Dear All,
Is it possible to calculate CHELPG charges for excited EOM-EE-CCSD states?
(In particular, using Q-Chem 4.4.)
Thank you.
I doubt it. They are probably only implemented for the underlying Hartree-Fock wavefunction in the ground state. Excited states would require the excited-state electrostatic potential, which I don’t think is readily available either.
Dear John,
Thank you very much for the prompt reply.
It seems that the ground state charges are either for HF or CCSD density matrix depending on CC_REF_PROP (false → HF, true → CCSD).
Interestingly, setting both CC_REF_PROP and STATE_ANALYSIS to true
results into different Mulliken charges (for the ground state), which are printed in the same output file.
Here is an example for H2O:
Mulliken Population Analysis
Atom Charge (e)
--------------------
1 O -0.231158
2 H 0.115579
3 H 0.115579
--------------------
Ground-State Mulliken Net Atomic Charges
Atom Charge (a.u.)
----------------------------------------
1 O -0.231647
2 H 0.115824
3 H 0.115824
----------------------------------------
Do you know why this is so?
[Differences are larger for larger molecules]
The input is below for completeness:
$molecule
0 1
O 0.00000 0.00000 0.11779
H 0.00000 0.75545 -0.47116
H 0.00000 -0.75545 -0.47116
$end
$rem
GUI = 2
SYM_IGNORE true
JOBTYPE sp
BASIS sto-2g
METHOD ccsd
!
CC_REF_PROP true
STATE_ANALYSIS true
!
MEM_STATIC 2000
MEM_TOTAL 165000
$end
It’s possible that this is using the CC relaxed density to compute the Mulliken charges, assuming that you’ve solved the Lambda equations to obtain the relaxed density. Mulliken charges require only the density matrix whereas CHELPG charges require the electrostatic potential. In principle you could compute this from the relaxed density but we never bothered when we coded the CHELPG charges. Differences should be very minor and I don’t trust these charge models to that level of detail to make the difference meaningful.
Focusing on the ground state Mulliken charges only:
Do you think that the block printed first (the one with the title Mulliken Population Analysis, which is triggered by STATE_ANALYSIS true)
corresponds to the unrelaxed CCSD density matrix
and the second block (Ground-State Mulliken Net Atomic Charges)
corresponds to the relaxed CCSD density matrix?
Check which one agrees with the corresponding Hartree-Fock calculation
HF Mulliken charges are yet different:
Ground-State Mulliken Net Atomic Charges
Atom Charge (a.u.)
----------------------------------------
1 O -0.252568
2 H 0.126284
3 H 0.126284
----------------------------------------
So, both sets of Mulliken charges in the CCSD output (CHELPG charges for excited EOM-EE-CCSD states - #3 by evgenii)
differ from the HF Mulliken charges.
And the question is why CCSD set 1 differs from CCSD set 2?
The HF input is below for completeness:
$molecule
0 1
O 0.00000 0.00000 0.11779
H 0.00000 0.75545 -0.47116
H 0.00000 -0.75545 -0.47116
$end
$rem
GUI = 2
SYM_IGNORE true
JOBTYPE sp
BASIS sto-2g
METHOD hf
!
MEM_STATIC 2000
MEM_TOTAL 165000
$end
I don’t really know, and maybe someone better acquainted with the CC code will step in to answer, but I would say two things
(1) Those first two sets of Mulliken charges are not meaningfully different. (Differences in the 4th decimal place could easily result from slightly different thresholds at two different places in the code.)
(2) The differences w.r.t. ground-state HF are meaningful and may reflect the use of the relaxed density. You could try to see how the relaxed density shifts things around and see if that’s consistent with these changes.
Thanks.
For a larger molecule differences are larger (in the 2nd decimal place).
Here are just the first three atoms:
Mulliken Population Analysis
Atom Charge (e)
--------------------
1 S -0.173539
2 O -0.457669
3 N -0.459491
Ground-State Mulliken Net Atomic Charges
Atom Charge (a.u.)
----------------------------------------
1 S -0.190108
2 O -0.474656
3 N -0.469341
I’m wondering if the both sets of charges are computed from the same density or not… ?
I think you are missing the “CC_FULLRESPONSE = 1” in your REM. When “CC_FULLRESPONSE = 1” is set, the orbital relaxation contribution will be included in the DM, and then the Mulliken Charges from the LIBWFA (STATE_ANALYSIS = true) printout should match the one listed at the end of the output file that corresponds to the relaxed CCSD DM. It does for me for water.
With the default “CC_FULLRESPONSE = 0”, I am not sure why the two sets of Mulliken Charges are different though—both these sets are obviously different from the set obtained from a relaxed-DM calculation.
Thank you, Kaushik!
I observe the same behavior.
Do you think that my input here
CHELPG charges for excited EOM-EE-CCSD states - #3 by evgenii
(i.e., without “CC_FULLRESPONSE = 1”)
results in use of the unrelaxed CCSD DM to calculate Mulliken charges?
Yes. Without “CC_FULLRESPONSE = 1”, the CCSD DM is not fully relaxed in that the orbital response is not computed. Also, you can compute Mulliken Charges for EOM target states by using the CC_STATE_TO_OPT REM keyword with “CC_EOM_PROP = true”. Again, set CC_FULLRESPONSE based on your needs for EOM target states.
Thanks.
Using CC_STATE_TO_OPT, it seems possible to calculate the CHELPG charges for excited states (see the original question: CHELPG charges for excited EOM-EE-CCSD states). But they are called “Ground-State” charges in the output.
For water I got the following (with Q-Chem 4.4):
Input (with CC_STATE_TO_OPT):
$molecule
0 1
O 0.00000 0.00000 0.11779
H 0.00000 0.75545 -0.47116
H 0.00000 -0.75545 -0.47116
$end
$rem
GUI = 2
SYM_IGNORE true
JOBTYPE sp
BASIS 6-31++G**
METHOD eom-ccsd
CCMAN2 true
N_FROZEN_CORE 0
!
EE_SINGLETS [1]
!
CC_REF_PROP true
CC_EOM_PROP true
CC_STATE_TO_OPT [1,1]
CC_FULLRESPONSE true
STATE_ANALYSIS true
CHELPG true
!
MEM_STATIC 2000
MEM_TOTAL 165000
$end
Output (with CC_STATE_TO_OPT):
Ground-State Mulliken Net Atomic Charges
Atom Charge (a.u.)
----------------------------------------
1 O 0.299656
2 H -0.149828
3 H -0.149828
----------------------------------------
Sum of atomic charges = 0.000000
Compute ChElPG charges using a head room of 2.80 A and a grid spacing of 0.300 A
Total Grid Volume 4782.182557 A^3
ChElPG fitting will utilize 12962 grid points
CHELPG time 1.27 (cpu) 0.08 (wall)
Ground-State ChElPG Net Atomic Charges
Atom Charge (a.u.)
----------------------------------------
1 O 0.103145
2 H -0.051146
3 H -0.051998
----------------------------------------
Sum of atomic charges = 0.000000
Input (without CC_STATE_TO_OPT):
$molecule
0 1
O 0.00000 0.00000 0.11779
H 0.00000 0.75545 -0.47116
H 0.00000 -0.75545 -0.47116
$end
$rem
GUI = 2
SYM_IGNORE true
JOBTYPE sp
BASIS 6-31++G**
METHOD eom-ccsd
CCMAN2 true
N_FROZEN_CORE 0
!
EE_SINGLETS [1]
!
CC_REF_PROP true
CC_EOM_PROP true
!CC_STATE_TO_OPT [1,1]
CC_FULLRESPONSE true
STATE_ANALYSIS true
CHELPG true
!
MEM_STATIC 2000
MEM_TOTAL 165000
$end
Output (without CC_STATE_TO_OPT):
Ground-State Mulliken Net Atomic Charges
Atom Charge (a.u.)
----------------------------------------
1 O -0.678043
2 H 0.339021
3 H 0.339021
----------------------------------------
Sum of atomic charges = 0.000000
Compute ChElPG charges using a head room of 2.80 A and a grid spacing of 0.300 A
Total Grid Volume 4782.182557 A^3
ChElPG fitting will utilize 12962 grid points
CHELPG time 1.29 (cpu) 0.08 (wall)
Ground-State ChElPG Net Atomic Charges
Atom Charge (a.u.)
----------------------------------------
1 O -0.802299
2 H 0.401223
3 H 0.401076
----------------------------------------
Sum of atomic charges = 0.000000
Again, setting CC_FULLRESPONSE false in Input (with CC_STATE_TO_OPT) results in two different sets of Mulliken charges (in contrast to CC_FULLRESPONSE true):
Mulliken Population Analysis
Atom Charge (e) h+ e- Del q
--------------------------------------------------------------------
1 O 0.328744 1.219176 -0.293203 0.925972
2 H -0.164372 0.135099 -0.598085 -0.462986
3 H -0.164372 0.135099 -0.598085 -0.462986
--------------------------------------------------------------------
Sum: 0.000000 1.489373 -1.489373 -0.000000
Ground-State Mulliken Net Atomic Charges
Atom Charge (a.u.)
----------------------------------------
1 O 0.302195
2 H -0.151097
3 H -0.151097
----------------------------------------
Sum of atomic charges = 0.000000
Compute ChElPG charges using a head room of 2.80 A and a grid spacing of 0.300 A
Total Grid Volume 4782.182557 A^3
ChElPG fitting will utilize 12962 grid points
CHELPG time 1.30 (cpu) 0.08 (wall)
Ground-State ChElPG Net Atomic Charges
Atom Charge (a.u.)
----------------------------------------
1 O 0.109926
2 H -0.054530
3 H -0.055396
----------------------------------------
Sum of atomic charges = 0.000000
So, two questions arise:
- Do you think it is correct to calculate the excited state CHELPG charges this way (i.e., using CC_STATE_TO_OPT)?
- Why different sets of Mulliken charges are produced in case of
CC_FULLRESPONSE falseand which one is correct?