This starting prompt: CC(C)=O.[H]C(C1=CC=C(C=C1)[N+]([O-])=O)=O.O=C(C2CCCN2)O, where all molecules are neutral and closed shell, occasionally result in intermediates with an overall wrong charge.
I have shown 4 examples below. In the first example, the overall charge of the system is +1, with 1 radical electron. It looks like that problem arises when H2O is elimintated from C-OH2+ and the charge is assigned to H2O (which then introduces a radical) rather than C: e.g. O=C([O-])C1CCCN1C(C)([OH2+])C. However, starting a new run from O=C([O-])C1CCCN1C(C)([OH2+])C, eliminates H2O correctly. So it's a bit random when the error occurs.
We are running with default parameters input parameters and have not retrained the model in any way.
This starting prompt: CC(C)=O.[H]C(C1=CC=C(C=C1)[N+]([O-])=O)=O.O=C(C2CCCN2)O, where all molecules are neutral and closed shell, occasionally result in intermediates with an overall wrong charge.
I have shown 4 examples below. In the first example, the overall charge of the system is +1, with 1 radical electron. It looks like that problem arises when H2O is elimintated from C-OH2+ and the charge is assigned to H2O (which then introduces a radical) rather than C: e.g. O=C([O-])C1CCCN1C(C)([OH2+])C. However, starting a new run from O=C([O-])C1CCCN1C(C)([OH2+])C, eliminates H2O correctly. So it's a bit random when the error occurs.
We are running with default parameters input parameters and have not retrained the model in any way.