Soft matter systems are extremely broad, encompassing polymers, colloids, liquid crystals, gels, and more. The competition between entropy and enthalpy within these systems leads to exceptionally complex microstructures at the mesoscopic scale. Numerous metastable structures in these systems capable of maintaining stability over extended periods. By regulating the evolution pathways of microstructural dynamics, we can obtain metastable structures with targeted structural features, effectively preventing the system from reaching its thermodynamically stable state of minimum free energy. We employed machine learning methods to propose research strategies and develop corresponding methodologies targeting the diffusion dynamics of microstructure formation in soft matter systems. We discovered that machine learning-based models demonstrate surprisingly superior performance in multiple aspects. Not only can they accurately extract the physical properties of non-equilibrium phase transitions, but they also significantly accelerate the solution process for the system’s dynamic evolution. Furthermore, they effectively extract the dynamic correlation information present in diffusion behavior. Based on these preliminary findings, we conclude that machine learning approaches can introduce novel perspectives to soft matter research, potentially establishing a transformative new research paradigm. This approach not only deepens our understanding of the physical mechanisms governing these systems but also holds promise for achieving precise quantitative predictions. Ultimately, it may bridge the gap from uncovering fundamental physical mechanisms to enabling experimentally controllable synthesis.
蒋滢，绍兴文理学院化学化工学院教授，获得国家优秀青年科学基金项目。主要从事软物质体系在微纳尺度相行为的理论研究，通过发展介观尺度的平衡态和动力学的理论与计算方法，阐释体系微结构的调控机制。研究体系包括：半刚性聚合物微相分离、表界面诱导液体定向输运、胶体溶液、软物质玻璃等。近期，聚焦于将机器学习方法引入软物质体系物理规律的研究，通过设计以及发展机器学习方法和模型，提升对体系非线性响应行为的预测准确性，同时，探索预测模型的物理可解释性，从机器学习模型的角度理解体系的微观机制。近年来，以第一/通讯作者发表SCI论文60余篇，包括：PRL, PNAS, Advanced Materials, Angew. Chem. Int. Ed., Small, ACS Macro Letters, Macromolecules, Soft Matter, PRE等。