[lammps-users] Question about indentation

LAMMPS LAMMPS Mailing List Mirror
1

Dear all,

I am doing indentation using minimize command as in the Lammps example. Suppose two following cases:

Case 1: Indent from step 1 to step 10, with equal indentation depth in each step; This progress run smoothly in my computer;

Case 2: If the computer happened to crash in step 6 for some reason, I used the atomic configuration as stored in Step 5 to restart the indentation(minimize) process.

I guess in these two cases, the output of step 6 (such as forces and other thermo output would be the completely the same) because minize is just a static command.

But the truth is not that simple. Why?

I have to indent some system into a deep distance(Generally, it will take a long time of more than a week), but the wall-time limit for our supercomputer is only 3-day. That means, I have to restart the indentation after that period with some already existed indented configuration. But as I told that the results of this restarting process are not exactly the same as the continuous indentation, are the outputs of this restarted calculation reasonable. How to solve this problem? Thanks.

2

In principle, what you describe should be possible. I assume you
have built some kind of loop where you minimize, move the indenter,
minimize, etc. So if you restart from the middle of that loop you
will need to do
it carefully. I'd imagine due to roundoff that LAMMPS could get slightly
different answers on a restart, especially since the minimizer stopping
criteria is sensitive to small changes. But if the answers are widely
different, then you probably didn't restart like you think. You might
test this procedure with running dynamics incrementally rather
than minimize and see if it works as you expect.

Steve