search

lsmo-epfl / aiida-lsmo

AiiDA workflows for the LSMO laboratory at EPFL
Other
9 stars 13 forks source link

cp2k binding energy calculations yield unphysical bsse #75

Closed mpougin closed 3 years ago

mpougin commented 3 years ago

The DFT energies calculated with the BindingSite workchain result in extremely high corrected binding energies (~ 2 to 3 * -10^5 kJ/mol) due to unphysical bsse. I experienced this in all of my calculations this year. I just run the test wc (https://github.com/lsmo-epfl/aiida-lsmo/blob/develop/examples/run_BindingSiteWorkChain_MOF74_CO2.py) using raspa-4467e14@fidis-serial andcp2k-5.1@helvetios. I get the following cp2k output dict: { "binding_energy_bsse": -2178989.9351219, "binding_energy_corr": 2178956.7936331, "binding_energy_dispersion": -21.361676400578, "binding_energy_raw": -33.141488855341, "binding_energy_unit": "kJ/mol", "motion_opt_converged": false, "motion_step_info": { "dispersion_energy_au": [ -0.16233019269837, -0.16234970471036, -0.16231227878718, -0.16224767612766, -0.16222322395047, -0.16221221242885 ], "energy_au": [ -829.91930582395, -829.92128447225, -829.92147671786, -829.92159367314, -829.921674636, -829.92167681991 ], "max_grad_au": [ null, 0.0067308872, 0.0044478517, 0.0067341529, 0.0008866158, 0.0018017187 ], "max_step_au": [ null, 0.0387788492, 0.0279071904, 0.0531515546, 0.0174067956, 0.0074783668 ], "rms_grad_au": [ null, 0.0007732628, 0.0005362628, 0.0007456283, 0.0001156248, 0.0002175644 ], "rms_step_au": [ null, 0.0052556901, 0.0042090882, 0.0048061243, 0.0021768516, 0.0010439191 ], "scf_converged": [ true, true, true, true, true, true ] } }

I think there is some intrinsic problem with the calculation, @ltalirz or @danieleongari maybe someone with a better understanding of the code can take a look

ltalirz commented 3 years ago

As mentioned in the forum this can happen when the GHOST atoms are missing.

This is my mistake - in the following commit I accidentally replaced get_kinds_with_ghost_section by get_kinds https://github.com/lsmo-epfl/aiida-lsmo/commit/e3b2babc39a8850dbee0466aee2b696924e8cac5#diff-eb5fed9f81158586ba48a37091048a7ec787ddb752f94cf853c710d8df9544ea

@mpougin Would you mind undoing this replacement in a pull request?

We do check the binding energy in a test https://github.com/lsmo-epfl/aiida-lsmo/blob/7f58b7b3fd15cfb60a4b97211cdc959652947278/examples/test_Cp2kBindingEnergy_CO2_MOF74.py#L82

but perhaps this particular key does not include the BSSE? Which key would include it?

ltalirz commented 3 years ago

Which key would include it?

The key would be binding_energy_corr - I'll add it to the test

mpougin commented 3 years ago

thank you @ltalirz