minimise & fix qeq

LAMMPS LAMMPS Mailing List Mirror
1

Dear Lammps Users
I have a problem with using minimisation command and using “fix qeq”.
I have used the Reax/c potential with “fix qeq” and attempt to minimise the system, but the minimisation stop due to “linesearch alpha is zero”. I dont have this problem when I used constant charge instead of “fix qeq” and everything is fine. I have changed the " min_style" and “min_modify” parameters but
the problem exist. please advise me.
thanks

2

This simply means that with and without equilibrium charges, the system is being brought to different minima - which is well expected.

p.s. there is no fix qeq, but only fix qeq/reax and fix qeq/comb.

Ray

3

Dear Ray

Thanks for your reply.

I used fix qeq/reax. is there any suggestion to minimise the system with equilibrium charges.

Thanks

4

No, it is actually better to use equilibrium charges since it is a variable charge potential.

Ray

5

who says, that there is a problem with your minimization? this is just
one of the possible conditions when it would stop. do you prefer it
will just jump around and get nowhere? have a look at the
documentation:

6

Dear Axel
Thanks for your reply. I am sorry for using the “problem” word. I have studied the documentation
previosly more for “linesearch alpha is zero”. I searched the most of LAMMPS mailing list for a technical suggestion to continue the minimisation if I am in a local minimum. this happend when I am using the ELASTIC sample file. for calculation of the elastic constants of crystal, I need to minimise the energy and force.
I want to know is there any technical suggestion.
thanks

7

Dear Axel
Thanks for your reply. I am sorry for using the "problem" word. I have
studied the documentation
previosly more for "linesearch alpha is zero". I searched the most of LAMMPS
mailing list for a technical suggestion to continue the minimisation if I am
in a local minimum. this happend when I am using the ELASTIC sample file.
for calculation of the elastic constants of crystal, I need to minimise the
energy and force.
I want to know is there any technical suggestion.

technical suggestion for what?

8

technical suggestion to avoid of local minimum and “linesearch alpha is zero” in minimisatin of a crystal. (Fe2O3)
Thanks

9

I’ll add one point which Ray can clarify. To do a fully
correct minimization with variable charge (e.g. one of
the fix qeq cases) you would need to treat the
charge on each atom as additional variables and compute a gradient
for how the energy changes as the charge changes.

Fix qeq/reax is not doing that so far as I know. It may be adjuring
the charge when the atom coordinates change, but
the minimizer will be confused by that b/c the new
energy will not reflect the limited gradient info it
knows about.

So I think it actually a bad idea to allow the charge
to vary via fix qeq when you perform a minimization.
The code should probably not allow you to do that.

Steve

10

That reminds me of another thing: one cannot minimize beyond the convergence of the charge equilibration.

Axel.

11

Steve, Axel, Aidan,

From methodology point of view, this may not be a problem. Please see the attached pdf document I just made.

Following the end of the document, I think it is okay to minimize the atomic positions of a variable charge while using QEq.

Ray

min_w_qeq.pdf (48.6 KB)

12

Yes, what Ray says is correct. Essentially, the forces on the charges are relaxed to zero at every iteration of the minimizer, and so they do not need to be treated explicitly by the minimizer.

Aidan

13

Mohammad,

Does the FIRE algorithm also fails? What are your minimization tolerances (send the input file maybe)

Carlos