tmQM

tmQM provides quantum-mechanical geometries and computed electronic properties for 86,665 mononuclear transition metal-organic complexes extracted from the Cambridge Structural Database to support data-driven studies and machine-learning investigations in transition metal chemistry.

Key Features:

• Dataset scope: 86,665 mononuclear transition metal-organic compounds covering Werner, bioinorganic, and organometallic complexes with diverse organic ligands and 30 transition metals across groups 3–12 (3d, 4d, 5d series).
• Cartesian geometries: Cartesian coordinates for each complex optimized at the semi-empirical DFTB(GFN2-xTB) level of theory.
• DFT-derived electronic properties: Single-point DFT computations with the TPSSh-D3BJ/def2-SVP functional yielding electronic and dispersion energies, HOMO and LUMO orbital energies, HOMO–LUMO gaps, and dipole moments.
• Metal-center charges: Natural charges at the metal center provided for electronic-structure analysis.
• Polarizabilities: Polarizabilities computed using DFTB(GFN2-xTB) to characterize electronic response to external fields.
• Chemical-space observations: Reported low correlations among quantum properties and examples of regions with large polarizabilities combined with wide HOMO–LUMO gaps and cases with low-energy HOMO orbitals alongside electron-rich metal centers.

Scientific Applications:

• Machine-learning model development: Training and benchmarking ML models for predicting electronic properties of transition metal complexes.
• Catalyst and catalyst-ligand design: Guiding data-driven discovery of catalysts by linking computed properties to reactivity and selectivity hypotheses.
• Materials and molecular design: Supporting design of materials and organometallic compounds with target electronic, optical, or polarizability characteristics.
• Chemical-space analysis: Enabling systematic exploration of structure–property relationships across diverse transition metal chemical space.

Methodology:

Structures were extracted from the Cambridge Structural Database, geometries optimized at the DFTB(GFN2-xTB) level, polarizabilities computed with DFTB(GFN2-xTB), and electronic properties obtained from DFT single-point calculations using TPSSh-D3BJ/def2-SVP (electronic and dispersion energies, HOMO/LUMO energies, HOMO–LUMO gaps, dipole moments, and natural charges at the metal center).