A 1752‑molecule benchmark set for molecular geometry optimizers on the GFN2‑xTB potential energy surface

This dataset is a Cartesian‑coordinate benchmark set of 1752 molecules (4–110 atoms) for evaluating molecular geometry optimizers on the semiempirical tight‑binding GFN2‑xTB potential energy surface. It collects molecules from the Baker geometry‑optimization test set, a broad selection of GMTKN55 subsets (conformers, reactions, non‑covalent complexes, main‑group and heavy‑element species), the 130 transition‑metal complexes of the tmQM database, organic and TURBOMOLE test molecules, and larger flexible drug‑like systems (Birkholz–Schlegel). The set spans 215 charged species and 114 open‑shell doublets. For each molecule the archive provides three items: (a) a distorted starting geometry, generated by perturbing the reference minimum, which serves as the optimizer input; (b) the corresponding GFN2‑xTB‑optimized minimum, which serves as the reference; and (c) the complete optimization trajectory from the starting geometry to the minimum. The two static geometries are given in .xyz format (Ångström) and in native TURBOMOLE coord format (atomic units / bohr); trajectories are multi‑frame .xyz files whose first frame is the starting geometry and whose final frame is the optimized minimum, with every frame annotated by its energy (Ha) and maximum gradient component. The file manifest.csv lists, for every molecule, its group, case name, atom count, chemical formula, total charge, spin multiplicity, the number of trajectory frames and optimization cycles, the maximum gradient component at the minimum, and source. Charges and multiplicities correspond to the automatic GFN2‑xTB low‑spin state (singlet for an even electron count, doublet for an odd one); the 130 tmQM complexes are closed‑shell singlets, matching the tmQM database. All structures were optimized on the GFN2‑xTB surface as implemented in TURBOMOLE's built‑in tight‑binding module (not the standalone xtb program), in the gas phase; the exact engine and settings are documented in METHOD.txt. To reproduce the benchmark, optimize each distorted starting geometry on the GFN2‑xTB surface and compare the result with the provided reference minimum. The convergence criterion is a maximum gradient component below 1e‑4 Ha/bohr, an RMS gradient below 5e‑4 Ha/bohr, and an energy change below 1e‑6 Ha between successive steps; every deposited minimum satisfies these criteria. Optimizer cost is measured as the number of GFN2‑xTB energy‑and‑gradient evaluations, equal to the trajectory length and listed per molecule in manifest.csv. Contents: xyz/ (starting and optimized geometries, Ångström), turbomole_coord/ (same geometries, native TURBOMOLE format), trajectories/ (complete optimization trajectories, multi‑frame Ångström .xyz), manifest.csv, METHOD.txt (engine and computational settings), README.md, and LICENSE. Released under the Creative Commons Attribution 4.0 International (CC‑BY‑4.0) license.