Dear Amber community,
I'm trying to use the POL3-LT forcefield from this paper
(
https://pubs.acs.org/doi/10.1021/jp212117d) - the authors mention
using AmberTools.
I've successfully constructed a topology using IPOL=4 to enable
Thole-linear damping (see prep.sh in the attachment). Using
AmberTools25, sander returns an energy and force, but they do not
agree with my own implementation of Thole's polarization model. The
total induced dipole does agree with my implementation, however.
Input & output are attached.
Furthermore, running with:
do_debugf=1,
atomn=1,2,3,4,5,6
reveals that sander's analytical gradients do not agree with numerical
gradients.
The same run with IPOL=1 (no Thole damping) yields energies and
gradients that agree with my own implementation and with numerical
gradients.
Note: this is for a water dimer artificially squeezed together with
aperiodic boundary conditions.
I realize that IPOL>1 is not documented in the manual, but thought I
should bring this to your attention, in case the behavior was not
known.
It would be good to know if a previous version of AmberTools does not
have this issue or if a fix might be in the works.
Best,
Will
William Glover
Associate Professor of Chemistry
Director, NYU-ECNU Center for Computational Chemistry
NYU Shanghai | 上海纽约大学
Room E622, 567 West Yangsi Road | 杨思西路567号，E622室
Pudong New Area, Shanghai | 上海市浦东新区
China
200124
Office (电话): +86 21 2059 5208
Website: wp.nyu.edu/glover
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Thu Feb 19 2026 - 05:30:02 PST
