Published November 7, 2026 | Version v4

TD-ωB97X-D excited-state energies and forces of 106,933 molecules at equilibrium, thermal, and excited-state geometries

Authors/Creators

  • 1. Karlsruhe Institute of Technology

Description

Dataset of the paper Toward universal excited-state machine-learned potentials: a local excitation Hamiltonian for size consistency and transferability (P. Friederich).

TD-DFT (ωB97X-D/6-31G*, full linear response, TURBOMOLE 7.5.1) labels for 303,061 geometries of 106,933 closed-shell organic molecules with C, H, N, O, F (and S in some dyes and ladder molecules): S0 energies and forces, the ten lowest singlet excitation energies with oscillator strengths, and analytic S1, S2, and S3 forces. Geometries: QCDGE minima, thermal geometries from GFN2-xTB dynamics at 300–1000 K, geometries from model-driven S1/S2 dynamics (four active-learning rounds), and test molecules beyond the training distribution (larger molecules from ZINC, dyes from two experimental databases, homologous ladder series up to 30 heavy atoms). Splits of the benchmark, dye-augmented, and production models of the paper.

Also included: S1–S3 forces at equilibrium geometries of 2,499 test molecules; pairs of molecules at 2.5–20 Å with TD-DFT dimer spectra; twelve TD-ωB97X-D S1 reference trajectories with initial conditions and a validity flag; Löwdin hole and electron populations and charge-transfer numbers of the ten roots for 25,808 geometries; and the TURBOMOLE input and output files (without MO files) of 200,445 calculations. The file README.md describes all files, fields, units, and splits.

Files

MANIFEST.json

Files (9.2 GB)

Name Size
md5:00b957b3b8e4905601d90cbe0a416cd6
4.0 MB Download
md5:b29b60bc388c3ef42eb8e3c8589cc116
763.1 MB Download
md5:0cce9c8e53bdf1016048afaf71a8e1e6
2.7 kB Preview Download
md5:78749172f350c7ed0084b60fcbb399c3
1.9 MB Download
md5:8fe7305f22c6f5ff0f48017abb789e07
385.8 MB Download
md5:347f1bc594466c958b6f3dda423ef0ed
497.3 MB Download
md5:f03b3f46b8185059f0f5636174731991
653.9 MB Download
md5:5575f5efb9db06673e6b3d912ab01cdf
654.9 MB Download
md5:578c8aa88b9da11e462a908c63aeba66
917.6 MB Download
md5:177a1aa18bd99449d5d75ca97a9a9016
916.1 MB Download
md5:bd7f809b21513ae9085e40360c8f5199
93.4 MB Download
md5:e036da7b4f40e94a30215fe93f126bb0
299.4 MB Download
md5:e542b935b9015ee5a1d87d98d3bb3f8f
3.5 GB Download
md5:f633908d9e45b3bf9a3c87095d462d10
11.4 kB Preview Download
md5:038d99754252375589782a2becd87e1b
15.1 MB Download
md5:4da5fd542024514e28d0f2992410633c
1.7 MB Preview Download
md5:6e02b5b244d33a84c15945399360e502
1.7 MB Preview Download
md5:d111d84878a4a85ed1a797d9b7e7ee1a
2.3 MB Preview Download
md5:5388c1b7851268cca6579964991a7e51
555.1 MB Download

Additional details

Related works

Is derived from
Journal article: 10.1038/s41597-024-03788-x (DOI)
Dataset: 10.6084/m9.figshare.c.7259125 (DOI)
Is supplement to
Software: https://github.com/aimat-lab/LEH (URL)