First principles calculation within DFT

some practical concerns

Ling-Ti Kong (-N)

EMail: konglt@sjtu.edu.cn
School of Materials Science and Engineering, Shanghai Jiao Tong University

Dec 24, 2010

1 / 19

Outline

k-sampling Plane wave cuto Pseudo-potential Another look on Born-von Karman PBC Summary: DFT with PP+PW Further readings

2 / 19

Kohn-Sham Equation for crystals

Schrdinger Equation Eigen problem o
2 |k m

+ Gm |2 m,m + vmm ci,m = i ci,m 2m

3 / 19

Brillouin zone integration

Properties like the electron density, total energy, etc., can be evaluated by integration over the rst Brillouin zone.

Electron density
n(r) = = = 1 Nk 1 FBZ 1 FBZ
FBZ occ.

ni,k (r),
k i occ.

ni,k (r) = |i,k (r)|2

ni,k (r)dk
i FBZ

fi,k ni,k (r)dk

4 / 19

k-point sampling: uniform vs non-uniform

Moreno and Soler (Phys. Rev. B 45(24):13891, 1992): A mesh with uniformly distributed k-points is preferred.

Example: A rectangular lattice (a) isotropic sampling (b) ner sampling vertically poor sampling horizontally

5 / 19

Monkhorst-Pack scheme
Monkhorst & Pack (Phys. Rev. B 13:5188, 1976.):
3

k=
i=1

2ni Ni 1 bi , 2Ni

ni = 1, , Ni

Example: 4 4 1 MP mesh for 2D square lattice 3 3 k1,1 = ( , ) 8 8 3 1 k1,2 = ( , ) 8 8 3 1 k1,3 = ( , ) 8 8 3 3 k1,3 = ( , ) 8 8

6 / 19

Shifted Monkhorst-Pack mesh

Centered on Centered around

25 k-points

Shifted by (1/8, 1/8, 0) 16 k-points

Yet the same k-point density.

7 / 19

Symmetry on k in the rst Brillouin zone

k+G = k Limit k in FBZ.

2me
2 2

+ve (r) uk (r)e ikr = uk (r)e ikr

2me
2 2

+ve (r) uk (r)e ikr = uk (r)e ikr

2me

+ve (r) uk (r)e ikr = uk (r)e ikr uk = uk

k = uk e ikr = uk e ikr
i,k

= k

i,k
8 / 19

Symmetry on k in the rst Brillouin zone

Sik (r) = ik (Sr) = uik (Sr)e ikSr = uik (r)e ik r k = S 1 k

occ.

n(r) =
k

k
i

ni,k (r)

k irreducible Brillouin zone k = Sik (r) = ik (Sr) # of sym. connected k total # of k in FBZ
9 / 19

Irreducible Brillouin zone

Example: MP-mesh for 2D square lattice 4 4 1 MP mesh = 16 k-points in the FBZ. 4 equivalent k4,4 = ( 3 , 3 ) w1 = 0.25 8 8 4 equivalent k3,3 = ( 1 , 3 ) w2 = 0.25 8 8 8 equivalent k4,3 = ( 3 , 1 ) w3 = 0.50 8 8 Brillouin zone integration 1 4
occ.

n(r ) =

ni,k4,4 +
i

1 4

occ.

ni,k3,3 +
i

1 2

occ.

ni,k4,3
i

10 / 19

Irreducible Brillouin zone

Symmetry break: Hexagonal cell Even meshes break the symmetry, while meshes centered on preserves it.
Shift the k-point mesh to preserve hexagonal symmetry!

k-sampling: rules of thumb Keep density of k constant in each direction; Ni bi = const. Denser k more precise results.
IBZ Nk does not necessarily scale with N. Symmetry matters

No k sampling needed for atoms or molecules! Generally need to do convergence test!

11 / 19

Cuto: nite basis set

Computationally, a complete expansion in terms of innitely many plane waves is not possible. The coecients, ci,m (k) decrease rapidly with increasing PW kinetic 2 2 energy |k+Gm | . 2m A cuto energy value, Ecut , determines the number of PWs (Npw ) in the expansion, satisfying,
2 |k

+ Gm |2 < Ecut 2m

Npw is a discontinuous function of the PW kinetic energy cuto, while depends only on the computational cell size and the cuto energy value.

12 / 19

Pseudopotential: why?
Reduction of basis set size eective speedup of calculation Reduction of number of electrons reduces the number of d.o.f. Unnecessary Why bother? They are inert anyway Inclusion of relativistic eects relativistic eects can be included partially

13 / 19

Pseudopotential: how?
Pseudopotential (PP) A smooth eective potential that reproduces the eect of the nucleus plus core electrons on valence electrons.

Norm conserving PP; Ultrasoft PP; Projector Augmented Wave PP;

14 / 19

Born-von Karman PBC: always a necessity?

Supercell approach

point defect

free surface

single molecule

Supercell must be suciently large to maintain isolation.

15 / 19

DFT: Ability and disability

Fundamentally, DFT can only predict the ground state electronic density and the ground state total energy of a set of electrons under an external potential. DFT cannot predict DFT can predict Total energy Excited state energies Forces Lattice constants Bond lengths Vibrational frequencies Phonon frequencies Electron density Static dielectric response Band gap Band structures Wave functions Fermi surface Superconductivity Excitons Electronic transport

16 / 19

Accuracy
Accuracies can be expected bond length bulk modulus phonon frequency energy dierence cohesive energy 3% too small 10% too high 10% too high > 1 mHartree very poor (much too high)

Accuracies for properties that DFT technically does not predict band gap band structure fermi surface 50% too small qualitatively reasonable qualitatively reasonable

17 / 19

Factors aecting accuracy

Born-Oppenheimer approximation; Density functional theory

LDA, GGA LSDA

Pseudopotential Kinetic energy cuto k-sampling convergence criterion

18 / 19

References and further readings

1 2 3 4

http://www.fhi-berlin.mpg.de/th/Meetings/FHImd2001/pehlke1.pdf
http://itp.tugraz.at/LV/ewald/TFKP/Literatur Pseudopotentiale/Roundy 05 DFT+PPsumm.pdf http://www.phys.sinica.edu.tw/TIGP-NANO/Course/2007 Spring/Class%20Notes/CMS.20070531.pseudo.pdf

J. Singleton, Band theory and electronic properties of solids, Oxford University press. http://cms.mpi.univie.ac.at/vasp/ http://www.pwscf.org/ http://www.abinit.org/ http://www.cpmd.org/ http://www.icmab.es/siesta/

5 6 7 8 9

19 / 19