3.

320: Lecture 6 (Feb 17 2005)

HARTREE-FOCK AND DFT

Photos of Hartree, Fock, Hohenberg, Kohn, and Sham

removed for copyright reasons.

Feb 17 2005 3.320 Atomistic Modeling of Materials -- Gerbrand Ceder and Nicola Marzari
 Matrix Formulation and
Variational Principle

cn m H n = Ecm
n =1, k

<|H | >
E [ ] = E0
<| >
Feb 17 2005 3.320 Atomistic Modeling of Materials -- Gerbrand Ceder and Nicola Marzari
 Energy of an Hydrogen Atom

H
E =

= C exp ( r )

C2 1 2 C2 1 C2
= , = = 2
3
2 2 r

Feb 17 2005 3.320 Atomistic Modeling of Materials -- Gerbrand Ceder and Nicola Marzari
 Two-electron atom
1 2 1 2 Z Z 1 r r r r
1 2 + r r (r1 , r2 ) = Eel (r1 , r2 )
2 2 r1 r2 | r1 r2 |

Feb 17 2005 3.320 Atomistic Modeling of Materials -- Gerbrand Ceder and Nicola Marzari
 Many-electron atom
1 Z 1 r r r r
i + r r (r1 ,..., rn ) = Eel (r1 ,..., rn )
2

2 i i ri i j > i | ri rj |

Feb 17 2005 3.320 Atomistic Modeling of Materials -- Gerbrand Ceder and Nicola Marzari
Atomic Units and Conversion Factors
(see handout)

1 a.u. = 2 Ry = 1 Ha
1 Ry = 13.6057 eV
1 eV = 23.05 kcal/mol

Feb 17 2005 3.320 Atomistic Modeling of Materials -- Gerbrand Ceder and Nicola Marzari
 Molecules and Solids:
Electrons and Nuclei
r r r r r r r r
H (r1 ,..., rn , R1 ,..., RN ) = Etot (r1 ,..., rn , R1 ,..., RN )

We treat only the electrons as quantum particles, in the

field of the fixed (or slowly varying) nuclei
This is generically called the adiabatic or Born-
Oppenheimer approximation
Adiabatic means that there is no coupling between
different electronic surfaces; B-O implies there is no
influence of the ionic motion on one electronic surface

Feb 17 2005 3.320 Atomistic Modeling of Materials -- Gerbrand Ceder and Nicola Marzari
 Energy of a collection of atoms
H = Te + Ve e + Ve N + VN N

( )
1 r r 1
Te = i2
Ve N = V RI ri
Ve e = r r

2 i i I i j >i | ri r j |

Te: quantum kinetic energy of the electrons

Ve-e: electron-electron interactions
Ve-N: electrostatic electron-nucleus attraction
(electrons in the field of all the nuclei)
VN-N: electrostatic nucleus-nucleus repulsion
Feb 17 2005 3.320 Atomistic Modeling of Materials -- Gerbrand Ceder and Nicola Marzari
Complexity of the many-body

Some form of approximation is essential, and this would

mean the construction of tables. The tabulation function of one
variable requires a page, of two variables a volume and of three
variables a library; but the full specification of a single wave
function of neutral iron is a function of 78 variables. It would be
rather crude to restrict to 10 the number of values of each
variable at which to tabulate this function, but even so, full
tabulation would require 1078 entries.

Feb 17 2005 3.320 Atomistic Modeling of Materials -- Gerbrand Ceder and Nicola Marzari
 Mean-field approach
Independent particle model (Hartree): each
electron moves in an effective potential,
representing the attraction of the nuclei and
the average effect of the repulsive
interactions of the other electrons
This average repulsion is the electrostatic
repulsion of the average charge density of
all other electrons
Feb 17 2005 3.320 Atomistic Modeling of Materials -- Gerbrand Ceder and Nicola Marzari
 Hartree Equations
The Hartree equations can be obtained directly from the variational
principle, once the search is restricted to the many-body
wavefunctions that are written as above as the product of single
orbitals (i.e. we are working with independent electrons)
r r r r r
(r1 ,..., rn ) = 1 (r1 ) 2 (r2 ) L n (rn )

1 2 r r r 2 1 r r r
i + I V ( RI ri ) + | j (rj ) | r r drj i (ri ) = i (ri )
2 j i | rj ri |

Feb 17 2005 3.320 Atomistic Modeling of Materials -- Gerbrand Ceder and Nicola Marzari
 The self-consistent field
The single-particle Hartree operator is self-
consistent ! It depends on the orbitals that are
the solution of all other Hartree equations
We have n simultaneous integro-differential
equations for the n orbitals
Solution is achieved iteratively
1 2 r r r 2 1 r r r
i + I V ( RI ri ) + | j (rj ) | r r drj i (ri ) = i (ri )
2 j i | rj ri |

Feb 17 2005 3.320 Atomistic Modeling of Materials -- Gerbrand Ceder and Nicola Marzari
 Iterations to self-consistency
Initial guess at the orbitals
Construction of all the operators
Solution of the single-particle pseudo-
Schrodinger equations
With this new set of orbitals, construct the
Hartree operators again
Iterate the procedure until it (hopefully)
converges

Feb 17 2005 3.320 Atomistic Modeling of Materials -- Gerbrand Ceder and Nicola Marzari
Differential Analyzer

Vannevar Bush and the Differential Analyzer.

Courtesy of the MIT Museum. Used with permission.

Feb 17 2005 3.320 Atomistic Modeling of Materials -- Gerbrand Ceder and Nicola Marzari
 Whats missing
It does not include correlation
The wavefunction is not antisymmetric
It does remove nl accidental degeneracy of
the hydrogenoid atoms

Feb 17 2005 3.320 Atomistic Modeling of Materials -- Gerbrand Ceder and Nicola Marzari
 Spin-Statistics
All elementary particles are either fermions
(half-integer spins) or bosons (integer)
A set of identical (indistinguishable)
fermions has a wavefunction that is
antisymmetric by exchange
r r r r r r r r r r
(r1 , r2 ,..., rj ,..., rk ,..., rn ) = (r1 , r2 ,..., rk ,..., rj ,..., rn )
For bosons it is symmetric

Feb 17 2005 3.320 Atomistic Modeling of Materials -- Gerbrand Ceder and Nicola Marzari
 Slater determinant
An antisymmetric wavefunction is constructed via a
Slater determinant of the individual orbitals (instead
of just a product, as in the Hartree approach)
r r r
(r1 ) (r1 ) L (r1 )
r r r
r r r 1 (r2 ) (r2 ) L (r2 )
(r1 , r2 ,..., rn ) =
n! M M O M
r r r
(rn ) (rn ) L (rn )

Feb 17 2005 3.320 Atomistic Modeling of Materials -- Gerbrand Ceder and Nicola Marzari
 Pauli principle
If two states are identical, the determinant
vanishes (i.e. we cant have two electrons in
the same quantum state)

Feb 17 2005 3.320 Atomistic Modeling of Materials -- Gerbrand Ceder and Nicola Marzari
 Hartree-Fock Equations
The Hartree-Fock equations are, again, obtained from the variational principle: we
look for the minimum of the many-electron Schroedinger equation in the class of all
wavefunctions that are written as a single Slater determinant
r r
(r1 ,..., rn ) = Slater

1 2 r r r
2 i + V ( RI ri ) (ri ) +
I
* r 1 r r r
(rj ) r r (rj ) drj (ri )
| rj ri |
* r 1 r r r r
(rj ) | rr rr | (rj )drj (ri ) = (ri )
j i
Feb 17 2005 3.320 Atomistic Modeling of Materials -- Gerbrand Ceder and Nicola Marzari
 Shell structure of atoms

Self-interaction free
Good for atomic properties
Start higher-order perturbation theory
Exchange is in, correlation still out
Feb 17 2005 3.320 Atomistic Modeling of Materials -- Gerbrand Ceder and Nicola Marzari
 Restricted vs. Unrestricted
Spinorbitals in the Slater determinant:
spatial orbital times a spin function
Unrestricted: different orbitals for different
spins
Restricted: same orbital part

Feb 17 2005 3.320 Atomistic Modeling of Materials -- Gerbrand Ceder and Nicola Marzari
 The Dissociation of H2
RHF
RELATIVE ENERGY

RH-H

UHF

EXACT

Figure by MIT OCW.

Feb 17 2005 3.320 Atomistic Modeling of Materials -- Gerbrand Ceder and Nicola Marzari
 Koopmans Theorems
Total energy is invariant under unitary
transformations
It is not the sum of the canonical MO orbital
energies
Ionization energy, electron affinity are
given by the eigenvalue of the respective
MO, in the frozen orbitals approximation

Feb 17 2005 3.320 Atomistic Modeling of Materials -- Gerbrand Ceder and Nicola Marzari
 What is missing
Correlations (by definition !)
Dynamical correlations: the electrons get too
close to each other in H.-F.
Static correlations: a single determinant
variational class in not good enough
Spin contamination: even if the energy is
correct (variational, quadratic) other
properties might not (e.g. the UHF spin is
an equal mixture of singlet and triplet)
Feb 17 2005 3.320 Atomistic Modeling of Materials -- Gerbrand Ceder and Nicola Marzari
 Faster, or better
The exchange integrals are the hidden cost
(fourth power). Linear-scaling efforts
underway
Semi-empirical methods (ZDO, NDDO,
INDO, CNDO, MINDO): neglect certain
multi-center integrals
Configuration interaction, Mller-Plesset

Feb 17 2005 3.320 Atomistic Modeling of Materials -- Gerbrand Ceder and Nicola Marzari
 Configuration Interaction
In normal Hartree Fock, the HF determinant is built from the lowest energy single-
electron states

N basis functions and K electrons -> fill up i with K lowest energies

HF
0
= 1 2 ... K
IDEA of CI is to mix in excited states
Issues:
e.g. exchange state i with K+1 Many possible excitations
HF
1
= 1 2 ... K+1 ... K (combinatorial problem)
Time consuming (scaling N7)
Define new variational wave function:

= c0 0 + c11 + c2 2 + ...

Feb 17 2005 3.320 Atomistic Modeling of Materials -- Gerbrand Ceder and Nicola Marzari
Image removed for copyright reasons.
Screenshot of online article.
Nobel Focus: Chemistry by Computer. Physical Review Focus, 21 October 1998.
http://focus.aps.org/story/v2/st19

Feb 17 2005 3.320 Atomistic Modeling of Materials -- Gerbrand Ceder and Nicola Marzari
 Density-functional theory
The external potential Vext and the number N of
electrons completely define the quantum problem
The wavefunctions are in principle ! uniquely
determined, via the Schrdinger Equation
All system properties follow from the wavefunctions
The energy (and everything else) is thus a functional
of Vext and N

Feb 17 2005 3.320 Atomistic Modeling of Materials -- Gerbrand Ceder and Nicola Marzari
 The Thomas-Fermi approach
Lets try to find out an expression for the
energy as a function of the charge density
E=kinetic+external+el.-el.
Kinetic is the tricky term: how do we get
the curvature of a wavefunction from the
charge density ?
Answer: local density approximation

Feb 17 2005 3.320 Atomistic Modeling of Materials -- Gerbrand Ceder and Nicola Marzari
 Local Density Approximation
We take the kinetic energy density at every
point to correspond to the kinetic energy
density of the homogenous electron gas
5
r r
T (r ) = A (r ) 3

r r
5
r r r r r 1 (r1 ) (r2 ) r r
ETh Fe [ ] = A (r )dr + (r )vext (r )dr + r r dr1dr2
3
2 | r1 r2 |

Feb 17 2005 3.320 Atomistic Modeling of Materials -- Gerbrand Ceder and Nicola Marzari
 Its a poor man Hartree
The idea of an energy functional is not
justified
It does not include exchange effects - but
Dirac proposed to add the LDA exchange
4
energy: r 3 r
C (r ) dr
It scales linearly, and we deal with 1
function of three coordinates !

Feb 17 2005 3.320 Atomistic Modeling of Materials -- Gerbrand Ceder and Nicola Marzari
 The Argon atom

Feb 17 2005 3.320 Atomistic Modeling of Materials -- Gerbrand Ceder and Nicola Marzari
The Hohenberg-Kohn theorems (1965)
The density as the basic variable: the
external potential determines uniquely the
charge density, and the charge density
determines uniquely the external potential.

Feb 17 2005 3.320 Atomistic Modeling of Materials -- Gerbrand Ceder and Nicola Marzari
 The universal functional F[]
The ground state density determines the
potential of the Schrdinger equation, and
thus the wavefunctions
r
F [ (r )] = T + Ve e

Its an emotional moment

Feb 17 2005 3.320 Atomistic Modeling of Materials -- Gerbrand Ceder and Nicola Marzari
 Second Hohenberg-Kohn theorem
The variational principle we have a new
Schrdingers-like equation, expressed in
terms of the charge density only
r r r r r
Ev [ (r )] = F [ (r )] + vext (r ) (r )dr E0

( determines its groundstate wavefunction, that can be taken as

a trial wavefunction in this external potential)

< | H | >=< | T + Ve e + vext | >= vext + F [ ]

Feb 17 2005 3.320 Atomistic Modeling of Materials -- Gerbrand Ceder and Nicola Marzari
 Euler-Lagrange equations

[ r r r r r r
F [ (r )] + vext (r ) (r )dr (r )dr N = 0 ( )]
r
F [ (r )] r
r + vext (r ) =
(r )

Feb 17 2005 3.320 Atomistic Modeling of Materials -- Gerbrand Ceder and Nicola Marzari
 References
W. Koch, M. C. Holthausen, A Chemists Guide to Density
Functional Theory
R. G. Parr, W. Yang, Density-Functional Theory of Atoms and
Molecules
W. Kohn, Nobel lecture
F. Jensen, Introduction to Computational Chemistry
J. M. Thijssen, Computational Physics
B. H. Bransden and C. J. Joachim, Physics of Atoms and
Molecules

Feb 17 2005 3.320 Atomistic Modeling of Materials -- Gerbrand Ceder and Nicola Marzari
 Software
Gaussian (http://www.gaussian.com) ($$) (chemistry, Hartree-Fock, DFT,
correlated approaches)
Gamess-UK (http://www.cse.clrc.ac.uk/qcg/gamess-uk/ ) ($) (chemistry,
Hartree-Fock, DFT, correlated approaches)

Materials Studio/Cerius (http://www.accelrys.com) ($$) (DFT, planewave,

ultrasoft)
Crystal (http://www.chimifm.unito.it/teorica/crystal) ($) (Hartree-Fock)
VASP (http://cms.mpi.univie.ac.at/vasp) ($) (DFT, planewave, ultrasoft,
PAW)

ESPRESSO (http://www.pwscf.org) (free) (DFT, planewave, ultrasoft, linear-

response theory, Car-Parrinello)
ABINIT (http://www.abinit.org) (free) (DFT, planewave, linear-response
theory, GW)
CPMD (http://www.cpmd.org) (free) (DFT, planewave, Car-Parrinello, time-
dependent DFT)

CASINO (http://www.tcm.phy.cam.ac.uk/~mdt26/cqmc.html ) (free)

(Quantum Monte Carlo)
Feb 17 2005 3.320 Atomistic Modeling of Materials -- Gerbrand Ceder and Nicola Marzari