I am simulating ramp compression of an Aluminum bar under the NVE ensemble with pps boundary. Additionally, wall/piston is used on the zlo, and wall/reflect is used on the zhi.
fix 17 all nve
fix 18 all wall/piston zlo vel 60 ramp units box
fix 19 all wall/reflect zhi EDGE units box
In this setting, zhi is a rigid momentum mirror that does not allow any atom to pass, and the elastic wave will reflect on this boundary. But in experiment, the zhi boundary is a LiF window that matches the impedance of aluminum and able to move back freely. So, is there any other boundary condition that is better suited for this scenario? I have tried pps without wall/reflect on zhi, but in that setting when the elastic wave reaches zhi, all atoms will “shoot out” with a rather high velocity (about the sound of speed in Al). Then the density of the structure starts to decrease, and compression will not happen until the ramp velocity reaches the atom velocity, which takes forever. Can anyone help me to identify the proper boundary condition that I need to use? Thank you in advance.
First of all, this is not really a LAMMPS question, but more a question about MD simulation of dynamic compression experiments. So a better way to get answers is to carefully read the literature on this topic. Interestingly, the second simulation you describe is the standard solution i.e. create a free surface at zhi and terminate the simulation when the compression reaches that surface, unless you are interested in studying the leftward-moving rarefaction wave that initiates at the free surface. You can delay this event by increasing the extent of your sample in the z direction. If you are interested in what happens at an LiF window, you could build a second slab of material that has the density and stiffness of LiF. I am not sure what you would learn by simulating that.
Thank you for the reply. Indeed, the goal for me is to look at what’s happening at high piston velocity (I am using ramp loading, so it takes time for the ramp to accelerate). But there is a limit to which I can extend the structure due to the computational cost. So originally, I was thinking if there is any way around this problem by using some “magical” boundary condition, but I guess that does not exist. Thank you.