MACE+short CP2K can give unrealistic structures and partial charges

[tdpham2]

Some MOFs from the current geometry optimization workflow (MACE followed by short CP2K run) have unrealistic bond lengths, which likely leads to unphysical DDEC6 charges and artificially high CO₂ uptake in GCMC simulations. Attached are example CIFs (attached as .txt files)

Example:

1. For mof-6355264e.txt:

• Oxygen atom O4 has a charge of +2.00
• Carbon atom C59 has a charge of -5.77

2. For mof-3078fad9.txt:

• C56 has a charge of -3.61
• N12 has a charge of -3.57

These values suggest that the current geometry optimization may not be sufficient to relax the structures properly.

mof-3078fad9.txt
mof-6355264e.txt