Heterogeneous thermal activation energy in Cu-Zr metallic glasses

<div>The zipped files include the atomic configurations of six Cu-Zr metallic glasses (MGs) and the activation energy of each atom calculated using the activation-relaxation technique (ART). The atomic configuration is in the format of LAMMPS dump file, and the activation energy is stored in the text file.</div><div><br></div><div>To generate these datasets, we first employ molecular dynamics (MD) simulation to synthesize six Cu-Zr model MGs: i) different compositions yet with the same cooling rates (Cu₅₀Zr₅₀, Cu₆₄Zr₃₆, Cu₈₀Zr₂₀ quenched from liquid at 10¹⁰ K/s), and ii) same composition but with different cooling rates (Cu₆₄Zr₃₆ MGs with the quenching rates of 10⁹ to 10¹² K/s). Samples containing 10,000 or 5,000 atoms (if 5,000, we will prepare two different samples at the same processing condition) were quenched to room temperature (300 K) from equilibrium liquids above the corresponding melting points.</div><div><br></div><div>We then apply the activation-relaxation technique (ART) to probe the energy barrier for thermally-activated events^[^1,2]. For each atom in those MGs, we initiate 50 independent events along random activation pathways upon atoms up to its nearest neighbors. Initial perturbations in ART were introduced by applying random displacement on a small group of atoms (an atom and its nearest-neighbors). The magnitude of the displacement was fixed, while the direction was randomly chosen. When the curvature of the PEL was found to overcome the chosen threshold, the system was pushed towards the saddle point using the Lanczos algorithm. The saddle point is considered to be found when the overall force of the total system is below 0.01 eV/Å. The corresponding activation energy is thus the difference between the saddle point energy and the initial state energy. The search is performed using ART nouveau package^[3]. For each group of atoms, we employed ~50 successful ART searches with different random perturbation directions. The average value of 50 independent activations around each center atom are sufficient to achieve a converged average activation energy, which contains key statistical information for thermal excitations on each local region (including centered atoms and its nearest neighbors).<br></div><div><br></div><div><br></div><div>References:</div><div>1. Barkema, G. T. & Mousseau, N. Event-Based Relaxation of Continuous Disordered Systems. Phys. Rev. Lett. 77, 4358–4361 (1996).</div><div>2. Rodney, D. & Schuh, C. Distribution of Thermally Activated Plastic Events in a Flowing Glass. Phys. Rev. Lett. 102, 235503 (2009).</div><div>3. Marinica, M.C., Willaime, F. & Mousseau, N. Energy landscape of small clusters of self-interstitial dumbbells in iron. Phys. Rev. B. 83, 9, (2011).</div><div><br></div><div><br></div><div><div><b>Note on data source</b></div><div><br></div><div>This data is part of our paper:</div><div>Qi Wang, Jun Ding and Evan Ma. Predicting the propensity for thermally activated β events in metallic glasses via interpretable machine learning (2020). </div><div><br></div><div>The quenching and ART simulations were carried out by Dr. Jun Ding. </div><div><br></div><div>If you found this dataset useful and would like to cite it in your work, please be sure to cite its original sources below rather than or in addition to this page.</div></div>