Particle size (volume) limit in fix particletemplate/sphere

msbentley's picture
Submitted by msbentley on Mon, 08/29/2011 - 17:58

Hi all,

I have so far been running LIGGGHTS with more-or-less "large" particles (mm/cm), but now want to move towards the size that interests me, around ~1 µm or so. However, if I try to use fix particletemplate/sphere and fix particledistribution/discrete to generate small particles I get an error which seems to arise from a hard-coded limit:

if(volume_expect < 1e-12) error->all("Fix template/sphere: Volume expectancy too small");

in fix_template_sphere.cpp.

Is there a fundamental reason for this limit, or is it an arbitrary sanity check assuming most people are interested in larger particles?

As a more general question then, are there problems I should expect modelling the behaviour of small particles?

Thanks! Mark

Forums:

ckloss's picture

ckloss | Tue, 08/30/2011 - 09:02

Hi Mark,

yes there is a fundamental limit - which is machine precision :-)
The obvious solution would be (a) coarse graining or (b) using another unit system such as cgs if you want to stick to small particles

>>As a more general question then, are there problems I should expect modelling the behaviour of small particles?
expect e.g. dominant cohesive/van der walls forces which makes it difficult to model

Christoph

msbentley's picture

msbentley | Tue, 08/30/2011 - 10:23

Thanks for the suggestion on units, I'll do that! Yes, cohesive and VdW forces are where the fun starts for me - especially when I turn down gravity and even large particles show these behaviours :)

I assume one can overlay the LAMMPS VdW implementation over the granular particles to do hybrid simulations including VdW attractions?

Cheers, Mark

ckloss's picture

ckloss | Tue, 08/30/2011 - 14:28

>>I assume one can overlay the LAMMPS VdW implementation over the granular particles
To my modest knowledge this is not possible since LAMMPS VdW is an atomistic model whereas for granular particles of mascroscopic size this should be some kind of "subgrid model". However, it would be possible to use the LAMMPS VdW model to developsuch a model.
But I am sure there are also models readily available in the literature

evansmuts | Thu, 09/22/2011 - 12:35

Hi Mark

I am also interested in looking at similar systems to you (i.e. particles around the 1 µm size) and so this issue of the particle size limit and combination of granular and vdW forces is also important to me. I am also interested in coupling the particles with fluid motion.

Now that I have moved to the latest version of LIGGGHTS and CFDEM, I have been hit with this particle size limit issue (it worked fine in the older versions). I haven't figured this one out yet. Looks like I will have to change unit systems.

However, I have managed to overlay the vdW forces (from LAMMPS) onto the granular atoms (LIGGGHTS). The hybrid/overlay command from LAMMPS works fine. I am still busy sorting out the material constants, so I haven't produced any usable data just yet. Also, my system is still pretty dilute as I am still setting up my model, but I have managed to get a stable model while including these forces. I am working on pushing up the concentrations at the moment. So it is possible.

Cheers
Evan

yangdm | Fri, 09/09/2011 - 18:10

In conventional granular methods, it is OK to model particles with a diameter in microns, but the model would run very slowly because very small timestep is required to achieve a computation stability.

My question is if MD could simulate the dynamics between atoms, why it could not simulate the interaction/contact between the particles at micron scale???

ckloss's picture

ckloss | Fri, 09/09/2011 - 18:15

>>My question is if MD could simulate the dynamics between atoms, why it could not simulate the
>>interaction/contact between the particles at micron scale???
there are groups which try to go in this direction, e.g. Marcus Bannerman (group of Prof. Pöschel/Univ Erlangen)
But to my knowledge true MD for a full contact of larger particles is hard because the largest systems that can be simulated with MD at the moment have the size of Lincoln's nose on a penny

Christoph

RAJ KASHYAP's picture

RAJ KASHYAP | Fri, 09/18/2015 - 09:36

Hye sir,

I am dealing with micro range particles.When I change the gravity value 9.81 to 981 to chage the unit from si to cgs,particles move with fast speed and leave the simulation domain.
can any one please sort it out.

Raj Kashyap

Daniel Queteschiner | Fri, 09/18/2015 - 14:42

Did you convert density values from kg/m³ to g/cm³?

RAJ KASHYAP's picture

RAJ KASHYAP | Sat, 09/19/2015 - 11:37

thanks a lot sir. I just forget to change the units of density. You judge accurately.
But after converting all units to cgs, the program runs like tortoise speed . Can u suggest something.?

Raj Kashyap

Daniel Queteschiner | Sat, 09/19/2015 - 13:27

I would recheck the skin distance specified in the neighbor command (crucial factors are: neighbors per particle and the frequency of neighborlist rebuilds) and the timestep used in the simulation.

SHUBHAM AGARWAL | Thu, 09/24/2020 - 22:35

Hi
I followed this interesting thread. I have a problem vary similar to this. I have to model fibers with particle diameter (~32 microns). As suggested, I converted the units to CGS, so I get the diameter of 32e-4.

I am getting the following error:
"Too many neighbor bins"

I am using the following bin setting:
neighbor 32.32e-4 bin
neigh_modify delay 0

Kindly help me with this

Thanks

mschramm | Fri, 09/25/2020 - 18:39

Try increasing your bin size.
neighbor 32.32e-4 bin
This command adds a "skin" to the pairwise interaction distance. So you are making your binsize to be, about, 2r.
You can specifiy what your bin size is with the following

neigh_modify delay 0 binsize 96.96e-4

This sets your bin sizes to be 3r.

SHUBHAM AGARWAL | Wed, 10/07/2020 - 07:43

I tried that, but now I am gettin the following error
ERROR on proc 2: Too many neighbor bins