How to improve elastic constant calculation

stress-strain.jpg

Dear All,

I often face the issue of what to do to improve the elastic constant calculation after seeing the warning message “….the convergence is already as good as it can be and my convergence criteria is too tight” (even though I choose "Fine" most of the time). An examination of the stress-strain pairs indicate that some are not linear or very scattered (an example is attached). Before calculating elastic constant, I usually do a geometry optimization with tested Ecut and k points and a selection of "Fine" in the elastic constant convergence and 6 steps for each strain with a maximum strain up to 0.006. Are there other parameters I have to adjust to improve these calculations? Thank in advance.

Leo