Energy inconsistent

Questions
#1

Hello! I was analyzing the frequency of the S1-state-benzene in water, and I also need to calculate the energy of S1-state-benzene. So I submitted two tasks for calculation. I found that the energy of S1-state-benzene in the two output files was inconsistent. What is the reason for this and which result should I use ?Here are my inputs.

49ae3716e8910824f41037a92578b1e
49ae3716e8910824f41037a92578b1e927×697 21.9 KB

here are my outputs
output1:
Excited state 1: excitation energy (eV) = 5.2907
Total energy for state 1: -232.05675062 au
Multiplicity: Singlet
Trans. Mom.: 0.0000 X -0.0001 Y -0.0000 Z

output2:
Excited state 1: excitation energy (eV) = 5.2801
Total energy for state 1: -232.05714100 au
Multiplicity: Singlet
Trans. Mom.: 0.0001 X -0.0002 Y -0.0000 Z

#2

Could you post the full input file (using preformatted text button “</>”) and not paste an image for our convenience? Are you using the same version of Q-Chem for the two outputs? Also, are the SCF energies in the two calculations different? If so, set “SCF_CONVERGENCE = 8” in both your input files. The default SCF_CONVERGENCE is 5 for “JOBTYPE = sp” and 8 for “JOBTYPE = freq”. This difference in defaults in different job types could cause the excitation energies to differ slightly, so setting the same SCF_CONVERGENCE might resolve the issue.

#3

That small of a different (0.01 eV) could easily be causes by the different convergence criteria, in my experience. By default, integral screening threshold and shell-pair drop tolerance are tied to the value of SCF_CONVERGENCE, which is different as Kaushik indicated.

#4

Thank you for your suggestion. After setting the scf_comvergence of the “jobtype=sp” to 8, the calculation results of them are still inconsistent.
here are my inputs.

$molecule
0 1
6       -0.767692585      0.026323675      0.000057693
6        0.661731661      0.026226200      0.000477435
6        1.376615305      1.264151602     -0.000060000
6        0.661824598      2.502219272     -0.001052779
6       -0.767626897      2.502280123     -0.001522463
6       -1.482431273      1.264284627     -0.000886217
1       -1.310394197     -0.913372235      0.000433853
1        1.203943501     -0.913720232      0.001340059
1        2.461693398      1.263879251      0.000313578
1        1.204364242      3.441916375     -0.001410105
1       -1.310360188      3.441904259     -0.002423918
1       -2.567567084      1.264116283     -0.001178136
$end

$rem
JOBTYPE         freq
METHOD          b3lyp
BASIS           6-31G*  
SYM_IGNORE      TRUE
CIS_N_ROOTS     3
CIS_SINGLETS    TRUE
CIS_TRIPLETS    FALSE
CIS_STATE_DERIV  1
SOLVENT_METHOD  PCM
DFT_D           D3_zero
xc_grid         3
mem_static      4000
mem_total       10000
CPSCF_NSEG      3
THRESH          12
$end

$pcm
Theory CPCM
Method SWIG
Solver Inversion
Radii Bondi
$end


$solvent
Dielectric 80.2 ! WATER
$end



$molecule
0 1
6       -0.767692585      0.026323675      0.000057693
6        0.661731661      0.026226200      0.000477435
6        1.376615305      1.264151602     -0.000060000
6        0.661824598      2.502219272     -0.001052779
6       -0.767626897      2.502280123     -0.001522463
6       -1.482431273      1.264284627     -0.000886217
1       -1.310394197     -0.913372235      0.000433853
1        1.203943501     -0.913720232      0.001340059
1        2.461693398      1.263879251      0.000313578
1        1.204364242      3.441916375     -0.001410105
1       -1.310360188      3.441904259     -0.002423918
1       -2.567567084      1.264116283     -0.001178136
$end

$rem
JOBTYPE         sp
METHOD          b3lyp
BASIS           6-31G*  
SYM_IGNORE      TRUE
CIS_N_ROOTS     3
CIS_SINGLETS    TRUE
CIS_TRIPLETS    FALSE
SOLVENT_METHOD  PCM
DFT_D           D3_zero
xc_grid         3
scf_convergence 8
mem_static      4000
mem_total       10000
THRESH          12
$end

$pcm
Theory CPCM
Method SWIG
Solver Inversion
Radii Bondi
$end


$solvent
Dielectric 80.2 ! WATER
$end
#5

It seems that this has to do with the LR-PCM parts. If you set TDDFT_LR_PCM = FALSE, both jobs afford same results. I suspect this discrepancy is somehow triggered by the fact that the frequency job is computing a relaxed density on an excited state and the LR-PCM code may be modified in preparation for that, but without going into the code I am not sure.

If you do set TDDFT_LR_PCM = FALSE, you still get solvent effects in the TDDFT part at what I’ve called the “0th-order” approximation (solvent-polarized MOs and energy levels).
https://doi.org/10.1002/wcms.1519