Hello all,
I am trying to simulate liquid-solid phase transition in Al-Si alloy and in Al metal matrix, with CNT (carbon nanotube) inserted within the metal matrix. CNT is not getting distorted when I am using Al-Si alloy, but, CNT get distorted when I am inserting it within Al-matrix. I am giving below the codes for both the simulations:
units metal
atom_style full
dimension 3
boundary p p p
kspace_style none
pair_style hybrid/overlay lj/cut 2.5 tersoff meam
pair_coeff 1 3 lj/cut 0.035078 3.0135
pair_coeff * * tersoff SiC.tersoff NULL Si C
pair_coeff * * meam library.meam Al Si AlSi.meam Al Si NULL
timestep 0.0002
thermo 1000
dump 3 all xyz 1000 a*.xyz
velocity all create 300 82557 dist gaussian
thermo_style custom step temp vol press density ke pe etotal enthalpy
restart 10000 data.restart
fix 1 all npt temp 300 1000 0.001 iso 0 0 2
run 200000
write_data system.data
…
for Al-metal matrix
…
units metal
atom_style full
dimension 3
boundary p p p
kspace_style none
thermo 1000
dump 3 all xyz 1000 a*.xyz
velocity all create 800 102557 dist gaussian
thermo_style custom step temp vol press density ke pe etotal enthalpy
restart 10000 data.restart
fix 1 all npt temp 800 800 0.001 iso 0 0 2
run 50000
…
The same problem arises when I have used AIREBO potential for C-C interaction
Kindly help me to resolve the problem.
Well, this is typically a risky combination, so it’s likely that the behavior you observe will primarily depend on the validity of this choice, rather than being a LAMMPS issue itself. Each potential is fitted to specific systems under certain assumptions. Mixing them can easily introduce inconsistencies unless you understand how the potentials interact and/or validate your simulations carefully.
Thank you very much for your reply. From two papers I took the values for C-C and Al-Al LJ parameters which are the following:
ϵ(c-c) = 0.0296 eV and Ϭ(C-C)=3.407 Ang
ϵ(Al-Al) = 0.4157 eV and Ϭ(Al-Al)=2.62 Ang
Then using Lorentz-Berthelot (LB) Rule, I calculated the parameters for Al-C interaction. Moreover, the cutoff distance was chosen arbitrarily.
I have confusion whether hybrid or hybrid/overlay (as in the following code) can define the interaction of Al-Al, Al-Si and Si-Si:
pair_style hybrid/overlay lj/cut 2.5 tersoff meam
pair_coeff 1 3 lj/cut 0.035078 3.0135
pair_coeff * * tersoff SiC.tersoff NULL Si C
pair_coeff * * meam library.meam Al Si AlSi.meam Al Si NULL
Kindly suggest me the possible corrections in my code.
I want to inform that after increasing the Al-C distance, the distortion is not coming.
So obviously, those are not great parameters for each of the elements, since otherwise you would be using Lennard-Jones throughout. On top of that, if you create parameters from mixing, they are usually worse than parameters specifically optimized for a specific setup. Keep in mind, that with empirical models the parameters are not very transferable. Biomolecular force fields, for example can have tens of parameters for carbon for example, and yet they are restricted to ambient conditions and molecular systems in water as a solvent.
That is a very bad idea. Every setting and parameter should be chosen for a good reason. 2.5 Angstrom is less than your parameter for sigma, so it does not even include all the repulsive part of your LJ potential. That is an extremely bad choice. The cutoff you chose is commonly used with reduced units. Instead you should use a cutoff in the range of what is documented for the parameters you found. That would typically be in the range of 10-12 Angstrom these days.
It should be clear from the documentation. I must be pair style hybrid. With hybrid/overlay you can add interactions to others, e.g. Coulomb to a non-Coulomb like lj/cut. In your case is is very bad, since you are double counting the Si-Si interactions. When combining two many-body potentials, there must not be any overlap of atom types and all the mixed interactions must be handled by pairwise additive interactions like lj/cut. This is all clearly documented in the LAMMPS manual.
It is not my job to correct your mistakes. You have to make your own choices based on your own judgement. If I just tell you what to do, you won’t learn much and with the next problem just come back and ask again.
This is part of your research and not relevant to LAMMPS. @simongravelle has already pointed out how unreliable your approach to getting force field parameters is. There is no way of knowing a priori whether this is a meaningful result or just random. You will have to figure out was to confirm its correctness. The fact, that you get the expected result is desireable but no proof.
In particular: an lj/cut 2.5 cutoff is often used in units lj where the Lennard Jones sigma is about 1 unit, so the command results in a typical 2.5 sigma cutoff.
In your simulation, units metal means that length units are in angstrom and you are cutting off the LJ interaction at 2.5 angstrom. This truncates a significant amount of interaction and renders your simulation completely invalid. A reviewer would expect at least an 8 angstrom cutoff.
Moreover, the cutoff is a force field parameter, it is not something you can change freely and expect to get valid results. That is, if a set of LJ parameters is validated at (for example) 10 angstrom cutoff, and you try the same parameters with 8 angstrom cutoff, you are definitively simulating a different system, and very likely getting unusable results.