Re: AMBER: Criteria for classical md being OK to move to QM-MM

From: M. L. Dodson <>
Date: Wed, 29 Aug 2007 13:24:43 -0500

Francesco Pietra wrote:
> Hi
> Thanks for meoh spreading out.
> xmgrace summary.TSOLVENT
> gives a straight line, which goes fro 100 to 400 (ordinate scale) vs
0 to 2000
> (abscissa). Similar trend for TSOLUTE. Which is confusing me, not
knowing the
> units 100-400 and 0-2000
> francesco

Do you mean the data points approach 400? (degrees Kelvin are the
units.) If so, that is wrong based on your tempi and temp0
settings. Since you have already equilibrated to temperature in a
constant volume run (I assume) before you switched to constant
pressure to equilibrate the density, you should get data points
that average to 300K.

Just to make sure, the classical equilibration recipe is something
like: minimization restraining solute, constant vol. dynamics to
heat up to desired temp., constant pressure dynamics to
equilibrate the density, then production. I always throw in a
200ps constant vol. run after this recipe before I start NVE
production. Different people do it different ways, but most will
use some variation of this kind of equilibration, I would think.

Following this recipe I have never observed temperature drift with
either classical dynamics (ff99SB) or QMMM for the <= 5ns runs I
have done (QMMM substantially shorter runs than 5ns). This is
with TIP3P water and protein and/or DNA solutes (neutralized with
Na+/Cl-). Other people report much longer NVE runs with stable
temperatures, and I believe Ross Walker has done ns length AMBER9
QMMM runs with stable temperatures. Note this is with PME.

For solvent cap or solvent box runs without PBC, you should not
expect energy conservation, so should use a thermostat, and even
then, you may have temperature drift with longer runs (depending
on how tightly the thermostat is coupled).

Bud Dodson

> --- "M. L. Dodson" <> wrote:
>> Francesco Pietra wrote:
>> > For a 98-atoms non polymeric molecule in meoh
>> >
>> > solvatebox EQE MEOHBOX 15
>> >
>> > I have carried out, in sequence:
>> > (1) minimization of solvent alone.
>> > (2) minimization whole system.
>> > (3) Heating in six steps from 0K to 300K.
>> > (4) Pressure equilibration according to:
>> >
>> > Equilibrate pressure and 30ps md
>> > &cntrl
>> > imin=0, irest=1, ntx=7,
>> > ntb=2, pres0=1.0, ntp=1,
>> > taup=2.0, cut=10.0, ntr=0,
>> > ntc=2, ntf=2,
>> > tempi=300.0, temp0=300.0,
>> > ntt=3, gamma_ln=1.0,
>> > nstlim=20000, dt=0.0015,
>> > ntpr=100, ntwx=100, ntwr=3000
>> > /
>> >
>> > After 150ps of the latter, I observe (perl script & xmgrace)
>> >
>> > Though, the solvent box (parallelepiped intended for later QM-MM) is
>> no more as
>> > neat as I expected: some meoh molecules are somewhat outside. Is
>> a warning
>> > that the system is not yet ready for QM-MM? If so, what next to do?
>> The solute
>> > is OK.
>> >
>> This is expected and is caused by the solvent molecules diffusing
>> out of the primary box of the periodic boundary conditions
>> treatment. See the discussion of the iwrap variable on pages 91
>> and 92 of the manual. Essentially, you can treat this two ways,
>> (1) set iwrap=1 which will mess up solvent diffusion calculations,
>> but is probably the best course if you are going toward QMMM. Or
>> (2) image the solvent back into the primary box with ptraj during
>> the analysis phase (see MANY postings on this list for guidance).
>> Post processing will be a problem if you use the trajectory file
>> to generate the starting structure for QMMM as trajectory files
>> have no velocities. Restart files have the velocities. Iwrap=1
>> is easier for your purpose, I believe. A single step of dynamics
>> with iwrap=1 should do the trick.
>> > Are somewhere guidelines how to interpret xmgrace plots for
>> summary_TSOLVENT
>> > summary_TSOLUTE?
>> Both should be stable, i.e., a straight, flat line, I would suppose.
>> Bud Dodson
>> >
>> > Thanks
>> >
>> > francesco pietra
>> >
>> >
M. L. Dodson
Email:	mldodson-at-houston-dot-rr-dot-com
Phone:	eight_three_two-56_three-386_one
The AMBER Mail Reflector
To post, send mail to
To unsubscribe, send "unsubscribe amber" to
Received on Sun Sep 02 2007 - 06:07:12 PDT
Custom Search