*3.3. Analysis of Dislocation-Mediated Delamination*

While we always observed lattice-trapped delamination in our simulations, dislocation nucleation was only observed in "high stress" simulations. Furthermore, if the stress was too high then dislocations would nucleate at the beginning of the simulation, much like the behavior observed in our molecular statics simulation. One important question is: what is the interplay between lattice-trapped and dislocation-mediated delamination? What governs the transition from one mode to the other? To gain insight into this question, we extracted the crack volume at the moment of dislocation nucleation from our simulations and present the average values in Figure 9. At lower stresses, the crack volume increases to a peak value and then subsequently decays to zero, indicating immediate nucleation of dislocations. We believe that these trends result from changes in the lattice-trapped delamination and dislocation nucleation rates with stress and temperature, and changes in the driving force for dislocation nucleation as the crack grows. We defer further analysis to the Discussion.

**Figure 9.** Average crack volume among 10 replicas at the moment of dislocation nucleation as a function of hydrostatic stress over temperatures ranging from 200 to 400 K.

In all simulations, as soon as the first dislocation nucleation event occurred, the delamination rate increased dramatically. The delamination rate was so large that a welldefined rate could not be determined. In some cases, dislocations nucleated from the "crack tip" near the particle interface, but not in all cases. For example, dislocation nucleation in the middle of the crack face in the Al matrix was also observed.
