Dataset: Cross-Geometry Transferability Assessment of Universal Machine Learning Interatomic Potentials: From Bulk Materials to Atomic Nanowires

This dataset contains zirconia (ZrO₂) atomic structures representing different geometries, coordination environments, stoichiometries, and defect configurations. The structures are organized into five directories according to their main geometry: - `bulk/`: periodic bulk structures;- `slab/`: surface and slab structures with different crystallographic orientations;- `particle/`: finite zirconia nanoparticles;- `neck/`: structures containing nanoscale necks between zirconia particles;- `wire/`: zirconia nanowires and structures generated during wire deformation or transition pathways. File naming convention Bulk structures Bulk filenames generally begin with `b_` or use the name `bulk`. The remaining terms identify the bulk configuration or defect type. Examples include: - `bulk-ZrO2.xyz`: reference bulk zirconia structure;- `b_sc-ZrO2.xyz`: bulk structure associated with a specific supercell configuration;- `b_vo3-ZrO2.xyz` and `b_vo4-ZrO2.xyz`: bulk structures containing oxygen-vacancy configurations;- `b_vzr-ZrO2.xyz`: bulk structure containing a zirconium vacancy. Particle structures Particle filenames begin with `p_`. The number following the prefix indicates the total number of atoms in the nanoparticle. Examples include: - `p_56_lowO-ZrO2.xyz`: a 56-atom nanoparticle with reduced oxygen content;- `p_72_stoich-ZrO2.xyz`: a 72-atom stoichiometric nanoparticle;- `p_107_highO-ZrO2.xyz`: a 107-atom nanoparticle with increased oxygen content. The labels `lowO`, `stoich`, and `highO` indicate oxygen-deficient, stoichiometric, and oxygen-rich compositions, respectively. Slab structures Slab filenames begin with `s` and are generally followed by the crystallographic surface orientation. Examples include: - `s001-ZrO2.xyz`;- `s010-ZrO2.xyz`;- `s011-ZrO2.xyz`;- `s100-ZrO2.xyz`;- `s110-ZrO2.xyz`;- `s111-ZrO2.xyz`;- `s211-ZrO2.xyz`. Negative Miller indices are represented using hyphens, as in `s-101`, `s-111`, or `s-11-1`. Additional terms may describe surface defects, grain-boundary structures, or intermediate construction steps. For example, `step3_surf_def_gb-ZrO2.xyz` represents a structure associated with a surface defect or grain-boundary configuration generated at a specific step. Neck structures Neck filenames begin with `neck_`. These files represent configurations containing a nanoscale connection between two zirconia particles or surfaces. For example: - `neck_case3v_step0-ZrO2.xyz`;- `neck_case3v_step11-ZrO2.xyz`. The `case` label identifies the structural setup, while `step` indicates the configuration number or stage of the structural evolution. Wire structures Wire filenames begin with `w_`. The remaining terms describe the simulation pathway, deformation stage, vacancy configuration, or structural step. Examples include: - `w_step1-ZrO2.xyz` to `w_step10-ZrO2.xyz`: sequential nanowire configurations;- `w_neb_01-ZrO2.xyz` to `w_neb_03-ZrO2.xyz`: structures associated with a nudged elastic band, or NEB, pathway;- `w_pull6vOstep3-ZrO2.xyz`: a nanowire configuration generated during a pulling procedure and containing oxygen vacancies. In general, `step` indicates the position of a structure within a sequence, `neb` identifies an NEB image, `pull` indicates a tensile or pulling process, and `vO` denotes oxygen vacancies.