The state of metastable equilibrium glass can carry an imprint of the past and exhibit memory effect. As a hallmark of glassy dynamics, memory effect can affect glassy behavior as it evolves further upon time. Even though the physical picture of the memory effect has been well studied, it is unclear whether a glass can recall as many pieces of information as possible, and if so, how the glass will accordingly behave. We report that by fractionizing temperature interval, inserting multistep aging protocols, and optimizing the time of each temperature step, i.e., by imposing a multistep “training” on a prototypical $\mathrm{P}{\mathrm{d}}_{40}\mathrm{N}{\mathrm{i}}_{10}\mathrm{C}{\mathrm{u}}_{30}{\mathrm{P}}_{20}$ metallic glass, the memory of the trained glass can be significantly strengthened, marked by a pronounced augment in potential energy. These findings provide a new guide for regulating the energy state of glass by enhancing the nonequilibrium behaviors of the memory effect and offer an opportunity to develop a clearer physical picture of glassy dynamics.

• Received 2 May 2017
• Revised 17 September 2017

DOI:https://doi.org/10.1103/PhysRevB.96.174204