[lammps-users] Metadynamics on cell vectors


Dear developers and users of Lammps:

I’m now trying to study the pressure-induced phase transition of CL-20 by using Metadynamics on the cell vectors. Plumed plug-in in lammps was employed, and Reaxff force field was used for interatomic interaction.

I have performed a NPT MD for 10 ps at 300K and 8Gpa, which is slightly higher than the experimental critical pressure of phase transition. Then metadynamics was employed to simulate the phase transition process. Each metastep consists of a NVT MD simulation of 1.0 ps. We hope that metadynamics can add a Gaussian potential to the system at each meta step and change the crystal cells, eventually leading to phase transition. However, no change in cell vector was observed. 

The related commands in input files for lammps and plumed are as followings:

Input for lammps :

fix             2 all nvt temp 300 300 20.0 tchain 1

thermo_style custom step temp press etotal pe ke vol density cella cellb cellc cellalpha cellbeta cellgamma

thermo 200

fix		pld all plumed plumedfile plumed.dat outfile p.log

run		100000

 

Input for plumed :

UNITS LENGTH=A ENERGY=kcal/mol

RESET_CELL TYPE=TRIANGULAR

cell: CELL

bias: METAD ARG=cell.ax,cell.bx,cell.by,cell.cx,cell.cy,cell.cz SIGMA=1,1,1,1,1,1 HEIGHT=5  NUMERICAL_DERIVATIVES PACE=200

 

Can someone tell me why Metadynamics doesn't change the cell vector or do I need to add extra commands to change it?

The version of lammps is 29-Oct-2020, and the example of plumed in lammps can be run correctly.

Thanks in advance.

Jidong


Dear developers and users of Lammps:

I’m now trying to study the pressure-induced phase transition of CL-20 by using Metadynamics on the cell vectors. Plumed plug-in in lammps was employed, and Reaxff force field was used for interatomic interaction.

[…]


 

Can someone tell me why Metadynamics doesn't change the cell vector or do I need to add extra commands to change it?

In order to change the simulation box dimensions, fix plumed has to be programmed accordingly.
However, when looking at the fix plumed sources, it communicates the box dimensions from LAMMPS to the Plumed library, but I cannot see any code for the opposite direction.
I am copying the developers that contributed the fix plumed code to LAMMPS. They may be able to provide more detailed information.

Ok, if fix plumed will not change the box directly, but instead contribute to the virial, then Jidong also has to change the input in order to see the effect from metadynamics on the cell.
The box would not change with fix nvt as shown, only with fix npt, or fix nph or by adding fix press/berendsen or similar.

Thank you Axel for forwarding this to me. There is indeed a bug in the off diagonal component of the virial. Therefore biasing off diagonal components of the cell matrix does not work properly. The fix is simple and I will send a pull request later today.

Best regards,

Pablo

Yes, you are right. Then the issue that Jidong is having is not related to the bug that I was referring to. I will take the opportunity to fix it nonetheless.

Pablo