Data-driven OLED candidate design: a generative model from independent-property domains to the comprehensive performance enhancement

Abstract

The discovery of high-performance organic light-emitting diode (OLED) materials is hindered by conventional human-aware design methodologies and the scarcity of pure organic luminescent scaffolds. Although machine learning models have improved the efficiency of high-throughput screening for OLED candidates, their effectiveness is still limited by the small size and low quality of available experimental datasets. In this study, we introduced LumiGen, an integrated framework for the de novo design of high-quality OLED candidate molecules with targeted photophysical properties. A sampling-screening iterative process was designed to gradually refine the molecular selection, enabling the transition from independent-property optimization to all-rounded OLED candidates. Among the collected high-quality OLED candidate molecules, computational estimates indicate that the optical properties of most molecules (approximately 80.2%) meet the required criteria. During the iterative training process, the Sampling Augmentor enhances the proportion of OLED candidate molecules by over threefold (from 6.56% to 21.13%). Additionally, we successfully synthesized a new molecular scaffold from the OLED candidates, achieving a photoluminescence quantum yield of up to 88.6%. According to the statistics, only 0.33% of the molecules in the dataset outperform our synthesized molecules in terms of overall optical performance. LumiGen demonstrates the ability to learn molecular distribution patterns from disjoint labeled datasets, enabling the direct generation of all-round OLED candidates, thereby advancing OLED material discovery.

Keywords

Machine learning / luminescent molecules / de novo design / OLED

References

[1] Xiong Z,Lv Y,Zhang Z.Hydrogen-bonded organic frameworks as an appealing platform for luminescent sensing.Adv Funct Mater2024;34:2403635

[2] Yang Q,Zhu S.Donor engineering for NIR-II molecular fluorophores with enhanced fluorescent performance.J Am Chem Soc2018;140:1715-24

[3] Hayashi S.Elastic organic crystals of a fluorescent π-conjugated molecule.Angew Chem Int Ed Engl2016;55:2701-4

[4] Wang C,Ahrens ET.Emergent fluorous molecules and their uses in molecular imaging.Acc Chem Res2021;54:3060-70

[5] Tu L,Li Z.Aggregation-induced emission: red and near-infrared organic light-emitting diodes.SmartMat2021;2:326-46

[6] Yoshida K,Kanibolotsky AL,Turnbull GA.Electrically driven organic laser using integrated OLED pumping.Nature2023;621:746-52 PMCID:PMC10533406

[7] Hong G,Leonhardt C.A brief history of OLEDs-emitter development and industry milestones.Adv Mater2021;33:e2005630

[8] Thakur K,Sachnik O.Effect of tert-butylation on the photophysics of thermally activated delayed fluorescence emitters.Adv Photonics Res2024;5:2400022

[9] Stavrou K,Danos A.Key requirements for ultraefficient sensitization in hyperfluorescence organic light-emitting diodes.Nat Photon2024;18:554-61

[10] Eggeman AS,Troisi A,Midgley PA.Measurement of molecular motion in organic semiconductors by thermal diffuse electron scattering.Nat Mater2013;12:1045-9

[11] Wei Q,Islam A.Small-molecule emitters with high quantum efficiency: mechanisms, structures, and applications in OLED devices.Adv Opt Mater2018;6:1800512

[12] Kuang C,Murtaza I.Enhanced horizontal dipole orientation by novel penta-helicene anthracene-based host for efficient blue fluorescent OLEDs.Small2024;20:e2311114

[13] Cho HH,Gillett AJ.Suppression of Dexter transfer by covalent encapsulation for efficient matrix-free narrowband deep blue hyperfluorescent OLEDs.Nat Mater2024;23:519-26 PMCID:PMC10990937

[14] Gómez-Bombarelli R,Hirzel TD.Design of efficient molecular organic light-emitting diodes by a high-throughput virtual screening and experimental approach.Nat Mater2016;15:1120-7

[15] Joung JF,Hwang J,Choi DH.Deep learning optical spectroscopy based on experimental database: potential applications to molecular design.JACS Au2021;1:427-38 PMCID:PMC8395663

[16] Shi H,Liang Z.Machine learning-enabled discovery of multi-resonance TADF molecules: unraveling PLQY predictions from molecular structures.Chem Eng J2024;494:153150

[17] Sun M,Su H.Enhancing chemistry-intuitive feature learning to improve prediction performance of optical properties.Chem Sci2024;15:17533-46 PMCID:PMC11457255

[18] Weiss T,Chakraborty S,Bronstein AM.Guided diffusion for inverse molecular design.Nat Comput Sci2023;3:873-82

[19] Zeni C,Zügner D.A generative model for inorganic materials design.Nature2025;639:624-32 PMCID:PMC11922738

[20] Alakhdar A,Washburn N.Diffusion models in de novo drug design.J Chem Inf Model2024;64:7238-56 PMCID:PMC11481093

[21] Gómez-Bombarelli R,Duvenaud D.Automatic chemical design using a data-driven continuous representation of molecules.ACS Cent Sci2018;4:268-76 PMCID:PMC5833007

[22] Zhang K,Wang Y.Artificial intelligence in drug development.Nat Med2025;31:45-59

[23] Moret M,Grisoni F,Schneider G.Generative molecular design in low data regimes.Nat Mach Intell2020;2:171-80

[24] Sumita M,Suzuki N.De novo creation of a naked eye-detectable fluorescent molecule based on quantum chemical computation and machine learning.Sci Adv2022;8:eabj3906 PMCID:PMC8906732

[25] Joung JF,Jeong M.Experimental database of optical properties of organic compounds.Sci Data2020;7:295 PMCID:PMC7478979

[26] Gong J,Wu B.ASBase: the universal database for aggregate science.Aggregate2023;4:e263

[27] Li P,Li W,Chen R.Design of thermally activated delayed fluorescence materials with high intersystem crossing efficiencies by machine learning-assisted virtual screening.J Phys Chem Lett2022;13:9910-8

[28] Kim H,Kim JH.Deep learning-based chemical similarity for accelerated organic light-emitting diode materials discovery.J Chem Inf Model2024;64:677-89

[29] Blaskovits JT,Vela S.Data-driven discovery of organic electronic materials enabled by hybrid top-down/bottom-up design.Adv Mater2024;36:e2305602

[30] Zdrazil B,Hunter F.The ChEMBL Database in 2023: a drug discovery platform spanning multiple bioactivity data types and time periods.Nucleic Acids Res2024;52:D1180-92 PMCID:PMC10767899

[31] Guo J,Zhao X.General graph neural network-based model to accurately predict cocrystal density and insight from data quality and feature representation.J Chem Inf Model2023;63:1143-56

[32] Medina-Franco JL,Bender A.Scaffold diversity analysis of compound data sets using an entropy-based measure.QSAR Comb Sci2009;28:1551-60

[33] Niu Y,Peng Q.MOlecular MAterials Property Prediction Package (MOMAP) 1.0: a software package for predicting the luminescent properties and mobility of organic functional materials.Mol Phys2018;116:1078-90

[34] Park Y,Jung DH.An aromatic imine group enhances the EL efficiency and carrier transport properties of highly efficient blue emitter for OLEDs.J Mater Chem2010;20:5930

[35] Kotaka H,Mizuno K.Synthesis and photoluminescence properties of π-extended fluorene derivatives: the first example of a fluorescent solvatochromic nitro-group-containing dye with a high fluorescence quantum yield.Tetrahedron Lett2010;51:181-4

[36] Liu X,Ding L.The study on two kinds of spiro systems for improving the performance of host materials in blue phosphorescent organic light-emitting diodes.J Mater Chem C2015;3:9053-6