Periodic Solids (UO2)

The UO2 example demonstrates a periodic solid workflow built from pre-existing CFG structures. It also uses NeuroChem/ANI for ML and a custom two-step VASP QM wrapper.

The files needed to run the example are located in examples/UO2.

Before You Run

This example is more specialized than the HIPPYNN examples.

Edit parsl_configs.py for the target cluster. The copied file is a generic CPU/GPU Slurm template and is not expected to run unchanged.
Keep custom_modules.py unless you replace the UO2 QM task. It defines the custom two-step VASP wrapper used by master_config.json.
Edit vasp_2step_config.json so both VASP input blocks and QM_run_command match the local VASP installation.
Edit ani_config.json for the NeuroChem/ANI training setup and make sure the NeuroChem dependencies are available.
Review the CFG files in fragment_library. These are pre-generated structures loaded by the builder rather than molecules assembled by ALF.

Workflow Pattern

Stage	Task
Builder	`alframework.builders.builders.simple_cfg_loader_task`
Sampler	`alframework.samplers.mlmd_sampling.simple_mlmd_sampling_task`
QM	`custom_modules.ase_calculator_task`
ML	`alframework.ml_interfaces.neurochem_interface.train_ANI_model_task`

The builder loads .cfg structures from fragment_library and applies the configured coordinate shake. Use this path when you already have periodic starting structures from another simulation or crystal-generation workflow.

The sampler uses use_potential_specific_code: "neurochem" and alframework.ml_interfaces.neurochem_interface.NeuroChemCalculator. It includes temperature and density fluctuation schedules for exploring the solid around the loaded configurations.

The QM task is intentionally custom. It runs one VASP input, removes WAVECAR files, runs a second VASP input, and records convergence from OUTCAR.

Recommended Bring-Up Commands

From examples/UO2:

python -m alframework --master master_config.json --test_builder
python -m alframework --master master_config.json --test_qm
python -m alframework --master master_config.json --test_ml
python -m alframework --master master_config.json --test_sampler
python -m alframework --master master_config.json

Configuration Files

File	Purpose
`master_config.json`	Connects the CFG loader, MLMD sampler, custom VASP task, ANI training task, output paths, queue controls, and Parsl configs.
`builder_config.json`	Points to the CFG structure library and coordinate-shake value.
`mlmd_config.json`	Defines MLMD schedule, density fluctuation, uncertainty thresholds, NeuroChem calculator hook, and metadata path.
`vasp_2step_config.json`	Defines the two VASP input stages used by `custom_modules.py`.
`ani_config.json`	Defines NeuroChem/ANI model and training settings.
`custom_modules.py`	Defines the custom two-step VASP QM task. This file is still required by the provided master config.
`parsl_configs.py`	Generic CPU/GPU Slurm template copied from the water example.

Expected Outputs

The run writes status.txt, h5store/data-*.h5, models/model-*, sampling/metadata-*.p, VASP scratch directories under QM_scratch_dir, and any optional plotting output configured in the master config.

See Builders for CFG loader details, ML Interfaces for NeuroChem/ANI notes, and QM Interfaces for the ASE/VASP integration pattern.