As one of the important simulation methods in computational catalysis, molecular dynamics (MD) simulation plays an important role in understanding the catalytic mechanisms and is critical to the design of efficient and stable catalysts. Classical MD simulation with empirical potentials has a high computational efficiency but a limited accuracy, particularly for systems involving chemical reactions, and the accurate first-principle methods suffer from heavy computational costs and become unaffordable in most cases. The existing emerging machine-learning force field (MLFF) method is proven with affordable computational cost and first-principle-level accuracy. MLFF-assisted MD simulation can offer an effective approach for dynamics simulation in nanoscale catalysis. This review represented the fundamental principle of two main MLFF methods, i.e., the Behler-Parrinello atom-centered neural network method and the embedded-network-based deep potential. The applications of MLFF-assisted dynamic studies related to nano-scale catalysis (i.e., structure reconstruction and reaction processes in catalysis) were described. In addition, some possible future challenges of MLFF methods in dynamics simulation were also given.

Materials discovery faces a huge and complex high-dimensional search space, from which the fast and effective selection of new materials with target properties is a major challenge in materials development. Machine learning can predict the performance of unexplored materials via establishing the relationship between features and target performance through algorithms based on the existing data. However, there are a relatively few known data for materials, and the machine learning models have a relatively low prediction accuracy, thus making it difficult to achieve an effective guidance for experiments or calculations. To address this problem, active learning was introduced for assistance, and the experimental design step was added to the traditional iterative feedback to select the experiments for target enhancement to supplement and to achieve the optimization of material performance. This review mainly represented recent progress on active learning-assisted materials development from three aspects, i.e., single-objective optimization, multi-objective optimization, and curve optimization.

Recent development of large-scale atomic simulation techniques based on machine learning has brought a great promise in chemistry. These simulations are featured by both high speed and high accuracy. This review outlined recent development on three key aspects of atomic simulation based on machine learning potential, i.e., machine learning models and structure descriptors, generation of global potential energy surface training sets, and automatic training of potential functions based on active learning. It is indicated that the designed structure descriptor and feedforward neural network model are suitable for generating a highly complex global potential energy surface. In addition, the applications of LASP software in material and reaction simulations were also selected to illustrate how ML-based atomic simulation could assist the discovery of novel materials and reactions.

It is ready to cause the interface delamination, macroscopic fracture and spalling failure of thermal barrier coatings (TBCs) due to their complexity of architecture and harsh service environment. In this paper, the failure process of TBCs under three-point bending (3PB) load was monitored in real time via acoustic emission (AE) technology, and the damage failure modes of TBCs were identified based on micro-morphology analysis of AE parameters and K-means cluster. The waveform characteristics of four failure modes were analyzed by fast Fourier transform and wavelet packet transform. The macroscopic fracture or spalling failure signals have no obvious frequency band, while the corresponding frequency components of substrate deformation, surface vertical crack, sliding interface crack and opening interface crack are 62.5-125.0 kHz, 187.5-250.0 kHz, 250.0-312.5 kHz and 375.0-437.5 kHz, respectively. The method of deep machine learning was used to process the in-situ acoustic emission signals. The wavelet energy coefficient was extracted as a characteristic vector of the Back propagation neural network, and the advantages and disadvantages of the model were evaluated by convergence curve, confusion matrix, Receiver operating characteristic curve and F₁ value, thus realizing the discrimination of failure modes of TBCs under 3PB test and providing a reference value for failure prediction and life assessment of thermal barrier coatings.

Alkali metal ions as some impurity ions in cement raw materials can affect the structure and properties of clinker minerals. The effect of alkali metal ions on the hydration activity and mechanical properties of ternesite was investigated with alkali-doped ternesite synthesized as analytical reagents by isothermal calorimetry, comprehensive thermal analysis, scanning electron microscopy and ²⁹Si solid state nuclear magnetic resonance. The results show that the solid solution of alkali metal ions in the crystal structure of ternesite reduces the crystallinity of formed crystals, forms the crystal defects, effectively improves the early hydration activity of ternesite, and promotes the rapid development of its early mechanical properties. Meanwhile, the addition of alkali metal can change the microstructure and structure of calcium silicate hydrates (C-S-H). Among them, Li₂O addition stabilizes the flocculent C-S-H, while Na₂O and K₂O addition can induce the fibrous growth of C-S-H. The degree of polymerization and mean silicon chain length of C-S-H increase due to the addition of alkali metal.

Low-dimensional iron-based nanomaterials are representative magnetic nanomaterials used in the field of medicine and healthcare. These nanomaterials have attracted recent attention. This review summarized recent research progress on the synthesis, formation mechanism and biomedical applications of high-performance iron-based nanoparticles, and discussed the controllable synthesis as well as themodulation of interface property. In addition, this review also gave some prospects for biomedical applications of typical two-dimensional iron based nanomaterials and composites.

Machine learning has become an important transformative method to explore novel materials, but the small sample size and high noise of material data bring a great challenge to data-driven research and development. To address the challenge, unsupervised learning was applied to discover perovskite materials with a high dielectric constant. Twenty perovskite materials with a high dielectric constant (i.e., BaHfO₃ and BiFeO₃) were screened out via iterative clustering. We performed dimensionality reduction analysis and descriptors analysis including elements, crystal structure and tolerance factors to find the underlying trend and the relationship between ABO₃ structure and dielectric constant. This method can provide an idea for solving the lack of material data labels, which can be also applied to screen other novel functional materials.

With the development of artificial intelligence technology, machine learning atomic interaction potential has become popular to solve a problem regarding the low accuracy of empirical potential. Machine learning atomic interaction potential avoids a low efficiency of conventional fitting method for empirical potential and becomes an emerging tool for material exploration and research. This review represented the characteristics of existing machine learning potential and the applications in phase change, intrinsic properties and interface researches. In addition, the challenge and development trends of machine learning atomic interaction potential were also prospected.

Thermoelectric materials are environmental-friendly energy conversion materials. Their performance optimization is a complex issue of multi-parameter coordination, which becomes a challenge. Although the computational simulation and experimental methods for thermoelectric materials have developed rapidly, the efficiency of searching thermoelectric materials still needs to be further improved. Machine learning has some advantages of low computational cost and high prediction speed, which can shorten the search process and accelerate the corresponding studies on the structure and performance optimization of thermoelectric materials. This review introduced the research progress on machine learning for small sample numerical data (data volume is about 10²), large sample numerical data (data volume >10⁴) and image data in thermoelectric materials from the perspective of data types. Moreover, different machine learning algorithm models used for the structure and performance of thermoelectric materials in different data types were discussed. In addition, the future development and application direction were also prospected.

The energy storage systems are an important basis for electric vehicles and electronic devices. The existing battery design based on machine learning is able to quickly connect the complex relationship among material microstructure, material properties, and battery macroscopic properties. This review represented the applications and prospects of machine learning in micro-material design and state estimation of batteries. The data sources of machine learning battery design, advantages and disadvantages of algorithms and their application scenarios in the battery field, related innovative work in recent years and their prospects were discussed. This review can provide a reference for machine learning in the macro-/micro-design of energy storage systems.