On Mon, May 11, 2015 at 4:43 AM, Marc van der Kamp <marcvanderkamp.gmail.com
> wrote:
> Dear Robert,
>
> I didn't write the manual (nor am I a developer), but here some comments
> that may clarify things:
>
> On 11 May 2015 at 07:13, Robert Molt <rwmolt07.gmail.com> wrote:
>
> > Good morning,
> >
> > I have some confusions on the qmcut keyword. On p.138, it states, in
> > describing the keyword: " By default this is the same as the value of
> > cut chosen for the classical region, and the default generally does not
> > need to be changed."
> >
>
> This is the case for when PBCs are used, I think - then one can say that
> qmcut doesn't have to be changed from the classical cut-off.
>
I would agree here. I would actually go so far as to say "qmcut should
rarely have to be changed". For non-periodic systems, I would suggest
always setting the cutoff to larger than the system size to avoid artifacts
in truncating the long-range electrostatic potential. If you simply can't
due to system size, then you may need to make some (carefully
investigated/validated) cutoff choice, but keeping the qmcut and cut values
equal seems best to me.
>
> > On p.149, it states:
> >
> > " A cutoff is applied to QM/MM interactions in QM/MM simulations using
> > electrostatic embedding with and without PBCs. This leads to
> > discontinuities in the potential energy surface and poor energy
> > conservation. In the case of QM/MM simulations without PBCs, this cutoff
> > (qmcut variable in the &qmmm namelist) can be set to a number that is
> > larger than the simulated system, thus effectively not applying a
> > cutoff. This is recommended."
> >
> > I thus infer that this addendum is an exception to the rule of thumb on
> > p.138.
> >
>
> Note "In the case of QM/MM simulations without PBCs" here.
>
>
> >
> > However, this leads to a confusion for me.
> >
> > 1.) The note is in reference to usage of Gaussian as a QM software. The
> > second quote above would be seemingly true of ANY software (if you have
> > a cutoff without a switching function, there will be a discontinuity).
> > Am I mistaken?
> >
> > Yes, it would be the case for any software, as well as Amber's built in
> (sqm) semi-empirical methods.
> But if you have PBCs with PME, this will work for QM/MM as well.
>
There is a caveat with QM/MM with PME -- the reciprocal space sum is done
using Mulliken charges, while the direct space sum is done using the
density matrix, IIRC. This is discussed in some detail in the Amber manual
(see section 1.4 of the QM/MM chapter -- chapter 10 in the Amber 15 manual).
> 2.) I am confused about setting qmcut to a number larger than the
> > simulated system. Doesn't this result in a "QM region + cutoff larger
> > than box" error, inherently?
> >
> >
> I suppose it would (easy to test if you're curious), but ONLY if you
> enforce PBCs, in which case you wouldn't really need to set qmcut to a
> large number.
>
>
Yes, it would be a problem (minimum image convention only works if the
cutoff is smaller than half the smallest box length), but as Marc mentions,
only for PBCs.
All the best,
Jason
--
Jason M. Swails
BioMaPS,
Rutgers University
Postdoctoral Researcher
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Mon May 11 2015 - 05:30:02 PDT
