PT Vgs Nist Tuchi

Acquisition of Diffusion Data from Ab Initio 
Comments on a “Good Usage” of the Pseudo-potential Method"
Patrice Turchi ([email protected])

NIST Diffusion Workshop, May 3-4, 2012, Gaithersburg, MD!
LLNL-PRES-553992
This work was performed under the auspices of the U.S. Department
of Energy by Lawrence Livermore National Laboratory under contract
DE-AC52-07NA27344. Lawrence Livermore National Security, LLC.
What Are We Talking About?"
There is a suite of popular pseudo-potential codes available:"
Abinit – www.abinit.org"
Pseudo-potentials and plane-wave basis set!
"
CAMPOS – www.fysik.dtu.dk/CAMP/CAMPOS_welcome.html"
Pseudo-potentials and plane-wave basis set."
"
CASTEP – www.tcm.phys.cam.uk/castep"
"
FHI96MD – www.fhi-berlin.mpg.de/th/fhimd"
Pseudo-potentials and plane-wave basis set,!
with molecular dynamics capabilities based on the Car-Parrinello technique!
"
PWSCF – www.pwscf.org "
Pseudo-potentials and plane-wave basis set.!
DFT codes include response function (phonon) features.!
!
Quantum Expresso – www.democritos.it/scientific.php"
Pseudo-potentials and LCAO basis set.!
Open-source package for research in electronic structure, simulation, and optimization. !
!
SIESTA – www.uam.es/depatamentos/ciencas/fismateriac/siesta"
Pseudo-potentials and LCAO basis set!
"
VASP – www.mpi.univie.ac.at/vasp"
The recommended standard potentials were used; in a few cases we opted for harder potentials
with semicore states treated as valence states, Bi_d, Ga_d, Ge_d, Hf_pv, Li_sv, Mo_pv, Nb_pv,
Pb_d, Sc_sv, Sn_d, Sr_sv, Ta_pv, Ti_sv, W_pv, Y_sv, and Zr_sv. For the lanthanides we selected
potentials with three f-like electrons treated as core states, Ce, Er, Gd, Ho, Lu, Nd, Pm, Pr,
Sm, Tb, and Tm, or two f-like electrons treated as core states, Eu. !
DFT, VASP-PAW-GGA!
D. Simonovic and M.H.F. Sluiter, “Impurity diffusion activation energies in Al from first principles”, Phys. Rev. B79,
054304 (2009)."
For the first-principles calculations we use the Vienna ab initio simulation package VASP with
projector augmented wave (PAW) potentials. The potentials used for each of the elements Al,
Mg, Si and Cu do not treat any semi-core states as valence.!
DFT, VASP-PAW-LDA/GGA!
M. Mantina, Y. Wang, L.Q. Chen, Z.K. Liu, and C. Wolverton “First-principles impurity diffusion coefficients”, Acta
Materialia 57, 4102 (2009)."
We used the Vienna ab initio Software Package (VASP )to perform DFT calculations with the
generalized gradient approximation (GGA) as parametrized by Perdew, Burke, and Ernzerfhof (PBE)
for the exchange-correlation functional. Inadequate repulsion between Zn 3d and conduction-
band levels in GGA calculations results in a significantly underestimated band gap. We
partially corrected this by using the GGA+U correction with a value of U−J =7. 5 eV so that the
valence-band and Zn 3d energy levels match experiments and self-interaction corrected
calculations. We used projector-augmented plane-wave (PAW) pseudopotentials with valence
configuration 3d 10, 4s 2 for Zn, 2s 2 2p 4 for O, and 4s 2 4p 3 for As, and a plane-wave energy
cutoff of 600 eV."
DFT, VASP-PAW-GGA+U"
B. Puchala and D. Morgan, “Atomistic modeling of As diffusion in ZnO”, Phys. Rev. B85, 064106 (2012)."
Why Pseudopotentials?"
v  Reduction of the basis set size.!

v  Reduction of the number of electrons.!
Critical Approximations:"
Ø  One-electron picture!
Ø  No significant overlap between core
and valence wavefunctions (frozen
core)!
The only role of the core electrons is to

provide an effective external potential (or
pseudopotential) in which the outer
electrons move.! !
SPEED versus ACCURACY and TRANSFERABILITY"

!
Results from a pseudo-potential-based code are as good as the
pseudo-potentials are.!
Norm-conserving pseudopotentials: outside rc the norm of each pseudo-

wavefunction is identical to its corresponding all-electron wavefunction.!
Ultra-soft pseudopotentials: relax the norm-conserving constraint to
reduce the basis set size (other related technique: PAW).!
Critical Approximations (true for most

DFT-based methods and codes):"
Ø  Exchange and correlation potential (LDA,
LSDA, GGA, SIC, GWM, DMFT) !
Ø  Relativistic effects (scalar relativistic, fully

relativistic, spin-orbit coupling)!
Ø  In most DFT-based codes, the zero-point

motion contribution to the total energy is
absent.! !
Why Construct a Pseudopotential?"
l  There is a handful of repositories with prefabbed pseudopotentials:!

l  Abinit's web site: http://www.abinit.org !
l  Rappe Group: http://lorax.chem.upenn.edu!
l  Vanderbilt's library of ultrasoft pseudopotentials: !
http://www.physics.rutgers.edu/~dhv/uspp/uspp-734.html!
l  But, it is very likely that you may want to make your own:!
!• All elements may not be available !!
!• Pseudopotential may not give accurate results!
!• Pseudopotential may be too expensive to use for your problem!
l  Furthermore, having the ability to construct pseudopotentials can be an !

important research tool.!
Pseudo-potential Construction: AE"
l  Get the all-electron eigenvalues and wavefunctions.
Cu AE wave-functions"
Pseudo-potential Construction: PS "
Select cut-off radii "
Construct pseudo-wavefunctions!
Invert the KS equation to find the semi-local pseudo-potential:!
Solve the KS equations using semi-local pseudo-potential!

Pseudo-potential Convergence"
Error in Total Energy vs. PSP Cut-off energy (Ecut)!
l  Cut-off energy is a measure of the!

size of the plane-wave basis!
!
l  Want to minimize Ecut !
!
l  Larger E longer run time!
cut
Optimized method delivers highly efficient pseudo-potentials !

Atomic Criteria"
● First step of construction is to design pseudo-wavefunctions!

● Choose reference configuration and cutoff radii (rc )"
5 rules for a “good” pseudo-potential "
1." True for all norm conserving "

pseudo-potentials"
2."
3."
Norm conservation
4."
guarantees"
5." But how good???"
D. R. Hamann, M. Schlüter, and C. Chiang, !
PRL 43, 1494 (1979)!
tail norm!
Pseudo-wavefunctions are nodeless
and satisfy (at least) the above criteria" log. derivative!
Eigenvalues and Tail Norm"
rc"
r"
ψ

rc"
r"
Must enforce conservation of both!"

What is a Good Pseudo-potential?"
l  For a pseudo-potential to be useful, it must be both accurate and efficient!

!
accuracy - reproduction of all-electron atomic properties!
"
efficiency - cost of using the pseudo-potential in solid-state calculations!
l  Need a self-contained program to construct and test the accuracy and !
efficiency of pseudo-potentials!
"
Open-source Pseudopotential Interface and Unification Module"
(J. Bennett & A.M. Rappe)"
http://opium.sourceforge.net/index.html "
What about 2nd-principles approaches to Alloy property
simulations? EAM Phase Diagrams"
Experimental" EAM Phase Diagram with"

Phase Diagram" The case of Fe-Cu alloys" Ackland s Potentials"
fcc Fe-Cu!
fcc Fe-Cu!
E. M. Lopasso, M. Caro, A. Caro, and P. E. A.

Turchi, “The phase diagram of Fe-Cu as
predicted by an empirical potential”, Phys. Rev. B
68, 214205-1 to 9 (2003)!
A. Caro, P. E. A. Turchi, M. Caro, and E. M.
Lopasso, “Thermodynamics of an Empirical EAM Phase Diagram
Potential Description of Fe-Cu Alloys”, J. of Nucl.
Mater. 336, 233-242 (2005).!
with Ludwig-Farkas
A. Caro, M. Caro, E. M. Lopasso, P. E. A. Turchi, Potentials"
and D. Farkas, “Thermodynamics of Fe-Cu Alloys
as Described by Classic Potential”, J. of Nucl.
Mater. 349, 317-326 (2006).!
Conclusions "
!Yes, there are plenty of “good” user friendly pseudopotential-

based codes out there
The question is: are the potentials for a given set of elements
designed to tackle a problem never looked at before?
• Pseudo-potentials must be checked out within a new
atomic configuration space (the same way EAM potentials
should) to make sure that accuracy is preserved
!There will always be a trade off between speed and accuracy

(and therefore transferability)

PT Vgs Nist Tuchi

Uploaded by

Copyright:

Available Formats

PT Vgs Nist Tuchi

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

PT Vgs Nist Tuchi

Uploaded by

Copyright:

Available Formats

Acquisition of Diffusion Data from Ab Initio

Comments on a “Good Usage” of the Pseudo-potential Method"

Patrice Turchi ([email protected])

v Reduction of the basis set size.!

The only role of the core electrons is to

SPEED versus ACCURACY and TRANSFERABILITY"

Norm-conserving pseudopotentials: outside rc the norm of each pseudo-

Critical Approximations (true for most

Ø Relativistic effects (scalar relativistic, fully

Ø In most DFT-based codes, the zero-point

l There is a handful of repositories with prefabbed pseudopotentials:!

l Rappe Group: http://lorax.chem.upenn.edu!

l Vanderbilt's library of ultrasoft pseudopotentials: !

l Furthermore, having the ability to construct pseudopotentials can be an !

l Get the all-electron eigenvalues and wavefunctions.

Select cut-off radii "

Invert the KS equation to find the semi-local pseudo-potential:!

Solve the KS equations using semi-local pseudo-potential!

l Cut-off energy is a measure of the!

Optimized method delivers highly efficient pseudo-potentials !

● First step of construction is to design pseudo-wavefunctions!

5 rules for a “good” pseudo-potential "

1." True for all norm conserving "

Must enforce conservation of both!"

l For a pseudo-potential to be useful, it must be both accurate and efficient!

Experimental" EAM Phase Diagram with"

E. M. Lopasso, M. Caro, A. Caro, and P. E. A.

!Yes, there are plenty of “good” user friendly pseudopotential-

!There will always be a trade off between speed and accuracy

You might also like

Acquisition of Diffusion Data from Ab Initio 

v  Reduction of the basis set size.!

Ø  Relativistic effects (scalar relativistic, fully

Ø  In most DFT-based codes, the zero-point

l  There is a handful of repositories with prefabbed pseudopotentials:!

l  Rappe Group: http://lorax.chem.upenn.edu!

l  Vanderbilt's library of ultrasoft pseudopotentials: !

l  Furthermore, having the ability to construct pseudopotentials can be an !

l  Get the all-electron eigenvalues and wavefunctions.

l  Cut-off energy is a measure of the!

l  For a pseudo-potential to be useful, it must be both accurate and efficient!