LAMMPS WWW Site - LAMMPS Documentation - LAMMPS Mailing List Archives
Re: [lammps-users] experiment on fix efield along with reax/c
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [lammps-users] experiment on fix efield along with reax/c


From: Mohammad Izadi <izadi0511@...24...>
Date: Sun, 27 Aug 2017 09:32:23 +0430

Dear Axel

 

You are right, about ReaxFF Fortran code experiment I didn’t bring any information.

Your suggestions are always interesting for me. Thanks a lot.

 

Best regard

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

Mohammad Ebrahim izadi,

Department of Chemistry,

Tehran University,

Islamic Republic of Iran,

Phone : +98 – 21 – 61113358

Fax :  +98 – 21 – 66409348


On Sun, Aug 27, 2017 at 12:36 AM, Axel Kohlmeyer <akohlmey@...24...> wrote:
below are comments on ray's remarks as well as on the original post.

On Sat, Aug 26, 2017 at 3:02 PM, Ray Shan <rshan@...1795...> wrote:

 

From: Mohammad Izadi <izadi0511@...1125.....>
Date: Saturday, August 26, 2017 at 11:09 AM
To: Axel Kohlmeyer <akohlmey@...24...>
Cc: "lammps-users@...42...e.net" <lammps-users@...4241......e.net>
Subject: [lammps-users] experiment on fix efield along with reax/c

 

Dear Axel



Frist of all, thanks for your benefit advices about testing the fix efield in lammps/reax/c. Following our previous discussions about fix efield in lammps/reax/c I setup some very simple systems as you said. The systems are as below:



1. Only one Na+1 atom is placed in the system and the fix efield is investigated when we have a L.J. potential.

2. Completely separated Na and O atoms with partial charges equal to that calculated by reax/c potential are placed in a system and this system is subjected for comparison the fix efield along with L.J. potential and reax/c potential in the Lammps software.  

3. Test the second experiment (in completely similar condition) in the ReaxFF Fortran code.



In all simulations I have used 0.01 volt per angstrom external electric field only in x direction and 0.01 K temperature. The input and output files are attached to the email. I calculate the force(s) on the atom(s) due to the external electric field with hand and by using the displacements from the simulation results I calculate the electric field energy imposed to the system then I compare my results with the results gained from the above simulations and the final results are as bottom:



1. The fix efield in the Lammps software along with both classical pair potentials (e.g. L.J.) and reactive potential (ReaxFF) act in the same manner absolutely equal to the thing mentioned in the Lammps manual. I think the external electric field is a classical homogenous electric field in the Lammps software.

This makes no sense at all. Your first style is pure LJ (lj/cut) which has no charge-dependent terms therefore any efield has no effect on pure LJ.  On the other hand, ReaxFF neutralizes the entire system then finds the equilibrium charges.  If you only have Na atoms in the system, then all of the Na atoms would have zero charges – which means any efield has no effect at all.


​well, one *could* do a simulation with atom style ​charge, assign charges, then use pair_style lj/cut (which will ignore the charges) and fix efield (which will use the charge information). whether this is a meaningful model, is a different question, but the conclusions are *extremely* obvious:
a) a pair style will compute the interactions for what it is written for (and what is documented). so lj/cut will compute a 12-6 lennard-jones. nothing more, nothing less.
b) fix efield does what is said in the docs, too, (and which i have pointed out repeatedly): it computes a force (E_xyz * q_i) for each atom i for each x-,y-, and z- component of the global E field vector.

 

2. In the Lammps software, in comparison with the classical potentials, when we use ReaxFF potential, we have smaller effects of the fix efield since the qeq/reax/c charge equilibration scheme produce partial charges on atoms during the simulations and this charges are very smaller than the initial charges at the beginning of the simulations but in the classical simulations we have constant charges (equal to the initial values) during the simulations. In a reax/c example, I set +1 and -1 charges on atoms but after the first step the new charges were approximately +0.3 and -0.3 respectively.

These results are expected and are reasonable.

3. When I use external efield in the ReaxFF Fortran code, It has some unpredictable results that are vague for me and also the results due to external efield are apparently different from the results gained from fix efield in the Lammps software in similar conditions. For example, the ratio of resulted delta_Epot in the ReaxFF to resulted delta_Epot in the Lammps package is about 120 for 100 steps. I will contact to ReaxFF developers about this subject.


​this is unscientific reporting of results: you are vague (what are "some unpredictable results"?) and provide no specifics or ways to reproduce this and not even quote the units of your observation. so any answer to that would have to be pure speculation.​ 
 

Do you mean you are using the stand-alone ReaxFF code from Prof. van Duin?  Make sure your efield has the same units.

Ray

Thank you for your beneficial reply to my questions. If you have any suggestion/correction or criticism on the above experiments and my descriptions, it's my pleasure.


​the obvious test that is missing here, is a test where fix efield is compared with a "real" electrostatic field. the latter can be modeled with explicit and non-modifiable charges. for example, consider a box, that is periodic in x-, and y- direction, but of fixed size in z-direction. now build a system that has a flat layer of atoms with positive or negative charges on the top or bottom side of the z-direction, respectively. now leave a little distance and then place a confining (LJ) wall on each side and do a slab system or liquid fil​m in between the confining walls. you set up the charges to correspond to the electrostatic field that you want to model with fix efield. mind you, the closer the charged points are to each other and thus the more points there are, the closer you get to the ideal scenario, but also the more computational cost, so there is an optimization there. 

at any rate, if the slab/film is thick enough and with a (model of) a substance that has a non-zero dielectric constant (in the model representation), you should see a difference in the effect of the electrostatic field as modeled by explicit point charges vs. fix efield. you can also quantify this by solving the poisson equation from a spatially (and time) averaged charge distribution information of your slab/film.

axel.


 



Thanks in advance

Best regard

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

Mohammad Ebrahim izadi,

Department of Chemistry,

Tehran University,

Islamic Republic of Iran,

Phone : +98 – 21 – 61113358

Fax :  +98 – 21 – 66409348


------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot
_______________________________________________
lammps-users mailing list
lammps-users@...655....net
https://lists.sourceforge.net/lists/listinfo/lammps-users




--
Dr. Axel Kohlmeyer  akohlmey@...24...  http://goo.gl/1wk0
College of Science & Technology, Temple University, Philadelphia PA, USA
International Centre for Theoretical Physics, Trieste. Italy.