Abstract
Artificial intelligence-guided closed-loop experimentation has emerged as a promising method for optimization of objective functions1,2, but the substantial potential of this traditionally black-box approach to uncovering new chemical knowledge has remained largely untapped. Here we report the integration of closed-loop experiments with physics-based feature selection and supervised learning, denoted as closed-loop transfer (CLT), to yield chemical insights in parallel with optimization of objective functions. CLT was used to examine the factors dictating the photostability in solution of light-harvesting donor–acceptor molecules used in a variety of organic electronics applications, and showed fundamental insights including the importance of high-energy regions of the triplet state manifold. This was possible following automated modular synthesis and experimental characterization of only around 1.5% of the theoretical chemical space. This physics-informed model for photostability was strengthened using multiple experimental test sets and validated by tuning the triplet excited-state energy of the solvent to break out of the observed plateau in the closed-loop photostability optimization process. Further applications of CLT to additional materials systems support the generalizability of this strategy for augmenting closed-loop strategies. Broadly, these findings show that combining interpretable supervised learning models and physics-based features with closed-loop discovery processes can rapidly provide fundamental chemical insights.
This is a preview of subscription content, access via your institution
Access options
Access Nature and 54 other Nature Portfolio journals
Get Nature+, our best-value online-access subscription
$29.99 / 30 days
cancel any time
Subscribe to this journal
Receive 51 print issues and online access
$199.00 per year
only $3.90 per issue
Buy this article
- Purchase on SpringerLink
- Instant access to full article PDF
Prices may be subject to local taxes which are calculated during checkout
Similar content being viewed by others
Data availability
The data that support the findings of this study are available online (https://github.com/TheJacksonLab/ClosedLoopTransfer), including experimental photostability measurement data for all molecules synthesized in this work, DFT- and RDKit-derived molecular featurizations and predicted photostabilities across the full space of 2,200 molecules. Regression models and scripts used to train and perform all analysis with the associated data are also provided. Datasets are available at Zenodo (https://doi.org/10.5281/zenodo.11580889)49. Source Data are provided with this paper.
Code availability
All supervised learning model codes are available at GitHub (https://github.com/TheJacksonLab/ClosedLoopTransfer). Gryffin is available at GitHub (https://github.com/aspuru-guzik-group/gryffin). Codes are available at Zenodo (https://doi.org/10.5281/zenodo.11580889)49.
References
Flores-Leonar, M. M. et al. Materials acceleration platforms: on the way to autonomous experimentation. Curr. Opin. Green Sustain. Chem. 25, 100370 (2020).
Peng, X. & Wang, X. Next-generation intelligent laboratories for materials design and manufacturing. MRS Bull. 48, 179–185 (2023).
Burger, B. et al. A mobile robotic chemist. Nature 583, 237–241 (2020).
Angello, N. H. et al. Closed-loop optimization of general reaction conditions for heteroaryl Suzuki-Miyaura coupling. Science 378, 399–405 (2022).
Koscher, B. A. et al. Autonomous, multiproperty-driven molecular discovery: from predictions to measurements and back. Science 382, eadi1407 (2023).
Chang, J. et al. Efficient closed-loop maximization of carbon nanotube growth rate using Bayesian optimization. Sci. Rep. 10, 9040 (2020).
Alem, S. et al. Degradation mechanism of benzodithiophene-based conjugated polymers when exposed to light in air. ACS Appl. Mater. Interfaces 4, 2993–2998 (2012).
Mateker, W. R. & McGehee, M. D. Progress in understanding degradation mechanisms and improving stability in organic photovoltaics. Adv. Mater. 29, 1603940 (2017).
Liu, Z.-X. et al. Molecular insights of exceptionally photostable electron acceptors for organic photovoltaics. Nat. Commun. 12, 3049 (2021).
Bekri, N., Asmare, E., Mammo, W. & Tegegne, N. A. Photostability of benzodithiophene based polymer: effect of PC60BM and intermolecular interactions. Mater. Res. Express 9, 055502 (2022).
Anderson, M. A., Hamstra, A., Larson, B. W. & Ratcliff, E. L. Distinguishing photo-induced oxygen attack on alkyl chain versus conjugated backbone for alkylthienyl-benzodithiophene (BDTT)-based push–pull polymers. J. Mater. Chem. A 11, 17858–17871 (2023).
Burlingame, Q. et al. Intrinsically stable organic solar cells under high-intensity illumination. Nature 573, 394–397 (2019).
Burlingame, Q., Ball, M. & Loo, Y.-L. It’s time to focus on organic solar cell stability. Nat. Energy 5, 947–949 (2020).
Kuramoto, N. in Physico-Chemical Principles of Color Chemistry (eds Peters, A. T. & Freeman, H. S.) 196–253 (Springer Netherlands, 1996); https://doi.org/10.1007/978-94-009-0091-2_6.
Kosco, J. et al. Enhanced photocatalytic hydrogen evolution from organic semiconductor heterojunction nanoparticles. Nat. Mater. 19, 559–565 (2020).
Zbyradowski, M. et al. Triplet-driven chemical reactivity of β-carotene and its biological implications. Nat. Commun. 13, 2474 (2022).
Tällberg, R., Jelle, B. P., Loonen, R., Gao, T. & Hamdy, M. Comparison of the energy saving potential of adaptive and controllable smart windows: a state-of-the-art review and simulation studies of thermochromic, photochromic and electrochromic technologies. Sol. Energy Mater. Sol. Cells 200, 109828 (2019).
Chan, C.-Y. et al. Stable pure-blue hyperfluorescence organic light-emitting diodes with high-efficiency and narrow emission. Nat. Photonics 15, 203–207 (2021).
Kim, J.-H. et al. Hydrophilic/hydrophobic silane grafting on TiO2 nanoparticles: photocatalytic paint for atmospheric cleaning. Catalysts 11, 193 (2021).
Eggeling, C., Widengren, J., Rigler, R. & Seidel, C. A. M. Photobleaching of fluorescent dyes under conditions used for single-molecule detection: evidence of two-step photolysis. Anal. Chem. 70, 2651–2659 (1998).
O’Brien, J., Wilson, I., Orton, T. & Pognan, F. Investigation of the Alamar Blue (resazurin) fluorescent dye for the assessment of mammalian cell cytotoxicity. Eur. J. Biochem. 267, 5421–5426 (2000).
Korycka-Dahl, M. & Richardson, T. Photodegradation of DNA with fluorescent light in the presence of riboflavin, and photoprotection by flavin triplet-state quenchers. Biochim. Biophys. Acta 610, 229–234 (1980).
Groeneveld, I., Kanelli, M., Ariese, F. & van Bommel, M. R. Parameters that affect the photodegradation of dyes and pigments in solution and on substrate – an overview. Dyes Pigments 210, 110999 (2023).
Kosumi, D., Horibe, T., Sugisaki, M., Cogdell, R. J. & Hashimoto, H. Photoprotection mechanism of light-harvesting antenna complex from purple bacteria. J. Phys. Chem. B 120, 951–956 (2016).
McNeill, K. & Canonica, S. Triplet state dissolved organic matter in aquatic photochemistry: reaction mechanisms, substrate scope, and photophysical properties. Environ. Sci. Process. Impacts 18, 1381–1399 (2016).
Distler, A. et al. Effect of PCBM on the photodegradation kinetics of polymers for organic photovoltaics. Chem. Mater. 24, 4397–4405 (2012).
Zheng, Q. & Lavis, L. D. Development of photostable fluorophores for molecular imaging. Curr. Opin. Chem. Biol. 39, 32–38 (2017).
Meng, Q.-Y. et al. Longevity gene responsible for robust blue organic materials employing thermally activated delayed fluorescence. Nat. Commun. 14, 3927 (2023).
Li, H. et al. Chemical stabilities of the lowest triplet state in aryl sulfones and aryl phosphine oxides relevant to OLED applications. Chem. Mater. 31, 1507–1519 (2019).
Zhang, K. et al. New insights about the photostability of DNA/RNA bases: triplet nπ* state leads to effective intersystem crossing in pyrimidinones. J. Phys. Chem. B 125, 2042–2049 (2021).
Guo, X. et al. Unraveling the important role of high-lying triplet–lowest excited singlet transitions in achieving highly efficient deep-blue AIE-based OLEDs. Adv. Mater. 33, 2006953 (2021).
Mukherjee, S., Fedorov, D. A. & Varganov, S. A. Modeling spin-crossover dynamics. Annu. Rev. Phys. Chem. 72, 515–540 (2021).
Strieth-Kalthoff, F. et al. Delocalized, asynchronous, closed-loop discovery of organic laser emitters. Science 384, eadk9227 (2024).
Wang, X. et al. Bayesian-optimization-assisted discovery of stereoselective aluminum complexes for ring-opening polymerization of racemic lactide. Nat. Commun. 14, 3647 (2023).
Häse, F., Aldeghi, M., Hickman, R. J., Roch, L. M. & Aspuru-Guzik, A. Gryffin: an algorithm for Bayesian optimization of categorical variables informed by expert knowledge. Appl. Phys. Rev. 8, 031406 (2021).
Li, J. et al. Synthesis of many different types of organic small molecules using one automated process. Science 347, 1221–1226 (2015).
Wang, W. et al. Rapid automated iterative small molecule synthesis. Nat. Synth. 3, 1031–1038 (2024).
Blair, D. J. et al. Automated iterative Csp3–C bond formation. Nature 604, 92–97 (2022).
Zhang, G. et al. Renewed prospects for organic photovoltaics. Chem. Rev. 122, 14180–14274 (2022).
Heitzer, H. M., Marks, T. J. & Ratner, M. A. Molecular donor–bridge–acceptor strategies for high-capacitance organic dielectric materials. J. Am. Chem. Soc. 137, 7189–7196 (2015).
Bronstein, H., Nielsen, C. B., Schroeder, B. C. & McCulloch, I. The role of chemical design in the performance of organic semiconductors. Nat. Rev. Chem. 4, 66–77 (2020).
Landrum, G. RDKit: open-source cheminformatics software, version 2021_09_01 (Q3 2021) (accessed 12 August 2024); www.rdkit.org.
Knapp, D. M., Gillis, E. P. & Burke, M. D. A general solution for unstable boronic acids: slow-release cross-coupling from air-stable MIDA boronates. J. Am. Chem. Soc. 131, 6961–6963 (2009).
Refaely-Abramson, S., Baer, R. & Kronik, L. Fundamental and excitation gaps in molecules of relevance for organic photovoltaics from an optimally tuned range-separated hybrid functional. Phys. Rev. B 84, 075144 (2011).
Abroshan, H., Coropceanu, V. & Brédas, J.-L. Hyperfluorescence-based emission in purely organic materials: suppression of energy-loss mechanisms via alignment of triplet excited states. ACS Mater. Lett. 2, 1412–1418 (2020).
Li, Y. et al. Non-fullerene acceptor organic photovoltaics with intrinsic operational lifetimes over 30 years. Nat. Commun. 12, 5419 (2021).
Yousif, E. & Haddad, R. Photodegradation and photostabilization of polymers, especially polystyrene: review. Springerplus 2, 398 (2013).
Fan, B. et al. Understanding the role of removable solid additives: selective interaction contributes to vertical component distributions. Adv. Mater. 35, 2302861 (2023).
TheJacksonLab. TheJacksonLab/ClosedLoopTransfer: ClosedLoopTransfer v1.0. Zenodo https://doi.org/10.5281/zenodo.11580889 (2024).
Acknowledgements
This work was supported by the Molecule Maker Lab Institute, an AI Research Institutes programme supported by the US National Science Foundation under grant no. 2019897 (to N.E.J., Y.D., C.M.S. and M.D.B.). A.A.-G. and A.H.C. acknowledge support from the Canada 150 Research Chairs Program and the Acceleration Consortium at the University of Toronto, as well as the generous support of A. G. Frøseth. T.C.T.-F., Y.D. and C.M.S. acknowledge support by the IBM–Illinois Discovery Accelerator Institute. Any opinions, findings and conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect those of the US National Science Foundation.
Author information
Authors and Affiliations
Contributions
N.H.A., D.M.F., C.H. and S.Y. contributed equally to this project. A.H.C. and T.C.T.-F. contributed equally to this project. The project was designed by N.H.A., D.M.F., C.H., S.Y., E.R.J., A.A.-G., M.D.B., C.M.S., Y.D. and N.E.J. Molecule synthesis was conducted by N.H.A., S.Y., T.C.T.-F., E.R.J. and W.W. Solution testing was conducted by C.H. BO and regression model training was conducted by D.M.F. and A.H.C. N.H.A., D.M.F., C.H., S.Y., A.A.-G., M.D.B., C.M.S., Y.D. and N.E.J. wrote the manuscript with contributions from all authors.
Corresponding authors
Ethics declarations
Competing interests
The University of Illinois has filed patent applications related to MIDA and TIDA boronates with M.D.B., N.H.A. and W.W. as inventors.
Peer review
Peer review information
Nature thanks Brett Savoie and the other, anonymous, reviewer(s) for their contribution to the peer review of this work.
Additional information
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Extended data figures and tables
Extended Data Fig. 1 Chemical design space of acceptor moieties.
Chemical diversity down-selected set of acceptor blocks used in populating the design space for the D-B-A motif utilized in this work.
Extended Data Fig. 2 Digital project manager.
Visualization of Streamlit web app used in this work showcasing integration of the Gryffin Bayesian optimizer, building block fragment assembler, and suggestion window for similar molecules as synthetic backups.
Extended Data Fig. 3 The small molecule synthesizer used in this work.
a, Picture of the hardware. b, Design schematic.
Extended Data Fig. 4 Triplet Density of States (TDOS) for all experimentally measured molecules.
DB08_A096 (the high performer in the predicted Bottom 7) is shown in red. All others are in gray, with the highest T80 in darker colors, and the lowest T80 in lighter colors.
Extended Data Fig. 5 Results from CLT phase I.
a, All support vector regression leave-one-out validation (LOOV) results for predicting T80 of the 30 molecules characterized in Phase I from 2-feature combinations. b, Comparison of the prediction strength of all possible 4-feature models containing either the TDOS at 4.0 eV or the T1 energy. Compare to Fig. 5b & c.
Extended Data Fig. 6 Physics-driven discovery in phase III.
a, The distribution of photostabilities of the 44 molecules synthesized through Phase II of the CLT campaign. b, The best 4 feature model for predicting T80 from Phase II. Note the similarities of the TDOS features to those in the original physics based T80 model c, The relative photostability of 3 molecules in CB, toluene, and decane, showing the improved photostability for all molecules in the absence of Dexter energy transfer, and that the improvement correlates with the TDOS at 4.0 eV. d, The impact of adding cyclooctatetraene (COT) triplet quencher to the CB solution (red) and chemically attaching it to DB_11_A_002 (blue). The structure shown is DB_11_A_002 (the highest performing molecule in Fig. 4a, the control), with the dodecyl side chain replaced with a hexyl-COT side chain. Results support the Dexter energy transfer hypothesis as explained in SI Section 6.
Supplementary information
Source data
Rights and permissions
Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
About this article
Cite this article
Angello, N.H., Friday, D.M., Hwang, C. et al. Closed-loop transfer enables artificial intelligence to yield chemical knowledge. Nature 633, 351–358 (2024). https://doi.org/10.1038/s41586-024-07892-1
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1038/s41586-024-07892-1