CP2K Cell Boundary Coordinates

[Dev Kumar]

Hi Experts, 

I am new cp2k user. I am trying to simulate a MD simulation on a metallic alloy and it keeps failing to start a NPT MD. I have a question about the structure boundaries in CP2K. I created the initial box in lammps and visualizing in ovito. Below are the box sizes. Am I converting it correctly for the cp2k simulation. 

# LAMMPS data file written by OVITO

208 atoms

3 atom types

-0.104644 20.6326618 xlo xhi

-19.8380928 19.8555032 ylo yhi

-0.0701645 2.082567 zlo zhi

! CP2K Format

 

 &CELL

      A   21.4280004501         0.0000000000         0.0000000000

      B   0.0000000000         42.9860000610         0.0000000000

      C   0.0000000000         0.0000000000        4.2329998016

      PERIODIC  XYZ 

    &END CELL

    &TOPOLOGY

      COORD_FILE_NAME AlCuMg.xyz

      COORD_FILE_FORMAT XYZ

    &END

Thanks

Dev Kumar

Brown University

Providence, RI 

[Matt Watkins]

I'm afraid I can't remember all Lammps' conventions - it would also depend on system of units used in the data file etc

Can you get ovito to also output a PDB or POSCAR or something you can check against.

Also, If you are running DFT you will want something closer to a cubic cell if possible - and it might need to be quite large if your allow is metallic...

Matt

[Dev Kumar]

HI Matt, 

Thanks for your reply. I attached the POSCAR. I am new to cp2k and I assumed that cp2k is taking the angles of the box 90 degrees and cubic. This is just my lack of knowledge. This is cubic fcc cell, I can give ALPHA_BETA_GAMMA 90.00000000 90.00000000  90.00000000   SYMMETRY CUBIC inside cell. I will try that out. Although I will still need to make sure cp2k taking all the atoms inside the boundary. I may need to leave some vacuum sometimes, so I should be able to manipulate the box size. Please let me know you comments. 

When I look at the output position file it does look like its printing out from the initial position I started with, but becomes mushy after few steps of geometry optimization. 

[Matt Watkins]

Well before you run NPT you need to check that you can converge the electron structure and that there are no large forces present in the system.
As for any MD I'd suggest a typical workflow would be at least a rough geometry optimization. Then NVT then switch to NPT.

NPT might need rather tight settings to have a well converged stress tensor.

Matt

[Dev Kumar]

Thanks for responding. Thats standard steps we use in LAMMPS MD simulations also. I will try it out. 

[Dev Kumar]

Hi, 

I tried a NPT MD , but it fails with an errors saying, An invalid set of cell vectors was specified. The determinant  det(h) is too small. I am attaching the input files, logs and other files. Will someone have a look to point out the error in the input file? 

Thanks

Dev

[Nam Tran]

Hi,

Why the cell vectors in your POSCAR are different with your CP2K input file?

Also it is not CUBIC cell.

Nam