Predicting Shear Transformation Events in Metallic Glasses

Shear transformation is the elementary process for plastic deformation of metallic glasses, the prediction of the occurrence of the shear transformation events is therefore of vital importance to understand the mechanical behavior of metallic glasses. In this Letter, from the view of the potential energy landscape, we find that the protocol-dependent behavior of shear transformation is governed by the stress gradient along its minimum energy path and we propose a framework as well as an atomistic approach to predict the triggering strains, locations, and structural transformations of the shear transformation events under different shear protocols in metallic glasses. Verification with a model ${\mathrm{Cu}}_{64}{\mathrm{Zr}}_{36}$ metallic glass reveals that the prediction agrees well with athermal quasistatic shear simulations. The proposed framework is believed to provide an important tool for developing a quantitative understanding of the deformation processes that control mechanical behavior of metallic glasses.

Figure 1

The distribution of triggering strains (between $-100.00\%$ and 100.00%) for $xy$ shear of all the events harvested. The inset highlights the distribution of triggering strain between 0.00% and 30.00%.

Figure 2

Loading (blue line, from shear strain 0.00% to 5.00%) and unloading (red line, from shear strain 5.00% to 0.00%) stress-strain curves for $xy$ shearing. The loading curve is shifted upwards by $+0.005\mathrm{GPa}$ for clarity. Four shear transformation events were found corresponding to the stress drops on the loading curve. The first two events were recovered during the subsequent unloading deformation.

Figure 3

The analytically predicted triggering strains [red circles, following Eq. (5)] and the numerically predicted triggering strains (blue squares, following the full numerical integration [19] versus the AQS values).

Figure 4

Comparison of the local configurations of the ST events between the predictions and AQS simulations. Atoms were colored according to their respective atomic von Mises strain [31] taking the initial undeformed configuration as a reference, and only those with an atomic von Mises strain greater than 0.01 are shown. For each case, the left one is identified by AQS loading and unloading, and the right one is predicted based on ARTn. The triggering strains calculated from Eq. (5) and simulated from AQS are annotated around local cluster. All the predicted triggering strains are close to simulation results, except for the right one in (e). The possible reason is discussed in the text.

Figure 5

The AQS triggering strains versus the predicted ones [following Eq. (5)] for all predicted events in the four samples investigated.