LAMMPS WWW Site - LAMMPS Documentation - LAMMPS Mailing List Archives
Re: [lammps-users] Problem with a simulation
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [lammps-users] Problem with a simulation

From: Axel Kohlmeyer <akohlmey@...24...>
Date: Sun, 10 Sep 2017 16:50:35 -0400

On Sun, Sep 10, 2017 at 3:18 PM, A. M.M <alaa.murar@...24...> wrote:
بسم الله الرحمن الرحيم


First, please see the attached article (especially the part of the Dendrimer model and simulation).

​please note, that a) by posting this article in a public forum, you probably violated the licensing/redistribution terms under which you got access to it, and b)​ i have no time to learn, study and review your research. that is your job or something you need to discuss with your adviser. i can tell you what i know to be incorrect or unusual. it is ultimately your choice whether you stick with your settings or change them, and it is going to be ultimately you that has to justify your choices when you present your work.

Second, the geometry should be a 3-D shape as in the article.

​it *very* obviously is not. just visualize it and look at the data file, it is *obvious* to see, that there are many lines with the exact same coordinates.​
this is easily proven by adding the following command to your input: 

delete_atoms overlap 0.01 all all

this will delete all atoms that have a distance of less than 0.01 \AA. ​you'll see in the output that​ this deletes the number of atoms that i claimed are redundant.

The article method has no simulation boundaries, so no periodic bc, but I chose this as a test.

​as mentioned before, if you do not want periodic interactions, don't use periodic boundaries. shrinkwrap boundaries in LAMMPS are specifically designed to set this up in the most effective way.​ 
The coul/debye potential in the article has no cutoff radius, so I chose this cutoff which is equal to the length of the box of simulation.

​that makes no sense at all. just make some simple back-of-the-envelope considerations: a) if you don't want interactions with perio​dic images of your base molecule, your cutoff *MUST NOT* be longer than *HALF* the shortest end-to-end length of your box. it is a *RADIUS*. b) however, if you do want all interactions *within* your molecule, your cutoff needs to be *at most* the maximum possible *extent of your molecule*.

As for morse potential I chose the same cutoff as in the article. I chose lj/coul 1.0 1.0 1.0 for special bond because I need to calculate all interaction as in this article.
I chose the dielectric constant of the water which is the implicit solvent and coul/debye for pair style as in the article.

​this makes no sense at all. this way you are include the screening effect from the solvent *TWICE*.​
The choice of nsq came from a problem faced and this choice solved it, but I made some changes so I think the default one will work fine now. (I forgot return it to default one).

​this is a very bad approach to problem solving. you cannot just change options without understanding what they do until your input runs. what if it will run only for the wrong reasons? and thus result in a bogus trajectories. please keep in mind, that a input that does not crash is a necessary condition for a correct simulation, but not a sufficient one.​

I hope that I help you guess the situation.

​i have summarized what i think. it is now your job to turn this into a meaningful simulation. it looks to me, you should seek some proper tutoring from a local person, that has experience in MD simulations and statistical mechanics. if you struggle as much with such basic issues, you will be in deep trouble, if you actually move on to something more complex.



2017-09-10 19:06 GMT+03:00 Axel Kohlmeyer <akohlmey@...24...>:

On Sun, Sep 10, 2017 at 7:44 AM, A. M.M <alaa.murar@...24...> wrote:
بسم الله الرحمن الرحيم


I attached to you two file ( Configuration and Input ) for a simulation of biomolecule called  "G6 EDA-cored PAMAM Dendrimer".

I faced many problems when I run this simulation and I solved all of them, but now when I run this simulation and after the first step it print "temp = -nan and all of energies = -nan", and when I check the xyz output file it contains only the initial positions that I configured and all new positions are equal to -nan.

Could you help me for find the wrong in this simulation?

​please check your geometry. you have a whopping 438 (out of 510) redundant positions, leaving only 72 unique positions.
have you visualized your starting geometry? is it supposed to be "flat".

there are many more questionable choices. please explain:

- the size of your simulation cell. if you don't want your molecule to interact with its periodic images, why not simply use fixed or shrinkwrap boundaries?
- the choice of your cutoff for coul/debye. your choice of box dimension suggests, that you don't want interaction with periodic images, by this will call in *a lot* of those. this is also in contrast with your morse potential cutoff, which is not exactly on the long side at 9 \AA. 
- the choice of special bonds. the FENE potential usually expects a different choice.
- the reason for setting a dielectric constant != 1 *and* using the coul/debye pair style.
- the reason for using the nsq neighbor list build method over the default binning.


Check out the vibrant tech community on one of the world's most
engaging tech sites,!
lammps-users mailing list

Dr. Axel Kohlmeyer  akohlmey@...24...
College of Science & Technology, Temple University, Philadelphia PA, USA
International Centre for Theoretical Physics, Trieste. Italy.

Dr. Axel Kohlmeyer  akohlmey@...24...
College of Science & Technology, Temple University, Philadelphia PA, USA
International Centre for Theoretical Physics, Trieste. Italy.