David Balcells

• Sandonas, Leonardo Medrano; Badia, David Balcells; Barnowsky, Tom; Bochkarev, Anton; Cole, Jacqueline & Deringer, Volker L. [Show all 24 contributors for this article] (2026). Towards sustainable exploration of chemical spaces with machine learning. Digital Discovery. doi: 10.1039/d6dd00157b. Full text in Research Archive Show summary

  Artificial intelligence is transforming molecular and materials science, but its growing computational and data demands raise critical sustainability challenges. In this perspective, we examine resource considerations across the AI-driven discovery pipeline–from quantum-mechanical (QM) data generation and model training to automated, self-driving research workflows–building on discussions from the “SusML workshop: towards sustainable exploration of chemical spaces with machine learning” held in Dresden, Germany. In this context, the availability of large quantum datasets has enabled rigorous benchmarking and rapid methodological progress, while also incurring substantial energy and infrastructure costs. We highlight emerging strategies to enhance efficiency, including general-purpose machine learning (ML) models, multi-fidelity approaches, model distillation, and active learning. Moreover, incorporating physics-based constraints within hierarchical workflows, where fast ML surrogates are applied broadly and high-accuracy QM methods are used selectively, can further optimize resource use without compromising reliability. Equally important is bridging the gap between idealized computational predictions and real-world conditions by accounting for synthesizability and multi-objective design criteria, which is essential for practical impact. Finally, we argue that sustainable progress will rely on open data and models, reusable workflows, and domain-specific AI systems that maximize scientific value per unit of computation, enabling efficient and responsible discovery of technological materials and therapeutics.

• Kneiding, Hannes & Badia, David Balcells (2026). Δ-Machine Learning for the Prediction of Metal Complex Properties. Chemistry - A European Journal. ISSN 0947-6539. doi: 10.1002/chem.71487. Full text in Research Archive
• Cibinel, Luca; Linjordet, Trond; Pensar, Johan; Badia, David Balcells; Bin, Riccardo De & Ell, Basil (2026). tmQM-RDF Data Set: A Knowledge Graph Representing Transition Metal Complexes. Journal of Chemical Information and Modeling. ISSN 1549-9596. 66(13), p. 7524–7538. doi: 10.1021/acs.jcim.6c01281. Full text in Research Archive Show summary

  Transition metal complexes (TMCs) have wide-ranging practical utility in chemistry, with possible applications that range from catalysis to medicinal chemistry. The study of TMCs and their properties is thus a field rich with potential, one in which machine learning and computational approaches can offer a substantial aid. For this reason, appropriate and accessible data sets, collecting a wide range of information, are required in order to facilitate the effective analysis and investigation of such compounds. This paper contributes to the data modeling effort via the introduction of the transition metal quantum mechanics RDF (tmQM-RDF) data set, a knowledge graph constructed using the Resource Description Framework (RDF) vocabulary which collects rich and detailed descriptions of approximately 60k TMCs. These descriptions are both qualitative and quantitative, encompassing the compositional nature of TMCs as well as the entirety of their molecular graphs. The Python package tmqmrdfdata is released together with the knowledge graph to provide a user-friendly interface. An example of the power of the proposed representation is presented, showcasing how the information available in tmQM-RDF can be exploited for TMC manipulation tasks, achieving promising performance even with relatively simple probabilistic models.

• Denjean, Aurore Evelyne Francine; Badia, David Balcells & Flores, Ainara Nova (2026). Impact of metal and active site configurations in hydrogenation reactions with N-doped graphene single atom catalysts. Journal of Materials Chemistry A. ISSN 2050-7488. 14(12), p. 6885–6893. doi: 10.1039/d5ta08094k. Full text in Research Archive Show summary

  Hydrogenation reactions are well-established transformations in both homogeneous and heterogeneous catalysis and are increasingly explored using single-atom catalysts (SACs). Despite this progress, a comprehensive understanding of the underlying reaction mechanisms remains limited, often restricted to specific systems. Moreover, the precise nature of the active sites is elusive, and their reactivity may be influenced by varying coordination numbers, hetero-atom doping, and other factors. To gain insight into hydrogenation reactions in nitrogen-doped graphene-based SACs, we conducted a thorough investigation into hydrogen transfer across Fe, Co, Mn, and Ru systems, considering different charges, spin states, pyrrolic and pyridinic sites. Our findings reveal substantial deviation from conventional homogeneous and heterogeneous systems, with SACs being strongly influenced by the nature of the active site. Analyses using Natural Bond Orbitals (NBO), natural charge, and natural decomposition analysis (NEDA) highlighted differences in nitrogen-metal interactions as a key factor driving the observed reactivity variations between Pyrr and Py systems, as well as between Ru and 1st-row metals.

• Safy, Mohamed E. A.; Rama, Raquel J.; Both, Niklas F.; Badia, David Balcells; Junge, Kathrin & Beller, Matthias [Show all 7 contributors for this article] (2026). Leveraging Diamines to Unlock the Mn‐MACHO Catalyst in the Reduction of CO2 to Methanol. Angewandte Chemie International Edition. ISSN 1433-7851. 65(10). doi: 10.1002/anie.202524012. Full text in Research Archive Show summary

  The conversion of CO 2 into energy‐dense liquid fuels, such as methanol, represents a cornerstone of sustainable chemistry; however, most homogeneous catalytic systems still rely on noble metals or Lewis acid additives. Here, we report the first protocol for amine‐assisted CO 2 hydrogenation to methanol using a Mn–MACHO catalyst without any Lewis acid co‐catalyst, achieving turnover numbers up to 45.2, the highest reported for Mn systems. A combined computational, microkinetic, and experimental study reveals that diamines dramatically enhance activity compared to monoamines by promoting a highly exergonic double amidation step. This thermodynamic driving force shifts the equilibrium away from formate resting states toward the active catalyst, thereby accelerating methanol formation. The correlation established between amidation free energies (Δ G amidation ) and methanol productivity provides a rational design principle for tailoring amine promoters across Ru‐ and Mn‐based MACHO catalysts. These insights advance the development of sustainable, base‐metal‐catalyzed CO 2 conversion strategies and open opportunities for integrated carbon capture and utilization.

• Meggio, Claudio; Pensar, Johan; Badia, David Balcells & Bin, Riccardo De (2026). A path-based boosting algorithm for exploring transition metal compounds. Statistical Modelling. ISSN 1471-082X. doi: 10.1177/1471082x251407458. Full text in Research Archive Show summary

  Motivated by the analysis of chemical metal compounds and their properties for catalysis, we developed a gradient boosting model that explores graph structures to perform prediction tasks. Taking advantage of the iterative nature of boosting, our novel approach, called PathBoost , explores the graphs to identify the relevant paths and simultaneously fits a prediction model. Advantages of PathBoost include automatic variable selection, as only relevant paths are kept in the model and explainability, as a measure of variable importance is provided. The novel algorithm is applied to the tmQM dataset, where the molecules are represented as graphs with atoms as nodes and bonds as edges. The goal is to predict specific quantum properties of the molecules, such as the highest occupied molecular orbital (HOMO) and lowest unoccupied molecular orbital (LUMO) gap. These properties usually require heavy computational power to be computed, while our model aims to provide comparable results using much fewer resources.

• Rasmussen, Maria H.; Strandgaard, Magnus; Seumer, Julius; Hemmingsen, Laura K.; Frei, Angelo & Badia, David Balcells [Show all 7 contributors for this article] (2025). SMILES all around: structure to SMILES conversion for transition metal complexes. Journal of Cheminformatics. 17. doi: 10.1186/s13321-025-01008-1. Full text in Research Archive Show summary

  We present a method for creating RDKit-parsable SMILES for transition metal complexes (TMCs) based on xyz-coordinates and overall charge of the complex. This can be viewed as an extension to the program xyz2mol that does the same for organic molecules. The only dependency is RDKit, which makes it widely applicable. One thing that has been lacking when it comes to generating SMILES from structure for TMCs is an existing SMILES dataset to compare with. Therefore, sanity-checking a method has required manual work. Therefore, we also generate SMILES two other ways; one where ligand charges and TMC connectivity are based on natural bond orbital (NBO) analysis from density functional theory (DFT) calculations utilizing recent work by Kneiding et al. (Digit Discov 2: 618–633, 2023). Another one fixes SMILES available through the Cambridge Structural Database (CSD), making them parsable by RDKit. We compare these three different ways of obtaining SMILES for a subset of the CSD (tmQMg) and find >70% agreement for all three pairs. We utilize these SMILES to make simple molecular fingerprint (FP) and graph-based representations of the molecules to be used in the context of machine learning. Comparing with the graphs made by Kneiding et al. where nodes and edges are featurized with DFT properties, we find that depending on the target property (polarizability, HOMO-LUMO gap or dipole moment) the SMILES based representations can perform equally well. This makes them very suitable as baseline-models. Finally we present a dataset of 227k RDKit parsable SMILES for mononuclear TMCs in the CSD. Scientific contribution We present a method that can create RDKit-parsable SMILES strings of transition metal complexes (TMCs) from Cartesian coordinates and use it to create a dataset of 227k TMC SMILES strings. The RDKit-parsability allows us to generate perform machine learning studies of TMC properties using ”standard” molecular representations such as fingerprints and 2D-graph convolution. We show that these relatively simple representations can perform quite well depending on the target property.

• H., Rasmussen, Maria; Magnus, Strandgaard,; Julius, Seumer,; K., Hemmingsen, Laura; Angelo, Frei, & Badia, David Balcells [Show all 7 contributors for this article] (2025). SMILES all around: structure to SMILES conversion for transition metal complexes | Journal of Cheminformatics | Springer Nature Link. Journal of Cheminformatics. 17. doi: 10.1186/s13321-025-01008-1. Full text in Research Archive Show summary

  We present a method for creating RDKit-parsable SMILES for transition metal complexes (TMCs) based on xyz-coordinates and overall charge of the complex. This can be viewed as an extension to the program xyz2mol that does the same for organic molecules. The only dependency is RDKit, which makes it widely applicable. One thing that has been lacking when it comes to generating SMILES from structure for TMCs is an existing SMILES dataset to compare with. Therefore, sanity-checking a method has required manual work. Therefore, we also generate SMILES two other ways; one where ligand charges and TMC connectivity are based on natural bond orbital (NBO) analysis from density functional theory (DFT) calculations utilizing recent work by Kneiding et al. (Digit Discov 2: 618–633, 2023). Another one fixes SMILES available through the Cambridge Structural Database (CSD), making them parsable by RDKit. We compare these three different ways of obtaining SMILES for a subset of the CSD (tmQMg) and find >70% agreement for all three pairs. We utilize these SMILES to make simple molecular fingerprint (FP) and graph-based representations of the molecules to be used in the context of machine learning. Comparing with the graphs made by Kneiding et al. where nodes and edges are featurized with DFT properties, we find that depending on the target property (polarizability, HOMO-LUMO gap or dipole moment) the SMILES based representations can perform equally well. This makes them very suitable as baseline-models. Finally we present a dataset of 227k RDKit parsable SMILES for mononuclear TMCs in the CSD. Scientific contribution We present a method that can create RDKit-parsable SMILES strings of transition metal complexes (TMCs) from Cartesian coordinates and use it to create a dataset of 227k TMC SMILES strings. The RDKit-parsability allows us to generate perform machine learning studies of TMC properties using ”standard” molecular representations such as fingerprints and 2D-graph convolution. We show that these relatively simple representations can perform quite well depending on the target property.

• Kneiding, Hannes & Badia, David Balcells (2025). tmQMg* Data Set: Excited State Properties of 74k Transition Metal Complexes. Journal of Chemical Information and Modeling. ISSN 1549-9596. 65(21), p. 11766–11777. doi: 10.1021/acs.jcim.5c01958. Full text in Research Archive Show summary

  The application of machine learning approaches to meaningful problems in chemistry and materials science is still challenged by the limited availability of data. In order to close this gap, we report the tmQMg* data set, which provides excited state properties for 74k mononuclear transition metal complexes extracted from the Cambridge Structural Database. All properties were computed at the TD-DFT ωB97xd/def2SVP level of theory. The strongest electron excitations in the ultraviolet, visible, and near-infrared ranges are included, together with the wavelengths and intensities of the first 30 excited states. Further, natural transition orbitals were computed for the strongest excitations in the visible range to determine the nature of the associated charge transfers. By computing the TD-DFT spectra in both gas phase and acetone, we quantified solvatochromic effects, which are also provided with the data set, in terms of both wavelength shifts and intensity changes. The tmQMg* data set will enable the development of discriminative and generative artificial intelligence models with respect to absorption spectra, charge transfer character, and solvatochromism, enabling novel advances in the field of transition metal photochemistry.

• Badia, David Balcells (2025). Co-intelligent Design of Catalysis Research with Large Language Models: Hype or Reality? ACS Catalysis. 15(18), p. 16412–16420. doi: 10.1021/acscatal.5c04238. Full text in Research Archive
• Strandgaard, Magnus; Linjordet, Trond; Kneiding, Hannes; Burnage, Arron Luc; Nova, Ainara & Jensen, Jan Halborg [Show all 7 contributors for this article] (2025). A Deep Generative Model for the Inverse Design of Transition Metal Ligands and Complexes. JACS Au. 5(5), p. 2294–2308. doi: 10.1021/jacsau.5c00242. Full text in Research Archive Show summary

  Deep generative models yielding transition metal complexes (TMCs) remain scarce despite the key role of these compounds in industrial catalytic processes, anticancer therapies, and the energy transition. Compared to drug discovery within the chemical space of organic molecules, TMCs pose further challenges, including the encoding of chemical bonds of higher complexity and the need to optimize multiple properties. In this work, we developed a generative model for the inverse design of transition metal ligands and complexes, based on the junction tree variational autoencoder (JT-VAE). After implementing a SMILES-based encoding of the metal–ligand bonds, the model was trained with the tmQMg-L ligand library, allowing for the generation of thousands of novel, highly diverse monodentate (κ1) and bidentate (κ2) ligands, including imines, phosphines, and carbenes. Further, the generated ligands were labeled with two target properties reflecting the stability and electron density of the associated homoleptic iridium TMCs: the HOMO–LUMO gap (ϵ) and the charge of the metal center (qIr). This data was used to implement a conditional model that generated ligands from a prompt, with the single- or dual-objective of optimizing either or both the ϵ and qIr properties and allowing for chemical interpretation based on the optimization trajectories. The optimizations also had an impact on other chemical properties, including ligand dissociation energies and oxidative addition barriers. A similar model was implemented to condition ligand generation by solubility and steric bulk.

• Burnage, Arron Luc; Kneiding, Hannes; Nova, Ainara & Badia, David Balcells (2025). Cross-coupling catalysts for inactive substrates from a three-dimensional virtual screening. Journal of Catalysis. ISSN 0021-9517. 452. doi: 10.1016/j.jcat.2025.116446. Full text in Research Archive Show summary

  We report a three-dimensional virtual screening method to search a chemical space formed of ligands from the tmQMg-L library for Pd Suzuki–Miyaura cross-coupling catalysts that has future potential to be applied to a broader range of catalytic reactions. It utilises three criteria related to important steps in the catalytic cycle of the reaction to distinguish between candidate Pd(0) species, namely the oxidative addition and reductive elimination barriers and the binding energy of acetonitrile. This method ultimately leads to identifying a region of chemical space that provides a good compromise between the three search criteria, highlighting a smaller set of ligands that merits further exploration. After the screening, several candidates are confirmed as catalysts based on literature searches. We also found Pd(0) complexes with high potential in catalytic cross-couplings involving less reactive substrates that require the oxidative addition of strong C–O bonds.

• Borys, Andryj. M.; Denjean, Aurore Evelyne Francine; Vedani, Luca; Badia, David Balcells & Hevia, Eva (2025). Biphenylene Coordination and Ring‐Opening by Alkali‐Metal Nickelates. Angewandte Chemie International Edition. ISSN 1433-7851. 64(19). doi: 10.1002/anie.202501995. Full text in Research Archive Show summary

  Abstract The pre‐coordination of substrates to heterobimetallic complexes plays a key role in facilitating challenging bond activations and catalytic reactions. Herein, we report a family of low‐valent phenyl‐alkali‐metal nickelates containing a η 4 ‐coordinated biphenylene ligand and demonstrate how the coordination and solvation preferences of the alkali‐metal cation (Li, Na, or K) influence the rate of C─C bond oxidative addition at nickel through combined structural, spectroscopic and computational studies. The rate of C─C bond cleavage can be further manipulated by replacing the two phenyl‐carbanionic ligands with a cyclic 2,2′‐biphenyl scaffold, which affords thermally stable and symmetric spirocyclic nickelates upon ring‐opening.

• Gonzalez, Lucia Moran; Burnage, Arron Luc; Flores, Ainara Nova & Badia, David Balcells (2025). AI Approaches to Homogeneous Catalysis with Transition Metal Complexes. ACS Catalysis. 15, p. 9089–9105. doi: 10.1021/acscatal.5c01202. Full text in Research Archive Show summary

  Artificial intelligence (AI) is transforming research in chemistry, including homogeneous catalysis with transition metals. Over the past 15 years, the number of publications combining AI with catalysis has increased exponentially, reflecting the interest and strength of this strategy in the field. Since this is a broad emerging discipline, it is essential to establish guidelines that clarify the diverse approaches already available. The complexity of the tasks that can be carried out with AI tools is directly linked to the nature of their components, including datasets, representations, algorithms, and high-throughput experimental and computational facilities. In parallel to the evolution of these tools, applications to catalysis have also advanced. Initially, models were developed to predict key aspects of the reaction mechanism, aiming at screening catalyst candidates. Subsequent studies have incorporated experimental data to optimize reaction conditions and yields. More recently, generative AI based on deep learning methods has enabled the inverse design of novel catalysts with predefined target properties. While most studies rely on computational data, recent advancements have improved the acquisition of experimental data, enabling AI-driven automated workflows. This Perspective gives a critical overview on selected studies that reflect the state of the art in the application of AI to homogeneous metal-catalyzed reactions, also highlighting future opportunities and challenges.

• Denjean, Aurore Evelyne Francine; Flores, Ainara Nova & Badia, David Balcells (2024). Borrowing Hydrogen Mechanism in Amine Alkylation by Single Atom Nickel Catalysts. ACS Catalysis. 14(15), p. 11332–11342. doi: 10.1021/acscatal.4c02182. Full text in Research Archive Show summary

  The alkylation of amines with alcohols by borrowing hydrogen (BH) is a versatile synthetic route to various relevant compounds, including drugs and agrochemicals. The atom economy of this reaction, with water as the sole byproduct, entails a complex mechanism in which alcohol activation can follow multiple competing pathways. The interest in single atom catalysts (SACs) combining the advantages of homogeneous and heterogeneous systems motivated this computational study on BH catalyzed by nickel SACs. We found that the oxidation, charge, and spin states of the system, as well as the geometry of the support, all have a strong influence on the BH mechanism. The pyrrolic support was the most reactive in a dicationic state with nearly degenerate singlet and triplet configurations. The free energy of the catalyst hydrogenation by H2 was strongly correlated to its reactivity, thus being a useful descriptor. Further, the mechanism followed an unusual reaction pathway in which the support takes and delivers the H equivalents, with the metal playing a spectator role. This study provides useful insights for the further development of BH SACs.

• Gonzalez, Lucia Moran; Betten, Jørn Eirik; Kneiding, Hannes & Badia, David Balcells (2024). AABBA Graph Kernel: Atom-Atom, Bond-Bond, and Bond-Atom Autocorrelations for Machine Learning. Journal of Chemical Information and Modeling. ISSN 1549-9596. 64(23), p. 8756–8769. doi: 10.1021/acs.jcim.4c01583. Full text in Research Archive Show summary

  Graphs are one of the most natural and powerful representations available for molecules; natural because they have an intuitive correspondence to skeletal formulas, the language used by chemists worldwide, and powerful, because they are highly expressive both globally (molecular topology) and locally (atom and bond properties). Graph kernels are used to transform molecular graphs into fixed-length vectors, which, based on their capacity of measuring similarity, can be used as fingerprints for machine learning (ML). To date, graph kernels have mostly focused on the atomic nodes of the graph. In this work, we developed a graph kernel based on atom-atom, bond-bond, and bond-atom (AABBA) autocorrelations. The resulting vector representations were tested on regression ML tasks on a data set of transition metal complexes; a benchmark motivated by the higher complexity of these compounds relative to organic molecules. In particular, we tested different flavors of the AABBA kernel in the prediction of the energy barriers and bond distances of the Vaska’s complex data set (Friederich et al., Chem. Sci., 2020, 11, 4584). For a variety of ML models, including neural networks, gradient boosting machines, and Gaussian processes, we showed that AABBA outperforms the baseline including only atom-atom autocorrelations. Dimensionality reduction studies also showed that the bond-bond and bond-atom autocorrelations yield many of the most relevant features. We believe that the AABBA graph kernel can accelerate the exploration of large chemical spaces and inspire novel molecular representations in which both atomic and bond properties play an important role.

• Kneiding, Hannes & Badia, David Balcells (2024). Augmenting Genetic Algorithms with Machine Learning for Inverse Molecular Design. Chemical Science. ISSN 2041-6520. doi: 10.1039/d4sc02934h. Full text in Research Archive Show summary

  Evolutionary and machine learning methods have been successfully applied to the generation of molecules and materials exhibiting desired properties. The combination of these two paradigms in inverse design tasks can yield powerful methods that explore massive chemical spaces more efficiently, improving the quality of the generated compounds. However, such synergistic approaches are still an incipient area of research and appear underexplored in the literature. This perspective covers different ways of incorporating machine learning approaches into evolutionary learning frameworks, with the overall goal of increasing the optimization efficiency of genetic algorithms. In particular, machine learning surrogate models for faster fitness function evaluation, discriminator models to control population diversity on-the-fly, machine learning based crossover operations, and evolution in latent space are discussed. The further potential of these synergistic approaches in generative tasks is also assessed, outlining promising directions for future developments.

• Kneiding, Hannes; Flores, Ainara Nova & Badia, David Balcells (2024). Directional multiobjective optimization of metal complexes at the billion-system scale. Nature Computational Science. 4, p. 263–273. doi: 10.1038/s43588-024-00616-5. Full text in Research Archive Show summary

  The discovery of transition metal complexes (TMCs) with optimal properties requires large ligand libraries and efficient multiobjective optimization algorithms. Here we provide the tmQMg-L library, containing 30k diverse and synthesizable ligands with robustly assigned charges and metal coordination modes. tmQMg-L enabled the generation of 1.37 million palladium TMCs, which were used to develop and benchmark the Pareto-Lighthouse multiobjective genetic algorithm (PL-MOGA). With fine control over aim and scope, this algorithm maximized both the polarizability and highest occupied molecular orbital–lowest unoccupied molecular orbital gap of the TMCs within selected regions of the Pareto front, without requiring prior knowledge on the objective limits. Instead of genetic operations on small ligand fragments, the PL-MOGA did whole-ligand mutation and crossover operations, which in chemical spaces containing billions of systems, yielded thousands of highly diverse TMCs in an interpretable manner.

• Luthra, Mahika; Castro, Abril C; Badia, David Balcells; Daasbjerg, Kim & Flores, Ainara Nova (2024). Metal-Dependent Mechanism of the Electrocatalytic Reduction of CO<inf>2</inf> by Bipyridine Complexes Bearing Pendant Amines: A DFT Study. ACS Organic & Inorganic Au. doi: 10.1021/acsorginorgau.4c00046. Full text in Research Archive Show summary

  In this study, the electrocatalytic reduction of carbon dioxide by MnI, ReI, and RuII bipyridine complexes bearing pendant amines is evaluated by DFT methods. Prior experimental studies showed that introducing pendant amines in the secondary coordination sphere of the catalyst shifts product selectivity from CO to HCOO– (in the presence of a proton source) in the case of Mn. In contrast, CO is the major product with Re and Ru. This work includes a comprehensive study of the pathways leading to CO, HCOO–, and H2 to elucidate the energetic preferences that underlie product selectivity. Our results show that switching the metal center leads to changes in the preferred mechanism. While with Mn, the reaction is preferred in an endo configuration, allowing the participation of amines in the hydride formation, reactivity on the exo configuration is preferred with Re. In addition, the distinct redox properties of Re allow for the formation of Re OCOCO2-bridged adducts that lead to CO without a proton source. Further, the ability of Ru to exchange the two Cl– anions changes the preferred coordination number of Ru compared to Mn and Re and, consequently, its reaction mechanism. Overall, this study provides the structure and reactivity insight needed for further catalyst design.

• Cao, Ning; Aguilera, Abril Castro; Badia, David Balcells; Olsbye, Unni & Flores, Ainara Nova (2024). Copper(II)-Oxyl Formation in a Biomimetic Complex Activated by Hydrogen Peroxide: The Key Role of Trans-Bis(Hydroxo) Species. Inorganic Chemistry. ISSN 0020-1669. 63(49), p. 23082–23094. doi: 10.1021/acs.inorgchem.4c01948. Full text in Research Archive Show summary

  Enzymes in nature, such as the copper-based lytic polysaccharide monooxygenases (LPMOs), have gained significant attention for their exceptional performance in C–H activation reactions. The use of H2O2 by LPMOs enzymes has also increased the interest in understanding the oxidation mechanism promoted by this oxidant. While some literature proposes Fenton-like chemistry involving the formation of Cu(II)–OH species and the hydroxyl radical, others contend that Cu(I) activation by H2O2 yields a Cu(II)–oxyl intermediate. In this study, we focused on a bioinspired Cu(I) complex to investigate the reaction mechanism of its oxidation by H2O2 using density functional theory and ab initio molecular dynamics simulations. The latter approach was found to be critical for finding the key Cu intermediates. Our results show that the highly flexible coordination environment of copper strongly influences the nature of the oxidized Cu(II) species. Furthermore, they suggest the favorable formation of trans-Cu(II)–(OH)2 moieties in low-coordinated Cu(II) species. This trans configuration hinders the formation of Cu(II)–oxyl species, facilitating intramolecular H–abstraction reactions in line with experimentally observed ligand oxidation processes.

• Hong, Wanwan; Luthra, Mahika; Jakobsen, Joakim B.; Madsen, Monica R.; Castro, Abril C & Hammershøj, Hans Christian D. [Show all 11 contributors for this article] (2023). Exploring the Parameters Controlling Product Selectivity in Electrochemical CO<inf>2</inf> Reduction in Competition with Hydrogen Evolution Employing Manganese Bipyridine Complexes. ACS Catalysis. p. 3109–3119. doi: 10.1021/acscatal.2c05951. Full text in Research Archive Show summary

  Selective reduction of CO2 is an efficient solution for producing nonfossil-based chemical feedstocks and simultaneously alleviating the increasing atmospheric concentration of this greenhouse gas. With this aim, molecular electrocatalysts are being extensively studied, although selectivity remains an issue. In this work, a combined experimental–computational study explores how the molecular structure of Mn-based complexes determines the dominant product in the reduction of CO2 to HCOOH, CO, and H2. In contrast to previous Mn(bpy-R)(CO)3Br catalysts containing alkyl amines in the vicinity of the Br ligand, here, we report that bpy-based macrocycles locking these amines at the side opposite to the Br ligand change the product selectivity from HCOOH to H2. Ab initio molecular dynamics simulations of the active species showed that free rotation of the Mn(CO)3 moiety allows for the approach of the protonated amine to the reactive center yielding a Mn-hydride intermediate, which is the key in the formation of H2 and HCOOH. Additional studies with DFT methods showed that the macrocyclic moiety hinders the insertion of CO2 to the metal hydride favoring the formation of H2 over HCOOH. Further, our results suggest that the minor CO product observed experimentally is formed when CO2 adds to Mn on the side opposite to the amine ligand before protonation. These results show how product selectivity can be modulated by ligand design in Mn-based catalysts, providing atomistic details that can be leveraged in the development of a fully selective system.

• Kneiding, Hannes & Badia, David Balcells (2023). Deep Learning Metal Complex Properties with Natural Quantum Graphs. Digital Discovery. Full text in Research Archive
• Kneiding, Hannes & Badia, David Balcells (2023). Deep Graph Learning for Molecules and Materials. Nordic Machine Intelligence (NMI). 3(2). doi: 10.5617/nmi.10517. Full text in Research Archive Show summary

  Machine learning approaches have become an important tool in chemistry and materials science for the accurate and efficient prediction of physical properties. Most notably among them are graph neural networks that leverage the inherent graph structure of molecules and materials in order to achieve state-of-the-art accuracy. In this perspective we give a brief introduction to the theoretical foundations of graph neural networks for molecular structures and their specific applications in chemistry and materials science. We conclude with a short outlook discussing remaining research questions as well as opportunities for further developments of the field.

• Kneiding, Hannes; Flores, Ainara Nova & Badia, David Balcells (2023). Directional Multiobjective Optimization of Metal Complexes at the Billion-Scale with the tmQMg-L Dataset and PL-MOGA Algorithm. ChemRxiv. ISSN 2573-2293. doi: 10.26434/chemrxiv-2023-k3tf2-v2. Full text in Research Archive Show summary

  Transition metal complexes (TMCs) play a key role in several areas of high interest, including medicinal chemistry, renewable energies, and nanoporous materials. The development of TMCs enabling these technologies remains challenged by the need to optimize multiple properties within very large chemical spaces, in which the thirty transition metals can be combined with a virtually infinite number of ligands. In this work, we provide the open tmQMg-L dataset including 30K TMC ligands, which combines large chemical diversity with synthesizability. The charge and metal-coordination mode of the ligands were robustly defined with a novel algorithm based on graph and natural bond orbital theories. The tmQMg-L dataset was leveraged in the automated generation of 1.37M TMCs resulting from all possible combinations between a square planar palladium(II) scaffold and a pool of 50 different ligands. This TMC space was used to benchmark a multiobjective genetic algorithm (MOGA) that optimized two properties over a Pareto front; namely the polarizability (alpha) and the HOMO-LUMO gap (epsilon). The MOGA evolved 130 TMC hits with maximal (alpha, epsilon) values in a way that could be easily rationalized by analyzing the nature of the ligands selected. Instead of the traditional mutation and crossover of fragments within a single ligand, this MOGA implemented full-ligand genetic operations acting on all coordination sites, maximizing chemical diversity. Further, we extended this MOGA algorithm with the Pareto-Lighthouse functionality (PL-MOGA), which allows for controlling both the aim and scope of the multiobjective optimization over the Pareto front. In explicit spaces containing billions of TMCs, the PL-MOGA enabled the explainable generation of thousands of novel and highly diverse TMC hits. We believe that the combined use of the tmQMg-L dataset and PL-MOGA algorithm will facilitate the discovery of TMCs with optimal properties within untapped chemical spaces.

• Skjelstad, Bastian; Helgaker, Trygve; Maeda, Satoshi & Badia, David Balcells (2022). Oxyl Character and Methane Hydroxylation Mechanism in Heterometallic M(O)Co3O4Cubanes (M = Cr, Mn, Fe, Mo, Tc, Ru, and Rh). ACS Catalysis. 12(19), p. 12326–12335. doi: 10.1021/acscatal.2c03748. Full text in Research Archive Show summary

  C–H activation in alkanes poses a major challenge in chemistry due to the inert character of this bond, manifesting the necessity of improved catalysts. Although various metal–oxo complexes are known to facilitate alkane hydroxylation, probing the mechanistic nature of the reaction is difficult due to the extremely fast rebound of the radical intermediate in the postulated oxygen-rebound pathway. Automated reaction mechanism discovery methods, such as the artificial force induced reaction (AFIR) method, enable the efficient exploration of both expected and unexpected reaction pathways, revealing the reaction mechanism. Here, we employed this approach combined with density-functional theory (DFT) to investigate the structure and reactivity of heterometallic cubane complexes similar to the oxygen-evolving complex of photosystem II. For a series of M(O)Co3O4 cubanes, where M(O) is a terminal oxo with M = Cr, Mn, Fe, Mo, Tc, Ru, and Rh, we computed the stability of the possible spin states and the radical (i.e., oxyl) character of the M(O) moiety as a measure of their potential activity in the catalytic hydroxylation of alkanes. DFT calculations on these reactions promoted by Ru(O)Co3O4 and Fe(O)Co3O4 suggest that the latter promotes the hydroxylation of methane with a rate-determining H-abstraction barrier of 24.6 kcal/mol. The moderate height of this barrier, together with the low cost and low toxicity of iron and cobalt, suggests that the Fe(O)Co3O4 cubane is a promising candidate for the catalytic oxidation of methane to methanol. AFIR calculations showed that the oxygen-rebound pathway yields the lowest-energy profile, thus validating this mechanism for the hydroxylation of alkanes by heterometallic cubanes. Furthermore, unexpected intermediates in which the methyl radical couples with either the metal center or the bridging oxo ligands were also observed.

• Kneiding, Hannes; Lukin, Ruslan & Badia, David Balcells (2022). Deep Learning Metal Complex Properties with Natural Quantum Graphs. ChemRxiv. ISSN 2573-2293. doi: 10.26434/chemrxiv-2022-fd43k. Full text in Research Archive
• Héron, Julie & Badia, David Balcells (2022). Concerted Cycloaddition Mechanism in the CuAAC Reaction Catalyzed by 1,8-Naphthyridine Dicopper Complexes. ACS Catalysis. 12(8), p. 4744–4753. doi: 10.1021/acscatal.2c00723. Full text in Research Archive Show summary

  Copper-catalyzed azide-alkyne cycloaddition (CuAAC) is one of the most versatile reactions in the “click chemistry” toolbox, and its development has made the synthesis of 1,4-triazole derivatives robust and efficient. In this work, we present a density functional theory (DFT) study on the mechanism of the CuAAC reaction catalyzed by a dicopper complex supported by a nonsymmetric 1,8-naphtyridine ligand bearing two different metal-coordinating substituents (i.e., −P(tBu)2 and −C(Me)(Py)2). The calculations showed that the cycloaddition of the azide to the alkyne occurs in a single concerted step, in contrast with the two-step mechanism proposed in the literature. The energies predicted for this step indicated that the 1,4-triazole isomer of the product is formed in a selective manner, in agreement with experiments. Further, the DFT results showed that there is a subtle and complex interplay between several variables, including the relative orientation of the two substrates, the position of the counter-anion, and the partial decoordination of the 1,8-naphtyridine ligand. A series of 90 transition state calculations showed that, on average, the impact of these variables is strong on the structures but soft on the energy barriers, highlighting the flexible nature of the bonding within the coordination sphere of the bimetallic core of the catalyst. The insight provided by this study will be valuable for the further development of dicopper catalysts for the CuAAC reaction.

• Somerville, Rosie J.; Borys, Andryj M.; Perez-Jimenez, Marina; Nova, Ainara; Badia, David Balcells & Malaspina, Lorraine A. [Show all 10 contributors for this article] (2022). Unmasking the constitution and bonding of the proposed lithium nickelate “Li3NiPh3(solv)3”: revealing the hidden C6H4 ligand†. Chemical Science. ISSN 2041-6520. 13, p. 5268–5268. doi: 10.1039/d2sc01244h. Full text in Research Archive Show summary

  More than four decades ago, a complex identified as the planar homoleptic lithium nickelate “Li3NiPh3(solv)3” was reported by Taube and co-workers. This and subsequent reports involving this complex have lain dormant since; however, the absence of an X-ray diffraction structure leaves questions as to the nature of the Ni–PhLi bonding and the coordination geometry at Ni. By systematically evaluating the reactivity of Ni(COD)2 with PhLi under different conditions, we have found that this classical molecule is instead a unique octanuclear complex, [{Li3(solv)2Ph3Ni}2(μ-η2:η2-C6H4)] (5). This is supported by X-ray crystallography and solution-state NMR studies. A theoretical bonding analysis from NBO, QTAIM, and ELI perspectives reveals extreme back-bonding to the bridging C6H4 ligand resulting in dimetallabicyclobutane character, the lack of a Ni–Ni bond, and pronounced σ-bonding between the phenyl carbanions and nickel, including a weak σC–Li → sNi interaction with the C–Li bond acting as a σ-donor. Employing PhNa led to the isolation of [Na2(solv)3Ph2NiCOD]2 (7) and [Na2(solv)3Ph2(NaC8H11)Ni(COD)]2 (8), which lack the benzyne-derived ligand. These findings provide new insights into the synthesis, structure, bonding and reactivity of heterobimetallic nickelates, whose prevalence in organonickel chemistry and catalysis is likely greater than previously believed.

• Chen, Shusen; Nielson, Taylor; Zalit, Elayna; Skjelstad, Bastian Bjerkem; Borough, Braden & Hirschi, William J. [Show all 9 contributors for this article] (2021). Automated Construction and Optimization Combined with Machine Learning to Generate Pt(II) Methane C–H Activation Transition States. Topics in catalysis. ISSN 1022-5528. doi: 10.1007/s11244-021-01506-0. Full text in Research Archive
• Dardir, Amira; Casademont-Reig, Irene; Badia, David Balcells; Ellefsen, Jonathan; Espinosa, Matthew & Hazari, Nilay [Show all 7 contributors for this article] (2021). Synthesis of Triarylmethanes via Palladium-Catalyzed Suzuki-Miyaura Reactions of Diarylmethyl Esters. Organometallics. ISSN 0276-7333. 40(14), p. 2332–2344. doi: 10.1021/acs.organomet.1c00085. Full text in Research Archive
• Nicolay, Amelie; Héron, Julie; Shin, Chungkeun; Kuramarohit, Serene; Ziegler, Micah & Badia, David Balcells [Show all 7 contributors for this article] (2021). Unsymmetrical Naphthyridine-Based Dicopper(I) Complexes: Synthesis, Stability and Carbon-Hydrogen Bond Activations. Organometallics. ISSN 0276-7333. 40(12), p. 1866–1873. doi: 10.1021/acs.organomet.1c00188. Full text in Research Archive
• Amtawong, Jaruwan; Skjelstad, Bastian; Handford, Rex; Suslick, Benjamin; Badia, David Balcells & Tilley, T. Don (2021). C−H Activation by RuCo3O4 Oxo Cubanes: Effects of Oxyl Radical Character and Metal-Metal Cooperativity. Journal of the American Chemical Society. ISSN 0002-7863. 143(31), p. 12108–12119. doi: 10.1021/jacs.1c04069. Full text in Research Archive
• Suàrez, Lluís Artús; Badia, David Balcells & Nova, Ainara (2021). Computational Studies on the Mechanisms for Deaminative Amide Hydrogenation by Homogeneous Bifunctional Catalysts. Topics in catalysis. ISSN 1022-5528. 65, p. 82–95. doi: 10.1007/s11244-021-01542-w. Full text in Research Archive Show summary

  The deaminative hydrogenation of amides is one of the most convenient pathways for the synthesis of amines and alcohols. The ideal source of reducing equivalents for this reaction is molecular hydrogen, though, in practice, this approach requires high pressures and temperatures, with many catalysts achieving only small turnover numbers and frequencies. Nonetheless, during the last ten years, this feld has made major advances towards larger turnovers under milder conditions thanks to the development of bifunctional catalysts. These systems promote the heterolytic cleavage of hydrogen into proton and hydride by combining a basic ligand with an acidic metal centre. The present review focuses on the computational study of the reaction mechanism underlying bifunctional catalysis. This review is structured around the fundamental steps of this mechanism, namely the C=O and C–N hydrogenation of the amide, the C–N protonolysis of the hemiaminal, the C=O hydrogenation of the aldehyde, and the competition between hydrogen activation and catalyst deactivation. In line with the complexity of the mechanism, we also provide a perspective on the use of microkinetic models. Both Noyori- and Milstein-type catalysts are discussed and compared.

• Levchenko, Vladimir; Nova, Ainara; Øien-Ødegaard, Sigurd; Balcells, David & Tilset, Mats (2020). Synthesis, Characterization, and Reactivity of Cyclometalated Gold(III) Dihalide Complexes in Aqua Regia. European Journal of Inorganic Chemistry (EurJIC). ISSN 1434-1948. 2020(34), p. 3249–3258. doi: 10.1002/ejic.202000529. Full text in Research Archive Show summary

  A range of N,C‐chelated, cyclometalated gold(III) complexes Au(ppyR)X2 have been prepared and characterized by spectroscopic, crystallographic, and computational means. Here, ppyH is 2‐phenylpyridine dicarboxylic acid (series 1), ppyEt is diethyl 2‐phenylpyridine dicarboxylate (series 2), and X is trifluoroacetate OAcF (a), Cl (b), Br (c), or I (d) anion. The dihalo complexes 1b‐d and 2b‐d are obtained when Au(ppyR)(OAcF)2 (1a and 2a) are treated with HNO3/HX mixtures (aqua regiaX). Good to high yields are obtained with short reaction times (< 30 min) and simple work‐up. Notably, the strongly acidic medium does not cause protolytic cleavage of the Au–C or Au–N bonds in the chelate, nor is ester hydrolysis of complexes 2b‐d seen. Ethylene inserts into an Au–O bond of 1a and 1b, and the resulting trifluoroacetoxyethyl–Au complexes can be further elaborated in aqua regia without cleavage of the two Au–C bonds in the molecule. Facile, mutual halide exchange reactions between complexes with different halides (1b and 1d, 1c and 1d) were observed and led to formation of mixed‐halide complexes Au(ppyH)(X)(Y). These exchange reactions occurred with complete stereoselectivity. The stereoisomer produced was the one expected based on the relative trans influence of the halides (I > Br > Cl), i.e. the highest trans influence halide was located trans to N which is the lowest trans influence end of the chelate. These thermodynamic preferences were also investigated by DFT computations.

• Suàrez, Lluís Artús; Jayarathne, Upul; Balcells, David; Bernskoetter, Wesley H.; Hazari, Nilay & Jaraiz, Martin [Show all 7 contributors for this article] (2020). Rational selection of co-catalysts for the deaminative hydrogenation of amides. Chemical Science. ISSN 2041-6520. 11(8), p. 2225–2230. doi: 10.1039/c9sc03812d. Full text in Research Archive Show summary

  The catalytic hydrogenation of amides is an atom economical method to synthesize amines. Previously, it was serendipitously discovered that the combination of a secondary amide co-catalyst with (^iPrPNP)Fe(H)(CO) (^iPrPNP = N[CH₂CH₂(PⁱPr₂)]₂−), results in a highly active base metal system for deaminative amide hydrogenation. Here, we use DFT to develop an improved co-catalyst for amide hydrogenation. Initially, we computationally evaluated the ability of a series of co-catalysts to accelerate the turnover-limiting proton transfer during C–N bond cleavage and poison the (^iPrPNP)Fe(H)(CO) catalyst through a side reaction. TBD (triazabicyclodecene) was identified as the leading co-catalyst. It was experimentally confirmed that when TBD is combined with (^iPrPNP)Fe(H)(CO) a remarkably active system for amide hydrogenation is generated. TBD also enhances the activity of other catalysts for amide hydrogenation and our results provide guidelines for the rational design of future co-catalysts.

• Castro, Abril C; Balcells, David; Repisky, Michal; Helgaker, Trygve & Cascella, Michele (2020). First-principles calculation of 1H NMR chemical shifts of complex metal polyhydrides: The essential inclusion of relativity and dynamics. Inorganic Chemistry. ISSN 0020-1669. 59(23), p. 17509–17518. doi: 10.1021/acs.inorgchem.0c02753. Full text in Research Archive Show summary

  ¹H NMR spectroscopy has become an important technique for the characterization of transition-metal hydride complexes, whose metal-bound hydrides are often difficult to locate by X-ray diffraction. In this regard, the accurate prediction of ¹H NMR chemical shifts provides a useful, but challenging, strategy to help in the interpretation of the experimental spectra. In this work, we establish a density-functional-theory protocol that includes relativistic, solvent, and dynamic effects at a high level of theory, allowing us to report an accurate and reliable interpretation of ¹H NMR hydride chemical shifts of iridium polyhydride complexes. In particular, we have studied in detail the hydride chemical shifts of the [Ir₆(IMe)₈(CO)₂H₁₄]²⁺ complex in order to validate previous assignments. The computed ¹H NMR chemical shifts are strongly dependent on the relativistic treatment, the choice of the DFT exchange–correlation functional, and the conformational dynamics. By combining a fully relativistic four-component electronic-structure treatment with ab initio molecular dynamics, we were able to reliably model both the terminal and bridging hydride chemical shifts and to show that two NMR hydride signals were inversely assigned in the experiment.

• Amtawong, Jaruwan; Skjelstad, Bastian; Balcells, David & Tilley, T. Don (2020). Concerted Proton-Electron Transfer Reactivity at a Multimetallic Co4O4 Cubane Cluster. Inorganic Chemistry. ISSN 0020-1669. 59(20), p. 15553–15560. doi: 10.1021/acs.inorgchem.0c02625. Full text in Research Archive
• Balcells, David & Skjelstad, Bastian (2020). tmQM Dataset—Quantum Geometries and Properties of 86k Transition Metal Complexes. Journal of Chemical Information and Modeling. ISSN 1549-9596. 60(12), p. 6135–6146. doi: 10.1021/acs.jcim.0c01041. Full text in Research Archive Show summary

  We report the transition metal quantum mechanics (tmQM) data set, which contains the geometries and properties of a large transition metal–organic compound space. tmQM comprises 86,665 mononuclear complexes extracted from the Cambridge Structural Database, including Werner, bioinorganic, and organometallic complexes based on a large variety of organic ligands and 30 transition metals (the 3d, 4d, and 5d from groups 3 to 12). All complexes are closed-shell, with a formal charge in the range {+1, 0, −1}e. The tmQM data set provides the Cartesian coordinates of all metal complexes optimized at the GFN2-xTB level, and their molecular size, stoichiometry, and metal node degree. The quantum properties were computed at the DFT(TPSSh-D3BJ/def2-SVP) level and include the electronic and dispersion energies, highest occupied molecular orbital (HOMO) and lowest unoccupied molecular orbital (LUMO) energies, HOMO/LUMO gap, dipole moment, and natural charge of the metal center; GFN2-xTB polarizabilities are also provided. Pairwise representations showed the low correlation between these properties, providing nearly continuous maps with unusual regions of the chemical space, for example, complexes combining large polarizabilities with wide HOMO/LUMO gaps and complexes combining low-energy HOMO orbitals with electron-rich metal centers. The tmQM data set can be exploited in the data-driven discovery of new metal complexes, including predictive models based on machine learning. These models may have a strong impact on the fields in which transition metal chemistry plays a key role, for example, catalysis, organic synthesis, and materials science. tmQM is an open data set that can be downloaded free of charge from https://github.com/bbskjelstad/tmqm.

• Friederich, Pascal; Gomes, Gabriel Dos Passos; Bin, Riccardo De; Aspuru-Guzik, Alán & Balcells, David (2020). Machine learning dihydrogen activation in the chemical space surrounding Vaska's complex. Chemical Science. ISSN 2041-6520. 11(18), p. 4584–4601. doi: 10.1039/d0sc00445f. Full text in Research Archive Show summary

  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.

• Wåhlander, Jakob Karl Kristoffer; Amedjkouh, Mohamed & Balcells, David (2019). A DFT Perspective on Diels-Alder Organocatalysts Based on Substituted Phosphoramides. European Journal of Organic Chemistry. ISSN 1434-193X. 2019(2-3), p. 442–450. doi: 10.1002/ejoc.201800844. Full text in Research Archive Show summary

  Phosphoramides are an interesting alternative to the thiourea motif that has long been used in organocatalysis. In the R1N(H)-P(O)(R3)-N(H)R2 formulation, phosphoramides give the advantage of having a third variable substituent, R3, compared to the two available in thioureas, and the possibility of introducing chirality with an asymmetric phosphorous center. In this work, we present a systematic computational study on the catalytic properties of these underexplored compounds. These studies show the catalytic potential of a series of substituted phosphoramides in Diels-Alder reactions, including the asymmetric [4+2] cycloaddition of pro-chiral substrates. The nature of the phosphorous-bound R3 substituent has a significant impact both on catalytic activity and enantioselectivity. Remarkably, several of the phosphoramides studied are predicted to be more active than the Schreiner’s thiourea organocatalyst.

• Barth, Emily; Davis, Ryan; Beromi, Megan Mohadjer; Walden, Andrew; Balcells, David & Brudvig, Gary W. [Show all 11 contributors for this article] (2019). Bis(dialkylphosphino)ferrocene-Ligated Nickel(II) Precatalysts for Suzuki-Miyaura Reactions of Aryl Carbonates. Organometallics. ISSN 0276-7333. 38(17), p. 3377–3387. doi: 10.1021/acs.organomet.9b00543. Full text in Research Archive
• Amtawong, Jaruwan; Balcells, David; Wilcoxen, Jarett; Handford, Rex; Biggins, Naomi & Nguyen, Andy I [Show all 8 contributors for this article] (2019). Isolation and Study of Ruthenium–Cobalt Oxo Cubanes Bearing a High-Valent, Terminal RuV–Oxo with Significant Oxyl Radical Character. Journal of the American Chemical Society. ISSN 0002-7863. 141(50), p. 19859–19869. doi: 10.1021/jacs.9b10320. Full text in Research Archive
• Suarez, Lluis Artus; Culakova, Zuzana; Balcells, David; Bernskoetter, Wesley H; Eisenstein, Odile & Goldberg, Karen I. [Show all 9 contributors for this article] (2018). The Key Role of the Hemiaminal Intermediate in the Iron-Catalyzed Deaminative Hydrogenation of Amides. ACS Catalysis. 8(9), p. 8751–8762. doi: 10.1021/acscatal.8b02184. Full text in Research Archive
• Grajeda, Javier; Nova, Ainara; Balcells, David; Bruch, Quinton J.; Wragg, David & Heyn, Richard H. [Show all 8 contributors for this article] (2018). Synthesis and characterization of stable gold(III) PNP pincer complexes. European Journal of Inorganic Chemistry (EurJIC). ISSN 1434-1948. 2018(26), p. 3113–3117. doi: 10.1002/ejic.201800019. Full text in Research Archive
• Thoresen, Eirik Mydske; Balcells, David; Øien-Ødegaard, Sigurd; Hylland, Knut Tormodssønn; Tilset, Mats & Amedjkouh, Mohamed (2018). Cyclometalated Ruthenium Complexes with Carboxylated Ligands from a Combined Experimental/Computational Perspective. Dalton Transactions. ISSN 1477-9226. 47(8), p. 2589–2601. doi: 10.1039/c7dt03935b. Full text in Research Archive
• Shopov, Dimitar Y; Sharninghausen, Liam S.; Sinha, Shashi Bhushan; Mercado, Brandon Q.; Balcells, David & Brudvig, Gary W. [Show all 7 contributors for this article] (2018). A Dinuclear Iridium(V,V) Oxo-Bridged Complex Characterized Using a Bulk Electrolysis Technique for Crystallizing Highly Oxidizing Compounds. Inorganic Chemistry. ISSN 0020-1669. 57(9), p. 5684–5691. doi: 10.1021/acs.inorgchem.8b00757. Full text in Research Archive
• Balcells, David & Nova, Ainara (2018). Designing Pd and Ni Catalysts for Cross-Coupling Reactions by Minimizing Off-Cycle Species. ACS Catalysis. 8(4), p. 3499–3515. doi: 10.1021/acscatal.8b00230. Full text in Research Archive Show summary

  This Perspective presents an overview on recent experimental and computational studies on the off-cycle reactions of palladium- and nickel-catalyzed cross-couplings. Several reactions entering or leaving the catalytic cycle have been characterized, including the activation of Pd(II) precatalysts by H-shift and the deactivation of Ni(II) precatalysts by comproportionation. A fundamental difference between the off-cycle chemistries of palladium and nickel is the larger diversity of species yielded by the latter, with a rich combination of different oxidation states, nuclearity, and ligand coordination modes. The molecular-level understanding of off-cycle reactions has been exploited in catalyst design, including the stereoelectronic fine-tuning of the ligands aimed at maximizing the activation of the precatalyst meanwhile preventing its deactivation. Despite several challenges, which concern both experiments (e.g., isolation and characterization of transient species) and computations (e.g., comprehensive mapping of the potential energy surface), this approach has already been applied with success in the optimization of popular catalytic platforms (e.g., NHC–Pd–allyl precatalysts) and shows promise for the development of highly active and robust catalysts based on nickel.

View all works in NVA

• Badia, David Balcells (2026). Generative AI for transition metal chemistry ( making the novel, exceptional ? ). Full text in Research Archive
• Safy, Mohamed E. A.; Flores, Ainara Nova & Badia, David Balcells (2026). How Mechanistic Understanding Guides the Design of New CO2-to-Methanol Catalysts. Full text in Research Archive
• González, Lucía Morán; Kneiding, Hannes; Badia, David Balcells & Flores, Ainara Nova (2026). High-Throughput Computational Screening of Bifunctional Catalysts for CO₂ Hydrogenation. doi: https:/ccsc2026.github.io/. Full text in Research Archive
• Badia, David Balcells (2025). Trying to Make the Novel Exceptional with Generative AI for Transition Metal Chemistry. Full text in Research Archive
• Badia, David Balcells (2025). Evolutionary Machine Learning for Transition Metal Chemistry. Full text in Research Archive
• Badia, David Balcells (2025). Evolutionary Machine Learning for Transition Metal Chemistry. Full text in Research Archive
• Badia, David Balcells (2025). Evolutionary and Deep Machine Learning for Transition Metal Chemistry. Full text in Research Archive
• Badia, David Balcells (2025). Generative AI for the Chemistry of the Transition Metals. Full text in Research Archive
• Kneiding, Hannes & Badia, David Balcells (2025). tmQMg* dataset: excited state properties of 74k transition metal complexes. Full text in Research Archive
• Kneiding, Hannes & Badia, David Balcells (2025). Graphs for machine learning applications in chemistry. Full text in Research Archive
• Kneiding, Hannes; Badia, David Balcells & Flores, Ainara Nova (2024). Directional Multiobjective Optimization of Metal Complexes In Vast Chemical Spaces. Full text in Research Archive Show summary

  Transition metal complexes (TMCs) are crucial for a multitude of applications including nanoporous materials, renewable energies, and medicinal chemistry. Their development time can be drastically reduced by identifying promising candidates in silico before validating hits with experiments. However, sufficiently large and diverse ligand libraries that enable the efficient computational optimization of TMCs in vast chemical spaces are still lacking. Here, we present our recent work on the open tmQMg-L [1] dataset including 30K experimentally relevant TMC ligands extracted from the Cambridge Structural Database. All ligands come with robustly defined charge and metal-coordination mode. We demonstrate how the tmQMg-L dataset can be leveraged in the multiobjective optimization of TMCs with respect to different properties of interest. In particular, we employ a genetic algorithm with full-ligand genetic operations and multiobjective selection rules to simultaneously maximize the HOMO-LUMO gap and polarizability of TMCs in chemical spaces based on a palladium(II) scaffold with up to 1B unique members. The final hits occupy, or are very close, to the Pareto front and their aim and scope can be controlled using our newly introduced Pareto-Lighthouse functionality. Furthermore, we discuss how the ligands in tmQMg-L can be used to train a variational autoencoder in order to generate novel ligands ready for their use in evolutionary optimization approaches. [1] Kneiding, H., Nova, A., Balcells, D. Nat. Comput. Sci. 2024, accepted

• Kneiding, Hannes & Badia, David Balcells (2024). Quantum-informed graph representations for the prediction of molecular properties of transition metal complexes. Full text in Research Archive Show summary

  Transition metal complexes (TMCs) are essential for a wide range of applications including renewable energies, medicinal chemistry, and nanoporous materials. Their investigation with high-throughput screenings for properties of interest using computational methods is still associated with prohibitively large computational costs rendering them impractical. Predictive machine learning (ML) can close this gap and enable the exploration of vast chemical spaces including millions of compounds. However, due to the complex chemistry present in TMCs, the construction of sufficiently accurate ML models is still impeded and an area of current research. Here, we present work in which we devised highly expressive graph representations of TMCs based on quantum chemistry information and used them in conjunction with graph neural networks in order to facilitate the prediction of quantum properties of interest. [1] In particular, these graph representations are based on Natural Bond Orbital [2] analysis data that provides a rich feature set to be leveraged in ML approaches leading. We show how they outperform methods from the literature based solely on structural features and furthermore demonstrate their application in Δ-learning approaches, in which properties are first calculated at a low level of theory to then predict the corresponding corrections to the target values at a high level of theory. Quantum information extracted from the necessary low level calculations will be used to featurize the TMC graph representations. [1] KNEIDING, Hannes, et al. Deep learning metal complex properties with natural quantum graphs. Digital Discovery, 2023, 2. Jg., Nr. 3, S. 618-633. [2] GLENDENING, Eric D.; LANDIS, Clark R.; WEINHOLD, Frank. NBO 7.0: New vistas in localized and delocalized chemical bonding theory. Journal of computational chemistry, 2019, 40.25: 2234-2241.

• Kneiding, Hannes & Badia, David Balcells (2024). Δ-Learning Graph Neural Networks for the Accurate Prediction of Quantum Properties of Transition Metal Complexes. Full text in Research Archive Show summary

  Transition metal complexes (TMCs) are crucial for a multitude of applications including medicinal chemistry, renewable energies, and nanoporous materials. Their investigation with high-throughput screenings for specific properties using computational methods is still associated with prohibitively large computational costs rendering them impractical. Predictive machine learning can close this gap and enable the exploration of vast chemical spaces including millions of compounds. However, due to the complex chemistry present in TMCs, the construction of sufficiently accurate machine learning models is still impeded and an area of current research. Here, we build on our previous work in which we devised highly expressive graph representations of TMCs based on quantum chemistry information and used them in conjunction with graph neural networks in order to facilitate the prediction of quantum properties of interest. [1] Instead of directly predicting these properties, we plan to first calculate them at a lower level of theory and then predict the corresponding corrections to the target values. Quantum information extracted from the necessary low level calculations will be used to featurize the TMC graph representations. Using this approach we expect a significant increase in predictive performance at comparably low computational cost. [1] KNEIDING, Hannes, et al. Deep learning metal complex properties with natural quantum graphs. Digital Discovery, 2023, 2. Jg., Nr. 3, S. 618-633.

• Kneiding, Hannes; Badia, David Balcells & Flores, Ainara Nova (2024). Directional Evolutionary Multiobjective Optimization of Metal Complexes at the Billion-Scale. Full text in Research Archive Show summary

  Transition metal complexes (TMCs) are crucial for a multitude applications in, including medicinal chemistry, renewable energies, and nanoporous materials. Computational methods can drastically reduce development time by identifying promising candidates in silico before validating hits with more expensive experiments. The efficient optimization of TMCs for specific properties however, is still hindered by a lack of sufficiently large and diverse ligand libraries used to span vast chemical spaces. In this work, we present the open tmQMg-L [1] dataset including 30K experimentally relevant TMC ligands extracted from the Cambridge Structural Database. All ligands come with robustly defined charge and metal-coordination mode. We show how the tmQMg-L dataset can be leveraged in the multiobjective optimization of TMCs with respect to different properties of interest. Setting up chemical spaces based on the palladium(II) scaffold with up to 1B unique members, we evolve ligands by simultaneously maximizing the HOMO-LUMO gap and polarizability using a genetic algorithm with full-ligand genetic operations and multiobjective selection rules. The final hits occupy or are very close to the Pareto front. We furthermore introduce the Pareto-Lighthouse functionality, which can be used to control the aim and scope of the optimization over the Pareto front. [1] Kneiding, H., Nova, A., & Balcells, D. Directional Multiobjective Optimization of Metal Complexes at the Billion-Scale with the tmQMg-L Dataset and PL-MOGA Algorithm, ChemRxiv 2023. DOI: 10.26434/chemrxiv-2023-k3tf2-v2

• Badia, David Balcells (2024). Machine Learning in the Metal-Organic Chemical Space. Full text in Research Archive
• Badia, David Balcells (2024). Generative Machine Learning in the Transition Metal Compound Space. Full text in Research Archive
• Badia, David Balcells (2024). Generative Machine Learning for Transition Metal Complexes. Full text in Research Archive
• Badia, David Balcells (2024). Machine Learning for Transition Metal Complexes: Deterministic, Evolutionary, and Generative. Full text in Research Archive
• Badia, David Balcells (2024). Generative Machine Learning for Transition Metal Chemistry. Full text in Research Archive
• Badia, David Balcells (2024). Machine Learning for Transition Metal Chemistry: Predictive, Evolutionary, and Generative. Full text in Research Archive
• Safy, Mohamed Eid Ahmed; Rama, Raquel Jimenez; Badia, David Balcells & Flores, Ainara Nova (2024). CO2 hydrogenation to methanol beyond Ru-MACHO: A DFT and microkinetic modeling study. Full text in Research Archive
• Kneiding, Hannes; Flores, Ainara Nova & Badia, David Balcells (2024). Directional Multiobjective Evolution of Metal Complexes at the Billion-Scale. Full text in Research Archive Show summary

  Transition metal complexes (TMCs) are important for applications in many different fields, ranging from cancer treatment to catalysis in industrial processes. The search for TMCs that exhibit optimal properties for their use in specific applications is inherently difficult due to the large size and high diversity of the chemical space to be explored. We show how methods from evolutionary learning can be employed in a database-driven approach to perform efficient, multiobjective optimization of TMCs with respect to relevant molecular properties. In particular, we illustrate how the optimization can find TMCs that occupy or are very close to the Pareto front in chemical spaces of up to 1 billion members. Furthermore, we propose an algorithm to control the aim and scope of the evolution towards specific regions of the search space, and discuss possibilities of augmenting the presented approach with neural networks.

• Denjean, Aurore Evelyne Francine; Flores, Ainara Nova & Badia, David Balcells (2024). Amine Alkylation with Single Atom Nickel Catalysts: Computational Insights into the Hydrogen Borrowing Mechanism. Full text in Research Archive
• Denjean, Aurore Evelyne Francine; Flores, Ainara Nova & Badia, David Balcells (2024). Graphene-based Nickel(II) Single-Atom Catalyst for Borrowing Hydrogen: a DFT study. Full text in Research Archive
• Denjean, Aurore Evelyne Francine; Flores, Ainara Nova & Badia, David Balcells (2024). Computational Insights into Graphene-Based Single Atom Nickel Catalysts for Amine Alkylation via the Borrowing Hydrogen Mechanism. Full text in Research Archive
• Badia, David Balcells (2024). Data-Driven Transition Metal Chemistry with Evolutionary & Deep Learning Approaches. Full text in Research Archive
• Badia, David Balcells (2024). Generative Artificial Intelligence for Transition Metal Chemistry. Full text in Research Archive
• Kneiding, Hannes & Badia, David Balcells (2024). Graphs for machine learning applications in chemistry. Full text in Research Archive
• Badia, David Balcells (2023). Machine Learning the Structure and Reactivity of Transition Metal complexes. Full text in Research Archive
• Badia, David Balcells (2023). Deep Graph Learning Approaches to Chemistry. Full text in Research Archive
• Badia, David Balcells (2023). Machine Learning for Metal-Organic Chemistry. Full text in Research Archive
• Badia, David Balcells (2023). Machine Learning the Chemistry of Transition Metals. Full text in Research Archive
• Badia, David Balcells (2023). Exploring the Metal-Organic Chemical Space with Evolutionary and Machine Learning Approaches. Full text in Research Archive
• Kneiding, Hannes; Flores, Ainara Nova & Badia, David Balcells (2023). Evolutionary Multiobjective Optimization for Transition Metal Complexes. Full text in Research Archive
• Kneiding, Hannes; Flores, Ainara Nova & Badia, David Balcells (2023). Evolutionary Multiobjective Optimization for Transition Metal Complexes. Full text in Research Archive
• Kneiding, Hannes & Badia, David Balcells (2023). Machine Learning Quantum Properties of Transition Metal Complexes with Natural Quantum Graphs. Full text in Research Archive
• Badia, David Balcells (2022). Machine Learning Organometallic Chemistry. Full text in Research Archive
• Badia, David Balcells (2022). Machine Learning Approaches to Transition Metal Chemistry. Full text in Research Archive
• Kneiding, Hannes & Badia, David Balcells (2022). Machine Learning Quantum Properties of Transition Metal Complexes using Graph Neural Networks. Full text in Research Archive
• Kneiding, Hannes & Badia, David Balcells (2022). Machine Learning Quantum Properties of Transition Metal Complexes from Graph Representations. Full text in Research Archive
• Kneiding, Hannes & Badia, David Balcells (2022). Graph Representation Learning for Transition Metal Complexes. Full text in Research Archive
• Balcells, David (2020). Machine Learning Reactivity at the Fundamental Steps of Transition Metal Catalysts. Full text in Research Archive
• Balcells, David (2020). Machine Learning Reactivity at the Fundamental Steps of Transition Metal Catalysis. Full text in Research Archive
• Héron, Julie Denise Josette & Balcells, David (2019). Catalyst Activation Mechanisms in the Cu2AAC reaction. Full text in Research Archive
• Héron, Julie Denise Josette & Balcells, David (2019). Computational Study of the Catalyst Activation of the Cu2AAC reaction. Full text in Research Archive
• Balcells, David; Clot, Eric; Macgregor, Stuart A.; Maseras, Feliu & Perrin, Lionel (2019). A career in catalysis: Odile Eisenstein. 9(11), p. 10375–10388. doi: 10.1021/acscatal.9b02498. Full text in Research Archive
• Suàrez, Lluís Artús; Balcells, David & Nova, Ainara (2019). Rational Cocatalyst Design for Amide Hydrogenolysis Based on DFT Calculations. Full text in Research Archive
• Balcells, David (2018). Minimizing Off-Cycle Species in Palladium- and Nickel-Catalyzed Cross-Coupling Reactions. Full text in Research Archive
• Balcells, David (2018). Minimizing Off-Cycle Species in the Design of New Catalysts for Cross-Coupling Reactions. Full text in Research Archive
• Balcells, David (2018). Rational Design of Pd and Ni Catalysts for Cross-Coupling Reactions by Minimizing Off-Cycle Species. Full text in Research Archive
• Safy, Mohamed E. A.; Flores, Ainara Nova; Olsbye, Unni & Badia, David Balcells (2026). Computationally guided CO2 to methanol catalysis: From a mechanistic investigation to a descriptor-based screening. Universitetet i Oslo. Full text in Research Archive Show summary

  CO2 hydrogenation to liquid fuels, especially methanol, is among the most attractive carbon capture and utilization techniques. This conversion can reduce CO2 emissions and produce green liquid fuels. Methanol is considered an attractive liquid organic hydrogen carrier and a valuable feedstock for important chemical industries. In basic media, Ru-MACHO-Ph combined with an amine co-catalyst is among the most active homogeneous systems reported for CO2 hydrogenation to methanol. However, the mechanistic origin of this high activity remains unclear, hindering the development of earth-abundant metal catalysts for direct CO2 hydrogenation to methanol. In this thesis, the first publication combined computational mechanistic investigations with microkinetic modeling to elucidate the mechanisms of amine-assisted CO2 hydrogenation to methanol. The theoretically obtained methanol turnover number from the microkinetic model was validated against experimental results. The model was used to analyze and compare competing mechanisms, thereby proposing the most consistent pathway and investigating the effect of the reactants’ initial concentrations on methanol production. The second publication explored why diamines exhibit enhanced activity relative to monoamines as co-catalysts and showed that a simple thermodynamic calculation can serve as a rational design principle for selecting effective amines. This work also shows that methanol can be produced in basic media using an earth-abundant metal catalyst (Mn-MACHO-Ph). The third paper integrated linear free energy scaling relationships with the energy span model to construct a molecular volcano plot, providing a predictive screening framework for bifunctional pincer complexes in CO2 hydrogenation to methanol. The resulting volcano plot correctly ranks the experimentally reported active catalyst among the most active candidates for amine-assisted CO2 hydrogenation to methanol. The model was also used to screen a newly designed Fe- and Mn-based chemical space, identifying promising systems for further computational and experimental evaluation. Overall, the thesis established a detailed mechanistic understanding and a computational model that links free energy calculations to catalytic performance, enabling the rational design of base-metal catalysts for amine-assisted CO2 hydrogenation to methanol.

• Skjelstad, Bastian; Balcells, David & Helgaker, Trygve (2020). A Computational Perspective on the Interplay Between Metal Dopant and Support in the Structure and Reactivity of Oxidized Cubanes. Universitetet i Oslo. Full text in Research Archive
• Reimann, Sarah; Fromager, Emmanuel; Gori-Giorgi, Paola & Balcells, David (2018). Treatment of Magnetic Fields in Density-Functional Theory. Universitetet i Oslo. Full text in Research Archive Show summary

  Magnetic properties are an important application area in quantum chemistry. However, for the most widely used method in electronic structure calculations, density-functional theory, it is still not understood how to rigorously include the magnetic field into the exchange--correlation functional. This work contributes to and further develops magnetic-field density-functional theory (BDFT), which is an alternative to, but less well-known, than current density-functional theory (CDFT). A major focus is to investigate the importance to introduce a magnetic-field dependence into density functional approximations (DFAs) to improve the computation of magnetic properties. The main conclusion is that the full benefit of a field dependence can only be realized if simultaneously the self-consistent density of present DFAs is improved. The investigations and results of this work provide valuable information both for further theoretical developments of BDFT and for the proper inclusion of magnetic field effects into DFAs, which will allow more accurate results for density-functional calculations involving magnetic fields.

View all works in NVA