Re: [AMBER] SCF convergence failed during QMMM

From: Brian Radak <radak004.umn.edu>
Date: Wed, 6 Aug 2014 09:32:58 -0400

What you describe sounds sensible, but unfortunately doesn't help to
diagnose the problem. Can you visualize the system up to the point where
things go wrong using the small print step as suggested?

This may sound crazy, but in the past the SCF procedure sometimes just
becomes unstable and it cannot be reproduced after a restart (everything
just continues "as expected"). I don't know if this is due to too much
stochastic noise (*i.e.* a bad integrator) or if the recent improvements to
SQM have fixed the problem (I certainly haven't seen this in over a year).

For what it is worth, I'm a bit skeptical of using QM/MM with SHAKE as the
constraint distances are not at all parameterized for the QM Hamiltonians.
I'd be twice as skeptical if you are using a 2 fs integration step (as that
may cause the integrator to become unstable). I'm sure these things could
be justified to a referee, but it is an unusual case nonetheless and would
require extra qualification.

Brian


On Wed, Aug 6, 2014 at 12:55 AM, Muthukumaran R <kumaran.bicpu.edu.in>
wrote:

> Dear Brian,
>
> thanks for your kind reply..
>
> We did two step minimization process, heating in constant volume followed
> by density equilibration under constant pressure. All these process were
> performed only with classical md methodology. After a 500ps classical md
> run QMMM was turned on. Two phase of 25ps QMMM simulations (with restart
> after 25ps) ran without any problems. Extension of the second phase
> resulted in the error. We observed some necessary conformational changes in
> the structure without major disturbances. We are performing QMMM to observe
> local changes only and not reaction mechanisms and hence SHAKE algorithm
> was used.
>
>
>
> On Tue, Aug 5, 2014 at 11:42 PM, Brian Radak <radak004.umn.edu> wrote:
>
> > This could be due to a lot of things. The usual advice is to decrease
> your
> > print step (*i.e.* set ntpr = ntwx = 1) and try to visualize what is
> > happening up until the divergence occurs. Usually you will see a rather
> > spectacular "explosion" of the QM region due to some specific event.
> >
> > Just glancing through your input you are turning of bond energies/forces
> > with hydrogens via SHAKE, including in the QM region. I would strongly
> > recommend setting "qmshake = 0". It is also rather perilous to do
> constant
> > pressure MD with QM/MM unless your system density is stable. I'd
> recommend
> > doing the density equilibration step with MM only or else with restraints
> > on the quantum region.
> >
> > Regards,
> > Brian
> >
> >
> > On Tue, Aug 5, 2014 at 12:23 PM, Muthukumaran R <kumaran.bicpu.edu.in>
> > wrote:
> >
> > > Dear amber users,
> > > I am running QMMM module implemented in AMBER 12. Here is my md.in
> > file.
> > >
> > > imin=0,irest=1,ntx=5,
> > > nstlim=25000,dt=0.001,
> > > ntc=2,ntf=2,
> > > cut=10, ntb=2, ntp=1, taup=1.0,
> > > ntpr=50, ntwx=50,
> > > ntt=3, gamma_ln=2.0,
> > > temp0=300.0,
> > > ifqnt=1,
> > > /
> > > &qmmm
> > > qmmask=":18-20,39-41,60-62",
> > > qmcharge=-5,
> > > qm_theory="PM3",
> > > qmcut=10.0,
> > > writepdb=1,
> > > /
> > >
> > > After running simulation for 50ps, while extending it to further time
> > > period ends in following error..
> > >
> > > QMMM: WARNING!
> > > QMMM: Unable to achieve self consistency to the tolerances specified
> > > QMMM: No convergence in SCF after 1000 steps.
> > > QMMM: Job will continue with unconverged SCF. Warning energies
> > > QMMM: and forces for this step will not be accurate.
> > > QMMM: E = -0.1349E+08 DeltaE = -0.2833E+05 DeltaP = 0.1770E+01
> > > QMMM: Smallest DeltaE = 0.2069E+03 DeltaP = 0.1570E+01 Step =
> 73
> > >
> > > How to solve this error??
> > >
> > > Thanks in advance..
> > > _______________________________________________
> > > AMBER mailing list
> > > AMBER.ambermd.org
> > > http://lists.ambermd.org/mailman/listinfo/amber
> > >
> >
> >
> >
> > --
> > ================================ Current Address =======================
> > Brian Radak : BioMaPS
> > Institute for Quantitative Biology
> > PhD candidate - York Research Group : Rutgers, The State
> > University of New Jersey
> > University of Minnesota - Twin Cities : Center for
> > Integrative Proteomics Room 308
> > Graduate Program in Chemical Physics : 174 Frelinghuysen Road,
> > Department of Chemistry : Piscataway, NJ
> > 08854-8066
> > radak004.umn.edu :
> > ====================================================================
> > _______________________________________________
> > AMBER mailing list
> > AMBER.ambermd.org
> > http://lists.ambermd.org/mailman/listinfo/amber
> >
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
>



-- 
================================ Current Address =======================
 Brian Radak                                            :     BioMaPS
Institute for Quantitative Biology
 PhD candidate - York Research Group       :     Rutgers, The State
University of New Jersey
 University of Minnesota - Twin Cities          :     Center for
Integrative Proteomics Room 308
 Graduate Program in Chemical Physics     :     174 Frelinghuysen Road,
 Department of Chemistry                          :     Piscataway, NJ
08854-8066
 radak004.umn.edu                                 :
====================================================================
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Wed Aug 06 2014 - 07:00:02 PDT
Custom Search