Problem with heating a non-periodic bicrystal to a certain temperature

Dear all lammps-users:

  As the title mentioned, my problem is about heating a non-periodic bicrystal to a certain temperature.

  Firstly, i constructed a bicrystal system with boundary condition(bc) 'p s p'(i.e. two free surfaces at the top and bottom of the bicrystal, and one grain boundary between two grains), and the dimensions for each direction are X-dimen=33.39, Y-dimen=70.4, Z-dimen=33.39.
  Secondly, "displace_atoms","fix box/reax" and "minimize" commands are used to enforce the bicrystal to reax to get a local minimum grain boundary energy, undoubtedly these commands will cause the chang of Y-dimen for non-periodic bc, and that is just why atoms arrangement of the bicrystal with minimun gb energy is slightly different from the initiallly constructed bicrystal.
  Thirdly, i dumped the bicrystal with minimun gb energy and heated it to a certain temperature by the "fix nvt" command, then the problem came.
  By observing atoms structure of the bicrystal visualized by the 'ATOMEYE', i found the grain boundary and the two free surface gradully disappeared with the process of heating, and finally there was no boundry in the bicrystal system, but this kind of phenomenon does not happen if i set the bc of the bicrystal as 'p p p'. So i think it is the non-periodic boundary condition in Y-direction causing the problem, is that ture? If yes, then how can i heat the bicrystal with bc 'p s p' without appearance of boundaries? Could anyone give me some ideas about this problem?

Dear all lammps-users:

  As the title mentioned, my problem is about heating a non-periodic bicrystal to a certain temperature.

  Firstly, i constructed a bicrystal system with boundary condition(bc) 'p s p'(i.e. two free surfaces at the top and bottom of the bicrystal, and one grain boundary between two grains), and the dimensions for each direction are X-dimen=33.39, Y-dimen=70.4, Z-dimen=33.39.
  Secondly, "displace_atoms","fix box/reax" and "minimize" commands are used to enforce the bicrystal to reax to get a local minimum grain boundary energy, undoubtedly these commands will cause the chang of Y-dimen for non-periodic bc, and that is just why atoms arrangement of the bicrystal with minimun gb energy is slightly different from the initiallly constructed bicrystal.
  Thirdly, i dumped the bicrystal with minimun gb energy and heated it to a certain temperature by the "fix nvt" command, then the problem came.
  By observing atoms structure of the bicrystal visualized by the 'ATOMEYE', i found the grain boundary and the two free surface gradully disappeared with the process of heating, and finally there was no boundry in the bicrystal system, but this kind of phenomenon does not happen if i set the bc of the bicrystal as 'p p p'. So i think it is the non-periodic boundary condition in Y-direction causing the problem, is that ture?

i would describe it differently. if your grain boundary vanishes upon
heating, that effectively means that your crystal is melting. now why
it doesn't happen with periodic boundary conditions, may be either due
to the open surface reducing the melting point,
or that the melting is delayed due to lack of nucleation. this is a
quite common phenomenon in molecular simulations.

If yes, then how can i heat the bicrystal with bc 'p s p' without appearance of boundaries?

you should rather ask, 'why does it melt?'. perhaps there are some
issues with your simulation or the choice of parameters or other
things, that we don't know about.

axel.

i would describe it differently. if your grain boundary vanishes upon
heating, that effectively means that your crystal is melting. now why
it doesn't happen with periodic boundary conditions, may be either due
to the open surface reducing the melting point,
or that the melting is delayed due to lack of nucleation. this is a
quite common phenomenon in molecular simulations.

Said another way, the periodicity constrains the defect (grain boundary
in this case) so it is not free to relax.

Steve

Dear Axel and Steve:
  Thanks for your response.
    According to your analysis, maybe the issuses with my simulation used to get minimum gb energy of bicrystal with 'psp', or the command 'fix nvt' cann't be used to heat the bicrystal with 'psp' as used to heat the bicrystal with 'ppp'. For the simulaton of gb energy of bicrystal, i am sure my input script for the 'ppp' boundary condition is right, and the input script for the 'psp' bicystal is the one, which i just make some changes to the script for the 'ppp' bicrystal, here i attache these two scritpts, please help me to check where is the problem with the 'psp' script. As analysed by Steve, melting of grain boundary is quite common in molecular simulations, so it seems quiet hard to heat the 'psp' bicrystal without boundary vanishing, but i believe there must be way to sovle this problem.

gb energy.ppp (9.13 KB)

gb energy.psp (9.5 KB)

No. What Steve and I are saying is that it is more likely your ppp boundary calculation that is more wrong. It has nothing to do with the input script as such but the statistical mechanics of your model. Axel.

Axel, about the simulaton of ppp boundary calculation, i want to say gb energies of different orienations calculated by my model are same as that calculated by other moedls, e.g. M. A. Tschopp. Would you please take a look at the two models posted in my last mail? For me, i just can't find where the problem is in the ppp model. According to the visualized picture of the atoms arrangements of both the psp and ppp boundaries, i found that for the psp boundary atoms at the middle grain boudary are less disordered than that for the ppp boundary(the attached two pictures show that), so i think maybe that's just why the psp boundary will melt when heated. I don't quite understand the statistical mechanics you mentioned, that should be relative to 'minimize','compute pe/atom' or 'fix box/reax'.

psp.jpg

ppp.jpg