PHYSICAL REVIEW MATERIALS 5, 075602 (2021) Exploring the crystallization path of lithium disilicate through metadynamics simulations Federica Lodesani , 1 Francesco Tavanti , 1, 2 Maria Cristina Menziani , 1 Kei Maeda, 3 Yoichi Takato, 4 Shingo Urata , 5 and Alfonso Pedone 1 , * 1 Department of Chemical and Geological Sciences, University of Modena and Reggio Emilia, via G. Campi 103, 41125 Modena, Italy 2 CNR-NANO Research Center S3, Via Campi 213/a, 41125 Modena, Italy 3 Department of Materials Science and Technology, Tokyo University of Science, Tokyo 125-8585, Japan 4 Innovative Technology Laboratories, AGC Inc., Yokohama, Kanagawa 230-0045, Japan 5 Planning Division, AGC Inc., Yokohama, Kanagawa 230-0045, Japan (Received 23 January 2021; accepted 16 July 2021; published 30 July 2021) Understanding the crystallization mechanism in silica-based materials is of paramount importance to com- prehend geological phenomena and to design novel materials for a variety of technological and industrial applications. In this work, we show that metadynamics simulations can effectively overcome a large energy barrier to crystallize from viscous oxide glass melts and can be used to identify the melt-to-crystal transition path of the lithium disilicate system. The accelerated atomistic simulation revealed of a two-step mechanism of the nanoscale crystal formation. First, a partially layered silica embryo appeared, and then a more ordered crystalline layer with size larger than the critical nucleus size was formed. Subsequently, lithium ions piled up around the silicate layer and triggered stacking of adjacent silicate layers, which eventually built a perfect crystal. Contrarily to previous hypotheses, no lithium metasilicate crystal was observed as a precursor of the homogeneous crystallization of lithium disilicate. DOI: 10.1103/PhysRevMaterials.5.075602 I. INTRODUCTION Understanding the nucleation and crystallization processes in silicate melts and glasses is essential to design novel glass- ceramics materials with enhanced optical, mechanical, and chemical properties [1,2] and to prevent uncontrolled crystal- lization (devitrification) during glass production. Several theories have been developed and refined to inter- pret the first key step, nucleation [3–8]. Classical nucleation theory (CNT), one of the most accepted, presumes that a critical crystal-like nucleus, with composition, structure, and properties of the final macroscopic phase, forms from random fluctuations of atomic arrangements in the melt. Therefore, in CNT the energy barrier for nucleation results from a change in size rather than composition and structure. This simplifica- tion limits the applicability to only simple crystallization, and it might be a reason why CNT severely underestimates the nucleation rates in silicate systems [3,9]. Alternative theories have been proposed to complement these limitations; in the generalized Gibbs approach (GGA) [10,11], it is assumed that the structure and properties of the nuclei can deviate from those of the macroscopic phases. This means that the nucleus structure can change during the growth, and, therefore, the final structure can be different from the original crystal structure. In GGA, the energy barrier for nucleation is not overcome by a change in size (as in CNT) but by a change in composition. Instead, in the two-step model [6], a temporal separation between the density and structural * Corresponding author: alfonso.pedone@unimore.it fluctuations is assumed. First, local disordered regions with composition and density different from those of the initial liquid phase are formed, and then a structural reorganization inside this region leads to the appearance of the crystals. In many cases, crystallization can evolve through a se- quence of intermediate metastable configurations formed before the thermodynamically stable final crystalline phase appears [12–15]. Although these intermediate phases are often only tran- sient, they play an important role in the crystallization pathway, and identifying their formation is crucial for con- trolling or inhibiting the formation of specific desired or undesirable phases, which would have a potential impact on the economic consequences as well for industry. Among silicates, lithium disilicate (LS 2 ) has attracted huge fundamental and technological interest since it nu- cleates homogeneously with relatively high rates (≈4.4 × 10 9 nuclei cm –3 s –1 at 500 ◦ C) and easily forms glass- ceramics products with enhanced chemical, mechanical, and thermal properties [16–19]. Several in situ characterization techniques have been applied to monitor the evolution of nucleation and crystallization processes to gain insight into fundamental aspects of crystallization in silicates [14,20–22]. However, contrary to the crystal growth, observing nucleation is extremely challenging since it occurs in the subnanometer length scales, which is often out of the experimental observ- able limit. Therefore, the crystallization pathway has not yet been understood from an atomistic viewpoint, and this major challenge remains active. The major controversy concerning the crystallization ki- netics in the LS 2 system is whether metastable phases appear 2475-9953/2021/5(7)/075602(9) 075602-1 ©2021 American Physical Society