|
|||||||||||||||||||||||||||||||||
AMBER Archive (2005)Subject: AMBER: broblem of replica exchange MD with polarizable force field
From: wenfei Li (wfli_at_biophy.nju.edu.cn)
Dear Amber users,
I'm doing a replica-exchange MD simulation with the polarizable force field (ff02). In calculating the induced dipole, the Car-Parinello scheme is used. However, the polarization part of the potential energy is always positive which is obviously unreasonable. Meanwhile, the polarization part of the potential energy is zero at the first time step of the MD. After reading the code carefully, I found that the varialble "initdip" which appears at Line 64 of ew_handle_dips.f is always zero after the first attempt of exchange:
"if ( indmeth < 3 .or. initdip == 1 )then)."
This results in that the initial value of the induced dipoles are always zero. I guess that the simultaneously use of the multisander and "data initdip/1/ " at Line 39 of ew_handle_dips.f may result in the above problem.
I tried to change the Line 64 of ew_handle_dips.f from "if ( indmeth < 3 .or. initdip == 1 )then)." to "if ( indmeth < 3 .or. nstep == 0 )then)."and added a common block "common /nstep_dip/ nstep" to both runmd.f and ew_handle_dips.f. After the above modifications,the results seems reasonable(The polarization part of the potential energy is negative). Is this a bug or something wrong with my input data?
Here is one of my input data:
Zinc finger : REMD simulation
My output data is as following,
NSTEP = 2 TIME(PS) = 550.042 TEMP(K) = 286.22 PRESS = 0.0
NSTEP = 3 TIME(PS) = 550.043 TEMP(K) = 287.20 PRESS = 0.0
...............
Thanks,
Wenfei Li
Institute of Biophysics,
-----------------------------------------------------------------------
| |||||||||||||||||||||||||||||||||
|