Skip to content

Commit

Permalink
add documentation
Browse files Browse the repository at this point in the history
  • Loading branch information
QuantumChemist committed Oct 25, 2024
1 parent 1555529 commit 23c130a
Showing 1 changed file with 23 additions and 2 deletions.
25 changes: 23 additions & 2 deletions docs/user/generation/data.md
Original file line number Diff line number Diff line change
Expand Up @@ -10,7 +10,7 @@ and demonstrates that a robust database of **crystalline structures** for MLIP r
be build by generating the single-atom displaced supercells using `phonopy` and combining them with a set of rattled supercell
structures, generated from the same unit cell models.

## `phonopy` generated structures
## Single-atom displaced supercell structures

The single-atom displaced supercell structures used in `autoplex` are generated by [phonopy](https://phonopy.github.io/phonopy/vasp.html) VASP-related routines,
that are collected in the `dft_phonopy_gen_data` flow
Expand Down Expand Up @@ -45,7 +45,7 @@ Where `pre_xyz_files` can also take a train and test database as argument, e.g.
`autoplex` is equipped with a [DFTPhononMaker](#autoplex.data.phonons.flows.DFTPhononMaker) class that inherits from the `atomate2` [PhononMaker](https://materialsproject.github.io/atomate2/reference/atomate2.vasp.flows.phonons.PhononMaker.html#atomate2.vasp.flows.phonons.PhononMaker) with
specific VASP input adjustments to guarantee high quality fit data. It can be used to run individual and customized `phonopy` workflows to generate MLIP fit data.

## Rattled structures
## Rattled supercell structures

There are several ways available in `autoplex` to rattle supercell structures,
that are collected in the `dft_random_gen_data` flow
Expand Down Expand Up @@ -96,7 +96,28 @@ As a counterpart to the `DFTPhononMaker` for generating data, `autoplex` include
that can be used to construct customized randomized structures workflows.
`autoplex` provides a variety of [utility](#autoplex.data.common.utils) subroutines to further customize a workflow.
## Adjust supercell settings
You can adjust the supercell settings by passing a dictionary containing your specific supercell settings for each
MP-ID to `CompleteDFTvsMLBenchmarkWorkflow`, e.g. like:
```python
mp_id = "mp-22905"
supercell_settings = {
mp_id: {
"supercell_matrix": [[0, 2, 0], [0, 0, 2], [2, 0, 0]]
},
"min_length": 11,
"max_length": 25,
"max_atoms": 200,
}
complete_flow = CompleteDFTvsMLBenchmarkWorkflow(
...,
supercell_settings=supercell_settings,
...).make(...)
```
To keep the calculations consistent, this will adjust the settings of single-atom displaced and rattled supercells.
## VASP settings
Expand Down

0 comments on commit 23c130a

Please sign in to comment.