Does 43-ccd.py work for CCD analytic gradients? #2681
-
|
The script examples/cc/43-ccd.py offers a simple solution to calculate the CCD energy using CCSD code. If I've performed such a calculation using different quantum chemistry packages, and I find that CCD analytic gradients of PySCF are somewhat different from those of other packages. This is my input script Here are the last few lines in the script Here are the CCD/cc-pVTZ analytic gradients from 3 packages (no frozen core): (2) CFOUR (3) PySCF Maybe there is some code to be modified regarding solving the lambda equation? Thanks for any help or suggestion. |
Beta Was this translation helpful? Give feedback.
Replies: 4 comments
-
|
Just to eliminate any other possible sources of disagreement, can you
confirm that if you use CCSD, you get better agreement with gradients
between codes?
…On Tue, Oct 24, 2023 at 9:10 AM jxzou ***@***.***> wrote:
The script examples/cc/43-ccd.py offers a simple solution to calculate the
CCD energy using CCSD code. If mc.Gradients() is appended to calculate
the CCD analytic gradients, will the results be correct?
I've performed such a calculation using different quantum chemistry
packages, and I find that CCD analytic gradients of PySCF are somewhat
different from those of other packages. This is my input script
h2o_rhf.zip <https://github.com/pyscf/pyscf/files/13114865/h2o_rhf.zip>
Here are the CCD/cc-pVTZ analytic gradients from 3 packages (no frozen
core):
(1) Gaussian
0.00000000E+00 0.00000000E+00 -4.16132400E-03
0.00000000E+00 -1.17245720E-02 2.08066200E-03
0.00000000E+00 1.17245720E-02 2.08066200E-03
(2) CFOUR
0.00000000E+00 1.10000000E-14 -4.16106176E-03
0.00000000E+00 -1.17247788E-02 2.08053088E-03
0.00000000E+00 1.17247788E-02 2.08053088E-03
(3) PySCF
0.00000000E+00 0.00000000E+00 -4.63005710E-03
0.00000000E+00 -1.14598343E-02 2.31502860E-03
0.00000000E+00 1.14598343E-02 2.31502860E-03
Maybe there is some code to be modified regarding solving the lambda
equation? Thanks for any help or suggestion.
—
Reply to this email directly, view it on GitHub
<#1918>, or unsubscribe
<https://github.com/notifications/unsubscribe-auth/ADGAJC3C45F753BNTMTFFLLYA643LAVCNFSM6AAAAAA6NVUTQ6VHI2DSMVQWIX3LMV43ASLTON2WKOZRHE2TSMRSGA2DKNQ>
.
You are receiving this because you are subscribed to this thread.Message
ID: ***@***.***>
|
Beta Was this translation helpful? Give feedback.
-
|
@tberkel Thank you for your reply. Yes. Here are CCSD/cc-pVTZ analytic gradients from 3 packages (no frozen core, the same geometry as that of CCD): (2) CFOUR (3) PySCF One can see the better agreement. As for the electronic energies, even better agreement can be observed, so I did not shown any energy data. |
Beta Was this translation helpful? Give feedback.
-
|
I have a hunch that it has to do with how the lambda equations are solved, Line 438 in 169257b |
Beta Was this translation helpful? Give feedback.
-
|
Still not 100% sure, but since we only have even excitations due to T in CCD, we should probably only have even deexcitations with lambda, i.e. I think probably l1 = 0 in CCD. |
Beta Was this translation helpful? Give feedback.
Still not 100% sure, but since we only have even excitations due to T in CCD, we should probably only have even deexcitations with lambda, i.e. I think probably l1 = 0 in CCD.