Dear GULP team,
I want to fit potential parameters for PZT, I set simultaneous fitting firstly.Then I constantly adjust the weights of parameters, When the square sum of errors can’t be reduced, I change to relax fitting.
But there are errors: Cell parameters has below allowed limit. Largest core-shell distance exceeds cutoff of cuts. Then I fix some parameters, but it’s useless.
Is the simultaneous fitting not sufficient? I should set a smaller function tolerance for optimisations?
My input contents are shown in the attachment. hope you can tell me the reason. Looking forward to your answer.
Thanks in advanced!
Xiaohan
PZT_10.grs (3.5 KB)
Dear Xiaohan,
I think the issue here is that you’re expecting a force field to give you a perfect fit to a lot of different properties for a material. This doesn’t generally happen for a complex system & depends on whether you’ve included all the right physics in the model. For example, for perovskites a shell model can reproduce a cubic or rhombohedral phase, but not a tetragonal distortion because this is driven by quadrupoles (which are not in a shell model). I would recommend start from literature values for potentials for the interactions in the system and then (assuming the system will relax successfully) do any fitting from there with relax. There’s no point playing with weights in such fine detail (you have values to many significant figures) - just use the defaults & only increase or decrease where there is a significant error (that the model has the capability to fundamentally handle). It’s important to remember that the properties only have meaning once the structure is at a minimum & so will only fit correctly for a relax fit (simultaneous fitting would work if the structure is easy to get right).
Regards,
Julian
1 Like
Dear Julian,
Thank you for your reply. By reading the manual, I know the properties only have meaning with the relax fit. There are errors if I use the relax fit directly, even though all the potential parameters are from the literature.
The GULP manual shows: you can only start a relax fit once you have a reasonable set of potential parameters. So I choose to start from a simul fit. I initially use the default weights, the attachment above is the result obtained after multiple iterations.
Is the shell model not suitable for tetragonal structure?（But all the literature I have seen used this model and GULP) Or my operation is incorrect?
The attachment is my initial set.
PZT_r.gin (1.6 KB)
Looking forward to your guidance.
Thanks in advanced!
Xiaohan
Dear Xiaohan,
If the initial literature parameters won’t optimise then I’d suggest manually adjusting the A parameter of the Buckingham for whichever distance is collapsing until the structure is stable. You can then use the relax fit to refine.
A shell model captures dipoles, but the tetragonal distortion in perovskites is caused by quadrupoles. Therefore a shell model can’t capture this distortion. It will minimise to either a cubic or rhombohedral structure depending on the polarisability of the ions.
Regards,
Julian