Dear Kirill,

The first part of my answer is about the reciprocal Bravais matrix. This is actually written out to the out file by the input generator. But it is overwritten once you perform a Fleur calculation. For example for Si inpgen out has a section that looks like

Bravais lattice vectors
                                               0.000000    5.130609    5.130609
                                               5.130609    0.000000    5.130609
                                               5.130609    5.130609    0.000000
reciprocal lattice vectors
                                              -0.612324    0.612324    0.612324
                                               0.612324   -0.612324    0.612324
                                               0.612324    0.612324   -0.612324
 
 

Be careful to interpret these matrices in the correct way. There may be a transposition present.

The second part is about the calculation of the cartesian coordinates of the k points:

We calculate the band band structure path in cartesian coordinates in the file dos/types_eigdos.F90 within the fleur source code folder. It is this loop:

      kx(1) = 0.0
      vkr_prev=matmul(kpts%bk(:,1),cell%bmat)
      DO k = 2, kpts%nkpt
        vkr=matmul(kpts%bk(:,k),cell%bmat)
        kx(k)=kx(k-1)+ sqrt(dot_product(vkr-vkr_prev,vkr-vkr_prev))
        vkr_prev=vkr
      ENDDO
 
 

kx is the x coordinate along the bandstructure plot. vkr is the cartesian coordinate of the respective k point stored in kpts%bk. cell%bmat is the reciprocal Bravais matrix though I am not sure whether the factor 2*pi is in it.

About compiling and linking to HDF5 I would like to add that (a) it is important to compile HDF5 with the same compiler you use for Fleur, (b) on my machine I use the following configure line for HDF5:

FC=/usr/local/intel/impi/2019.3.199/intel64/bin/mpiifort CC=/usr/local/intel/impi/2019.3.199/intel64/bin/mpiicc CXX=/usr/local/intel/impi/2019.3.199/intel64/bin/mpiicpc ./configure --enable-fortran --enable-parallel
 
 

Of course your compilers differ, but one important point here is the --enable-fortran switch. (c) You need to specify the HDF5 library path in your LD_LIBRARY_PATH environment variable. For me this looks like

export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:~/hdf5/current/hdf5/lib
 
 

Of course, for you this is a different path.

Please wait with these calculations for about a week. Right now we have a problem with the output in this calculation mode. You will not be happy with the results you see there. I will tell you once it is fixed.

The error message can be overcome by suppressing the detection of symmetries that are used to group symmetry equivalent atoms. There is an inpgen command line switch "-nosym" that suppresses this detection.

Oh wait, what did you actually do to compile with HDF5? You certainly called the configure script with the respective options. But did you call the two commands

make clean
make
 
 

in your build directory afterwards? This is needed. Whenever you change something in the configuration or the code changes you also have to run make again.

Best, Gregor.

Ok, that actually looks good. Can you also attach the out and out.xml files you obtain when running fleur and the output of

echo $LD_LIBRARY_PATH
 
 

The terminal output when running fleur would also be useful.

Maybe you are right, I only had a short look at your input. Only a band structure calculation could tell us whether it is an insulator or a metal. But in the end, as you just mentioned, this is not really relevant. The charge at the surface stays and leads to a very slow convergence with respect to the number of layers.

With respect to the band gap I could imagine that a finite band gap actually is something artificial in the output here. Maybe with a finer k mesh it will just disappear. As mentioned I actually expect to see a conductive layer on the surface.

Dear Terry,

Could you please also paste the valenceDensity and coreStates sections for the last iteration of the largest setup from the out.xml file? It looks like you accumulate charge at the surfaces. Another thing that may become visible in this data is whether there are problems due to the very small MT radii you have to use due to the small bonding lengths. It may be that your observation has a physical origin and is not related to a problem of the calculation.

On what kind of processor is this running? Is this an Intel or an AMD processor? For AMD processors you have to be careful in the compilation process. If you don't use the optimal compiler options you will get a bad performance on AMD processors.

Could you provide an input file to get a better impression? Is this actually a film calculation or a periodic slab calculation with a large space in between the slabs? Is your film/slab setup symmetric or asymmetric? How thick is the slab, what kind of chemical element is involved. Depending on the quantity and the involved elements surface properties can either be converged with just a few layers or you may need several tens of atom layers.

In periodic slab calculations you also have the electrostatic interaction between the slabs that may require a very large vacuum region.

The reciprocal lattice is written out to the banddos.hdf file. In the user guide you can find instructions on how to print a band structure from that file. This implies usage of the reciprocal lattice vectors.