Band energies in exciting BXSF files

[Patrick Rourke]

Dear exciting experts,

I have developed an algorithm for analyzing band energies in order to
extract quantum oscillation information such as frequencies and
effective masses, as well as band electronic density of states (DOS)
contributions -- quantities that can be useful for comparing DFT
calculations to experimental data. You can read about it in P.M.C.
Rourke and S.R. Julian, Computer Physics Communications 183, 324
(2012) (also http://arxiv.org/abs/0803.1895) and download the current
code release, named SKEAF v1.1.0 from http://www.wien2k.at/reg_user/unsupported/
.

I initially developed this program for use with BXSF files generated
using Wien2k + XCrysDen, but if possible I would like my code to work
with other DFT packages as well!

While preparing my next release of the SKEAF program, I have been
looking at compatibility with ELK and exciting, using calculations of
the band energies of Cu as a test.

I found that BXSF files generated using ELK differ in 3 ways from the
ones generated by Wien2k + XCrysDen that my program is expecting:
#1. the energies are specified on an incorrect Periodic Grid, rather
than a General Grid (see http://www.xcrysden.org/doc/XSF.html#__toc__14
for details)
#2. reciprocal lattice vectors include the factor of 2*Pi
#3. energies in units of Hartrees, rather than Rydbergs

But these are all easy to fix, and once I have appropriately corrected
the ELK Cu BXSF file, the SKEAF results agree with those from Wien2k
and with the experimental data for Cu.

I found that BXSF files generated using exciting helium-3 +
xmlfermis2bxsf.xsl have the same 3 problems as ELK, but have some
additional issues too:
#4. the sign (+ve/-ve relative to the 0 of the Fermi energy) of the
band energies has been switched --> this swaps around hole and
electron orbits
#5. band energies have been multiplied by a factor 10^(-minexp), where
minexp changes from calculation to calculation
#6. another mysterious problem that I don't understand with the band
energies that causes incorrect DOS and effective mass values to be
obtained, but does not affect the obtained frequencies (see below, I
need your help!)

Problem #4 is easy to fix and problem #5 can be fixed with some trial
and error (need to guess at the value of minexp, since as far as I
know it is not recorded by exciting anywhere). Then, for frequencies
only, the exciting SKEAF results agree with the ELK/Wien2k/
experimental ones.

Unfortunately, even after doing these corrections, the exciting Cu
effective masses and DOS contribution do not agree with what I find
using ELK & Wien2k -- the exciting values are too big by roughly a
factor of 1.5-2, indicating that the exciting band energies are too
small by this factor. Furthermore, the disagreement factor varies from
orbit to orbit (& the DOS), so it looks like this factor is not
uniform in k-space (i.e. it is different for different band energies
on the k-grid)... but despite this, it does not seem to affect the
positions of the Fermi energy crossings.

I looked through the exciting code, but do not understand it well
enough to see what might be causing this problem #6. I would be very
thankful for any insights that the exciting community may have about
this issue!

Note that I am a new user of exciting, so I may well have made a
mistake in my test calculation. I have copied my Cu input.xml file
below, and would appreciate learning of any errors on my part.

Thank you very much for your help,

Patrick


-------------------------------------------------------------------------------------------------
<?xml version="1.0" encoding="UTF-8"?>
<?xml-stylesheet href="http://xml.exciting-code.org/inputfileconverter/
inputtohtml.xsl" type="text/xsl"?>
<input xsi:noNamespaceSchemaLocation="http://static.exciting-code.org/
hydrogen/excitinginput.xsd"
xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance"
>
<title>Cu</title>
<structure speciespath="http://xml.exciting-code.org/species/">
<crystal scale="3.4156815" >
<basevect>1.0 1.0 0.0</basevect>
<basevect>1.0 0.0 1.0</basevect>
<basevect>0.0 1.0 1.0</basevect>
</crystal>
<species speciesfile="Cu.xml">
<atom coord="0.0 0.0 0.0" bfcmt="0.0 0.0 0.0"></atom>
</species>
</structure>
<groundstate ngridk="28 28 28" mixer="msec"></groundstate>
<properties>
<fermisurfaceplot/>
</properties>
</input>
-------------------------------------------------------------------------------------------------