Page 116 - 《摩擦学学报》2021年第3期

第 3 期 周峰, 等: 两类润滑剂物性参数和摩擦系数的高通量分子动力学计算 405

the existing framework. The starting point for generating the molecular model library is to obtain the simplified
molecular input line entry specification (SMILES) string of the functional groups of the ionic liquid anion, cation and
ester compounds. Although there are many ionic liquids and ester compounds with different molecular structures, the
number of functional groups is very limited and can be easily obtained from existing chemical libraries. The SMILES
string of common carbon chain structure has been built in the program. After entering the SMILES string of the tube
energy group, the program will automatically add the carbon chain structure to the functional group, and then combine to
generate ionic liquid and ester compounds to quickly obtain the molecular geometric topology of a large number of
lubricants Based on graph theory, the optimized potentials for liquid simulations all-atom (OPLS-AA) force field is
automatically assigned to lubricant molecules, and finally a molecular model library containing a large number of
lubricants is generated. The OPLS-AA force field can accurately describe the liquid phase thermodynamic
characteristics and dynamic behavior of the lubricant. The 1.14*CM1A method is used to calculate the partial charge.
This method can generate accurate partial charges for both the neutral lubricant molecules and the anions and cations in
the non-neutral ionic liquid.
Finally, based on the established lubricant molecular dynamics model library, a two-layer high-throughput parallel
computation method of lubricant physical parameters and friction coefficients is carried out. In high-throughput
computations, the amount of computation is concentrated in two aspects: first, there are a large number of lubricants in
the library, and computations need to be carried out for each; second, the computation of a single lubricant adopts a full-
atom scale simulation to ensure the computation accuracy. In order to solve the problem of the amount of computation, a
two-layer parallel strategy is adopted to improve the computation efficiency. The first layer divides the lubricant files in
the library into blocks, and computes the lubricant in each block concurrently. In a certain block, the lubricants are
calculated sequentially. The second level of parallel setting is to use multiple CPUs for parallel molecular dynamics
computations for a certain lubricant. The computation method can achieve tens of thousands of concurrent computations.
The accuracy of the proposed high-throughput computation method is verified by taking ionic liquid as the object. This
high-throughput computation framework points out a feasible direction for realizing the high-throughput rational design
of the whole process lubricant, and lays a theoretical foundation for the design of new lubricating materials.
Key words: highthroughput computation; molecular dynamics; ionic liquids; ester compounds; physical property;
friction coefficient

制造业的升级换代对现有润滑剂的供给和品质 化，快速获得大量润滑剂的分子几何拓扑结构；然后
[11]
提出了更高的要求. 润滑剂的需求必然趋向于个性 基于图论将OPLSAA力场 自动分配给润滑剂分子，
化、差异化和多样化. 当前，航空领域等高端润滑油依 并最终生成包含大量润滑剂的分子模型库；最后以分
旧过度依赖进口，我国润滑剂的研发速度和品质落后 子模型库为基础设计了两层高通量物性参数和摩擦
市场需求. 为应对未来市场对材料的要求，美国于 系数计算方法，可实现万级并发计算. 以离子液体为
2011年提出了材料基因组计划，我国也提出了自己的 对象验证了所提出高通量计算方法的准确性. 这一高
材料基因组规划 . 材料基因组计划的主要核心目标 通量计算框架为实现全流程润滑剂的高通量理性设
[1]
将研发周期缩短一半，研发成本降低一半 ，将极大促 计指出了可行的方向，为新型润滑材料的设计奠定了
[1]
进新材料的研发速度和效率. 对我国走出润滑剂研发 理论基础.

不足的困境具有重要的战略意义. 材料基因计划要实
1 高通量计算算法
现目标将依赖于高通量计算方法和计算资源，以实现
[2]
材料的理性设计 . 近年来，国内已经开展了大量材料 1.1 润滑剂分子模型库的生成
的高通量计算研究 [3-10] ，但是润滑材料的高通量计算 离子液体和酯类化合物是两类主要的润滑剂，所
尚属空白. 以目前生成的分子模型库中包含这两类润滑剂，其他
本研究中以离子液体和酯类化合物为对象提出 种类润滑剂在现有框架下可方便地添加. 生成分子模
了液体润滑剂的高通量分子动力学计算框架. 该方法 型库的起点是获得离子液体阴阳离子和酯类化合物
首先以离子液体和酯类化合物官能团的简化分子线 官能团的SMILES字符串. 虽然不同分子结构离子液
性输入规范(simplified molecular input line entry specifi- 体和酯类化合物有很多，但是官能团的个数非常有
cation，SMILES)为起点，对官能团进行碳链延伸和支 限，可以很方便地从已有的化学库中获得. 如图1所示