High-Throughput_Workflow_for_Polaron-Resolved_Modeling_of_Reduced_Ceria_dataset

# Oxygen vacancies and Ce3+ polarons on CeO2 slabs: DFT and MACE dataset Companion dataset to the article by Royo et al. "High-Throughput Workflow for Polaron-Resolved Modeling of Reduced Ceria". The dataset contains the atomic structures (POSCAR, CONTCAR), DFT+U energies and site-resolved magnetisations, and the MACE pre-relaxed structures used to down-select the configurations submitted to DFT. Configurations were generated with **TAPAS** (Topology-Aware PolaronArrangement Sampler): https://github.com/pafervi/TAPAS . ## Scope - Surfaces: CeO2(110) p(1x1) Ce32O64, and CeO2(111) p(2x2) Ce28O56.- Oxygen vacancy counts (nVO): 1, 2, 3, 4, 6.- For every (surface, nVO): the full set of symmetry-inequivalent configurations produced by TAPAS, each with its MACE pre-relaxation summary and its DFT+U result. ## Folder layout ```zenodo_data/├── surface_110/│ ├── mace/│ │ ├── {1,2,3,4,6}vO/│ │ │ └── cfg_NNNNN/│ │ │ ├── POSCAR # as-generated by TAPAS (La proxy)│ │ │ ├── CONTCAR # after MACE pre-relaxation│ │ │ ├── config_info.txt # TAPAS provenance block│ │ │ └── mace_opt.summary.txt # model, device, final E, fmax│ └── vasp/│ ├── inputs_vasp/│ │ ├── INCAR│ │ ├── KPOINTS│ │ └── POTCAR_info.txt # POTCAR identification (see below)│ ├── {1,2,3,4,6}vO/│ │ └── cfg_NNNNN/│ │ ├── POSCAR # input to VASP (La back-substituted to Ce)│ │ ├── CONTCAR # DFT+U optimised│ │ ├── OSZICAR│ │ ├── energy_and_magnetizations.txt│ │ └── config_info.txt└── surface_111/ └── (same layout as surface_110)``` The `cfg_NNNNN` indices are consistent between the `mace/` and `vasp/`subtrees of the same surface, so pairs (MACE pre-relaxation, DFT+Urefinement) can be matched directly by folder name. ## File formats - **POSCAR / CONTCAR**: standard VASP5 format, direct coordinates, selective dynamics not used.- **config_info.txt**: plain-text provenance block written by TAPAS. For each configuration it records the parent POSCAR, nVO, pair_type, dedup hash, vacancy indices (0-based in the original centered slab), the Ce3+ sites represented as La in the POSCAR, the NN and NNN pools per vacancy, and the La indices in the written POSCAR (0- and 1-based). The schema is documented in the TAPAS repository.- **mace_opt.summary.txt**: path to the MACE model, device, final energy in eV and final maximum force in eV/A.- **energy_and_magnetizations.txt**: extracted from the OUTCAR at the final ionic step. Contains the convergence flag, number of ionic steps, TOTEN, E(sigma to 0), total magnetisation in muB, and the per-ion (s, p, d, f, tot) magnetisation projection. The two Ce3+ polarons per vacancy are the Ce rows with tot close to 1 muB.- **OSZICAR**: VASP electronic/ionic step log, kept for ionic convergence inspection. ## DFT setup (summary; full input files in `inputs_vasp/`) - Code: VASP 6.x- Functional: PBE + U (Dudarev, LDAUTYPE = 2), Ueff(Ce 4f) = 4.5 eV- Plane-wave cutoff: ENCUT = 420 eV- k-points: see KPOINTS- Spin-polarised (ISPIN = 2), Gaussian smearing (ISMEAR = 0, SIGMA = 0.05 eV)- Electronic SCF: EDIFF = 1e-6 eV, ALGO = All, LREAL = Auto- Ionic optimisation: VTST LBFGS (IBRION = 3, IOPT = 1), force convergence EDIFFG = -0.005 eV/A, NSW = 3000- Charge-density mixing for f electrons: LMAXMIX = 6- Initial magnetic moments: 1 muB on every Ce, 0 on every O (MAGMOM = 32*1 63*0 for the 110 slab; adjusted per slab) ## MACE setup - Model: `mace-omat-0-medium` (MACE-MP-0 family)- The exact model path is recorded in each `mace_opt.summary.txt`.- La is used as a proxy for Ce3+ during the MACE pre-relaxation and is back-substituted to Ce before the DFT+U refinement. ## POTCAR VASP pseudopotentials are proprietary and cannot be redistributed. In`inputs_vasp/POTCAR_info.txt` we list, for each species in the order theyappear in the POSCAR: - the TITEL line (pseudopotential name and generation date),- the VRHFIN line (reference valence configuration),- the LEXCH line (exchange-correlation flavour),- the SHA256 of that species block, and- the SHA256 of the full concatenated POTCAR. Anyone with a VASP licence can reconstruct the POTCAR from their owninstallation and verify the SHA256 matches. The script`make_potcar_info.sh` (included in this dataset) regenerates`POTCAR_info.txt` from a local POTCAR. ## Reproducing a configuration 1. Clone TAPAS: `git clone https://github.com/pafervi/TAPAS`.2. Regenerate the configuration set from the pristine slab POSCAR used in the paper (provided in the TAPAS repository or on request) with the same command recorded in each `config_info.txt`.3. For MACE: run the pre-relaxation with the model listed in `mace_opt.summary.txt`.4. For DFT+U: back-substitute La to Ce and run VASP with the INCAR, KPOINTS and POTCAR identified in `inputs_vasp/`. ## Licence - Structures, energies and magnetisations in this dataset: CC-BY 4.0.- TAPAS source code and helper scripts: MIT (see the TAPAS repository).- VASP pseudopotentials are NOT part of this dataset. ## Contact Pau Ferri-Vicedo, [e-mail], Instituto de Tecnologia Quimica (ITQ, UPV-CSIC),Valencia, Spain.