Dynamic mesophase transition induces anomalous suppressed and anisotropic phonon transport revealed by unified machine learning potential

The physical/chemical properties undergo significant transformation in the different states arising from phase transition. However, owing to the lack of a dynamic perspective, transitional mesophases are largely underexamined, which is limited by the high resources burden of first-principles. Here, using molecular dynamics (MD) simulations empowered by advanced unified machine learning (ML) potential, we proffer an innovative paradigm for phase transition: regulating the thermal transport properties via the transitional mesophase triggered by a uniaxial force field. We investigate the mechanical, electrical, and thermal transport properties of the novel two-dimensional carbon allotrope of Janus-graphene with strain engineered phase transition. Notably, we found that the transitional mesophase significantly suppresses the thermal conductivity and induces strong anisotropy near the phase transition point. ML-driven MD simulations meticulously recapitulate the atomic-scale dynamic metamorphosis exhibited in Janus-graphene, where thermal vibration-induced intermediate amorphous or interfacial phases induce strong and anisotropic interfacial thermal resistance, which eludes capture from traditional first-principles methods. The investigation not only endows us with a novel perspective on mesophases during phase transitions but also augment our holistic comprehension of the evolution of material properties.