Dear all,
I’m testing different parameters for EAM potential. Some guesses are really bad and I have “NaN” for potential energy in a minimization run. I would like to avoid such bad trials by simply running LAMMPS for 1 step.
variable pe equal pe
run 0 post no
if " ${pe} > 0 " then “jump SELF end”
Unfortunately, if statement fails with the error:
ERROR: Invalid Boolean syntax in if command (…/variable.cpp:4042)
Do you have any suggestions how to check potential energy value and safely exit program.
Regards,
Konstantin
If I add these lines to the bottom of bench/in.lj,
it loops forever (the pe stays < 0):
label end
variable pe equal pe
run 10
if " ${pe} < 0 " then “jump SELF end”
Are you using the current version of LAMMPS?
Steve
Hi Steve,
Probably, you didn’t understand my point.
The command if " ${pe} < 0 " then “jump SELF end” works fine for me, if potential energy is a float. But when pe = nan, that error happens.
Regards,
Konstantin
If you’re wanting LAMMPS to recover and do something specific
when your simulation generates NaNs, b/c of a bad model,
I don’t think that’s easy to do.
If you defined
variable pe equal pe/2.0
or
variable foo equal v_pe/2.0
and then tried using those in your if statement, the code would throw an error
in different places. The variable.cpp file does a lot of complex parsing
of formulas of different kinds and with different elements in
the equations. Trying to make each of them aware of the possibility
that a NaN might be present would be very difficult. And for
many of them, there is not the option of simply returning “false”
as a result. They would need to generate an error.
It would be better if you thought about how to avoid running
a model with bogus EAM parameters. E.g. in your driver
script or program, do some simple tests of parameters you
generate to see if the EAM formulas will work vs croak.
Steve