mirror of https://gitlab.com/QEF/q-e.git
More documentation on PP generation
git-svn-id: http://qeforge.qe-forge.org/svn/q-e/trunk/espresso@6877 c92efa57-630b-4861-b058-cf58834340f0
This commit is contained in:
parent
8526e302e1
commit
7caac40cd7
Binary file not shown.
Binary file not shown.
Binary file not shown.
|
@ -16,6 +16,8 @@ Universit\`a di Udine and IOM-Democritos, Trieste\\
|
|||
URL: {\tt http://www.fisica.uniud.it/$\thicksim$giannozz}}
|
||||
\maketitle
|
||||
\date
|
||||
\tableofcontents
|
||||
|
||||
\section{Introduction}
|
||||
|
||||
When I started to do my first first-principle calculation
|
||||
|
@ -662,7 +664,8 @@ from the origin, in the energy range \texttt{eminld=-2.0} Ry to
|
|||
In the \texttt{\&inputp} namelist, we specify the we want a single-projector,
|
||||
NC-PP (\texttt{pseudotype=1}), with nonlinear core correction
|
||||
(\texttt{nlcc=.true.}), using the $l=1$ channel as local (\texttt{lloc=1}).
|
||||
The output PP wil be written in UPF format to file \texttt{Ti.pbe-rrkj.UPF}.
|
||||
The output PP wil be written in UPF format to file \texttt{Ti.pbe-n-rrkj.UPF}
|
||||
(following the {\sc quantum ESPRESSO} convention for PP names).
|
||||
Following the two namelists, there is a list of states used for pseudization:
|
||||
the 4S state, with pseudization radius $r_c=2.9$ a.u.; the 3D state,
|
||||
$r_c=1.3$ a.u.; the 4P, $r_c=2.9$ a.u., listed as last because it is
|
||||
|
@ -675,7 +678,7 @@ the channel to be chosen as local potential.
|
|||
/
|
||||
&inputp
|
||||
pseudotype=1, nlcc=.true., lloc=1,
|
||||
file_pseudopw='Ti.pbe-rrkj.UPF',
|
||||
file_pseudopw='Ti.pbe-n-rrkj.UPF'
|
||||
/
|
||||
3
|
||||
4S 1 0 2.00 0.00 2.9 2.9
|
||||
|
@ -783,7 +786,7 @@ namelist):
|
|||
iswitch=2
|
||||
/
|
||||
&test
|
||||
file_pseudo='Ti.pbe-rrkj.UPF',
|
||||
file_pseudo='Ti.pbe-n-rrkj.UPF',
|
||||
nconf=1, configts(1)='3d2 4s2 4p0',
|
||||
ecutmin=50, ecutmax=200, decut=50
|
||||
/
|
||||
|
@ -823,7 +826,7 @@ So our data for PP generation are as follows:
|
|||
/
|
||||
&inputp
|
||||
pseudotype=1, nlcc=.true., lloc=0,
|
||||
file_pseudopw='Ti.pbe-rrkj.UPF',
|
||||
file_pseudopw='Ti.pbe-n-rrkj.UPF',
|
||||
/
|
||||
3
|
||||
4P 2 1 0.00 0.00 2.9 2.9
|
||||
|
@ -877,7 +880,7 @@ allows to test several configurations in the following way:
|
|||
iswitch=2
|
||||
/
|
||||
&test
|
||||
file_pseudo='Ti.pbe-rrkj.UPF',
|
||||
file_pseudo='Ti.pbe-n-rrkj.UPF',
|
||||
nconf=10
|
||||
configts(1)='3d2 4s2 4p0'
|
||||
configts(2)='3d2 4s1 4p1'
|
||||
|
@ -985,8 +988,108 @@ on eigenvalues$<5$ mRy. Configurations with different $3d$ occupancy,
|
|||
however, have errors one order of magnitude higher. For the extreme
|
||||
case of Ti$^{4+}$, the error is $\sim$ 0.1 Ry.
|
||||
|
||||
\newpage
|
||||
In order to better understand what is going on, let us have a
|
||||
look at the AE vs PS orbitals and logarithmic derivatives for
|
||||
configuration 10 (i.e. for the bare PP). Let us add a line
|
||||
like this:
|
||||
\begin{verbatim}
|
||||
rlderiv=2.90, eminld=-4.0, emaxld=0.0, deld=0.01, nld=3,
|
||||
\end{verbatim}
|
||||
and plot files \texttt{ld10ps.wfc}, \texttt{ld10.dlog},
|
||||
\texttt{ld10ps.dlog} using \texttt{gnuplot} as above :
|
||||
|
||||
\includegraphics[width=7.5cm]{pseudo-gen-fig3.pdf}
|
||||
\includegraphics[width=7.5cm]{pseudo-gen-fig4.pdf}
|
||||
|
||||
Both the orbitals and the logarithmic derivatives (note the
|
||||
different energy range) start to exhibit some visible
|
||||
discrepancy now.
|
||||
|
||||
One can try to fiddle with all generation parameters,
|
||||
better if one at the time, to see whether things improve.
|
||||
Curiously enough, the pseudization radius for the core
|
||||
correction, which in principle should be as small as
|
||||
possible, seems to improve things if pushed slightly
|
||||
outwards (try \texttt{rcore=2.0}). Also surprisingly,
|
||||
a smaller pseudization radius for the $3d$ state, 0.9
|
||||
or 1.0 a.u., doesn't bring any visible improvement
|
||||
to transferability
|
||||
(but it increases a lot the required cutoff!).
|
||||
Changing the pseudization radii for $4s$ and $4p$ states
|
||||
doesn't affect much the results.
|
||||
|
||||
A different local potential -- a pseudized version
|
||||
of the total self-consistent potential -- can be chosen
|
||||
by setting \texttt{lloc=-1} and setting \texttt{rcloc}
|
||||
to the desired pseudization radius (a.u.). For small
|
||||
\texttt{rcloc} ghosts re-appear, but \texttt{rcloc=2.5}
|
||||
yields slightly better results. Note that the PP so
|
||||
generated will also have a $s$ projector, while the previous
|
||||
ones had only $p$ and $d$ projectors.
|
||||
|
||||
One could also generate the PP from a different electronic
|
||||
configuration. Since Ti tends to lose rather than to attract
|
||||
electrons, it will be more easily found in a ionized state than
|
||||
in the neutral one. One might for instance use the electronic
|
||||
configuration of the Bachelet-Hamann-Schl\"uter paper\cite{BHS}:
|
||||
$3d^2 4s^{0.75} 4p^{0.25}$. This however doesn't seem to improve
|
||||
much. Finally we end up with these generation data:
|
||||
\begin{verbatim}
|
||||
&input
|
||||
atom='Ti', dft='PBE', config='[Ar] 3d2 4s2 4p0',
|
||||
iswitch=3
|
||||
/
|
||||
&inputp
|
||||
pseudotype=1, nlcc=.true., rcore=2.0, lloc=-1, rcloc=2.5,
|
||||
file_pseudopw='Ti.pbe-n-rrkj.UPF'
|
||||
/
|
||||
3
|
||||
4P 2 1 0.00 0.00 2.9 2.9
|
||||
3D 3 2 2.00 0.00 1.3 1.3
|
||||
4S 1 0 2.00 0.00 2.9 2.9
|
||||
\end{verbatim}
|
||||
|
||||
\subsection {Single-projector, norm-conserving, with semicore states}
|
||||
|
||||
The results of transferability tests suggest that a Ti PP with only
|
||||
$3d$, $4s$, $4p$ states have limited transferability to cases with
|
||||
different $3d$ configurations. In order to improve it, a possible
|
||||
way is to put semicore $3s$ and $3p$ states in valence. The maximum
|
||||
for those states (0.87 a.u. and 0.90 a.u. respectively) is in the
|
||||
same range as for $3d$ (0.98 a.u.). Let us try thus the following:
|
||||
\begin{verbatim}
|
||||
&input
|
||||
atom='Ti', dft='PBE', config='[Ar] 3d2 4s2 4p0',
|
||||
rlderiv=2.90, eminld=-4.0, emaxld=2.0, deld=0.01, nld=3,
|
||||
iswitch=3
|
||||
/
|
||||
&inputp
|
||||
pseudotype=1, lloc=-1, rcloc=2.5, rho0=0.001
|
||||
file_pseudopw='Ti.pbe-sp-rrkj.UPF'
|
||||
/
|
||||
3
|
||||
3S 1 0 2.00 0.00 1.1 1.1
|
||||
3P 2 1 6.00 0.00 1.2 1.2
|
||||
3D 3 2 2.00 0.00 1.3 1.3
|
||||
&test
|
||||
configts(1)='3s2 3p6 3d2 4s2 4p0',
|
||||
/
|
||||
\end{verbatim}
|
||||
Note the presence of the \texttt{\&test} namelist: it is used in this
|
||||
context to supply the electronic valence configuration, to be used
|
||||
for unscreening. As a first step, we do not include the core correction
|
||||
|
||||
\newpage
|
||||
\subsection{Testing in molecules and solids}
|
||||
|
||||
Even if our PP looks good
|
||||
(or not too bad) on paper based on the results of atomic calculations,
|
||||
it is always a good idea to test it in simple molecular or solid-state
|
||||
systems, for which all-electron data (i.e. calculations performed with
|
||||
the same XC functional but without PP's, such as e.g. FLAPW, LMTO,
|
||||
Quantum Chemstry calculations) is available. The comparison with
|
||||
experiments is of course interesting, but the goal of PP's is to
|
||||
reproduce AE data, not to improve DFT.
|
||||
|
||||
|
||||
\appendix
|
||||
|
|
Loading…
Reference in New Issue