Relaxation refused to converge

[perfetti]

Dear Every one,
       I am trying to relax a CNT(10,0) with 10 copper chains on it. However, after long time relaxation, the structure failed to converge, and gave a meaningless structure at the end(not converged).  
       I am wondering if you can pointed out one way for me to get this structure converged, if not, can you give me some advice to let the CNT with 6 chains successfully relaxed?  
       The input script for CNT with 6 chains/10 chains have been attatched, the trajectory file for the CNT with 10 Cu chains is too large to upload, but I can send it by email at asking.
      

[zh]

Several aspects in the initial geometries may lead to slow convergence in your geometry optimization:
1) The Cu-Cu bond length is nearly equal to that of C-C bond length and it may be too short.
2) The Cu atoms are adsorbed on the top site of carbon atoms. Have you done some test calculations for the adsorption of single Cu atom on CNT to check the most stable adsorption site?
3) The CNT is covered by so many Cu chains. If the Cu chain has strong interaction with CNT, the CNT would be destroyed the covering of Cu chains.
4). You should optimize the structure of an isolated Cu chain to determine its lattice constant. And then, you choose a proper length for CNT and Cu chain to make them be matched according to the  lattice constants of CNT and Cu chain.

[perfetti]

Thanks Mr Zh.
       The above you said is aimed at convergence, since my other relaxations did converge,
could I say my other relaxations are correct? 

 

Several aspects in the initial geometries may lead to slow convergence in your geometry optimization:
1) The Cu-Cu bond length is nearly equal to that of C-C bond length and it may be too short.
2) The Cu atoms are adsorbed on the top site of carbon atoms. Have you done some test calculations for the adsorption of single Cu atom on CNT to check the most stable adsorption site?
3) The CNT is covered by so many Cu chains. If the Cu chain has strong interaction with CNT, the CNT would be destroyed the covering of Cu chains.
4). You should optimize the structure of an isolated Cu chain to determine its lattice constant. And then, you choose a proper length for CNT and Cu chain to make them be matched according to the  lattice constants of CNT and Cu chain.

[Nordland]

Hey.

I haven't look into in details, but you could try to use the FIRE method instead.

[perfetti]

Thanks Mr. Nordland.

I have changed the method to FIRE, but the force get as close as to 0.0702119 eV/Ang , and then it flips away to higher force.
 
It was on the 73 step, and I could see the configuration in trajectory files.

Could I use this configuration as a new start, or could I just use this configuration to do following calculations?

I couldn't find the way to throw this configuration into builder window, however.

Hey.

I haven't look into in details, but you could try to use the FIRE method instead.

[zh]

I have run one of your test files without modification.  The geometry optimization reached convergence at the 154th step and the "Max force" is  0.0495122 eV/Ang. The optimized structure is attached. The Cu chain exhibits zigzag shape rather the straight one.  As I mentioned in the previous reply, the initial atomic positions of Cu atoms on the surface CNT may be not properly set up.  For a free standing Cu chain, the zigzag one may be more stable the equally spaced one due to the Peierls distortion (or Perierls transition).

[perfetti]

Thanks, Mr Zh.
You really helped me out.

I just killed it before it has a chance to converge, maybe I cared it too much. Lol.
 

Thank you very much.

I have run one of your test files without modification.  The geometry optimization reached convergence at the 154th step and the "Max force" is  0.0495122 eV/Ang. The optimized structure is attached. The Cu chain exhibits zigzag shape rather the straight one.  As I mentioned in the previous reply, the initial atomic positions of Cu atoms on the surface CNT may be not properly set up.  For a free standing Cu chain, the zigzag one may be more stable the equally spaced one due to the Peierls distortion (or Perierls transition).