My Community

Dear all,

I have calculated Hubbard U parameter by using linear response.

However, I don't know how to implement DFT+U method using the U value that I gained.

Now I got U value, but I don't have J value.

Is there any specific process that calculates J value?

Or should I just implement DFT+U method using 'LDAUTYPE=3' ?

Thanks for reading.

However, I don't know how to implement DFT+U method using the U value that I gained.

After having calculated U, you can use it in a DFT+U calculation: LDAUTYPE=1 or 2 (4 is probably not recommended) and LDAUU=U, where U is your calculated value.

Now I got U value, but I don't have J value.
Is there any specific process that calculates J value?

It is in principle also possible to calculate J from linear response, as mentioned at page 6 of https://doi.org/10.1103/PhysRevB.71.035105, but this is probably rarely done and not necessary because J is typically smaller than U and is of less importance except if you are looking at very fine details. Usually, either U is redefined as U=U-J or J is set to some reasonable value like J=0.75 eV. See page 3 of https://doi.org/10.1103/PhysRevB.71.035105.

Or should I just implement DFT+U method using 'LDAUTYPE=3' ?

This question is not clear.