| 摘要: |
| 以层状材料为代表的固体润滑涂层如石墨、二硫化钼（MoS2）等具有极低的摩擦因数和磨损速率，是当前优异的润滑材料。但是，摩擦作为典型的表界面过程，固体润滑涂层的性能在很大程度上会受到所处环境的影响，例如湿度等。由于试验方法在原位实时检测摩擦界面动态演变过程时会遇到极大的技术挑战，理论计算研究在揭示材料润滑行为和机制中起到越来越重要的作用。从经典的理论摩擦分析模型出发，在回顾这些模型构建思路的基础上，总结目前常用的原子级理论研究方法，包括经典分子动力学模拟、第一性原理静态势能面计算以及第一性原理分子动力学（AIMD）模拟。并且强调量子力学方法（第一性原理计算）在探索涉及复杂电子相互作用摩擦问题时不可替代的重要作用，由此基于自由能微分提出一种有效的 AIMD 模拟方法来精确模拟界面滑动过程，从而揭示摩擦性能演变的电子级起源。该方法还可以很好地结合机器学习力场加速，大幅增加模拟尺度并减少模拟时间。研究结果对理论模拟方法的总结和展望将有助于未来更好地探索材料在复杂环境下的微观润滑机制，并指导设计高性能润滑涂层。 |
| 关键词: 润滑 层状材料 环境作用 第一性原理计算 机器学习 |
| DOI：10.11933/j.issn.1007-9289.20240102004 |
| 分类号:TG156；TB114 |
| 基金项目:国家自然科学基金（U21A20127，22272192）；国家重点研发计划（2022YFB3402803）；中国科学院战略性先导科技专项 B 类（XDB0470103） |
|
| Theoretical Research on Solid Lubrication Coatings: From Classical Models to Quantum-mechanical Simulations |
|
HAO Yu，HUANG Liangfeng，WANG Liping
|
|
Key Laboratory of Advanced Marine Materials, Ningbo Institute of Materials Technology and Engineering,Chinese Academy of Sciences, Ningbo 315201 , China
|
| Abstract: |
| Solid lubricating coatings represented by layered materials, such as graphite and molybdenum disulfide (MoS2), have extremely low friction coefficients and wear rates and are currently considered superior lubricating materials in many fields. However, as a typical interfacial process, the performance of solid lubricants is largely dependent on environmental factors such as humidity. Due to the significant technical challenges encountered by experimental methods in real-time and in situ detection of the dynamic evolution process of friction interfaces, theoretical research plays an increasingly important role in revealing material lubrication behavior and mechanisms. Starting with classical friction analytical models, this paper reviews and summarizes commonly used atomic-level theoretical research methods, including classical molecular dynamics (MD) simulations, first-principles static potential energy surface (PES) calculations, and ab initio molecular dynamics (AIMD) simulations. Classical analytical models, such as the Prandtl-Tomlinson and Frenkel-Kontorova-Tomlinson models, are the foundations for understanding the friction behavior of materials. Although these models ignore many realistic factors of materials, they can clearly reflect the basic physical characteristics of friction, such as specific stick-slip and continuous low-dissipation sliding behaviors. Based on analytical models, a classical MD simulation is further introduced, which can better consider atomic details and further investigate the friction, adhesion, wear, and lubrication behaviors of materials. The accuracy of classical MD depends on the preset force field, and the simulation results are usually based on the material structure and morphology, such as the puckering effect and evolution of interface quality. However, due to the complexity of realistic environments, the friction process often involves many electronic interaction mechanisms that classical MD cannot handle, such as complex coupling behaviors among the substrate, lubricating film, and environmental substances. Density functional theory (DFT) calculation is suitable for revealing these electronic interaction mechanisms, as it can provide all the ground-state properties of materials by solving the electronic wavefunction. In friction research, DFT can be used to simulate the PES of sliding interfaces, reflect the difficulty of sliding, and reveal the electronic origin of PES fluctuations. Therefore, exploring the complex interactions between various environmental substances, material substrates, and defect sites as well as the effects of environmental substance adsorption, aggregation, and diffusion on the long-term stability of coatings is necessary. However, although static DFT calculations can accurately reveal the electronic interaction mechanism, they cannot consider the dynamic evolution process involved in friction, such as the dynamic disturbance of environmental water molecules and the internal stress generated by interface sliding. AIMD simulations are ideal in that they consider both electronic interactions and the dynamic evolution of friction. In addition to reviewing existing AIMD simulation models, this work introduces a new “slow growth” AIMD simulation method and conducts systematic verification calculations in representative MoS2 systems, revealing the effectiveness and accuracy of the “slow growth” method in studying material friction behavior. Compared with classical MD, AIMD simulations have the problems of small scale and slow speed, which severely limit their application. With the development of artificial intelligence technology, machine-learning methods have been used to train MD force fields based on AIMD calculation data, opening the door to a comprehensive exploration of macroscopic engineering problems using computational simulations. Accordingly, this work introduces an “on the fly” machine-learning method, which can continuously train and improve the force field during the AIMD calculation process, greatly accelerating the calculation speed while ensuring the accuracy of the calculation to a certain extent. The summary and outlook of the theoretical simulation methods in this work can help to better understand the microscopic lubrication mechanism of materials in complex environments in the future and guide the design of advanced lubricating coatings. |
| Key words: lubricantion layered material environmental effect first-principles calculation machine learning |
