Design and Assembly of Virtual Homogeneous Catalyst Libraries –Towardsin silico Catalyst Optimisation

Leveraging natural language processing to curate the tmCAT, tmPHOTO, tmBIO, and tmSCO datasets of functional transition metal complexes

The breadth of transition metal chemical space covered by databases such as the Cambridge Structural Database and the derived computational database tmQM is not conducive to application-specific modeling and the development of structure-property relationships. Here, we employ both supervised and unsupervised natural language processing (NLP) techniques to link experimentally synthesized compounds in the tmQM database to their respective applications. Leveraging NLP models, we curate four distinct datasets: tmCAT for catalysis, tmPHOTO for photophysical activity, tmBIO for biological relevance, and tmSCO for magnetism. Analyzing the chemical substructures within each dataset reveals common chemical motifs in each of the designated applications. We then use these common chemical structures to augment our initial datasets for each application, yielding a total of 21 631 compounds in tmCAT, 4599 in tmPHOTO, 2782 in tmBIO, and 983 in tmSCO. These datasets are expected to accelerate the more targeted computational screening and development of refined structure-property relationships with machine learning.

Paving the road towards automated homogeneous catalyst design

In the past decade, computational tools have become integral to catalyst design. They continue to offer significant support to experimental organic synthesis and catalysis researchers aiming for optimal reaction outcomes. More recently, data-driven approaches utilizing machine learning have garnered considerable attention for their expansive capabilities. This Perspective provides an overview of diverse initiatives in the realm of computational catalyst design and introduces our automated tools tailored for high-throughput in silico exploration of the chemical space. While valuable insights are gained through methods for high-throughput in silico exploration and analysis of chemical space, their degree of automation and modularity are key. We argue that the integration of data-driven, automated and modular workflows is key to enhancing homogeneous catalyst design on an unprecedented scale, contributing to the advancement of catalysis research.

Mesokinetics as a Tool Bridging the Microscopic-to-Macroscopic Transition to Rationalize Catalyst Design

Heterogeneous catalysis is the workhorse of the chemical industry, and a heterogeneous catalyst possesses numerous active sites working together to drive the conversion of reactants to desirable products. Over the decades, much focus has been placed on identifying the factors affecting the active sites to gain deep insights into the structure-performance relationship, which in turn guides the design and preparation of more active, selective, and stable catalysts. However, the molecular-level interplay between active sites and catalytic function still remains qualitative or semiquantitative, ascribed to the difficulty and uncertainty in elucidating the nature of active sites for its controllable manipulation. Hence, bridging the microscopic properties of active sites and the macroscopic catalytic performance, that is, microscopic-to-macroscopic transition, to afford a quantitative description is intriguing yet challenging, and progress toward this promises to revolutionize catalyst design and preparation.In this Account, we propose mesokinetics modeling, for the first time enabling a quantitative description of active site characteristics and the related mechanistic information, as a versatile tool to guide rational catalyst design. Exemplified by a pseudo-zero-order reaction, the kinetics derivation from the Pt particle size-sensitive catalytic activity and size-insensitive activation energy suggests only one type of surface site as the dominant active site, in which the Pt(111) with almost unchanged turnover frequency (TOF₁₁₁) is further identified as the dominating active site. Such a method has been extended to identify and quantify the number (N_i) of active sites for various thermo-, electro-, and photocatalysts in chemical synthesis, hydrogen generation, environment application, etc. Then, the kinetics derivation from the kinetic compensation effects suggests a thermodynamic balance between the activation entropy and enthalpy, which exhibit linear dependences on Pt charge. Accordingly, the Pt charge can serve as a catalytic descriptor for its quantitative determination of TOF_i. This strategy has been further applied to Pt-catalyzed CO oxidation with nonzero-order reaction characteristic by taking the site coverages of surface species into consideration.Hence, substituting the above statistical correlations of N_i and TOF_i into the rate equation R = ∑N_i × TOF_i offers the mesokinetics model, which can precisely predict catalytic function and screen catalysts. Finally, based on the disentanglement of the factors underlying Pt electronic structures, a de novo strategy, from the interfacial charge distribution to reaction mechanism, kinetics, and thermodynamics parameters of the rate-determining step, and ultimately catalytic performance, is developed to map the unified mechanistic and kinetics picture of reaction. Overall, the mesokinetics not only demonstrates much potential to elucidate the quantitative interplay between active sites and catalytic activity but also provides a new research direction in kinetics analysis to rationalize catalyst design.

Machine Learning for the Discovery, Design, and Engineering of Materials

Machine learning (ML) has become a part of the fabric of high-throughput screening and computational discovery of materials. Despite its increasingly central role, challenges remain in fully realizing the promise of ML. This is especially true for the practical acceleration of the engineering of robust materials and the development of design strategies that surpass trial and error or high-throughput screening alone. Depending on the quantity being predicted and the experimental data available, ML can either outperform physics-based modes, be used to accelerate such models, or be integrated with them to improve their performance. We cover recent advances in algorithms and in their application that are starting to make inroads toward (a) the discovery of new materials through large-scale enumerative screening, (b) the design of materials through identification of rules and principles that govern materials properties, and (c) the engineering of practical materials by satisfying multiple objectives. We conclude with opportunities for further advancement to realize ML as a widespread tool for practical computational materials design. Expected final online publication date for the Annual Review of Chemical and Biomolecular Engineering, Volume 13 is October 2022. Please see http://www.annualreviews.org/page/journal/pubdates for revised estimates.

Computational Discovery of Transition-metal Complexes: From High-throughput Screening to Machine Learning

Transition-metal complexes are attractive targets for the design of catalysts and functional materials. The behavior of the metal-organic bond, while very tunable for achieving target properties, is challenging to predict and necessitates searching a wide and complex space to identify needles in haystacks for target applications. This review will focus on the techniques that make high-throughput search of transition-metal chemical space feasible for the discovery of complexes with desirable properties. The review will cover the development, promise, and limitations of "traditional" computational chemistry (i.e., force field, semiempirical, and density functional theory methods) as it pertains to data generation for inorganic molecular discovery. The review will also discuss the opportunities and limitations in leveraging experimental data sources. We will focus on how advances in statistical modeling, artificial intelligence, multiobjective optimization, and automation accelerate discovery of lead compounds and design rules. The overall objective of this review is to showcase how bringing together advances from diverse areas of computational chemistry and computer science have enabled the rapid uncovering of structure-property relationships in transition-metal chemistry. We aim to highlight how unique considerations in motifs of metal-organic bonding (e.g., variable spin and oxidation state, and bonding strength/nature) set them and their discovery apart from more commonly considered organic molecules. We will also highlight how uncertainty and relative data scarcity in transition-metal chemistry motivate specific developments in machine learning representations, model training, and in computational chemistry. Finally, we will conclude with an outlook of areas of opportunity for the accelerated discovery of transition-metal complexes.

Machine learning dihydrogen activation in the chemical space surrounding Vaska's complex

Homogeneous catalysis using transition metal complexes is ubiquitously used for organic synthesis, as well as technologically relevant in applications such as water splitting and CO2 reduction. The key steps underlying homogeneous catalysis require a specific combination of electronic and steric effects from the ligands bound to the metal center. Finding the optimal combination of ligands is a challenging task due to the exceedingly large number of possibilities and the non-trivial ligand-ligand interactions. The classic example of Vaska's complex, trans-[Ir(PPh3)2(CO)(Cl)], illustrates this scenario. The ligands of this species activate iridium for the oxidative addition of hydrogen, yielding the dihydride cis-[Ir(H)2(PPh3)2(CO)(Cl)] complex. Despite the simplicity of this system, thousands of derivatives can be formulated for the activation of H2, with a limited number of ligands belonging to the same general categories found in the original complex. In this work, we show how DFT and machine learning (ML) methods can be combined to enable the prediction of reactivity within large chemical spaces containing thousands of complexes. In a space of 2574 species derived from Vaska's complex, data from DFT calculations are used to train and test ML models that predict the H2-activation barrier. In contrast to experiments and calculations requiring several days to be completed, the ML models were trained and used on a laptop on a time-scale of minutes. As a first approach, we combined Bayesian-optimized artificial neural networks (ANN) with features derived from autocorrelation and deltametric functions. The resulting ANNs achieved high accuracies, with mean absolute errors (MAE) between 1 and 2 kcal mol-1, depending on the size of the training set. By using a Gaussian process (GP) model trained with a set of selected features, including fingerprints, accuracy was further enhanced. Remarkably, this GP model minimized the MAE below 1 kcal mol-1, by using only 20% or less of the data available for training. The gradient boosting (GB) method was also used to assess the relevance of the features, which was used for both feature selection and model interpretation purposes. Features accounting for chemical composition, atom size and electronegativity were found to be the most determinant in the predictions. Further, the ligand fragments with the strongest influence on the H2-activation barrier were identified.

A Quantitative Model for Alkane Nucleophilicity Based on C-H Bond Structural/Topological Descriptors

A first quantitative model for calculating the nucleophilicity of alkanes is described. A statistical treatment was applied to the analysis of the reactivity of 29 different alkane C-H bonds towards in situ generated metal carbene electrophiles. The correlation of the recently reported experimental reactivity with two different sets of descriptors comprising a total of 86 parameters was studied, resulting in the quantitative descriptor-based alkane nucleophilicity (QDEAN) model. This model consists of an equation with only six structural/topological descriptors, and reproduces the relative reactivity of the alkane C-H bonds. This reactivity can be calculated from parameters emerging from the schematic drawing of the alkane and a simple set of sums.

Computational Ligand Descriptors for Catalyst Design

Ligands, especially phosphines and carbenes, can play a key role in modifying and controlling homogeneous organometallic catalysts, and they often provide a convenient approach to fine-tuning the performance of known catalysts. The measurable outcomes of such catalyst modifications (yields, rates, selectivity) can be set into context by establishing their relationship to steric and electronic descriptors of ligand properties, and such models can guide the discovery, optimization, and design of catalysts. In this review we present a survey of calculated ligand descriptors, with a particular focus on homogeneous organometallic catalysis. A range of different approaches to calculating steric and electronic parameters are set out and compared, and we have collected descriptors for a range of representative ligand sets, including 30 monodentate phosphorus(III) donor ligands, 23 bidentate P,P-donor ligands, and 30 carbenes, with a view to providing a useful resource for analysis to practitioners. In addition, several case studies of applications of such descriptors, covering both maps and models, have been reviewed, illustrating how descriptor-led studies of catalysis can inform experiments and highlighting good practice for model comparison and evaluation.

Predictive and mechanistic multivariate linear regression models for reaction development

Multivariate Linear Regression (MLR) models utilizing computationally-derived and empirically-derived physical organic molecular descriptors are described in this review. Several reports demonstrating the effectiveness of this methodological approach towards reaction optimization and mechanistic interrogation are discussed. A detailed protocol to access quantitative and predictive MLR models is provided as a guide for model development and parameter analysis.

Predicting electronic structure properties of transition metal complexes with neural networks

High-throughput computational screening has emerged as a critical component of materials discovery. Direct density functional theory (DFT) simulation of inorganic materials and molecular transition metal complexes is often used to describe subtle trends in inorganic bonding and spin-state ordering, but these calculations are computationally costly and properties are sensitive to the exchange-correlation functional employed. To begin to overcome these challenges, we trained artificial neural networks (ANNs) to predict quantum-mechanically-derived properties, including spin-state ordering, sensitivity to Hartree-Fock exchange, and spin-state specific bond lengths in transition metal complexes. Our ANN is trained on a small set of inorganic-chemistry-appropriate empirical inputs that are both maximally transferable and do not require precise three-dimensional structural information for prediction. Using these descriptors, our ANN predicts spin-state splittings of single-site transition metal complexes (i.e., Cr-Ni) at arbitrary amounts of Hartree-Fock exchange to within 3 kcal mol-1 accuracy of DFT calculations. Our exchange-sensitivity ANN enables improved predictions on a diverse test set of experimentally-characterized transition metal complexes by extrapolation from semi-local DFT to hybrid DFT. The ANN also outperforms other machine learning models (i.e., support vector regression and kernel ridge regression), demonstrating particularly improved performance in transferability, as measured by prediction errors on the diverse test set. We establish the value of new uncertainty quantification tools to estimate ANN prediction uncertainty in computational chemistry, and we provide additional heuristics for identification of when a compound of interest is likely to be poorly predicted by the ANN. The ANNs developed in this work provide a strategy for screening transition metal complexes both with direct ANN prediction and with improved structure generation for validation with first principles simulation.

Predicting the performance of oxidation catalysts using descriptor models

Mix & match: we show that combining simple heuristic models with experimental validation is an effective method for optimising supported mixed oxide catalysts.

Lost in chemical space? Maps to support organometallic catalysis

Descriptors calculated from molecular structures have been used to map different areas of chemical space. A number of applications for such maps can be identified, ranging from the fine-tuning and optimisation of catalytic activity and compound properties to virtual screening of novel compounds, as well as the exhaustive exploration of large areas of chemical space by automated combinatorial building and evaluation. This review focuses on organometallic catalysis, but also touches on other areas where similar approaches have been used, with a view to assessing the extent to which chemical space has been explored.

Graphical abstract

Effective donor abilities of E-t-Bu and EPh (E = O, S, Se, Te) to a high valent transition metal

Amido rotation in the chromium(vi), d(0)-system NCr(NPr(i)2)2X is under investigation as a method for the parameterization of ligands for their donor properties toward high valent metals. In this study, two new series were prepared and studied based on chalcogenide ligands, X = EBu(t) and EPh and where E = O, S, Se, Te; the OPh and SPh compounds were previously reported. The ligand donor parameters for these ligands correlate with the Cr-E-C angles in these chalcogenide series. In addition, it was found that NBO calculated overlaps and DFT calculated bond dissociation enthalpies correlate within X = halide-, EBu(t)- and EPh-series. All of the new complexes were characterized by X-ray diffraction.

Screening substituent and backbone effects on the properties of bidentate P,P-donor ligands (LKB-PP(screen))

We present a computational exploration of the effect of systematic variation of backbones and substituents on the properties of bidentate, cis-chelating P,P donor ligands as captured by calculated parameters. The parameters used are the same as reported for our ligand knowledge base for bidentate P,P donor ligands, LKB-PP (Organometallics 2008, 27, 1372-1383; Organometallics 2012 31, 5302-5306), but calculation protocols have been streamlined, suitable for an extensive evaluation of ligand structures. Analysis of the resulting LKB-PP(screen) database with principal component analysis (PCA) captures the effects of changing backbones and substituents on ligand properties and illustrates how these are complementary variables for these ligands. While backbone variation is routinely employed in ligand synthesis to modify catalyst properties, only a limited subset of substituents is commonly accessed and here we highlight substituents which are likely to generate new ligand properties, of interest for the design and improved sampling of bidentate ligands in homogeneous organometallic catalysis.

High throughput technology: approaches of research in homogeneous and heterogeneous catalysis

High throughput experimentation (HTE) approaches and the choice of the design of experiment (DoE) tools are discussed with regard to their convenience and applicability in homogeneous and heterogeneous catalysis as a concerted workflow. Much attention is given to diverse methodologies and strategies, which are fundamental for the experimental planning. For two target reactions in two case studies presented in this chapter, HTE methods were applied to create and evaluate catalyst libraries. A homogeneous catalyst case study is illustrated first, which deals with parallel synthesis and screening of organometallic catalysts in the polymerisation of ethylene. The second case study (heterogeneous catalysis) focuses on coherent synthesis and testing of dopant effects on the performance of oxidation catalysts in a reaction of transformation of n-butane to maleic anhydride. Supporting examples from the literature described here show that careful planning of libraries and test conditions is vital in high throughput experimentation in order to deliver meaningful results leading to performance improvements or disruptive new findings.

Expansion of the Ligand Knowledge Base for Chelating P,P-Donor Ligands (LKB-PP)

We have expanded the ligand knowledge base for bidentate P,P- and P,N-donor ligands (LKB-PP, Organometallics2008, 31, 1372-1383) by 208 ligands and introduced an additional steric descriptor (nHe8). This expanded knowledge base now captures information on 334 bidentate ligands and has been processed with principal component analysis (PCA) of the descriptors to produce a detailed map of bidentate ligand space, which better captures ligand variation and has been used for the analysis of ligand properties.

Evaluation of donor and steric properties of anionic ligands on high valent transition metals

Synthetic protocols and characterization data for a variety of chromium(VI) nitrido compounds of the general formula NCr(NPr(i)(2))(2)X are reported, where X = NPr(i)(2) (1), I (2), Cl (3), Br (4), OTf (5), 1-adamantoxide (6), OSiPh(3) (7), O(2)CPh (8), OBu(t)(F6) (9), OPh (10), O-p-(OMe)C(6)H(4) (11), O-p-(SMe)C(6)H(4) (12), O-p-(Bu(t))C(6)H(4) (13), O-p-(F)C(6)H(4) (14), O-p-(Cl)C(6)H(4) (15), O-p-(CF(3))C(6)H(4) (16), OC(6)F(5) (17), κ(O)-N-oxy-phthalimide (18), SPh (19), OCH(2)Ph (20), NO(3) (21), pyrrolyl (22), 3-C(6)F(5)-pyrrolyl (23), 3-[3,5-(CF(3))(2)C(6)H(3)]pyrrolyl (24), indolyl (25), carbazolyl (26), N(Me)Ph (27), κ(N)-NCO (28), κ(N)-NCS (29), CN (30), NMe(2) (31), F (33). Several different techniques were employed in the syntheses, including nitrogen-atom transfer for the formation of 1. A cationic chromium complex [NCr(NPr(i)(2))(2)(DMAP)]BF(4) (32) was used as an intermediate for the production of 33, which was produced by tin-catalyzed degredation of the salt. Using spin saturation transfer or line shape analysis, the free energy barriers for diisopropylamido rotation were studied. It is proposed that the estimated enthalpic barriers, Ligand Donor Parameters (LDPs), for amido rotation can be used to parametrize the donor abilities of this diverse set of anionic ligands toward transition metal centers in low d-electron counts. The new LDPs do not correlate well to the pK(a) value of X. Conversely, the LDP values of phenoxide ligands do correlate with Hammett parameters for the para-substituents. Literature data for (13)C NMR chemical shifts for a tungsten-based system with various X ligands plotted versus LDP provided a linear fit. In addition, the angular overlap model derived e(σ) + e(π) values for chromium(III) ammine complexes correlate with LDP values. Also discussed is the correlation with XTiCp*(2) spectroscopic data. X-ray diffraction has been used used to characterize 31 of the compounds. From the X-ray diffraction data, steric parameters for the ligands using the Percent Buried Volume and Solid Angle techniques were found.

Steric nature of the bite angle. A closer and a broader look

The bite angle (ligand-metal-ligand angle) is known to greatly influence the activity of catalytically active transition-metal complexes towards bond activation. Here, we have computationally explored how and why the bite angle has such effects in a wide range of prototypical C-X bonds and palladium complexes, using relativistic density functional theory at ZORA-BLYP/TZ2P. Our model reactions cover the substrates H(3)C-X (with X = H, CH(3), Cl) and, among others, the model catalysts, Pd[PH(2)(CH(2))(n)PH(2)] (with n = 2-6) and Pd[PR(2)(CH(2))(n)PR(2)] (n = 2-4 and R = Me, Ph, t-Bu, Cl), Pd(PH(3))X(-) (X = Cl, Br, I), as well as palladium complexes of chelating and non-chelating N-heterocyclic carbenes. The purpose is to elaborate on an earlier finding that bite-angle effects have a predominantly (although not exclusively) steric nature: a smaller bite angle makes more room for coordinating a substrate by bending away the ligands. Indeed, the present results further consolidate this steric picture by revealing its occurrence in this broader range of model reactions and by identifying and quantifying the exact working mechanism through activation strain analyses.

Finding Furfural Hydrogenation Catalysts via Predictive Modelling

We combine multicomponent reactions, catalytic performance studies and predictive modelling to find transfer hydrogenation catalysts. An initial set of 18 ruthenium-carbene complexes were synthesized and screened in the transfer hydrogenation of furfural to furfurol with isopropyl alcohol complexes gave varied yields, from 62% up to >99.9%, with no obvious structure/activity correlations. Control experiments proved that the carbene ligand remains coordinated to the ruthenium centre throughout the reaction. Deuterium-labelling studies showed a secondary isotope effect (k(H):k(D)=1.5). Further mechanistic studies showed that this transfer hydrogenation follows the so-called monohydride pathway. Using these data, we built a predictive model for 13 of the catalysts, based on 2D and 3D molecular descriptors. We tested and validated the model using the remaining five catalysts (cross-validation, R(2)=0.913). Then, with this model, the conversion and selectivity were predicted for four completely new ruthenium-carbene complexes. These four catalysts were then synthesized and tested. The results were within 3% of the model's predictions, demonstrating the validity and value of predictive modelling in catalyst optimization.