 
&RASSCF
Title= The CASSCF energy of acetone
nActEl= 6 0 0
Inactive= 9 0 0 6
Ras2= 0 3 2 0 
Symmetry= 4; Spin= 1 
CIRoot= 3 3 1
LumOrb
&CASPT2
Multistate= 3 1 2 3 
MaxIter= 80
Title = CASPT2 on gs

Furan (MS-CASPT2, both irreps)
State   Energy (Eh)     ΔE (Eh)      ΔE (eV)       Expt (eV)   Expt−Calc (eV)
gs      -229.450981
1B2     -229.2094824    0.24149860   6.571511682   6.43        -0.141511682
2A1     -229.2120184    0.23896260   6.502503607   6.52         0.017496393
3A1     -229.1508759    0.30010510   8.166275790   8.22         0.053724210

&GATEWAY
coord=/home/pxp479/thiel2008XYZs/furan.xyz
Group = XY Y
basis=def-tzvp
&SEWARD
&SCF
Charge = 0
Spin= 1 
&RASSCF
Title= The CASSCF energy of acetone
nActEl= 6 0 0
Inactive= 9 0 0 6
Ras2= 0 3 2 0 
Symmetry= 4; Spin= 1 
CIRoot= 3 3 1
LumOrb
&CASPT2
MaxIter= 80
Title = CASPT2 on gs

Furan (SS-CASPT2 for B2)
State   Energy (Eh)     ΔE (Eh)      ΔE (eV)       Expt (eV)   Expt−Calc (eV)
gs      -229.450981
1B2     -229.2062364    0.24474460   6.659845155   6.43        -0.229845155
2A1     -229.2120184    0.23896260   6.502503607   6.52         0.017496393
3A1     -229.1508759    0.30010510   8.166275790   8.22         0.053724210

Ignacio

Administrator

From: Uppsala

Registered: 2015-11-03

Posts: 1,165

Re: OpenMolcas CASPT2 vertical excitations for furan vs. Thiel benchmark

You're missing the SS calculation for the ground state:

&GATEWAY
coord=furan.xyz
Group = XY Y
basis=def-tzvp
&SEWARD
&SCF
Charge = 0
Spin= 1 
&RASSCF
Title= The CASSCF energy of furan
nActEl= 6 0 0
Inactive= 9 0 0 6
Ras2= 0 3 2 0 
Symmetry= 1; Spin= 1 
LumOrb
&CASPT2
MaxIter= 80
Title = CASPT2 on gs

Furan (SS-CASPT2 for gs, MS-CASPT2 for B2)
State   Energy (Eh)     ΔE (Eh)      ΔE (eV)       Expt (eV)   Expt−Calc (eV)
gs      -229.444412
1B2     -229.2094824    0.23492921   6.392749746   6.43        -0.141511682

PS. Your B2 calculations say "acetone"... That's why I find the TITLE keyword often does more harm than good.