We integrate massconserving saddle dynamics with solution landscape construction to study triblock copolymer selfassembly under soft confinement. First, we develop an accelerated projection saddle dynamics (APSD) method for a multicomponent THC model, prove its linear stability, and construct solution landscapes, bifurcation diagrams, and heatmaps. A wealth of extreme points is classified into Layer, Flower, Circle, and CoreShell classes; their connections reveal novel structures and transition pathways. Second, we present a unified saddle dynamics framework under different inner products (projected $L^2$ and $H^{-1}$), prove the equivalence of indexkk saddle points to stable steady states, and show that the innerproduct choice may enriches the solution landscape, exposing previously unreported saddle points. 

徐真, 北京工业大学数学统计学与力学学院, 副研究员。2019年博士毕业于北京师范大学, 2019年-2021年在北京大学北京国际数学研究中心展开博士后研究。从事复杂流体的计算与分析方向, 研究相场模型高效数值算法, 解景观算法和应用。相关成果发表在 SIAM. J. Appl. Math., Neural Network, J. Sci. Comput., Commun. Comput. Phys., Comput. Phys. Comm., Phys. Rev. E 等期刊。入选北京工业大学高层次人才队伍建设"青年优秀人才项目”。目前主持国家自然科学基金青年项目, 参与国家自然科学基金面上项目, 北京市自然科学基金青年重点项目等。