Re: [AMBER] Question about usage of external QM in QM/MM

From: Dr. Andreas W. Goetz <>
Date: Thu, 4 Aug 2016 12:17:15 -0700

Dear Fabricio,

Regarding triplet states: sqm (the semiempirical QM and DFTB implementation in Amber) at present can only handle closed shell systems. For open-shell systems you would need to use an external QM program, irrespective of the QM model that you plan to use.

Let me clarify QM/MM with external QM programs and PBC: PME is used for all MM particles but not for any electrostatic interactions with the QM region. Thus, if you use electronic embedding (which is default), only MM charges within the cutoff (qmcut) are included in the QM Hamiltonian. The QM region does not interact with particles beyond that cutoff. MM particles beyond this cutoff will see a vacuum bubble instead of QM atoms. This means that the electric field experienced by the QM region due its surrounding changes abruptly whenever MM particles enter/leave the cutoff region. Equivalent for MM particles entering/leaving the QM cutoff, leading to discontinuities in the energy surface.

The problem is less pronounced if you increase the QM/MM cutoff (qmcut). While it is not ideal, you are probably fine if you use a sufficiently large cutoff and thermostat.

If you use a non-periodic system with solvent shell and cutoff (MM and QM/MM) that is larger than your system, then all electrostatic interactions are included and there are no discontinuities. Amber employs a soft half-harmonic potential at the boundary to avoid solvent evaporation (no hard wall) so energy is conserved. There is a limit to the pair list size in sander, which prevents such calculations for larger systems. One thing to note is that the computational cost for the MM portion of the calculation can become significant for very large systems in combination with a small QM region and/or fast QM method/implementation.

All the best,

Dr. Andreas W. Goetz
Assistant Project Scientist
San Diego Supercomputer Center
Tel: +1-858-822-4771

> On Aug 4, 2016, at 10:59 AM, Fabrício Bracht <> wrote:
> Hello. I have a few doubts regarding the use of Gaussian to perform qm/mm.
> First let say that the choice of using an external qm involves the fact the
> the catalytic group is a copper complex in the triplet state, and thus, I
> am not sure whether semi-empirical methods would give good reliable
> results. But any comment on this would be appreciated.
> My question may be a silly one, but it is regarding the use of periodic
> conditions or not and the use of the large cutoff. I have performed QM/MM
> simulations before but never considered the use of non periodic systems.
> The amber 16 manual states that PBC are supported with extQM but that PME
> is not and the use of the cutoff leads to discontinuities in the energy
> surface. So the manual advises you to use non-periodic simulations. Well,
> I do not understand why that is a better choice. From what I understood,
> the use of a non periodic system means that there is no imaging and thus,
> every molecule that hits the border interacts with the border. Wouldn't
> this affect the energy conservation more than the discontinuity caused by
> the large cutoff in a periodic system?
> Do you have any recommendation as to how to setup the system. A have the
> structure from a very long classic MD simulation solvated in a cubic box.
> Thank you
> Fabrício Bracht
> _______________________________________________
> AMBER mailing list

AMBER mailing list
Received on Thu Aug 04 2016 - 12:30:02 PDT
Custom Search