Hi Da, I think if you are interested in the self-diffusion of atoms in interfacial region, you can just group the atoms in that area and then simulate a long enough time. During the diffusion process, the atoms will probably move out of the region you defined, but still they belong to the region you defined before diffusion (LAMMPS can do that). The warning is that you need average the MSD overall multiple time origins to obtain statistical accurate results (you can find more details in this document: http://utkstair.org/clausius/docs/mse614/pdf/diffusion_intro_v02.pdf). Thanks
Jiaqi