Markus

Yes, there is an example for CaO in the examples folder. You need the core states as valence, i.e. you have to model them with local orbitals. There is a special species file for Ca in the example which shows you how to do it for the 2p states. These calculations are extremely expensive and need lots of RAM. Carefully optimize the initial and final states to keep the matrix size as small as possible and start with a very small number of k-points. XMCD is also possible, compare with the XMCD example....

Yes, there is an example for CaO in the examples folder. You need the core states as valence, i.e. you have to model them with local orbitals. There is a special species file for Ca in the example which shows you how to do it for the 2p states. These calculations are extremely expensive and need lots of RAM. Carefully optimize the initial and final states to keep the matrix size as small as possible and start with a very small number of k-points. Best wishes, Markus

Hi Hanley, you can find it in the code, start from occupy.f90: https://elk.sourceforge.io/doxygen/html/occupy_8f90_source.html Then look for stheta.f90 and stheta_fd.f90: https://elk.sourceforge.io/doxygen/html/stheta_8f90_source.html and https://elk.sourceforge.io/doxygen/html/stheta__fd_8f90_source.html The argument of the standard Fermi-Dirac distribution is (efermi-e)/swidth, see line 64 of occupy.f90. This is the distribution, then the approximation to the smooth delta function is specified...

Hi, your input is wrong, that is not UN, you made UN2 (with the wrong lattice constant). Replace the N position with 0.5 0.0 0.0, that gives you the correct structure and stoichiometry. Cheers, Markus

Hi, did you comment this line: #SRC_W90S = w90_stub.f90 above the lines you have to uncomment? Is Wannier90 compiled with the same compiler? Cheers, Markus

Dear Hong Tang, 2D systems are generally quite demanding in terms of convergence with GGA and mGGA. Instead of playing with all the basis set and numerics parameters by hand, I would recommend to try highq=.true. and vhigh=.true. first. There can be discontinuities in the density at the MT boundaries, which can be resolved with tighter basis set parameters. These spoil the gradient calculation and lead to numerical instability. Also, try with smaller k-point grid first, so you get a feeling for the...

... if I remember correctly, you have to use libxc 7.0.0 for elk 10.8.1. Please try this. Remember to compile libxc according to your preferred linking type (static or dynamic).

What are you comparing to? Is it FLAPW calculations or maybe some sort of pseudopotential calculation? The basis sets play a big role in the description of localized states. You can tweak your convergence parameters by just setting high or vhighq and see what happens. Easier than testing all parameters individually. Once you get what you expect, you can try to find the responsible parameter.