search

prisms-center / CASMcode

First-principles statistical mechanical software for the study of multi-component crystalline solids
Other
105 stars 70 forks source link

displacement degree of freedom #247

Open wshao1995 opened 2 years ago

wshao1995 commented 2 years ago

Hi developers, I recently read an study of vibrational cluster expansion. DOI: 10.1103/PhysRevB.88.214111.

I have some questions about this article:

  1. This article contains an initial structure, according to the formula in the article, is its cluster known? If so, how should I import the known cluster information into CASM code?
  2. Author fitted the ECIs using 1484 configuration， where 1184 configurations contain homogeneous deformation, is this equivalent to the strain degrees of freedom? 300 supercells both contain the local ionic displacements and homogeneous strains，Is this equivalent to considering both the strain degrees of freedom and the displacement degrees of freedom?
  3. This article is mainly about anharmonic approximation. If I only consider (quasi-)harmonic approximation, do I only need to consider pair cluster?
  4. If it is for a given configuration, how should the bspecs.json file be written?

I hope I can get your help.

bpuchala commented 2 years ago

This article contains an initial structure, according to the formula in the article, is its cluster known? If so, how should I import the known cluster information into CASM code?

I'm not sure I understand this question, but perhaps it will be answered by the following.

Author fitted the ECIs using 1484 configuration， where 1184 configurations contain homogeneous deformation, is this equivalent to the strain degrees of freedom? 300 supercells both contain the local ionic displacements and homogeneous strains，Is this equivalent to considering both the strain degrees of freedom and the displacement degrees of freedom?

The homogeneous deformations are equivalent to just strain degrees of freedom. The other configurations include strain and displacement degrees of freedom.

Note however that the work in that paper uses a formulation for the Hamiltonian that is not included in the current public version of CASM. The cluster functions are constructed after removing the rigid rotation associated with the cluster, which is not currently done in the public version of CASM. Removing the rigid rotations before constructing the functions allows the fit to handle much larger atomic deformations.

This article is mainly about anharmonic approximation. If I only consider (quasi-)harmonic approximation, do I only need to consider pair cluster?

Yes.

If it is for a given configuration, how should the bspecs.json file be written?

A new project can be constructed using a particular configuration to construct a new prim.json file. The ccasm init method with the --write-prim,--confignames, --dofs options can be used to do this. For example: ccasm init --write-prim --confignames SCEL2_2_1_1_0_0_0/2 --dofs disp Hstrain would write a new prim.json file in the training_data directory for the SCEL2_2_1_1_0_0_0/2 configuration with only disp and Hstrain degrees of freedom, and not occupation degrees of freedom.

A new project can be constructed inside the existing project with the ccasm init --sub command, or the new prim.json file can be copied and a new project constructed elsewhere.

The bspecs.json file for the new project can be constructed following the format here. For disp and strain continuous DoF, the main parameter is global_max_poly_order, but you can also control the polynomial order by "branch" (number of sites in the cluster) using orbit_branch_max_poly_order.