Thanks for the reply. There models I used have the same sizes with the paper, containing 1152, 9216, and 73728 atoms. All the system became triclinic when cooled down.
Do you have any advice on energy minimization? Actually, Zhong minimized their system by using stress-controlled conjugate gradient energy minimization. Is there any lammps command to realize this, i.e. stress-controlled cg? I minimized the system by min_style cg and fix box/relax command, different from Zhong.