Deciphering the Complexity of Step Profiles on Vicinal Si(001) Surfaces Through Multiscale Simulations

The behavior of vicinal Si(001) surfaces are a subject of intense research for years, yet the mechanism behind its step modulation remains unresolved. Step B, in particular, can meander randomly or form a periodic zigzag profile, a surface phenomenon that has eluded explanation due to the lack of appropriate simulation tools. Here, a multiscale simulation strategy, enhanced by machine learning potentials are proposed, to investigate this mesoscale behavior. The study reveals a phase transition in the step profile on vicinal Si(001) surfaces from random meandering to a zigzag wave pattern as the miscut angle decreases. This step-profile transition is corroborated by Monte Carlo simulations and experimental observations. Remarkably, this transition is robust across various surface conditions, including bare, hydrogen-saturated, boron-doped, or strained surfaces. The findings resolve the long-standing puzzle of step polymorphism on vicinal Si(001) surfaces and pave the way for exploring mesoscale phenomena using multiscale simulations.