Juzar, Some systems are more "sensible" than others when it comes to atomic displacements - unfortunately I cannot provide you with much more information than this empirical fact. One thing to do is to restrain the atomic displacements. In the case of CG, you can go down from Siesta's default for MaxCGDisplacement - 0.4 Bohr - to a much lower 0.04 Bohr, or even lower, and this will limit the atomic displacements at all steps, something that in my experience helps for more difficult cases. In the case of Broyden, check the flag that sets the value of the initial Jacobian and set it to 0.1 or something lower then the default, which is 1. For Broyden, the limiting of the atomic displacements is indirect, since the Jacobian, built over the subsequent steps, depends on its initial value.
However... have you let the system evolve for more than two steps? The initial steps of the relaxation will naturally be more unstable, and when the system approaches its ground state, you should see much lower oscillations in the maximum force in the system. Good luck, Marcos On Sun, Sep 26, 2010 at 12:33 PM, Juzar Thingna <[email protected]> wrote: > Hello Everyone, > > I'm trying to geometry optimize graphene on Cu (111) surface using LDA > pseudos. > At the 0th CG step the max constrained force is around 7 eV/A but at > the 1st CG step it increases to to more than 1000 eV /A. I'm not quite > sure why there should be such a large change. I also tried Broyden > (One thread on the forum says broyden is more stable than CG) instead > of Cg but the result remains the same. Please can someone explain why > the maximum force should change so abruptly. Also what would be a good > way to optimize such a structure? > > My input files are attached. > > Regards, > > Juzar Thingna > Department of Physics, > Center for Computation Science and Engineering, > National University of Singapore. > > > "I almost wish I hadn't gone down the rabbit-hole ---and yet---and > yet--- it's rather curious you know, this sort of life!" Alice >
