Re: [ESPResSo] Question about P3M and MMM2D

[Axel Arnold]

Am Freitag, 9. Oktober 2009 18:10 schrieb Salvador Herrera Velarde:
> Hi,
>
> I am interested in simulate a charged bilayer membrane. First, I need a 2D
> electrostatic method + NPT ensamble. If I understood well, this is only
> possible with MMM2D. So, I put the center of the bilayer as close as 
> possible to the bottom of the simulation box and choose and appropiete
> value for layered cell (>4). Then, I need to calculate the surface tension
> of the membrane (I want to obtain the tensionless state) but when I tried
> with stress_tensor I got the next message:
>
> WARNING: pressure calculated, but MMM2D pressure not implemented => stress
> tensor of membranes
>
> Is there another way in ESPresSO to calculate the stress_tensor or the
> surface tension?

Unfortunately, no. There is (to my knowledge) currently no 2D electrostatics 
methods for which the pressure tensor is known. The problem with 
electrostatics in general is that a virial approach does not work, since 
the "pair forces" in fact also contain contributions from periodic images. 
Therefore, a virial sum would need to be redone analytically, and as far as I 
know that has only been published for P3M. We also tried to determine 
something like the pressure tensor for ELC long time ago, but gave up... With 
MMM2D determining the pressure tensor might be easier, but still requires 
quite some analytical work, which simply hasn't been done yet.

> Besides, once I know the tensionless state of the membrane, I want to use
> the P3M + ELC method for the same system. However, I am not sure what is
> the criterion to choose the best value for the parameter "skin". I run some
> test simulations in my desktop computer
>
> #P3M + ELC  skin 0.4
>    background_errors 0 {001 number of cells 1 is smaller than minimum 8
> (interaction range too large or max_num_cells too small)}


Do you have any other (short range) interactions in your system? Normally, the 
number of cells gets too small not if the skin is too small, but rather, if 
it is too large. However, if no short ranged interaction is on at all, then 
the number of cells is 1, and P3M tuning fails. In any case, you should tune 
P3M only _after_ switching on all interactions you plan to use, because the 
speed measurements obviously depend also on the other other interactions, and 
the effects are not simply additive.

>
> #P3M + ELC  skin 1.0
>    {coulomb 1.55 p3m 6.562365701286867 32 6 0.4436576322607176
> 7.641701920650234e-5} {coulomb epsilon  metallic n_interpol 32768 mesh_off
> 0.5 0.5 0.5} {coulomb elc 0.0001 13.808598559818787 0.06611687661638957}
>
> #P3M + ELC  skin 1.5
>    {coulomb 1.55 p3m 6.062365701286867 32 6 0.48136997545253307
> 9.22176560109132e-5} {coulomb epsilon  metallic n_interpol 32768 mesh_off
> 0.5 0.5 0.5} {coulomb elc 0.0001 13.808598559818787 0.06611687661638957}
>
> #P3M + ELC  skin 2.0
>    {coulomb 1.55 p3m 5.562365701286867 32 7 0.5259642582393248
> 9.693819928448824e-5} {coulomb epsilon  metallic n_interpol 32768 mesh_off
> 0.5 0.5 0.5} {coulomb elc 0.0001 13.808598559818787
>
> Should I use the minimum value of skin where the system in well behaved (it
> does not crash)? any advice how should I choose this value?
> Furthermore, If I choose a specific value for the skin using MMM2D (P3M +
> ELC), the same value can be used for MMM2D (P3M+ELC) and get the same
> results ?

The skin value never enters any physical results, it is purely a tuning 
parameter of the Verlet list. Typically, 0.4*sigma, where sigma is the 
particle diameter, is a good value.

Axel

-- 
Dr. Axel Arnold 
Fraunhofer SCAI
Schloss Birlinghoven, 53754 Sankt Augustin, Germany
Tel: +49 2241 14 2575