Molcas Forum

GJC

Member

Registered: 2024-06-13

Posts: 3

Gradient Convergence Issues in QM/MM Optimizations

Dear All,

Our group has been using OpenMolcas/Tinker (version 24.06 with tag 19-g44e44d571), which resolves the issue with the 'rHidden' keyword, to perform QM/MM optimizations.

We carried out our optimizations at the HF/3-21G level, followed by CASSCF/3-21G, treating exclusively the chromophore as QM. We used the CHARMM36 and AMBER94 force fields to treat the MM region (protein, solvent, membrane, and ions).

The QM/MM calculations were set up identically, except for the force field used. The optimizations using the AMBER94 force field proceed smoothly and converge quickly compared to those using CHARMM36.

We are employing the hydrogen link atom approach at the QM/MM boundary and have assigned zero charges to the atoms involved in the C–C bond. In this case, our QM region is positively charged (+1).

We have observed a problem with the gradient convergence. For AMBER94, the gradients converge gradually and smoothly. However, using CHARMM36, we observe high oscillations. For instance, the threshold (default) for the optimization goes back and forth, in cycles, until convergence.

We would like to know if anyone in the community has encountered a similar issue. What could be the reason for this discrepancy (except the atomic charges)? Do we need to use an additional keyword in the OpenMolcas/Tinker input?

Is this trend expected?

Please find our input section below:

*********************************************************************
*********************************************************************
&GATEWAY &END
 Tinker
 Basis=3-21G
 Group=nosymm
 RICD
*********************************************************************
*********************************************************************
> EXPORT CONT=1
> EXPORT MOLCAS_MAXITER = 100
> DO WHILE <
*********************************************************************
*********************************************************************
&SEWARD &END
*********************************************************************
*********************************************************************
&ESPF &END
 External = Tinker
 LAMorok
*********************************************************************
*********************************************************************
&SCF &END
 CHArge = 1
*********************************************************************
*********************************************************************
&ALASKA &END
*********************************************************************
*********************************************************************
&SLAPAF &END
 Cartesian
 rHidden = 10
*********************************************************************
*********************************************************************

Thank you very much in advance.

Best regards,

GJC