I encountered such a problem when simulating the interaction system of rigid body and explicit solvent particles:
I constructed a spherical structure of atomic stacking, half of which are type1 and the other half are type2. At the same time, a certain number of explicit solvent particles type3 are placed in the simulation box. I hope to make the forces on the left and right sides of this spherical structure asymmetric by setting the weak interaction between type1 and type3 (epsilon=0.1) and the strong interaction between type2 and type3 (epsilon=10.0), so that it moves in one direction. However, I found in the simulation that this rigid spherical structure only performs a rapid three-dimensional rotation at the initial position, and does not produce directional movement. How can this problem be solved? Here is my in file:
units lj
dimension 3
boundary p p p
atom_style atomic
read_data CPS-lammps.data
group Colloid type 1 2
group ExceptRigid type 3 4
pair_style lj/expand 1.12
pair_modify shift yes
pair_coeff * * 0.1 1.0 0.0 1.12
pair_coeff 1 3 0.1 1.0 0.0 1.12
pair_coeff 2 3 10.0 1.0 0.0 1.12
neighbor 1.0 bin
neigh_modify delay 0 every 1 check yes
neigh_modify exclude group Colloid Colloid
neigh_modify exclude group ExceptRigid ExceptRigid
thermo 100
thermo_style custom step temp ke pe etotal
dump 1 Colloid custom 2000 output.lammpstrj id type x y z ix iy iz
fix step1 Colloid rigid/nvt single temp 1.0 1.0 1.0
fix nve_1 ExceptRigid nve
fix step2 ExceptRigid langevin 1.0 1.0 1.0 35628736
timestep 0.01
run 500000
unfix step1
unfix nve_1
unfix step2
clear
At the same time, I found that the potential energy of the system is very low:
Per MPI rank memory allocation (min/avg/max) = 17.61 | 18.98 | 21.41 Mbytes
Step Temp KinEng PotEng TotEng
0 0.66678549 0.99833133 0.0047890597 1.0031204
100 0.9559068 1.4312125 0.00012851137 1.431341
200 0.99371995 1.4878274 0.00017593835 1.4880034
300 0.9979491 1.4941595 0.00021253792 1.494372
400 0.99992648 1.49712 0.00020112572 1.4973212
500 1.0003921 1.4978171 0.00018859753 1.4980057
600 0.99957758 1.4965977 0.00020192778 1.4967996
700 1.0004672 1.4979296 0.00016616739 1.4980958
800 1.0006566 1.4982132 0.00018545524 1.4983987
900 0.99951732 1.4965074 0.00017468191 1.4966821
1000 1.0016245 1.4996624 0.00021770173 1.4998801
I don’t even understand what the specific problem is, since there is no sketch or visualization describing your system and no convincing explanation why there should be some translation (and for which particles and how).
Fully periodic systems are generally expected to be position invariant (hence the subtraction of 3 degrees of freedom when computing the global temperature).
Furthermore, your interactions are all purely repulsive and I don’t see explicit initialization of velocities (unless that is already present in the data file).
Please note that for reduced units all thermodynamic data is normalized (thermo_modify norm yes) while for all other units settings it is by default not. Please see the LAMMPS manual for details.
Thank you very much for your reply. For my model, the solvent particles on the right have a stronger repulsive force on the rigid sphere. Should they move to the left as expected? If the velocity of the rigid sphere in its initial conformation is 0, will it also move under the action of asymmetric forces? Does this time scale need to be very long to be observable?
These are all questions about the science of your model and thus something you need to discuss with people that care about your research like your adviser or tutor or similar. It is not the job of this forum to explain your research to you. We focus on LAMMPS issues and there is no LAMMPS issue here.
The only thing that can be said be quickly glancing at your input and model is that all the LJ particles in your sphere have the same effective diameter (i.e. the same sigma) but different steepness of the potential (different epsilon), when interacting with the solvent. So the solvent “sees” the same size particles, but some of them are way softer than others. Since you have only repulsive interactions, those will only apply to the solvent atoms right next to them. I don’t expect that that should have a lasting impact. At best, the center of mass of the rigid object and the solvent would move a small distance until they are in equilibrium.
What is very unusual in your model is that you remove the solvent/solvent interactions from the neighbor list. So the solvent will for the most part behave like an ideal gas and only get repelled from the rigid object. That is not really a solvent then.
Please also, note that if we could predict the simulation results and answer your questions up front, then there would be no more need to do a simulation. Outside some unexpected choices, there doesn’t seem to by a major issue with the commands. They will do what you ask them to do. Whether this is the expected result is not a LAMMPS issue. Hence your original subject line, that there are abnormal interactions is clearly an incorrect assessment.
Thanks for your answer, I understand the problem. I initially thought there was something wrong with my lammps command.