AMBER: QM/MD energy conservation

From: Evan Kelly <>
Date: Mon, 11 Dec 2006 20:17:59 -0700

Hi Everyone,

I am hoping to use Amber9 to perform some QM/MD simulations of a
molecule (treated quantum mechanically) in a box of classical (TIP3P)
water molecules. After warming and equilibrating my system, I
decided to test my timestep by running the system under NVE
conditions (ie. ntb=1, ntt=0, ntp=0). I am surprised by the results:
during a 20ps run, the total energy steadily rises by about 120kcal/
mol (this is about 100 from the potential energy and about 20 from
the kinetic energy). My guess is that this indicates a problem
somewhere. The problem appears to be with the QM portion of the
simulation, as the identical run with ifqnt=0 conserves energy almost
perfectly. The QM energy does seem to be more or less conserved,
though. The same problem exists for dt = 1, 1.5, and 2, as well as
for both the AM1 and PM3 QM flavors. Nothing unusual is visible upon
viewing the trajectory of the simulation. I have included my input
file below.

I'd appreciate any suggestions or comments anyone can make about this
potential problem. Thanks!

-----input file-------

Minimization of water molecules of VLB (1Z2B structure) in TIP3P box
   imin = 0,
   irest = 1,
   ntx = 7,
   ntb = 1, !Constant volume simulation while testing E conservation
   pres0 = 1.0,
   ntp = 0, !no barostat while testing E conservation
   taup = 2.0,
   ntc = 2, !says to SHAKE H-X bond lengths, needed for TIP3P water
   ntf = 2, !says to skip H-X bond calculations, again needed for
TIP3P water
   cut = 10,
   dielc = 1,
   tempi = 300.0,
   temp0 = 300.0,
   ntt = 0, !constant energy simulation to test conservation w/ dt
   nstlim = 10000, dt = 0.002,
   ifqnt = 1, !TURN ON QM/MM
   ntpr = 50, ntwx = 100, ntwr = 500, ntwe = 50,
   qmmask = ':VLB', !which atoms are treated with QM
   qmcharge = 0,
   qmtheory = 1, !use the PM3 hamiltonian
   qmshake = 1, !use shake on the QM atoms
   qm_ewald = 1,
   qm_pme = 1,
   spin = 1,
   printcharges = 1,
   writepdb = 1,

Evan Kelly

The AMBER Mail Reflector
To post, send mail to
To unsubscribe, send "unsubscribe amber" to
Received on Wed Dec 13 2006 - 05:21:51 PST
Custom Search