**Disclaimer: This task may be too "easy", but their are many possible modifications to increase its complexity/ difficulty. The main drawback of many of these modifications is that the design/ evaluation of the metric may become more difficult.**

### Task Track #1: Verifiable Scientific Workflow Task

#### Science domain: materials science
#### Science macro: first-principles materials design
#### Task name: determine ideal solar thermochemical water splitting material

#### Task description:

Identify the best perovskite candidate material for an isothermal two-step thermochemical water splitting process operating at 1400 ℃ with pressure swings between $10^{-5}$ and $10^{-4}$ bars. Consider the $LaCrO_{3-\delta}$, $LaMnO_{3-\delta}$, $LaFeO_{3-\delta}$, $LaCoO_{3-\delta}$, and $LaNiO_{3-\delta}$ compositions and perovskite cubic (space group #221), rhombohedral (#167), tetragonal (#99), and orthorhombic (#62) structures. Assuming no prior knowledge of these materials, estimate and rank their $H_2$ yield. Consider dilute neutral oxygen vacancy formation energies calculated using the `uma-s-1p2` foundation potential (FP) and configurational entropy in the model estimating $H_2$ yield.

#### End goal:

Estimate $H_2$ yield for each material under the stated conditions and rank them from lowest to highest performing.

#### Evaluation metric:

Possible metrics:

- Error in calculated dilute neutral oxygen vacancy formation energies (e.g., $|E_{calc} - E_{ref}|$)
- Error in estimated $H_2$ yield (e.g., $|Y_{calc} - Y_{ref}|$)
- Correctness of ranking (e.g., Kendall's $\tau$)

#### Simulation list:

- FP-driven structural relaxation
- Determination of ground-state polymorph
- FP-driven dilute (and neutral) oxygen vacancy calculations
- Construction of defect thermodynamic model to prediction $H_2$ yield

#### Workflow:

1) Pull the 4 relevant perovskite structure prototypes from Materials Project (MP)
2) Decorate these $ABO_3$ structures with the appropriate $A$ and $B$ site elements
3) Run FP-driven relaxation with fixed symmetry
4) For each composition, determine the 0 K ground-state structure to use in subsequent analysis
5) For each composition, using the ground-state structure, determine dilute neutral oxygen vacancy formation energies
	1) Construct various supercell sizes
	2) Enumerate all unique single oxygen vacancies
	3) Run FP-driven relaxations with fixed cell (only allowing atom positions to relax)
	4) Calculate oxygen vacancy formation energies (for the chemical potential of oxygen, use the `omat-24` DFT calculated value with the appropriate $O_2$ overbinding correction)
	5) Comparing oxygen vacancy formation energies vs 1 / # atoms in supercell, project out to y-intercept to determine dilute neutral oxygen vacancy formation energy for each composition
6) Taking the dilute neutral oxygen vacancy formation energies and assuming ideal entropy for the defects, calculate $\Delta \delta = \delta^{red} - \delta^{ox}$ and convert to $H_2$ yield in μmol/g of material

#### Constraint:

- `ase` python package
- `pymatgen` python package
- `mp-api` python package/ access to MP database
- `fairchem-core` python package/ access to `uma-s-1p2` model

#### Dataset:

- at present, I believe as long as the agent has access to MP and the fairchem ecosystem (python package, hugging face datasets and model), no explicit dataset needs to be provided