Screening 400 emitter candidates used to mean — 400 DFT jobs
Without running a separate excited-state calculation for each candidate, without manually inspecting HOMO/LUMO maps, without a DFT queue. Provide SMILES. Get emission wavelength, TADF suitability, and device role classification — in minutes, not days.
“Find me a blue OLED emitter from this library of 400 candidates, ranked by singlet-triplet gap for TADF suitability.”
How it works
Provide SMILES
Geometry optimizes via ANI-2x or MACE in milliseconds. No conformer setup, no input file preparation.
Frontier orbitals and device classification
predict_frontier_orbitals returns HOMO, LUMO, gap, S1/T1 energies, oscillator strength, and device role — emitter, charge transport, host, or not emissive. 14 OLED-relevant functional-group motifs detected automatically (carbazole, triphenylamine, anthracene, pyrene, oxadiazole, triazine, Ir/Pt complexes).
Excited-state ladder for top candidates
run_excited_states returns the full singlet/triplet ladder via sTDA-xTB with oscillator strengths. Screen for TADF suitability (small singlet-triplet gap) or phosphorescent emitters.
Proof
Anthracene S1: 3.249 eV computed vs 3.3 eV experimental (0.05 eV error). T1: 1.805 eV vs 1.8 eV (0.005 eV error).
Carbazole: correctly classified as UV emission / charge transport. Ethanol: correctly returns not_emissive (no OLED motifs).
14 detected motifs: carbazole, triphenylamine, anthracene, pyrene, oxadiazole, triazine, Ir/Pt complexes, and others.
Tool chain: optimize_geometry_nnp → predict_frontier_orbitals → run_excited_states.
Screen your emitter library
S1 within 0.05 eV. 14 motifs detected. Sign up and screen your first candidates.