Page 1 of 1
LMBJ calculations
Posted: Fri Aug 04, 2023 9:27 am
by rishikanta_m
Hi,
I would like to know the INCAR file tags for LMBJ calculation. Is the tag below enough to include LMBJ:
METAGGA = LMBJ
LASPH = .TRUE.
LMAXMIX = 4
Re: LMBJ calculations
Posted: Fri Aug 04, 2023 10:38 am
by rishikanta_m
Yet, another question is:
1. If the calculation with LMBJ is not converged within the wall time or the given NELM, how can we restart the calculation of LMBJ from the point it stops? Is there any recipe for it?
2. Another in relation to MBJ: If we run MBJ calculation with fixed c parameter (CMBJ), and it is aborted using STOPCAR (before it hits the wall time) to save the unconverged WAVECAR. Is it reasonable/correct to restart the MBJ calculation using this unconverged WAVECAR?
Re: LMBJ calculations
Posted: Fri Aug 04, 2023 3:09 pm
by alexey.tal
Dear rishikanta_m,
These tags should be sufficient to enable the LMBJ functional.
1. If the calculation with LMBJ is not converged within the wall time or the given NELM, how can we restart the calculation of LMBJ from the point it stops? Is there any recipe for it?
As far as I know, with LMBJ it is not possible to restart a calculation without repeating the self-consistent-field procedure. Our expert on xc functionals is currently on vacation, but we will post an update in this thread later if he knows of a way to achieve that.
2. Another in relation to MBJ: If we run MBJ calculation with fixed c parameter (CMBJ), and it is aborted using STOPCAR (before it hits the wall time) to save the unconverged WAVECAR. Is it reasonable/correct to restart the MBJ calculation using this unconverged WAVECAR?
This should work if you restart the calculation with
CMBJ from the last iteration before the calculation was interrupted.
Re: LMBJ calculations
Posted: Sat Aug 05, 2023 5:27 am
by rishikanta_m
Thanks for the clarification.
Would like to know more about LMBJ calculations!!
Re: LMBJ calculations
Posted: Thu Aug 10, 2023 3:38 pm
by fabien_tran1
Hi,
A few things about MBJ and LMBJ:
-MBJ should not be used if your system has vacuum or an interface, because the average (integral) to calculate CMBJ = c = Eq. (3) in
http://dx.doi.org/10.1103/PhysRevLett.102.226401 has no meaning for such systems. However, there are two solutions: (a) Use MBJ, but with a fixed value of CMBJ (to be specified in INCAR) or (b) use LMBJ.
-For bulk systems with no vacuum and no interface, MBJ and LMBJ should lead to the same results.
-Using LMBJ is a bit dangerous, because I am not sure at 100% that the implementation is bug free.