Thanks for your advice. I used GB potential for liquid crystal simulation.
In experiment, a system confined in a cylinder wall anything looks natural. The wall is infinite compared with the liquid inside and will not change their position. If we want to model them in Lammps, the things turn to confused.
The IDs for liquid crystal inside, wall and entire system are "GB", "wall" and "AllCell" respectively. The wall and liquid crystal are described by two kinds of particles in my simulation.
In order to let the wall not move, we can use "fix W wall freezel" or "fix W wall setforce 0.0 0.0 0.0" and "velocity wall set 0.0 0.0 0.0". Maybe it also can be realized just by using the group ID setting in "fix npt GB npt/asphere". It is because the information of the wall will be always staying at first step with the command "fix npt GB npt/asphere". Am I right?
As to the infinite of the length of cylinder wall, it is a trouble waiting for salvations. Dose anybody has any experience on the confinement with different density changing as temperature. When the temperature decreases, the box length should decrease. How the redundant wall atoms go? They are necessary for high temperature.