1 DFT Introduction

1
Introduction to Density Functional Theory
Prof.P. Ravindran,
Department of Physics, Central University of Tamil
Nadu, India
http://folk.uio.no/ravi/CMT2015
P.Ravindran, Condensed Matter Physics July 2015 Introduction to Density Functional Theory
2
3
4
Ab initio Computational
Based Design
5
Conducting a Computational Project
 These questions should be answered

– What do you want to know?
– How accurate does the prediction need to be?
– How much time can be devoted to the
problem?
– What approximations are being made?
 The answers to these questions will determine

the type of calculation, model and basis set to
be used.
* from D. Young P.Ravindran, Condensed Matter Physics July 2015 Introduction to Density Functional Theory
6
7
Computational Models
 A model is a system of equations, or computations used to

determine the energetics of a molecule/material.
 Different models use different approximations (or levels of
theory) to produce results of varying levels of accuracy.
 There is a trade off between accuracy and computational
time.
 There are two main types of models; those that use
Schrödinger's equation (or simplifications of it) and those
that do not.
8
Computational Models
 Types of Models
(Listed in order from most to least accurate)
– Ab initio
 uses Schrödinger's equation, but
with approximations
– Semi Empirical
 uses experimental parameters and
extensive simplifications of
Schrödinger's equation
– Molecular Mechanics
 does not use Schrödinger's equation Simulated (12,0) zigzag carbon nanotube
9
Ab Initio
 Ab initio translated from Latin means “from first principles.” This refers
to the fact that no experimental data is used and computations are based
on quantum mechanics.
 Different Levels of Ab Initio Calculations

– Hartree-Fock (HF)
 The simplest ab initio calculation
 The major disadvantage of HF calculations is that electron

correlation is not taken into consideration.
– The Møller-Plesset Perturbation Theory (MP)
– Density Functional Theory (DFT)
Take into consideration
– Configuration Interaction (CI) electron correlation
10
Ab Initio
 Approximations used in some ab initio calculations
– Central field approximation: integrates the

electron-electron repulsion term, giving an average
effect instead of an explicit energy
– Linear combination of atomic orbitals (LCAO): is

used to describe the wave function and these
functions are then combined into a determinant.
This allows the equation to show that an electron
was put in an orbital, but the electron cannot be
specified.
11
Density Functional Theory
 Considered an ab initio method, but different from other ab initio

methods because the wave function is not used to describe the solid,
instead the electron density is used.
 Three types of DFT calculations exist:
– local density approximation (LDA) – fastest method, gives less

accurate geometry, but provides good band structures
– gradient corrected - gives more accurate geometries
– hybrids (which are a combination of DFT and HF methods) - give

more accurate geometries
12
Semi Empirical
 Semi empirical methods use experimental data to

parameterize equations
 Like the ab initio methods, a Hamiltonian and wave function

are used
– much of the equation is approximated or eliminated
 Less accurate than ab initio methods but also much faster
 The equations are parameterized to reproduce specific results,

usually the geometry and heat of formation, but these methods
can be used to find other data.
13
Molecular Mechanics
 Simplest type of calculation

– Used when systems are very large and approaches that are more
accurate become too costly (in time and memory)
 Does not use any quantum mechanics instead uses parameters derived
from experimental or ab initio data
– Uses information like bond stretching, bond bending, torsions,
electrostatic interactions, van der Waals forces and hydrogen bonding
to predict the energetics of a system
– The energy associated with a certain type of bond is applied
throughout the molecule. This leads to a great simplification of the
equation
 It should be clarified that the energies obtained from molecular mechanics

do not have any physical meaning, but instead describe the difference
between varying conformations (type of isomer). Molecular mechanics
can supply results in heat of formation if the zero of energy is taken into
account.
14
Basis Sets
 In chemistry a basis set is a group of mathematical functions used to

describe the shape of the orbitals in a molecule, each basis set is a different
group of constants used in the wavefunction of the Schrödinger equation.
 The accuracy of a calculation is dependent on both the model and the type
of basis set applied to it.
 Once again there is a trade off between accuracy and time. Larger basis
sets will describe the orbitals more accurately but take longer to solve.
 General expression for a basis function = N * e(- * r)

– where: N is the normalization constant,  is the orbital exponent, and r
is the radius of the orbital in angstroms.
15
Is numerical simulation necessary?
 Experiments have limitations under certain circumstances

(e.g., samples, techniques, signals, etc) so that many
physical properties can not be measured accurately.
 Microscopic process (e.g., growth and phase transition

mechanism, etc) and the understanding of physics are not
experimentally measurable.
Numerical computation must be relied on
16
Computational Physics
Performs idealized "experiments" on
Computation
the computer by solving physical
models numerically
predicts
Theory tests Experiments

Construction of idealized Quantitative measurement of
models through mathematical physical properties
(analytical) analysis of
physical principle to describe
nature
17
18
19
20
21
22
Crystal Structure is the basis for understanding materials

and their properties
Electronic
property
Optical Thermodynamic
property property
Crystal structure
Mechanical
property Magnetism
23
What Density Functional Theory Can Do
 Applies to atoms, molecules, surfaces and periodic structures...

 Self-consistent total energies, forces, stresses (ground state)
 Structural optimization
 Elastic modulus
 Electron band structure of many-body systems
 Phonon frequencies and eigenvectors at any wave vector
 Full phonon dispersions
 Electron-phonon interactions
 ab-initio MD
 etc..
24
What Density Functional Theory Can NOT Do?
 DFT Calculations are not exact solutions of the full Schrodinger equation
(exact functional unknown).
 The Hohenberg-Kohn theorems apply only to ground state energies.
Excited states can be predicted, but these predictions are not on the same
footing as predictions for ground-state energies.
 There is a well-known inaccuracy in DFT involving the underestimation of
calculated band gaps in semiconducting and insulating materials (changed
recently).
 DFT gives inaccurate results involving weak van der Waals attractions,
which are a direct result of long range electron correlation (changed
recently).
 It is computationally expensive to perform DFT calculations. Simulations
with hundreds of atoms are about the largest you can get.
25
First principles: the whole picture
The equipment Application
Scientific
problem-
solving
“Base Implementation
Theory” (the algorithms Setup model, Research
(DFT) and program) run the code output
“Analysis
Theory”
26
What is Density Functional Theory (DFT)?
 “Density Functional Theory is a phenomenally successful

approach to finding solutions to the fundamental equation
for the quantum behavior of atoms and molecules, the
Schrodinger equation, in settings of practical value.”
 Problem: Solving the Schrodinger equation is a many-body

problem (3N bodies to be exact).
 Solution: Electrons are indistinguishable. So the quantity

of physical interest is instead the probability that N
electrons in any order have specific coordinates. So the
electron density, which is a function of only 3 coordinates,
contains a great deal of information that is physically
observable from the full solution to the Schrodinger
equation.
27
The T in DFT
 The entire field of Density Functional Theory rests on two

fundamental mathematical theorems proven by Kohn and
Hohenberg, and the derivation of a set of equations by Kohn
and Sham.
 Theorem #1: There exists a 1-to-1 mapping between the

ground state wave function and the ground state electron
density.
 Theorem #2: The electron density that minimizes the energy of
the above functional is the true electron density corresponding
to the full solution to the Schrodinger equation.
Guess the functional form and vary the electron density iteratively
until the energy is minimized!
28
The Math
 A useful way to write down the functional described by Theorem

#1 is in terms of the single electron wave functions (according
to the Hartree Product approximation). The Schrodinger equation
then becomes the Kohn-Sham equations:
29
30
Approximation 1
Separation of electrons into

valence electrons and core electrons
Ion cores = core electrons + nuclei
e.g. Si: [Ne] 3s23p2
32
Approximation 2
Born-Oppenheimer or adiabatic approximation
To electrons, ions are essentially stationary.
Ions only see a time-averaged adiabatic electronic potential.
ionic motion
electron-phonon
 Lattice vibration (phonons) interaction
electronic motion
34
Many Particle problem in Solids
35
36
37
Approximation 3
Mean-field approximation: Every electron experiences the

same average potential V(r)
V(r): by first principle (ab initio), or by semi-empirical

approach
39
40
41
42
Hartree Method for many particle Solids
Hartree Hamiltonian 43
SCF procedure 44
45
46
47
Introduction of Exchange (Spin) : Hartree Fock Method
48
ν2(i,j) =1/r12
49
Failure of HF Method
50
Exchange and Correlation
51
52
53
54
What is functional ?
55
56
57
58
59
How to get an approximate XC functional?
60
Approximations to EXC
 Local density approximation (LDA)
– Assumption: EXC depends locally on ρ( r )
LDA
E XC    (r ) XC
hom
(r )dr
– Recommended LDA functional: Perdew-Wang (PRB 45, 13244,
1992)
– LDA is currently being used to study fundamental problems in

physics, chemistry, geology, material science and pharmacy.
61
LDA-LSDA: Trends And Limitations
 Favors more homogeneous electron densities
 Overbinds molecules and solids (Hartree-Fock underbinds)
 Geometries, bond lengths and angles, vibrational
frequencies reproduced within 2-3%
 Dielectric constants overestimated by about 10%
 Bond lengths too short for weakly bound systems (H-bonds,
VDW)
 Correct chemical trends, e.g. ionization energies
 Atoms (core electrons) poorly described (HF is much better)

 XC potential decays exponentially into vacuum regions. It should
decay as –e2/r. Hence, it is poor for dissociation and ionization
 Poor for metallic surfaces and physisorption
 Very poor for negatively charged ions (self-interaction error)
 Poor for weakly bound systems: H-bonds (), VDW (non-local)
 Band gap in semiconductors too small (~40%)
 Poor for strong on-site correlations (d and f systems, oxides, UO2)
Beyond The LDA
 Inhomogeneities in the density

 Self-interaction cancellation
 Non-locality in exchange and correlation
 Strong local correlations
 Gradient expansions
 Weighted density approximation
 Exact exchange in DFT (OEP local vs HF non-local)
 DFT-HF hybrids
 Self-interaction correction
 Van der Waals and RPA functionals
 LSDA+U
 Multi-reference Kohn-Sham
 GW approximation (Many-body)
64
 Generalized gradient approximation (GGA)
– Assumption:
GGA
E XC    (r ) XC (  (r ),  (r ))dr
– Recommended GGA functional: Perdew-Burke-
Ernzerhof (PBE) [PRL 77, 3865 (1996)].
– GGA is found to improve the binding energies, but not

the band gaps.
Properties of the GGA
 Improves atomization and surface energies

 Favors density inhomogeneities
 Increases lattice parameters of metals
 Favors non-spherical distortions
 Improves bond lengths
 Improves energies and geometries of H-bonded systems
 There is error cancellation between X and C at short range
 XC potential still decays exponentially into vacuum regions

 Some improvement in band gaps in semiconductors
 What was correct in LDA is worsened in GGA
 Still incorrect dissociation limit. Fractionally charged
fragments
 Inter-configurational errors in IP and EA
 Error cancellation between X and C is not complete at long-
range. X hole is more long-ranged than XC hole
66
 Meta-GGA (MGGA)
– Assumption:
MGGA
E XC    (r ) XC (  (r ),  (r ), (r ))dr
– Here, τ is the kinetic energy density

occ
 (r )   | i (r ) |2
i
– Recommended MGGA functional: Toa-Perdew-

Staroverov-Scuseria (TPSS) [PRL 91, 146401 (2003)]
– Self-interaction free correlation. Not well tested yet.
67
Main problem with LDA, GGA and MGGA

– They allow for spurious self-interaction (SI).
– Exact DFT is SI free:
1 v*k (r )vk (r ' )v 'k ' (r ' )v*'k ' (r )

E x     drdr '
.
2 vv 'kk ' |r r | '
For vk  v 'k ' ,

1  vk (r )  vk (r ' )
E    self
drdr '   EHSelf
|r r |
x
2
 Exact cancelation between self VH and VX
 This is not the case for LDA and GGA.
68
69
Hybrid Functionals
 Combine GGA local exchange with Hartree-Fock non-local exchange:
EHYBRID [  ]   E XGGA [  ]  (1   ) E XHF [  ]  ECGGA [  ]

 Parameter  fitted to experimental data for molecules (~0.75), or
determined from known properties.
 PBE0, B3LYP, HSE06
 Properties:
1. Quite accurate in many respects, e.g. energies and geometries
2. Improve on the self-interaction error, but not fully SI-free
3. Improve on band gaps
4. Improve on electron affinities
5. Better quality than MP2
6. Fitted hybrids unsatisfactory from the theoretical point of view
71
Hybrid functionals (mixing of LDA with HF)
72
 Theory of Exact-exchange (EXX)

[Stadele et al. PRB 59, 10 031 (1999)]
– Total energy
1  (r )  (r ' )
Etot [  ]  To  Ee i   drdr '

2 |r r | '
1  v*k (r ) vk (r ' ) v 'k ' (r ' ) v*'k ' (r )


2 vv 'kk '  |r r | '
drdr '
 EC [  ]
– Single-particle equations
 1 2 
   Vion  VH [  ]  VX [  ]  VC [  ] i   i i , with
 2 
E X  E X [  ] vk (r ' )  VKS (r '' )
VX (r )       c.c.
 (r ) vk  vk (r ) VKS (r )
' ''
  (r )
73
Some key points about DFT
 DFT is a description of interacting electrons in the ground

state, including exchange and correlation
 The basic variable is the density rather than the wave
function
 The theory is simple and the implementations efficient
compared with other methods
 Implementations scale at least as well as N2
 It offers an excellent balance between accuracy and scale of
calculation
Self-interaction Correction (SIC)
 Self-interaction can be removed at the level of classical electrostatics:
1  (r)  (r' )
2  | r  r' |
EH  drdr'
1 N  (r )  n (r ' )
ESIC  EH    n drdr '
2 n 1 | r  r'|
 (r' )
Perdew-Zunger 1982 VSIC (n)
(r )  VH (r )   n dr '
Mauri, Sprik, Suraud | r  r ' |
 Potential is state-dependent. Hence it is not an eigenvalue problem
anymore, but a system of coupled PDEs
 Orthogonality of SIC orbitals not guaranteed, but it can be imposed
(Suraud)
 Similar to HF, but the Slater determinant of SIC orbitals is not
invariant against orbital transformations
 The result depends on the choice of orbitals (localization)
Dynamical Correlation: VDW
 Van der Waals (dispersion) interactions: are a dynamical non-

local correlation effect
 Dipole-induced dipole interaction due to quantum density

fluctuations in spatially separated fragments
1(r,t) 2(r,t)
 Functional (Dion et al 2004):
1(t) E1(t).2(t)
EVDW    (r ) (r, r ' )  (r ' ) drdr '  = VDW kernel fully non-local.
Depends on (r) and (r’)
 Expensive double integral
 Efficient implementations (Roman-Perez and Soler 2009)
 Good approximations based on dynamical response theory
 Beyond VDW: Random Phase Approximation (Furche)
Dynamical Correlation: Vdw And Beyond
 Empirical approaches:
C6IJ ( RIJ )
P
EVDW   f IJ ( RIJ )
IJ RIJ6
 With f a function that removes smoothly the singularity at R=0, and
interferes very little with GGA (local) correlation.
 Grimme (2006): C6 parameters from atomic calculations. Extensive

parameterization: DFT-D.
 Tkatchenko and Scheffler (2009): C6 parameters dependent on the
density.
 Random Phase Approximation (RPA): captures VDW and beyond.

Can be safely combined wit exact exchange (SIC). Infinite order
perturbation (like Coupled clusters in QC).
Furche (2008); Paier et al (2010); Hesselmann and Görling (2011)
Strong Static Correlation: LSDA+U
 Strong onsite Coulomb correlations are ot captured by LDA/GGA
 These are important for localized (d and f) electronic bands, where
many electrons share the same spatial region: self-interaction problem
 Semi-empirical solution: separate occupied and empty state by an

additional energy U as in Hubbard’s model:
1 1
 E LSDA
This  E
induces
U a  U
splitting
LSDAU N
in( N 
the 1)
KS  U  f i f j f =orbital occupations
eigenvalues:
2 2 i j i
E 1   iocc   iLSDA  U / 2
 i  LSDAU   iLSDA  U   f 
f i 2   iempty   iLSDA  U / 2
Summary Of DFT Approximations
79
80
Density Functional Theory (DFT) is a formally exact representation

of the N electron Schrödinger Equation
Schrödinger view DFT view
Formally
equivalent
electron Kohn-Sham particle
interaction (non-interacting)
external potential effective potential
Hard problem to solve “Easy” problem to solve
(scales like N5 ) ( scales like N3 or better)
Hohenberg and Kohn proved this (1964)
Nobel prize in Chemistry for Kohn in 1998
81
1998 Nobel Prize Laureates, Chemistry
"for his development of "for his development of

the density-functional computational methods in
theory" quantum chemistry"
Advent of Quantitative Theory and Modeling

82
83
84
Exchange-Correlation Potential
85
86
Bloch’s Theorem
From translational symmetry
 For a periodic potential V(x), the eigenfunctions of H (wavefunctions

of e-) always have the form:
ψk(x) = eikx uk(x) with uk(x) = uk(x+a)
 “BLOCH FUNCTIONS”
 Recall, for a free e-, the wavefunctions have the form:
ψfk(x) = eikx (a plane wave)

 A Bloch function is the generalization of a plane wave for an e- in
periodic potential. It is a plane wave modulated by a periodic function
uk(x) (with the same period as V(x) ).
Periodic Boundary Conditions and Bloch’s Theorem
 Periodic Boundary Conditions: Finite systems are assumed to be

periodically repeated to fill the whole space
An efficient recipe to study

atoms, molecules, surfaces,
Interfaces, … etc
 Bloch’s Theorem: The wave-functions of the electrons moving in a periodic potential are
given as
 nk (r )  unk (r )eik .r
Unk(r) have the same periodicity as the potential.
n is the band index
k is a wave-vector inside the 1st BZ.
87
88
 This transforms the problem into calculating few wavefunctions for, in principle, infinite
number of k points.
 The great simplification comes from the fact that Ψnk are weakly varying functions with
respect to k … only few carefully chosen k-points (known as special k-points) are
required.
 Convergence test: Si in the diamond structure
No. special E (H) Lattice Bulk

Mesh K-points (a=10.4 Bohr) constant (Å) modulus (Mbar)
2x2x2 2 -7.930764 5.392 0.959
4x4x4 10 -7.936765 5.384 0.956
8x8x8 60 -7.936879 5.384 0.954
Expt. 5.431 0.99

Example: 2x2 mesh
For 2D square lattice
89
SCF procedure
1. Define an initial, trial electron density.

2. Solve the Kohn-Sham equations using the trial electron
density to find the single particle wave functions.
3. Calculate the electron density defined by the Kohn-Sham
single particle wave functions from step 2.
4. Compare the calculated electron density with the electron
density used to solve the Kohn-Sham equations. If they
agree with some specified error, then compute the total
energy from this density, otherwise update the trial
electron density in some way.
5. Repeat.
This iterative method can lead to a solution of the Kohn-Sham

equations that is self-consistent!
90
The Kohn-Sham equations must be solved self-consistently The potential
(input) depends on the density (output)
Initial guess
Calculate effective potential
Solve the KS equation
No
Compute electron density
Yes Output quantities
Self-consistent? Energy, forces,
stresses …
91
92
93
94
95
96
97
98
99
100
101
Success and failure of “standard” DFT in solids
 Standard LDA (GGA) gives good description of structural and electronic

properties of most solids (lattice parameters within 1-2%, at least
qualitatively correct bandstructure, metal-insulator, magnetism,…)
 Problems: “localized” (correlated) electrons
– late 3d transition metal oxides/halides
 metals instead of insulators (FeO, FeF2, cuprates, …)
 nonmagnetic instead of anti-ferromagnetic (La2CuO4,

YBa2Cu3O6)
– 4f, 5f electrons
 all f-states pinned at the Fermi energy, “always” metallic
 orbital moments much too small
– “weakly” correlated metals

 FeAl is ferromagnetic in theory, but nonmagnetic
experimentally
 3d-band position, exchange splitting,…
102
Is LDA/GGA repairable ?
Ab initio methods
 GGA: usually improvement, but often too small.
 Hartree-Fock: completely neglects correlation, very poor in solids
 Exact exchange: imbalance between exact X and approximate (no) C
 Hybrid-Functionals: mix of HF + LDA; good for insulators, poor for
metals
 GW: gaps in semiconductors, but ground state? expensive!
 Quantum Monte-Carlo: very expensive
Not fully ab initio
 Self-interaction-correction: vanishes for Bloch states
 Orbital polarization: Hund’s 2nd rule by atomic Slater-parameter
 LDA+U: strong Coulomb repulsion via external Hubbard U
parameter
 DMFT: extension of LDA+U for weakly correlated systems
103
104
105
Control and conditions
 We can manipulate the model system: complete control

– Move and place atoms
– Apply strains
– Try configurations
 Any conditions and situations are accessible
– High pressures and temperatures
– Buried interfaces, porous media, nanostructures
106
Taking advantage
 Calculate quantities for other theories

– Transition states and barriers
– Defect energies
 Use unphysical routes, e.g. free energy calculations
– Switch from reference system to full simulation
– Transmute elements
107
108
109
110
Computational approaches
All-electron: Pseudopotential:
- all the electron are explicitly included
- electrons = valence+
- the space is separated in core are core.
interstitial regions.
- only Valence electrons
are explicitly included.
interstitial
- effective potential
(pseudopotential)
- Two main approaches core
due to the nucleus
I- LAPW {partial waves (core) and PW are the core electrons
(interstitial)}
II- LMTO {partial waves (core) and Hankel - PW basis sets to
functions (interstitial)} expand Ψnk
Lecture Summary
111
 Density functional theory is an exact reformulation of many-body

quantum mechanics in terms of the probability density rather than
the wave function
 The ground-state energy can be obtained by minimization of the

energy functional E[n]. All we know about the functional is that
it exists, however, its form is unknown.
 Kohn-Sham reformulation in terms of single-particle orbitals helps

in the development of approximations and is the form used in
density functional calculations today.
112
Summary
 DFT has been the mainstay of electronic structure

calculations more than 25 years.
 Because approximate functional provide a useful balance
between accuracy and computational cost.
 This allowed much larger systems to be treated.
 DFT is a completely different way of approaching any
interacting problem, by mapping it exactly to a much
easier-to-solve non-interacting problem.
 Its methodology is applied into a large variety of systems as
atoms ,molecules, clusters and solids.
 Limitations of LDA/GGA were partly overcome by Hybrid
functionals.
113
11
4
115

1 DFT Introduction

Uploaded by

Copyright:

Available Formats

1 DFT Introduction

Uploaded by

Document Information

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

1 DFT Introduction

Uploaded by

Copyright:

Available Formats

1

Introduction to Density Functional Theory

 These questions should be answered

 The answers to these questions will determine

 A model is a system of equations, or computations used to

 Different Levels of Ab Initio Calculations

 The major disadvantage of HF calculations is that electron

 Approximations used in some ab initio calculations

– Central field approximation: integrates the

– Linear combination of atomic orbitals (LCAO): is

Density Functional Theory

 Considered an ab initio method, but different from other ab initio

 Three types of DFT calculations exist:

– local density approximation (LDA) – fastest method, gives less

– gradient corrected - gives more accurate geometries

– hybrids (which are a combination of DFT and HF methods) - give

 Semi empirical methods use experimental data to

 Like the ab initio methods, a Hamiltonian and wave function

 Less accurate than ab initio methods but also much faster

 The equations are parameterized to reproduce specific results,

 Simplest type of calculation

 It should be clarified that the energies obtained from molecular mechanics

 In chemistry a basis set is a group of mathematical functions used to

 General expression for a basis function = N * e(- * r)

Is numerical simulation necessary?

 Experiments have limitations under certain circumstances

 Microscopic process (e.g., growth and phase transition

Numerical computation must be relied on

Theory tests Experiments

Crystal Structure is the basis for understanding materials

What Density Functional Theory Can Do

 Applies to atoms, molecules, surfaces and periodic structures...

What Density Functional Theory Can NOT Do?

First principles: the whole picture

The equipment Application

What is Density Functional Theory (DFT)?

 “Density Functional Theory is a phenomenally successful

 Problem: Solving the Schrodinger equation is a many-body

 Solution: Electrons are indistinguishable. So the quantity

 The entire field of Density Functional Theory rests on two

 Theorem #1: There exists a 1-to-1 mapping between the

 A useful way to write down the functional described by Theorem

Separation of electrons into

Ion cores = core electrons + nuclei

e.g. Si: [Ne] 3s23p2

Born-Oppenheimer or adiabatic approximation

To electrons, ions are essentially stationary.

Ions only see a time-averaged adiabatic electronic potential.

Many Particle problem in Solids

Mean-field approximation: Every electron experiences the

V(r): by first principle (ab initio), or by semi-empirical

Exchange and Correlation

– Assumption: EXC depends locally on ρ( r )

– LDA is currently being used to study fundamental problems in

 Atoms (core electrons) poorly described (HF is much better)

 Inhomogeneities in the density

 Generalized gradient approximation (GGA)

– GGA is found to improve the binding energies, but not

 Improves atomization and surface energies

1 vk (r )vk (r ' )v 'k ' (r ' )v'k ' (r )

1  vk (r ) vk (r ' ) v 'k ' (r ' ) v'k ' (r )