Exploring the nature of the liquid–liquid transition in silicon: a non-activated transformation

Abstract

In contrast to other glass formers, silicon exhibits a thermodynamic discontinuity between its liquid and amorphous solid states. Some researchers have conjectured that a first-order phase transition occurs between two forms of liquid silicon: the high-density liquid (HDL) and the low-density liquid (LDL). Despite the fact that several computer simulations have supported a liquid–liquid phase transition (LLPT) in silicon, recent work based on surface free energy calculations contradicts its existence and the authors of this work have argued that the proposed LLPT has been mistakenly interpreted [J. Chem. Phys., 2013, 138, 214504]. A similar controversy has also arisen in the case of water because of discrepancies in the calculation of its free energy surface [Nature, 2014, 510, 385; J. Chem. Phys., 2013, 138, 214504]. Current evidence supporting or not supporting the LLPT is mostly derived from the thermodynamic stability of the LDL phase. Provided that the HDL–LDL transition is a first-order transition, the formation of LDL silicon should be an activated process. Following this idea, the nature of the LLPT should be clarified by tracing the kinetic path toward LDL silicon. In this work, we focus on the transformation process from HDL to LDL phases and use the mean first passage time (MFPT) method to examine thermodynamic and dynamic trajectories. The MFPT results show that the presumed HDL–LDL transition is not characterized by a thermodynamic activated process but by a continuous dynamic transformation. LDL silicon is actually a mixture of the high-density liquid and a low-density tetrahedral network. We show that the five-membered Si–Si rings in the LDL network play a critical role in stabilizing the low-density network and suppressing the crystallization.