An excess of low-frequency vibrational states beyond the Debye prediction—known as the boson peak—is a universal feature of disordered materials, yet its microscopic origin and connection to atomic-scale structural disorder remain unresolved. Here we study the microscopic origin and real time evolution of the boson peak during crystallization of amorphous Ge-Sb-Te by carrying out large-scale molecular dynamics simulations using machine learning interatomic potentials. The boson peak frequency is found to match with the transverse Ioffe-Regel crossover frequency throughout crystallization. As crystallization proceeds, the boson peak shifts to higher frequencies and its intensity progressively weakens, while the macroscopic shear modulus increases concurrently. The square-root scaling of the boson peak frequency with the shear modulus is more pronounced in mechanically soft and interfacial regions than in stiff regions. Atom-resolved vibrational anomaly highlights the population of quasilocalized low-frequency modes in the mechanically soft regions. The anticorrelation between the shear modulus and the boson peak intensity persists even at atomic scales in both amorphous and partially crystalline phases. Our large-scale atomistic simulations of structural ordering provide a microscopic perspective on the interplay between structural ordering, local elastic rigidity, and vibrational anomalies in real materials.

Close

PrevNext

View figure in article

• Elastic modulus
• Phonons
• Amorphous materials
• Machine learning