Rational design of Pd-catalysts for C-H activation and Ru-catalysts for C=O hydrogenation: from operando X-ray absorption spectroscopy identification of metal complexes to multi-technique machine learning-based characterization

Rational design of Pd-catalysts for C-H activation and Ru-catalysts for C=O hydrogenation: from operando X-ray absorption spectroscopy identification of metal complexes to multi-technique machine learning-based characterization

РНФ. Бельгия. 2020-2022
From 01.01.2020 till 31.12.2022
Grant holder: Alexander Soldatov
Responsible: Aram Bugaev
Members: Yury Rusalev, Vera Butova, Alexander Guda, Andrei Tereshchenko, Sergey Guda, Oleg Usoltsev, Alina Skorynina, Elizaveta Kozyr

Modern pharmaceutical and fine chemical industries aim at producing chemical compounds of ever increasing complexity, via cost-effective routes that have low environmental impact. Key motifs in many new compounds are vinylated aromatic molecules, which are prepared through catalytic C=C bond formation. The interesting and still unsolved problem is the direct activation of aromatic C-H bonds [1], instead of the currently used use ‘pre-oxidized’ aromatic precursors (e.g. aryl halides) results in stoichiometric formation of halide salt waste. Another promising direction is deoxygenations of biobased oxygenated compounds, in particular, defunctionalization of polyols to C3, C4, C5, C6 olefins [2,3], which can be used as building blocks for the chemical industry. However, there so far no catalysts that can defunctionalize polyols to olefins with a satisfactory selectivity and substrate scope.

Rational design of novel catalysts require deep understanding of the structure of active site and their evolution along the catalytic reactions, which requires a complex approach exploiting state-of-the-art methods of chemical synthesis, operando characterization and computer modelling. Although the importance of such approach is obvious, its practical application remains challenging and the structure of metal coordination complexes, which play the key role in catalysis, remains unresolved for many catalytic processes. This project is aimed to delopping of novel industrially relevant Ru- and Pd-based catalysts for C-H activation and C-O hydrogenation and their multi-technique XANES/EXAFS/FTIR operando characterization based on machine learning algorithms. This multi-stage interdisciplinary project involves not only chemical synthesis of new functional materials for fine chemical and pharmaceutical industries, but also achieving fundamental understanding of the structure-reactivity relationships of their active metal sites, as well as developing novel innovative machine learning-based algorithms, that can be be further applied for a wide range of problems in materials science.

In particular, within this project, we will develop a method based on machine learning (ML) for quantitative analysis of X-ray absorption near-edge structure (XANES), extended X-ray absorption fine structure (EXAFS) and Fourier-transform infrared (FTIR) spectra. Combination of these spectroscopies allows descripting both the structure of the catalytically active metal center and adsorbed organic species. The main challenge of the analysis of such data for real catalytic systems is the dynamic evolution of active metal sites along the reaction, so that their structures may not correspond to any known reference structures from the crystallographic databases. For the unknown structures, comparison with theoretically calculated spectra is, in principle possible, but requires (especially for XANES and FTIR spectra) utilization of hypothetical structures with huge number of variable parameters. For such multiparametric problem, ML algorithms are highly appropriate. However, the application of ML for analysis for X-ray absorption spectra have been shown only recently by few research groups [4,5], including the team of this project [6]. Thus, development of sustainable algorithms and its further delivery as a user-friendly software to the scientific community will become a considerable and innovative achievement in the field of material science. Moreover, unlike the existing methods in quantitative spectroscopy, our algorithm will perform simultaneous analysis of three different spectral regions, which will improve stability and precision of the obtained results. We will establish a database of theoretical XANES and FTIR spectra, describing more than 15000 of different coordinations of Pd- and Ru-complexes in different charge state, train ML algorithm and further apply it to real experimental data.

The developed algorithms will be applied to determine fundamental structure-reactivity relationships in two industrially relevant reactions: Ru-catalyzed selective deoxygenation of polyols to olefins and Pd-catalyzed, dehydrogenative coupling of arene C-H bonds to olefinic C-H bonds resulting in the creation of new C-C bonds, for which Leuven group has designed a new class of heterogeneous catalysts [1]. We will develop a specialized operando cell for measurements of solid and liquid samples under pressures up to 20 bar. Operando experiments will be performed using synchrotron radiation sources, including the leading European synchrotron source – ESRF, which undergoes upgrade program and will be reopened for users in 2020. To identify both active and inactive species we aim to explore wide ranges of experimental conditions and sample types. Then, applying multivariate statistical methods we will decompose the whole experimental dataset into the spectra of “pure” states of metal complexes and their concentration profiles, which will be correlated to the reaction products monitored by means of online mass spectrometry and infrared spectroscopy, while the structure of the corresponding metal sites will be determined using ML algorithms according to the calculated spectral database.

Finally, based on the obtained structure-reactivity relationships, we will develop new catalytic materials with improved properties. In particular, for Ru-catalyst for hydrogenative deoxygenation of erythritol, we aim for a TON number ≥ 500 and a selectivity of at least 85 % for butenes. And for Pd-catalysts for arene C-H activation, we will improve its resistance towards deactivation, resulting in overall 2-3 times increase of TON and positional selectivity over 75% in the activation of C-H bonds on the arene ring. This materials will act as prototypes for real catalysis for fine chemical and pharmaceutical industries. Thus, within this project we will be solving both fundamental and methodological problems, as well as developing new materials for practical applications.

[1] Van Velthoven, Smolders, De Vos, et al. Chem Sci, 10, 3616-3622 (2019)
[2] Huber, et al. Angew. Chem. Int. Ed., 43, 1549-1551 (2004)
[3] Tomishige, et al. ChemSusChem, 8, 1114-1132 (2015)
[4] Timoshenko, et al. Nano Lett., 19, 520-529 (2019)
[5] Zheng, et al. Comput. Mater., 4, 9 (2018)
[6] Guda, Soldatov, Bugaev, et al. Catal. Today, 336, 3-21 (2019)