Bond parameter fitting

Dear Julian,

I am confused by the fitting procedure of gulp to parameterize the bond terms (K and r0 in harmonic functional form).

My goal is to release the stress to 0 GPa and to keep the volume the same as the experimental one by varying the bond parameters, so my observable is just the stress of 0 GPa, and I think it is not necessary to use other observables, like elastic constant. Meanwhile, I used ‘conv’ to hold the volume.

I start firing only one parameter and fit others subsequently in the restarts. The first parameterization always gives a lower stress at the end of the fit (from initial -1.015 to -0.228), but the subsequent fittings resulted in more enormous stress than the initial one.

I tried with ‘relax’, ‘simplex’, and conventional fit with the above steps, but ‘relax’ fits always show a warning of 'Not all configurations optimised successfully in relaxed fit ', and all of the fitting algorithms give the same increased stress after several turns of fit.

Please let me know if there is anything not suitable. Thank you in advance.

Here is my gin file:
fit_pb1.gin (9.3 KB)

All the best
wang

Dear Wang,
I’m afraid I can’t give you a detailed discussion of your fitting process since this is the role of the academic lead for the project (your supervisor?) to decide how to conduct the work. I can point out that if you want to fit the volume then it makes no sense to use “conv” since this means you’re only fitting the internal coordinates. If you use “conp” then the cell will naturally be fitted. Again, I’d recommend making sure you fully understand the content of the Phil. Mag. B paper on fitting with GULP since this would make things much easier.
Regards,
Julian