Open Challenges in Developing Generalizable Large-Scale Machine-Learning Models for Catalyst Discovery

Fine-Tuning Graph Neural Networks via Active Learning: Unlocking the Potential of Graph Neural Networks Trained on Nonaqueous Systems for Aqueous CO2 Reduction

Graph neural networks (GNNs) have revolutionized catalysis research with their efficiency and accuracy in modeling complex chemical interactions. However, adapting GNNs trained on nonaqueous data sets to aqueous systems poses notable challenges due to intricate water interactions. In this study, we proposed an active learning-based fine-tuning approach to extend the applicability of GNNs to aqueous environments. The geometry optimization and transition state search workflows are designed to reduce computational costs while maintaining DFT-level accuracy. Applied to the CO2 reduction reaction, the workflow delivers a 2-3-fold acceleration in geometry optimization through a relaxed force threshold combined with DFT refinement. The versatility of the transition state search algorithm was demonstrated on key C-C coupling pathways, pinpointing *CO-*COH as the most energetically favorable pathway in aqueous systems of Cu and Cu-based Ag, Au, and Zn alloys. The Brønsted-Evans-Polanyi relationship remains robust under water-induced fluctuations, with alloyed metals such as Al, Ga, and Pd, along with Ag, Au, and Zn, exhibiting coupling efficiency comparable to that of Cu. Additionally, perturbation-based training on forces and energies extends the application of GNNs to aqueous ab initio molecular dynamics simulations, enabling efficient modeling of dynamical trajectories. This work presents novel approaches to adapting nonaqueous models for application in aqueous systems, highlighting GNNs' potential in solvated environments and laying a foundation for accelerating predictions of catalytic mechanisms under realistic conditions.

Application of graph neural network in computational heterogeneous catalysis

Heterogeneous catalysis, as a key technology in modern chemical industries, plays a vital role in social progress and economic development. However, its complex reaction process poses challenges to theoretical research. Graph neural networks (GNNs) are gradually becoming a key tool in this field as they can intrinsically learn atomic representation and consider connection relationship, making them naturally applicable to atomic and molecular systems. This article introduces the basic principles, current network architectures, and datasets of GNNs and reviews the application of GNN in heterogeneous catalysis from accelerating the materials screening and exploring the potential energy surface. In the end, we summarize the main challenges and potential application prospects of GNNs in future research endeavors.

Enhancing chemical synthesis research with NLP: Word embeddings for chemical reagent identification-A case study on nano-FeCu

Nanoparticle synthesis is complex, influenced by multiple variables including reagent selection. This study introduces a specialized corpus focused on "Fe, Cu, synthesis" to train a domain-specific word embedding model using natural language processing (NLP) in an unsupervised environment. Evaluation metrics included average cosine similarity, visual analysis via t-distributed stochastic neighbor embedding (t-SNE), synonym analysis, and analogy reasoning analysis. Results indicate a strong correlation between learning rate and cosine similarity, with enhanced chemical specificity in the tailored model compared to general models. The framework facilitates rapid identification of potential reagents for nano-FeCu synthesis, enhancing precision in nanomaterial research. This innovative approach offers a data-driven pathway for chemical material synthesis, demonstrating significant interdisciplinary applications.

Scalable crystal structure relaxation using an iteration-free deep generative model with uncertainty quantification

In computational molecular and materials science, determining equilibrium structures is the crucial first step for accurate subsequent property calculations. However, the recent discovery of millions of new crystals and super large twisted structures has challenged traditional computational methods, both ab initio and machine-learning-based, due to their computationally intensive iterative processes. To address these scalability issues, here we introduce DeepRelax, a deep generative model capable of performing geometric crystal structure relaxation rapidly and without iterations. DeepRelax learns the equilibrium structural distribution, enabling it to predict relaxed structures directly from their unrelaxed ones. The ability to perform structural relaxation at the millisecond level per structure, combined with the scalability of parallel processing, makes DeepRelax particularly useful for large-scale virtual screening. We demonstrate DeepRelax's reliability and robustness by applying it to five diverse databases, including oxides, Materials Project, two-dimensional materials, van der Waals crystals, and crystals with point defects. DeepRelax consistently shows high accuracy and efficiency, validated by density functional theory calculations. Finally, we enhance its trustworthiness by integrating uncertainty quantification. This work significantly accelerates computational workflows, offering a robust and trustworthy machine-learning method for material discovery and advancing the application of AI for science.

Artificial intelligence driven molecule adsorption prediction (AIMAP) applied to chirality recognition of amino acid adsorption on metals

Predicting the adsorption structure of molecules has long been a challenging topic given the coupled complexity of surface binding sites and molecule flexibility. Here, we develop AIMAP, an Artificial Intelligence Driven Molecule Adsorption Prediction tool, to achieve the general-purpose end-to-end prediction of molecule adsorption structures. AIMAP features efficient exploration of the global potential energy surface of the adsorption system based on global neural network (G-NN) potential, by rapidly screening qualified adsorption patterns and fine searching using stochastic surface walking (SSW) global optimization. We demonstrate the AIMAP efficiency in constructing the Cu-HCNO6 adsorption database, encompassing 1 182 351 distinct adsorption configurations of 9592 molecules on three copper surfaces. AIMAP is then utilized to identify the best adsorption structure for 18 amino acids (AAs) on achiral Cu surfaces and the chiral Cu(3,1,17) S surface. We find that AAs chemisorb on copper surfaces in their highest deprotonated state, through both the carboxylate-amino skeleton and side groups. The chiral recognition is identified for the d-preference of Asp, Glu, and Tyr, and l-preference for His. The physical origin for the enantiospecific adsorption is thus rationalized, pointing to the critical role of the competitive adsorption between functional side groups and the carboxylate-amino skeleton at surface low-coordination sites.

Designing asymmetrical TMN4 sites via phosphorus or sulfur dual coordination as high-performance electrocatalysts for oxygen evolution reaction

The development ofhighly efficient oxygen evolution reaction (OER) catalysts based on more cost-effective and earth-abundant elements is of great significance and still faces a huge challenge. In this work, a series of transition metal (TM)embedding a newly-defined monolayer carbon nitride phase is theoretically profiled and constructed as a catalytic platform for OER studies. Typically, a four-step screening strategy was proposed to rapidly identified high performance candidates and the coordination structure and catalytic performance relationship was thoroughly analyzed. Moreover, the eliminating criterion was established to condenses valid range based on the Gibbs free energy of OH*. Our results reveal that the as-constructed 2FeCN/P exhibits superior activity toward OER with an ultralow overpotential of 0.25 V, at the same time, the established 3FeCN/S configuration performed well as abifunctional OER/ORR electrocatalysis with extremely low overpotential ηOER/ηORR of 0.26/0.48 V. Overall, this work provides an effective framework for screening advanced OER catalysts, which can also be extended to other complex multistep catalytic reactions.

The spin polarization strategy regulates heterogeneous catalytic activity performance: from fundamentals to applications

In recent years, there has been significant attention towards the development of catalysts that exhibit superior performance and environmentally friendly attributes. This surge in interest is driven by the growing demands for energy utilization and storage as well as environmental preservation. Spin polarization plays a crucial role in catalyst design, comprehension of catalytic mechanisms, and reaction control, offering novel insights for the design of highly efficient catalysts. However, there are still some significant research gaps in the current study of spin catalysis. Therefore, it is urgent to understand how spin polarization impacts catalytic reactions to develop superior performance catalysts. Herein, we present a comprehensive summary of the application of spin polarization in catalysis. Firstly, we summarize the fundamental mechanism of spin polarization in catalytic reactions from two aspects of kinetics and thermodynamics. Additionally, we review the regulation mechanism of spin polarization in various catalytic applications and several approaches to modulate spin polarization. Moreover, we discuss the future development of spin polarization in catalysis and propose several potential avenues for further progress. We aim to improve current catalytic systems through implementing a novel and distinctive spin engineering strategy.

The Potential of Neural Network Potentials

In the next half-century, physical chemistry will likely undergo a profound transformation, driven predominantly by the combination of recent advances in quantum chemistry and machine learning (ML). Specifically, equivariant neural network potentials (NNPs) are a breakthrough new tool that are already enabling us to simulate systems at the molecular scale with unprecedented accuracy and speed, relying on nothing but fundamental physical laws. The continued development of this approach will realize Paul Dirac's 80-year-old vision of using quantum mechanics to unify physics with chemistry and providing invaluable tools for understanding materials science, biology, earth sciences, and beyond. The era of highly accurate and efficient first-principles molecular simulations will provide a wealth of training data that can be used to build automated computational methodologies, using tools such as diffusion models, for the design and optimization of systems at the molecular scale. Large language models (LLMs) will also evolve into increasingly indispensable tools for literature review, coding, idea generation, and scientific writing.

Invariant Molecular Representations for Heterogeneous Catalysis

Catalyst screening is a critical step in the discovery and development of heterogeneous catalysts, which are vital for a wide range of chemical processes. In recent years, computational catalyst screening, primarily through density functional theory (DFT), has gained significant attention as a method for identifying promising catalysts. However, the computation of adsorption energies for all likely chemical intermediates present in complex surface chemistries is computationally intensive and costly due to the expensive nature of these calculations and the intrinsic idiosyncrasies of the methods or data sets used. This study introduces a novel machine learning (ML) method to learn adsorption energies from multiple DFT functionals by using invariant molecular representations (IMRs). To do this, we first extract molecular fingerprints for the reaction intermediates and later use a Siamese-neural-network-based training strategy to learn invariant molecular representations or the IMR across all available functionals. Our Siamese network-based representations demonstrate superior performance in predicting adsorption energies compared with other molecular representations. Notably, when considering mean absolute values of adsorption energies as 0.43 eV (PBE-D3), 0.46 eV (BEEF-vdW), 0.81 eV (RPBE), and 0.37 eV (scan+rVV10), our IMR method has achieved the lowest mean absolute errors (MAEs) of 0.18 0.10, 0.16, and 0.18 eV, respectively. These results emphasize the superior predictive capacity of our Siamese network-based representations. The empirical findings in this study illuminate the efficacy, robustness, and dependability of our proposed ML paradigm in predicting adsorption energies, specifically for propane dehydrogenation on a platinum catalyst surface.

Modelling local and general quantum mechanical properties with attention-based pooling

Atom-centred neural networks represent the state-of-the-art for approximating the quantum chemical properties of molecules, such as internal energies. While the design of machine learning architectures that respect chemical principles has continued to advance, the final atom pooling operation that is necessary to convert from atomic to molecular representations in most models remains relatively undeveloped. The most common choices, sum and average pooling, compute molecular representations that are naturally a good fit for many physical properties, while satisfying properties such as permutation invariance which are desirable from a geometric deep learning perspective. However, there are growing concerns that such simplistic functions might have limited representational power, while also being suboptimal for physical properties that are highly localised or intensive. Based on recent advances in graph representation learning, we investigate the use of a learnable pooling function that leverages an attention mechanism to model interactions between atom representations. The proposed pooling operation is a drop-in replacement requiring no changes to any of the other architectural components. Using SchNet and DimeNet++ as starting models, we demonstrate consistent uplifts in performance compared to sum and mean pooling and a recent physics-aware pooling operation designed specifically for orbital energies, on several datasets, properties, and levels of theory, with up to 85% improvements depending on the specific task.

First-Row Transition Metals for Catalyzing Oxygen Redox

Rechargeable zinc-air batteries are widely recognized as a highly promising technology for energy conversion and storage, offering a cost-effective and viable alternative to commercial lithium-ion batteries due to their unique advantages. However, the practical application and commercialization of zinc-air batteries are hindered by the sluggish kinetics of the oxygen reduction reaction (ORR) and oxygen evolution reaction (OER). Recently, extensive research has focused on the potential of first-row transition metals (Mn, Fe, Co, Ni, and Cu) as promising alternatives to noble metals in bifunctional ORR/OER electrocatalysts, leveraging their high-efficiency electrocatalytic activity and excellent durability. This review provides a comprehensive summary of the recent advancements in the mechanisms of ORR/OER, the performance of bifunctional electrocatalysts, and the preparation strategies employed for electrocatalysts based on first-row transition metals in alkaline media for zinc-air batteries. The paper concludes by proposing several challenges and highlighting emerging research trends for the future development of bifunctional electrocatalysts based on first-row transition metals.

Machine Learning Methods for Small Data Challenges in Molecular Science

Small data are often used in scientific and engineering research due to the presence of various constraints, such as time, cost, ethics, privacy, security, and technical limitations in data acquisition. However, big data have been the focus for the past decade, small data and their challenges have received little attention, even though they are technically more severe in machine learning (ML) and deep learning (DL) studies. Overall, the small data challenge is often compounded by issues, such as data diversity, imputation, noise, imbalance, and high-dimensionality. Fortunately, the current big data era is characterized by technological breakthroughs in ML, DL, and artificial intelligence (AI), which enable data-driven scientific discovery, and many advanced ML and DL technologies developed for big data have inadvertently provided solutions for small data problems. As a result, significant progress has been made in ML and DL for small data challenges in the past decade. In this review, we summarize and analyze several emerging potential solutions to small data challenges in molecular science, including chemical and biological sciences. We review both basic machine learning algorithms, such as linear regression, logistic regression (LR), k-nearest neighbor (KNN), support vector machine (SVM), kernel learning (KL), random forest (RF), and gradient boosting trees (GBT), and more advanced techniques, including artificial neural network (ANN), convolutional neural network (CNN), U-Net, graph neural network (GNN), Generative Adversarial Network (GAN), long short-term memory (LSTM), autoencoder, transformer, transfer learning, active learning, graph-based semi-supervised learning, combining deep learning with traditional machine learning, and physical model-based data augmentation. We also briefly discuss the latest advances in these methods. Finally, we conclude the survey with a discussion of promising trends in small data challenges in molecular science.

Bayesian-optimization-assisted discovery of stereoselective aluminum complexes for ring-opening polymerization of racemic lactide

Stereoselective ring-opening polymerization catalysts are used to produce degradable stereoregular poly(lactic acids) with thermal and mechanical properties that are superior to those of atactic polymers. However, the process of discovering highly stereoselective catalysts is still largely empirical. We aim to develop an integrated computational and experimental framework for efficient, predictive catalyst selection and optimization. As a proof of principle, we have developed a Bayesian optimization workflow on a subset of literature results for stereoselective lactide ring-opening polymerization, and using the algorithm, we identify multiple new Al complexes that catalyze either isoselective or heteroselective polymerization. In addition, feature attribution analysis uncovers mechanistically meaningful ligand descriptors, such as percent buried volume (%V_bur) and the highest occupied molecular orbital energy (E_HOMO), that can access quantitative and predictive models for catalyst development.