Incorrect Polarizability Assignment for PC Molecule in poltype2 (Possibly Related to Aromatic Ring Detection Issue)

[KeitoHASEGAWA]

Hello,

I am using poltype2 to generate AMOEBA parameters for a PC (propylene carbonate) molecule.
However, I found that incorrect polarizability values are being assigned, especially for atoms within the ring (O and C atoms). For example, poltype2 currently assigns values such as:

polarize 404 1.6620 0.3900 (should be Oar 0.832260743)
polarize 405 1.6620 0.3900 (should be Oar 0.832260743)
polarize 407 1.6620 0.3900 (should be Osp2c 0.913766949)

These incorrect values suggest that the PC ring atoms are being improperly detected or assigned.
This may be related to the previously reported issue:
Bug Report: Incorrect Aromatic Ring Detection and Parameter Assignment #82

Additional Concern
I also noticed a slight discrepancy between polarizability values assigned by poltype2 and the values in amoeba21polarize.prm. For example:

polarize 406 2.0730 0.3900 404 405 407

But according to amoeba21polarize.prm, the value should be approximately 2.064511770 (~2.0645).
Why does poltype2 assign a different value here?

I have attached the related files for reference.
Thank you for your support.

bug.zip

[KeitoHASEGAWA]

Hello again,
I would like to provide additional details regarding the issue I am experiencing with AMOEBA atom-type assignment in poltype2.

I suspect that the incorrect bond, angle, stretch-bend, out-of-plane bend, and torsion parameters are caused by incorrect internal assignment of atom types during the parameter generation process.
Specifically, it appears that poltype2 may be incorrectly assigning AMOEBA atom classes based on amoeba21smartstoclass.txt. This issue may be related to the previously reported problem:
“Incorrect Aromatic Ring Detection and Parameter Assignment #82.”

For the PC molecule, I believe the correct atom-class assignment should be as follows:

401 35, 402 40, 403 40, 404 177, 405 177, 406 83, 407 166, 408 16, 409 16, 410 16

The numbers on the left correspond to atom indices in the generated final.key file,
and the numbers on the right correspond to atom classes defined in amoeba21smartstoclass.txt.

If these classes were assigned properly, the bond parameters should instead map to the correct reference entries. For example:

bond 402 404 211.9971 1.4342
↓
177 33 249.687105

bond 403 405 211.9971 1.4360
↓
177 34 326.046328

bond 401 402 244.2139 1.4929
↓
33 35 280.353180

Similar assignment errors likely affect angle, strbnd, opbend, and torsion parameter selection as well.

Additional Note: VDW parameter
VDW parameters were mostly assigned correctly, but the carbonyl carbon was assigned:

vdw 4 3.8200 0.1060

However, I believe the most appropriate value should come from "Acetic Acid C=O":
62 3.7800 0.1060

Is such precise VDW atom-class recognition currently beyond the intended scope of poltype2, or could this also be resolved by correcting atom-type assignment logic?