Re: [ESPResSo-users] dielectric interfaces simulation

[Stefan Kesselheim]

Hi Jimmy,
welcome to the Espresso users. Please register for the mailing list, if you 
have not done that already.

On Feb 6, 2014, at 12:22 AM, Xikai Jiang <address@hidden> wrote:

> Hello everyone:
> 
> I'm learning Espresso and very interested in the dielectric interfaces 
> simulations. Currently I'm trying to simulate two dielectric surfaces with 
> non-zero electric potential difference between them, but I got some questions:
> 
> 1st  question: 
> To simulate a non-zero electrical potential difference between two dielectric 
> surfaces, do I need to input the external electric field (from the solution 
> of Laplace equation) just to every surface elements? or to every discrete 
> points in the domain that confined by two dielectric surfaces?

Which algorithm are you using? Induced Charges or Image Charges. Details are 
found here:
http://www.mdpi.com/1099-4300/15/11/4569
I'm now answering for the case with Induced Charges.

> 2nd question:
> I've done a simple case of two parallel plates, which are separated by a 
> certain distance in z-direction. Two charges are put between the plates and I 
> measure the force they experience.
> 
> When the potential difference between two plates is zero, the MD results 
> matches very well with the analytical solution.

That is nice. I had trouble doing that with ICCP3m because the P3M tuning works 
not very well. This is because the system is not at all homogenous and you also 
don't know the charges that will appear.
> 
> When the potential difference between two plates is set to 1V. I applied the 
> solution (electric field at every atoms of both plates with nonzero potential 
> difference) of Laplace equation to every surface elements. The MD results I 
> got are about 4% percent off the analytical solutions. Furthermore, when I 
> vary the box size in z-direction, the force I measured increases linearly as 
> box size increases. Does anyone knows why this happens?

You can superimpose a solution of the Laplace equation and the ICCP3M result. 
For two plates this is just a homogenous field of magnitude V/L. But you must 
not apply this field to the induced charges. I'm not quite sure what it would 
do, but you can think of it as follows: You need the induced charges to keep 
the potential created by the charges in the system constant, and on top of that 
you can stack a second potential which creates the potential difference. But 
that one has already the right boundary conditions.

By the way: The analytic solution should be that of 0V plus V/L. Right?

> In all above simulations, I used P3M solver for electrostatic interactions 
> with periodicity 1 1 1.
What about ELC? Probably not...

I hope that helps
Cheers
Stefan