Highlights
-
The process of machine learning is introduced in detail.
-
Recent developments in machine learning for low-dimensional electrocatalysts are briefly reviewed.
-
Future directions and perspectives for machine learning in hydrogen evolution reaction are critically discussed.
Abstract
Efficient electrocatalysts are crucial for hydrogen generation from electrolyzing water. Nevertheless, the conventional "trial and error" method for producing advanced electrocatalysts is not only cost-ineffective but also time-consuming and labor-intensive. Fortunately, the advancement of machine learning brings new opportunities for electrocatalysts discovery and design. By analyzing experimental and theoretical data, machine learning can effectively predict their hydrogen evolution reaction (HER) performance. This review summarizes recent developments in machine learning for low-dimensional electrocatalysts, including zero-dimension nanoparticles and nanoclusters, one-dimensional nanotubes and nanowires, two-dimensional nanosheets, as well as other electrocatalysts. In particular, the effects of descriptors and algorithms on screening low-dimensional electrocatalysts and investigating their HER performance are highlighted. Finally, the future directions and perspectives for machine learning in electrocatalysis are discussed, emphasizing the potential for machine learning to accelerate electrocatalyst discovery, optimize their performance, and provide new insights into electrocatalytic mechanisms. Overall, this work offers an in-depth understanding of the current state of machine learning in electrocatalysis and its potential for future research.
Similar content being viewed by others
Avoid common mistakes on your manuscript.
1 Introduction
With the urgency of achieving carbon neutrality, green energy sources are becoming more important in promoting economic and social growth [1,2,3,4,5]. Hydrogen, as an ideal renewable energy, has gained great attention [6,7,8]. Hydrogen generation from the electrolyzed water is the most efficient and sustainable approach to producing high-purity hydrogen [9,10,11,12,13,14]. However, hydrogen evolution reaction (HER) requires an electrocatalyst with high activity to reduce overpotential substantially. Therefore, more effort is needed in electrocatalyst discovery. Although electrocatalysts play a critical role in the HER, many are still developed through a trial-and-error approach in experimental settings [15,16,17]. However, given the vast materials space, it is challenging to explore and design exceptional electrocatalysts [18]. Furthermore, the identification of ultrafast catalytic HER atomically is difficult and expensive. Consequently, effective methods for screening electrocatalysts and understanding the HER mechanism are highly needed to be developed.
Low-dimensional materials present a distinctive edge in the context of the HER. Here are some key advantages: (1) High surface-to-volume ratios: these unique geometries endow a multitude of active sites for the HER. The consequent amplification in surface area bolsters catalytic activity, thereby augmenting the efficiency of hydrogen evolution; (2) Enhanced charge transport: low-dimensional materials often exhibit excellent electron mobility, allowing for efficient charge transport. This leads to a decrease in overpotentials, reducing the energy required to drive the reaction and thereby enhancing its efficiency; (3) Tailorable electronic structure: the electronic properties of low-dimensional materials can be finely tuned through chemical modification or by introducing defects. This customization allows for optimized binding energies for hydrogen adsorption and desorption, critical for the HER process; (4) Structural flexibility: the mechanical properties of low-dimensional materials can also be tailored to support optimal structural dynamics. This flexibility can enhance the reaction kinetics, thus making the HER process faster; (5) Integration and synergy: the adaptability of low-dimensional materials allows them to be seamlessly amalgamated with other catalysts or support materials to forge hybrid structures. This engenders synergistic effects, where the ensemble of diverse materials magnifies the overall HER performance. By leveraging these advantages, low-dimensional materials have the potential to significantly improve the efficiency and sustainability of hydrogen production through the HER.
Machine learning has emerged as an important tool for data mining and analysis, and is progressively transforming the way we collect, analyze, and discover materials [19,20,21,22], which can aid in the design and discovery of novel low-dimensional electrocatalysts with optimized properties for this important reaction. The increasing popularity of machine learning has revolutionized the predictions of new electrocatalysts, optimal composition, adsorption energy, active sites, electrocatalytic activity, and HER mechanism, enabling researchers to identify these catalysts in a faster and more cost-effective manner compared to traditional experience-based methods [8, 23,24,25,26]. For instance, Liu et al. utilized machine learning to investigate the HER process on single-atom catalysts (SACs) doped on 2D GaPS4, and illustrated the potential of SACs in catalyzing HER on 2D GaPS4 [27]. It provides an effective and economical method to predict HER performance for various low-dimensional catalysts. Wexler et al. adopted machine learning to identify key factors affecting the HER performance of Ni2P doped with nonmetal [28]. It was found that the Ni–Ni bond length was the most significant descriptor, indicating that nonmetal doping could enhance reactivity by inducing a chemical pressure effect on the Ni3-hollow site. These results provide important insights for the design of new HER electrocatalysts. Moreover, Pandit et al. employed enhanced eXtreme Gradient Boost Regression models to select NiCoCu alloy-based catalysts for the HER, and demonstrated that this approach successfully screened the most active HER catalysts from a vast array of candidates [29]. To investigate structure/property relationships, Parker et al. analyzed 1300 platinum ensembles by machine learning. Their findings revealed that small particles were conducive to disordered materials, while ultra-large (110) surface areas were supportive of ordered materials to achieve efficient hydrogen evolution [30]. Besides, machine learning can be utilized to design experiments for fabricating materials instead of designing materials themselves [31,32,33]. Despite considerable research on low-dimensional electrocatalysts for the HER with machine learning, a comprehensive summary of this approach is still lacking. Hence, it is crucial to review recent advancements in the machine learning applications to low-dimensional HER electrocatalysts.
In this work, we first present a detailed introduction of the general scheme of machine learning, encompassing data collection, feature engineering, machine learning algorithms, as well as model optimization. Subsequently, we summarize the latest advancements in low-dimensional electrocatalysts guided by machine learning, specifically focusing on their application to the HER. A particular emphasis is placed on comprehending the descriptor performance and enhancing the scope of the application and design effectiveness. Finally, future development prospects and directions for machine learning methods in HER research are proposed.
2 General Process of Machine Learning
When compared to traditional hard-coded screening methods, in which algorithms were prearranged by human specialists prior to their use, computers are capable of learning from training data in the middle of a machine-learning process, allowing them to screen electrocatalytic materials automatically [34,35,36]. A machine learning model can be constructed as a result of linear or nonlinear relationships that exist between features and material properties, and it is through this fundamental step that the applicability and feasibility of the model can be evaluated [37,38,39]. Subsequently, using the correct algorithms, models are constructed and predictions are made about reaction mechanisms or properties based on the data presented [40]. Finally, after the training and test sets have been established, the machine learning models are validated, assessed, and optimized [41]. In Fig. 1, we illustrate the process of using machine learning-accelerated computations and designing electrocatalysts for the HER. Subsequently, the utilization of machine learning will be described with detailed elaboration.
2.1 Data Collection
In machine learning, the most critical step is the collection of data [42]. To obtain the desired data, a digital format is used, based on the data gathered from the experiment, the calculation, and the simulation, as well as from the database [43]. Furthermore, to integrate data from a variety of sources during the collection process, data fusion is performed [44]. As a result, the merged data is more comprehensive and informative as compared with the original data. Nevertheless, the collected data often contains noise, incorrect, irrelevant, and even missing data, which can significantly compromise the quality of the machine learning model [45]. The quality of the data is evaluated based on various indexes, including validity, accuracy, completeness, consistency, and uniformity [46,47,48,49]. Fortunately, databases for materials science have been consistently built over the past century and can provide large raw data, bringing great convenience for machine learning [50,51,52]. Materials databases, for example, the Inorganic Crystal Structure Database, the Pauling File, the Crystallography Open Database, and similar databases are widely used as data sources [20, 53,54,55]. Additionally, there is no doubt that the use of databases of computed materials, such as those developed by the Material Project and Open Quantum Materials, is becoming increasingly common [9, 56,57,58,59]. Databases for specific applications, including Materials Web online database, Cambridge Structural Database, and Catalyst Acquisition by Data Science, are also adopted [13, 60,61,62,63,64]. Furthermore, low-dimensional materials databases are specialized repositories of information and data related to various low-dimensional materials, such as 2D materials (e.g., graphene, transition metal dichalcogenides) and 1D nanowires (e.g., carbon nanotubes, nanowires). Renowned databases include but are not limited to Materials Project, 2D Materials Database [65], Nanomaterial Registry [66], and Computational 2D Materials Database [67]. These databases serve as valuable resources for researchers and scientists working in the field of materials science and nanotechnology. More importantly, rapid access to materials databases is crucial for data collection in machine learning.
After the data collection process, the collected data may be incomplete, inconsistent, or even spurious, and may not be compatible with the working environment [68]. As a precaution to make sure that predicting models will perform well, irrelevant and duplicate data should be removed. Machine learning models require accurate data that is free of errors, redundant, erroneous, or duplicate information within a dataset, and data cleaning can eliminate or correct such errors [21, 69,70,71]. In data cleaning, several steps are involved, such as sampling, processing abnormal values, discretizing data, and normalizing data [59]. With data sampling, highly accurate prediction models can be derived with fewer data, and abnormal values must be eliminated to ensure model accuracy. It is known that data discretization reduces the possible values for a continuous feature, while data normalization defines the data magnitudes at the same level to optimize several machine learning algorithms. It is possible that the dataset may be messy even after reducing noise, data redundancy, as well as abnormal values through these steps [72,73,74].
To ensure the representativeness and unbiased nature of the data used for model training, a multi-faceted approach is adopted. This includes defining clear objectives, selecting diverse and reliable data sources, using random sampling techniques, building a large and diverse dataset, preprocessing and cleaning the data, handling missing data appropriately, and detecting and mitigating biases. It must be acknowledged that achieving the complete elimination of biases remains a formidable challenge, however, a systematic and robust approach, coupled with regular assessments and monitoring, can substantially enhance the prospects of fairness and accuracy in the trained models.
2.2 Feature Engineering
After collecting enough available data, appropriate features are obtained from the raw data and tasks, which is a key step in predicting targeted properties without redundancy [33, 75]. Features are also known as descriptors, which enable machine learning algorithms to perform at their best [76, 77]. Generally, an input descriptor can take the form of a number, a vector, a matrix, or a character string, and represents a series of input data assigned to the characteristic properties of the materials [78, 79]. If appropriate features are selected during feature extraction, the model can be made more understandable and accurate while reducing its dimensionality and complexity. It involves a cautious and comprehensive collection of reasonable descriptors [80,81,82]. In addition, feature selection not only simplifies, accelerates, and improves the interpretability of the learner models, but also allows for a deeper exploration of material characteristics through the use of feature engineering tools (Columbus, DeepDive, and Explorekit) [83,84,85]. Notably, the number of features should be considered before choosing features for a model [86,87,88]. Raw scientific data is converted into useful features that encode material information, including the density of state, Voronoi tessellations, chemical species, and one-hot encoded composition-based feature vectors [89,90,91,92,93]. For electrocatalytic materials, features related to the crystal and geometrical structure, element compositions, and electronic properties are used to predict and disclose destined properties containing activity, stability, and selectivity [19, 44, 94,95,96,97,98,99]. The electrocatalytic properties of materials can be predicted by converting features into descriptors, which should be broad and efficient in describing such properties [100]. However, the initial guess structure is a starting point, it may not invariably yield accurate structural descriptors. To enhance precision, a multifaceted approach often necessitates a synergy of additional strategies, iterative refinement, and optimization. For instance, in the landmark study by Chen et al. [101], a novel machine learning framework was devised to optimize local structures through a local machine- learning potential (MLP). This pioneering method enabled the extraction of precise structure descriptors, leading to the identification of 43 high-performance alloys as potential HER electrocatalysts from a pool of 2973 candidates. Several top candidates were further corroborated experimentally, with the AgPd alloy being systematically scrutinized using ab initio calculations under realistic electrocatalytic conditions to attest to the framework's accuracy. This approach epitomizes the fusion of computational efficiency and precision using optimized local MLP structural descriptors, offering a path forward in the design of high-performance electrocatalysts.
An electrocatalytic reaction, for example, is characterized by the bond energy formed between adsorbed hydrogen and the electrocatalyst, and this is expressible by the hydrogen adsorption energy (EHad), and the d-band structure. These two factors show the catalytic activity of the reaction [102]. Additionally, electronic and structural properties, such as Fermi level, work function, electron affinity, coordination numbers, and atomic radial distribution functions, are also important features [4, 20, 57, 96, 103,104,105,106,107,108,109,110]. With the rapid development of deep learning, automated feature engineering is now extensively utilized [111]. Using deep learning, computers are capable of learning features automatically from data and combining them during model construction, reducing the inadequacy of manual feature engineering [112]. A wide range of applications are possible with this technology including drug delivery, batteries, bioinformatics, and nanotechnology [113,114,115,116].
2.3 Machine Learning Algorithms
Machine learning is an efficient and powerful strategy towards predicting electrocatalytic performance due to its ability to correlate input features (properties of the electrocatalyst) with output parameters (electrocatalytic performance) [117, 118]. The selection of an appropriate machine learning algorithm is crucial for achieving high prediction and generalization ability. Several mathematical theories have been applied to the creation of machine learning algorithms, including Markov chains, least squares methods, and Gaussian processes. In materials science, machine learning algorithms can be grouped into three types: classification, regression, and clustering (Fig. 2), with each algorithm offering unique advantages and limitations. While the details of each algorithm have been reviewed and summarized elsewhere [31, 41, 94], we present here some representative algorithms that have been employed for establishing deep structure–activity relationships.
Material science has become increasingly interested in kernel-based algorithms because they are capable of handling complicated regression issues beyond the capabilities of simple linear models. As a result of the kernel trick, ridge regression, and classification are often combined to yield kernel ridge regression (KRR), which can predict the electrochemical activity of electrode materials according to their structural characteristics [119]. As a binary classification and regression algorithm, the support vector machine (SVM) classifies training data into two distinct classes with a low error by using a hyperplane. For instance, SVM can classify high-entropy alloy solid solution phases across feature ensemble sizes [20]. In both SVM and KRR, the hyperparameter settings have a significant impact on performance. SVM takes less time in prediction, while KRR spends less time to fit medium datasets. Despite this, the training time of SVM and KRR cannot be exclusively attributed to the dataset size. Rather, it emerges as a function of various intertwined factors, encompassing the complexity of the algorithms, the specific attributes of the dataset, and potentially other underlying considerations that contribute to the overall computational demand. An artificial neural network (ANN), which simulates biological neural networks, is often included in machine learning models. Because of its self-learning and self-adaptive abilities, it can adjust its internal structure as it learns and adapts to external information. As a reliable algorithm, it has been utilized to forecast the rest usable lifetime of Li-ion batteries online and screen catalysts for their effectiveness as a reliable algorithm for large datasets [58, 120, 121]. Moreover, Rice and coworkers utilized a high-dimensional neural network, along with sampling methodologies and DFT calculations, to explore the efficiency of the hydrogen coupling reaction at the H2O/Pt (111) interface [122]. The aim was to obtain a detailed atomistic understanding of how the presence of an aqueous medium impacts the structure and reactivity of the HER. By using cutting-edge tools and techniques, the research team was able to gain valuable insights into the underlying mechanisms of the reaction, which can inform the development of more effective HER electrocatalysts. This approach provides a detailed and comprehensive picture of the hydrogen coupling reaction at the interface, which can serve as a basis for further research and advancement in the field of the HER. ANN is capable of addressing complex nonlinear problems with large datasets, but requires a deeper understanding of machine learning, which is more difficult to operate than SVM and KRR owing to the weight setting [123, 124]. Furthermore, the K-nearest neighbor algorithm is applied extensively to identify and predict materials with excellent performance [125,126,127]. In the case of large datasets, however, it may require a considerable amount of time and memory. Also, other algorithms, like logistic regression and decision trees, are utilized, which require strict mathematical reasoning, strong interpretation, and fast running speeds, making them suitable for small datasets [128, 129].
2.4 Model Optimization
Data sets are crucial to evaluating the stability and performance of machine learning models in material research. Accurately dividing these data sets into training and test sets is a critical step in assessing the model's effectiveness. Typically, 80% of the original data is found in the training set, while a test set includes the remainder. The process of evaluating errors in both the training and test sets provides valuable insight into the extent of underfitting or overfitting of a model. Overfitting occurs when training errors are much smaller than validation errors, whereas underfitting occurs when both errors are high but the gap between them is small. To prevent overfitting or underfitting within the models, several strategies can be employed. These encompass techniques such as cross-validation to assess how the results of a statistical analysis will generalize to an independent dataset, regularization to add some form of penalty to the loss function, feature selection to choose relevant predictive variables, increasing training data to promote a richer understanding of underlying patterns, early stopping to terminate training when validation performance deteriorates, ensemble methods to combine predictions from multiple models, and data augmentation to increase the diversity of the training set. It is important to note that the selection of an appropriate strategy depends on the specific problem, as well as the characteristics and nuances of the underlying data. Through careful and considered application of these techniques, it is feasible to mitigate the risks of overfitting and underfitting, thereby enhancing both the performance and the generalizability of the model. For preventing overfitting and effectively using limited data, cross-validation methods are commonly utilized. Typically, these methods involve running the model multiple times with different data sets, and splitting each set in a way that maximizes the use of the data available in each set. There are several methods of cross-validation, including k-fold cross-validation, repeated random subsampling analysis, Bootstrap cross-validation, and leave-out cross-validation [130,131,132,133]. Machine learning models can be assessed and compared by utilizing various metrics, such as mean-absolute-error (MAE), mean squared error (MSE), and root mean square error (RMSE) [134].
3 Machine Learning Application for the HER
Materials science and engineering researchers are committed to developing materials that have desired properties rationally and efficiently [31, 135,136,137]. However, traditional trial-and-error methods have significant drawbacks, including high costs and time investments, which can impede progress and limit the exploration of advanced materials [138,139,140]. As machine learning advances, researchers can develop models that can make accurate predictions with greater efficiency than pure DFT calculations [141]. This revolutionary trend has greatly enhanced the development of the HER, as evidenced by the publication of over 1741 related articles related to machine learning methods to the HER during the past 8 years (Fig. 3a). In this section, an in-depth discussion of how machine learning techniques can be applied to the advancement of low-dimensional electrocatalysts will be presented (Fig. 3b), including noble metals, metal alloys, MXenes, carbon-based materials, metal phosphides, metal dichalcogenides, and others, and examine their performance. In the research for the HER, machine learning techniques play an essential role, and their application will significantly accelerate the progress toward developing a rational design of advanced electrocatalysts, thereby speeding up the research process.
3.1 0D Electrocatalysts
It is one of the crucial issues to discover excellent multi-metallic alloy (MMA) electrocatalysts that contain an optimal component and composition for the HER [80]. Recently, through the use of active machine learning in conjunction with experimentation, an optimal MMA catalyst was found. The model resulting from the experiment was successfully built solely by using the composition of the mixed precursor as input data, which resulted in an enhanced electrocatalytic performance (Fig. 4a) [142]. Interestingly, the strategy could be applied extensively by adjusting an electrocatalyst composed of an optimal component. As displayed in Fig. 4b, only binary data were used to train the model, which resulted in high uncertainties. A significant reduction in the uncertainty of the Pt-Ru-Ni sample has been achieved by updating the model in an extensive range of compositional possibilities with a minimum overpotential. Also, from Fig. 4c, d, we can see that there is the less expected improvement in model performance with additional iterations than at the beginning of the loop. Furthermore, as presented in Fig. 4e, several points were located. Consequently, the HER overpotential of an optimal Pt0.65Ru0.30Ni0.05 electrocatalyst was 54.2 mV, even exceeding a pure Pt catalyst. This work simplifies the challenges associated with identifying efficient catalyst components and compositions and can be extended to other catalytic reactions.
In addition, nanoclusters possess alterable catalytic activity at the nanoscale, rendering them promising candidates for catalysis applications [143,144,145,146]. Nevertheless, computational screening methods face difficulties owing to numerous relevant atomic sites in nanoclusters [102]. To address this issue, new techniques for efficient exploration are required. In this regard, as a means of replacing noble metals with eco-friendly and cost-effective HER electrocatalysts, Cu-based alloy nanoclusters with 7924 candidate structures were screened (Fig. 5a, b) [147]. High-throughput DFT computations and machine learning were employed to identify the most stable core–shell configurations for Cu55-nMn (M = Co, Ni, Ru, and Rh, and n ≤ 22) nanoclusters. A descriptor that eliminated the burdensome computations was developed to assess the activity of the HER on nanoclusters. Subsequently, an ANN was trained with a large DFT database to quickly and precisely predict the EHad on the surface of the nanoclusters (Fig. 5c–e). This work provided an effective strategy to quickly discover HER electrocatalysts with great potential based on metal alloy nanoclusters. Additionally, machine learning was found to lower the consumption of modeling various adsorption site structures aided by descriptors [148]. Advanced structural descriptors, including Many-Body Tensor Representation, Smooth Overlap of Atomic Positions, and Atom-Centered Symmetry Functions, were investigated by Jager et al. for predicting the Gibbs free energy of hydrogen adsorption (ΔGH*) on the nanoclusters [102]. The accuracy of the descriptors in KRR was evaluated by potential energy scans of hydrogen on the nanocluster surfaces. Analyzing the data sets of 91 molybdenum disulphide nanoclusters and 24 copper–gold nanoclusters, machine learning could reduce MAE by learning diverse nanoclusters simultaneously as opposed to sequentially. Furthermore, they observed a marked reduction in the fitting of potential energy surfaces when data from different nanoclusters were merged. Also, their group demonstrated a machine learning-based workflow for nanocluster configurations and adsorption energy screening [149]. The results presented that the adsorption was exemplified on the HER, and the maximum of the d-band Hilbert-transform ϵu was associated with the EHad at the nanocluster level.
3.2 1D Electrocatalysts
Through ab initio simulations of reaction and activation energies, the HER kinetics of various electrocatalysts have been extensively studied [150]. The effect of the water static layers, hydrogen-bond networks, adsorbed species, and electric double layers increase the uncertainties in the energetics related to the DFT methods [151,152,153]. Although the Tafel slope and the EHad are frequently applied to study the HER kinetics, methods for revealing the HER kinetics remain challenging. Therefore, simulating and interpreting kinetics with adequate rate expressions is an essential step for understanding HER mechanisms [122]. Gu et al. utilized an end-to-end avenue to aid in simulating the kinetics of jagged Pt nanowires by machine learning multiscale method (Fig. 6a), which was consistent with the experimental results in alkaline media [154]. They demonstrated that the optimal ΔGH* value in alkaline solution for the overall rate was lower than that of acidic condition, and the jagged Pt nanowires exhibited an auto-bifunctional mechanism, namely, protons were adsorbed on the stronger binding sites and hydrogen was activated on weaker binding sites. Meantime, unsupervised machine learning model results indicated that ΔGH* was interrelated with the coordination number, as well as the sites with CN ≤ 7 showed great activity (Fig. 6b, c). Additionally, as displayed in Fig. 6d, the plot of the ΔrGads values and coordination numbers of top sites from the Pt nanowire proportion indicated that exposed sites with low coordination were nearer to the optimal ΔrGads value. This work is conducive to comprehending complex kinetic processes on the nanoscale and HER mechanism.
Adsorbed hydrogen, as an intermediate, plays a crucial role in the HER [155], and EHads are commonly served as descriptors in machine learning. For example, Kronberg et al. investigated and interpreted the hydrogen adsorption performance on defective nitrogen-doped carbon nanotubes (NCNTs) using DFT simulations and Shapley additive explanations (SHAP) analysis based on machine learning (Fig. 7a) [156]. The authors achieved significant results, as evidenced by the MAE and RMSE results presented in Fig. 7b, which showed highly accurate predictions on the training set, attaining a chemical accuracy with an R2-score of 0.99. According to Fig. 7c–e, the attributions of each feature were summarized based on the mean magnitude of the SHAP values. More importantly, by using the SHAP strategy, they examined several chemical, structural, and electronic properties of around 6500 different NCNTs in relation to hydrogen adsorption. Some factors are considered to be responsible for the increased hydrogen adsorption strength of the high spin polarization as well as the dopant-adsorption sites, narrow gap between the highest occupied and lowest unoccupied molecular orbitals, and diverse angles and coordination effects. The method also allowed for the investigation of catalyst durability problems, as indicated by the breakage of pyrrolic nitrogen bonds, highlighting the balance between defect- and dopant-caused activation and extreme instability (Fig. 7f–i). Moreover, four feature pairs interactions were investigated (Fig. 7j–m). The results showed that different features would affect each other, potentially impacting the model output. Prominently, it has been shown that analogous features could affect the hydrogen adsorption and HER performance on defective NCNTs, which may exert a positive influence on the future design of HER electrocatalysts.
3.3 2D Electrocatalysts
The use of machine learning to screen large material databases for appropriate electrocatalyst properties offers significant advantages in terms of reduced experimental cycles and costs, as well as addressing the challenge of material selection [157]. 2D materials with more active sites show great potential in large-scale hydrogen production, which is expected to replace Pt catalysts [158,159,160,161,162]. Nevertheless, the rapid discovery of high-performance 2D HER catalyst is greatly depressed owing to the long experiment cycles and high costs associated with high-throughput calculations for adsorption energies [163]. Particularly, MXene-based materials are gaining considerable attention because of their potential use in the HER [5, 155, 164]. For instance, the 2D metal carbide semiconductor Ti2CO2 MXenes with a cost-effective basal plane has been used as an electrocatalyst for the HER, and the corresponding HER performance could be modulated by immobilizing single transition metal (STM) atoms onto Ti vacancies (Fig. 8a–c) [165]. Accordingly, machine learning methods were used to identify 27 different MXenes Ti2CO2-STMs and 81 HER catalytic active sites using a facile descriptor, which was then applied to explain the trends in Ti2CO2-STM HER catalysis and search for other highly active HER catalysts (Fig. 8d). As exhibited in Fig. 8e, the training error decreased unremittingly before overfitting occurred. Subsequently, two descriptors of dM1–O and Ef were selected for constructing a machine-learning model with KRR towards the HER (Fig. 8f, g). From the symbolic regression, which used dM1–O, Ef, covalent radii rO, rTi, rC, and rH as input, and the expression for ΔGH* could be obtained (Fig. 8h, i). To verify the validity of the descriptor, the ΔGH* values of Ti2CO2-STM and Zr2CO2-STM were calculated and showed small R2 scores. (Fig. 8j, k). A fitting coefficient was defined to further prove the descriptor's ability to explore promising HER catalysts, which successfully predicted all 81 points of Ta2CO2-STM, as exhibited in Fig. 8l. The corresponding results implied that STM doping not only optimized the ΔGH* but also transformed semiconductors into conductors, and hybridization of the p-d orbitals between STM and C/O resulted in the rearranging of electrons close to the Fermi level, improving HER performance. In addition, the adsorption energy of hydrogen on the active sites is a primary factor determining HER activity [19]. For instance, as compared to traditional methods, Zheng et al. showed that machine learning models were more effective in predicting the ΔGH* of S-terminated and bare MXenes [166]. Four machine learning models, including Elman ANN, support vector regression, KRR, as well as random forest algorithms (RFA), were selected to predict ΔGH*. According to the results, ΔGH* values predicted via the RFA had a high-level accuracy, and testing RMSE was 0.27 eV. Furthermore, the ΔGH* values of Os2B- and S-terminated Scn+1Nn (n = 1, 2, 3) approached zero within a wide hydrogen coverage, resulting in excellent HER. In addition, HER performance was improved through the regulation of antibonding states by S functional groups. Moreover, the successful development of machine learning models coupled with DFT calculations has enabled the prediction and design of HER electrocatalysts from numerous bare and single-atom doped B-based MXenes [167]. Utilizing a support vector algorithm with structural and elemental features, the ΔGH* values were computed for 271 B-based MXenes. This approach allowed for the prediction of diverse active catalysts, including Co/Ni2B2, Pt/Ni2B2, Co2B2, Os/Co2B2, and Mn/Co2B2 with small ΔGH* values. Among these materials, Co2B2 and Mn/Co2B2 were identified as the most stable electrocatalysts, as they exhibited small ΔGH* over a broad hydrogen coverage. This study indicated that machine learning models were exciting tools for developing advanced HER electrocatalysts. Besides, the integration of high-throughput DFT and machine learning models have garnered significant attention for determining activity trends in 2D MXenes and guiding HER catalyst design [168]. This approach has led to the screening of 188 ideal HER catalysts with high mechanical and thermal stability from 2520 candidates that can be experimentally synthesized. Notably, 110 of these 2D MXenes exhibited exceptional thermos-stability and HER activity, surpassing even the ideal Pt catalyst. Additionally, extensive accuracy was achieved in predicting the HER activity for 2D MXenes using the AdaBoost ensemble learning method, suggesting that this approach is a promising strategy for developing superior 2D MXene catalysts.
In recent studies, it has been demonstrated that certain 2D graphene and graphdyine-based SACs present exceptional activity for the HER [80, 169,170,171,172,173]. To explore the efficient HER catalysts, Fung et al. applied machine learning to systematically screen SACs anchored on 2D graphene [174]. It has been demonstrated that the dimensionality and size of the substrate can affect hydrogen adsorption on metals, resulting in modifications in the electronic structure of the metal. Furthermore, when compared to graphene-based SACs including V, Rh, and Ir, nanographene containing these elements displayed improved HER performance. This strategy not only facilitates the rapid exploration of better HER catalysts by modulating and doping the nanographene support but can also be easily extended to other catalytic reactions. Analogously, 104 graphene-supported SACs and their applications for the HER were described using machine learning models, along with a quantitative evaluation of their importance [175]. The outstanding catalytic activities of these machine learning-recommended SACs were documented by DFT calculations, and the best-performing catalysts delivered an ultralow overpotential of 3 mV for the HER, exceeding the noble metal Pt. As part of this machine learning approach, factors such as geometry optimization, total energy calculation, and analysis of reaction pathways were not included. Significantly, this method can be employed to screen and construct other catalysts for N2 and CO2 reduction reactions as well. Moreover, Sun et al. employed machine learning models to perform theoretical calculations and design potential graphdyine (GDY)-atomic catalysts (AC) electrocatalysts, while also investigating the HER mechanism, reaction pathway, and adsorption energies by considering multiple parameters [176]. This study represents the first comprehensive calculation of the EHads on different active sites of GDY-M (transition/lanthanide metals) during the HER. The electronic structures of the GDY-AC electrocatalysts were also evaluated and screened out. Additionally, the authors utilized the bagged-tree approach as the machine learning algorithm, based on the fuzzy model for data separation, to predict the EHads for various AC systems. This work provides innovative theoretical comprehension and direction for GDY-based AC for the HER.
In addition, recent reports have highlighted other metallic alloy catalysts with 2D active surfaces. Bimetallic alloy catalysts containing dominant (100) facet exhibited decent catalytic performance and durability [177,178,179,180]. However, there has been less attention paid to the alloying effects at (100) surfaces compared to surfaces that are closely packed. By means of high-throughput DFT calculations and machine learning methods, Li et al. demonstrated that certain (100) surfaces alloyed with strong-binding metals (Pd and Pt) and weak-binding metals (Ag, Au, and Cu) are capable of operating as HER catalysts in acidic solutions (Fig. 9a) [178]. To systematically predict the HER activity of other bimetallic alloys, a machine-learning model was developed by the DFT-calculated database. Before machine learning modeling, feature selection was conducted (Fig. 9b, c), determining the most important input variables to predict H binding energy. Subsequently, a BPNN model was constructed by the selected features. The results, shown in Fig. 9d–i, indicated excellent accuracy for both the training and test sets. For different combinations, the scores of the training and test all surpassed 0.97 with small RMSEs (less than 0.01). This study found that PdxAg1−x and PdxAu1−x, due to their highly active fourfold ensembles, showed encouraging HER activities on their surfaces (100), relative to their monometallic counterparts. It is anticipated that this work will offer valuable direction in designing and fabricating bimetallic alloys for the HER.
3.4 Others
In the electrocatalysis field, the active sites on the surfaces of 0D, 1D, and 2D materials have been found to exhibit effective catalytic performance for the HER [181,182,183]. However, the bulk nature of these materials leads to relatively low numbers of active sites, which limits their potential as compelling catalysts, thus constraining the exploration of HER performance in bulk materials using machine learning methods. Nevertheless, some bulk materials, including metal–organic framework (MOF) and porous graphene, possess a multiscale porous structure that can also act as catalytic active sites [140]. For instance, Li et al. developed a ruthenium-adapted MOF electrocatalyst that distributed atoms at an atomic level. This electrocatalyst can efficiently adjust the MOF metal center's electronic structure, and it convincingly catalyzes the HER reaction [184,185,186]. However, the theoretical calculations involve multi-dimensional interactions of larger systems, resulting in very few studies by means of machine learning. This presents an opportunity to explore and better understand the electrocatalytic performance of these materials coupled with machine learning in the future.
4 Conclusions and Perspectives
Without large-scale trial-and-error experiments or theoretical calculations, machine learning can efficiently explore electrocatalysts and predict their properties. This technique has been widely used to rapidly and effectively screen low-dimensional electrocatalysis systems, providing a fundamental understanding of electrochemical reactions from the atomic scale. By elucidating the development of input characteristics, structures, and descriptors, as well as learning algorithms for performance prediction and electrocatalyst screening, this work highlights the advances in machine learning methods for investigating, comprehending, and optimizing the HER. Despite the great progress that has been achieved in the HER through the implementation of machine learning techniques, there are some trends in developing high-performance HER electrocatalysts via machine learning, as illustrated in Fig. 10.
-
1.
In the field of electrocatalysis for the HER, the development of efficient and effective datasets is crucial. Accurate prediction of the activity, selectivity, and stability of different electrocatalysts is essential in designing superior materials for HER electrocatalysis. Machine learning can aid in this process by analyzing large datasets and identifying patterns and correlations between various properties of the electrocatalysts. An effective dataset for HER electrocatalysts should include a diverse set of electrocatalysts with varying properties, such as composition, morphology, crystal structure, and surface area, among others. The dataset should also include precise experimental measurements of key properties, such as overpotential, exchange current density, and Tafel slopes. A lack of precise and accurate data can lead to incorrect predictions and unreliable models. Accurate datasets with precise measurements of key properties can aid in the development of reliable machine learning models and enable the discovery of novel electrocatalysts with improved HER performance.
-
2.
With the emergence of intelligent robots and 3D printing technology, the development of electrocatalysts has rapidly advanced. In fact, it has become possible for robots to predict electrocatalysts in their self-driving laboratories. To further enhance research and development efficiency, an innovative robotic platform will be developed by integrating machine learning with AI chemists and high-throughput experiments. This platform will aim to automatically optimize experimental design and improve research efficiency. It will also incorporate data-driven robotic synthesis, robot-assisted controllable preparation, and HER performance-oriented inverse design to effectively enhance the development of electrocatalysts.
-
3.
Aside from the current effective descriptors, low-cost computation, and environmentally-friendly preparation techniques used for the HER, it is essential to establish more powerful multi-objective optimization models. These models are crucial in screening and predicting suitable electrocatalysts for HER, and will greatly improve the efficiency and accuracy of the screening process. By incorporating these models, the development of HER electrocatalysts can reach new heights, resulting in a more sustainable and efficient energy production.
-
4.
The electrocatalytic activity of HER is mainly influenced by the electronic and geometrical structures of active sites at an atomic level. However, other macroscopic factors such as solvents and electrical fields can also impact the HER performance. To improve the predictability of novel electrocatalysts, it is essential to develop cross-scale models that incorporate experimental and environmental parameters. By doing so, the predictive power of electrocatalysts can be enhanced, leading to the development of more efficient HER electrocatalysts. Such advancements will facilitate the design and production of electrocatalysts that are better suited for HER, resulting in more sustainable and efficient energy production. However, incorporating experimental and environmental parameters into cross-scale models can dramatically increase the complexity of the modeling framework. Such an increase in complexity necessitates the employment of more advanced algorithms and substantial computational resources. Consequently, this can lead to a considerable escalation in both computational cost and the time required for simulations.
In summary, machine learning is a versatile and comprehensive quantitative methodology that holds significant potential for advancing research in the field of HER. Its adaptable and flexible nature makes it a promising tool for further progress in this domain. This review aims to inspire theoretical and experimental investigations into the use of machine learning, ultimately driving the field forward. Nevertheless, we emphasize the crucial role of rigorous scientific inquiry in unlocking the full potential of this approach. Therefore, this review aims to facilitate the widespread deployment of machine learning in high-performance HER electrocatalysts.
References
J. Xiong, D. Xu, Relationship between energy consumption, economic growth and environmental pollution in China. Environ. Res. 194, 110718 (2021). https://doi.org/10.1016/j.envres.2021.110718
S. Ligani Fereja, P. Li, Z. Zhang, J. Guo, Z. Fang et al., W-doping induced abundant active sites in a 3D NiS2/MoO2 heterostructure as an efficient electrocatalyst for urea oxidation and hydrogen evolution reaction. Chem. Eng. J. 432, 134274 (2022). https://doi.org/10.1016/j.cej.2021.134274
B. Zhao, J. Liu, C. Xu, R. Feng, P. Sui et al., Hollow NiSe nanocrystals heterogenized with carbon nanotubes for efficient electrocatalytic methanol upgrading to boost hydrogen co-production. Adv. Funct. Mater. 31(8), 2008812 (2020). https://doi.org/10.1002/adfm.202008812
A. Mistry, A.A. Franco, S.J. Cooper, S.A. Roberts, V. Viswanathan, How machine learning will revolutionize electrochemical sciences. ACS Energy Lett. 6(4), 1422–1431 (2021). https://doi.org/10.1021/acsenergylett.1c00194
Y. Wu, W. Wei, R. Yu, L. Xia, X. Hong et al., Anchoring sub-nanometer Pt clusters on crumpled paper-like mxene enables high hydrogen evolution mass activity. Adv. Funct. Mater. 32(17), 2110910 (2022). https://doi.org/10.1002/adfm.202110910
J. Zhang, R. Cui, C. Gao, L. Bian, Y. Pu et al., Cation-modulated HER and OER activities of hierarchical VOOH hollow architectures for high-efficiency and stable overall water splitting. Small 15(47), 1904688 (2019). https://doi.org/10.1002/smll.201904688
G. Qian, J. Chen, T. Yu, J. Liu, L. Luo et al., Three-phase heterojunction NiMo-based nano-needle for water splitting at industrial alkaline condition. Nano-Micro Lett. 14, 1–15 (2022). https://doi.org/10.1007/s40820-021-00744-x
C. Pylianidis, V. Snow, H. Overweg, S. Osinga, J. Kean et al., Simulation-assisted machine learning for operational digital twins. Environ. Modell. Softw. 148, 105274 (2022). https://doi.org/10.1016/j.envsoft.2021.105274
J. Li, J. Luo, H. Chen, B. Qin, C. Yuan et al., Strong electronic metal-support interaction between Pt and stainless mesh for enhancing the hydrogen evolution reaction. Chem. Commun. 58(71), 9918–9921 (2022). https://doi.org/10.1039/d2cc02988j
J. Li, J. Wang, Y. Liu, C. Yuan, G. Liu et al., Sodium tungsten bronze-supported Pt electrocatalysts for the high-performance hydrogen evolution reaction. Catal. Sci. Technol. 12(14), 4498–4510 (2022). https://doi.org/10.1039/d2cy00577h
S. Anantharaj, Ru-tweaking of non-precious materials: the tale of a strategy that ensures both cost and energy efficiency in electrocatalytic water splitting. J. Mater. Chem. A 9(11), 6710–6731 (2021). https://doi.org/10.1039/d0ta12424a
C. Walter, P.W. Menezes, M. Driess, Perspective on intermetallics towards efficient electrocatalytic water-splitting. Chem. Sci. 12(25), 8603–8631 (2021). https://doi.org/10.1039/d1sc01901e
T. Gu, R. Sa, L. Zhang, D.-S. Li, R. Wang, Engineering interfacial coupling between Mo2C nanosheets and Co@NC polyhedron for boosting electrocatalytic water splitting and zinc-air batteries. Appl. Catal. B: Environ. 296, 120360 (2021). https://doi.org/10.1016/j.apcatb.2021.120360
J. Zhang, J. Li, H. Huang, W. Chen, Y. Cui et al., Spatial relation controllable di-defects synergy boosts electrocatalytic hydrogen evolution reaction over VSe2 nanoflakes in all pH electrolytes. Small 18(47), 2204557 (2022). https://doi.org/10.1002/smll.202204557
J. Li, J. Zhang, J. Shen, H.-H. Wu, H. Chen et al., Self-supported electrocatalysts for hydrogen evolution reaction. Mater. Chem. Front. 7, 567–606 (2023). https://doi.org/10.1039/d2qm00931e
Y. Zhao, S. Wei, K. Pan, Z. Dong, B. Zhang et al., Self-supporting transition metal chalcogenides on metal substrates for catalytic water splitting. Chem. Eng. J. 421, 129645 (2021). https://doi.org/10.1016/j.cej.2021.129645
Y. Zhao, S. Wei, L. Xia, K. Pan, B. Zhang et al., Sintered Ni metal as a matrix of robust self-supporting electrode for ultra-stable hydrogen evolution. Chem. Eng. J. 430, 133040 (2022). https://doi.org/10.1016/j.cej.2021.133040
C.-Y. Pei, T. Li, M. Zhang, J.-W. Wang, L. Chang et al., Synergistic effects of interface coupling and defect sites in WO3/InVO4 architectures for highly efficient nitrogen photofixation. Sep. Purif. Technol. 290, 120875 (2022). https://doi.org/10.1016/j.seppur.2022.120875
J. Chen, M. Qin, S. Ma, R. Fan, X. Zheng et al., Rational construction of Pt/PtTex interface with optimal intermediate adsorption energy for efficient hydrogen evolution reaction. Appl. Catal. B-Environ. 299, 120640 (2021). https://doi.org/10.1016/j.apcatb.2021.120640
G.L.W. Hart, T. Mueller, C. Toher, S. Curtarolo, Machine learning for alloys. Nat. Rev. Mater. 6(8), 730–755 (2021). https://doi.org/10.1038/s41578-021-00340-w
N. Artrith, K.T. Butler, F.X. Coudert, S. Han, O. Isayev et al., Best practices in machine learning for chemistry. Nat. Chem. 13(6), 505–508 (2021). https://doi.org/10.1038/s41557-021-00716-z
J. Peng, D. Schwalbe-Koda, K. Akkiraju, T. Xie, L. Giordano et al., Human- and machine-centred designs of molecules and materials for sustainability and decarbonization. Nat. Rev. Mater. 7(12), 991–1009 (2022). https://doi.org/10.1038/s41578-022-00466-5
C.H. Chan, M. Sun, B. Huang, Application of machine learning for advanced material prediction and design. EcoMat 4(4), 12194 (2022). https://doi.org/10.1002/eom2.12194
P. Riley, Three pitfalls to avoid in machine learning. Nature 572(7767), 27–29 (2019). https://doi.org/10.1038/d41586-019-02307-y
Z.H. Zhou, Open-environment machine learning. Nat. Sci. Rev. 9(8), 123 (2022). https://doi.org/10.1093/nsr/nwac123
Z. Yao, Y. Lum, A. Johnston, L.M. Mejia-Mendoza, X. Zhou et al., Machine learning for a sustainable energy future. Nat. Rev. Mater. 8, 202–215 (2023). https://doi.org/10.1038/s41578-022-00490-5
T. Liu, X. Zhao, X. Liu, W. Xiao, Z. Luo et al., Understanding the hydrogen evolution reaction activity of doped single-atom catalysts on two-dimensional GaPS4 by DFT and machine learning. J. Energy Chem. 81, 93–100 (2023). https://doi.org/10.1016/j.jechem.2023.02.018
R.B. Wexler, J.M.P. Martirez, A.M. Rappe, Chemical pressure-driven enhancement of the hydrogen evolving activity of Ni2P from nonmetal surface doping interpreted via machine learning. J. Am. Chem. Soc. 140(13), 4678–4683 (2018). https://doi.org/10.1021/jacs.8b00947
N.K. Pandit, D. Roy, S.C. Mandal, B. Pathak, Rational designing of bimetallic/trimetallic hydrogen evolution reaction catalysts using supervised machine learning. J. Phys. Chem. Lett. 13(32), 7583–7593 (2022). https://doi.org/10.1021/acs.jpclett.2c01401
A.J. Parker, G. Opletal, A.S. Barnard, Classification of platinum nanoparticle catalysts using machine learning. J. Appl. Phys. 128(1), 14301 (2020). https://doi.org/10.1063/5.0009129
J. Kim, D. Kang, S. Kim, H.W. Jang, Catalyze materials science with machine learning. ACS Mater. Lett. 3(8), 1151–1171 (2021). https://doi.org/10.1021/acsmaterialslett.1c00204
R. Ding, Y. Ding, H. Zhang, R. Wang, Z. Xu et al., Applying machine learning to boost the development of high-performance membrane electrode assembly for proton exchange membrane fuel cells. J. Mater. Chem. A 9(11), 6841–6850 (2021). https://doi.org/10.1039/d0ta12571g
Y. Jia, X. Hou, Z. Wang, X. Hu, Machine learning boosts the design and discovery of nanomaterials. ACS Sustain. Chem. Eng. 9(18), 6130–6147 (2021). https://doi.org/10.1021/acssuschemeng.1c00483
J. Thiyagalingam, M. Shankar, G. Fox, T. Hey, Scientific machine learning benchmarks. Nat. Rev. Phys. 4(6), 413–420 (2022). https://doi.org/10.1038/s42254-022-00441-7
X. Zhang, Y. Tian, L. Chen, X. Hu, Z. Zhou, Machine learning: a new paradigm in computational electrocatalysis. J. Phys. Chem. Lett. 13(34), 7920–7930 (2022). https://doi.org/10.1021/acs.jpclett.2c01710
R. Pederson, B. Kalita, K. Burke, Machine learning and density functional theory. Nat. Rev. Phys. 4(6), 357–358 (2022). https://doi.org/10.1038/s42254-022-00470-2
J.A. Keith, V. Vassilev-Galindo, B. Cheng, S. Chmiela, M. Gastegger et al., Combining machine learning and computational chemistry for predictive insights into chemical systems. Chem. Rev. 121(16), 9816–9872 (2021). https://doi.org/10.1021/acs.chemrev.1c00107
O.A. von Lilienfeld, K.-R. Müller, A. Tkatchenko, Exploring chemical compound space with quantum-based machine learning. Nat. Rev. Chem. 4(7), 347–358 (2020). https://doi.org/10.1038/s41570-020-0189-9
F. Strieth-Kalthoff, F. Sandfort, M.H.S. Segler, F. Glorius, Machine learning the ropes: principles, applications and directions in synthetic chemistry. Chem. Soc. Rev. 49(17), 6154–6168 (2020). https://doi.org/10.1039/c9cs00786e
Y. Chen, S. Wang, J. Xiong, G. Wu, J. Gao et al., Identifying facile material descriptors for charpy impact toughness in low-alloy steel via machine learning. J. Mater. Sci. Technol. 132, 213–222 (2023). https://doi.org/10.1016/j.jmst.2022.05.051
Y. Liu, O.C. Esan, Z. Pan, L. An, Machine learning for advanced energy materials. Energy and AI 3, 100049 (2021). https://doi.org/10.1016/j.egyai.2021.100049
K. Gulati, S. Saravana Kumar, R. Sarath Kumar Boddu, K. Sarvakar, D. Kumar Sharma et al., Comparative analysis of machine learning-based classification models using sentiment classification of tweets related to covid-19 pandemic. Mater. Today: Proc. 51, 38–41 (2022). https://doi.org/10.1016/j.matpr.2021.04.364
X. Li, S. Yi, A.B. Cundy, W. Chen, Sustainable decision-making for contaminated site risk management: a decision tree model using machine learning algorithms. J. Clean Prod. 371, 133612 (2022). https://doi.org/10.1016/j.jclepro.2022.133612
Y. Lai, Z. Zhang, Z. Zhang, Y. Tan, L. Yu et al., Electronic modulation of Pt nanoclusters through tuning the interface of Pt-SnO2 clusters for enhanced hydrogen evolution catalysis. Chem. Eng. J. 435, 135102 (2022). https://doi.org/10.1016/j.cej.2022.135102
M. Tamtaji, H. Gao, M.D. Hossain, P.R. Galligan, H. Wong et al., Machine learning for design principles for single atom catalysts towards electrochemical reactions. J. Mater. Chem. A 10(29), 15309–15331 (2022). https://doi.org/10.1039/d2ta02039d
Z. Xu, G. Zhu, N. Metawa, Q. Zhou, Machine learning based customer meta-combination brand equity analysis for marketing behavior evaluation. Inform. Process Manag. 59(1), 102800 (2022). https://doi.org/10.1016/j.ipm.2021.102800
Y. Ding, L. Fan, X. Liu, Analysis of feature matrix in machine learning algorithms to predict energy consumption of public buildings. Energy Build. 249, 111208 (2021). https://doi.org/10.1016/j.enbuild.2021.111208
J. Li, L. Liu, T.D. Le, J. Liu, Accurate data-driven prediction does not mean high reproducibility. Nat. Mach. Intell. 2(1), 13–15 (2020). https://doi.org/10.1038/s42256-019-0140-2
M. Meuwly, Machine learning for chemical reactions. Chem. Rev. 121(16), 10218–10239 (2021). https://doi.org/10.1021/acs.chemrev.1c00033
W. Sha, Y. Li, S. Tang, J. Tian, Y. Zhao et al., Machine learning in polymer informatics. InfoMat 3(4), 353–361 (2021). https://doi.org/10.1002/inf2.12167
A. Bender, N. Schneider, M. Segler, W. Patrick Walters, O. Engkvist et al., Evaluation guidelines for machine learning tools in the chemical sciences. Nat. Rev. Chem. 6(6), 428–442 (2022). https://doi.org/10.1038/s41570-022-00391-9
C. Gao, X. Min, M. Fang, T. Tao, X. Zheng et al., Innovative materials science via machine learning. Adv. Funct. Mater. 32(1), 2108044 (2022). https://doi.org/10.1002/adfm.202108044
V. Venkatraman, P.A. Carvalho, On the value of popular crystallographic databases for machine learning prediction of space groups. Acta Mater. 240, 118353 (2022). https://doi.org/10.1016/j.actamat.2022.118353
R.M. Geilhufe, B. Olsthoorn, A.V. Balatsky, Shifting computational boundaries for complex organic materials. Nat. Phys. 17(2), 152–154 (2021). https://doi.org/10.1038/s41567-020-01135-6
Y. Iwasaki, R. Sawada, E. Saitoh, M. Ishida, Machine learning autonomous identification of magnetic alloys beyond the slater-pauling limit. Commun. Mater. 2(1), 31 (2021). https://doi.org/10.1038/s43246-021-00135-0
Q. Tao, P. Xu, M. Li, W. Lu, Machine learning for perovskite materials design and discovery. npj Comput. Mater. 7(1), 23 (2021). https://doi.org/10.1038/s41524-021-00495-8
A.S. Rosen, S.M. Iyer, D. Ray, Z. Yao, A. Aspuru-Guzik et al., Machine learning the quantum-chemical properties of metal-organic frameworks for accelerated materials discovery. Matter 4(5), 1578–1597 (2021). https://doi.org/10.1016/j.matt.2021.02.015
N. Zhang, B. Yang, K. Liu, H. Li, G. Chen et al., Machine learning in screening high performance electrocatalysts for CO2 reduction. Small Methods 5(11), 2100987 (2021). https://doi.org/10.1002/smtd.202100987
J. Cai, X. Chu, K. Xu, H. Li, J. Wei, Machine learning-driven new material discovery. Nanoscale Adv. 2(8), 3115–3130 (2020). https://doi.org/10.1039/d0na00388c
A.A.B. Baloch, S.M. Alqahtani, F. Mumtaz, A.H. Muqaibel, S.N. Rashkeev et al., Extending shannon’s ionic radii database using machine learning. Phys. Rev. Mater. 5(4), 043804 (2021). https://doi.org/10.1103/PhysRevMaterials.5.043804
P. Nguyen, D. Loveland, J.T. Kim, P. Karande, A.M. Hiszpanski et al., Predicting energetics materials’ crystalline density from chemical structure by machine learning. J. Chem. Inf. Model 61(5), 2147–2158 (2021). https://doi.org/10.1021/acs.jcim.0c01318
C. Backes, S. Bartus, C. Casiraghi, A. Ferrari, A.R. Kamali et al., Applications in opto-electronics: general discussion. Faraday Discuss. 227, 184–188 (2021). https://doi.org/10.1039/d1fd90006d
K.T. Winther, M.J. Hoffmann, J.R. Boes, O. Mamun, M. Bajdich et al., Catalysis-hub.Org, an open electronic structure database for surface reactions. Sci. Data 6(1), 75 (2019). https://doi.org/10.1038/s41597-019-0081-y
E. Komp, N. Janulaitis, S. Valleau, Progress towards machine learning reaction rate constants. Phys. Chem. Chem. Phys. 24(5), 2692–2705 (2022). https://doi.org/10.1039/d1cp04422b
J. Deng, J. Pan, Y.F. Zhang, S. Du, Database construction of two-dimensional charged building blocks for functional-oriented material design. Nano Lett. 23(10), 4634–4641 (2023). https://doi.org/10.1021/acs.nanolett.3c01237
A. Tropsha, K.C. Mills, A.J. Hickey, Reproducibility, sharing and progress in nanomaterial databases. Nature nanotech. 12(12), 1111–1114 (2017). https://doi.org/10.1038/nnano.2017.233
F.A. Rasmussen, K.S. Thygesen, Computational 2D materials database: electronic structure of transition-metal dichalcogenides and oxides. J. Phys. Chem. C 119(23), 13169–13183 (2015). https://doi.org/10.1021/acs.jpcc.5b02950
H. Lu, X. Ma, Hybrid decision tree-based machine learning models for short-term water quality prediction. Chemosphere 249, 126169 (2020). https://doi.org/10.1016/j.chemosphere.2020.126169
A. Mahmood, J.-L. Wang, Machine learning for high performance organic solar cells: current scenario and future prospects. Energy Environ. Sci. 14(1), 90–105 (2021). https://doi.org/10.1039/d0ee02838j
Y. Zhang, P.J. Thorburn, Handling missing data in near real-time environmental monitoring: a system and a review of selected methods. Future Gener. Comp. Sy. 128, 63–72 (2022). https://doi.org/10.1016/j.future.2021.09.033
S.K. Baduge, S. Thilakarathna, J.S. Perera, M. Arashpour, P. Sharafi et al., Artificial intelligence and smart vision for building and construction 4.0: machine and deep learning methods and applications. Automat. Constr. 141, 104440 (2022). https://doi.org/10.1016/j.autcon.2022.104440
A. Turetskyy, J. Wessel, C. Herrmann, S. Thiede, Battery production design using multi-output machine learning models. Energy Storage Mater. 38, 93–112 (2021). https://doi.org/10.1016/j.ensm.2021.03.002
S. Guo, J. Popp, T. Bocklitz, Chemometric analysis in Raman spectroscopy from experimental design to machine learning-based modeling. Nat. Protoc. 16(12), 5426–5459 (2021). https://doi.org/10.1038/s41596-021-00620-3
W. Zhao, H. Zhang, J. Zheng, Y. Dai, L. Huang et al., A point prediction method based automatic machine learning for day-ahead power output of multi-region photovoltaic plants. Energy 223, 120026 (2021). https://doi.org/10.1016/j.energy.2021.120026
Z. Yao, Y. Lum, A. Johnston, L.M. Mejia-Mendoza, X. Zhou et al., Machine learning for a sustainable energy future. Nat. Rev. Mater. 8(3), 202–215 (2023). https://doi.org/10.1038/s41578-022-00490-5
D. Roman, S. Saxena, V. Robu, M. Pecht, D. Flynn, Machine learning pipeline for battery state-of-health estimation. Nat. Mach. Intell. 3(5), 447–456 (2021). https://doi.org/10.1038/s42256-021-00312-3
B. Wang, F. Zhang, Main descriptors to correlate structures with the performances of electrocatalysts. Angew. Chem. Int. Ed. 61(4), 2111026 (2022). https://doi.org/10.1002/anie.202111026
M. Ducamp, F.-X. Coudert, Prediction of thermal properties of zeolites through machine learning. J. Phys. Chem. C 126(3), 1651–1660 (2022). https://doi.org/10.1021/acs.jpcc.1c09737
S. Liu, B.B. Kappes, B. Amin-ahmadi, O. Benafan, X. Zhang et al., Physics-informed machine learning for composition-process-property design: shape memory alloy demonstration. Appl. Mater. Today 22, 100898 (2021). https://doi.org/10.1016/j.apmt.2020.100898
L. Zhang, J. Wen, Y. Li, J. Chen, Y. Ye et al., A review of machine learning in building load prediction. Appl. Energy 285, 116452 (2021). https://doi.org/10.1016/j.apenergy.2021.116452
W. Li, N. Sengupta, P. Dechent, D. Howey, A. Annaswamy et al., Online capacity estimation of lithium-ion batteries with deep long short-term memory networks. J. Power Sources 482, 228863 (2021). https://doi.org/10.1016/j.jpowsour.2020.228863
D.P. Finegan, J. Zhu, X. Feng, M. Keyser, M. Ulmefors et al., The application of data-driven methods and physics-based learning for improving battery safety. Joule 5(2), 316–329 (2021). https://doi.org/10.1016/j.joule.2020.11.018
J. Westermayr, M. Gastegger, M. Menger, S. Mai, L. Gonzalez et al., Machine learning enables long time scale molecular photodynamics simulations. Chem. Sci. 10(35), 8100–8107 (2019). https://doi.org/10.1039/c9sc01742a
Y. Xu, D. Verma, R.P. Sheridan, A. Liaw, J. Ma et al., Deep dive into machine learning models for protein engineering. J. Chem. Inf. Model. 60(6), 2773–2790 (2020). https://doi.org/10.1021/acs.jcim.0c00073
A. Harari, G. Katz, Automatic features generation and selection from external sources: a DBpedia use case. Inform. Sci. 582, 398–414 (2022). https://doi.org/10.1016/j.ins.2021.09.036
J. Xiong, S.-Q. Shi, T.-Y. Zhang, Machine learning of phases and mechanical properties in complex concentrated alloys. J. Mater. Sci. Technol. 87, 133–142 (2021). https://doi.org/10.1016/j.jmst.2021.01.054
C. Janiesch, P. Zschech, K. Heinrich, Machine learning and deep learning. Electron. Mark. 31(3), 685–695 (2021). https://doi.org/10.1007/s12525-021-00475-2
B.B. Gupta, K. Yadav, I. Razzak, K. Psannis, A. Castiglione et al., A novel approach for phishing urls detection using lexical based machine learning in a real-time environment. Comput. Commun. 175, 47–57 (2021). https://doi.org/10.1016/j.comcom.2021.04.023
D.S. Wigh, J.M. Goodman, A.A. Lapkin, A review of molecular representation in the age of machine learning. WIREs Comput. Mol. Sci. 12(5), 1603 (2022). https://doi.org/10.1002/wcms.1603
G.G.C. Peterson, J. Brgoch, Materials discovery through machine learning formation energy. J. Phys. Energy 3(2), 022002 (2021). https://doi.org/10.1088/2515-7655/abe425
S.Y. Louis, Y. Zhao, A. Nasiri, X. Wang, Y. Song et al., Graph convolutional neural networks with global attention for improved materials property prediction. Phys. Chem. Chem. Phys. 22(32), 18141–18148 (2020). https://doi.org/10.1039/d0cp01474e
M. Sun, T. Wu, A.W. Dougherty, M. Lam, B. Huang et al., Self-validated machine learning study of graphdiyne-based dual atomic catalyst. Adv. Energy Mater. 11(13), 2003796 (2021). https://doi.org/10.1002/aenm.202003796
K. Fukami, R. Maulik, N. Ramachandra, K. Fukagata, K. Taira, Global field reconstruction from sparse sensors with voronoi tessellation-assisted deep learning. Nat. Mach. Intell. 3(11), 945–951 (2021). https://doi.org/10.1038/s42256-021-00402-2
H. Mai, T.C. Le, D. Chen, D.A. Winkler, R.A. Caruso, Machine learning for electrocatalyst and photocatalyst design and discovery. Chem. Rev. 122(16), 13478–13515 (2022). https://doi.org/10.1021/acs.chemrev.2c00061
A. Mosallanezhad, C. Wei, P. Ahmadian Koudakan, Y. Fang, S. Niu et al., Interfacial synergies between single-atomic Pt and CoS for enhancing hydrogen evolution reaction catalysis. Appl. Catal. B-Environ. 315, 121534 (2022). https://doi.org/10.1016/j.apcatb.2022.121534
R. Ding, Y. Chen, P. Chen, R. Wang, J. Wang et al., Machine learning-guided discovery of underlying decisive factors and new mechanisms for the design of nonprecious metal electrocatalysts. ACS Catal. 11(15), 9798–9808 (2021). https://doi.org/10.1021/acscatal.1c01473
M. Zhong, K. Tran, Y. Min, C. Wang, Z. Wang et al., Accelerated discovery of CO2 electrocatalysts using active machine learning. Nature 581(7807), 178–183 (2020). https://doi.org/10.1038/s41586-020-2242-8
L. Chen, X. Zhang, A. Chen, S. Yao, X. Hu et al., Targeted design of advanced electrocatalysts by machine learning. Chinese J. Catal. 43(1), 11–32 (2022). https://doi.org/10.1016/s1872-2067(21)63852-4
M. Umer, S. Umer, M. Zafari, M. Ha, R. Anand et al., Machine learning assisted high-throughput screening of transition metal single atom based superb hydrogen evolution electrocatalysts. J. Mater. Chem. A 10(12), 6679–6689 (2022). https://doi.org/10.1039/d1ta09878k
L. Giordano, K. Akkiraju, R. Jacobs, D. Vivona, D. Morgan et al., Electronic structure-based descriptors for oxide properties and functions. Acc. Chem. Res. 55(3), 298–308 (2022). https://doi.org/10.1021/acs.accounts.1c00509
L. Chen, Y. Tian, X. Hu, S. Yao, Z. Lu et al., A universal machine learning framework for electrocatalyst innovation: a case study of discovering alloys for hydrogen evolution reaction. Adv. Funct. Mater. 32(47), 2208418 (2022). https://doi.org/10.1002/adfm.202208418
M.O.J. Jäger, E.V. Morooka, F. Federici Canova, L. Himanen, A.S. Foster, Machine learning hydrogen adsorption on nanoclusters through structural descriptors. npj Comput. Mater. 4(1), 37 (2018). https://doi.org/10.1038/s41524-018-0096-5
S. Jiao, X. Fu, H. Huang, Descriptors for the evaluation of electrocatalytic reactions: D-band theory and beyond. Adv. Funct. Mater. 32(4), 2107651 (2021). https://doi.org/10.1002/adfm.202107651
S. Lu, J. Cao, Y. Zhang, F. Lou, Z. Yu, Transition metal single-atom supported on PC3 monolayer for highly efficient hydrogen evolution reaction by combined density functional theory and machine learning study. Appl. Surf. Sci. 606, 154945 (2022). https://doi.org/10.1016/j.apsusc.2022.154945
C. Zou, J. Li, W.Y. Wang, Y. Zhang, D. Lin et al., Integrating data mining and machine learning to discover high-strength ductile titanium alloys. Acta Mater. 202, 211–221 (2021). https://doi.org/10.1016/j.actamat.2020.10.056
P.O. Dral, M. Barbatti, Molecular excited states through a machine learning lens. Nat. Rev. Chem. 5(6), 388–405 (2021). https://doi.org/10.1038/s41570-021-00278-1
C. Altintas, O.F. Altundal, S. Keskin, R. Yildirim, Machine learning meets with metal organic frameworks for gas storage and separation. J. Chem. Inf. Model. 61(5), 2131–2146 (2021). https://doi.org/10.1021/acs.jcim.1c00191
A. Nandy, C. Duan, M.G. Taylor, F. Liu, A.H. Steeves et al., Computational discovery of transition-metal complexes: from high-throughput screening to machine learning. Chem. Rev. 121(16), 9927–10000 (2021). https://doi.org/10.1021/acs.chemrev.1c00347
C. Lv, X. Zhou, L. Zhong, C. Yan, M. Srinivasan et al., Machine learning: an advanced platform for materials development and state prediction in lithium-ion batteries. Adv. Mater. 34(25), 2101474 (2022). https://doi.org/10.1002/adma.202101474
D. Lemm, G.F. von Rudorff, O.A. von Lilienfeld, Machine learning based energy-free structure predictions of molecules, transition states, and solids. Nat. Commun. 12(1), 4468 (2021). https://doi.org/10.1038/s41467-021-24525-7
Y. Guan, C.W. Coley, H. Wu, D. Ranasinghe, E. Heid et al., Regio-selectivity prediction with a machine-learned reaction representation and on-the-fly quantum mechanical descriptors. Chem. Sci. 12(6), 2198–2208 (2020). https://doi.org/10.1039/d0sc04823b
K. Choudhary, B. DeCost, C. Chen, A. Jain, F. Tavazza et al., Recent advances and applications of deep learning methods in materials science. npj Comput. Mater. 8(1), 59 (2022). https://doi.org/10.1038/s41524-022-00734-6
T.H. Pham, Y. Qiu, J. Zeng, L. Xie, P. Zhang, A deep learning framework for high-throughput mechanism-driven phenotype compound screening and its application to covid-19 drug repurposing. Nat. Mach. Intell. 3(3), 247–257 (2021). https://doi.org/10.1038/s42256-020-00285-9
J. Tian, R. Xiong, W. Shen, J. Lu, F. Sun, Flexible battery state of health and state of charge estimation using partial charging data and deep learning. Energy Storage Mater. 51, 372–381 (2022). https://doi.org/10.1016/j.ensm.2022.06.053
Y. Li, C. Huang, L. Ding, Z. Li, Y. Pan et al., Deep learning in bioinformatics: introduction, application, and perspective in the big data era. Methods 166, 4–21 (2019). https://doi.org/10.1016/j.ymeth.2019.04.008
Y. Ji, S. Ma, S. Lv, Y. Wang, S. Lu et al., Nanomaterials for targeted delivery of agrochemicals by an all-in-one combination strategy and deep learning. ACS Appl. Mater. Interfaces 13(36), 43374–43386 (2021). https://doi.org/10.1021/acsami.1c11914
H. Masood, C.Y. Toe, W.Y. Teoh, V. Sethu, R. Amal, Machine learning for accelerated discovery of solar photocatalysts. ACS Catal. 9(12), 11774–11787 (2019). https://doi.org/10.1021/acscatal.9b02531
G. Pan, F. Wang, C. Shang, H.-H. Wu, G. Wu et al., Advances in machine learning- and artificial intelligence-assisted material design of steels. Int. J. Min. Met. Mater. 30, 1003–1024 (2023). https://doi.org/10.1007/s12613-022-2595-0
S. Manna, D. Roy, S. Das, B. Pathak, Capacity prediction of K-ion batteries: a machine learning based approach for high throughput screening of electrode materials. Mater. Adv. 3(21), 7833–7845 (2022). https://doi.org/10.1039/d2ma00746k
T. Gao, W. Lu, Machine learning toward advanced energy storage devices and systems. iScience 24(1), 101936 (2021). https://doi.org/10.1016/j.isci.2020.101936
Y. Guan, D. Chaffart, G. Liu, Z. Tan, D. Zhang et al., Machine learning in solid heterogeneous catalysis: recent developments, challenges and perspectives. Chem. Eng. Sci. 248, 117224 (2022). https://doi.org/10.1016/j.ces.2021.117224
P.S. Rice, Z.P. Liu, P. Hu, Hydrogen coupling on platinum using artificial neural network potentials and DFT. J. Phys. Chem. Lett. 12(43), 10637–10645 (2021). https://doi.org/10.1021/acs.jpclett.1c02998
H.-H. Wu, L.-S. Dong, S.-Z. Wang, G.-L. Wu, J.-H. Gao et al., Local chemical ordering coordinated thermal stability of nanograined high-entropy alloys. Rare Met. 42(5), 1645–1655 (2022). https://doi.org/10.1007/s12598-022-02194-9
F. Wang, H.-H. Wu, L. Dong, G. Pan, X. Zhou et al., Atomic-scale simulations in multi-component alloys and compounds: a review on advances in interatomic potential. J. Mater. Sci. Technol. 165, 49–65 (2023). https://doi.org/10.1016/j.jmst.2023.05.010
S. Mehrabi-Kalajahi, A.O. Moghaddam, F. Hadavimoghaddam, M.A. Varfolomeev, A.L. Zinnatullin et al., Entropy-stabilized metal oxide nanoparticles supported on reduced graphene oxide as a highly active heterogeneous catalyst for selective and solvent-free oxidation of toluene: a combined experimental and numerical investigation. J. Mater. Chem. A 10(27), 14488–14500 (2022). https://doi.org/10.1039/d2ta02027k
Y. Zhang, T.C. Peck, G.K. Reddy, D. Banerjee, H. Jia et al., Descriptor-free design of multicomponent catalysts. ACS Catal. 12(17), 10562–10571 (2022). https://doi.org/10.1021/acscatal.2c02807
K.E. McCullough, D.S. King, S.P. Chheda, M.S. Ferrandon, T.A. Goetjen et al., High-throughput experimentation, theoretical modeling, and human intuition: lessons learned in metal-organic-framework-supported catalyst design. ACS Cent. Sci. 9(2), 266–276 (2023). https://doi.org/10.1021/acscentsci.2c01422
A. Chen, J. Cai, Z. Wang, Y. Han, S. Ye et al., An ensemble learning classifier to discover arsenene catalysts with implanted heteroatoms for hydrogen evolution reaction. J. Energy Chem. 78, 268–276 (2023). https://doi.org/10.1016/j.jechem.2022.11.035
Z. Yang, W. Gao, Applications of machine learning in alloy catalysts: rational selection and future development of descriptors. Adv. Sci. 9(12), 2106043 (2022). https://doi.org/10.1002/advs.202106043
Z. Xiong, Y. Cui, Z. Liu, Y. Zhao, M. Hu et al., Evaluating explorative prediction power of machine learning algorithms for materials discovery using k-fold forward cross-validation. Comput. Mater. Sci. 171, 109203 (2020). https://doi.org/10.1016/j.commatsci.2019.109203
D. Perez-Guaita, G. Quintas, J. Kuligowski, Discriminant analysis and feature selection in mass spectrometry imaging using constrained repeated random sampling-cross validation (corrs-cv). Anal. Chim. Acta 1097, 30–36 (2020). https://doi.org/10.1016/j.aca.2019.10.039
S.Y. Ho, K. Phua, L. Wong, W.W. Bin Goh, Extensions of the external validation for checking learned model interpretability and generalizability. Patterns 1(8), 100129 (2020). https://doi.org/10.1016/j.patter.2020.100129
N.O. Dogan, U. Bozuyuk, P. Erkoc, A.C. Karacakol, A. Cingoz et al., Parameters influencing gene delivery efficiency of pegylated chitosan nanoparticles: experimental and modeling approach. Adv. NanoBiomed Res. 2(1), 2100033 (2021). https://doi.org/10.1002/anbr.202100033
M. Karthikeyan, D.M. Mahapatra, A.S.A. Razak, A.A.M. Abahussain, B. Ethiraj et al., Machine learning aided synthesis and screening of HER catalyst: present developments and prospects. Catal. Rev. (2022). https://doi.org/10.1080/01614940.2022.2103980
D. Sui, M. Yao, L. Si, K. Yan, J. Shi et al., Biomass-derived carbon coated SiO2 nanotubes as superior anode for lithium-ion batteries. Carbon 205, 510–518 (2023). https://doi.org/10.1016/j.carbon.2023.01.039
G. Liu, T. Zhang, X. Li, J. Li, N. Wu et al., MoS2@C with S vacancies vertically anchored onV2C-MXene for efficient lithium and sodium storage. Inorg. Chem. Front. 10(5), 1587–1602 (2023). https://doi.org/10.1039/d2qi02389j
Y. Liu, P. Vijayakumar, Q. Liu, T. Sakthivel, F. Chen et al., Shining light on anion-mixed nanocatalysts for efficient water electrolysis: fundamentals, progress, and perspectives. Nano-Micro Lett. 14(1), 43 (2022). https://doi.org/10.1007/s40820-021-00785-2
C. Wang, X. Wei, D. Ren, X. Wang, W. Xu, High-throughput map design of creep life in low-alloy steels by integrating machine learning with a genetic algorithm. Mater. Design 213, 110326 (2022). https://doi.org/10.1016/j.matdes.2021.110326
Z. Liu, S. Ha, Y. Liu, F. Wang, F. Tao et al., Application of Ag-based materials in high-performance lithium metal anode: a review. J. Mater. Sci. Technol. 133, 165–182 (2023). https://doi.org/10.1016/j.jmst.2022.06.015
X. Zhao, H. Qiu, Y. Shao, P. Wang, S. Yu et al., Silver nanoparticle-modified 2D MOF nanosheets for photothermally enhanced silver ion release antibacterial treatment. Acta Phys-Chim. Sin. 1, 2211043 (2023). https://doi.org/10.3866/pku.whxb202211043
L. Wu, T. Guo, T. Li, Machine learning-accelerated prediction of overpotential of oxygen evolution reaction of single-atom catalysts. iScience 24(5), 102398 (2021). https://doi.org/10.1016/j.isci.2021.102398
M. Kim, M.Y. Ha, W.B. Jung, J. Yoon, E. Shin et al., Searching for an optimal multi-metallic alloy catalyst by active learning combined with experiments. Adv. Mater. 34(19), 2108900 (2022). https://doi.org/10.1002/adma.202108900
Y.L. Wu, X. Li, Y.S. Wei, Z. Fu, W. Wei et al., Ordered macroporous superstructure of nitrogen-doped nanoporous carbon implanted with ultrafine Ru nanoclusters for efficient pH-universal hydrogen evolution reaction. Adv. Mater. 33(12), 2006965 (2021). https://doi.org/10.1002/adma.202006965
Y. Xu, R. Wang, J. Wang, J. Li, T. Jiao et al., Facile fabrication of molybdenum compounds (Mo2C, MoP and MoS2) nanoclusters supported on N-doped reduced graphene oxide for highly efficient hydrogen evolution reaction over broad pH range. Chem. Eng. J. 417, 129233 (2021). https://doi.org/10.1016/j.cej.2021.129233
Y. Li, Y. Luo, Z. Zhang, Q. Yu, C. Li et al., Implanting Ru nanoclusters into N-doped graphene for efficient alkaline hydrogen evolution. Carbon 183, 362–367 (2021). https://doi.org/10.1016/j.carbon.2021.07.039
C. Li, H. Jang, S. Liu, M.G. Kim, L. Hou et al., P and Mo dual doped Ru ultrasmall nanoclusters embedded in P-doped porous carbon toward efficient hydrogen evolution reaction. Adv. Energy Mater. 12(23), 2200029 (2022). https://doi.org/10.1002/aenm.202200029
M.O.J. Jager, Y.S. Ranawat, F.F. Canova, E.V. Morooka, A.S. Foster, Efficient machine-learning-aided screening of hydrogen adsorption on bimetallic nanoclusters. ACS Comb. Sci. 22(12), 768–781 (2020). https://doi.org/10.1021/acscombsci.0c00102
D.X. Zhu, K.M. Pan, Y. Wu, X.Y. Zhou, X.Y. Li et al., Improved material descriptors for bulk modulus in intermetallic compounds via machine learning. Rare Met. 42, 2396–2405 (2023). https://doi.org/10.1007/s12598-023-02282-4
X. Mao, L. Wang, Y. Xu, P. Wang, Y. Li et al., Computational high-throughput screening of alloy nanoclusters for electrocatalytic hydrogen evolution. npj Comput. Mater. 7(1), 46 (2021). https://doi.org/10.1038/s41524-021-00514-8
Q. Yang, H. Liu, P. Yuan, Y. Jia, L. Zhuang et al., Single carbon vacancy traps atomic platinum for hydrogen evolution catalysis. J. Am. Chem. Soc. 144(5), 2171–2178 (2022). https://doi.org/10.1021/jacs.1c10814
H. Prats, K. Chan, The determination of the HOR/HER reaction mechanism from experimental kinetic data. Phys. Chem. Chem. Phys. 23(48), 27150–27158 (2021). https://doi.org/10.1039/d1cp04134g
A. Gross, S. Sakong, Ab initio simulations of water/metal interfaces. Chem. Rev. 122(12), 10746–10776 (2022). https://doi.org/10.1021/acs.chemrev.1c00679
S.M.N. Jeghan, D. Kim, Y. Lee, M. Kim, G. Lee, Designing a smart heterojunction coupling of cobalt-iron layered double hydroxide on nickel selenide nanosheets for highly efficient overall water splitting kinetics. Appl. Catal. B-Environ. 308, 121221 (2022). https://doi.org/10.1016/j.apcatb.2022.121221
G.H. Gu, J. Lim, C. Wan, T. Cheng, H. Pu et al., Autobifunctional mechanism of jagged Pt nanowires for hydrogen evolution kinetics via end-to-end simulation. J. Am. Chem. Soc. 143(14), 5355–5363 (2021). https://doi.org/10.1021/jacs.0c11261
H. Jin, H. Yu, H. Li, K. Davey, T. Song et al., MXene analogue: a 2D nitridene solid solution for high-rate hydrogen production. Angew. Chem. Int. Ed. 61(27), 2203850 (2022). https://doi.org/10.1002/anie.202203850
R. Kronberg, H. Lappalainen, K. Laasonen, Hydrogen adsorption on defective nitrogen-doped carbon nanotubes explained via machine learning augmented DFT calculations and game-theoretic feature attributions. J. Phys. Chem. C 125(29), 15918–15933 (2021). https://doi.org/10.1021/acs.jpcc.1c03858
J. Mao, J. Miao, Y. Lu, Z. Tong, Machine learning of materials design and state prediction for lithium ion batteries. Chinese J. Chem. Eng. 37, 1–11 (2021). https://doi.org/10.1016/j.cjche.2021.04.009
H. Jin, T. Song, U. Paik, S.-Z. Qiao, Metastable two-dimensional materials for electrocatalytic energy conversions. Accounts Mater. Res. 2(7), 559–573 (2021). https://doi.org/10.1021/accountsmr.1c00115
Y. Ding, M. Zeng, Q. Zheng, J. Zhang, D. Xu et al., Bidirectional and reversible tuning of the interlayer spacing of two-dimensional materials. Nat. Commun. 12(1), 5886 (2021). https://doi.org/10.1038/s41467-021-26139-5
H.S. Kim, Computational design of a switchable heterostructure electrocatalyst based on a two-dimensional ferroelectric In2Se3 material for the hydrogen evolution reaction. J. Mater. Chem. A 9(19), 11553–11562 (2021). https://doi.org/10.1039/d0ta09738a
L. Najafi, R. Oropesa-Nunez, S. Bellani, B. Martin-Garcia, L. Pasquale et al., Topochemical transformation of two-dimensional VSe2 into metallic nonlayered VO2 for water splitting reactions in acidic and alkaline media. ACS Nano 16(1), 351–367 (2021). https://doi.org/10.1021/acsnano.1c06662
Y. Cui, X. Guo, J. Zhang, X. Li, X. Zhu et al., Di-defects synergy boost electrocatalysis hydrogen evolution over two-dimensional heterojunctions. Nano Res. 15(1), 677–684 (2021). https://doi.org/10.1007/s12274-021-3545-2
S. Wu, Z. Wang, H. Zhang, J. Cai, J. Li, Deep learning accelerates the discovery of two-dimensional catalysts for hydrogen evolution reaction. Energy Environ. Mater. 1, 1–7 (2022). https://doi.org/10.1002/eem2.12259
H. He, Y. Chen, C. Yang, L. Yang, Q. Jiang et al., Constructing 3D interweaved mxene/graphitic carbon nitride nanosheets/graphene nanoarchitectures for promoted electrocatalytic hydrogen evolution. J. Energy Chem. 67, 483–491 (2022). https://doi.org/10.1016/j.jechem.2021.10.019
C. Wang, X. Wang, T. Zhang, P. Qian, T. Lookman et al., A descriptor for the design of 2D MXene hydrogen evolution reaction electrocatalysts. J. Mater. Chem. A 10(35), 18195–18205 (2022). https://doi.org/10.1039/d2ta02837a
J. Zheng, X. Sun, C. Qiu, Y. Yan, Z. Yao et al., High-throughput screening of hydrogen evolution reaction catalysts in MXene materials. J. Phys. Chem. C 124(25), 13695–13705 (2020). https://doi.org/10.1021/acs.jpcc.0c02265
X. Sun, J. Zheng, Y. Gao, C. Qiu, Y. Yan et al., Machine-learning-accelerated screening of hydrogen evolution catalysts in Mbenes materials. Appl. Surf. Sci. 526, 146522 (2020). https://doi.org/10.1016/j.apsusc.2020.146522
X. Wang, C. Wang, S. Ci, Y. Ma, T. Liu et al., Accelerating 2D MXene catalyst discovery for the hydrogen evolution reaction by computer-driven workflow and an ensemble learning strategy. J. Mater. Chem. A 8(44), 23488–23497 (2020). https://doi.org/10.1039/d0ta06583h
R. Liu, Z. Gong, J. Liu, J. Dong, J. Liao et al., Design of aligned porous carbon films with single-atom Co-N-C sites for high-current-density hydrogen generation. Adv. Mater. 33(41), 2103533 (2021). https://doi.org/10.1002/adma.202103533
A. Baby, L. Trovato, C. Di Valentin, Single atom catalysts (SAC) trapped in defective and nitrogen-doped graphene supported on metal substrates. Carbon 174, 772–788 (2021). https://doi.org/10.1016/j.carbon.2020.12.045
H. Choi, D.H. Kim, M.H. Han, H.-S. Oh, J. Heo et al., Prediction of the catalytic site of single-atom Ni catalyst using the hydrogen evolution reaction as a model platform. Electrochim. Acta 431, 141138 (2022). https://doi.org/10.1016/j.electacta.2022.141138
C. Rong, X. Shen, Y. Wang, L. Thomsen, T. Zhao et al., Electronic structure engineering of single-atom Ru sites via Co-N4 sites for bifunctional pH-universal water splitting. Adv. Mater. 34(21), 2110103 (2022). https://doi.org/10.1002/adma.202110103
T. Jin, X. Liu, Q. Gao, H. Zhu, C. Lian et al., Pyrolysis-free, facile mechanochemical strategy toward cobalt single-atom/nitrogen-doped carbon for highly efficient water splitting. Chem. Eng. J. 433, 134089 (2022). https://doi.org/10.1016/j.cej.2021.134089
V. Fung, G. Hu, Z. Wu, D.-E. Jiang, Descriptors for hydrogen evolution on single atom catalysts in nitrogen-doped graphene. J. Phys. Chem. C 124(36), 19571–19578 (2020). https://doi.org/10.1021/acs.jpcc.0c04432
S. Lin, H. Xu, Y. Wang, X.C. Zeng, Z. Chen, Directly predicting limiting potentials from easily obtainable physical properties of graphene-supported single-atom electrocatalysts by machine learning. J. Mater. Chem. A 8(11), 5663–5670 (2020). https://doi.org/10.1039/c9ta13404b
M. Sun, A.W. Dougherty, B. Huang, Y. Li, C.H. Yan, Accelerating atomic catalyst discovery by theoretical calculations-machine learning strategy. Adv. Energy Mater. 10(12), 1903949 (2020). https://doi.org/10.1002/aenm.201903949
C. Guo, T. Zhang, X. Lu, C.L. Wu, Rational design and effective control of gold-based bimetallic electrocatalyst for boosting CO2 reduction reaction: a first-principles study. Chemsuschem 14(13), 2731–2739 (2021). https://doi.org/10.1002/cssc.202100785
H. Li, S. Xu, M. Wang, Z. Chen, F. Ji et al., Computational design of (100) alloy surfaces for the hydrogen evolution reaction. J. Mater. Chem. A 8(35), 17987–17997 (2020). https://doi.org/10.1039/d0ta04615a
H. Liu, K. Liu, P. Zhong, J. Qi, J. Bian et al., Ultrathin Pt-Ag alloy nanotubes with regular nanopores for enhanced electrocatalytic activity. Chem. Mater. 30(21), 7744–7751 (2018). https://doi.org/10.1021/acs.chemmater.8b03085
H. Wu, F. Zhuo, H. Qiao, L. Kodumudi Venkataraman, M. Zheng et al., Polymer-/ceramic-based dielectric composites for energy storage and conversion. Energy Environ. Mater. 5(2), 486–514 (2022). https://doi.org/10.1002/eem2.12237
L. Ding, Z. Xie, S. Yu, W. Wang, A.Y. Terekhov et al., Electrochemically grown ultrathin platinum nanosheet electrodes with ultralow loadings for energy-saving and industrial-level hydrogen evolution. Nano-Micro Lett. 15(1), 144 (2023). https://doi.org/10.1007/s40820-023-01117-2
J. Chen, Y. Ha, R. Wang, Y. Liu, H. Xu et al., Inner Co synergizing outer Ru supported on carbon nanotubes for efficient pH-universal hydrogen evolution catalysis. Nano-Micro Lett. 14(1), 186 (2022). https://doi.org/10.1007/s40820-022-00933-2
Z. Wu, T. Liao, S. Wang, J.A. Mudiyanselage, A.S. Micallef et al., Conversion of catalytically inert 2D bismuth oxide nanosheets for effective electrochemical hydrogen evolution reaction catalysis via oxygen vacancy concentration modulation. Nano-Micro Lett. 14(1), 90 (2022). https://doi.org/10.1007/s40820-022-00832-6
Y. Sun, Z. Xue, Q. Liu, Y. Jia, Y. Li et al., Modulating electronic structure of metal-organic frameworks by introducing atomically dispersed Ru for efficient hydrogen evolution. Nat. Commun. 12(1), 1369 (2021). https://doi.org/10.1038/s41467-021-21595-5
Y. Gao, J. Wang, Y. Yang, J. Wang, C. Zhang et al., Engineering spin states of isolated copper species in a metal-organic framework improves urea electrosynthesis. Nano-Micro Lett. 15(1), 158 (2023). https://doi.org/10.1007/s40820-023-01127-0
X. He, Fundamental perspectives on the electrochemical water applications of metal-organic frameworks. Nano-Micro Lett. 15(1), 148 (2023). https://doi.org/10.1007/s40820-023-01124-3
Acknowledgements
This work was supported by the National Natural Science Foundation of China (Grant No. 22008098, 52122408), the Program for Science & Technology Innovation Talents in Universities of Henan Province (No.22HASTIT008), the Programs for Science and Technology Development of Henan Province, China (No.222102320065), the Key Specialized Research and Development Breakthrough (Science and Technology) in Henan Province (No.212102210214), the Natural Science Foundations of Henan Province (No. 222300420502), and the Key Scientific Research Projects of University in Henan Province (No. 23B430002).
Funding
Open access funding provided by Shanghai Jiao Tong University.
Author information
Authors and Affiliations
Corresponding authors
Ethics declarations
Conflict of interest
The authors declare no interest conflict. They have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Li, J., Wu, N., Zhang, J. et al. Machine Learning-Assisted Low-Dimensional Electrocatalysts Design for Hydrogen Evolution Reaction. Nano-Micro Lett. 15, 227 (2023). https://doi.org/10.1007/s40820-023-01192-5
Received:
Accepted:
Published:
DOI: https://doi.org/10.1007/s40820-023-01192-5