AMBER Archive (2000)

Subject: Re: Decreasing PE during MD

From: Thomas Cheatham (cheatham_at_cgl.ucsf.edu)
Date: Wed Apr 26 2000 - 11:49:16 CDT


> We have been doing long dynamics runs (2-3 ns) under constant
> pressure. During these runs everything looks pretty good for the
> first 1ns (temp, pressure, energy all settle down). After the first
> 1ns, the potential energy begins to fall at an increasing rate so that
> after 2ns it 0-1ns you can see a slight downward slope. Nothing else
> appears to be changing (temp, pressure, volume, or kinetic energy.

> ndfmin = 0, ntcm = 0, nscm = 0,

See:

  Harvey et. al, "The flying ice cube..." J. Comp. Chem. 19, 726-740 (1998)
  Chiu et al., "Collective motion artifacts..." J. Comp. Chem. 21, 121-131 (2000)

This is a hallmark of the "flying block of ice problem" and lack of
complete energy conservation (i.e. energy drain) which leads to a
increase in center of mass translation when velocity scaling is applied...

To remove the artifact, periodically remove the COM translation (both at
the beginning of the run after velocity assignment and periodically during
the run) using NTCM = 1, NCSM = 5000, (or so).

To get better energy conservation, use AMBER6 which has heuristic pairlist
updating by default and tighter SHAKE tolerances (TOL=0.000001). Even
without complete energy conservation however, just removing the COM
translation should remove the artifact...

e-mail if you have additional questions,

--tom

Thomas E. Cheatham, III
Assistant Research Professor
Department of Medicinal Chemistry & Center for High Performance Computing
University of Utah INSCC 418
30 South 2000 East, Room 201 155 South 1452 East
Salt Lake City, Utah 84112-5820 Salt Lake City, Utah 84112-0190
tom.cheatham_at_pharm.utah.edu cheatham_at_chpc.utah.edu
FAX: (801) 585-9119 FAX: (801) 585-5366
phone: (801) 587-9652