[gmx-users] box size and neighbor list cut-off

[Justin Lemkul]

On 8/17/15 2:02 PM, Johnny Lu wrote:
> Dear Users,
>
> Hi.
>
> Box vector length:
>        3.448, 3.448, 3.448 nm dodecahedron.
>        (Box volume is 28.98 nm^3)
>
> Diameter of tripeptide:
>        1.248 nm
>
> Volume of tripeptide:
>        0.35503 nm^3 (from genbox output)
>
> Short range Neighbor list cut-off:
>        1.1 nm (set manually by rlist=1.1)
>
> Other short range cut off (VdW and electrostatic):
>        1.0 nm
>
> According to 5.2 manual (manual-5.2-dev.pdf),
>        "This means that the length of each box vector must exceed the length
> of the macromolecule in the direction of that edge plus two times the
> cut-off radius R_c"
>
> Box vector length >= Diameter of peptide + 2 x cut-off
>        3.448 = 1.248 + 2 x 1.1
>
> Do I have to leave any leeway in the box size? (say, a box size of 3.5 nm)
>

The point is to set a minimum such that you avoid minimum image violations. 
This is tricky if the peptide might extend.

> I will run a NPT simulation (Berendsen barostat). Then continue the
> simulation with a NVT simulation. And finally continue the NVT simulation
> with a NVE simulation.
>
> Last time when I ran a double precision NVE simulation, I set rlist=1.3 and
> used a 4.848^3 nm^3 dodecahedron box. The energy conservation was good. If
> I use a 1.1nm rlist, will I get an energy conservation that is nearly as
> good as the rlist = 1.3nm simulation?
>

I would say let the Verlet algorithm tune rlist.  That's its job.

-Justin

-- 
==================================================

Justin A. Lemkul, Ph.D.
Ruth L. Kirschstein NRSA Postdoctoral Fellow

Department of Pharmaceutical Sciences
School of Pharmacy
Health Sciences Facility II, Room 629
University of Maryland, Baltimore
20 Penn St.
Baltimore, MD 21201

jalemkul at outerbanks.umaryland.edu | (410) 706-7441
http://mackerell.umaryland.edu/~jalemkul

==================================================