Our simulation data is much more detailed for the system where we generated various visualizations to create video movies of the fracture dynamics for experiments 1 and 2. Because of the greater detail, especially in the time evolution, we will describe this size system and mention facts about the larger system. Figure 1 and Figure 2 graphically summarize our nonzero strain rate simulation (experiment 1) for

Figure 1 (12kB). (a) The crack tip position (in units of reduced length) as a function of reduced time. The slope is the limiting speed in reduced units which corresponds to 0.57 of the Rayleigh speed . (b) The crack tip speed (in units of the Rayleigh sound speed) as a function of reduced time; (c) An expanded view of the crack tip speed for early time.

Figure 1 shows (a) the crack tip position (in units of reduced length) and (b) the crack tip speed (in units of the Rayleigh sound speed), both as a function of reduced time. Figure 1(c) is an expanded view of the crack tip speed for early time. From Fig. 1a, we find that the crack tip achieves a limiting speed equal to 0.57 of the Rayleigh speed However, the ``instantaneous" tip velocity is very erratic (Fig. 1b) after reaching a speed of Before a time of about 70 and a speed less than the acceleration of the crack tip is quite smooth (Fig. 1c); but with the onset of the erratic fluctuations of the tip speed, there is significant deceleration of the propagating crack. Each of these features in Figs. 1a, b and c are obtained for experiment 2 and for the larger system (), and they are in agreement with Fineberg's et al. experiments (3,4). Like in the laboratory experiment, the influence of physical boundaries are a concern when sound and dynamical defects reflect from them. It should be noted that the onset of the instability relative to tip motion ( ) occurs significantly earlier than it takes sound to travel from the tip to a lateral boundary and return ( ). Hence, the transition seems to be an intrinsic instability. This is like the laboratory experiments (3,4), and in terms of scaled crack distances adopted by Fineberg et. al. (4), the scaled crack extension at this arrival time is about the same for our computer simulation and the laboratory experiment.